Long and Very Long Lamellar Phases in Model Stratum Corneum Lipid Membranes

Membrane models of the stratum corneum (SC) lipid barrier, either healthy or affected by recessive X-linked ichthyosis, constructed from ceramide (Cer; non-hydroxyacyl sphingosine CerNS24 alone or with omega- O- acylceramide CerEOS), free fatty acids (C16-24), cholesterol (Chol) and cholesteryl sulfate (CholS) were investigated. X-ray diffraction (XRD) revealed a previously unreported polymorphism of the membranes. In the absence of Cer EOS, the membranes formed a short lamellar phase (SLP; the repeat distance d = 5.3 nm), a medium lamellar phase (MLP; d = 10.6 nm) or very long lamellar phases (VLLP; d = 15.9 and 21.2 nm). An increased CholS-to-Chol ratio modulated the membrane polymorphism, although the CholS phase separated at ≥ 7 weight % (of total lipids). The presence of CerEOS led to the stable long lamellar phase (LLP) with d = 12.2 nm and prevented VLLP formation. Our XRD results agree well with recently published cryo-electron microscopy data for vitreous skin sections, while also revealing new structures. Thus, lamellar phases with long repeat distances (MLP and VLLP) may be formed in the absence of omega- O- acylceramide, whereas these ultralong Cer species likely stabilize the final SC lipid architecture of LLP by riveting the adjacent lipid layers.


INTRODUCTION
The stratum corneum (SC) has evolved to protect the body from desiccation and thus to ensure the terrestrial life of mammals, including humans (1). This outermost skin layer also hampers the entry of possibly harmful substances from the environment. SC consists of several layers of cornified cellscorneocytes, and extracellular lipid matrix, which represents the major skin permeability barrier (2). The SC extracellular lipid matrix consists mainly of ceramides (Cer), free fatty acids (FFA) and cholesterol (Chol) at an ~1:1:1 molar ratio (Supplemental Figure S1). These lipids are highly organized in the lamellae aligned parallel to the skin surface (3,4). Skin Cer are a class of sphingolipids comprising, to date, 15 classes of free Cer including the ultralong ω-O-acylceramides (ω-O-acylCer), which contain 30-34C acyl chains with a linoleic acid ester-linked to the ω-hydroxyl terminus (5)(6)(7)(8)(9), for a review, see (10).
The lipid matrix of isolated SC has a lamellar characteristic and a long repeat distance (d) of ~13 nm, as was revealed by electron microscopy (11)(12)(13) and/or by X-ray diffraction (XRD) (12,14,15). The so-called long periodicity phase with d ~13 nm was reconstructed in vitro from isolated skin lipids (16)(17)(18) and lipids synthesized in a laboratory (19). A separated Chol with d ~3.4 nm and a short periodicity phase with d ~6.4 nm were found to coexist with the long periodicity phase in isolated SC by XRD (20).
Chol also separates in vitro in model SC lipid membranes, creating a distinct phase that is detectable by XRD (21,22). A key role of ω-O-acylCer for formation of the long periodicity phase was demonstrated in vitro in model lipid membranes with isolated pig skin Cer (18). A correlation was found ex vivo between a decreased fraction of ω-O-acylCer and changes in the XRD patterns of SC in healthy humans (23) and in patients with atopic dermatitis (24). Several models of the molecular arrangement of the SC extracellular lipid domains have been proposed; however, they remain under discussion (25)(26)(27)(28).
In this work, we investigated the lamellar organization in a membrane model of recessive Xlinked ichthyosis (RXLI). RXLI is a genetic skin disease with impaired processing of CholS to Chol. In healthy skin, CholS is gradually distributed across the SC, from 5 weight % at the stratum granulosum/SC boundary to 1 weight % in the outer SC (29,30). In RXLI, mean CholS levels in the SC are ~5-10-fold by guest, on April 28, 2019 www.jlr.org Downloaded from elevated, whereas free Chol levels are ~50% reduced in comparison to healthy skin (4). The disruption of CholS desulfation leads to an abnormal permeability barrier, abnormal desquamation, and the presence of nonlamellar domains within the SC extracellular spaces (31,32). A high CholS level in model lipid mixtures induces the formation of an additional less ordered phase (33) and alterations in the electron spin probe microenvironment (34). Rehfeld et al. proposed that different H-bonding of CholS relative to Chol may play a role in RXLI pathogenesis (34). However, the contributions of increased CholS and decreased Chol to the abnormalities in lipid organization and permeability in RXLI are unclear.
Thus, we aimed at studying the effects of CholS on SC lipid membranes based on N-tetracosanoyl D-erythro-sphingosine (CerNS24). During the initial experiments, we observed an extraordinary and yet unreported polymorphic behavior of our SC lipid membranes, which was modulated by the Chol/CholS content variation and sample preparation method. Here, we show, for the first time, that SC lipids without ω-O-acylCer can self-organize in patterns with not only a short repeat distance but also with medium or very long repeat distances. The formation of these structures is further compared with the structural behavior of the SC model containing N- (32-

Dropping at 50 °C:
The 17-60 µL of 2:1 hexane/96% ethanol (v/v) was added per 1 mg of dry lipid. The mixture was homogenized using an ultrasound bath at 50 °C until a homogeneous transparent viscous liquid was created. The concentration depended on the solubility of each lipid sample. The liquid, which was continuously heated to 50 °C, was applied in several subsequent steps on a wafer preheated to 50 °C using a 100-µL gastight syringe preheated to 50 °C. Before each application, the syringe and wafer were preheated again.
The prepared lipid membranes were dried overnight under a vacuum over P 4 O 10 and solid paraffin in a desiccator and then annealed according to the following two protocols: Annealed at 90 °C: The samples were heated to 90 °C, equilibrated at this temperature for 10 min and slowly (~3 h) cooled to room temperature.

Annealed at 70 °C/H 2 O:
The samples underlaid with aluminum rings were sealed in aluminum containers with distilled water at the bottom (water was not in contact with the lipids) and heated to 70 °C, equilibrated at this temperature for 10-30 min and slowly (~3 h) cooled to room temperature.
The prepared membranes were not specifically hydrated and were stored at 2-6 °C. Before the XRD measurements, they were equilibrated at room temperature for 24 h.

X-ray diffraction (XRD)
The XRD data were collected at ambient room temperature and humidity with an X'Pert PRO θ-θ

Polymorphism of CerNS24-based model membranes: formation of MLP and VLLP in the absence of ω-O-acylCer
First, we constructed simple SC lipid models using CerNS24, FFA (16)(17)(18)(19)(20)(21)(22)(23)(24) and Chol in equimolar ratios with 1 and 5 weight % of CholS (samples labeled 1/0.03 and 1/0.13, respectively, Table 1). The lipids were deposited by spraying and were annealed at 90 °C (17). The XRD pattern of the 1/0.13 sample The MLP formation was further reproduced using a modified preparation protocol (Figure 1c). To avoid the overlap of Chol and MLP reflections, the Chol content was decreased to 45%, and no CholS was used (0.45/0 sample). The membrane was prepared by dropping the lipid solution at 50 °C instead of spraying at room temperature, followed by annealing at 70 °C/H 2 O for 10 min. This sample enabled us to resolve the 3 rd order MLP peak from Chol (only two shoulders at the 3 rd and 6 th order MLP reflections indicated separated Chol) to further confirm this lamellar structure.  Table ST1).
Analysis revealed that all these reflections belonged to a lamellar phase as the 1/d dependence on the order (h) was clearly linear (Figure 2b). These reflections were assigned the orders III v , IV v , V v …up to XXX v and they provided a very long d = 15.9 nm. This d was three times longer than d SLP , and we denoted this lamellar phase as a very long lamellar phase (VLLP 15.9 , subscripts "v" in the order numbers mark the assignment to VLLP). The formation of VLLP 15.9 instead of MLP in these membranes was most likely induced by the longer annealing at 70 °C/H 2 O. The region of the short-range arrangement contained a weak reflection at the repeat distance of 0.41 nm and an insufficiently resolved reflection at the repeat distance of 0.37 nm.
Chol was reduced to 45% to prevent an overlap with the LLP reflections. The sample was prepared by dropping at 50 °C and annealing once or twice at 70 °C/H 2 O for 10 min or 10 and 20 min, respectively. In the XRD patterns of this sample ( Figure 5), we detected reflections of the orders II, III, IV… up to XIV, providing LLP with d = 12.2 nm after the first (Supplemental Table ST1) and second annealing. There was only a small indication that an additional structure, most likely VLLP, formed after the second annealing (Figure 5b). Weak reflections of separated Chol and very weak reflections of the orthorhombic polymethylene chain packing (unresolvable in Figure 5a) were also found.

DISCUSSION
The XRD and electron microscopy results have provided essential information about the SC lipid barrier structure, i.e., that the dominant repeating unit is ~13 nm long (14,15,42) and contains repeating patterns of broad/narrow/broad electron-lucent bands bordered by electron-dense segments (25). A recent study using cryo-electron microscopy of vitreous skin sections without staining showed a different pattern, consisting of narrow (4.5 nm) and broad (6.5 nm) lucent bands, together revealing an asymmetric repeating unit of 11 nm (10-12 nm) (26).
Until recently, ω-O-acylCer were thought to be essential for the formation of lamellar lipid phases with long d, as only complex model membranes containing ω-O-acylCer formed a long periodicity phase with d ~ 13 nm along with other phases (18,43). Here, we report, for the first time, that CerNS24-based model membranes in the absence of ω-O-acylCer form not only the SLP but also MLP and VLLP and that their lamellae can be arranged with even longer repeat distances than in membranes with ω-O-acylCer.
The formation of MLP has recently also been observed in model lipid systems based on CerNH (44) (26). The broad electron-lucent band in the SC micrographs was 6.5 nm wide, which is 0.44 nm longer than the calculations based on our XRD data. If we subtract the width of the sphingosine domain (4.54 nm) from the LLP (d = 12.2 nm) in the membrane with CerEOS, we obtain a length of 7.66 nm, which is more than 1 nm larger than the broad lucent band in the SC micrographs reported by Iwai et al. (26). However, the CerEOS fraction in our model was increased to 30 molar % (of total Cer pool) because this composition ensures LLP formation without other lamellar phases. The physiological content of ω-O-acylCer is approximately 10 molar % of the total Cer fraction (47). Thus, if we assume a linear relationship between the average lipid chain length and d, we can The mechanism of the CholS effect on the SC model polymorphism is unknown and requires further studies. In addition, CholS separation was already detected at 7 weight % CholS and became more pronounced at 14 weight % CholS. These abilities of increased CholS concentrations to modulate the SC lipid architecture and to phase separate appear to be consistent with the altered organization of the SC barrier in RXLI patients (31,32). CholS applied topically onto the isolated murine stratum corneum caused extensive nonlamellar domain formation and disruption of the lamellar membrane structure (31). Lamellar-phase separation was also detected in RXLI patients (32). The phase-separation of CholS in our model could be relevant to the nonlamellar domains and phase-separation in the skin, which is attributed to the skin barrier abnormality. However, the pathophysiology of RXLI patient skin is more complex, comprising the inhibition of serine proteases by CholS and, consequently, diminished degradation of the corneodesmosomes that bind corneocytes, leading to abnormal desquamation (48). Furthermore, CholS acts as an acidifier of the SC interstices in RXLI patients relative to the pH in healthy individuals (49).
Increased Ca 2+ levels in the SC of RXLI patients also contribute to corneocyte retention by increasing corneodesmosome and interlamellar cohesion (32). Although our model simulated some important