The Cornified Envelope-Bound Ceramide Fraction is Altered in Patients with Atopic Dermatitis.

The lipids covalently bound to the cornified envelope of corneocytes are essential for the human skin barrier (Swartzendruber et al., 1987Swartzendruber D.C. Wertz P.W. Madison K.C. Downing D.T. Evidence that the corneocyte has a chemically bound lipid envelope.J Invest Dermatol. 1987; 88: 709-713Abstract Full Text PDF PubMed Scopus (303) Google Scholar). Together with the unbound lipids, they form a matrix regulating the permeability for water and pathogens through the stratum corneum (SC). Three main lipid classes are observed in the SC: sterols, fatty acids, and ceramides. Bound ceramides are a subset of the unbound ceramides and have an ultra-long omega-hydroxyl chain (OCer) (Farwanah et al., 2007Farwanah H. Pierstorff B. Schmelzer C.E. Raith K. Neubert R.H. Kolter T. et al.Separation and mass spectrometric characterization of covalently bound skin ceramides using LC/APCI-MS and nano-ESI-MS/MS.J Chromatogr B Analyt Technol Biomed Life Sci. 2007; 852: 562-570Crossref PubMed Scopus (51) Google Scholar, Hill et al., 2006Hill J. Paslin D. Wertz P.W. A new covalently bound ceramide from human stratum corneum -omega-hydroxyacylphytosphingosine.Int J Cosmet Sci. 2006; 28: 225-230Crossref PubMed Scopus (19) Google Scholar). They are biosynthetically related to unbound linoleate esterified omega-hydroxyl ceramides (EOCers). Ceramide binding is selective, where ceramides with an unsaturated acyl chain (MuCers) and shorter chains are preferentially bound (Boiten et al., 2019Boiten W. Helder R. van Smeden J. Bouwstra J. Selectivity in cornified envelop binding of ceramides in human skin and the role of LXR inactivation on ceramide binding.Biochim Biophys Acta Mol Cell Biol Lipids. 2019; 1864: 1206-1213Crossref PubMed Scopus (9) Google Scholar). Yet, as both bound and unbound ceramides originate from the same precursors, compositional changes of precursors can affect both. Previously, changes in the unbound ceramide composition of patients with atopic dermatitis (AD) were observed. These changes correlated to decreased barrier function and altered lipid organization (Ishikawa et al., 2010Ishikawa J. Narita H. Kondo N. Hotta M. Takagi Y. Masukawa Y. et al.Changes in the ceramide profile of atopic dermatitis patients.J Invest Dermatol. 2010; 130: 2511-2514Abstract Full Text Full Text PDF PubMed Scopus (241) Google Scholar, Janssens et al., 2012Janssens M. van Smeden J. Gooris G.S. Bras W. Portale G. Caspers P.J. et al.Increase in short-chain ceramides correlates with an altered lipid organization and decreased barrier function in atopic eczema patients.J Lipid Res. 2012; 53: 2755-2766Crossref PubMed Scopus (300) Google Scholar). Changes in bound ceramide composition were observed in autosomal recessive congenital ichthyosis and psoriasis and in atopic dogs and mice (Fujii et al., 2018Fujii M. Ohyanagi C. Kawaguchi N. Matsuda H. Miyamoto Y. Ohya S. et al.Eicosapentaenoic acid ethyl ester ameliorates atopic dermatitis-like symptoms in special diet-fed hairless mice, partly by restoring covalently bound ceramides in the stratum corneum.Exp Dermatol. 2018; 27: 837-840Crossref PubMed Scopus (7) Google Scholar, Grond et al., 2017Grond S. Eichmann T.O. Dubrac S. Kolb D. Schmuth M. Fischer J. et al.PNPLA1 deficiency in mice and humans leads to a defect in the synthesis of omega-O-Acylceramides.J Invest Dermatol. 2017; 137: 394-402Abstract Full Text Full Text PDF PubMed Scopus (67) Google Scholar, Popa et al., 2011Popa I. Remoue N. Hoang L.T. Pin D. Gatto H. Haftek M. et al.Atopic dermatitis in dogs is associated with a high heterogeneity in the distribution of protein-bound lipids within the stratum corneum.Arch Dermatol Res. 2011; 303: 433-440Crossref PubMed Scopus (19) Google Scholar, Wertz et al., 1989Wertz P.W. Madison K.C. Downing D.T. Covalently bound lipids of human stratum corneum.J Invest Dermatol. 1989; 92: 109-111Abstract Full Text PDF PubMed Scopus (178) Google Scholar). It is hypothesized that the bound ceramide composition of patients with AD deviates and contributes to AD pathology. To our knowledge, no detailed liquid chromatography–mass spectrometry analysis of bound ceramides in patients with AD has been reported. Here, we applied a quantitative liquid chromatography–mass spectrometry method to SC samples of patients with AD (Boiten et al., 2016Boiten W. Absalah S. Vreeken R. Bouwstra J. van Smeden J. Quantitative analysis of ceramides using a novel lipidomics approach with three dimensional response modelling.Biochim Biophys Acta. 2016; 1861: 1652-1661Crossref PubMed Scopus (39) Google Scholar). Samples were obtained after institutional approval and written informed consent, according to the Declaration of Helsinki. From eight healthy volunteers and 16 patients with AD, a 1.6-cm diameter nonlesional site on the ventral forearm was sequentially tape-stripped. A second tape-strip sample was obtained from lesional sites at the ventral forearm of eight patients with AD. Tapes were measured with a SquameScan (Heiland Electronic, Wetzlar, Germany) to determine the amount of SC stripped. Both bound and unbound lipids were extracted from tapes 9–12, as described elsewhere (Boiten et al., 2019Boiten W. Helder R. van Smeden J. Bouwstra J. Selectivity in cornified envelop binding of ceramides in human skin and the role of LXR inactivation on ceramide binding.Biochim Biophys Acta Mol Cell Biol Lipids. 2019; 1864: 1206-1213Crossref PubMed Scopus (9) Google Scholar). Before tape-stripping, transepidermal water loss was measured to determine the skin barrier function. Liquid chromatography–mass spectrometry analysis did not show a decreased bound ceramide amount in AD, contrary to what others observed using thin layer chromatography (Macheleidt et al., 2002Macheleidt O. Kaiser H.W. Sandhoff K. Deficiency of epidermal protein-bound omega-hydroxyceramides in atopic dermatitis.J Invest Dermatol. 2002; 119: 166-173Abstract Full Text Full Text PDF PubMed Scopus (197) Google Scholar). Figure 1a depicts the total detected bound ceramide amount corrected by SquameScan. To examine the bound ceramide composition in more detail, ceramides were grouped using three distinguishing features of their molecular structures (Figure 1b): MuCers percentage, sphingosine-subclass (S-subclass) percentage, and mean carbon chain length (MCL). Supplementary Text S1 explains the linear mixed modeling. The models compare patients with AD to healthy controls and a lesional site to the same patient's nonlesional site. Figure 1c and Supplementary Table S1 show that patients with AD had a 6.0% increase of bound MuCers, which further increased by 7.1% at their lesional sites. EOCers had similar increases in MuCers. In the unbound ceramides of patients with AD, increased percentages of S-subclass ceramides were observed (Ishikawa et al., 2010Ishikawa J. Narita H. Kondo N. Hotta M. Takagi Y. Masukawa Y. et al.Changes in the ceramide profile of atopic dermatitis patients.J Invest Dermatol. 2010; 130: 2511-2514Abstract Full Text Full Text PDF PubMed Scopus (241) Google Scholar, Janssens et al., 2011Janssens M. van Smeden J. Gooris G.S. Bras W. Portale G. Caspers P.J. et al.Lamellar lipid organization and ceramide composition in the stratum corneum of patients with atopic eczema.J Invest Dermatol. 2011; 131: 2136-2138Abstract Full Text Full Text PDF PubMed Scopus (85) Google Scholar). Figure 1d and Supplementary Table S2 depict that in AD the bound S-subclass ceramides were increased with 4.5% and an additional 2.1% in lesional skin. Again, these changes were similar in unbound EOCers. Examining the chain length distribution, it was observed that at lesional sites, relatively more short chain ceramides were bound, thereby depleting short-chain EOCers. To compare MCLs of OCers and EOCers, 18 carbons were subtracted from the latter, compensating for the linoleic acid moiety. The MCLs of saturated and unsaturated ceramides differed significantly and were examined as separate parameters but were affected similarly by AD and at lesional skin (Figure 1e and Supplementary Table S3). At lesional sites, the MCL of bound ceramides significantly decreased; however, nonlesional sites showed no significant decrease. The changes in the MCL because of AD and lesional skin were inverse and significantly different between bound OCers and EOCers. Supplementary Figure S2 shows the saturated ceramide chain length distribution of bound OCers and EOCers. Although the overall amount of EOCers decreased, short-chain EOCers were further depleted. In AD, decreased amounts of EOCers are thought to play an role in decreasing barrier function (van Smeden et al., 2014van Smeden J. Janssens M. Gooris G.S. Bouwstra J.A. The important role of stratum corneum lipids for the cutaneous barrier function.Biochim Biophys Acta. 2014; 1841: 295-313Crossref PubMed Scopus (3) Google Scholar). It was examined if changes in the composition of bound ceramides coincided with changes in barrier function. The three distinguishing ceramide features were compared with transepidermal water loss. Table 1 shows the Pearson r and corresponding P-value (Supplementary Figure S2 graphically displays the correlations). The increased MuCer and S-subclass percentages correlated to reduced barrier function. Previously, it was demonstrated that unbound unsaturated lipids can contribute to a more disorganized lipid matrix, thereby decreasing barrier function (Mojumdar et al., 2014Mojumdar E.H. Helder R.W. Gooris G.S. Bouwstra J.A. Monounsaturated fatty acids reduce the barrier of stratum corneum lipid membranes by enhancing the formation of a hexagonal lateral packing.Langmuir. 2014; 30: 6534-6543Crossref PubMed Scopus (46) Google Scholar). Although bound ceramides already have an increased MuCer percentage compared with unbound ceramides, a further increased percentage of MuCers could have a destabilizing effect, increasing SC permeability. To examine the correlation with depletion because of shorter chain ceramide binding, the difference in MCL between the bound OCers and EOCers was calculated. Saturation was included as a separate parameter. The difference in saturated ceramide MCL between EOCers and OCers correlated to SC barrier function.Table 1Correlation Coefficients and P-Values of Bound Ceramide Composition Compared with Barrier FunctionFeatureTEWL (g/cm2/h)Pearson rP-valueMuCer (%)0.731<0.001S-subclass (%)0.692<0.001Difference MCL saturated (carbons)0.5610.001Difference MCL MuCer (carbons)0.2310.203Abbreviations: EOCer, linoleate esterified omega-hydroxyl ceramide; MCL, mean carbon chain length; MuCer, ceramide with an unsaturated acyl chain; OCer, omega-hydroxyl ceramide; S-subclass, sphingosine subclass; TEWL, transepidermal water lossFour different distinguishing features were used to describe the bound ceramide composition: MuCer percentage, the percentage of S-subclass ceramides, and the difference in MCL between the unbound EOCers and bound OCers. The MCL differences of saturated and unsaturated ceramides were examined separately, because MuCers had a higher chain length and an increased MuCer percentage would increase the MCL. Open table in a new tab Abbreviations: EOCer, linoleate esterified omega-hydroxyl ceramide; MCL, mean carbon chain length; MuCer, ceramide with an unsaturated acyl chain; OCer, omega-hydroxyl ceramide; S-subclass, sphingosine subclass; TEWL, transepidermal water loss Four different distinguishing features were used to describe the bound ceramide composition: MuCer percentage, the percentage of S-subclass ceramides, and the difference in MCL between the unbound EOCers and bound OCers. The MCL differences of saturated and unsaturated ceramides were examined separately, because MuCers had a higher chain length and an increased MuCer percentage would increase the MCL. In summary, the bound ceramide composition of patients with AD contained more MuCers and S-subclass ceramides in lesional and nonlesional sites than healthy SC and at lesional sites, the chain length was reduced. These alterations in composition correlated to a decreased barrier function. Although individual alterations correlated, this does not imply a direct causality. It is feasible that all changes in the bound and unbound lipid composition of patients with AD combined would compromise the integrity of the lipid organization and thereby decrease SC barrier function. Nonetheless, it is thought that the bound lipids function as a scaffold for the unbound lipids. The altered state of the bound ceramide composition could impair this scaffold function. Both changes in the synthesis of precursor lipids and changes in the binding process could have influenced the bound ceramide composition in AD (Boiten et al., 2019Boiten W. Helder R. van Smeden J. Bouwstra J. Selectivity in cornified envelop binding of ceramides in human skin and the role of LXR inactivation on ceramide binding.Biochim Biophys Acta Mol Cell Biol Lipids. 2019; 1864: 1206-1213Crossref PubMed Scopus (9) Google Scholar). Treatment in dogs and mice has shown to increase bound ceramides (Fujii et al., 2018Fujii M. Ohyanagi C. Kawaguchi N. Matsuda H. Miyamoto Y. Ohya S. et al.Eicosapentaenoic acid ethyl ester ameliorates atopic dermatitis-like symptoms in special diet-fed hairless mice, partly by restoring covalently bound ceramides in the stratum corneum.Exp Dermatol. 2018; 27: 837-840Crossref PubMed Scopus (7) Google Scholar, Popa et al., 2011Popa I. Remoue N. Hoang L.T. Pin D. Gatto H. Haftek M. et al.Atopic dermatitis in dogs is associated with a high heterogeneity in the distribution of protein-bound lipids within the stratum corneum.Arch Dermatol Res. 2011; 303: 433-440Crossref PubMed Scopus (19) Google Scholar), making bound ceramides a possible target for barrier repair treatment of AD. Datasets related to this article can be found at: https://dx.doi.org/10.17632/n48kv8ssy9.1 hosted at Mendeley data (Boiten et al., 2019Boiten W. Helder R. van Smeden J. Bouwstra J. Selectivity in cornified envelop binding of ceramides in human skin and the role of LXR inactivation on ceramide binding.Biochim Biophys Acta Mol Cell Biol Lipids. 2019; 1864: 1206-1213Crossref PubMed Scopus (9) Google Scholar, "EO and O ceramides, TEWL in atopic dermatitis stratum corneum"). Walter Boiten: http://orcid.org/0000-0001-9060-6222 Jeroen van Smeden: http://orcid.org/0000-0002-2728-2832 Joke Bouwstra: http://orcid.org/0000-0002-7123-6868 The authors state no conflict of interest. This research was financially supported by the Dutch Fountdation TTW (grant no. 12400). Conceptualization: WB, JB; Data Curation: WB; Formal Analysis: WB; Funding Acquisition: JB; Investigation: WB, JVS; Methodology: WB, JVS; Project Administration: JB, JVS; Resources: JB; Software: WB; Supervision: JB; Validation: WB; Visualization: WB; Writing - Original Draft Preparation: WB; Writing - Review and Editing: WB, JVS, JB Statistical analysis was performed using linear mixed modeling in SPSS, version 24 (IBM, Armonk, NY). Volunteers' and patients' numbers were set as random variable (1 to 24); this enabled a paired comparison of the nonlesional and lesional sites within a patient. In all models, atopic dermatitis (AD) was set as a fixed variable, being either 0 or 1, indicating if the skin was healthy or from a patient, respectively. Lesional was set as a fixed variable nested within AD: AD(lesional). Hereby, the model takes into account that changes because of lesions can only occur in patients. This gives a separate value that indicates if lesional sites were different from nonlesional. A fixed variable named EOCers was made; this indicates if there was a difference between the groups of bound omega-hydroxyl ceramides (OCers) and unbound linoleate esterified omega-hydroxyl ceramides (EOCers). For the statistical analysis of the percentage of ceramides with an unsaturated acyl chain, a model was made including all specifications described previously. Interaction terms between the fixed variables EOCers and AD and EOCers and the nested lesional effect were tested. These interactions describe if a change in ceramides with an unsaturated acyl chain because of AD or lesions was different in the EOCers compared with the OCers. Supplementary Figure S1 gives a schematic overview of this model. After running the model, the interactions had nonsignificant contributions. This showed that changes in ceramides with an unsaturated acyl chain because of AD and lesions were similar in bound OCers and EOCers. The interactions were excluded for the final model. The parameter estimates obtained with the model are shown in Supplementary Table S1. The intercept represents the healthy bound OCer group. The same model that was used for the percentage of ceramides with an unsaturated acyl chain was used for the percentage of sphingosine subclass ceramides. Again, a nonsignificant contribution of interaction was observed, indicating that the changes in sphingosine subclass percentage did not differ for bound OCers and EOCers. The parameter estimates observed after excluding the interactions are shown in Supplementary Table S2. To examine the mean carbon chain length (MCL), a fixed factor for saturation was added to the model. This indicated if there was a difference in MCL between the saturated and unsaturated ceramides. This gives an additional interaction and another level of three-way interactions. The interaction between AD and lesional skin and saturation was added and the interaction between saturation and EOCers was added. The three-way interactions between AD and lesional skin with EOCer and saturation were added. This describes if the change in MCL because of AD or lesions in either the bound OCers or EOCers was different because of saturation. By running the model, it was shown that the three-way interactions were not significant and they were excluded. The interactions of saturation with AD and lesional skin were nonsignificant and were excluded as well. Supplementary Table S3 depicts the results of the final model. The interactions between the EOCers and AD and EOCers and lesions were significant. This showed that the change in MCL because of AD and lesions was different in the bound OCers compared with the unbound EOCers. Furthermore, the interaction between EOCers and saturation was significant. This showed that the change in MCL between OCers and EOCers was different between saturated and unsaturated ceramides.Supplementary Figure S2Comparing the chain length distribution. The carbon chain length distribution of the bound OCers and unbound EOCers is depicted. For the EOCers, 18 carbons were subtracted to compensate for the linoleic acid. The amount is given in fmol per amount of stripped SC measured as SquameScan. The data is depicted as mean ± SD (healthy n = 8, nonlesional n = 16, lesional n = 8). EOCer, linoleate esterified omega-hydroxyl ceramide; OCer, omega-hydroxyl ceramide; SC, stratum corneum; SD, standard deviation; SQ, SquameScan.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Supplementary Figure S3Graphical depiction of correlations with barrier function. Each graph shows the TEWL compared with a different parameter. The line depicts a linear regression between the points indicating the correlation coefficient. EOCer, linoleate esterified omega-hydroxyl ceramide; MCL, mean carbon chain length; MuCer, ceramide with an unsaturated acyl chain; OCer, omega-hydroxyl ceramide; S-subclass, sphingosine subclass; TEWL, transepidermal water loss.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Supplementary Table S1Linear Mixed Model Output for MuCer %MuCer %EstimateP-value95% confidence intervalLowerUpperIntercept (Healthy bound)20.58<0.00116.5424.62EOCer−12.96<0.001−15.29−10.62AD6.030.0151.2710.80AD(Lesional)7.07<0.00013.9410.20Abbreviations: AD, atopic dermatitis; EOCer, linoleate esterified omega-hydroxyl ceramide; MuCer, ceramide with an unsaturated acyl chain.The table shows the output for the fixed variables of the linear mixed model. The estimate of the intercept and effect sizes is given. The P-value and 95% intervals are given to interpret the output. Open table in a new tab Supplementary Table S2Output of the LMM Used to Compare the Percentage of S-SubclassesS-Subclass %EstimateP-value95% confidence intervalLowerUpperIntercept (Healthy bound)71.43<0.00169.0273.84EOCer−20.18<0.001−21.66−18.70AD4.500.0031.677.32AD(Lesional)2.640.0100.674.61Abbreviations: AD, atopic dermatitis; EOCer, linoleate esterified omega-hydroxyl ceramide; LMM, linear mixed model; S-subclass, sphingosine subclass.The table shows the output for the fixed variables of the linear mixed model. The estimate of the intercept and effect sizes is given. The P-value and 95% intervals are given to interpret the output. Open table in a new tab Supplementary Table S3Output of the LMM for the MCL DataMCLEstimateP-value95% confidence intervalLowerUpperIntercept49.84<0.00149.6650.03AD−0.070.511−0.280.14AD(Lesional)−0.230.008−0.40−0.06EOCer0.80<0.0010.591.01Saturation2.09<0.0011.972.22Interactions AD*EOCer0.280.0160.050.51 AD(Lesional)*EOCer0.370.0020.150.60 Saturation* EOCer0.38<0.0010.200.57Abbreviations: AD, atopic dermatitis; EOCer, linoleate esterified omega-hydroxyl ceramide; LMM, linear mixed model; MCL, mean carbon chain length.The table shows the output for the fixed variables of the linear mixed model. The estimate of the intercept and effect sizes is given. The P-value and 95% intervals are given to interpret the output. Open table in a new tab Abbreviations: AD, atopic dermatitis; EOCer, linoleate esterified omega-hydroxyl ceramide; MuCer, ceramide with an unsaturated acyl chain. The table shows the output for the fixed variables of the linear mixed model. The estimate of the intercept and effect sizes is given. The P-value and 95% intervals are given to interpret the output. Abbreviations: AD, atopic dermatitis; EOCer, linoleate esterified omega-hydroxyl ceramide; LMM, linear mixed model; S-subclass, sphingosine subclass. The table shows the output for the fixed variables of the linear mixed model. The estimate of the intercept and effect sizes is given. The P-value and 95% intervals are given to interpret the output. Abbreviations: AD, atopic dermatitis; EOCer, linoleate esterified omega-hydroxyl ceramide; LMM, linear mixed model; MCL, mean carbon chain length. The table shows the output for the fixed variables of the linear mixed model. The estimate of the intercept and effect sizes is given. The P-value and 95% intervals are given to interpret the output.