Five ceramide synthases (CerS2–CerS6) are expressed in mouse skin. Although CerS3 has been shown to fulfill an essential function during skin development, neither CerS6- nor CerS2-deficient mice show an obvious skin phenotype. In order to study the role of CerS4, we generated CerS4-deficient mice (Cers4−/−) and CerS4-specific antibodies. With these biological tools we analysed the tissue distribution and determined the cell-type specific expression of CerS4 in suprabasal epidermal layers of footpads as well as in sebaceous glands of the dorsal skin. Loss of CerS4 protein leads to an altered lipid composition of the sebum, which is more solidified and therefore might cause progressive hair loss due to physical blocking of the hair canal. We also noticed a strong decrease in C20 1,2-alkane diols consistent with the decrease of wax diesters in the sebum of Cers4−/− mice. Cers4−/− mice at 12 months old display additional epidermal tissue destruction due to dilated and obstructed pilary canals. Mass spectrometric analyses additionally show a strong decrease in C20-containing sphingolipids.
CerS (ceramide synthases) form a family of six distinct proteins. They catalyse N-acylation of (dihydro-) sphingosine to (dihydro-) ceramide. Specificity towards acyl-CoA chain length and a distinct expression pattern are characteristic features of this class of enzymes [1,2]. CerS proteins are located in the endoplasmic reticulum and harbour at least five transmembrane domains [3,4]. The catalytically active region is the so-called lag1 motif [5–7]. Shorter acyl-CoAs (C14:0/C16:0) are mainly processed by CerS5 and CerS6. CerS1 and CerS4 were described to show specificity towards C18 and C20 respectively . CerS2 catalyses the synthesis of ceramides with an acyl chain length of C22–C24, whereas CerS3 generates ultra-long-chain ceramides [1,8]. Ceramides are the building blocks of complex sphingolipids such as sphingomyelin or glycosylceramide. They fulfill distinct roles in signal transduction with different functions depending on their acyl chain length [9,10] and they are precursors of other signalling molecules such as ceramide-1-phosphate [11,12]. Four strains of CerS-deficient mice have been described so far [13–18]. Among these, only the CerS3-deficient mice show a skin-specific phenotype , although CerS2 and CerS6 are also expressed in skin. CerS3-deficient mice die shortly after birth due to a defect in the transepidermal water barrier. The barrier function is implemented by a specific composition of lipids . Altered lipid metabolism in the skin can also lead to abnormalities in the pilo-sebaceous unit, which may cause alopecia, i.e. hair loss. Many mouse mutations have been generated or have arisen spontaneously that cause alopecia due to a plethora of diverse alterations, some of which are involved in the production of skin lipids . Malfunctions in the sebaceous glands lead to alterations in the composition of the sebum. The FA2H (fatty acid 2-hydroxylase) protein, for example, is required for the synthesis of 2-hydroxylated sphingolipids . Deficiency in this enzyme alters the biochemical and physical properties of the sebum and additionally leads to cycling alopecia .
In the present study we have generated and characterized CerS4-deficient mice (Cers4−/−). In order to differentiate between knockout and wild-type tissue and to study the tissue- and cell type-specific expression pattern of the endogenous CerS4 protein, we generated CerS4-specific antibodies. We found that CerS4 was expressed in the suprabasal cell layer of the epidermis and also in the sebaceous glands. The Cers4−/− mice suffer from progressive hair loss due to alterations in the composition of the sebum, which appears more viscous. Thus the sebum blocks and dilates the hair shaft leading to a partial loss of hair growth, an increased size of the hair cone and a progressive form of alopecia.
All animal experiments performed were in line with the guidelines of the corresponding local and state authorities. The mice were kept under standard housing conditions, including a 12 h light/dark cycle with food and water ad libitum. If not mentioned otherwise, for tissue preparation animals were anaesthetized and transcardially perfused with PBS as described previously . Generation of Cers4−/− mice is described in the Supplementary Online Data (at http://www.biochemj.org/bj/461/bj4610147add.htm).
CerS activity assay
The assay was performed with epidermis and dermis of 20-week-old mice. The back skin of the mice was incubated with DispaseII (Sigma–Aldrich, working solution 2 units/ml in PBS) overnight to separate dermis from epidermis . Afterwards the resulting parts of the skin were frozen in liquid nitrogen and ground in a porcelain mortar. The assay itself was performed with nitrobenzoxadiazole sphingosine as described previously .
Immunofluorescence and immunoblotting analyses
Protein lysates for immunoblotting analyses and protein determination were carried out as described previously . For protein separation via gel electrophoresis, 25 μg of each lysate (dermis and epidermis, 50 μg) in urea buffer (5% SDS, 9 M urea, 0.04 M Tris/HCl, 0.01% Bromophenol Blue, 1 mM EDTA and 5% 2-mercaptoethanol, pH 6–8) was separated by SDS/PAGE (10% gels). Proteins were blotted for 1 h with a transfer buffer (20 mM TrisBase, 150 mM glycine and 20% methanol) at 100 V on to nitrocellulose membrane (Hybond-ECL, GE Healthcare). To prevent unspecific binding of the antibodies, membranes were blocked with 5% (w/v) non-fat dried skimmed milk powder in TBS-T (TBS containing 0.1% Tween) for 1 h. CerS2-specific antibodies were used as described previously . Anti-CerS4 antibodies, raised against two C-terminally located peptide sequences of mouse CerS4 protein (LHKGQMTEDIRSDVEEPD and DDEPVSEGPQLKNGMAR) by Pineda Antibody Service (Berlin, Germany), were used at a 1:500 dilution for immunoblotting at 4°C overnight, followed by three washing steps with TBS-T. Incubation with secondary antibodies was performed for 1 h at room temperature [horseradish peroxidase-conjugated antibodies (1:10 000 dilution; Dianova)]. Signals were detected using the SuperSignal West Pico Chemiluminescence detection kit (Pierce) and membranes were developed using the VersaDoc Imaging System.
For immunofluorescence analyses, we used either cryosections of frozen tissues or paraffin-embedded tissue. For each cryosection, frozen tissues were sectioned (14 μm, Leica CM3050 S), mounted on to Superfrost plus slides (Menzel-Gläser), fixed with 4% paraformaldehyde for 5 min and washed three times (50 mM Tris/HCl, 1.5% NaCl and 0.3% Triton X-100, pH 7.6). Sections were blocked using a 4% BSA, 1% NGS (normal goat serum) and 0.3% Triton X-100 solution for 1 h at room temperature and incubated with the appropriate antibodies [anti-CerS4 (1:300 dilution), anti-CerS6 (1:500 dilution), anti-K14 (Santa Cruz Biotechnology; 1:50 dilution), anti-Ki67 (CloneTEC3; DakoCytomation; 1:100 dilution), anti-loricrin (BAbCO, U.S.A.; 1:500 dilution), anti-fillagrin (BAbCO; 1:500 dilution) or anti-K1 (Covance; 1:500 dilution)] at 4°C overnight. Secondary Alexa Fluor® 488- or Alexa Fluor® 546-labelled antibodies (Invitrogen) were incubated for 1 h at room temperature. Then sections were mounted with Mounting Medium (Dako). Immunofluorescence staining of paraffin-embedded sections for FA2H was performed as described in .
Histological analyses of the skin
Animals were killed by cervical dislocation and skin biopsies were taken from the thoracic region near the dorsal midline. The tissue was fixed by immersion in Bouin's solution for 4 h at room temperature and then incubated in 70% ethanol overnight. The biopsies were embedded in methylmethacrylate Technovit 7100 (Heraeus Kulzer) according to the manufacturer's recommendations.
Lipid extraction and tandem MS
Lipid extraction was performed as described previ-ously [13,17,18]. Sphingolipids were dissolved in methanol/chloroform/300 mM ammonium acetate (665:300:35, by vol.) , internal standards were added and sphingolipids were analysed using Q-TOF (quadrupole–time-of-flight) MS [17,18].
A wax diester fraction was isolated by immersing killed mice in acetone (20 min). After evaporation under nitrogen, the skin surface lipids were dissolved in n-hexane and purified on a silica column (Phenomenex). Wax diesters were eluted with hexane/diethylether (95:5, v/v), and further purified by TLC and lipid isolation from the silica material. Subsequently, the wax diesters were cleaved by methanolysis in chloroform/0.3 M methanolic NaOH (1:1, v/v) at 42°C for 4 h. Water was added to a final concentration of 3:3:2.4 (v/v/v). The lipophilic phase was purified on an aminopropyl column (Chromabond; Macherey-Nagel). The fatty acid methyl ester and 1,2-alkane diol fractions were then analysed using an Agilent 7890 GC (gas chromatograph) with mass spectrometer. Fatty acid methyl esters were measured directly, whereas 1,2-alkane diols were derivatized using MSTFA [N-methyl-N-(trimethylsilyl)trifluoroacetamide] prior to measurement.
TLC analyses of sebum lipids
Mice were killed by cervical dislocation and total surface lipids were isolated as described above. Lipids were analysed by TLC on silica gel 60 HPTLC plates (Merck) using toluene/n-hexane (1:1, v/v) as described previously . The following lipid standards were added to identify the different lipid species: cholesteryl palmitate, oleic acid behenyl ester, triacylglycerol mix (C8:0–C16:0) and cholesterol (all from Sigma). Lipids were visualized using cupric sulfate in aqueous phosphoric acid as described previously . The temperature range in which the isolated surface lipids change from a solid to fluid appearance was determined as described in . Briefly, the acetone surface lipid extract was filtered through glass wool and dried under nitrogen. After evaporation, the lipids were taken up in a small volume of n-hexane and transferred into melting point capillaries. The ‘melting temperature’ was determined with a capillary type melting point apparatus (Buchi Melting Point Model 545; Büchi Labortechnik).
Data were analysed by means of repeated measures ANOVA and Student's t test. The P values were considered to be significant when P<0.05 (*), P<0.01 (**) and P<0.001 (***).
The genome locus of Cers4 is approximately 33 kbp in length. The first exon is not translated. Therefore, in order to achieve a total loss of CerS4 protein expression, we decided to exchange the coding region (E2-11; 8.2 kbp) with an NLS (nuclear localization sequence)-lacz reporter gene. The splice acceptor site of intron 1 remained as a part of the 5′ homology region.
The final vector consists of the NLS-lacz reporter gene and a downstream located neomycin resistance cDNA, which is flanked by frt sites and under the control of the SV40 promoter. These building blocks are surrounded by a 5 kb 5′ homology region and a 1.8 kb 3′ homology region (Supplementary Figure S1A at http://www.biochemj.org/bj/461/bj4610147add.htm). Transfection of HM1 embryonic stem cells with the vector led to homologous recombination  into the genomic locus of CerS4. The resulting G418-resistant clones were then tested via Southern blot and PCR analyses (results not shown).
Chimaeras were obtained by injecting correctly recombined embryonic stem cell clones into mouse blastocysts and bred with C57BL/6 mice. The heterozygous offspring were backcrossed first with deleter Flp mice to lose the neomycin resistance cDNA and afterwards bred consecutively with C57BL/6 to obtain a genetic background of at least 87.5%. To verify correct insertion into the genomic locus, we performed PCR and Southern blot analyses (Supplementary Figures S1B and S1C).
The loss of the whole coding region of the CerS4 protein was presumed to result in a loss (or significant reduction) of activity towards the CerS4-specific acyl-CoA substrate in Cers4−/− mouse tissue. Therefore we performed CerS activity assays as described previously [17,28]. CerS activity assays with tissue lysates of epidermis and dermis confirmed a high activity towards C20 acyl-CoAs in wild-type lysates. We observed a highly significant decrease in the activity to approximately 10% in the epidermis and to approximately 20% in the dermis of Cers4−/− mice (Supplementary Figure S1D).
It has been shown that Cers4 mRNA is nearly ubiquitously expressed [1,2]. To analyse the organ-specific expression at the protein level and to show that the CerS4 protein is no longer expressed in Cers4−/− mice, we performed immunoblotting analyses. Therefore we generated antibodies against two C-terminally located peptides of the protein. As a control for the specificity of the newly derived antibodies, we generated CerS4-overexpressing HEK (human embryonic kidney)-293 cells.
With these affinity-purified antibodies, the CerS4 protein was detected in immunoblotting analyses in all tested tissues, whereas no band at the corresponding size was observed in the tissue from Cers4−/− animals. Strong CerS4 expression was found in the spleen, lung, liver and heart (Supplementary Figure S2 at http://www.biochemj.org/bj/461/bj4610147add.htm).
Cers4−/− mice suffer from a progressive hair loss starting at approximately 7 weeks of age. Heterozygous mice did not show any obvious phenotypic abnormalities. Although there is no obvious difference on P10 (postnatal day 10) between wild-type and Cers4−/−, starting around P50 the Cers4−/− mice appear to have spiky hair (Figure 1). The phenotype worsened in further stages when mice start losing their hair, beginning mainly around the neck, but also at the ventral abdomen. Not every Cers4−/− mouse displays a similar degree of hair loss. Bald regions appear at the dorsal and ventral sides of some of the mice at approximately 5 months of age, and the hair loss progresses in very old mice (Figure 1).
Progressive hair loss in
Immunofluorescence analyses of the skin
We studied the cell-type-specific expression of CerS4 in the skin to elucidate the cause for the hair loss. To do this, we performed immunofluorescence analyses with the newly generated antibodies. Strong signals were observed in the sebaceous glands, whereas no signals were observed in Cers4−/− mice (Figures 2A–2F). Sections were prepared from the back skin of mice. Additionally we found CerS4 mainly expressed in the suprabasal cell layer of the epidermis atop of the basal cell layer indicated by the expression of K14 (Figures 2G–2L) in the paws of mice.
Immunofluorescence analyses of the skin
Immunohistochemically, the expression of CerS4 protein in murine skin is restricted to the stratum spinosum and stratum granulosum of the epidermal layers and the sebaceous gland. To study the effect of the loss of functional CerS4 protein on the development of alopecia, we undertook a histological investigation of the dorsal skin of Cers4−/− mice and wild-type controls during early and late phases of life when the alopecia was evident. During hair follicle morphogenesis, the mutant and wild-type control skin biopsies showed no difference with respect to hair follicle development (staging according to ) (Figures 3A–3D). The area occupied by the sebaceous glands in 15 tissue sections on P10 was 20% larger in Cers4−/− mice than the corresponding area of phenotypically wild-type controls (Figures 3C and 3D), but not different when compared with adult (P50) wild-type sections (results not shown). The number of layers in the epidermis was not conspicuously altered. By P53 (Figures 3E–3H), however, functional defects of the sebaceous glands in Cers4−/− mice were evident, as the pilary canals were dilated and filled with sebum plugs (Figure 3H). Frozen sections stained with Oil Red O (Figures 3I–3L) revealed that these plugs were composed of lipids derived from the sebaceous glands (Figures 3K and 3L). In the wild-type pilary canal, there was no accumulation of Oil Red O-positive material (Figures 3I and 3J), presumably because the sebum forms a thin lipid film on the surface of the hair shaft and the epidermis. It is also important to note that, after the end of the first hair cycle on P53 (according to ), when alterations of the sebaceous gland function were evident, no increased cellularity was seen in the dermis of Cers4−/− mice, which would be indicative of an inflammatory response (Figures 2G and 2H). Thus development of the Cers4−/− skin phenotype begins during the first hair cycle with the accumulation of sebum lipids in the pilary canals, which causes their dilation and can presumably block the exit of hair shafts.
Early stages of alopecia in
Cers4−/− dorsal skin
On P86 during the second cyclic anagen, hair follicles in both Cers4+/+ and Cers4−/− skin were in anagen V (according to ), indicating that the lack of functional CerS4 does not affect the stem cells in the bulge region (Figures 4A–4C). The dilation of the mutant pilary canal by sebum plugs had increased to an extent that interposed tissue was pinched (Figures 4E and 4F) and tissue remnants were expulsed into the stratum corneum (Figure 4E), whereas the wild-type hair follicles were partly in anagen and partly in telogen (Figure 4D). The increased thickness of the epidermis in these specimens may represent a reaction to these skin ulcerations. Since loss of a CerS might lead to alterations in proliferation, we additionally investigated the expression of the proliferation marker Ki67 in the skin, but found no significant alterations (Supplementary Fig-ure S3 at http://www.biochemj.org/bj/461/bj4610147add.htm). We obtained similar results by immunohistochemical analyses with antibodies against phospho-histone 3, another known proliferation marker (results not shown). Differentiation in the epidermis was also investigated, using the markers fillagrin and loricrin, with no obvious alterations between Cers4−/− and wild-type epidermis (Supplementary Figure S4 at http://www.biochemj.org/bj/461/bj4610147add.htm).
Late stages of alopecia in
Cers4−/− dorsal skin
Skin biopsies taken from 1-year-old mice showed telogen hair follicles in the wild-type (Figures 4G and 4J) and dystrophic hair follicles in the Cers4−/− mutant (Figures 4H, 4I, 4K and 4L). Melanocytes with active melanogenesis were associated with epithelial structures or scattered in the dermis (Figure 4K), which is typical of dystrophic catagen . Thus, by the end of the first year of life, the skin of Cers4−/− mice is characterized by tissue destruction as a consequence of dilated and obstructed pilary canals and due to dystrophic catagen. In skin sections of one 2-year-old Cers4−/− mouse, hair follicles were rare and associated with only a few sebocytes, whereas collagenous scar tissue predominated (Figures 4I and 4L).
Lipid profile of the skin
To analyse the consequences of the loss of the CerS4 protein in more detail, we performed mass spectrometric analyses of dermal and epidermal lipids of Cers4−/− and wild-type animals (n=4 for each group, P143). In the epidermis of the Cers4−/− mice we observed a strong decrease in the relative C20 ceramide levels as expected from previous results regarding the specificity of CerS4 (Figure 5A) . Additionally we found a surprisingly strong relative decrease in C18 ceramide in the epidermis of Cers4−/− animals compared with their wild-type controls. These results were not seen in the dermis (Figure 5B). The strong relative decrease in C18 and C20 ceramides of the epidermis appeared to be compensated by a significant relative increase in very-long-chain ceramides such as C24, h24:1 and h24 (Figure 5A).
Mass spectrometric analyses of neutral sphingolipids of epidermis and dermis
Both epidermis and dermis of Cers4−/− mice showed a severe decrease in the relative amounts of C18- and C20-containing hexosylceramides (Figures 5C and 5D). These losses were compensated by increases in hexosylceramides with very long chain fatty acids such as h24 in the epidermis and C22 in the dermis. The strong decreases in the relative levels of C20-containing sphingomyelin levels were similar between epidermis and dermis as well as the increase in short chain sphingomyelin (C16) (Figures 5E and 5F).
Since CerS2 is described to have a substrate specificity overlapping with that of CerS4  and since we observed an increase in CerS2-specific substrates, especially regarding the long-chain ceramides in the epidermis, we wondered whether an increase of CerS2 is responsible for the compensational increase of very-long-chain ceramides. We decided to investigate the expression pattern of CerS2 in both tissues of the skin in comparison with CerS4 expression. Therefore we analysed dermis and epidermis separately by immunoblotting analyses and found that CerS2 is not expressed in the epidermal, but in the dermal layer of the skin. Regarding the relative expression level we found no alterations of CerS2 in Cers4−/− mice (Figure 6). To complete the analyses we also checked for the expression level of CerS6 and found it expressed in both layers of the skin; however, the relative expression levels also remained unchanged in Cers4+/+ compared with Cers4−/− mice (Figure 6).
Immunoblot analyses of CerS4 protein expression in epidermis and dermis
In histological analyses we showed that the properties of the sebum of Cers4−/− animals were altered in comparison with wild-type. Thus we analysed the sebum lipid composition. Total surface lipids, most of which are derived from sebum , were extracted by dipping killed mice into acetone and the lipid extracts were examined by HPTLC (Figure 7A). Densitometric analysis showed a significant decrease in the amount of wax diesters in Cers4−/− mice, which was accompanied by a significant increase in wax monoesters and cholesterol (Figure 7B). To test the hypothesis that these changes in the sebum composition affect its viscosity, we determined the temperature range in which the sebum lipid extract changes from a solid to a fluid appearance (‘melting temperature’). Surface lipids from wild-type mice started to become fluid at 36.0±0.2°C (mean±S.D., n=3) and were completely fluid at 42.4±0.5°C. In contrast, lipids from Cers4−/− mice started to become fluid at 40.2±1.6°C (mean±S.D., n=3) and were completely fluid at 52.4±1.0°C (Figure 7C). The changes in the composition of surface lipids and the increased ‘melting temperature’ of the lipid extract are reminiscent of changes observed in FA2H-deficient mice . We therefore explored the possibility of an FA2H deficiency in sebaceous glands of Cers4−/− mice. However, immunofluorescence staining of P10 dorsal skin sections from wild-type and Cers4−/− mice showed the presence of FA2H in both genotypes (Supplementary Figure S5 at http://www.biochemj.org/bj/461/bj4610147add.htm).
Alterations in sebum composition of
Since the sebum composition was drastically altered, we performed further mass spectrometric analyses of the neutral sphingolipids of the sebum. However, we observed only minor alterations (Figure 8A). Therefore we additionally investigated the constituents of the strongly decreased wax diester fraction. We found no obvious alterations of the fatty acid composition of wax diesters (Figure 8B). However, we have shown for the first time that the major diol of the wax diester fraction in the sebum of wild-type mice is the C20-1,2-alkane diol. In Cers4−/− mice the relative amount of the C20-1,2-diol was significantly reduced, whereas C16-1,2-diol was increased (Figure 8B).
Neutral sphingolipids and wax diesters in sebum
Mammalian CerS differ from each other by their specificity towards acyl-CoAs of different chain length and by their cell-type-specific expression pattern. To study the biochemical and biological functions of CerS4, we generated a CerS4-deficient mouse line. To do this we exchanged the coding region of the endogenous Cers4 gene with the NLS-lacz reporter cDNA via homologous recombination. The enzymatic assays performed confirmed the decrease of activity towards C20 acyl-CoA. Thus we concluded that the CerS4 protein was functionally ablated. Despite the strong loss of activity, some residual activity towards C20 acyl-CoA in Cers4−/− mice remained and the relative amount of C20 ceramide in the dermis was not altered. The CerS2 protein, which might be capable of synthesizing C20 ceramide, was detected in the dermis of wild-type mice by immunoblotting, but the amount of CerS2 protein in Cers4−/− mice remained unaltered. These findings were in contrast with the epidermal fraction, where the relative amounts of C18 and C20 ceramides were strongly decreased in accordance with the activity assays, and CerS2 was not detected by immunoblotting analyses. Therefore CerS2 may be capable of compensating for the loss of ceramide production by CerS4 if expressed in the same tissue.
Further results, i.e. the loss of C18- and C20-containing sphingomyelin or hexosylceramide in the dermis, strongly suggest that, although CerS2 is able to compensate in part for the loss of CerS4 regarding ceramide production, this is not sufficient for all sphingolipids with the corresponding chain lengths. Another observation was the significant increase in C16-containing sphingomyelin in both dermal and epidermal layers. Thus we also checked for the expression pattern of CerS6, but found no alterations in the protein level in Cers4−/− mice. We conclude that the increase in the amount of sphingolipids with longer and shorter chain length is not due to a compensatory increase in the protein amount of other CerS, but due to the accumulation of free sphingosine not used for ceramide synthesis in Cers4−/− mice.
By immunofluorescence analyses we investigated the localization of the protein in the different skin layers. CerS4 is expressed in the stratum spinosum and the stratum granulosum of the epidermis and is most prominent in the sebaceous glands of the dermis. The Cers4−/− phenotype described in the present paper involves alterations of the sebaceous glands and hair follicle dystrophy that causes a progressive scarring alopecia. The earliest morphological alteration identified in Cers4−/− mice skin was an increase in the size of the sebaceous glands on P10, when the pelage of the mutant was inconspicuous. This may indicate increased proliferation of the sebocyte precursor cells, delayed maturation of the terminally differentiating sebocytes or even increased allocation of Lrig1 (leucine-rich repeats and immunoglobulin-like domains protein 1)-expressing stem cells of the hair germ to the sebocyte lineage . Enlargement of the sebaceous glands in Cers4−/− mice may reflect an alteration of the lipid metabolism in sebocytes due to the Cers4 deficiency, but similar enlargements of the sebaceous glands have been observed in other mouse mutants harbouring gene defects affecting lipid production, such as Elovl3 , FA2H , Acbp (acyl-CoA binding protein) , palmitoyl acyl transferase Zdhhc21 (zinc finger DHHC domain-containing protein 21) creating the depilated phenotype , and a transgenic mouse with keratinocyte-specific ablation of the genes Insig1 (insulin-induced gene 1) and Insig2, which regulate cholesterol homoeostasis . Furthermore, mice with mutations in genes encoding proteases such as Ctsl (cathepsin L) [38,39] and proteins involved in signal transduction such as Plcd1 (phospholipase Cδ1)  have also been reported to exhibit enlarged sebaceous glands. Thus enlargement of sebaceous glands may not hint at a specific defect, but merely represent a sensitive indicator of malfunction of the developing pilo-sebaceous unit.
However, we did find sebum plugs in the pilary canals of Cers4−/− mice. These plugs not only widened the opening and presumably blocked the exit of hair shafts, but apparently bruised skin tissue between adjacent pilary canals, which was subsequently expelled on to the epidermal surface. This suggests that at least for some time a skin wound was induced, which would explain the reactive thickening of the epidermis which is not due to alterations in the differentiation and may have opened an entry for microbes from the skin surface. Although we did not find prominent inflammatory infiltrates in the mutant skin at any of the time points examined, we cannot exclude that the final dystrophic catagen  observed in 1-year-old Cers4−/− mice at least contributed to inflammatory tissue destruction. In this context, Cers4−/− mice resemble flake mutant mice carrying a missense mutation of Scd1 (stearoyl-CoA desaturase 1) with respect to onset and spread of alopecia . This phenotype is supposed to evolve from the absence of anti-bacterially active lipids due to the defective sebaceous gland [41,42]. However, even in those skin sections exhibiting dystrophic catagen dystrophy in 1-year-old Cers4−/− mice, sebaceous glands were preserved, contradicting the principle of ‘sebaceous glands first’ in the development of scarring alopecia . Only in the single 2-year-old Cers4−/− mouse examined was scarring alopecia evident and the remaining hair follicles showed only a few associated sebocytes.
Previously, there has been growing awareness for the importance of lipids in hair follicle formation and cycling, not only the lipids derived from the sebaceous glands, but also regarding lipids in the various layers of the human hair [44,45]. Since intrinsic lipids of both the hair shaft and the inner root sheath include roughly 25% ceramides , at least the late phases of hair follicle dystrophy in the Cers4−/− phenotype could be attributed to lipid-specific defects in the inner root sheath or hair shaft itself. However, neither immunohistochemistry using CerS4-specific antibodies nor whole-mount lacZ staining to visualize Cers4 promoter activity provided evidence for CerS4 protein localization in the pilo-sebaceous unit other than the sebaceous glands. The importance of sebum secretion for normal hair follicle development was emphasized by studies on the asebia (Scd1) mutant and mice with skin-specific ablation of Scd1 that show a marked hypoplasia of the sebaceous glands [46–48]. Unlike asebia-type mouse mutants, Cers4−/− mice did not show hair shaft malformations at weaning, which can lead to trichogranuloma and inflammatory foreign body reactions. Thus the Cers4−/− mouse is similar to the asebia-type mouse mutants, but apparently follows a different path towards the final stage of scarring alopecia.
However, with respect to the formation of sebum plugs, the Cers4−/− phenotype highly resembled the alopecia caused by functional inactivation of FA2H , which is exclusively expressed in the sebaceous glands. In the latter, CerS4 expression in the skin was increased (which at least in part can be explained by the enlarged sebaceous glands) and no significant changes in the levels of non-hydroxylated and omega-hydroxylated hexosylceramides were observed . However, FA2H deficiency resulted in the absence of 2-hydroxylated C20-hexosylceramides and a strong reduction in 2-hydroxylated C22- to C24-hexosylceramides . Interestingly, there were no changes in the amount of free ceramide species (including 2-hydroxylated ceramides). In the dermis of Cers4−/− mice, we found similar results, mainly decreased levels of C20-hexosylceramides and C20-sphingomyelin. Only the C20-ceramide levels in the dermis remained unaltered.
The similar phenotype with respect to the sebaceous glands and sebum of CerS4- and FA2H-deficient mice may therefore indicate a specific role for normal sebum production of (C20-) hexosylceramide (or sphingomyelin), which is strongly reduced in both mouse mutants. In line with these findings, both mouse mutants share another interesting feature. Investigations of the sebum composition in Cers4−/− mice revealed a strong increase in wax monoesters, a decrease in type II wax diesters and a higher melting point of the sebum. The strong decrease in wax diesters, in combination with impaired secretion from sebaceous glands and progressive hair loss, is shared by another mouse mutant, which is deficient in DGAT1 (acyl CoA:diacylglycerol acyltransferase 1) , the enzyme that synthesizes type II wax diesters . Therefore the amount of the type II wax diester seems to be critical for the occurrence of alopecia in these mice.
Our measurements of the wax diester components revealed a strong decrease in the amount of C20 1,2-alkane diol in the Cers4−/− mice in contrast with the amount of C20-fatty acids, which was not conspicuously altered. In the current literature it is assumed that FA2H uses free fatty acids as a substrate, which are used as a substrate in the FA2H in vitro enzymatic assay . However, other putative substrates (e.g. sphingolipids) have, to our knowledge, not been tested. Studies in yeast suggested that ceramide is the substrate for the yeast FA2H homologue . If this holds true in mammals, the specific reduction in C20 alkane-1,2-diol in Cers4−/− mice could be due to the decrease in C20-containing sphingolipids, especially C20-hexosylceramide, a possible substrate for FA2H .
The C20-fatty acid residue of these sphingolipids might no longer be hydroxylated to a sufficient amount by FA2H. According to this hypothesis, breakdown of hydroxylated ceramides by ceramidases releases hydroxylated fatty acids, which are finally reduced to long-chain 1,2-alkane diols. In the absence of CerS4, synthesis of C20 1,2-alkane diols is therefore strongly decreased. Despite the strong decrease in C20 1,2-alkane diols, there is still a residual amount which might be due to a possible dual function of FA2H; FA2H, as discussed above, was suggested to be capable of using free fatty acid as a substrate.
In the present study we demonstrate that CerS4 is a ubiquitously expressed protein whose loss in mice results in alopecia. We conclude that CerS4 or its lipid products play a crucial role in the generation of proper sebum by producing C20 sphingolipids, which possibly are precursors for C20 1,2-alkane diols required for wax diester synthesis. Further studies with this mouse mutant will shed light on to the role of CerS4 for the generation of proper sebum and also the pathogenesis of scarring alopecia.
Klaus Willecke and Philipp Ebel conceived and designed the study. Philipp Ebel, Silke Imgrund, Katharina vom Dorp, Helena Maier, Helena Drake, Matthias Eckhardt, Thomas Franz and Kristina Hofmann acquired the data. Philipp Ebel, Matthias Eckhardt, Thomas Franz and Peter Dörmann analysed and interpreted the data. Philipp Ebel, Thomas Franz and Matthias Eckhardt drafted the paper. Klaus Willecke, Peter Dörmann, Matthias Eckhardt, Thomas Franz, Katharina vom Dorp, Silke Imgrund, Philipp Ebel and Joachim Degen critically revised the paper prior to submission.
We thank Professor Maarten Egmond (University of Utrecht, Utrecht, The Netherlands) for synthesis and purification of the immunogenic CerS4 peptides. Also the authors are grateful for the expert technical assistance of G. Eversloh, B. Blanck and M. Michels. The excellent technical assistance of Brita Wilhelm and Helga Peisker is also gratefully acknowledged.
This study was supported by the German Research Foundation through the Collaborative Research Centre (SFB-645) [project numbers B2 (to K.W.), B5 (to M.E.) and Z4 (to P.D.)].