Profiling and Characterizing Skin Ceramides Using Reversed-Phase

Nov 23, 2011 - Liquid Chromatography−Quadrupole Time-of-Flight Mass. Spectrometry ... stripping of human skin, were analyzed by reversed-phase liqui...
2 downloads 0 Views 5MB Size
Article pubs.acs.org/ac

Profiling and Characterizing Skin Ceramides Using Reversed-Phase Liquid Chromatography−Quadrupole Time-of-Flight Mass Spectrometry Ruben t’Kindt,† Lucie Jorge,† Emmie Dumont,‡ Pauline Couturon,§ Frank David,‡ Pat Sandra,‡ and Koen Sandra*,† †

Metablys, President Kennedypark 26, 8500 Kortrijk, Belgium Research Institute for Chromatography, President Kennedypark 26, 8500 Kortrijk, Belgium § Chemistry Department, K.U. Leuven, Celestijnenlaan 200f, 3001 Heverlee, Belgium ‡

S Supporting Information *

ABSTRACT: An LC-MS based method for the profiling and characterization of ceramide species in the upper layer of human skin is described. Ceramide samples, collected by tape stripping of human skin, were analyzed by reversed-phase liquid chromatography coupled to high-resolution quadrupole time-of-flight mass spectrometry operated in both positive and negative electrospray ionization mode. All known classes of ceramides could be measured in a repeatable manner. Furthermore, the data set showed several undiscovered ceramides, including a class with four hydroxyl functionalities in its sphingoid base. High-resolution MS/MS fragmentation spectra revealed that each identified ceramide species is composed of several skeletal isomers due to variation in carbon length of the respective sphingoid bases and fatty acyl building blocks. The resulting variety in skeletal isomers has not been previously demonstrated. It is estimated that over 1000 unique ceramide structures could be elucidated in human stratum corneum. Ceramide species with an even and odd number of carbon atoms in both chains were detected in all ceramide classes. Acid hydrolysis of the ceramides, followed by LC-MS analysis of the endproducts, confirmed the observed distribution of both sphingoid bases and fatty acyl groups in skin ceramides. The study resulted in an accurate mass retention time library for targeted profiling of skin ceramides. It is furthermore demonstrated that targeted data processing results in an improved repeatability versus untargeted data processing (72.92% versus 62.12% of species display an RSD < 15%).

C

moieties but also in the chain length and the number, position, and stereochemistry of double bonds, hydroxyl groups, and other functionalities of their building blocks.1,2,8 Typical SBs detected in SC comprise dihydrosphingosine (or sphinganine), sphingosine, phytosphingosine, and 6-hydroxysphingosine (Scheme 1).8 SBs predominantly have a chain length of 18 carbons, but variation in the chain length from C12 to C26 has been reported.9−11 The coupled FA part (R1) is typically saturated with a chain length from C14 to C26, although longer FA chains have already been detected in the skin surface. αHydroxy FAs are also fairly common, next to esterified ωhydroxy FAs.12−14 Because of the huge structural variety, skin CER profiling and characterization requires the use of a highly efficient chromatographic step before mass spectrometric detection. Both normalphase (NP) liquid chromatography (LC) and reversed-phase (RP) LC have been used to separate CER species.1,2,15,16 NP-

eramides (CERs), in se only a subclass of sphingolipids, represent an enormous structural diversity that is consequently translated into a large number of existing CER species.1,2 CERs are found in all tissues of the human body, but the largest diversity of CERs (up to now) is found in the human skin, more precisely in the uppermost layer of the epidermis, i.e., the stratum corneum (SC). This layer of the epidermis consists of enucleate cells or corneocytes, which are embedded in a matrix of mainly CERs (>50% by weight), cholesterol and free fatty acids (FAs) (between 15 and 20 wt %), and minor fractions of triacylglycerols and phospholipids.3−6 This lipid-rich intercellular matrix is the only continuous domain through the SC, defining the pathway through which molecules can diffuse across the SC. As a consequence, CERs in SC play a prominent role in the barrier and permeability properties or water barrier functions of the skin.5 More recently, CERs have also been linked to skin diseases and other pathologies.7 CERs are produced by coupling FA acyl chains onto sphingoid bases (SB) by an amide binding. The resulting molecules do vary not only in the combination of FA and SB © 2011 American Chemical Society

Received: October 20, 2011 Accepted: November 23, 2011 Published: November 23, 2011 403

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

Scheme 1. Structure of CER Classes Already Discovered in Stratum Corneuma

a

If all possible combinations are considered, 16 CER classes can be listed: CER[NDS], CER[NS], CER[NP], CER[NH], CER[ADS], CER[AS], CER[AP], CER[AH], CER[ODS], CER[OS], CER[OP], CER[OH], CER[EODS], CER[EOS], CER[EOP], and CER[EOH]. For detailed nomenclature see Supporting Information, S-1.

Collection of Skin Samples. Stratum corneum samples (tape stripped) were collected from healthy volunteers. All of these individuals enrolled with informed consent. The lower part of the inner forearm, just above the wrist, is first cleaned with ethanol to remove interfering skin surface lipids. After cleaning with ethanol, the skin of the inner forearm of the individuals was stripped with a patch coated with an adhesive layer, Corneofix (Courage + Khazaka electronic GmbH, Köln, Germany). Contact time was five seconds. To collect one sample, three patches were placed on a 1 cm distance on the lower forearm skin, pushed down, and removed with tweezers. Ceramide Extraction Method. The three skin patches were immersed in 5 mL of methanol, vortexed, and sonicated for 10 min. After sonication, the patches were removed and the remaining methanol fractions were combined in one tube. The combined methanol fraction was dried at 37 °C under a nitrogen stream. For measurements in the negative ESI mode, this extract was injected after dissolving in 300 μL of isopropanol/chloroform 50/50 (v/v). For measurements in positive ESI mode, a solid-phase extraction (SPE) step using aminopropyl silica cartridges (100 mg, 3.0 mL from Alltech, Deerfield, IL, USA) was included. The dried extract was therefore reconstituted in 300 μL of 11/1 hexane/isopropanol (v/v). SPE columns were preconditioned with 2 mL of hexane. After the sample load, the cartridge was washed with 2 mL of hexane and CERs were eluted using 2 mL of a hexane/ methanol/chloroform 80/10/10 (v/v) mixture. The eluted fraction was dried under nitrogen and dissolved in 300 μL of isopropanol/chloroform 50/50 (v/v). Liquid Chromatography−Mass Spectrometry. A 1200 RRLC system (Agilent Technologies, Waldbronn, Germany) was used for RP-LC measurements. Intact CERs were analyzed on an XBridge BEH C18 Shield column (2.1 × 100 mm; 1.8 μm; Waters, Milford, MA, USA) placed in a Polaratherm 9000 series oven (Selerity Technologies, Salt Lake City, UT, USA) at 80 °C. Elution was carried out with a gradient of (A) 20 mM ammonium formate pH 5 and (B) methanol, starting from 70% B to 100% B in 75 min. The flow rate was 0.5 mL/min, and the injection volume was 10 μL. The whole system was allowed to re-equilibrate under starting conditions for 20 min. FAs were analyzed using a shorter version of the method. The gradient was shortened to 45 min and started at 30% B. SBs were analyzed on an Acquity HSS T3 column (2.1 × 100 mm; 1.8 μm; Waters), at a flow rate of 0.4 mL/min, and at 40 °C.

LC distinguishes CERs according to their hydrophilic functionalities. It is typically used to separate the CERs in their representative classes but lacks sensitivity when coupled to electrospray ionization (ESI) mass spectrometry. Moreover, dozens of molecular species of the same class are eluted over a narrow time range, which makes the identification of minor species very difficult, especially when low-resolution mass spectrometry is used.17 RP-LC enables the separation of CERs based on their hydrophobic properties, depending mainly on the FA chain length, the length of the SB acyl chain, and the number of unsaturated bonds. The use of RP instead of NP might substantially increase the separation efficiency for the CERs, as their intrinsic hydrophilic properties are weak compared to their hydrophobic properties. Likewise, shotgun phospholipid analysis already confirmed that RP-LC resulted in a higher peak capacity for lipid species compared to NP-LC.17 In continuation of the blood plasma lipidomics platform previously described, the aim of the present study was to develop a comprehensive skin CER profiling and characterization platform.18 In that perspective, CERs, extracted from the inner forearm skin using adhesive patches, were subjected onto a hyphenated setup composed of high resolution RP-LC on sub-2 μm particles and Jetstream ESI quadrupole-time-offlight (Q-TOF) mass spectrometry. Due to the large complexity within the skin CER pool and in response to the current trend in life sciences, the complete setup was developed as a holistic approach, typically used in “omics” analytical procedures.



EXPERIMENTAL SECTION Materials. Water (ULC-MS grade), methanol (ULC-MS grade), isopropanol (HPLC grade), hexane (Pesti-S grade), acetonitrile (HPLC-S grade), chloroform (HPLC grade), and formic acid (LC-MS grade) were purchased from Biosolve BV (Valkenswaard, The Netherlands). Ammonium formate was purchased from Sigma-Aldrich (Bornem, Belgium). Hydrochloric acid 6 N was obtained from Thermo Scientific (Rockford, IL, USA). Standard CERs N-lignoceroyl-D-erythrosphingosine or CER[N(24:0)S(18)], N-lignoceroyl-D-erythrosphinganine or CER[N(24:0)DS(18)], N-(2′-(R)-hydroxystearoyl)-D-erythro-sphingosine or CER[A(18:0)S(18)], and N-stearoyl 4-hydroxysphinganine or CER[N(18:0)P(18)] were obtained from Avanti Polar Lipids (Alabaster, AL, USA). 404

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

Figure 1. (a) Total ion chromatogram of a CER extract in negative ESI mode. Regions of nonesterified and esterified ω-OH FAs can be distinguished. (b) Extracted compound chromatograms (ECC) representing the range of CER features present in CER[NS] (general structure is shown) with a carbon atom number from 32 to 54. CER with an odd number of carbon atoms are also present but in a lower concentration compared to CER with an even number of carbon atoms. (c) Extracted compound chromatograms (ECC) representing CER species with C42 from different CER classes. CER[AS] and CER[NP] and CER[ADS] and CER[NS] are (partially) coeluting, but these can be distinguished by their diagnostic masses and subsequent MS/MS fragmentation pattern.

capabilities is a prerequisite. In continuation of our recently described blood plasma lipidomics platform, the present study uses a Q-TOF MS system complemented with an RP-LC methodology utilizing small chromatographic particles (sub-2 μm) and elevated temperature (80 °C).18 In contrast to the lipidomics methodology, that requires a complex elution profile for the simultaneous detection of (lyso)phospholipids, sphingolipids, cholesterol esters, and mono-, di-, and triacylglycerol species; the CER LC methodology is based on a linear gradient between 70 and 100% methanol. A typical total ion current (TIC) LC-MS chromatogram of a skin CER extract obtained in negative ESI is shown in Figure 1a. To obtain a better view on the overall profile, the corresponding two-dimensional LC-MS plot is shown in Figure 2a. An enormous complexity is revealed, illustrating the power of the LC-MS methodology. A smooth transition between the lower molecular weight nonesterified CERs and the heavier esterified CERs is observed. It has been reported in an earlier study that RP-LC using methanol has difficulties in eluting the more hydrophobic CERs such as esterified ω-hydroxy FA containing species (CER[EO]).1 The current protocol allows eluting CER[EO] using methanol due to the use of an elevated column temperature of 80 °C. The benefit of the use of RP-LC can be seen when extracted compound chromatograms (ECC) from a CER class are withdrawn from the data, as is shown for CER[NS] with C32 to C54 in Figure 1b (identification strategy is described further). Each individual CER elutes on a fixed retention time distance from the other CERs of its class. The unique elution pattern, originating from the differences in the number of CH2 groups, facilitates the detection of the representatives of all the CER classes. RP-LC also gives a proper separation of the different CER classes, as is shown in

Gradient elution started from 100% water + 0.1% formic acid to 100% acetonitrile + 0.1% formic acid in a time frame of 20 min. High-resolution accurate mass measurements were obtained on an Agilent 6530 Q-TOF mass spectrometer (Agilent Technologies) equipped with a Jetstream ESI source. The instrument was operated in both positive and negative ion electrospray mode. Needle voltage was optimized to ±3.5 kV; the drying and sheath gas temperatures were set to 300 °C, and the drying and sheath gas flow rates were set to 6 and 8 L/min, respectively. Data were collected in centroid mode at an acquisition rate of 1 spectrum/s in the extended dynamic range mode (2 GHz), offering a resolution of ±10 000 fwhm in the CER m/z range. MS/MS experiments were performed in the data dependent acquisition mode (DDA). A survey MS scan was alternated with three DDA MS/MS scans resulting in a cycle time of 4 s. Singly charged precursor ions were selected on the basis of abundance. After being fragmented twice, a particular m/z value was excluded for 30 s, allowing the MS/MS fragmentation of chromatographically resolved CER isomers. The quadrupole was operated at medium resolution, and the collision energy was fixed at either 20 or 35 eV, respectively. The data processing step is extensively described in S-2, Supporting Information.



RESULTS AND DISCUSSION LC-MS Based Ceramide Profiling Platform. The substantial structural diversity in skin CERs makes it challenging to develop a comprehensive analytical method. The use of high resolution chromatography combined with high-end mass spectrometric equipment with inherent high resolution, accurate mass measurement, and tandem MS 405

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

Figure 2. (a) Two-dimensional representation of the raw CER analysis, which plots retention time versus m/z of the detected features (negative ESI mode). Esterified and nonesterified CERs can be distinguished easily. (b) Profile plot of identified CER species in stratum corneum after targeted data processing, representing 264 CER species with their adducts detected in negative ESI mode. (c) The present CER distribution for each class. [Color intensity is decreasing from N < A < O; DS CER are colored black, S are blue, P are green, H are red, and T are yellow; EO CER have identical color symbols as O CER but are color filled; CER containing a double bond in the SB part (i.e., S and H) have no fill; [M − H]− are plotted as circles, [M + Cl]− as triangles, [M + HCOO]− as lozenges, [M + CF3COO]− as squares].

This retention time difference is in accordance with previous reports on hydroxy FAs.19−21 In accordance with the literature, CERs can be detected in both ionization modes.1,22,23 For measurements in negative ESI, methanol extracts can be subjected onto the LC-MS methodology without further pretreatment apart from concentration via evaporation. Solid-phase extraction (SPE), however, is a definite requirement when analyzing skin CER samples in positive ESI mode. Interfering substances originating from the patch and skin are removed by the normal-phase based SPE step. Since elaborate sample preparation imparts precision and there is no immediate complementarity between positive and negative ESI (apart from the difference in MS/MS behavior), measurements in negative ESI without SPE purification are preferred for profiling purposes. Under the analytical conditions used, CERs predominantly appear as protonated species in

Figure 1c for CER species with equal total carbon number (i.e., CER[NDS]C42, CER[ADS]C42, CER[NS]C42, CER[AS]C42, CER[NP]C42, CER[AP]C42, CER[NH]C42, and CER[AH]C42). The elution order is CER[AH] < CER[NH] < CER[AP] < CER[NP] ≈ CER[AS] < CER[ADS] ≈ CER[NS] < CER[NDS]. The coeluting CERs (i.e., CER[NP] and CER[AS]; CER[ADS] and CER[NS]) can be easily distinguished by their masses. The chromatographic step clearly separates structural isomers that originate from the different positioning of a hydroxyl group between distinct CER classes such as CER[NP] and CER[ADS] or CER[AS] and CER[NH] (i.e., position isomerism or class isomerism). However, MS/MS fragmentation is mandatory to appoint these CERs to their respective classes (see further). CERs containing ω-hydroxy FAs (not esterified) elute about 8 min earlier than the position isomeric CERs coupled to α-hydroxy FAs (data not shown). 406

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

Figure 3. Fragmentation spectrum of CER[NP] with a total carbon number of 45 at a collision energy of 35 eV. Due to skeletal isomerism, CER[NP]C45 can be defined as the result of various CER species varying in carbon length of SB and FA building blocks, i.e., CER[N(29:0)P(16)], CER[N(28:0)P(17)], CER[N(27:0)P(18)], CER[N(26:0)P(19)], CER[N(25:0)P(20)], CER[N(24:0)P(21)], CER[N(23:0)P(22)], and CER[N(22:0)P(23)].

carboxylate ions in MS/MS enables a direct identification of the FA substituent(s).24 Elaborate Skin Ceramide Characterization. Intact Ceramide Analysis. In the present work, efforts have been undertaken to identify a substantial portion of the detected CER species. The mass accuracy and MS/MS capabilities of the Q-TOF system and the intraspecies elution pattern proved to be highly valuable. The identification strategy started from a triplicate LC-MS analysis in negative ESI of a skin CER sample followed by the generation of a feature list for every run. A feature, in principle corresponding to a unique molecular entity, results from a feature extraction algorithm that localizes and combines related covarying ions with the generation of a single mass (median), retention time (peak apex), and an abundance (sum of ions in clusters). A peak height intensity filter hereby allowed a faster processing of the feature extraction algorithm. Adducts were considered as separate molecular entities to keep precision high. A data matrix was subsequently generated, this by aligning the features from the replicate runs, followed by abundance RSD filtering (RSD < 15%), maintaining 1392 reproducible features of the 2704 aligned features detected. An in-house build database that included the known skin CER classes described in the literature was built to annotate the features within a 5 ppm mass accuracy window. On the basis of the theoretical database, 591 features could be annotated as CER compounds, only considering the already known skin CER classes. The number of unique CER species detected is predicted lower, as adducts (chlorine, formate, and TFA) are considered as separate molecular entities for reason of precision. After annotation, at least two representatives of each CER class were studied in detail using MS/MS fragmentation in both positive and negative ESI mode and identified as such using reported fragmentation mechanisms/ spectra.1,2,10,23−28 In the MS/MS spectra, practically all fragments of CER species could be unraveled with a mass

positive ESI. In negative ESI, a typical CER spectrum displays ions with an increasing signal intensity from [M − H]− < [M + HCOO]− ≈ [M + CF3COO]− < [M + Cl]−. This adduct formation occurred independently of the sample (standard CERs, SPE purified or raw skin extracts) and instrument used (both the JetStream ESI source as the classical ESI source give similar spectra) and was described earlier for chlorine adducts.23,24 Formate adducts are originating from the mobile-phase modifier, while trifluoroacetic acid (TFA) is present in the Agilent TOF tuning mix and is used frequently on the LC-Q-TOF instrument as LC modifier in biopharmaceutical experiments. Experiments were performed to push the adduct formation toward one single adduct or to enhance the ionization of the molecular ion, by postcolumn addition of fortifying or weakening modifiers such as 0.01% TFA, 0.1% FA, or different types of amine containing bases, all this with limited success. In contrast, atmospheric-pressure chemical ionization (APCI) caused mainly formic acid and chlorine adducts but with an even more reduced signal of the original molecular ion compared to ESI (see S-3, Supporting Information, for comparison between JetStream ESI and APCI). As the presence of the molecular ion is a necessity for identification of the CERs through MS/MS fragmentation (collision induced dissociation of adducts do not yield valuable MS/MS spectra), Jetstream ESI was chosen as ionization source. Previous experiments already described the increased sensitivity of JetStream ESI ionization compared to classical ESI ionization in RP-LC Q-TOF-MS lipidomics analysis.18 The excellent sensitivity of CER [M + H]+ ions in positive ESI mode, together with the rich fragmentation spectra providing information of both the FA and the SB part, makes it the preferred choice for identification purposes. The adduct formation in negative ESI mode perturbs fragmentation of the lower present [M − H]− ions, although the formation of FA 407

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

accuracy lower than 5 ppm. An example of an MS/MS spectrum is displayed in Figure 3, representing the fragmentation pattern of CER[NP]C45. Fragmentation spectra and fragmentation mechanisms for all other classes are included in Supporting Information (S-4 for standard CERs and S-5 for CERs in skin SC extracts). Identified CERs are plotted in Figure 2b. CER[NP] is the most abundant CER class, covering 22.10% of all identified CERs. The distribution of CER classes in the upper layer of human skin is shown in Figure 4 (note

that potential differences in ionization efficiency between CER classes have not been taken into account). The total chain length distribution for each CER is listed in Figure 2c as well, presenting a larger carbon chain distribution than ever reported. Practically all CER classes linked with saturated FAs show a partition from C32 to C54, while CERs coupled with α-hydroxy FAs were generally present from C32 to C52. The relative distribution of CER species within each class appoints CERs with a total chain length of C42, C44, C46, and C48 as the most abundant species (S-6, Supporting Information). CERs with an odd number of carbon atoms are also present but in a lower concentration compared to CERs with an even number of carbon atoms. A very small fraction of the CER content (i.e., 1.33%) could be identified as CERs coupled with ω-hydroxy FAs. The ω-hydroxy FAs are known as long chain FAs, which is confirmed by the higher carbon atom number from C28 up to C34.10 CER[ODS] is the only CER class which is not detected in the data set. This is not surprising since CER[EODS] species are as well detected at low intensities. All esterified CER classes showed a distribution from C62 to C76, except for CER[EOH] which was detected only up to C73. The most abundant CER[EO] species contain a total carbon atom number of 66, 68, 70, and 72. Fragmentation spectra of CER[EO] in negative ESI mode revealed only linoleic acid as the ester-linked FA in the CER[EO] classes, as is exemplified for CER[EODS]C68 in S-5, Supporting Information. However, Hinder et al. recently described that other FAs can be esterified in the ω-OH-position with chain lengths from C17:2 to C20:2.23 Analysis of the FAs following CER hydrolysis highlights these species at low intensities (see further). The ratio of CERs with an even number of carbon atoms to CERs with an odd number of carbon atoms was 65/35.

Figure 4. Pie chart representing the distribution of identified CER classes in human stratum corneum (potential differences in ionization efficiency have not been taken into account). Some classes are appointed as letters: a, CER[ADS] 1.63%; b, CER[EODS] 0.40%; c, CER[OP] 0.17%; and d, CER[EOP] 1.14%.

Figure 5. Fragmentation spectrum of a yet undiscovered CER class annotated as dihydroxy dihydrosphingosine or dihydroxy sphinganine and abbreviated as CER[NT]. The presented MS/MS spectrum results from a CER[NT]C46 at a collision energy of 35 eV. The proposed structure of the new SB is shown. Due to skeletal isomerism, CER[NT]C46 can be defined as the result of CER species CER[N(28:0)T(18)], CER[N(26:0)T(20)], and CER[N(24:0)T(22)]. 408

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

Figure 6. (a) FA distribution and (b) SB distribution obtained after hydrolysis of CER skin extracts. The signal intensity of each species is scaled to unit variance within the class. The size of the circles is proportional with the abundance of the hydrolysis products. [(1) E(18:2)O(30:0); (2) O(30:0); (3) A(26:0); (4) A(24:0); (5) N(18:2); (6) N(18:1); (7) N(16:1); (8) N(28:0); (9) N(26:0); (10) N(24:0); (a) T(18); (b) H(18); (c) P(18); (d) S(18); (e) DS(26)].

higher in CERs containing saturated FAs (N) than in CERs linked to α-hydroxy FAs or ω-hydroxy FAs. This skeletal isomerism only enlarges the already described structural heterogeneity of CERs and the number of CER species present in the upper layer of human skin. A lot of reproducible features could not be linked to CER classes reported in literature. As these nonidentified features show typical CER MS/MS fragmentation spectra, confirming their CER nature, potentially undiscovered CER classes are detected in the profiling method. One potentially new CER class has been unraveled (Figure 5). The new class contains 4 hydroxyl groups on the base, as can be clearly derived from the loss of four water molecules in the fragmentation spectrum (identified as O′, O″, O″′, and O″″), both of the molecular ion as the SB fragment. Annotations of the different fragments and their mass accuracy are represented in S-7, Supporting Information. This new SB building block can as such be annotated as dihydroxy-dihydrosphingosine or dihydroxysphinganine (abbreviated as T because of two hydroxyl groups on the sphingosine base). The FA part consists of saturated FAs (N). The base in conjugation with hydroxy FAs has not been observed. This new CER class is displayed in Figure 2b as well. Further in-depth confirmation of the exact location of the different hydroxyl groups on the SB is necessary by complementary techniques. Glucosylceramides have been described earlier as major precursors of several stratum corneum ceramides.29 These species have not been detected in the SC extracts. All identified CERs have been listed in an Accurate Mass Retention Time (AMRT) CER library for targeted data processing, containing nomenclature, molecular formula, accurate mass, and retention time.30 On the basis of this constructed database (.csv format), the Agilent MassHunter Qualitative Analysis Find By Formula option allows extracting the signal of individual CER species in all samples, within a given mass accuracy and retention time window. The exclusion of a peak height cutoff in the targeted data processing enabled

CER class isomerism due to different positioning of hydroxyl groups is unraveled in the RP-LC step (see Figure 1c and Figure 2). Apart for CERs with either α- or ω-hydroxy FAs, isomers can readily be discriminated on the basis of MS/MS information (see Figure S-5, Supporting Information.). Note that, in the absence of accurate mass information, true isomers CER[NH] and CER[AS] will be accompanied by CER[NDS] with one additional CH2 group only displaying a mass difference of approximately 50 ppm. All this emphasizes the importance of combining high resolution LC with accurate mass spectrometry and subsequent MS/MS fragmentation. Besides this class isomerism, MS/MS fragmentation spectra revealed several structural isomers varying in carbon length of SB and FA building blocks within each CER species (i.e., due to skeletal/chain isomerism). Figure 3 represents the fragmentation spectrum of CER[NP] with total carbon number of 45. The peak originating from CER[NP]C45 is actually composed out of (at least) 8 different skeletal isomers, being CER[N(29:0)P(16)], CER[N(28:0)P(17)], CER[N(27:0)P(18)], CER[N(26:0)P(19)], CER[N(25:0)P(20)], CER[N(24:0)P(21)], CER[N(23:0)P(22)], and CER[N(22:0)P(23)]. Previously reported fragmentation mechanisms have been confirmed for most CER classes on the basis of accurate mass of the observed fragments. As the [M − H2O+H]+ ion was fragmented for CER[NH], the MS/MS spectrum in negative ESI mode is also shown to confirm its identification (Figure S-5, Supporting Information.). Some potentially new fragmentation mechanisms are observed, e.g., for CER[AH], appointed as a loss of ammonia in the 6-hydroxy sphingosine part. The presence of CER subspecies with an odd number of carbon atoms in more than one chain, which recently has been described for CER[EOS] and CER[NP], can be extended for all other CER classes (see representative fragmentation spectra of CER[NDS]C48, CER[ADS]C34, CER[AS]C34, CER[OS]C52, CER[EOS]C66, CER[AP]C42, CER[EOP]C70, CER[NH]C42, CER[AH]C42, and CER[EOH]C70 in Figure S-5, Supporting Information).23 The observed skeletal isomerism is apparently 409

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

presence of its base T with carbon atom number ranging from 16 to 22. Apart for DSs, which contains a higher number of carbon atoms, C18 and C20 are the predominant species. The observed ratio of even chain SBs to odd chain SBs is approximately 3 to 1. Platform Performance from an Omics Perspective. Next to obtaining a detailed insight into the CERs present in skin, the ultimate goal of the platform is to apply it to address biological questions, this by revealing CER up- or downregulation in comparative experiments. A profiling method relies on high repeatability of the analytical method used in order to detect subtle biological differences between samples. To calculate the precision of the analysis, both a triplicate injection of the same skin extract (defines LC-MS repeatability) and a triplicate sampling of the same individual (defines extraction repeatability) were tested for the negative ESI mode procedure. Features were extracted according to the protocol described in the experimental part. The relative standard deviation (RSD%) was calculated for features that have 100% frequency, yielding 1368 features for the LC-MS replicates and 1172 features for the extraction replicates. 71.68% and 62.12% of the detected features had an RSD < 15% in the LC-MS and extraction replicates, respectively, which confirms the repeatability of the CER profiling method. The full RSD distribution of features with 100% frequency in both LC-MS replicates and extraction replicates is displayed in S-9, Supporting Information. However, these figures of merit are an underestimation of the true repeatability, as probably not all features that are detected by LC-MS are CER species. In that perspective, a targeted data processing covering only CERs using an AMRT library is much more reliable and time-saving compared to an untargeted data processing approach. This targeted data processing extracted 650 CERs (adducts included; representing 226 unique CERs) with 100% frequency out of the extraction replicates, of which 72.92% showed an RSD < 15%; (note the deviation from the 264 unique CERs reported earlier that resulted from an LC-MS triplicate experiment). The resulting RSD distribution is shown in S-9, Supporting Information, which clearly illustrates the benefit of targeted data processing versus untargeted data processing in a profiling method. On the other hand, new CER classes will be missed if a targeted approach is used. The combined use of both targeted and untargeted data processing is hereby justified.

the identification of more CERs (although present in very low amounts). All identified CERs are presented in Figure 2b and comprise a number of 264 unique CER species. However, when all skeletal isomers are taken into account, the figure of 264 is truly outnumbered and the number of unique CER species in skin SC can be estimated in the range of a thousand unique molecular species. Analysis of FA and SB Building Blocks Following CER Hydrolysis. To confirm the diversity of the FA groups and SBs, the CERs have been split up in their building blocks by breaking the amide bond. Although no enzymes are commercially available to catalyze this reaction, acid hydrolysis of CERs has been reported.31 In our hands, this procedure (described in S-8, Supporting Information) resulted in a limited hydrolysis yield (10−60% depending upon the CER class). Nevertheless, interesting qualitative information could be generated. To prevent interference from free FAs, cholesteryl ester, and triacylglycerols associated FAs, hydrolysis of CERs was performed on SPE purified skin SC extracts. The SPE extraction yield of the standard CERs ranged from 80% to 100%. Free FAs are not eluted under the given conditions and stay retained on the SPE cartridge, while cholesteryl esters and triacylglycerols are completely removed by the washing step (data not shown). LC-MS analyses covering the FAs and SBs individually have been performed on the hydrolyzed skin CER extracts. Figure 6 plots the retention time against m/z for all building blocks that have been identified. The FA analysis revealed the presence of 53.66% of saturated FAs, 3.42% of monounsaturated FAs, 1.63% of diunsaturated FAs, 41.30% of α-hydroxy FAs, and 9.00% of ω-OH FAs. The resulting within class FA distribution correlates well with the distribution of FAs found at the intact CER level (see above). C24 to C28 are the most abundant chain lengths observed in saturated and α-OH FAs, while the ratio of even chain FAs to odd chain FAs is approximately 4 to 1. ω-OH FAs can only be detected as long chain FAs with a carbon atom number from C28 to C34. This is again in accordance with observations at an intact CER[EO] level. ω-OH FAs can be discriminated from α-OH FAs on the basis of their reduced retention (Figure 6a).19−21 The presence of other esterified FA in CER[EO], previously described by Hinder et al. (see above), can be inferred from these experiments.23 Linoleic acid (N(18:2)) represented about 90.9% of the diunsaturated FAs, while traces of other FAs that can be esterified on the ω-OH-position were detected (N(16:2), 0.2%; N(20:2), 4.6%; N(22:2), 1.2%). The contribution of diunsaturated FAs to the overall FA pool is unexpectedly low. It is worthwhile to mention that these FA species are present in an ester linkage instead of an amide linkage. Under the given conditions, hydrolysis might proceed at a different rate. Consequently, ω-OH FAs esterified with linoleic acid have been found intact as well (i.e., the complete FA part from CER[EO]) from E(18:2)O(28:0) up to E(18:2)O(34:0). It is remarkable that monounsaturated FAs are observed (e.g., N(18:1)). Although their origin could not be inferred from experiments performed at an intact CER level, these FA might be presented as esterified FAs in CER[EO], as Wertz et al. claimed before.32 Oleic acid linked to ω-OH FAs (E(18:1)O) could not be detected in the partial hydrolysate. GC/MS analysis of the FA part corroborated all acquired LCMS results (data not shown).33 The SB analysis shows a partitioning of bases in decreasing order: P 43.23%, DS 37.52%, S 10.86%, T 4.64%, and H 3.75%. The new CER class CER[NT] could be confirmed here by the



CONCLUSIONS An analytical platform based on RP-LC Q-TOF MS for profiling and characterization of SC CERs is described. The high-resolution RP-LC enabled the separation of practically all CER classes, including CERs present as position isomers due to the different location of a hydroxyl group. MS/MS experiments were performed to confirm the accurate mass based identifications of all CER classes and to distinguish the position isomers. Furthermore, fragmentation of CER species unraveled several skeletal isomers eluting as one peak in the RP-LC, differing in the combination of SB and FA chain length. Both even- and odd-numbered chains were present in the CER building blocks. The hydrolysis of CERs confirmed the observed patterns within the intact CER analysis. Also, the base of the newly identified CER[NT] showed up after hydrolysis, which corroborates its presence among stratum corneum CERs. The repeatability of the analysis enabled us to build an AMRT CER library containing the molecular formula, accurate mass, and retention time of all identified CER species. 410

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411

Analytical Chemistry

Article

(15) Masukawa, Y.; Tsujimura, H.; Narita, H. J. Lipid Res. 2006, 47, 1559−1571. (16) Masukawa, Y.; Narita, H.; Sato, H.; Naoe, A.; Kondo, N.; Sugai, Y.; Oba, T.; Homma, R.; Ishikawa, J.; Takagi, Y.; Kitahara, T. J. Lipid Res. 2009, 50, 1708−1719. (17) Houjou, T.; Yamatani, K.; Imagawa, M.; Shimizu, T.; Taguchi, R. Rapid Commun. Mass Spectrom. 2005, 19, 654−666. (18) Sandra, K.; dos Santos Pereira, A.; Vanhoenacker, G.; David, F.; Sandra, P. J. Chromatogr., A 2010, 1217, 4087−4099. (19) Lin, J.; McKeon, T. A.; Stafford, A. E. J. Chromatogr., A 1995, 699, 85−91. (20) Santiago-Vazquez, L. Z.; Mydlarz, L. D.; Pavlovich, J. G.; Jacobs, R. S. J. Chromatogr., B 2004, 803, 233−236. (21) Eberhard, J.; Jepsen, S.; Albers, H. K.; Acil, Y. Anal. Biochem. 2000, 280, 258−263. (22) Masukawa, Y.; Tsujimura, H. Lipids 2007, 42, 275−290. (23) Hinder, A.; Schmelzer, C. E. H.; Rawlings, A. V.; Neubert, R. H. H. Skin Pharmacol. Physiol. 2011, 24, 127−135. (24) Hsu, F.; Turk, J. J. Am. Soc. Mass Spectrom. 2002, 13, 558−570. (25) Hsu, F.; Turk, J.; Stewart, M. E.; Downing, D. T. J. Am. Soc. Mass Spectrom. 2002, 13, 680−695. (26) Camera, E.; Picardo, M.; Presutti, C.; Catarcini, P.; Fanali, S. J. Sep. Sci. 2004, 27, 971−976. (27) Ann, Q.; Adams, J. Anal. Chem. 1993, 65, 7−13. (28) Lee, M. H.; Lee, G. H.; Yoo, J. S. Rapid Commun. Mass Spectrom. 2003, 17, 64−75. (29) Hamanaka, S.; Hara, M.; Nishio, H.; Otsuka, F.; Suzuki, A.; Uchida, Y. J. Invest. Dermatol. 2002, 119, 416−423. (30) Ding, J.; Sorensen, C. M.; Jaitly, N.; Jiang, H.; Orton, D. J.; Monroe, M. E.; Moore, R. J.; Smith, R. D.; Metz, T. O. J. Chromatogr., B 2008, 871, 243−252. (31) Johnson, S. B.; Brown, R. E. J. Chromatogr. 1992, 605, 281−286. (32) Wertz, P. W.; Swartzendruber, D. C.; Madison, K. C.; Downing, D. T. J. Invest. Dermatol. 1987, 89, 419−425. (33) Bicalho, B.; David, F.; Rumplel, K.; Kindt, E.; Sandra, P. J. Chromatogr., A 2008, 1211, 120−128.

The library can be used in the MassHunter Qualitative Analysis software to integrate CER species in new studies in an automated manner. The data can as well be subjected to untargeted profiling to detect differential entities not listed in the library. The developed CER platform can be a valuable tool in skin type classification, providing cosmetic companies leads for customized skin products matching the clients’ skin profile. Furthermore, it can be useful for tracking biomarkers for various skin diseases.



ASSOCIATED CONTENT S Supporting Information * Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION Corresponding Author *E-mail: [email protected]. Fax: +32 56 20 48 59.



ACKNOWLEDGMENTS The authors thank Steve Fischer (Agilent Technologies, Santa Clara, CA) and Andris Jankevics (Institute of Molecular, Cell and Systems Biology, University of Glasgow, Glasgow, UK) for their valuable input. This research is sponsored by the Flemish agency for Innovation by Science and Technology (IWT).



REFERENCES

(1) Masukawa, Y.; Narita, H.; Shimizu, E.; Kondo, N.; Sugai, Y.; Oba, T.; Homma, R.; Ishikawa, J.; Takagi, Y.; Kitahara, T.; Takema, Y.; Kita, K. J. Lipid Res. 2008, 49, 1466−1476. (2) Van Smeden, J.; Hoppel, L.; Van der Heijden, R.; Hankemeier, T.; Vreeken, R. J.; Bouwstra, J. A. J. Lipid Res. 2011, 52, 1211−1221. (3) Weerheim, A.; Ponec, M. Arch. Dermatol. Res. 2001, 293, 191− 199. (4) Lampe, M. A.; Burlingame, A. L.; Whitney, J.; Williams, M. L.; Brown, B. E.; Roitman, E.; Elias, P. M. J. Lipid Res. 1983, 24, 120−130. (5) Mizutani, Y.; Mitsutake, S.; Tsuji, K.; Kihara, A.; Igarashi, Y. Biochimie 2009, 91, 784−790. (6) Holleran, W. M.; Takagi, Y.; Uchida, Y. FEBS Lett. 2006, 580, 5456−5466. (7) Ishikawa, J.; Narita, H.; Kondo, N.; Hotta, M.; Takagi, Y.; Masukawa, Y.; Kitahara, T.; Takema, Y.; Koyano, S.; Yamazaki, S.; Hatamochi, A. J. Invest. Dermatol. 2010, 130, 2511−2514. (8) Pruett, S. T.; Bushnev, A.; Hagedorn, K.; Adiga, M.; Haynes, C. A.; Sullards, M. C.; Liotta, D. C.; Merrill, A. H. J. Lipid Res. 2008, 49, 1621−1639. (9) Haynes, C. A.; Allegood, J. C.; Park, H.; Sullards, M. C. J. Chromatogr., B 2009, 877, 2696−2708. (10) Farwanah, H.; Pierstorff, B.; Schmelzer, C. E.; Raith, K.; Neubert, R. H.; Kolter, T.; Sandhoff, K. J. Chromatogr., B 2007, 852, 562−570. (11) Stewart, M. E.; Downing, D. T. J. Invest. Dermatol. 1995, 105, 613−618. (12) Fahy, E.; Subramanian, S.; Brown, H. A.; Glass, C. K.; Merrill, A. H. Jr.; Murphy, R. C.; Raetz, C. R. H.; Russell, D. W.; Seyama, Y.; Shaw, W.; Shimizu, T.; Spener, F.; van Meer, G.; VanNieuwenhze, M. S.; White, S. H.; Witztum, J. L.; Dennis, E. A. J. Lipid Res. 2005, 46, 839−861. (13) Quehenberger, O.; Armando, A. M.; Brown, A. H.; Milne, S. B.; Myers, D. S.; Merrill, A. H.; Bandyopadhyay, S.; Jones, K. N.; Kelly, S.; Shaner, R. L.; Sullards, C. M.; Wang, E.; Murphy, R. C.; Barkley, R. M.; Leiker, T. J.; Raetz, C. R. H.; Guan, Z.; Laird, G. M.; Six, D. A.; Russell, D. W.; McDonald, J. G.; Subramanian, S.; Fahy, E.; Dennis, E. A. J. Lipid Res. 2010, 51, 3299−3305. (14) Levy, M.; Futerman, A. H. IUBMB Life 2010, 62, 347−356. 411

dx.doi.org/10.1021/ac202646v | Anal. Chem. 2012, 84, 403−411