The skin barrier: an extraordinary interface with an exceptional lipid organization

Abstract

The barrier function of the skin is primarily located in the stratum corneum (SC), the outermost layer of the skin. The SC is composed of dead cells with highly organized lipid lamellae in the intercellular space. As the lipid matrix forms the only continuous pathway, the lipids play an important role in the permeation of compounds through the SC. The main lipid classes are ceramides (CERs), cholesterol (CHOL) and free fatty acids (FFAs). Analysis of the SC lipid matrix is of crucial importance in understanding the skin barrier function, not only in healthy skin, but also in inflammatory skin diseases with an impaired skin barrier. In this review we provide (i) a historical overview of the steps undertaken to obtain information on the lipid composition and organization in SC of healthy skin and inflammatory skin diseases, ii) information on the role CERs, CHOL and FFAs play in the lipid phase behavior of very complex lipid model systems and how this knowledge can be used to understand the deviation in lipid phase behavior in inflammatory skin diseases, iii) knowledge on the role of both, CER subclasses and chain length distribution, on lipid organization and lipid membrane permeability in complex and simple model systems with synthetic CERs, CHOL and FFAs, iv) similarity in lipid phase behavior in SC of different species and complex model systems, and vi) future directions in modulating lipid composition that is expected to improve the skin barrier in inflammatory skin diseases.

1. INTRODUCTION

1.1. Human skin

The physical skin barrier resides in the upper layer of the skin, the stratum corneum (SC). The SC consists of corneocytes, dead keratinized cells, embedded in a lipid matrix. This lipid matrix is highly organized and the only continuous pathway of externally applied chemicals in the skin. Therefore chemicals always have to cross the intercellular lipid regions. This makes the lipid matrix an important contributor to the skin barrier function [1]. The main lipid classes in this matrix are ceramides (CERs), cholesterol (CHOL) and free fatty acids (FFAs). No phospholipids are present in the SC [2, 3]. Due to the small head groups of the CERs and FFAs and the long hydrocarbon chains, strong hydrophobic forces exist in this lipid matrix [4]. These forces play a prominent role in the tight packing of the lipids in this three-dimensional arrangement and therefore form crystalline lipid lamellae.

The thin SC layer (thickness 10–30 μm) is the final product of a differentiation process of keratinocytes that takes place in the viable epidermis (thickness 80–200 μm) underlying the SC. Together the SC and viable epidermis, which are both avascular, form the epidermis. This process of differentiation is only possible due to the intimate contact of the viable epidermis with the vascularized dermis.

The dermis is a thick layer (thickness 1 to 4 mm) that consists of an extracellular matrix of collagen that acts as a structural support for the epidermis and as a supply of nutrients for the viable epidermis. Its main functions are temperature regulation, hydration and support of the viable epidermis. Besides the vascular bed, several cell types are located in this layer. Dermal collagen is produced by the fibroblasts, a major cell type in the dermis. The tight connection between the dermis and epidermis is created by the basal membrane. In the basal layer of the viable epidermis, the keratinocytes, the most prominent cell type of the epidermis, are generated. The keratinocytes are programmed to travel to the surface of the skin. On their way, which takes approximately 4 weeks, the keratinocytes change in morphology and start to produce different keratins, growth factors and cytokines. They are intimately connected by desmosomes and tight junctions [5]. Finally the keratinocytes become corneocytes, which is the final differentiation product. In this way the epidermis renews itself continuously by a balanced proliferation and differentiation process. A schematic cross-section of the epidermis is provided in Figure 1.

Figure 1.

Schematic view of the epidermal cross section, the lamellar body extrusion process and lipid lamellae formation.

The epidermis is composed of 4 layers. The basal layer is the stratum basale in which keratinocytes divide. On top of the basal layer is stratum spinosum. In the stratum spinosum the keratinocytes start to differentiate and the formation of lamellar bodies, containing the precursors of the barrier lipids, is initiated. In the next viable layer, stratum granulosum, synthesis of these precursor lipids is intensified and the concentration of lamellar bodies increases. When the keratinocytes arrive at the superficial part of the stratum granulosum, the final differentiation product, the SC, is generated. The bounding membranes of the lamellar bodies fuse into the plasma membrane of the uppermost granular cells, and the contents of the lamellar bodies are extruded into the intercellular space between stratum granulosum and SC, and they release their content into the intercellular space. During this extrusion process the phosphoglycerides are converted into fatty acids, while glucosylceramides and sphingomyelin are converted into CERs. Enzymes catalyzing these chemical modifications are also stored in the lamellar bodies. Simultaneously with the changes in lipid composition, the lamellar disks fuse together and form the crystalline lipid lamellae. Most probably the lipids bound to the cornified envelope serve as a template for this fusion process. Adapted from [273].

The epidermis consists of four layers. In the stratum basale the keratinocytes proliferate. When the cells escape from this layer, they move into the stratum spinosum and start to differentiate. The spinous layer has a spinous appearance due to the presence of many desmosomal connections between the cells. During this differentiation process the keratinocytes move gradually from the stratum spinosum to the stratum granulosum. The biosynthesis of precursor barrier lipids starts in the stratum spinosum [6]. The glucosylceramides and sphingomyelin (both precursors of CERs) and phospholipids (precursors of FFAs) are synthesized and then stored in lamellar bodies [7–9]. When the keratinocytes enter the stratum granulosum the lipid synthesis is intensified and the density of lamellar bodies is increased.

When the SC is created at the interface between stratum granulosum and SC, the keratinocytes are transformed into the keratin-containing corneocytes and simultaneously the densely cross-linked cornified envelope, the surface layer of the corneocytes, is formed. After a monolayer of non-polar lipids is esterified to the cornified envelope (referred to as the bound lipids), the intercellular lipid matrix, composed of non-polar lipids, is formed. These non-polar free lipids and the bound lipids are made available by extensive enzymatic processing of precursor barrier lipids that are extruded along with the lipid catabolizing enzymes from the lamellar bodies, see Figure 1. The bound lipid monolayer serves as a template for the free lipids that form lipid lamellae oriented approximately parallel to the corneocytes [10]. The final step in differentiation is desquamation (sheading of superficial cells). This step is necessary for maintaining a constant thickness of SC.

The composition and organization of the extracellular matrix is crucial for the skin barrier function whether the cornified envelope, together with the bound lipid monolayer, does or does not act as a significant obstacle for compounds to enter the corneocytes; in the former situation, permeation across the SC is restricted to the intercellular lipid matrix (as observed in visualization studies [11, 12]). In the latter situation, permeation can also proceed through the corneocytes in series with the intercellular matrix that surrounds the corneocytes (as suggested in NMR and transport modelling studies) [13–16]. For this reason in the next paragraphs an introduction is provided on the history in establishing the lipid composition and lipid organization in normal (healthy) skin. This barrier function is impaired in many skin diseases, in section 2 the changes in lipid composition and organization in inflammatory skin diseases are discussed along with some of the underlaying changes in protein levels or activity of enzymes involved in the lipid synthesis. Three inflammatory skin diseases have been selected, atopic dermatitis (AD), Netherton syndrome (NTS) and psoriasis, which show similar deviations in lipid composition. In order to obtain more details on the role the lipid classes or subclasses play in the lipid organization in the SC and its impaired barrier function, model systems have been developed. These are mostly composed of CERs, FFAs and CHOL. In section 3 the studies with model systems are reviewed, including systems containing isolated CERs as well as synthetic CERs. Both, simple systems composed of only three components and multicomponent systems (maximum is around 12–14 components) are described. In section 4 the findings obtained with intact SC including the clinical studies and lipid model systems will be discussed. Finally, an integration of the results of model systems and of SC will be provided. Based on these findings future directions will be given on skin barrier repair.

1.2. Lipid composition in stratum corneum

The lipids of the SC consist primarily of three main lipid classes, CERs, FFAs and CHOL. CERs are very heterogeneous in nature. Currently there are around 25 CER subclasses identified [17]. Most of these CERs have a sphingoid base and an acyl chain linked by an amide bond. In this paper CER subclasses have been denoted by the 4-letter nomenclature introduced by Motta et al [18]. The first one or two letters denote the type of acyl chain, while the last one or two letters identify the type of sphingoid base. This is referred to as CER XY. The acyl chain (X) is designated as non-hydroxy (N), α-hydroxy (A), ω-hydroxy (O), esterfied- ω-hydroxy (EO) or β-hydroxy acyl (B), see Figure 2. In human SC the sphingoid bases (Y) of the CERs are dihydroxysphingosine (dS), sphingosine (S), phytosphingosine (P), 6-hydroxy-sphingosine (H) and 4,14 sphingadiene (SD). Furthermore, a particular chain length of the CER acyl chain and also the FFA chain is denoted by CXX. Thus, CER NS C24 and FFA C24 denotes sphingosine-based CER with a non-hydroxy fatty acid of 24 carbon atoms and a fatty acid chain length of 24 carbon atoms in length, respectively. Most of the CER subclasses identified in human SC are illustrated in Figure 2. In addition to the variation in subclasses, their total chain length (sphingoid base+acyl chain) varies substantially between around 32 and 72 carbon atoms. The variation in the length of the sphingoid base is much less than the variation in acyl-chain length [17, 19, 20]. The CER EO subclasses are defined as all subclasses having a very long ω-hydroxy acyl chain ester linked to a fatty acid chain with 18 carbon atoms, the majority of which is a linoleate, but oleate and slightly shorter saturated chain length have also been reported [21]. The CER EO group consists of CER EOS, CER EOP, CER EOH and CER EOdS, see Figure 2, and are only observed in the epidermis (In papers these are also referred to as acylCERs). If not further specified, the sphingoid base in the synthetic CERs is 18 carbons in length. The FFAs also have a large variation in chain length, varying between 16 and 36 carbon atoms with 24 and 26 carbon atoms the most abundant ones. Most of the FFAs are saturated, only a small fraction is monounsaturated (muFFAs) [22, 23]. FFAs with an additional OH group are also present, but in minor amounts.

Figure 2.

Stratum corneum ceramide subclasses, nomenclature and structure.

The CERs are composed of a large number of subclasses. All CERs consist of a sphingoid base and an acyl chain. According to the nomenclature introduced by Motta, the acyl chains and the sphingosine chain are indicated by 1 or 2 letters [18]. The acyl chains are either the non-hydroxy (N), α-hydroxy (A), β-hydroxy (B), ω-hydroxy (O) or linoleic acid esterified to an ω-hydroxy acyl chain (EO). The sphingoid base is either dihydrosphingosine (dS), sphingosine (S), phytosphingosine (P), 6-hydroxysphingosine (H) or 4, 14 sphingadiene (SD) and dihydro dihydrospingosine (T). For example, CER NS is a non-hydroxy acyl chain linked to a sphingosine, CER AP is an α-hydroxy acyl chain linked to an phytosphingosine base and CER EOH is an linoleic esterified to an ω-hydroxy acyl chain linked to an 6-hydroxysphingosine. The subclasses with the esterified ω-hydroxy acyl chain (EO), that is CER EOS, CER EOH, CER EOP and CER EOdS, are referred to as CER EO. The subclasses with the ω-hydroxy acyl chain are referred to as CER O. In each CER subclass there is a distribution of acyl chain lengths and sphingoid base chain length. In general, the acyl chain length distribution is much broader than the sphingoid base chain length distribution. All CER subclasses identified in human stratum corneum that consist of a sphingoid base and one acyl chain are depicted in the figure. In addition oleate and linoleate linked to the ω-hydroxy acyl chain are depicted separately.

1.3. A brief history of discovery of free lipids in stratum corneum

CERs:

In this section we focus primarily on papers in which new CER subclasses were reported as the focus is the history of identifying the CER subclasses. The first studies were performed by Gray and Yardley [24, 25]. They performed lipid analysis of different epidermal layers in porcine and human skin. : An accumulation was reported of glucosylceramides in the differentiated viable epidermal layers, while the final differentiated layer (SC) contained a higher percentage of CERs than in the viable layers [24–26]. Gray et al. identified a major linoleate-containing glucosylceramide with an unusually long, but unidentified amide-linked fatty acid [27]. The first study depicting the glucosylceramide with an amide-linked ω-hydroxy acyl chain was published by Wertz and Downing [28]. In 1983 Lampe et al. reported a detailed lipid composition of the epidermal layers and confirmed the CER enrichment in SC when compared to the various layers in the viable epidermis, but the CER subclasses were not further determined [29].

In the same year Wertz and Downing reported a detailed analysis of CER subclasses and chain lengths in porcine SC [30]. The porcine CER composition was determined by thin layer chromatography (TLC) in combination with gas chromatography and occasionally Nuclear Magnetic Resonance (NMR). They reported 5 CER subclasses plus an additional CER with uncertain structure. The 5 CER subclasses differed in their sphingoid bases (S and P) and/or in their acyl chain (N, A and EO), were in sequence of increasing hydrophilicity on the TLC plate: CER EOS, CER NS (most abundant CER subclass), CER NP, CER AS and CER AP, see Table 1. For the first time the extraordinary structure of CER EOS was identified with a high proportion of linoleic acid esterified to the ω-hydroxyl acyl chain. The amide linked acyl chain length of the CERs (not including CER EOS) varied between 14 and 34 carbon atoms with 20, 22, and 26 carbon atoms most prominently present, while the sphingoid base varied between 16 and 24 carbon atoms with the most abundant lengths 18, 20 and 22 carbon atoms. A fraction of the CER AS subclass had a clearly shorter acyl chain length, typically 16 carbon atoms. The ω-hydroxy chain length of CER EOS varied between 20 and 34 carbon atoms with the most prominent length containing 30 carbon atoms. The spots at the TLC plate of the sphingosine-based CER subclasses contained also CERs with an unsaturated base, which could be identified as CERs with a dihydrosphingosine base. These were later reported as CER AdS, CER NdS and CER EOdS (see below).

Table 1.

CER subclass composition either in w/w% or m/m% of porcine, mouse and canine SC

	Porcine w/w [30] Wertz et al.	Porcine w/w [44] Mojumdar et al.	Canine w/w [41] Chermprapai et al.	Canine w/w [43] Reiter et al.	Canine w/w [42] Yoon et al.	Mouse m/m [17] Kawana et al.	Mouse m/m [40] Martins Cardoso et al.
EOSa	7.7	13.8	13.51	13.2	14	16.35	14.26
NdS						9.53	10.6
NS	42.4	44.2b	35.87	30.2	29b	57.87	42.5
NP	10.2	8.8	4.87	4.8	9.6	1.10	3.08 (+AdS)
EOP		0.9	0.18		4	0.04
AdS		16.8				0.13
AS	22.6	5.9	35.64	10.3	24c	0.52	29.6
AP	15.5	9.7	2.55		4
AH			3.97		7.2
EOdS						0.49
EOH			3.40	10.5	8.5
NSD						0.40
ASD						0.01
BSd						4.87
OS						8.09
OdS						0.52
OP						0.04
EOSD						0.03

^a= abbreviations of the CER subclasses in the tables exclude CER, so CER EOS = EOS, etc. The abbreviations are explained in Figure 2 and the abbreviation list at page 2.

^b= includes CER NdS, but a smaller fraction as NS

^c= CER AS+ CER NH

^d= β-hydroxy acyl chain

In a second study Wertz and Downing focused on the CER composition in human SC [21]. In this study, all 5 CER subclasses present in porcine SC were also present in human SC, but the relative CER subclass composition was different. The esterified acyl chain of CER EOS was mainly unsaturated, while the amide linked acyl chains had a very long saturated acyl chain (majority 24 or 26 carbon atoms) and the most prominent ω-hydroxy chain length was 30 carbon atoms. In general, the acyl chain lengths in human SC were slightly longer than in porcine SC. Several years later Downing and colleagues reported two additional CER subclasses, CER OH and CER AH, bringing the number of identified CER subclasses equal to 7 [31]. This composition is provided in Table 2. A further search for CER subclasses resulted in the identification of CER NH [32]. In 2003 Ponec et al. reported an additional CER subclass, CER EOP using HPTLC and NMR, see Table 2, and confirmed that the most abundant chain length of the non-hydroxy and α-hydroxy acyl chains are 24 and 26 carbon atoms [2].

Table 2.

Ceramide subclass composition in human SC in % (w/w or m/m). [245]. For abbreviations CER subclasses, see Figure 2 and abbreviation list at page 2

	leg w/w [31] Robson et al.	breast w/w [2] Ponec et al.	forearm w/w [19] Masukawa	forearm m/m[109] van Smeden et al.	forearm m/m [36] ‘t Kind et al.	breast/abdomen m/m [38] Helder et al.	forearm m/m [17] Kawana et al.	forearm m/m [39] Suzuki et al.
EOSa	8	8.3	4.12	3.76	6.46	4.51	2.11	7.7
NdS			6.9	9.48	9.83	10.97	6.17	11.3
NS	21	20.9	6.35	6.88	7.44	7.29	5.16	3.4
NP	13	18.0	26.45	26.5	22.10	27.48	24.21	29.4
EOP		6.4	1.43	1.43		1.83	1.03	1.1
AdS			0.87	1.09		1.27	0.91	0.9
AS	27	19.5b	2.65	4.57	9.58	3.31	4.29	4.0
AP	4	8.6	13.57	14.79	8.78	11.72	9.16	6.4
NH			21.78	14.01	14.51	16.17	23.74	23.4
AH	22	12.9	12.63	13.07	12.63	10.53	17.96	9.1
EOH	4	5	3.08	4.06	4.26	3.20	3.10	2.0
EOdS				0.37		0.89	0.09
NT					1.73
OS					0.73	0.24	0.56	0.9
OH					0.43	0.41	0.62	0.2
OdS						0.19	0.07
OP						0.42	0.33	0.2
NSD							0.13	0.005
ASD							0.15	0.9
BS							0.17
OSD							0.02	0.002
EOSD							0.02	0.002

^a= abbreviations CER subclasses in the tables exclude CER, so CER EOS = EOS, etc

^b= NH also included. NH was not determined by Robsen et al.

Until 2003 most CER subclasses were analyzed and identified by TLC or HPTLC in combination with NMR and gas chromatography. Chain length distributions were primarily determined by gas chromatography and the most abundant CER subclasses reported in human SC were CER NS, CER NP and CER AS. In the meantime, LC/MS methods were developed to examine the CER subclasses and chain length distribution in just one run. This is extremely important as much more details on the CER subclasses and chain length could be obtained. In principle it is possible to see which acyl chain is linked to the sphingoid base. Farwanah et al. published the detection of the 9 CER subclasses with LC/MS, in which also chain length distributions were provided [33]. However, it was the group of Masukawa that for the first time reported a quantitative LC/MS method to analyze the CER subclasses and the corresponding chain length distributions (including the individual lengths of the sphingoid base and acyl chain) using a large number of CER standards [19, 34]. The group presented 2 CER subclasses, CER NdS and CER AdS, which were already reported by Wertz et al, but using a TLC plate they could not be identified as an isolated fraction, because CER NS and CER NdS, and also CER AS and CER AdS, were located at almost the same spot [21, 30]. As CER NdS is present in quite high quantities in human SC, all previously reported TLC results showed a too high fraction of CER NS subclass.

Some years later van Smeden et al. identified the 12^th CER subclass, CER EOdS, together with a less time-consuming LC/MS method to determine the CER profile in detail [35]. However, also this CER subclass was already reported in the studies of Wertz et al., but again not shown as an isolated fraction on the TLC plate [21, 30]. Simultaneously t’Kindt et al. also provided a comprehensive analysis of human CERs and detected a non-hydroxy CER with two OH groups (T) in the sphingoid base (dihydroxy-dihydrosphingosine) referred to as CER NT [36]. The reported CER composition of t ‘Kindt et al. and van Smeden et al. are also provided in Table 2. Also, at the same time 2 additional CER subclasses were reported by Rabionet et al., the acyl-O-CERs being present at around 5% of the total CER EO fraction [37]. No CER subclass composition was provided. Although the acyl-O-CERs occur in only small amounts, due to their lipophilic character these CERs may have a function in improving the skin barrier. In the meantime it also became clear that low amounts of non-esterfied ω-hydroxy CER subclasses, referred to as CER O, are present in the free lipid mixture: CER OS, CER OP, CER OH and CER OdS [38]. A CER composition of human SC including those 4 CER subclasses is also provided in Table 2. Still 8 other very low level CER subclasses having an β-hydroxy acyl chain (B) or having an 4,14-sphingadiene (SD) as a base were identified in mouse SC. Those that are present in human SC are provided in Table 2 [17]. Overall 21 subclasses together with CER NT and the two acyl-O-CER subclasses have been identified in human SC. This brings the total number identified CER subclasses on 24. Very recently another detailed CER analysis of human SC has been provided, including the individual chain length distributions of the sphingoid base as well as the acyl chain of the CER subclasses [39]. When comparing human CER profiles of the LC/MS methods provided in Table 2, the profiles are quite similar, although some variations are noticed, especially with respect to the CER NH quantities.

For comparison with human SC, besides the porcine SC CER composition, the CER composition of canine SC and mouse SC are also shown in Table 1 [40–43]. Clearly there is a major difference between porcine, canine, mouse on the one hand and human CER composition on the other hand. While in human SC, CER NP and AP are the dominant CER subclasses, in SC of porcine, canine and murine, CER NS (often including CER NdS) is the dominant CER subclass. LC/MS analysis showed that CER NS is dominant to CER NdS in porcine SC [44]. When considering mouse SC, CER NS is even more dominant than in porcine SC. However, as this comprehensive analysis has been performed using 1-day old C57B1/6 mice, this CER composition has been compared with a CER composition of the same mice strain around 8 weeks old, although much less detailed analyzed, see Table 1. This showed a similar level of CER NS, but higher levels of CER NdS and CER AS. Although difference in CER composition between murine, canine and porcine and human SC is substantially, the lamellar organization is remarkably similar, as explained in section 1.6.

FFAs:

The SC FFA composition has been examined less frequently than the CER composition, although this lipid class plays a prominent role in the lipid organization and thus skin barrier function. The first report focusing on the chain length of FFA dates from 1983: Lampe et al. reported a chain length distribution of FFAs, in which FFA C16 was prominently present together with FFA C18, FFA C18:1 and FFA C18:2 [29]. However, in time this appeared not to be correct. Wertz et al. reported that the most abundant chain lengths FFAs in porcine as well as in human SC are C24 and C26, with a chain length distribution similar to the acyl chains of CERs (excluding CER EO), see Table 3 [23]. In addition, the amounts of unsaturated FFA with 16 and 18 carbon atoms were low. Several subsequent studies have confirmed that 24 and 26 carbon atoms are the most abundant FFA chain lengths [2, 22, 45, 46]. Van Smeden et al. reported very small amounts of unsaturated FFA for almost all chain lengths [22].

Table 3.

FFA composition in SC of either human or porcine skin.

FFA carbon atoms Nr		abdomen m/m% [29] Lampe et al.	porcine w/w%[23] Wertz et al.	cysts Human w/w%[23] Wertz et al.	forearm m/m% [45] Norlen et al.	breast w/w% [46] Vicanova et al.	breast w/w% [2]Ponec et al.	ex vivo skina [22] m/m% van Smeden et al.
12.0		0.1
14.0		2.7		1.5		0.9
15:0				1.5
16:0		21.1	5.1	7.4		8.3	10	4
16:1		7.3				0		0.2
17:0				0.8				0.1
17.1								0.1
18:0		24.7	5.4	9.1		4.8	7	5.6
18:1		24.1	4.7	5.7		3.0	10	2.3
18:2		14.7	1.5	1.4		5.1	6
19:0				1.1	1.2			0.1
20:0		0.3	4.3	5.9	5	3.1	5	0.7
20:1								0.1
20.2		0.12
21:0			3.2	1.9	0			0.1
22:0		0.4	31.4	15.3	11	3.1	5	3.8
23:0			7.1	6.2	Is			2.8
24:0		1.6	24.8	26.9	39	34.3	30	33.7
24:1								0.1
25:0			1.6	5.0	10	6.1		8.0
25:1								0.1
26:0			3.3	8.5	23	25.8	25	25.2
26.1								0.1
27:0			0.5		3			2.3
28:0			1.9	2.7	8	5.4	5	7.1
28:1								0.1
29:0					1			0.7
30:0					2			0.1
30.1								0.2

^a=In a previous review, by accident, tape-stripped in vivo skin was mentioned as skin source. [158] This was incorrect.

1.4. The variation in ratio between ceramides:cholesterol:free fatty acids lipid classes

When using HPTLC or TLC, it is possible to determine the weight fractions of CERs, CHOL and FFAs from one HPTLC or TLC plate. However, to obtain a reasonable accuracy, a high number of standards need to be run on the same plate using a variation in the lipid concentrations of all samples. In 2002 Weerheim et al. summarized a large number of weight fractions of CERs, CHOL and FFAs from literature [47]. Due to either the inaccuracy of the method, a lack of proper standards, variation in extraction solvents and methods or variations in skin site, a large variation in the ratio of the fractions is noticed, see supplemental Table S1. To make comparing compositions easier, in this table the relative amount of the total fraction of CERs is assigned to 1, and CHOL and FFA fraction are reported relative to the CER fractions. Some ratios published more recently are included as well. When calculating the mean lipid mass ratios of SC harvested from the forearm, leg or isolated SC (breast or abdomen), these CER:CHOL:FFA ratios are 1:0.5:0.98 (forearm, mostly in vivo extraction), 1:0.59:0.64 (leg, in vivo extraction) and 1:0.71:0.66 (SC, in vitro extraction). Taking into account that the molecular weight of CERs is around twice the molecular weight of CHOL and FFAs, the 1:1:1 CER:CHOL:FFA molar weight ratio often used in lipid model systems is a reasonable approximation, although the observed FFA and CHOL levels seems to be somewhat higher.

Using LC/MS to determine ratios has the problem that absolute data on the CER, CHOL and FFA quantities are required. Absolute data on CERs are not frequently published due to a lack of a large number of CER standards. Furthermore, the quantification of all CERs and all FFAs generally requires two different LC/MS methods. Therefore, no data on CER:CHOL:FFA m/m ratios is available, only CER:FFA m/m ratios were reported, but vary widely. These are CER:FFA 1:0.61 (Forearm, SC extraction), CER:FFA 1:1.7 (SC, abdomen/breast) and 1:1.1 (SC, abdomen/breast) [38, 48, 49].

1.5. Short history of identification of lipid classes bound to the cornified envelope

In the present paper a short description on the discovery of the various bound lipids is provided. The synthetic pathways of the CERs bound to the cornified envelope will not be discussed. Although this is of major interest, this has been reviewed very recently by Wertz [50].

The first studies on bound lipids were initiated when it was discovered that a significant fraction of the unusual linoleate-rich acylglucosylceramides (precursors of CER EO) present in the viable epidermis appeared to be missing from the free lipids in SC. Specifically the relative quantities of CER EO to all CERs in SC (around 10 w/w %) is much less than the relative quantities of acylglucosylceramides of the total glucosylceramide fraction in the viable epidermis (around 50 w/w%). Intrigued by this loss, Wertz and Downing performed a mild saponification of the porcine SC remaining after its lipid extraction, followed by another lipid extraction. In the resulting extract they identified CER OS (that is CER EOS without the linoleate-rich esterified acyl chain) [51]. As this fraction could only be detected after saponification, it was suggested that CER OS was ester bound to the SC. Almost simultaneously the group reported an electron microscopic study showing that CER OS was attached to the cornified envelope, most probably acting as a template to orient the lipid matrix lamellae along the corneocyte surfaces [52]. In the next study focusing on human SC not only CER OS but also CER OH was for the first time isolated by Wertz et al. and later on also identified by Robson et al. [31, 53]. This is not surprising as in the related free CER subclass composition, CER EOS is predominantly present in the CER EO fraction of porcine SC, but is less dominant in the CER EO fraction of human SC. Besides these bound CER subclasses, two bound fatty acids were also detected: non-hydroxy FFAs and ω-hydroxy FFAs. The acyl chains of these esterified CERs and the fatty acids were very long up to 34 carbon atoms.

In 2007 another study reported the CER profile of the bound lipids in human SC [54]. Two CER subclasses were identified being CER OS and CER OH, in agreement with the early studies. The chain length distributions of the sphingoid bases (12 to 22 carbon atoms) as well as the acyl chains were determined (28 to 36 carbon atoms). Hill et al. was able to identify a 3^rd bound CER subclass, namely CER OP, which is present in human SC in smaller quantities than CER OH and CER OS [55]. In the meantime, CER subclasses with a dihydroxy sphingoid base were discovered and of course it is no surprise that very low levels of CER OdS are also bound to the cornified envelope in human SC [56]. Furthermore, in the same paper bound and unbound CER subclasses were described; it was reported that unsaturated CERs and CERs with a relatively shorter total chain length are selectively bound to the cornified envelope. It is striking that the level of bound CER OS is around 72 m/m% in the total bound CERs, while the fraction of CER EOS is around 50 m/m% of al CER EO classes. This indicates that there is a slight preference for CER EOS to bind to the cornified envelope compared to the other CER EO subclasses. The same observation is reported by Suzuki et al. in mouse SC: while the fraction of CER EOS in the total CER EO is around 71 m/m %, in the bound CERs, CER OS is around 82 m/m % [39]. Seasonal and age influence on the amount of bound CERs was also studied and it was shown that in the winter season the amount of bound CERs is around 50% of that during the summer season [57]. Furthermore, in individuals at 50 years of age or older, there is also a drop in the amount of bound CERs. These variations in bound lipid quantities raise the question on whether the bound lipid composition affects the barrier function. Several studies report a correlation between an increased trans epidermal water loss (TEWL) and a reduction in the amount of bound CERs or with a change in the composition of bound CERs, such as the level of unsaturated bound CERs [58–60]. This demonstrates that CERs bound to the cornified envelope may contribute to the skin barrier. However, it cannot be excluded that changes in bound CERs occur coincidentally with changes in free lipid composition, which also contributes to the reduced TEWL.

1.6. Lipid organization in the stratum corneum

The lipid matrix in human SC is organized in crystalline lipid lamellae as represented schematically in Figure 3. Below a brief history is provided about the elucidation of the lipid organization in this extraordinary lipid matrix.

Figure 3.

A schematic illustration of the lamellar and lateral organization in stratum corneum.

The stratum corneum (SC) consists of corneocytes with a lipid matrix in the intercellular space (A). The lipids in the intercellular space are arranged in two lamellar phases: the short periodicity phase (SPP) with a repeat distance of around 6 nm (B) and the long periodicity phase (LPP) with a repeat distance of around 13 nm (C). A lamellar phase consists of a series of repeating units referred to as the unit cell with a length equal to the repeat distance. A schematic presentation of the unit cell is provided in figure D and E for the SPP and LPP, respectively. F. Within the plane perpendicular to the unit cell direction, the lipids can form either a very dense orthorhombic packing, a less dense hexagonal packing or a disordered liquid packing. The orthorhombic packing has two distances between the lattice planes (a lattice plane is a plane with a high electron density) of approximately 0.42 and 0.37 nm. In this dense packing the lipids are not able to rotate along their longest axis. In the hexagonal packing, the distance between the lattice planes is around 0.42 nm and the lipids are able to rotate along their longest axis. Finally, in the fluid phase the distance between the lattice planes vary to some extent, but is approximately 0.46 nm.

The first studies using X-ray diffraction to elucidate the lipid organization in isolated human SC were reported around 60 years ago by Schwanbeck and Wilkes [61–63]. These studies using isolated SC were performed with a conventional small angle X-ray diffraction camera and revealed a similar pattern as published more than 20 years later by White et al. [64]. The interpretations of the X-ray diffraction curves were extremely difficult as no information was available on a structural level. A short explanation of the X-ray diffraction technique is provided in supplemental Figure S1. The X-ray patterns were interpreted as lipids that form tubes surrounding keratin filaments. These interpretations were based on calculated X-ray diffraction patterns (by Fourier transforming the experimental data using guessed diffraction phase information) derived from models describing lipids in hollow tubes. It took at least 10 years before additional information became available using freeze fracture electron microscopy [65–67]. The SC was visualized, and for the first time the existence of lipid lamellae was reported. This provided important new insights into the lipid organization in SC. Although freeze fracture electron microscopy is an excellent technique, it provides a shadow pattern of the real structure, which makes it difficult to interpret the freeze fracture images in details: after cryofixation the tissue is fractured. Subsequently Pt vapor is evaporated from a fixed angle on the freshly obtained surface (at cryo conditions). This results in a shadowing pattern.

Additional information was obtained from images of thin sections treated with fixation and staining material, which is either viable epidermis, dermatomed skin or full thickness skin. A problem with conventional electron microscopic studies using osmium tetroxide as oxidizing reagent in early studies was the absence of lipid lamellae staining due to its weak reactivity with saturated SC lipids.

By changing the reagent to ruthenium tetroxide which reacts strongly with the saturated SC lipids, Madison et al. reported the visualization of lipid lamellae, which were characterized by dark electron-dense bands separated by a repeating pattern of broad-narrow-broad electron lucent bands [10, 68]. The distance of this broad-narrow-broad repeating pattern (referred to as the repeat distance of the unit cell in X-ray and neutron diffraction studies, see Figure 3) was around 12–13 nm [10]. The first models of the lipid organization were built on this pattern [69]. Importantly the lamellae were oriented primarily parallel to the corneocyte surfaces and thus parallel to the SC surface [10]. Disadvantages of this technique are that the entire SC cannot be visualized due a poor penetration of the ruthenium tetroxide reagent and the pattern is a derivative of the molecular organization, as staining patterns are visualized and not the lipids themselves. Nevertheless, electron microscopic visualization with ruthenium tetroxide stained SC lamellae was a very big step forward in elucidating the lipid organization. These micrographs showed clearly that lipid lamellar phases exist in SC oriented parallel to the corneocyte cell surfaces. The existence of lipid lamellar phases with a distinctive repeating pattern suggested a unique molecular organization.

Other details became available using X-ray diffraction and Fourier transformed infrared spectroscopy (FTIR). While electron microscopy obtains local information, with a field of view in the nm-range, X-ray diffraction and FTIR are bulk techniques obtaining “average” structural information.

After the initial studies of Swanbeck and Wilkes [61–63], the next studies focusing on the lamellar organization were performed by White et al. using small angle X-ray diffraction of hairless mouse SC [64]. They obtained a series of sharp equidistant peaks that could be attributed to a single lamellar phase with a repeat distance of 13.1 nm (see supplemental Figure S1 for explanation of technique and interpretation). Later on this phase was referred to as the long periodicity phase (LPP). They also performed wide-angle X-ray diffraction and observed an orthorhombic packing of the lipids, which was in agreement with an earlier X-ray study of Elias et al. using neonatal mouse skin [64, 70]. As one of the reflections attributed to the orthorhombic packing overlaps with the reflection attributed to a hexagonal packing, the hexagonally packed lipids could also be present. In addition to the orthorhombic packing the reflection of liquid phase was also observed. The source of this liquid phase might also be sebum, which is present as a thick layer at the SC surface of hairless mouse skin [71].

In other studies using mouse SC besides the LPP (indicated by very sharp peaks in the diffraction pattern), reflections attributed to a second lamellar phase were also detected with a repeat distance of around 6 nm, referred to as the short periodicity phase (SPP) [72, 73]. Both, the LPP and SPP are formed by the intercellular extractable lipids; after lipid extraction of SC the reflections of the LPP and SPP disappeared [64, 72]. Bouwstra et al. showed that the 13.4 nm lamellar phase (LPP) already started to disappear at round 45 °C together with the orthorhombic packing: only a hexagonal packing was observed at 45 °C [72]. The peaks attributed to the SPP remained, showing that indeed the two lamellar phases coexist in the SC, and that these phases behave differently as a function of temperature. In another study ruthenium tetroxide staining and X-ray diffraction were combined. Both methods resulted in the same repeat distance of around 13 nm (for explanation repeat distance, see Figure 3) [68]. In 1991 the lamellar phases in human SC were reported using small X-ray diffraction at synchrotron radiation facilities [74]. Due to the strong intensity of the beam, the diffraction pattern of structures that weakly scatter can be measured. Broad peaks that were partly overlapping were detected. To interpret these peaks required additional information, which was obtained by heating the SC to 120 °C and recrystallizing the lipids. This resulted in X-ray patterns with sharp peaks at equidistant positions, similarly as observed in the diffraction pattern of mouse SC. The peak positions indicated a lamellar phase with a repeat distance of around 13 nm. The characteristic intensity distribution of the peaks attributed to this phase was also observed: the 2^nd order diffraction peak had a higher intensity than the 1^st and 3^rd order peaks. When comparing the original X-ray diffraction pattern at room temperature with the pattern obtained after recrystallization, it was concluded that a 2^nd lamellar phase is present in human SC with a repeat distance of around 6 nm, being the SPP that was also observed in mouse skin [73].

Also in 1991, Garson et al. reported an X-ray pattern of human SC together with its orthorhombic lateral packing showing a preferred orientation of the lamellae parallel to the SC surface [75]. They reported two main reflections attributed to the lamellar phases, located at spacings of 6.5 nm and 4.5 nm, similar to the studies by Bouwstra et al. [74]. However, due to the X-ray diffraction settings, it was not possible for them to detect the 1^st order reflection of the LPP being close to the primary beam. This is why they could not attribute the 6.5 and 4.5 nm reflections (2^nd and 3^rd order reflection) to the LPP with a 1^st order reflection corresponding to a spacing at around 13 nm. The LPP and SPP were also detected in porcine SC with repeat distances of 13.1 and 6 nm [76]. When heating human SC, the lipid lamellae were still present at 60 °C and were therefore thermally more stable than in mouse SC. Hydration did not affect the peak positions in the diffraction patterns of porcine, mouse and human SC, indicating an absence of swelling of the lamellae [68, 72, 74, 76]. However, Ohta et al. showed that in mouse skin the SPP repeat distance increased at elevated water levels, while the repeat distance of the LPP did not change [77]. It is not clear why these differences in swelling behavior were observed. However, in lipid model systems, the SPP as well as the LPP in a fully hydrated state showed only 1 or 2 water molecules per lipid molecule; thus, no swelling has been observed in these model systems [78–80]. Essential to mention is that the absence of swelling is important for a proper barrier when SC is at a higher hydration level.

Besides the lamellar phases, the lateral packing in human and porcine SC was examined using wide angle X-ray diffraction. In human SC an orthorhombic packing was observed, while in porcine SC a hexagonal packing was dominantly present [75, 76, 81]. Again, in these studies relying on X-ray diffraction, it was not possible to determine whether beside an orthorhombic phase in human SC, a part of the lipids was forming a hexagonal lateral packing. FTIR is a technique to provide additional information. In the first study using FTIR, Gay et al. investigated the lateral packing of human SC as function hydration using the scissoring frequencies [82]. The conformational ordering was examined by the asymmetric and symmetric stretching frequencies, see supplemental Figure S2 for a brief explanation. In the case of an orthorhombic packing the hydrocarbon tails are so close that a splitting of energies occurs in the scissoring vibrations resulting in two peaks: an orthorhombic packing is characterized by peak positions at frequencies 1463 cm⁻¹ and 1473 cm⁻¹. When the lipids are in a hexagonal packing (also referred to as a gel-phase), the hydrocarbon tails do not have a short range interaction and a single peak appears at a frequency of around 1468 cm⁻¹. When measuring the infrared spectra as a function of temperature, the orthorhombic to hexagonal packing transition can be monitored. Gay et al. observed this transition at around 65 °C in dry isolated SC; a weak transition to a more disordered phase was also observed at around 35 °C (monitored by an increase in the wavenumber of the stretching frequencies). They associated this shift with an increase of lipids in a fluid phase. For the first time they also observed that an increase in hydration lowered the orthorhombic to hexagonal transition temperature regions in the lipid matrix. Why the reported orthorhombic transition temperatures are quite high in this study compared to others is not clear. Although they used cadaver leg skin, it is unlikely that the origin of the skin caused these high transition temperatures. In a more recent study Caussin et al. reported a transition temperature region between 40 and 50 °C in dry SC, which is in line with the disappearance of the orthorhombic reflections in the wide angle X-ray diffraction studies. In addition the increase in stretching frequencies was observed in the same temperature region [83]. They also observed that an increase in hydration level reduced these transition temperatures confirming the studies by Gay et al. However, FTIR did not yet provide information about the fraction of lipids forming either an orthorhombic or a hexagonal lateral packing, as the extracted SC also resulted in a peak close to 1468 cm⁻¹.

In 2010 Damien and Boncheva used the width of the scissoring vibrations measured by ATR-FTIR (ATR: attenuated total reflectance) to determine the fraction of lipids in an orthorhombic packing [84]. This method was used in a clinical study in which the increased fraction of lipids forming an orthorhombic packing correlated with a reduced TEWL [85]. Very recently Ohnari et al. used a novel FTIR technique: focal plane array (FPA)-based Fourier transform (FT) IR imaging [20]. This method enabled them to measure, based on stretching vibrations and high resolution imaging, the distribution of lipids forming a hexagonal, liquid or orthorhombic packing in SC samples collected on glued grids (These are grids made sticky. When removing the grids from the skin a small thin sheet of SC is also removed) in a clinical study. A higher fraction of liquid packing with a lower fraction of orthorhombic packing could be correlated with higher TEWL values in sampled volunteers, while a lower TEWL correlated with a higher lipid fraction forming an orthorhombic packing. However, as the lateral packing was obtained from stretching frequencies instead of scissoring frequencies with a pure substance as a reference for the phase changes, the absolute values of the obtained lipid fractions forming a particularly lateral packing are uncertain.

Further details on the lipid packing were obtained using electron diffraction. The essential difference with X-ray diffraction is that a very small sample area is exposed to the electron beam (1 μm²) compared to the area exposed in a X-ray diffraction experiment (approximately 1–4 mm²). For this reason in X-ray studies the number of small single crystals exposed to the X-ray beam is large. Because these crystals have different orientations, the X-ray diffraction pattern shows rings in the diffraction plane and no spots. The existence of these rings is why it is difficult to detect a hexagonal packing in the presence of an orthorhombic packing (overlapping 0.42 nm reflections). The situation is different in electron diffraction. Because only one or a limited number of crystals is exposed to the electron beam, the diffraction pattern shows spots instead of rings and therefore contains more detailed information. In the case of hexagonal packing, the angle between the spots perpendicular to the electron beam are 60 °. Furthermore all 6 spots are at an equal distance to the location of the primary beam. In the case of orthorhombic packing, these angles are not equal to 60 ° and two out of 6 spots are closer to the primary beam, see Figure 4. This makes it possible to distinguish more easily between the hexagonal and orthorhombic packing.

Figure 4.

Electron diffraction patterns obtained from stratum corneum.

Electron diffraction patterns can be used to distinguish between an orthorhombic and hexagonal packing when only a single crystal or a few crystals are exposed to the electron beam. (A) When the incident beam is perpendicular to the paper plane, the hexagonal packing results in a hexagonal pattern with all angles between two adjacent spots and the primary beam equal to 60 °. In case of an orthorhombic lattice the orthorhombic packing is also characterized by 6 spots, but the angle between two adjacent spots is not equal to 60 °. Furthermore 2 out of 6 spots have a longer distance (yellow spots) to the primary beam. (B) Frequently an orthorhombic packing was observed based on three crystals, oriented such that the spots were rotated over a fixed angle. When the incident beam is perpendicular to the plane of the paper, this give rise to the formation of 6 doublet spots as shown in the figure. (C) An example of diffraction pattern from a hexagonal packing. (D) An example of an orthorhombic diffraction pattern that can be explained by the three orientations. Adapted and modified from [86, 128].

Electron diffraction studies used glued grids for SC sampling. SC collected from ex vivo human skin measured at 25 °C revealed that only an orthorhombic packing is present in 38% of the patterns, while an orthorhombic packing is present and a hexagonal packing cannot be excluded in 54% of the analyzed diffraction patterns [86]. In the remaining 8% of the collected patterns, a hexagonal packing is certainly present. Increasing the temperature to 32 °C showed no large shift in the frequency of detecting the various patterns. SC samples harvested in vivo by glued grids showed a similar ratio at 25 °C. However, at 32 °C, a hexagonal packing was present in around 50% of the patterns. The reasons for these differences in the ex vivo and in vivo samples are not clear. However, in the same study it was shown that the hexagonal patterns were mainly observed in samples collected superficially (3 strips) while in deeper layers of the SC (11–17 strips), only diffraction patterns in which an orthorhombic packing is present were observed. Differences in the ex vivo and in vivo stripping depth may partly account for these differences. The superficial presence of a hexagonal packing is in agreement with observations by Bommannan et al., in which the lipids close to the surface of the skin showed a higher conformational disorder which might be due to the sebum mixing with the barrier lipids [87]. Interestingly, studies of the electron diffraction patterns revealing an orthorhombic packing often showed reflections attributed to 3 orthorhombic crystals with a fixed angle over which these crystals were rotated. This led to the conclusion that the 3 orientations of the crystals are related to each other. Possibly, these differently oriented crystals may be related to the three lipid layers in the unit cell of the LPP (see Figure 3, Figure 4B and Figure 9A–D): the crystal belonging to one orientation should be located in the first lipid layer, the 2^nd orientation of crystals to the central lipid layer, while the 3^rd orientation is attributed to the crystalline structure in the 3^rd lipid layer. Several papers report that a large fraction of CERs are in linear conformation [69, 88–90]. Consequently, the orientation of the lateral packing in each layer of the LPP may not be independent, but dictated by the CERs in linear conformation.

Figure 9.

The lipid arrangement in the repeating unit (unit cell) of the LPP as proposed by various molecular models.

(A) The molecular model based on the broad-narrow-broad lucent ruthenium tetroxide pattern [69]. (B). A revised model also based on the broad-narrow-broad lucent ruthenium tetroxide pattern obtained with lipid mixtures [228]. (C) A molecular model based on the electron density profile obtained from X-ray diffraction [186]. (D) A molecular model based on the location of the lipids in the unit cell obtained by neutron diffraction [97]. (E) A molecular model based on the pattern obtained by cryo-electron microscopy combined with simulations [99].

With respect to wide angle X-ray diffraction, Doucet et al. performed additional studies using a microfocus beam with a small cross section of 1–2 μm² as a function of depth in isolated SC sheets [91]. Due to the small beam size, it was possible to obtain information as a function of depth with the beam oriented approximately parallel to the SC surface. Although the beam cross section is small, many crystals are exposed due to the sample thickness. The authors report that the orthorhombic packing is more prevalently present in the central part of human SC compared to the superficial and deep regions in SC confirming earlier studies, such as those of Gay et al. using FTIR and Pilgram et al. using electron diffraction [82, 86]. With respect to the lamellar phases no changes were observed as a function of depth. They also observed a preferred orientation of the lipid lamellae and lateral packing as reported in earlier papers indicating that lamellae were parallel to the SC surface [8, 10, 75, 81].

A related question is whether the hexagonal packing or orthorhombic packing are present in both the LPP and SPP, or that one of these lamellar phases is linked with only an orthorhombic or only a hexagonal packing. The group of Hatta studied in detail the lamellar phase changes and lateral packing as function of temperature. Based on detailed temperature dependent X-ray studies they concluded that most likely the LPP is linked to the hexagonal packing [73]. However, this conclusion differs from several observations using isolated SC and also lipid model systems. The observations of Doucet et al. that lamellar phases do not change in depth in SC, while the orthorhombic lateral packing is most prevalent in the central part of the SC does no coincide with an LPP being only related to a hexagonal packing [91]. Furthermore, when focusing on lipid model systems forming only the LPP using CER mixtures mimicking closely the CER composition of intact SC, it was reported that a hexagonal, orthorhombic and even a liquid lateral packing coexists in this phase [44, 92, 93].

In 2005 the first paper appeared in which cryo-electron microscopy of vitreous sections (i.e. water in the sample is in non-crystalline amorphous solid) of fresh fully hydrated, non-cryo-protected human skin was used to visualize the SC lipid matrix. In this technique, electron microscopy of rapidly frozen samples is performed at temperatures at which water vapor pressures are extremely low, thereby eliminating the need to dehydrate the sample [94]. Very detailed images were obtained. When focusing on the intercellular lipid matrix, lipid lamellae were visualized oriented primary parallel to the corneocytes. An advantage of the cryo-fixed tissue is that the visualization of the lipid lamellae throughout the whole SC is based on the electron density variations in the structure in samples of normal hydration rather than the electron density of staining agents, such as ruthenium tetroxide in dehydrated samples. Stacks of lipid lamellae were visualized as electron dense lines oriented primarily parallel to the corneocytes. The observed distances between these lines were frequently 4.1, 4.2 and less frequently 6 nm in one stack. This suggested an irregular pattern of subsequent electron dense and electron lucent layers. These observations did not match with the X-ray diffraction patterns obtained from SC, as sharp diffraction peaks can only be obtained when a repeating pattern of the same structure (referred to as unit cell) exists in the tissue. Whether this difference in results was due to a different handling (isolated SC versus cryofixed skin) is not known. However X-ray patterns obtained from fresh epidermis were similar to the X-ray diffraction pattern of isolated SC [95].

In a follow-up study, the presented cryo-pattern, derived using a quantitative pixel intensity analysis of very high magnification micrographs, was slightly different [88]. A regular pattern of electron lucent and electron dense bands with an alternating distance of 6.5 and 4.5 nm was reported. This was observed in all images of the SC and corresponds to a regular repeating structure that is in agreement with an X-ray diffraction pattern consisting of equidistant peaks. The images would result in a repeating pattern of around 11 nm, which is short compared to the 12–13 nm repeat distance by X-ray diffraction. However, a more striking difference in result is a two layer asymmetric pattern, which is not in agreement with the trilayer symmetric unit cell obtained using X-ray diffraction and neutron diffraction of isolated SC and lipid model systems [96, 97].

In a third study using cryo-electron microscopy of vitreous sections by the same group, two lamellar repeats were reported, one with a repeat distance between 11–12 nm and another with repeat distance around 6 nm [98]. This might correspond to the LPP and SPP, respectively. However, the difference in unit cell (asymmetric two layer unit cell versus a symmetric trilayer unit cell) remained. Furthermore, while the cryo-electron microscopy study reports that the 6 nm repeat is mainly in the lower layers of the SC, the X-ray diffraction studies do not show a change in lamellar organization as function of depth, this means that no change in small angle X-ray diffraction pattern was observed as function of depth [91]. Based on the cryo-electron microscopic images a molecular model has been built and further refined using simulations [99]. This model will be described in section 3.2.7.

2. IMPACT OF INFLAMMATORY SKIN DISEASES ON STRATUM CORNEUM LIPID PROPERTIES

2.1. Atopic dermatitis

2.1.1. Lipid composition and organization

The lipid composition of all early studies was analyzed with HPTLC or convention TLC, sometimes in combination with gas chromatography and/or NMR. The first publication that showed a reduction in the amount of CERs in AD was from the group of Plewig in 1990, which described lipids extracted from in vivo skin [100]. Only one year later Imokawa et al. reported more details on changes in the SC lipid profiles of patients with AD. At that time 6 CER fractions could be separated. It was reported that the amount of all CER fractions reduced in AD patients compared to controls. The most drastic reduction was observed for CER EOS and more dramatically in lesional skin (inflamed red skin) than in non-lesional skin [101]. In the same year Yamamoto reported relative amounts of the various CER fractions: the level of fraction 1 (CER EOS) and fraction 2 of all CER fractions were reduced in non-lesional skin by at least 50% compared to the controls, but fraction 2 was not reduced significantly. The relative distribution among the other CER fractions remained similar [102]. The fatty chain esterified to the ω-hydroxy acyl chain of CER EOS was primarily 18 carbons. However, surprisingly a large fraction of these fatty acid chains was saturated in both the control and AD skin, which is not in agreement with other studies [2, 21]. Furthermore, the fraction of esterified oleic acid was increased in AD patients compared to control. In 1998 Di Nardo et al. reported reductions in CER EOS and CER NP in both, non-lesional and lesional skin [103]. A decrease in cholesterol sulphate was also noticed, while CHOL levels were increased. They also reported a reduced skin barrier: as assessed by TEWL, which was correlated with a reduced amount of CER NP [103].

A few years later for the first time a change in CER chain length was reported: in the HPTLC plate a separate fraction contained shorter CER AS (C16/C18 acyl chain) in non-lesion AD skin, which was not observed in control skin. [104] Subsequently using LC/MS that can provide information on the chain length distribution and CER profile, the group of Neubert and Wohlrab did not notice a change in CER subclass or chain length profiles maybe due to the small patient group and the fact that only non-lesional skin was investigated [105].

In 2002 the first analysis of FFAs in AD lesional skin showed a strong reduction in the FFA fraction with a chain length longer than 24 carbon atoms was reported [106]. Interestingly, labelling studies performed to examine the rate of synthesis of CER subclasses showed that CER NP and CER EOH were synthetized much slower in AD skin than in controls, while CER NS synthesis was slowed marginally. These observations relate very well with quantitative changes in the 11 CER subclasses reported by Ishikawa et al. using LC/MS [107]. CER AS and CER NS (not significantly) were increased, while CER NP, CER NH and the CER EO subclasses were reduced. The chain length distribution of CER NS was also reported: the short-chain fraction of CER NS (total chain length < 40 carbon atoms) was increased, while CER NS with longer chains (total chain length > 50 carbon atoms) were reduced. The CER NS fraction with a total chain length of 34 carbon atoms was exceptionally high (referred to as CER NS-C34). Correlations were observed between the disturbed barrier recorded by TEWL and changes in both the levels of the individual CER subclasses and the levels of chain length of CER NS. These results demonstrated that the changes in CER composition do play a role in the reduced skin barrier in AD with more pronounced changes in lesional skin than in non-lesional skin.

A few years later Janssens et al. and van Smeden et al. published results from an extensive series of studies that examined CER chain length distribution, relative amounts of CER subclasses, FFA chain lengths and lipid organization in AD non-lesional and lesional skin compared to controls [108–110]. Observed changes in CER subclass composition were similar to those reported by Ishikawa, but in this study CER AS and CER NS were significantly increased [107]. The chain length distribution of 12 CER subclasses were also reported. In AD skin the level of long chain CERs was decreased and the level of short chain CERs, especially those with a total chain length of 34 carbon atoms, which were most abundantly in CER NS and CER AS. All these changes were seen again most drastically in lesional skin.

Concerning the lipid organization, the lamellar phases were affected in SC of non-lesional AD compared to controls: the peak position in the diffraction pattern shifted (see Figure 5), suggesting a shorter repeat distance. Simultaneously the shoulder attributed to the 3^rd order diffraction peak of the LPP was in a subpopulation of patients not present. A correlation was observed between the presence of this shoulder and the presence of CER EO: when this shoulder was not present, a lower CER EO fraction was noticed (8.8±2.2 m/m% versus 5.4±1.8 m/m% ) indicating that the presence of the LPP is related to the fraction of CER EO (see Table 4). Such a relationship was not observed with any of the other CER subclasses. This relationship was also noticed in a cohort that included samples of dry and control skin [95]. Furthermore a lower fraction of lipids adopted the very dense orthorhombic packing in AD skin compared to controls. Concerning the chain length distribution: the short chain FFAs were more abundant and the long chain FFAs were reduced in quantity in lesional and non-lesional skin. Finally strong correlations with barrier function (TEWL) were observed for chain packing density (caused by a lower fraction of lipids in orthorhombic packing) and also chain length of the FFAs + CERs combined. However, whether these lipid properties are also an important underlying factor cannot be judged from these studies, but have been reported using lipid model systems, see section 3.2.4 [111, 112].

Figure 5.

Diffraction patterns of controls and atopic dermatitis patients.

Diffraction patterns of (A) control subjects, and (B) atopic dermatitis patients non-lesional skin and (C) boxplot showing the variance of the main peak position in the diffraction pattern. 1^st and 3^rd indicate the 1^st and 3^rd order diffraction peaks of the long periodicity phase (LPP). ① and ② indicate patients with an altered diffraction profile. In both diffraction patterns the main peak position shifted to higher q-value. (* and **) and in ② the shoulder attributed to 3^rd order diffraction peak of the LPP is not present. # indicates a diffraction peak attributed to crystalline CHOL. The main peak is caused by the 1^st order diffraction peak of the short periodicity phase and the 2^nd order of the LPP, see supplement Figure S1 for explanations. Reproduced from Janssens et al. [108].

Table 4.

Percentage (m/m) CER EO in the presence and absence of the shoulder n the X-ray pattern in atopic dermatitis (AD) patients non-lesional skin from AD patients and control skin, see Figure 5. The shoulder is attributed solely to the 3^rd order diffraction peak of the LPP. For abbreviation CER subclasses, see Figure 2 and abbreviation list at page 2.

	AD patients, no 3rd order peak present	AD patients, 3rd order peak present	Control skin, 3rd order peak present
w/w % CER EO	5.4 ± 1.8%	8.8 ± 2.2 %	9.6 ± 2.6%

Another study reported that certain ratios of CER subclasses, CER NP/CER NS, CER NP/CER AS, CER NH/CER NS and CER NH/CER AS were reduced in non-lesional and lesional AD skin [113]. In particular CER NP/CER NS ratio was shown to be strongly correlated with the TEWL values. This is in agreement with the studies of Ishikawa et al., Janssens et al. and van Smeden et al. [107–109]. With respect to absolute values of CERs and FFA, a recent paper reported the absolute total amount of CERs and of FFAs in the same cohort [49]. The CERs as well as the FFAs were reduced in quantity in non-lesional and lesional AD skin. Moreover, in lesional skin, the FFA reduction was stronger than CER reduction. This strong reduction in FFAs may contribute to the reduced presence of the orthorhombic phase in these samples [109].

Several additional studies have focused on other aspects of AD.

1. The group of Imokawa studied whether there is a difference in response to barrier disruption induced by tape-stripping, and showed that in control skin the CER production is increased after tape-stripping, but not in non-lesional AD skin [114].

2. A few studies have examined the effect of Staphylococcus aureus (S. Aureus) on lipid synthesis. Different levels of some CER subclasses (including members of the CER EO) with a specific chain length were observed in S. aureus-rich AD patients compared to S. aureus-poor AD patients, suggesting that S. aureus affects CER synthesis [115]. In the same publication it was reported that CER AH with 34 carbon atoms and CER AP with 34 carbon atoms were significantly elevated in AD skin compared to control. Surprisingly, no information was provided about CER NS and CER AS, which other studies have reported as being the CERs with the highest abundance of 34 carbon atoms chain [107–109]. In a recent study of CER composition Emmert et al. reported positive correlation between the presence of S. aureus in AD skin and the amounts of sphingosine subclasses CER AS, CER AdS, CER NS and CER NdS [116]. Concerning the body site they observed substantial differences in CER subclass composition especially between the forehead and lower forearm with an increase in CER NS and a decrease in CER NH in forehead SC. The short (C16 and C18) chain fatty acids showed a negative correlation with S. aureus, but these results should be taken with caution as volunteers were asked to stop using topical treatments only 24 h before samples were harvested.

3. Very recently the effect of Mallasezia Fulfur ( M. Fulfur) colonization on Head and Neck Dermatitis was reported. CER profiling showed a substantial decrease in the level of most CER subclasses. The levels of CER EOS and CER OS were most dramatically decreased. [117] The CER levels were much lower than in patients with only AD. The outcome of the study suggests that the reduced CER level might proliferate M. Fulfur level in the skin, thereby enhancing the pro-inflammatory cytokine levels.

4. CER profiles in AD children have been reported as well [118]. However, the reported CER profiles are extraordinary as the levels of CER NH and CER NdS are exceptionally high in both control and AD skin: more than 95% of all CER subclasses and 10 to 100 times higher than CER NP, which in most studies is the most abundant subclass in healthy human skin. Although mass spectrometry was used, there was no information on chain length distribution. In agreement with most other studies, CER NS and CER AS were reported to be more abundant in lesional compared to control skin [118].

In summary, most studies have shown clearly different CER profiles in AD skin compared to the control skin. In AD skin, the FFA and CER chain length is reduced with an increase in CERs with a total chain length of 34 carbon atoms. Often these changes in lipid composition correlated with a change in TEWL. When relating CER composition to enzyme activity, it is important to use absolute CER amounts, but these absolute amounts are reported infrequently. Of the studies that used LC/MS, absolute amounts are reported by Ishikawa et al., Boer et al., Berdyshev et al. and Ito et al. [49, 107, 119, 120]. These papers all show a strong reduction in CER NP, CER NH and CER EO in AD (lesional) skin and a less drastic but significant increase in CER AS. Boer et al. and Berdyshev et al. also showed an increase in CER NS, while Ishikawa et al. and Ito et al. reported no difference in CER NS [49, 107, 119, 120].

2.1.2. Enzymes playing a role in ceramide and free fatty acid compositional changes in atopic dermatitis skin

In the nineties several research groups initiated studies to unravel the underlying factors for a reduced CER level. In 1999 the group of Ohnishi observed that ceramidase is secreted by colonized bacteria resulting in a higher activity of this enzyme when harvested from AD skin surface (lesional and non-lesional SC) compared to SC of controls [121]. As this enzyme cleaves the fatty moiety from the sphingoid base, a higher level and activity may indeed reduce the CER amount in the SC, but should result in a higher level of sphingoid base. However, Arikawa et al. showed a reduction in sphingosine base in SC of AD patients and showed a correlation between a reduced activity of acid ceramidase and the level of sphingosine, suggesting that ceramidase might play a less prominent role in the reduced level of ceramides in AD[6, 122]. Almost simultaneously a high activity of two novel enzymes in AD in non-lesional and lesional skin were reported, which appeared to be the same enzyme: sphingomyelin glucosylceramide deacylase. This enzyme, cleaves fatty acid from sphingomyelin and glucosylceramide, respectively, yielding sphingosine phosphocholine or glucosylsphingosine. [123–126] Sphingomyelin serves as a precursor of CER AS and CER NS, while glucosylceramide is the precursor of all CER subclasses, including CER EO [127]. A higher activity on sphingomyelin would reduce primarily the absolute levels of CER AS and CER NS. However, in general this has not been reported in AD skin [49, 107, 128]. In addition, the group of Imokawa showed an increased level of glucosyl-sphingosine indicating that sphingomyelin glucosylceramide deacylase might be responsible for the degradation of CER EOS in AD skin [125]. However, when the amount of the product glucosylsphingosine (around 10 ng/mg SC) is considered compared to the amount of CER EOS (between 1 and 6 μg/mg SC), other enzymes should also play a role in the reduced level of CER EOS. In addition, it has been shown that neutral and acid sphingomyelinase (aSMASE), catalyzing the last step from precursor sphingomyelin to CER NS or CER AS, reduced in activity in AD skin compared to control. The reduction in activity was more drastic in lesional skin than in non-lesional skin. [129] Again, this is expected to reduce the level of CER NS and CER AS, while the absolute values have been reported to be increased [49, 107, 120]. This shows the complexity of the altered lipid synthesis in AD. The location of the active enzymes β-glucoceribrosidase-1 (GBA-1 is one enzyme of the total group of β-glucoceribrosidase (GBA) enzymes) and aSMASE was also examined [49]. The change in localization of the active aSMASE was related to the quantity of CER [NS]+CER[AS] in SC. In addition, the quantity of the remaining CERs (all CERs except CER NS+CER AS) was related to the location of the activity of GBA-1. Kezic et al. reported that GBA activity (not specified to GBA-1) and glucosylcholesterol correlated with TEWL and levels of interleukin (IL)-1a and IL-18 [130]. Although an excellent relation was shown between GBA activity (again not specifically GBA-1) and glucosylcholesterol concentration, the GBA activity is extremely high. The fact that GBA has been extracted from the SC-strips and is therefore not in its natural environment (pH, presence of saponins may play a role) and in addition the non-specificity of the method may contribute to this high activity. Furthermore, the observed total amount of CER H (including CER NH, CER AH and CER EOH) was around 15% per mg protein in lesional skin, being higher than in non-lesional skin. In almost all other studies absolute values of CER H in non-lesional skin is higher than in lesional skin [49, 107, 119].

Although many questions are unanswered, it is clear that a joint effort of several enzymes are responsible for the reduced amount of CERs in AD skin compared to control. The extent to which each of these enzymes contributes might depend on the properties of their local microenvironment, such as pH, the concentration of Ca²⁺, an imbalance in cell proliferation and differentiation, inflammation, and a change in the microbiome composition. To support this statement, it has been shown that bacterial derived ceramidase was effective in degrading CERs in the presence of S. aureus [131, 132]. Furthermore, metabolites of CERs (such as sphingosine-1-phosphate stimulate the production of inflammatory mediators (tumor necrosis factor (TNF)-α and IL-8) [133] and it has been reported that TNF-α and INF-γ affect the biosynthesis of lipids in the skin [134, 135].

Besides enzymes affecting CER subclass composition in AD skin, studies focused also on the enzyme expression in relation to the chain length and degree of unsaturation of the FFAs and CERs. An interesting study was published by Ito et al. in 2016 [119]. In that study it was shown that the increase in short chain CER NS-C34 (total chain length 34 carbon atoms) in lesional skin compared to non-lesional skin, correlated with the gene and protein level of ceramide synthase 4 (CerS4), the enzyme catalyzing acylation of C16, C18, C20, C22 and C24 fatty acid to the sphingoid base. However, this does not explain the higher levels of particularly CER NS-C34 in AD skin. It is suggested that this may be due to the higher level of especially FFA C16 and FFA C18 in lesional skin [119]. In another study, the gene and protein expression of other enzymes were reported to be altered: i) ceramide synthase 3 (CerS3), involved in the acylation of very long fatty acids (≥ C26) to the sphingoid backbone, ii) stearoyl CoA-desaturase (SCD-1) catalyzing the step of the transformation of saturated fatty acids to unsaturated fatty acids that can be rate limiting, iii) Elongase 1 (ELOVL1, elongation of fatty acids from C20 to C26) and iv) elongase 6 (ELOVL6, elongation of fatty acid C16 and C18) [136]. The expression of SCD-1 was clearly increased in lesional skin of AD compared to controls and could explain the higher level of muFFAs in lesional skin of AD patients [109]. The ELOVL1 expression was reduced in the uppermost layers of the viable epidermis, which may relate to a reduced chain length of the FFAs. An altered expression of CerS3 in involved skin was observed, which may contribute to the lower level of CER EOS in involved skin of AD patients. However, in this study only relative amounts of CERs were determined, and in addition no enzyme activity was examined. This makes correlation with the lipid composition uncertain.

Although some important insights are already provided concerning the change in expression, activity and localization of several enzymes involved in the synthesis of CERs and FFAs in AD skin compared to control skin, many steps need to be undertaken to fully understand the change in lipid synthesis in AD skin.

2.1.3. Filaggrin mutations and lipid composition

Filaggrin ïs a barrier protein and is a building block of the cornified envelope. Filaggrin mutation is an important predisposing factor for AD [137–139]. For this reason the relation between filaggrin mutations and the lipid properties in human skin was examined. Jungested et al. was the first to report about the effect of filaggrin mutation on the CER subclass composition [140]. All studies were in non-lesional forearm skin. Reduced levels were observed in CER EOS, CER EOH and CER AH in AD skin compared to the control group, but no clear correlation could be established between filaggrin mutations and changes in CER composition. This study already indicated that filaggrin mutations do not affect CER subclass composition dramatically. In the studies of Janssens et al. and Danso et al. similar results were obtained, but now including both, chain length distribution and CER subclass composition [108, 136]. However, Janssens et al. showed a correlation between lipid properties and the level of natural moisturizing factor, a degradation product of filaggrin (only non-lesion skin was analyzed). This may suggest a common underlying factor being responsible for reduced natural moisturizing factor levels and changes in the lipid properties. The relation between filaggrin mutation and lipid metabolism was further investigated in subsequent clinical studies. In pediatric patients mechanistic pathways were identified using transcriptomic analysis and it was observed that lipid metabolism pathways were important in AD patients, but were independent of filaggrin genotype showing the absence of a relationship [141]. Emmert et al. also saw almost no effect of filaggrin mutation on the CER subclass composition, but noticed that levels of FFA C22:1 and FFA C24:1 were significantly increased in filaggrin mutated patients. [116] Another study reporting filaggrin mutations and lipid properties is that of Angelova-Fischer et al. [142]. When focusing on the ceramide composition filaggrin mutations induced a reduction in relative and absolute CER EOH level only in lesional skin. Human skin equivalents have also been used to study the effect of filaggrin knock down on the barrier and lipid properties. Mildner et al. was the first to study this relationship [143]. They did not observe a relationship between filaggrin mutations and lipid composition in SC, but a reduced skin barrier in filaggrin knock-down tissue was observed [143]. However, this conclusion was not based on kinetic (permeability) studies, but on the localization of a fluorescent label in the skin at a single time-point, which in principle does not provide information on the skin barrier. In more recent studies, the effect of filaggrin mutation on the lipid synthesis was also studied by van Drongelen et al. [144]. This study reported that filaggrin mutation did not affect the lipid composition, lipid organization (lamellar phases and lateral packing) and the skin barrier function. The latter was investigated by measuring permeation of n-butyl p-aminobenzoate.

Summarizing, there is strong evidence that filaggrin mutations do not affect the lipid composition and organization in SC of AD patients. Although many times proposed, the question is still whether filaggrin mutations contributes in a direct manner significantly to a reduced barrier in AD patients: there is only a slight increase in TEWL in ichthyosis vulgaris patients with filaggrin null mutations [145].

2.1.4. Inflammation and lipid composition in atopic dermatitis

In a very interesting study lipid abnormalities were identified using a meta-analysis of four different AD clinical datasets to determine the differentially expressed genes, and lipid abnormalities were identified [146]. Genes of important epidermal lipid enzymes such as ELOVL3 (encoding elongase 3 important for the elongation of fatty acids with chain lengths between 16 and 20 carbon atoms), FA2H (encoding fatty acid 2-hydroxylase involved in the synthesis of 2-hydroxysphingolipids) and FAR2 (encoding the enzyme Acyl-CoA Reductase 2), were strongly downregulated and correlated with upregulation of Th2 genes. These observations showed a strong relation between inflammation and epidermal lipid synthesis in clinical settings. In another very interesting study that combined SC lipid analysis of AD patients with studies of IL-13 specific transgenic mice, a higher concentration of short chain CER NS, especially those having a sphingoid base of C18 and a fatty acid of C16, was detected [147]. In addition a similar change in sphingomyelin was reported, sphingomyelin being a precursor of CER NS and CER AS. Furthermore, a strong increase in the level of short chain lysophosphatidyl cholines and a strong reduction in the long chain lysophosphatidylcholines were reported. This illustrates again the relationship between Th2 cytokines and changes in the lipid composition. In the same study the expression of elongases was also assessed in patients. In lesional AD, the expression of ELOVL1 and ELOVL4 was increased, while the expression of ELOVL3 and ELOVL6 was reduced. However, these changes on mRNA level cannot explain the changes in lipid profiles as ELOVL3 and ELOVL6 elongate the shorter fatty acids (C16:0 to C20:0), while ELOVL1 and ELOVL4 increase the very long and ultra-long fatty acids (C20:0 to C32:0 or even longer) [106, 109, 136, 148]. Very illustrative were the studies using the IL-13 transgenic mice, showing the same changes in lipid profile and thus demonstrating that inflammation affects lipid synthesis. The effect of inflammation on the lipid biosynthesis was also examined by van Smeden et al.: in the same patient alterations in FFA chain length and disordering of the lipids deviated more in inflamed lesional skin than in non-lesional skin [109].

Several studies were undertaken using cultured systems, in which the culture medium was supplemented with cytokines. Hatano et al. studied the effect of TNF-α (pro-inflammatory cytokine) and IFN-γ (Th1 cytokine) on the three enzymes involved in the ceramide synthesis or ceramide breakdown; acid-ceramidase, aSMASE and GBA [149]. Supplementation of TNF-α and INF-γ to organ skin cultures or living skin equivalents reduced the expression of ceramidase, while it increased the gene expression for GBA-1 and aSMASE. In the same study an increased quantity in the CER levels was observed. This is remarkable as it opposes all findings in clinical settings [49, 101, 107]. In an in vitro study using human skin equivalents, the effect of TNF-α and Th2 cytokine (IL-4, IL-13, IL-31) supplementation on the lipid SC composition and organization was investigated [134]. TNF-α alone or in combination with Th2 cytokines reduced the chain length of FFAs, the level of CER EO and affected the lipid organization. The expression of both CerS3 and ELOV1, which are involved in synthesis of CER EO, were reduced.[134]. In another simultaneously published paper it was shown that IFN-γ also affected lipid metabolism [135]. IFN-γ reduced the level of ELOVLs thereby reducing the chain length of CERs. One may conclude that indeed inflammation affects lipid biosynthesis in lesional AD skin substantially and this concerns enzymes involved in the biosynthesis of the CER subclasses as well as those synthesizing the FFAs.

2.1.5. Lipids esterified to the cornified envelope in stratum corneum of atopic dermatis

Since the covalently bound CERs are derived from the free CER EOS, CER EOH and CER EOP in AD patients, it would be expected that the composition of the CERs esterified to the cornified envelope would change when the composition of the precursors altered. Macheleidt et al. studied the bound lipids in non-lesional and lesional AD skin and observed a reduced level of bound CER O relative to the bound fatty acids [106]. An increased activity of the enzyme ceramidase may play a role, but this was not investigated [121]. In a recent study by Boiten et al., no changes were observed in the total amount of CERs bound to the cornified envelope in AD skin compared to controls. Because these studies were performed using LC/MS, the chain length and the degree of unsaturation could also be assessed [58]. Four CER subclasses were detected, namely CER OdS, CER OS, CER OP and CER OH. In the bound CER composition there was an increased level of CER OS and a higher fraction of unsaturated CERs similar to observations in the free CER composition. These changes were more pronounced in lesional skin compared to non-lesional skin. Furthermore, there was a preference of shorter chain CERs O to be linked to the cornified envelope as also observed in healthy skin and human skin equivalents [56]. Changes in bound CER profile also correlated with a reduced barrier function in AD.

2.2. Psoriasis

Motta et al. [18] reported for the first time the CER profile in the SC of psoriasis patients together with a reduced barrier in lesional skin monitored by TEWL. Two new CER fractions were discovered, but not further identified (referred to as CER 2_II and CER 5_I). In lesional skin CER EOS, CER NP, CER NP C16/C18 and CER AH were strongly reduced, while together with the two new fractions, the amount of CER NS was strongly increased. Alessandrini et al. studied enzymes that directly affect the CER profile [150, 151]. In lesional and non-lesional skin, a reduction in prosaponin mRNA and protein expression was observed [150]. Prosaponin is the precursor of saponin, which is a cofactor for the enzyme GBA. This means that the absence of saponin probably reduces the amount of CERs in SC. Further studies of the same group showed that levels of GBA on mRNA and protein levels were decreased in non-lesional psoriatic skin, but higher and differently located in lesional skin [150, 151]. Alessandrini et al. suggested that an impaired skin barrier triggers the synthesis of GBA-1 to compensate for this reduced barrier. In psoriasis skin an increase in the level of CER NS is observed (with glucosylceramides and sphingomyelin as precursors), while e.g. the amount of CER NP and CER EO (with glucosylceramide precursor) were reduced [18]. Tawada et al. studied CER subclass levels using LC/MS, and analyzed the amount and chain length distribution of CER NH, CER (AdS+NP) and CER AP [135]. In lesional psoriasis skin, the percentages of short-chain CERs increased, while the percentages of long-chain CERs decreased. The supplementation of IFN-γ to the medium of three-dimensional skin cultures resulted in an effective reduction of mRNA levels of ELOVL1 and CerS3. It was shown that the long chain CER NS reduced in a dose dependent manner, suggesting that in psoriasis, especially the Th1 cytokine IFN-γ affects CER synthesis in the epidermis. Yokose et al. studied the ratios of various CER subclasses in a clinical setting and showed that CER NP and CER NH subclass ratios relative to CER AS or CER NS were reduced in non-lesional skin, and reduced even more in lesional skin [113]. The observed CER subclass ratios showed the same trends as in AD, especially the CER NP/CER NS ratio, which is correlated negatively with the TEWL values.

2.3. Netherton Syndrome

Netherton Syndrome (NTS) is a rare genetic skin disease with a deficiency of lymphoepithelial kazal-type related inhibitor (LEKTI) and shows severe atopic manifestations. In absence of inhibition of serine proteases by LEKTI, the skin shows severe scaling and erythema. The first data on lipid analysis (samples harvested from legs) was published in 1994 [152]. Differences in the quantities of free CER subclasses reported together with 2 subclasses of bound CERs for NTS and control skin were small. Unfortunately, no information was provided on the severity of the disease in these patients. Fartasch et al. visualized the NTS skin using electron microscopy. Disturbed premature lamellar body secretion, severe parakeratosis and lipid droplets in the corneocytes were reported [153]. Furthermore, the lipid lamellae of the SC were disturbed. All these changes indicated severe disturbance in lipid synthesis.

In 2014 van Smeden et al. reported changes in lipid composition and organization [154]. Changes were observed similar to those observed in AD patients, although the variation in CER and FFA composition was much more pronounced: a decrease in FFA chain length was observed along with a drastic increase in the fraction of unsaturated FFA compared to control skin that occurred most prominently in chain lengths of 20 and 24 carbon atoms. Although CERs were not quantified, the LC/MS chromatograms showed almost no CER EO in some patients, and an increase in CER AS and CER NS concentrations, while the CER NP concentration was strongly reduced. These changes in lipid composition had a dramatic effect on the lamellar organization in some patients: in 2 out of 8 patients no lamellar ordering could be detected, while in others the lamellar organization changed substantially from, or was similar to the lamellar phases in control skin. The conformational ordering of the lipid chains monitored by FTIR decreased. In the same study the expression of several enzymes involved in CER lipid synthesis was also assessed. The expression of SCD-1 was observed in all viable epidermis layers, while in control skin it is only detected in the basal layer. With respect to GBA-1 and aSMASE, the GBA-1 protein expression was more diffuse and less confined to the interface between SC and SG, while the protein expression of aSMASE varied substantially between NS patients. How this was related to the changes in lipid composition was not clear. A follow up paper focused more on the expression and activity of enzymes GBA-1 and aSMASE in situ. The activity of both enzymes was measured in their natural environment [155]. In control skin the GBA-1 activity was mainly located at the interface between viable epidermis and SC and to a lesser extent also in the intercellular regions in the SC. In NTS patients there was a large variation in the GBA-1 activity and location in SC. In some patients the GBA-1 activity was virtually absent, in other patients it was homogeneously distributed in the SC intercellular layers, while in some NTS patients the GBA-1 activity was present in both the intercellular regions as well as in corneocytes. Activity of aSMASE in control skin was primarily present at the interface and to a lesser extent in the intercellular regions in a patchy appearance. In NTS patients either an aSMASE activity spread homogenously in SC, or aSMASE activity was almost absent. The activity of GBA-1 and aSMASE was counteracting: a high activity of GBA-1 was observed in regions with low activity of aSMASE. The CER subclass quantification was also reported: high concentrations of CER AS and CER NS were observed, while CER NP and CER EO subclass concentrations were reduced in most patients. The NTS patients with scaly erythroderma showed the highest change in enzyme activity score and the most dramatic changes in CER subclass levels.

3.0. LIPID MODEL SYSTEMS

When comparing diseased skin with control skin in clinical studies and also in animal studies several changes in lipid composition may occur simultaneously. It is therefore very difficult to determine which of these changes are crucial for a change in lipid organization that results in barrier dysfunction. In order to study the effect of a single change in lipid composition, lipid model systems are an attractive tool. However, when using them, it is crucial that they represent similar lipid interactions as encountered in SC. One of the challenges is the preparation of these mixtures as the lipids form crystalline phases. In vivo the crystalline lipid organization is created in a temperature gradient between 30°C and 37°C. Within this temperature range the lipid membranes turn gradually from a liquid phase to a crystalline phase [156]. This phase change occurs due to a drastic change in lipid composition (glucosylceramides, sphingomyelin and phosphoglycerides turn into CERs and FFAs). In the in vitro situation using lipid model systems, a liquid phase is created by elevating the temperature of the mixtures, after which the lipids self-assemble during cooling. Several steps in this preparation of the samples are crucial as explained below.

In most cases, the first step in preparing crystalline lipid mixtures is spraying a lipid solution onto a support. Ideally, during this spraying small droplets should be created consisting of lipids dissolved in an organic solvent. A part of the solvent should evaporate before it reaches the support to minimize the droplet size on the support. This will reduce the chance for phase separation during the drying process. This means that the spraying rate (we use 14 s/μL), the lipid concentration and the distance between the nozzle and support (we use 1 mm) are critical parameters. After spraying, the lipid sample will be equilibrated at elevated temperatures, typically between 60 to 90 °C, depending on the lipid composition and the material of the support (e.g. mica, glass, polymer filter, silicon). This step is also crucial and should be carried out while avoiding phase separation (forming additional phases not encountered in SC). For this it is important that the lipid system is equilibrated close to the order (solid)-disorder (liquid) phase transition temperature region. Furthermore, it is also important that no contraction of the sample occurs during heating and equilibration. Contraction will result in large droplets creating excellent conditions for phase separation. Contraction depends on the equilibration temperature, the composition of the lipid mixture and the support material and is most often encountered when using silicon supports (used for neutron diffraction). After equilibration at elevated temperatures, the lipid temperature should be reduced in a controlled manner. Next, the hydration of the sample is performed. This should be carried out for a sufficient long time at slightly elevated temperature (higher than room temperature) to facilitate the diffusion of water into the sample, preferably above the orthorhombic-hexagonal phase transition temperature. Finally, the sample needs to be equilibrated at room temperature. Some studies show that it can take days or a week until a thermodynamic equilibrium has been reached [157].

In this review we will first discuss the early studies followed by the systems containing CER EO. Finally fundamental studies on simpler mixtures will be discussed. This review will not discuss papers focusing on molecular simulations. Recently, this has been reviewed elsewhere [158].

3.1. Early studies of model systems

The first studies focusing on single CERs were published by the group of Pascher [159–162]. Several important properties of CERs were reported; such as i) the thermal history of the samples is an important determinant for the obtained crystal structure, ii) fork-shaped arrangements in crystal structures of CERs were reported (fork-shaped indicates an angle between the sphingoid base and acyl chain less than 180 °, but larger than 90 °), with the acyl chain and sphingoid base localized in different layers of the structure, and iii) hydrogen bonding in the head group regions were strong and played an important role in the ceramide arrangement. Many years later Moore and Mendelsohn reported findings that were in line with these results; using FTIR they observed that the acyl chain and sphingoid base of CER NS showed a different phase behavior suggesting different locations in the structure [163]. They also noticed the importance of hydrogen bonding in the head group regions and detected only a low level of water in these structures. At elevated temperatures, enhanced hydration of the head groups was observed. In a follow up study using a single CER, the group compared sphingosine based CERs (CER NS and CER AS) with phytosphingosine based CERs (CER NP and CER AP) and reported that the sphingosine based CERs were more densely packed (orthorhombic packing) compared to their phytosphingosine (hexagonal packing) counterparts, while the hydrogen bonding network of the phytosphingosine based CERs was much denser resulting in a higher order-disorder transition temperature of the latter [164]. Furthermore, the addition of an α-hydroxy group (CER AS or CER AP) resulted in water permeation (detected by proton-deuterium exchange) into the polar head group regions below the order-disorder transition.

The group of Shipley also studied the phase behavior of single CERs, in particular CER NS and CER AS [165, 166]. They also observed that hydration reduced the order-disorder phase transition temperature. In addition, a complex phase behavior was observed for CER NS. Several publications of Raudenkolb et al. focused on CER NP, CER AS and CER AP and also reported polymorphism that depended on the thermal history [167–170]. Furthermore, hydration was only observed after heating above the melting temperature of the CERs and did not result in swelling, but in pure CER AS samples hydration affected crystallinity.

Several of these early studies used isolated CERs. In studies published by Wertz and Abraham stable liposomes were formed using CERs isolated from porcine SC, CHOL, FFA C16 and a relatively high level of cholesterol sulphate. The presence of CaCl₂ induced fusion of the liposomal bilayers, while the presence of CHOL increased the thermal stability of the bilayers [171–173]. The liposomal fusion was compared with the fusion of extruded disks in the intercellular space between stratum granulosum and SC and showed that Ca²⁺ might accelerate this fusion. In another very remarkable study it was shown that CER EO and glucosylceramides EO (precursors of CERs) could link bilayers [174]. This was the first illustration of the exceptional role of CER EO in the SC lipid organization. These results were interpreted in line with a molecular model the group proposed based on the broad-narrow-broad lucent pattern observed in ruthenium tetroxide fixed SC. [10, 69] Simultaneously, electron spin resonance was used to probe the mobility of the lipids in the intercellular space in membrane couplets and showed that these lipids were highly ordered [175]. Parrot and Turner reported that acetone extracted porcine SC lipids resulted in the formation of the LPP with a repeat distance of 13.5 nm together with crystalline CHOL similar to the phases observed in porcine SC [76, 176]. When using chloroform/methanol extraction, a lipid phase with a repeat distance of 5.4 nm was detected. Unfortunately, no lipid composition of the extracted lipids was provided. Importantly, they showed for the first time that the LPP could be formed in the absence of proteins. When using a combination of brain CERs (CER NS and CER AS, but having a larger percentage unsaturated acyl chains compared to CER AS and CER NS in SC) a doublet single layer was observed with a repeat distance of around 10.5 nm. Many years later this phase was also reported by Pullmannova et al. [177].

3.2. Lipid model systems including ceramide EO

As CER EO, see Figure 2, have an exceptional structure and they are thought to play an important role in the formation of the lamellar phases in SC, the phase behavior of mixtures that include CER EO will be reviewed separately from those studies in which no CER EO was included in the mixtures.

3.2.1. Models with isolated porcine ceramides

In 1996 the first studies were performed, in which samples with lamellar stacks were prepared with an from SC isolated porcine CER fraction (porcine CERs) enabling the formation of similar lamellar phases as observed in SC [178, 179]. The porcine CER composition reported by Bouwstra et al. is provided in Table 5. In these studies a mixture of FFA C14, FFA C16 and FFA C18 (saturated and unsaturated) was selected. McIntosh et al. did not report the CER composition and combined porcine CERs and CHOL with FFA C16. [180] The porcine CER:CHOL:FFA mixtures were studied by X-ray diffraction. Both publications report the effect of the CHOL:CER ratio, the CER composition and the presence of FFA C16 or the FFA mixture on the formation of the lamellar phases and lateral packing. In the presence of porcine CERs (including all subclasses), CHOL and FFAs, the LPP was formed with a repeat distance of approximately 12–13 nm. Both studies reported that the formation of the LPP was not very sensitive to the ratio of porcine CER:CHOL and the presence of these short chain FFAs. However, in the absence of CHOL hardly any lipids assembled into the LPP. The intensity distribution of the peaks, with the 2^nd order being most intense, in the diffraction patterns were very similar to that observed in intact SC, signifying that the phase observed was the same as that detected in intact SC; for more details see in the paragraph at the end of this section [64, 72, 181]. The lateral packing was mostly hexagonal as shown by a reflection at around 0.415 nm in the wide-angle X-ray diffraction pattern. Hydration did not affect the phase behavior, as indicated by an almost complete absence of swelling of the lamellae, and thus a minimal amount of water in the head group regions. Only in the studies of Bouwstra et al. a SPP and a sharp peak with a spacing of 4.2 nm were also detected probably because they used a highly collimated synchrotron radiation X-ray beam in combination with high resolution detection [178]. When the FFA mixture was replaced with a mixture of very long chain FFAs (most abundant FFAs are FFA C22 and FFA C24, referred to as FFA_mix), the 4.2 nm phase was not formed suggesting an improved mixing, the LPP repeat distance increased and an orthorhombic packing was observed [178]. The studies demonstrate that long chain FFAs are crucial for the formation of the orthorhombic packing, while CERs and CHOL are crucial for the formation of the lamellar phases. In addition, the FFA_mix enhanced the solubility of CHOL in the two lamellar phases. Interestingly, in porcine SC, the lipids formed mainly a hexagonal lateral packing. The reasons for this are not clear, but CERs with shorter chains may play a role, especially of CER AS [30]. Importantly, neither proteins, nor bound lipids were required for the formation of the SPP and LPP, but bound lipids are of eminent importance for the orientation of the lipid lamellae in SC [182].

Table 5.

Composition of synthetic human and porcine ceramide (CER) mixtures. These mixtures were used in model systems. Only mixtures having at least 3 CER subclasses are included. The composition is expressed in percentages. All percentages of the synthetic CER mixtures are in m/m, while the isolated human and porcine CER mixtures are in w/w. Synth= synthetic; Iso: isolated. For abbreviations CER subclasses, see Figure 2 and abbreviation list page 2.

	Synth porcine CER mixture				Iso. porcine CER mixture		Synth human CER mixture			Iso. Human CER mixture
CER	[196] de Jager et al.a,b m/m%	[210] Janssens et al. m/m%			[44] Mojumdar et al. w/w%	[178] Bouwstra et al. w/w%	[220] Uche et al. m/m%	[206] Opalka et al.d m/m%	[217] Opalka et al.e w/w%	[198] Bouwstra et al. w/w%	[93] Sochorova et al. w/w%
EOSc	15	15		15	13.8	7.8	15	10	0	9.4	5.1 EOdS
NdS							13	8.2	8.86
NS	51	78.8		57	44.2	55.4	13	7.5	8.10	31.7	17.8 +NdS
NP	16		45.3	17.9	8.8	17.6	30	22.8	25.27	21.6	37.3
AS	4	6.2			5.9	3.6	13	23.4	25.92	13.4f	4.8 + AdS
AP	5		14.2		9.7	5.6	16	26.8	30.45		6.1
EOH										8.4	4.8
NP C16	9		25.5	10.1	16.8AS short	9.9 AS short
AH											7.2
AdS								1.3	1.40
EOP			15								2.7
NH										15.3	13.3
NS	66	93.8		72	62	63.2	28	17.5	17.8	41.1	22.9

^a= Variation in CER subclass composition has been examined. Many other papers used this composition, also by changing the CER EOS content to 40%.

^b= AS with short chains have been replaced by NP C16 as a that time synthetic CER AS C16 was not yet commercially available

^c= In the isolated mixtures the CER subclasses have a wide chain length distribution. In the synthetic mixtures CER EO has an acyl chain length of 30 carbon atoms, except for CER EOP in Janssens et al having 27 carbon atoms. Other CERs in the synthetic mixtures have an acyl chain length of 24 carbon atoms, except for CER NP C16.

^d= Levels of CER EOS content varied and replaced by other CER EO subclasses as well, 10% CER EOS is one of the selected CER EO contents. The ratio between the other CERs is kept equal.

^e= Levels of CER EO varied between 0 and 10 m/m%. CER EO consists of either EOS, EOP, EOdS, EOH, or a mixture of them.

^f= This also includes CER AP + CER AH

Changes in CER subclass composition were also investigated. McIntosh et al. reported that when using only CER EOS, a lamellar phase with a repeat distance of 9.5 nm was detected [179]. Other drastic variations in CER composition were mixtures with only CER EOS and CER NS, and a mixture excluding the most hydrophilic subclass (CER AP). Lipids in both mixtures formed both the LPP and SPP, demonstrating that the phase behavior of mixtures prepared with these isolated CERs was not very sensitive to the CER subclass composition [183, 184]. This is differently from that reported in early studies from the group of Mendelsohn and Moore using a single CER subclass showing high sensitivity of lipid organization to the type of CER subclass [164, 185]. In additional studies that examined particularly the effect of CER EOS, it was shown that without CER EOS only a small fraction of lipids formed the LPP [179, 186]. For the first time it was demonstrated that CER EOS is crucial for the formation of the LPP in line with the studies of Abraham et al. using liposomes [174]. Although at skin surface temperature (around 32 °C) and below the phase behavior of porcine CER:CHOL:FFA (in equimolar ratio) was very similar to the lipid organization in SC, at elevated temperatures, between 45°C and 50°C, an additional strong peak started to appear in the diffraction pattern of the equimolar porcine CER:CHOL:FFA mixture, which is less dominantly present in intact SC [187]. In all the described mixtures with FFA_mix and a varying CER composition an orthorhombic packing was detected using wide angle X-ray diffraction.

As a pH gradient, Ca²⁺ gradient and a change in cholesterol sulphate concentration are observed in SC, it was also of interest to determine the effect of these parameters on the lipid phase behavior [47, 188–192]. It was shown that cholesterol sulphate, a component with a low concentration in the lipid matrix in SC, is important for the solubility of CHOL in the lamellar phases. Furthermore, cholesterol sulphate reduced the fraction of lipids forming the SPP and this phase even disappeared at very high concentrations of cholesterol sulphate [193]. Ca²⁺ counteracted the effect of cholesterol sulphate. Increasing the pH, which is expected to increase the charge density in the head group region due to the ionization of FFAs in the mixtures, had almost no effect on the phase behavior. Importantly, after the addition of cholesterol sulphate the phase behavior at elevated temperatures mimicked more closely that observed in human and porcine SC: the strong peak appearing between 45°C and 50°C drastically reduced [193]. In a study of the chain length distribution of the CER subclasses, in a more recent paper compared the phase behavior of isolated porcine CERs mixed with CHOL and FFA_mix with mixtures of synthetic CERs with the same subclass composition, but having a uniform CER chain length [194]. Both mixtures formed the SPP and the LPP. However, the ordering of the lipids in mixtures prepared with isolated porcine CERs is much lower than that of lipids in mixtures prepared with synthetic porcine CER mixture.

Remark on correctly identifying the LPP.

The diffraction pattern attributed to the LPP is characterized by an exceptional intensity distribution: the 1^st and 3^nd order diffraction peaks have a lower intensity than the 2^nd order diffraction peak [178]. The intensity distribution of the X-ray diffraction pattern (including also higher order diffraction peaks) is determined by the electron density distribution of the repeating unit (that is the repeating building block, also referred to as the unit cell) of the lamellar phase, see Figure 3. Therefore, only a diffraction pattern with this characteristic intensity distribution and a corresponding repeat distance between approximately 12 and 13.5 nm can be assigned to an LPP. This intensity distribution has been observed in the SC of porcine, human, mouse skin and SC of human skin equivalents, as well as in mixtures prepared with human and porcine CERs and in many mixtures prepared with synthetic CER mixtures [64, 72, 74, 178, 179, 181, 195, 196]. However, the lamellar phase with a periodicity of around 10.5 nm as reported by Parrot and Turner and later on by Pullmannova et al. and Kovacik et al. are based on diffraction peaks with a different intensity distribution and therefore these lamellar phases are not the same as the LPP [176, 177, 197].

3.2.2. Models with isolated human ceramides

Until the beginning of the 21^th century only isolated porcine CERs were used as porcine SC was more readily available than human SC. However, in 2001 and 2002 a few studies using isolated human CERs were reported [198, 199]. At that time 8 CER subclasses were identified in human SC, including CER EOS and CER EOH, both present at around 8–9 w/w% each in the CER mixture, see Table 5. The effect of CHOL on the lamellar phases was examined. It was shown that with only CHOL and the isolated human CER mixture mainly the LPP could be formed, although the 3^rd order phase attributed to the LPP was high in intensity compared to the 1^st and 2^nd order diffraction peaks. This may indicate an additional phase in this mixture. When FFA_mix was added to these mixtures, both the LPP and the SPP were formed, very similar to that of the porcine CER containing mixtures as indicated by a similar intensity distribution of the diffraction peaks (see section 3.2.1 for more details). In addition, besides the very abundant presence of the orthorhombic lateral packing, a fluid phase could also be identified. An increased level of FFA_mix enhanced the solubility of CHOL in the lamellar phases, similarly to that in isolated porcine CER:CHOL:FFA_mix mixtures. A reduction of FFA_mix as well as a 50% replacement of FFA_mix by FFA C16 did not affect the lateral packing [199]. This is slightly different from the findings with the porcine CER mixtures, in which a full replacement of FFA_mix by FFA C16 resulted in a hexagonal packing. This difference might be due to the longer acyl chains in human CERs compared to those in porcine SC, which may enhance the formation of the orthorhombic lateral packing [21, 30, 44]. A 50% replacement of FFA_mix by FFA C16 is probably close to the maximum fraction that can be incorporated into the lamellar phases as reported in a recent study using a synthetic porcine CER mixture [111]. Furthermore, besides an orthorhombic and probably a hexagonal packing, also a liquid packing was encountered.

To study the effect of CER EOS on the phase behavior, the same CER subclass mixture, but without CER EOS was mixed with CHOL and FFA_mix. This resulted in primarily the formation of the SPP. [198] This demonstrated again that CER EOS is crucial for the formation of the LPP. Importantly, in the absence of CER EOS a liquid packing could not be detected. For the first time this showed that there is a relationship between the formation of a liquid packing and the presence of CER EOS in these lipid mixtures.

When CER EOS is replaced by its synthetic counterpart with either an esterified oleate or linoleate, the LPP was abundantly present. However, when CER EOS was replaced by its synthetic counterpart with an esterified stearate, the LPP was absent and no liquid lateral packing could be detected [198]. These studies demonstrated the importance of the unsaturated C18 chain esterified to the ω-hydroxy acyl chain of CER EOS for the formation of the LPP. The importance of an unsaturated C18 acyl chain for the formation of the LPP has also been demonstrated in mixtures with a synthetic porcine CERs mimicking composition [200].

Recently Sochorova et al. also used isolated human CERs, see Table 5 for the subclass composition. Using an equimolar ratio of isolated human CERs, CHOL and FFA_mix formed both the LPP and SPP confirming the earlier studies [93, 201]. However, in the studies of Sochorova et al. the percentage of total CER EO was lower, which resulted in a more prominent formation of the SPP as can be concluded from the peak intensities of their diffraction patterns. When the CHOL concentration was reduced varying between an 1:1:1 human CER:CHOL:FFA and 1:0:1 ratio, no changes were induced concerning the lateral packing, chain ordering and phase transitions [93]. This is different from mixtures prepared with synthetic porcine CER composition that showed in the absence of CHOL, mainly a hexagonal lateral packing [92]. However, Sochorova et al. added cholesterol sulphate to their mixtures, which may take over the role of CHOL [202]. In the other study from Sochorova et al., an increasing amount of glucosylceramides with a fatty acid chain length of C18 replaced a portion of isolated human CERs in the equimolar mixture of CERs, CHOL and FFA. An abundant phase separation was noticed [201]. However, as human CERs have a wide chain length distribution and a longer acyl chain length compared to the uniform and much shorter incorporated glucosylceramides, it is likely that the difference in chain length plays an important role in the observed phase separation.

In conclusion, whether porcine CERs or human CERs are used, the phase behavior in the mixtures is very similar: CER EO are required for the formation of the LPP, while long chain FFA are required for the formation of the orthorhombic lateral packing. The lamellar phases can be formed in the absence of membrane proteins and bound lipids and mimic the SPP and LPP observed in SC.

3.2.3. Models with synthetic porcine ceramide composition

3.2.3.1. Initial studies and preparation method

In the beginning of the 21^th century, the first studies were carried out with synthetic CERs in which CER EOS was also included. Using synthetic CER mixtures, forming the LPP appeared to be much more difficult than with isolated CERs. The thermal history appeared to be crucial as also observed by the group of Pascher for single CERs [159, 160]. In the initial experiments an equilibration temperature at around 60 °C was selected similar to that of mixtures with isolated CERs. However, no LPP was formed. [203] In follow up studies, the equilibration temperature used in preparing mixtures was varied between 60 °C and 100 °C [204]. In these studies, the CER mixtures contained CER EOS (around 15 m/m% of the CERs), CER NP, and brain CER AS (brain CER IV with variation in acyl chain length and partially unsaturated acyl chains) [204]. This CER composition was used because several of the CER subclasses present in porcine SC were not yet available. In the absence of FFA, the LPP was formed most abundantly after equilibration at 100 °C. When adding FFA_mix, the most optimal equilibration temperature for the formation of the LPP was reduced to around 80 °C. This undoubtedly shows that in preparing model mixtures with synthetic CERs, an optimal equilibration temperature that depends on the lipid composition is an important factor for the formation of the LPP.

Besides the formation of the LPP, lipids also assembled in the coexisting SPP. In addition separate crystalline phases were formed, most probably CER NP rich phases as this CER subclass forms a very stable crystal structure [161]. The intensity distribution of the diffraction peaks attributed to the LPP with a repeat distance of around 12 nm was very similar to that of isolated porcine lipid mixtures, demonstrating the similarity in lamellar organization of the mixtures (see section 3.2.1 for more details). In an additional study, the concentration of CER EOS could be directly related to the ability of forming the LPP, confirming yet again that CER EOS is crucial for formation of the LPP [205]. Partial replacement of CER EOS by CER EOP did not affect the lamellar phase behavior. However, when CER EOP was used as the only CER EO class, the formation of the LPP was less abundant compared to mixtures with CER EOS as the only member of the CER EO [205]. More recently this was also reported by Opalka et al. [206].

3.2.3.2. Mixtures with synthetic porcine ceramide composition

Between 2000 and 2005 most of the CER subclasses that are also present in porcine SC became available as synthetic counterparts, while several CER subclasses specific for human SC remained unavailable. As a consequence, model mixtures could be prepared mimicking closely the lipid composition in porcine SC, but not that observed in human SC [196]. At that time CER AS with a short acyl chain was not yet available. For this reason, the short chain CER AS was replaced by CER NP C16. Using a synthetic porcine CER composition, CHOL and FFA_mix the lipids formed the LPP and SPP without additional crystalline phases, see Table 5 for the CER composition used. The mixtures mimicked closely the lamellar organization in porcine SC, as well as that observed in mouse SC, human SC and in SC of human skin equivalents [64, 72, 74–76, 207]. Furthermore CER EOS was required for the formation of the LPP just as it is in isolated CERs and CER mixtures prepared with CER NP and brain CER AS [204, 205]. A variation in the fraction of other CER subclasses specifically, a large reduction in CER NS and an increase in CER NP and CER AP, while keeping the CER EOS fraction constant, did not affect the phase behavior dramatically. This shows again, like the mixtures prepared with isolated CERs, that the phase behavior in these complex mixtures is insensitive to the CER composition, with exception of CER EO. This is not the case for simple systems with only 1 or 2 CER subclasses and no CER EO [179, 183, 208, 209]. Furthermore, the X-ray diffraction profile suggested a similar lipid arrangement in the unit cell of the LPP of the mixtures prepared with either synthetic or isolated porcine mixtures (see for details in section 3.2.1). The lipids also formed an orthorhombic lateral packing similar to the mixtures with isolated CERs, but different from that in porcine SC [183, 184, 196, 210].

Skin lipid model systems can provide very important information about the impact of a specified modification in lipid composition for the lipid organization and lipid barrier function as these can be examined in the same model, see Figure 6. For studies of barrier function, the lipids are sprayed onto a porous membrane, which can be clamped between the donor and acceptor compartment in a diffusion cell. The first study reporting such an experiment compared the presence and absence of CER EOS on the barrier function and lipid organization in a model with the synthetic porcine CER mixture [211]. These studies revealed that in the presence of 15 m/m% CER EOS of the total amount of CERs, the LPP and SPP were both formed, resulting in an improved barrier compared to mixtures in the absence of CER EOS formed only the SPP.

Figure 6.

Lipid membranes studied by different methods.

After spraying and equilibration, the resulting lipid membrane can be studied using various methods.

Lipids are sprayed on a support. After equilibration and hydration, the sprayed lipid membrane model can be used for either permeability studies of a model compound including water (TEWL measurements). The lipid models can also be used to examine the lateral packing, conformational ordering, hydrogen bonding network, or mixing properties of the lipids. Finally, the sprayed lipids can be examined by diffraction to determine the long range ordering, lateral packing, and the position of water or the lipids in the unit cell.

In another study synthetic CERs, which have a rather uniform chain length and isolated porcine CERs, which have a wide chain length distribution, composed of almost the same CER subclass composition were mixed with CHOL and FFA_mix [44]. Both mixtures formed the LPP and SPP, but the mixture with synthetic porcine CERs showed an increased fraction of lipids forming an orthorhombic lateral packing and a much higher conformational ordering than the mixture prepared with isolated CERs (lamellar ordering can be obtained from the width of the diffraction peaks at half maximum). In permeability measurements of the two membranes the hydrocortisone flux was reduced 13-fold in the mixture with synthetic CER composition [44]. This shows that chain length distribution has a drastic effect on the permeability of the membranes.

3.2.3.3. Increase in ceramide EOS fraction enhances the formation of the long periodicity phase

The presence of CER EOS is crucial for the formation of the LPP [179, 186, 196]. However, the studies of Mojumdar et al. showed for the first time that increasing the CER EOS concentration to 40 m/m% (of the total synthetic porcine CER fraction), produced lipid models that only formed the LPP [79]. At 30m/m% the LPP was also the only lamellar phase, but then the 4^th order diffraction peak was at the same position as the diffraction peak of crystalline CHOL, which hampers analysis using neutron diffraction. Crucially, the intensity distribution of the diffraction peaks attributed to the LPP is the same in the 15 m/m% and 40 m/m% CER EOS models. [212] This shows that an increase in the percentage of CER EOS results in a higher fraction of lipids forming the LPP, while the electron density distribution and thus the lipid arrangement in the unit cell of the LPP remains very similar. For more details, see section 3.2.1.

3.2.3.4. The importance of unsaturated fatty acids esterified in ceramide EOS.

Several studies have been performed to provide more details on the role the unsaturated esterified acyl chain plays in the phase behavior. Importantly, for mixtures with porcine synthetic CERs, CER EOS with an esterified unsaturated acyl chain is required, as seen in mixtures prepared with isolated human CERs [198, 200]. However, important details were obtained from these mixtures by polymerized transfer solid-state NMR. It was shown that the esterified linoleate of CER EOS was in an isotropic phase, while the remainder CER EOS molecule, and also the other lipids, were in a crystalline state. This resulted in an unusually low mobility of the CER EOS in the lipid structure [213].

In a related study, CER NS (60 m/m%) and CER EOS (40 m/m%) with an esterified oleate were used instead of synthetic porcine CERs. Solid state ²H NMR, FTIR and Raman spectroscopy were performed on equimolar mixtures of these two CERs with CHOL and FFA_mix in which either the oleate of CER EOS, the acyl chain of CER NS or the FFA were deuterated. Isotropic nanodroplets of deuterated oleate of CER EOS were detected, while the other lipids (CER NS and FFA) remained in a crystalline packing [210, 214]. The lipid droplets remained liquid even at a temperature of −30 °C, showing that due to environmental constrains, the oleate was not able to solidify. These findings undoubtedly show that liquid droplets are present in the LPP unit cell surrounded by a crystalline environment. The presence of the isotropic phase offers the possibility of the linoleate to fold back in the LPP. This has been suggested based on the location and width of the position of linoleate in the LPP unit cell using neutron diffraction [97]. This also explains why it was impossible to form an LPP when CER EOS with an esterified stearate was incorporated in the mixture [198, 200].

3.2.3.5. Variation in lipid class ratio of ceramides, cholesterol and free fatty acids

Variation in CHOL levels:

increasing the CHOL level in the synthetic porcine CER:CHOL:FFA molar ratio to 1:2:1 did not affect the phase behavior, but resulted in a high fraction of CHOL that phase separated [215]. However, it reduced the flux of benzoic acid across these membranes. The same results were obtained by Basse et al. in membranes with a CER composition closer to that of psoriatic SC [216]. Mojumdar et al. performed additional studies varying the CHOL content between 1:0:1 and 1:1:1 molar ratio for the synthetic porcine CER:CHOL:FFA_mix mixture with a 40 m/m% CER EOS of the total CER content to form only the LPP [92]. In the absence of CHOL, there was no formation of the LPP and mainly a hexagonal packing was observed. This illustrates that CHOL is required for the formation of the LPP as was also observed for mixtures prepared with isolated porcine CERs and CER NS [92, 178, 179, 217]. When the CHOL molar ratio was raised to 1:0.2:1 only the LPP was formed. FTIR studies revealed an excellent mixing between CER and FFA_mix. A further increase in CHOL concentration did not change the phase behavior, although CHOL phase separates into crystalline domains when CHOL concentrations are higher than 1:0.5:1 CER:CHOL:FFA.

Variation in FFA or CER levels:

when increasing either the CER levels keeping the CER composition the same or increasing FFA levels, the basic lipid organization remained the same, albeit a small part of either the CERs or FFAs phase separate into CER-rich and FFA-rich domains [215].

3.2.3.6. Variation in free fatty acid chain length distribution and degree of unsaturation

As both, an increase in the degree of unsaturation and a reduction in fatty acid chain length are encountered in inflammatory skin diseases, the effect of these changes on the lipid organization was examined using a synthetic porcine CER composition, but with 40 m/m% CER EOS to form exclusively the LPP [111]. The FFA C16 content of the FFA_mix was gradually increased keeping the ratio between the other FFA chain lengths equal. Only after m/m% of FFA C16 was increase to 40% of the total fraction of FFA_mix phase separation could be detected in the lamellar phases. Furthermore, a lower fraction of lipids formed the orthorhombic packing, while the size of the orthorhombic domains remained the same. Although this shows the lamellar phases are not very sensitive to an increase in the FFA C16 concentration, barrier function is: TEWL values showed that at 40 m/m% FFA C16 the lipid barrier is reduced significantly. Most probably this reduction in barrier function is due to a higher conformational disordering of the FFA C16 and a reduction in the fraction of lipids forming an orthorhombic packing. In previous studies a reduction in the fraction of lipids adopting an orthorhombic packing also caused a reduced lipid barrier [44, 111, 218, 219].

An increase in the concentration of muFFAs has also been studied using the synthetic porcine CER mixture [219]. When either FFA C22 or FFA C24 is replaced by their monounsaturated counterparts, the lamellar phases did not change, but the fraction of lipids forming a hexagonal packing increased and an increase in conformational disordering was noticed. As a consequence, the flux of water and hydrocortisone across the membranes increased.

3.2.4. Human ceramide composition

Studies with a synthetic CER composition mimicking that in human SC have also been performed [206, 217, 220]. Two human CER mixtures have been used in three studies; the CER compositions for these mixtures are provided in Table 5. In all three studies the lipids formed both, a SPP and an LPP, as in the mixtures prepared with synthetic porcine CERs. In the studies of Opalka et al. when a single CER EO (either CER EOdS, CER EOS, or CER EOP) was used at least 20% m/m CER EO was required to form the LPP, while in the study of Uche et al., the LPP was clearly formed at 7% m/m CER EOS [206, 220]. The reasons for this difference are not known. Almost certainty this is not caused by a difference in CER composition; the LPP has been formed at low CER EOS levels in mixtures with only CER NS, with CER NS and brain CER NP, as well as with the human CER mixture [205, 221]. Most probably the preparation method used in the group of Vavrova was suboptimal for the formation of the LPP. However, besides these differences, there are also similarities; for example both groups showed that in the presence of CER EOP the formation of the LPP was less efficient than in the presence of CER EOS. [205, 206] Opalka et al. reported the important finding that a mix of CER EOS, CER EOdS and CER EOP enhanced the formation of the LPP more efficiently than each of the individual CER EO subclasses. The study of Uche et al. focused on the changes in lipid composition mimicking that in AD skin [220]. They showed that a reduction in FFA chain length diminished the fraction of lipids forming the orthorhombic packing, while a reduction in the level of CER EOS reduced the formation of the LPP. Finally, the reduction in CER NS acyl chain length resulted in an increase in the conformational disordering of the CER NS acyl chain. When comparing the changes in composition, the reduction in FFA chain length was most effective in reducing the lipid barrier [220].

3.2.5. Reduction in number of ceramide subclasses

Janssens et al. studied the effect of a reduction in the various CER subclasses on the lateral packing and lamellar phases [210]. In these studies the synthetic porcine CER composition was compared with a CER composition consisting of only phytosphingosine-based CERs (CER EOP, CER NP and CER AP), or only sphingosine-based CERs (CER EOS, CER NS and CER AS) or non-hydroxy-based CERs (CER EOS, CER NS and CER NP). The CER EO level was 15 m/m% in all compositions, while the remaining CERs in the mixture keeping the same ratio as in the original porcine CER composition, see Table 5. All three mixtures formed the LPP and SPP demonstrating that the formation of the lamellar phases was not sensitive to shifting the composition from one type of subclass (e.g. non-hydroxy CERs) to another type of subclass (e.g. phytosphingosine CERs), although in the phytosphingosine-based mixtures additional phases were encountered. However, there is a clear difference in ordering: the non-hydroxy-based CER mixtures showed by far the highest lamellar ordering [212]. When focusing on the conformational ordering of the lipid chains, at 20 °C the highest ordering was obtained for the phytosphingosine-based CER containing mixtures, but in these mixtures a higher fraction of lipids formed a hexagonal packing. This showed that a dramatic change in head group architecture affects the lipid ordering as well as the lateral packing in the LPP, but has less effect on the lamellar phases. Phytosphingosine CERs have been shown to form a higher density of hydrogen bonding, than their sphingosine counterparts. This leads to more ordering, but also implies a larger sized head group due to the increased number of OH groups (enhancing the hexagonal lateral packing) as was already observed by Rerek et al. [164]. The formation of a hexagonal packing might be caused by the presence of CER NP C16 and CER AP C24, as when CER EOS is combined with only CER NP C24, an orthorhombic packing was observed [222]. In these mixtures no excessive phase separation was encountered: deuterated FFA showed some splitting of the scissoring vibrations, indicating domain formation, but this splitting is resolved between 30 and 40 °C, which is during the orthorhombic to hexagonal transition of these systems. In fact, this study showed that a reduced number of synthetic CER subclasses can form similar lamellar phases, but the lateral packing and conformation ordering can be affected by the CER subclass choice. This was also observed in mixtures prepared with isolated CERs [179, 183, 184].

Several studies were performed with further reduction in the number of CER subclasses. Simpler systems have the advantage that more detailed information can be obtained on the mixing properties of the lipids, the arrangement of lipids in the unit cell of the lamellar phases and the conformational ordering of individual lipids. For this, selective deuteration of the lipids is required. The first study that combined CER EOS with a single CER subclass that formed the LPP was reported in 2016 [217]. In that study CER EOS was at 30 m/m % of the total amount of CERs and the CER NS comprised at 70 m/m % were combined in an equimolar mixture with CER:CHOL:FFA + 5 w/w% cholesterol sulphate. Besides the LPP, phase separated CHOL was detected [217]. Reducing the CHOL concentration below the equimolar CER:CHOL:FFA ratio resulted in the formation of both the LPP and SPP indicating that CHOL enhanced the formation of the LPP [93]. In another study 40 m/m% of CER EOS together with CER NS (60 m/m%) was used in an equimolar ratio of CER:CHOL:FFA and again only a LPP was formed [212]. The diffraction peaks attributed to the LPP had a similar peak intensity distribution as at lower CER EOS levels indicating that the unit cell of the lamellar phase did not change, but that there was an increase in the fraction of lipids forming the LPP; see section 3.2.1 [212]. The majority of the lipids displayed an orthorhombic lateral packing.

In the study of Opalka et al. the permeability of model compounds and water (monitored by TEWL) appeared to be higher in the presence of CER EOS than in its absence [217]. This would suggest that adding CER EOS reduces the lipid barrier function contrary to the results of de Jager et al. using a synthetic porcine CER mixture [211]. However, when comparing the orthorhombic to hexagonal packing transition temperature range (detected by the stretching frequencies in the FTIR spectrum) of the mixture in the absence of CER EOS, the temperature of the flux studies (32°C) is below the orthorhombic-hexagonal phase transition, while in the presence of CER EOS, it is within the orthorhombic-hexagonal phase transition temperature range and thus a higher fraction of lipids are hexagonally packed. Because generally the dense orthorhombic packing improves the barrier function compared to a hexagonal packing, the shift in transition temperature could contribute to the difference in results between the two studies.

In another study 40 % m/m CER EOS was combined 60 m/m % of either CER NS C24, or CER AS C24, or CER NP C24 or CER AP C24 in an equimolar CER:CHOL:FFA_mix ratio [222]. The lamellar organization, lateral packing and permeability of the membranes was assessed. In agreement with previous studies using simple systems in membranes with CER EOS, a large subfraction of the lipids were in an orthorhombic packing, except for the CER AP-containing membranes, in which the hexagonal packing was more prominent [164]. Both the CER NS- and CER AS-containing membranes formed only the LPP and phase separated crystalline CHOL, while in the CER NP- and CER AP-containing membranes additional phases were detected. Flux studies revealed that phytosphingosine-containing membranes showed an improved barrier properties compared to the sphingosine-based membranes. This is likely due to a stronger hydrogen bonding network in the head group regions that counteracts the effect of an increased lipid fraction forming a hexagonal phase that is expected to reduce the barrier.

The mixture with 40 m/m% EOS and 60 m/m% CER NS was also used to study the effect of increased concentration of CER NS C16 (which has a total chain length of 34 carbon atoms that is often encountered in diseased skin) at the expense of CER NS C24 in an equimolar CER:CHOL:FFA C24 ratio [112]. Again the LPP was quite insensitive to the increased concentration of CER NS C16; only a change in the diffraction pattern was encountered when 75 m/m% of CER NS C24 was replaced by CER NS C16, while a smaller fraction of lipids formed an orthorhombic lateral packing [112]. In much earlier papers, CER EOS was combined with CER AP C18 together with CHOL and a single FFA varying in chain length [223, 224]. In these studies the LPP did not form, although the CER EOS level was quite high, typically around 50 m/m% of the total CER content or even higher. Several explanations can be provided for this. First, in all other studies CER subclasses with longer acyl chains (primarily 24 carbon atoms) were used, while in this study CER AP C18 was selected. This shorter acyl chain might reduce the ability of the lipids to form the LPP. Second, the use of CER AP C18 might, together with CHOL and FFA C16 (or FFA C18), form a very stable crystalline structure due to the extensive network of hydrogen bonds as reported by the authors [224]. In fact, the results of these studies strongly suggest that CER AP C18 dictates the final structure and not CER EOS, in which longer FFAs interdigitate and CER EOS had almost no effect on the lamellar phase periodicity. Recently it was shown that even CER AP C24 in combination with CER EOS forms the LPP less abundantly than the combination of CER EOS and CER NS C24 [222]. Third, the choice of the equilibrium temperature (not specified in details, but higher than 60 °C) might have been too low to form the LPP. In a follow up study from the same group the LPP was formed with an esterified methyl branched acyl (palmitic acid) CER EOS and CER AP C18 [225]. In that study the equilibration temperature was around 70–75 °C and a small fraction of lipids formed the LPP, although it is expected that the methyl branched CER EOS is less efficient in forming the LPP than its unsaturated counterpart (see section 3.4.3.2). This indicates that a lower equilibration temperature might indeed have played a role.

3.2.6. The localization of lipid subclasses in the unit cell of the long periodicity phase

To understand in detail the role each of the lipid classes and subclasses play in the lipid organization, it is very important to localize the lipids within the unit cell of the lamellar phase. The crystallographic unit cell is defined as the basic building block from which the large crystallographic lattice can be constructed (for further explanations of the unit cell, see Figure 3). X-ray diffraction relies on the electron density contrast between the different components and provides information on the electron density distribution. This includes the localization of the head group with a higher electron density, the swelling of the lipid lamellae and sometimes the location of the lipid tails. Additional detailed information can be provided by neutron diffraction using deuterated compounds, which have a high neutron scattering contrast with respect to protiated compounds, which allows identification of the position in the unit cell.

The first studies that provided electron density profiles of the unit cell of the LPP in model systems was reported in 1998 using isolated porcine CERs [186]. The electron density profile consisted of a series of high and low electron density blocks indicating a three layer arrangement in the unit cell. Electron densities were calculated in an iterative process until the calculated intensities of the diffraction peaks matched the intensities of the measured diffraction pattern. In 20.03 McIntosh was able to induce swelling of the unit cell by replacing CHOL with cholesterol sulphate in a lipid mixture of isolated porcine CERs with FFA C16 [202]. He reported a symmetric electron density profile of the unit cell, in which CHOL and cholesterol sulphate were located asymmetrically in the unit cell. The electron density profile consisted of two lipid layers with a swollen layer in between. This layer was suggested to be water, see Figure 7. Although this was a big step forward in understanding the lipid arrangement, the resolution of the electron density profile was poor. In 2009 Groen et al. used a methodology similar to the swelling method as small changes in the length of the repeating unit were introduced by selecting slightly different compositions, while assuming that the electron density profiles of the unit cell remained similar [96]. The obtained resolution of the electron density profile was much higher than the results reported by McIntosh [202]. This approach resulted in a unit cell with an electron density profile consisting of three layers of low electron density indicating the localization of the hydrocarbon tails and four regions of high electron density suggesting the localization of the head groups. The layer of water proposed by McIntosh appeared to be also a low electron density region (but not as low as the other electron density regions). When comparing the electron density profiles of these two studies, one has to realize that the center of the unit cell in the studies of Groen et al. was shifted half a unit cell, thus the unit cell center in Groen et al. is at the unit cell boundary of McIntosh, see Figure 7 [96]. However, the choice of the center of the unit cell is arbitrary and does not affect the result. Groen et al. built a molecular model based on this electron density distribution with CER EOS head group at the cell boundaries spanning the two outer lipid layers. The localization of the other lipids remained very uncertain, as no information could be obtained by X-ray diffraction about their specific locations.

Figure 7.

Similarity of the calculated electron density profiles of the long periodicity phase unit cell.

A comparison of the electron density profiles of the LPP unit cell observed by McIntosh (top; 2:1:1 molar ratio of isolated porcine CERs:CHOL:FFA C16) and Groen et al. (bottom, 2:1:1 isolated porcine CERs:CHOL:FFA_mix) [96, 202]. The layer suggested by McIntosh as a water layer in the electron density profile is the same as the central hydrocarbon layer in the model of Groen et al. In both models, the electron density is higher than the other two hydrocarbon layers, but lower than in the head group regions. The dashed lines indicate equal positions with high electron density in two profiles. The position of CHOL in both unit cells is the same in both studies (shown by an arrow in both unit cells, see also Figure 8). The CHOL position in the McIntosh unit cell model was determined by the swelling method, while the position of CHOL in the Groen unit cell was determined by neutron diffraction (see Figures 8E and 8F [97]. The figures are adapted from [96, 202].

To determine the locations of the various lipids, neutron diffraction studies are very useful. The advantage of neutron diffraction is that the localization of water can be obtained by a variation in the D₂O/H₂O ratio of hydrated samples. In addition, the location of deuterated moieties of lipids or perdeuterated lipids can be determined by measuring the diffraction pattern of the protiated mixture and that of the mixture with the deuterated moiety/perdeuterated lipid of interest, after which the scattering length density (SLD) profiles, equivalent concept as the X-ray derived electron density profile, can be calculated. Kiselev et al. were the first to perform such studies using a simple system discussed in section 3.3.4 [80]. Introducing this technique for studying SC lipid systems was a big step forward in understanding the detailed arrangement of the lipids in the unit cell of the LPP and SPP.

In all neutron diffraction studies the first step in the analysis is the localization of water in the unit cell. This implies that the unit cell consists of three lipid layers, a central lipid layer located between two boundary lipid layers, similar to that observed in electron density profiles obtained from the X-ray studies (compare Figure 8 and Figure 7). In the next step, by using (partially) deuterated lipids and the fully protiated mixture, the deuterated lipids are localized in the unit cell by subtracting the SLD profile of the protiated mixture from the profile containing a (partial) deuterated lipid [97]. This provides the position of the deuterated lipid. Thus for the lipid mixture containing synthetic porcine CERs, CHOL and FFA_mix, Mujamdar et al. localized CHOL in the outer layers with the hydroxyl group close to the inner head group regions, see Figure 8 [97]. This is at the same position as McIntosh identified cholesterol sulphate by his swelling experiment and shows that the results match, see arrows in Figure 7 [202].

Figure 8.

The location of the water and lipids based in the unit cell of the long periodicity phase based on the scattering length density (SLD) profiles.

A 1:1:1 CER:CHOL:FFA m/m ratio was used in all studies. (A) The SLD profile of water [79]. A higher SLD value means a higher concentration of water. The water is expected to be located at the lipid head group in the unit cell. (B) The SLD profile of the lipid mixture containing synthetic porcine CERs with deuterated linoleate esterified to the ω-hydroxy acyl chain of CER EOS. The corresponding location of linoleate is also shown. [97] (C) The SLD profile of the lipid mixture containing synthetic porcine CERs with the terminally deuterated sphingosine chain of CER EOS. (D) The SLD profile with perdeuterated FFA C24 [79]. (E) The SLD profile of a lipid mixture containing synthetic porcine CERs with deuterated head group of CHOL [97]. (F) The SLD profile of a lipid mixture containing synthetic porcine CERs with deuterated tail of CHOL [97]. (G) The SLD profile of the lipid mixture containing CER NS either perdeuterated acyl chain or perdeuterated terminal moiety of the sphingoid base [90]. For abbreviations CER subclasses, see Figure 2.

Mojumdar also reported the position of the esterified linoleate chain (using deuterated linoleate) of CER EOS using contrast variation (now changing the deuterated/protiated linoleate ratio): linoleate is positioned in the inner head group region and slightly protrudes into the central lipid layer of the unit cell. This is slightly shifted from that proposed by Groen et al. [96]. The width of the domains suggest that linoleate is folding back and must therefore be highly flexible. This is in agreement with the existence of liquid droplets of the esterified oleate or linoleate versions of CER EOS, see also section 3.2.3.4 [213, 214]. This study also confirmed the location of the EOS head group proposed by Groen et al. at the cell boundary of the unit cell [96]. Very recently CER EOS with terminally deuterated sphingosine chain became available. As shown in Figure 8, these deuterated moieties are located in the outer layers of the unit cell (unpublished results), confirming that the sphingosine chain of CER EOS is present in the outer layers of the LPP.

Neutron diffraction studies using either perdeuterated acyl chain of CER NS or the perdeuterated FFA C24 revealed that both chains are primarily located in the inner lipid layer of the three layer unit cell of the LPP [79]. Based on the width of the inner layer, the FFA C24 chains and acyl chains of CER NS C24 are interdigitating. The studies of Mojumdar et al. did not provide information on whether CER NS, is in a hairpin or linear conformation. However, very recently making use of CER NS with a perdeuterated acyl chain and CER NS with a terminally deuterated sphingosine chain, it was shown that CER NS molecules with head groups in the inner head group region are primarily in an extended conformation: the acyl chains are mainly in the central lipid region, while the terminal sphingosine moiety is in the two outer layers [89, 226]. This was confirmed by FTIR studies showing that the deuterated acyl chain of CER NS together with deuterated FFA form large domains in the unit cell of the LPP. This excludes the possibility that a high fraction of CER NS is in the hairpin conformation, which would require sphingosine chain localization in the inner layer of the unit cell. Whether CER NS is in a simple system comprised of CER EOS, CER NS, CHOL and FFA_mix or in a complex system prepared with a synthetic porcine CER composition does not affect the outcome of the studies [89]. Therefore, the arrangement of the CER NS is not very sensitive to the CER composition. In a very recent study, CER NP was also localized in the LPP. In this study CER NS and CER NP each were present at 30 m/m% and CER EOS was 40 m/m % of the CERs. The acyl chains of CER NS and CER NP were located primarily in the inner layer confirming earlier studies. Based on FTIR studies, it was also shown that CER NP is primarily present in a linear arrangement [90].

The effect of head group interactions was also investigated. When replacing CER NS by separate sphingosine and fatty acid molecules, the sphingosine and CER NS were located at the same position. However, when CER NS was replaced by two fatty acids, losing its specific head group interactions, the fatty acid was evenly distributed in the unit cell, demonstrating that specific head group interactions are important to concentrate the sphingosine moiety at a specific position in the LPP unit cell [227].

3.2.7. Molecular models for the unit cell of the long periodicity phase

Several models are reported to describe the lipid organization in systems with CER EO. The first published model from Swartzendruber el al. was based on the ruthenium tetroxide pattern using electron microscopy. [69] At that time there was no further information available, except the linking of liposome membranes by CER EOS [174]. This model is provided in Figure 9A. Around 10 years later two additional models were presented based on ruthenium tetroxide staining and X-ray diffraction. These were developed to include i) the role of CER EOS as linking lipid layers and required for the LPP formation, and ii) the electron density in the unit cell of the LPP showing a trilayer arrangement of the lipids in the LPP (see Figure 9B and 9C) [179, 186]. In both models CHOL is located in the central layer together with short chain CERs and the linoleate of CER EOS [186, 228]. This is different from the more recent models, but at that time no information was present on the localization of the individual lipid (sub)classes and their arrangement. In 2012 another model was presented based on the cryo-electron microscopy images that show an asymmetric pattern that was refined by simulations, see Figure 9D [88]. This model shows CERs in a linear asymmetric repeating arrangement. The sphingosine chain is located next to CHOL, while fatty acid is neighboring the acyl chain of the CERs. No specific role is attributed to CER EO in linking the various layers. This asymmetric unit cell is different from all other models. More recently based on X-ray diffraction and neutron diffraction studies, a trilayer model was presented that includes lipid localization in the unit cell of the LPP as observed by the (partially) deuterated lipids. The arrangement of the CERs is linear, see Figure 9E [97]. In this model CER EO links the various layers similar to the models presented in Figure 9B and 9C. Like the model if Figure 9D the sphingoid base is located next to CHOL and the FFAs and the acyl chains of the CERs are also neighboring. In the model presented in Figure 9E, the role of CER EOS in linking the various layers explains the correlation between the fraction of CER EOS and the observed abundance of lipids forming the LPP in clinical studies and in vitro. The disagreement between these two models is the two layer asymmetric arrangement (Figure model 9D) and the trilayer symmetric arrangement in Figure 9E.

3.3. Lipid model systems using (semi) synthetic ceramides in the absence of ceramide EO

In the absence of CER EO, most studies were performed using only a few (semi) synthetic CER subclasses in the model systems. This has the advantage that a much more detailed analysis can be performed. Most of the early studies were performed using brain CERs (derived from cerebrosides by removal of the galactose from the primary hydroxyl group of the CER) as these CERs were already commercially available in the 1980s. However, although the head group is very similar to CER NS (brain CER III) or CER AS (brain CER IV), the chain length composition has important differences compared to the CERs in SC [229]. While in SC the CERs are mainly present with saturated hydrocarbon chains, in brain CERs there is a large population of CERs containing an unsaturated acyl chain [229]. Furthermore FFA C16 was used often in these early studies. Probably this fatty acid was selected because it was reported to be a major fraction of FFA in SC at that time. More recent publications report that FFA C22, FFA C24 and FFA C26 are the dominant chain lengths [22, 23, 29, 45, 230]. Often studies were performed with (partly) deuterated lipids. This is especially an advantage when using techniques like FTIR, ²H NMR, Raman spectroscopy and neutron diffraction, as the deuterated part in the mixture can be followed exclusively due to the unique physical properties. In this sense a simple composition is also an important advantage. In many studies biophysical techniques are frequently combined with permeability measurements.

3.3.1. Studies with brain ceramides

Kitson and Thewalt studied lipid dispersions and oriented systems prepared from either sphingomyelin or brain CER NS at pH 5.2 with CHOL and FFA C16 [156, 231]. These systems are of interest because at the interface between the stratum granulosum and SC the lamellar bodies are extruded and their membrane disks with sphingomyelin and glucosylceramides are converted into lamellar phases with CERs. Studies were performed with ¹H-NMR and ²H-NMR and showed a remarkable phase shift from the liquid-ordered phase for sphingomyelin containing membranes to a polymorphic solid phase of membranes with brain CER NS. Furthermore, FFA C16 and CHOL were also present in solid phases. Above 50 °C, the solid phase disappeared and became liquid ordered and at higher temperatures an isotropic phase was encountered. The authors concluded that the interaction of CERs with CHOL are entirely different from those observed between CHOL and phospholipids. In the mixtures of phospholipids with at high CHOL content, liquid ordered phases often form [232]. The crystalline polymorphism at room temperature was confirmed by studies using X-ray diffraction. [233]. The presence of crystalline phases in the CER containing mixtures correlated very well with those observed in intact SC and showed that the intercellular crystalline phase behavior was different from that of cellular membrane systems [64].

The effect of pH was also studied and it was shown that a variation between pH 5.2 and pH 7.4 of the brain CER NS:CHOL:FFA C16 mixture revealed a polymorphism that depended strongly on pH, but below 50 °C remained crystalline [156]. At elevated temperatures for mixtures at pH 7.4 an inverted hexagonal phase was observed. Moore et al. and Velkova and Lafleur examined the same composition using FTIR [234, 235]. They also observed polymorphism and showed the presence of phase separated domains of FFA C16 and brain CER NS with both adopting an orthorhombic packing at physiological temperatures. The symmetric stretching vibrations in the FTIR spectra at elevated temperatures indicated an isotropic liquid phase or a liquid ordered phase, depending on the CHOL levels. In the group headed by Lafleur Raman micro-imaging showed microdomains of FFA C16 and brain CER NS. The phase separation was strongly enhanced by hydration [236]. Moore et al. compared the phase separation of mixtures with brain CER NS and brain CER AS and showed that both brain CER NS and brain CER AS resulted in phase separation when mixed with CHOL and FFA C16 in an equimolar ratio [208]. The effect of FFA chain length on the brain CER NS:CHOL:FFA organization was also extensively investigated. It was noticed that an increased FFA chain length clearly shifted the phase transitions from crystalline to liquid ordered and finally to an isotropic phase at higher temperatures, except for the mixture with FFA C22, in which almost no liquid ordered phase was present [237]. In these studies often deuterated FFA were used. The splitting of the CD₂ scissoring vibrations showed that phase separation occurred.

3.3.2. Studies with ceramide NS

Around 1995–2000 several well-defined CER subclasses became commercially available. The group of Moore and Mendelsohn reported studies in which CER NS C18 was used [238]. Both the acyl chain of CER NS C18 and FFA C18 were also available perdeuterated. They showed that CHOL enhanced the miscibility of CER NS C18 and FFA C18 and concluded that these lipids together with CHOL were miscible, while CER NS C18 and FFA C18 were not. Unfortunately, in the three component system no scissoring vibrations were measured with FFA C18 or CER NS with the acyl chain perdeuterated. This would have provided more details about miscibility. Several years later the same mixture was studied using ²H NMR [239]. That study showed that the equimolar CER NS C18:CHOL:FFA C18 mixture adopts a crystalline phase until 32 °C, while at 50 °C most of the lipids are in a liquid ordered phase. However, whether only one mixed crystalline phase or several solid phases coexisted, was unclear. This has been investigated for a very similar composition comprised of equimolar CER NS C16:CHOL:FFA C16, which has a hydrophobic match in chain lengths [240]. It was shown that crystalline phases coexist at temperatures below 32 °C. Phase separation in FFA C16 rich and CER NS C16 rich phases were clearly observed by Raman spectroscopy, infrared spectroscopy and ²H NMR. Interestingly upon heating the FFA C16 together with CHOL formed liquid ordered phases, in which CER NS C16 became progressively solubilized when the temperature increased. In a 2020 study of the same composition using Raman microspectroscopy, phase separation was confirmed, and an increase in storage time increased the ability to phase separate [157]. This is an important result, because it shows that at least several days of storage at room temperature might be required between sample preparation and measurement to obtain equilibrated samples. The fact that phase separation was encountered in CER NS C16:CHOL:FFA C16 samples, also means that the CER NS C18:CHOL:FFA C18 mixture likely does not entirely form homogeneous solid mixtures.

In 2013 and 2014 Skolova et al. published three studies using CER NS with an acyl chain length varying between C2 and C24 in equimolar mixtures with CHOL and FFAs [241–243]. The lateral packing as well as the lamellar phases were examined. These studies revealed phase separated lamellar phases and crystalline domains at room temperature, except for the CER NS C24:CHOL:FFA C24 mixture, which formed a homogeneous mixture showing that matching of very long chains enhances mixing. Additional studies using CER NS C16:CHOL:FFA C24 showed that FFA C24 phase separated [243]. Oguri et al. and Paz Ramos et al. compared the phase behavior of CER NS C24 in equimolar mixtures with CHOL and either FFA C16 or FFA C24. [157, 244] Mixtures with FFA C16 showed some phase separation that was clearly enhanced by hydration, confirming the results of Percot et al. [236], while CER NS C24:CHOL:FFA C24 formed homogeneous mixtures. Similar results were obtained with FFA C24 that was replaced by FFA_mix [244]. The difference in mixing properties between the FFA C16 and FFA C24 containing membranes was evident: large FFA-enriched domains were encountered in the FFA C16 containing mixtures, while the FFA C24 containing mixtures showed homogenous mixing.

In another paper ²H NMR combined with FTIR and specific deuteration made it possible to exclusively study the sphingoid base of CER NS [245]. The conformational ordering of the sphingosine chain (FTIR) was lower than in the protiated mixture. Based on the ²H NMR spectra a fluid phase was proposed. However, this is not the same as the fluid phase observed for esterified oleate and linoleate of CER EOS in the LPP system, which is isotropic as illustrated by the sharp single peak in this ²H NMR spectrum, even at −30 °C [214]. The difference is also established with FTIR. While in case of the deuterated sphingoid base the infrared stretching vibrations were around 2090 cm⁻¹ at 32 °C, compared with 2096 cm⁻¹ to 2098 cm⁻¹ for the deuterated esterified linoleate and oleate of CER EOS demonstrating a much higher conformational disordering of the latter [210, 214].

Two publications deserve additional attention [242, 245]. In these studies it is shown by specific deuteration of the FFAC24 and the acyl chain of CER NS C24 that the chains are neighboring in the orthorhombic lattice. These results demonstrate that a large fraction of the CER NS C24 is in linear arrangement resulting in an asymmetric arrangement in the unit cell. This contrasts the neutron diffraction results showing a symmetric arrangement. This will be discussed further in section 3.3.4.

The permeability of membranes prepared from CER NS, CHOL and FFA was also studied. When varying the acyl chain length of CER NS, the highest fluxes were obtained for systems prepared with CER NS C4 and CER NS C6 and thus showed that very short chain CERs may have a detrimental effect on the skin barrier. Although these very short chain lengths are not encountered in SC, they may be candidates as penetration enhancers [241].

In another paper, CER NS C24 models prepared with CHOL and either FFA C24 or FFA_mix were compared. The flux of ethyl-PABA through these membranes was approximately a factor 7 higher in the FFA_mix membrane, even though the lamellar phases and lateral packing were not different [246]. By specific deuteration of the short chain FFA, it was demonstrated that the high ethyl para-amino benzoate flux was probably due to a higher mobility of the short chain FFAs, even though FFA C16 + FFA C18 was less than 10 m/m% of the FFA mixture. In 2017 Pullmannova et al. examined the same systems (equimolar CER NS C24: CHOL with either FFA C24 or FFAmix) and also found higher fluxes through the FFA_mix membrane: approximately 2.3 fold for TEWL; about 3-fold theophyline; about 3.8 fold for indomethacin [243]. Replacing CER NS C24 by CER NS C16 resulted in a less drastic increase in flux.

Summarizing, only CER NS C24 forms homogeneous mixtures with CHOL and FFA_mix or FFA C24. FTIR data indicate a linear arrangement of the CER NS molecules, in which the acyl chain is neighboring FFA C24. The sphingosine chain together with CHOL exhibit a higher conformational disordering than the acyl chain of CER NS.

3.3.3. Variation in head group architecture of the ceramide subclasses

In 2005 the first publication focusing on the effect of head group architecture of CERs was reported by Rerek et al. [209] CER NP C18 or CER AP C18 were mixed with CHOL and FFA C18. Both mixtures showed a hexagonal packing at room temperature, while similar mixtures with brain CER NS or brain CER AS exhibited an orthorhombic packing [208, 234]. NMR studies showed that in these mixtures CER NP destabilized the orthorhombic packing in agreement with the findings of Rerek et al. [239]. Melting of mixtures with CER AP occurred at a higher temperature than the melting of mixtures with CER NP, most probably due to a more extensive hydrogen bonding network in the head group region of the CER AP mixture, although the hydrogen bonding in the mixtures was less strong than seen in individual pure CERs [209]. Together these studies showed that the head group architecture affects the phase behavior of these simple mixtures. Surprisingly in these CER AP and CER NP containing mixtures with a rather short chain length (C18) a homogenous mixture was encountered. This behavior was very different from mixtures prepared with (brain) CER NS, brain CER AS and CER NS C16. The latter showed phase separation in mixtures with CHOL and FFA C18 or FFA C16 [157].

Three studies from the group of Vavrova reported the role of the trans double bond in the head group region in CER NS. To study the effect of the double bond, membranes were prepared with either CER NS C24, CER NdS C24 or CER NP C24 mixed with CHOL and FFA C24 in an equimolar ratio [247–249]. In these mixtures only the membranes with CER NS formed a crystalline homogenous mixture, while CER NP and CER NdS showed phase separation. In contrast to the studies from Rerek et al. in all three systems an orthorhombic packing was observed probably due to the longer acyl and FFA chain lengths compared to the previous study [209]. Maybe the tight packing in CER NP C24:CHOl:FFA C24 enhanced the phase separation compared to CER NP C18:CHOL:FFA C18. Another possibility is a different arrangement of the lipids in the unit cell; In case of CER NP C18, the sphingoid and acyl chain have an equal length, which is not the case for CER NP C24: a sphingoid base of 18 carbons and an acyl chain of 24 carbons. These results show that head group architecture affects the miscibility of the lipid subclasses, which was also observed in membranes prepared with only 2 CER subclasses (CER EOS with either CER NS C24, CER AS C24, CER NP C24 or CER AP C24) [222].

In another paper from the Vavrova group the same systems were reported, but prepared with FFA_mix [197]. In addition CER NH C24:CHOL:FFA_mix was included. Membranes of the newly examined CER NH C24:CHOL:FFA_mix mixture showed phase separation and the highest flux of indomethacin, but also one of the lowest fluxes of theophyline, compared to the membranes prepared with the other CERs. A peculiar observation was that CER NH-containing mixture formed a lamellar phase with a repeat distance of 10.7 nm, which is a twice the repeat distance of that seen in CER NS C24:CHOL:FFA_mix mixtures. However, in contrast to what is mentioned in their studies, this is not the LPP as discussed earlier in section 3.2.1.

Recently another study compared the α-hydroxy CER subclasses, CER AS C24, CER AdS C24 and CER AP C24 in equimolar mixtures with FFA_mix and CHOL to their non-hydroxy CER counterparts [250]. Mixtures with both the α-hydroxy and non-hydroxy CER subclasses formed various crystal structures at low temperature, except CER NS C24. In a subsequent study CER NS C24 was combined with either CER NH C24 or CER AP C24 in an equimolar mixture with CHOL and FFA C24 (CER NH C24:CER NS C24:CHOL:FFA C24 and CER AP C24:CER NS C24:CHOL:FFA C24) to examine whether homogenous mixtures were formed [251]. The ²H NMR measurements at 25 °C and 32 °C revealed mainly crystalline phases for all components. When the temperature was increased, the different components underwent different thermotropic phase behavior indicating that the mixtures were not homogenous. X-ray data showed that CER NS C24 enhances mixing as only one lamellar phase was detected with a repeat distance of either 10.7 nm (NH containing mixture) or 5.4 nm (AP containing mixture).

Two other studies were performed using neutron diffraction. Schmitt et al. studied mixtures of CER AP C24 with either CER NS C24 or CER NP C24 at 2:1 and 1:2 ratios. These CERs were mixed with CHOL and FFA C24 in a 1:0.7:1 molar ratio. In all compositions they observed one lamellar phase with a repeat distance of approximately 5.4 nm [252, 253]. However, as the lateral packing was not investigated, different crystalline domains might still exist in these systems.

In summary when comparing the various CER subclasses, CER NS C24 with CHOL and FFA (FFA_mix or FFA C24) forms a single lamellar phase, while CER NdS C24, CER AP C24, CER AS C24, CER AdS C24, CER NP C24 or CER NH C24 mixed with CHOL and either FFA_mix or FFA C24 results in phase separation.

3.3.4. The position of ceramides, cholesterol and free fatty acids in the unit cell

Several studies were performed using neutron diffraction to obtain more details about the lipid arrangement in the unit cell. In a first series of studies CER AP C18 was used. Kiselev et al. and Ruettinger et al. studied the equimolar CER AP C18:CHOL:FFA mixture, in which the FFA chain length varied between FFA C16 and FFA C26 [80, 254]. CER AP was selected because this is one of the most prevalent CERs in human SC, although it has a much longer acyl chain in SC. For the mixture with FFA C16 the various lipids were found to all fit into one lamellar phase. The neutron analysis was very detailed. The SLD profile showed clearly the location of the head groups close to the boundary of the unit cell and also demonstrated little hydration of the head group regions, which is very different from phospholipid systems [255]. When FFA chain length was increased to FFA C18 or FFA C22, two lamellar phases were detected [254]. Previously Rerek et al. observed no phase separation for the CER AP C18:CHOL FFA C18 mixture, in a study of primarily the lateral packing and conformational disordering [209]. Further increasing the FFA chain length to FFA C24 and FFA C26 resulted in three coexisting lamellar phases, indicating that a longer FFA chain enhances phase separation in agreement with previous studies for CER NP-containing mixtures [209, 248]. These studies demonstrated that CER AP dictated the position of the FFA in the most prominent lamellar phase. In studies with deuterated CHOL in equimolar CER AP C18:CHOL:FFA C16 mixture, the CHOL head group was located close to the head group regions, while the CHOL tail was located in the center [256]. Schroeter et al. examined the interdigitation of the partially deuterated FFA C22 or FFA C26. [257] The location of the deuterated moiety of the two FFAs could be determined with high accuracy in the unit cell of the two lamellar phases in each composition. In both lamellar phases interdigitation of the FFA C22 and FFA C26 was shown [257]. In summary these studies show that CER AP is dictating the lipid arrangement and that FA interdigitation enhances with increasing chain length. A minimum amount of water is located in the structure.

In subsequent studies CER EOS was added to the mixture, but no LPP was formed. Possible reasons for the absence of the LPP were discussed in section 3.2.5 [224, 225]. In a follow up study, the CER component was changed from CER AP C18 to CER NP C24 [258]. This chain length is abundantly present in the lipid composition of the SC. Neutron diffraction was combined with ²H NMR. Neutron diffraction revealed the co-existence of two lamellar phases with repeat distance of 5.42 nm and 4.3 nm at a physiological skin temperature. Hardly any water was present in the head group regions and a symmetric arrangement of the lipids. When increasing the temperature to 50 °C there was hardly any change in SLD profile and repeat distance. The 4.3 nm phase might be crystalline CER NP in a fork-like arrangement as suggested by Pascher et al. [161]. The 5.4 nm phase has a similar repeat distance to the SPP. The NMR studies showed a highly ordered phase at 25 and 32 °C for all three components

Two additional studies were performed by the same group focusing on compositions in a CER:CHOL:FFA molar ratio of 1:0.7:1, in which CER is CER NS:CER AP in a 2:1 or 1:2 molar ratio [252, 253]. Neutron diffraction studies showed that these systems formed single lamellar phases with interdigitation of the CER acyl chains that are slightly tilted. The mixtures with CER NP:CER AP in a 2:1 and 1:2 ratio were also studied, and slight differences in the behavior of the two CER species were noticed. A single lamellar phase with a repeat distance of 5.45 nm was observed with interdigitation of the CER NP long acyl chains and to a lesser extent the CER AP acyl chain. The higher CER AP composition resulted in a higher tilting of the two CER species.

Neutron diffraction was also employed in studies by Groen et al. and Mojumdar et al. of a complex synthetic lipid system mimicking the porcine lipid composition, but in the absence of CER EOS [78, 259]. This composition revealed a lamellar phase with a repeat distance of 5.4 nm and confirmed the low hydration levels in these systems. Contrast variation using deuterated CHOL, perdeuterated FFA and perdeuterated acyl chain of CER NS C24 revealed interdigitation of the CER acyl chain as well as the FFA C24 tail. In addition, CHOL was localized in a similar way to that described by Kessner et al. [256]. The studies of Groen et al. and Mojumdar et al. could not distinguish whether the CERs were in a hairpin or in a linear conformation.

All neutron diffraction studies of simple systems containing CER AP or CER NP and the complex compositions with synthetic porcine CER composition showed a symmetric structure with the head groups at the boundary of the unit cell and interdigitation of the acyl chain of CERs and the FFAs. However, whether the CERs in the SPP showed a linear or hairpin conformation could not be concluded from these neutron diffraction data. Two studies from the group of Vavrova and Huster showed convincingly that the acyl chain of CER NS C24 is neighboring the FFA C24 demonstrating that in the SPP unit cell most of the CER NS is in a linear conformation [242, 245]. This linear arrangement of CER NS C24 is also observed in the LPP. [89] Although no neutron diffraction data are available for the CER NS:CHOL:FFA C24 mixture, it is expected that this system will also show a symmetric SLD profile. This contrasts with the findings using FTIR, which will be discussed briefly in section 4.1.

3.3.5. Molecular models for the unit cell of the SPP

The molecular arrangement in the SPP has been explored, based either on X-ray diffraction, neutron diffraction or FTIR studies. Figure 10A is a SPP model in which CER AP dominates the structure and CER EOS adjusts to this structure [224]. This model is based on studies in which CER EOS was included in the composition, but no LPP formed. All CERs are in hairpin conformation. In Figure 10B the SPP model of the porcine CER composition are shown. Although all CERs are drawn in hairpin conformation, the authors stated that whether the CERs were in hairpin or linear conformation could not be deduced from their studies [259]. In Figure 10C the most recent model is presented, in which the CERs are in a linear conformation as concluded from FTIR studies using deuterated lipids [245].

Figure 10.

The lipid arrangement in the repeating unit (unit cell) of the short periodicity phase as proposed by various molecular models.

(A) The model based on the neutron diffraction studies. The model is based on the presence of CER AP C18 in the mixture, which has a dominant effect on the formation of the unit cell. Partially deuterated FFA C22 is localized in this structure. CER EOS is arranged in the short periodicity phase spanning a whole bilayer [224]. (B) The model is based on the localization of FFA C24, CHOL head group and tail and the acyl chain of CER NS using neutron diffraction [259]. At that time it was not yet known whether the CERs were in a linear or hairpin conformation. (C) The model of the unit cell based on NMR and FTIR studies. The FTIR studies showed that the acyl chains and fatty acid form large domains suggesting a linear arrangement of the CER NS [245]. For abbreviations CER subclasses, see Figure 2.

4. DISCUSSION

During the last 40 years many more details have become available regarding the lipid composition and lipid organization in the SC of normal (healthy) skin and inflammatory skin diseases mainly focused on human skin. In this review we aimed to provide an overview of this knowledge, in which we combined information on lipid model systems with the lipid composition and organization seen in SC of healthy and inflammatory skin diseases. Below a short discussion is provided on some of the findings.

4.1. Simple model systems in the absence of ceramide EO

When using simple model systems with one or two CER subclasses mixed with CHOL and FFA, the phase behavior is very sensitive to the head group architecture. Especially in these simple systems when deuterated moieties or perdeuterated lipids are used, detailed information can be obtained on the phase behavior and even on the position of the molecules in the unit cell. In addition, information on the hydrogen bonding network can be obtained. First some general remarks about these studies.

Frequently, physical property comparisons of different lipid compositions, are performed only at 32 °C. This temperature is a logical choice as it is close to the skin surface temperature and permeability studies are often carried out at this temperature. However, choosing only 32 °C for a detailed comparison limits the interpretation because for several of the simple mixtures 32 °C is in the temperature range at which the phase changes occurs from orthorhombic to hexagonal lateral packing. Therefore, differences in lateral phase behavior or chain conformational ordering between different samples might be due to the orthorhombic-hexagonal phase change rather than to different compositions. An example is the frequently studied equimolar mixture of CER NS C24:CHOL:FFA C24, which exhibits this phase change between 30 and 40 °C [260]. Another observation is the symmetric stretching vibrations reported at elevated temperatures (typically above 70 °C): the symmetric stretching vibrations of the deuterated FFA in mixtures prepared with synthetic CER NS or brain CER NS is around 2095 cm⁻¹, suggesting an isotropic phase being supported by NMR studies [235, 240, 260]. Only at very high CHOL concentrations the symmetric stretching vibrations are at lower frequencies, typically around 2092 cm⁻¹, suggesting a liquid ordered phase [235]. An isotropic phase has also been observed in other mixtures, but using the asymmetric stretching vibrations of deuterated FFA. [164, 185] Different high temperature frequencies are reported by the group of Vavrova: symmetric stretching frequencies of deuterated FFA (close to 100 °C) are around 2090–2092 cm⁻¹ suggesting a liquid ordered phase. [242, 247, 248, 261] This is not only observed for CER NS-containing mixtures, but also for mixtures prepared with CER NP, CER AP or CER AdS [250]. Why these variations in symmetric stretching vibration at elevated temperatures occur is not clear, but could be influenced by high local CHOL levels, that may result in a liquid ordered phase [235].

Despite these hurdles some important conclusions can be made for the CER subclasses with a C24 acyl chain length, which is frequently encountered in SC, combined with CHOL and FFAs at equimolar ratios. When using one CER subclass, CER NS C24 forms a homogenous mixture with FFA_mix (or FFA C24). For all other CER subclasses (CER NdS, CER NP, CER NH, CER AP, CER AS, CER AdS) various crystalline domains are reported at temperatures up to 32 °C [247, 248, 250, 258]. When shortening the acyl chain to CER NS C16 combined with either FFA C16 or FFA C24 phase separation occurs [243, 260]. These observations show that a hydrophobic match in chain length does not always lead to proper mixing. However, when CER NS C16 in an equimolar ratio with CHOL and FFA C16 is replaced by CER AP C18 or CER NP C18 with FFA C18, no phase separated crystalline domains are detected [209, 257]. The reason for the absence of phase separation is not clear. Possibly the increased head group size and the formation of a hexagonal lateral packing plays a role. Another important finding is that the CER structure is dictating the lamellar organization, while FFAs adjust their position through interdigitation in the structure and CHOL facilitating the mixing of CER and FFA [238, 257]. Therefore, each of the lipid classes has their role in the formation of the lamellar phase.

Neutron diffraction studies using SC lipid model systems showed convincingly that a small amount of water is located close to the head group regions [80, 224]. Concerning the arrangement of the lipids in the repeating unit, using CER NS with either a deuterated acyl chain or sphingoid base chain, FTIR showed that CER NS is in a linear conformation with its acyl chain neighboring FFA C24, while the sphingoid base is neighboring CHOL, see Figure 10C [245]. However, to match with the observation of a symmetric structure from X-ray diffraction and neutron diffraction including those with deuterated lipids, it is likely that this asymmetric structure mirrors itself within the lamellar stacks. Only then both X-ray diffraction and neutron diffraction can monitor the average structure and thus detect a symmetric structure.

4.2. Similarity in lipid phase behavior in stratum corneum of human, porcine, canine and mouse

Great similarity has been observed in the SC lipid organization in several (animal) species. X-ray diffraction results show a very similar lamellar organization in human, porcine, mouse and canine SC. Ruthenium tetroxide staining also shows great similarity in lamellar pattern for porcine, human and mouse skin [10, 68, 262, 263]. This is quite remarkable, as the CER subclass composition of each species is very different (see Table 1, 2 and 5). In human SC only a small fraction (below 10 % m/m) consists of CER NS, while the subclasses CER NP, CER AP and CER NH are most abundantly present. In mouse and porcine SC CER NS comprises a large fraction of CER content [17, 21, 30, 35, 107]. These results show clearly that the lamellar organization has a large capacity to accommodate changes in the CER subclass composition.

The changes in SC lipid organization as function of temperature are slightly different in the various species. The lamellar phases in human SC are more thermally stable than in porcine and mouse SC. In porcine SC this could be due to a lower level of hydrophilic CERs (such as CER AP and CER NP) resulting in a less dense hydrogen bonding network between head groups [17, 21, 30, 108]. The difference in acyl chain lengths could also contribute to the difference in thermotropic stability: porcine SC in particular has a large fraction of CER AS with short acyl chains [30]. Concerning mouse SC, the reduced levels of hydrophilic CERs compared to human SC may play a role, similar to porcine SC. However, in mouse SC the acyl chains of CERs and FFA chains are not shorter than in human SC [17, 154, 264]. Possibly a higher level of unsaturated FFA in mouse SC may contribute to the lower thermostability of the lamellar phases [22, 264]. Most studies show very limited or no swelling of the lipid lamellae when increasing the water level in the SC of the various species [68, 72, 74, 76]. Probably the hydrogen bonding network is too strong to incorporate additional water molecules in close proximity to the head groups, which would force the head groups of different lipid layers to separate [80, 164, 209]. Of course, the linear arrangements of CERs also inhibit swelling. Practically this observation is very important as otherwise skin hydration during e.g. swimming would lead to a tremendous reduction in skin barrier function. Remarkably, packing in porcine SC at 32 °C is predominantly hexagonal, while in human and mouse SC an orthorhombic packing is predominant at 32 °C [76]. It is not clear why this happens, since higher CER NS content in porcine SC compared to human SC should favor an orthorhombic packing [30]. The shorter chains of the CERs, in particularly CER AS may play a role.

4.3. Similarities in phase behavior between complex model systems and the stratum corneum

One of the elements that needs to be discussed is the similarity and dissimilarity between lipid model systems and the lipid organization in the SC. The most complex mixtures are those prepared with either isolated porcine or human CERs. Mixtures prepared with these CER mixtures showed the LPP and SPP, even though there is a substantial difference in the CER subclass composition. Isolated human CERs used in the reported studies contained less than 30 w/w% CER EOS and CER NS (see Table 5), whereas these two CER subclasses accounted for more than 60 w/w% in porcine SC [93, 198, 201]. Yet, the lamellar phases and lateral packing are very similar in these mixtures. This shows that the lipid phase behavior is relatively insensitive to the changes in CER subclass composition. This is also observed in mixtures with synthetic CER subclasses chosen to represent the compositions of isolated porcine and human CERs. However, in all these mixtures CER NS is still present, see Table 5 [196, 217, 220]. Differences in phase behavior are observed with extreme changes in CER subclass composition. For example, formation of additional phases and/or increase in the fraction of lipids in hexagonal packing were observed in studies, in which all sphingosine based CERs were replaced by phytosphingosine based CERs, or when CER NS was replaced entirely by either CER NP or CER AP in binary CER mixtures with CER EOS combined with equimolar amounts of CERs and FFA_mix [210, 222]. These observations demonstrate that the phase behavior in the presence of CER EO is relatively insensitive towards less extreme changes in CER subclasses, except for changes in the fraction of CER EO (see below). These observations are in agreement with comparisons with human, porcine, canine and mouse SC, showing a similar phase behavior despite significant different CER subclass composition, see section 4.2. However, similar phase behavior does not necessarily mean that the permeability of model membranes prepared at different lipid composition are also similar. As shown most frequently by the group of Moore and Mendelsohn in simple lipid systems, an increase in more hydrophilic CERs, such as CER NP, CER AP, but probably also CER AH, leads to a tighter network of hydrogen bonded head groups that can increase the barrier [163, 164, 209]. This was also shown recently by Uche et al. in systems with CER EOS and only one other CER subclass mixed with CHOL and FFA_mix: substituting CER NS with CER AP reduced the membrane permeability to ethyl-PABA by a factor of 1.6, most probably due to the increased hydrogen bonding density between the head groups [222]. The permeability reduction was significant, but not large, perhaps because the fraction of lipids forming a hexagonal lateral packing also increased, which might have offset some of the effect of a higher hydrogen network density on the permeability.

Phase behavior does change with changes in the CER EO concentration, which affects directly the formation of the LPP: an increase in CER EO (frequently CER EOS has been used) enhances the fraction of lipids forming the LPP, while the lipid arrangement in the unit cell of the LPP remains the same. [212] At 30 m/m% CER EOS of the CERs only the LPP has been formed [93, 265]. The relationship between the presence of the LPP and the fraction of CER EO has also been observed in clinical studies: when the shoulder of the strong complex diffraction peak (see Figure 5) (attributed to the 3^rd order diffraction peak of the LPP), is almost absent, the level of CER EO is also reduced, see Table 4. This has been observed in AD patients as well as in healthy volunteers including a population with dry skin [95].

Furthermore, a more pronounced presence of the LPP has been observed in the SC of cultured in vitro models (human skin equivalents) and in mouse SC [64, 72, 77, 266]. This may have been driven by their higher levels of CER EO compared to native human skin. In the in vitro cultured models (human skin equivalents, primarily the epidermal model (LEM) and fibroblast derived model (FDM)), the levels of CER EO in SC are higher than in native human skin [17, 267]. In the in vitro cultured skin models, the CER EO percentage is approximately 50 w/w% in the epidermal model and around 44 w/w% in the fibroblast derived model [267]. This is approximately 30 and 25 m/m% of the total CERs and thus close to the 30 m/m% observed in the synthetic lipid mixtures that form exclusively the LPP. In mouse SC, as far as we know, the lamellar phase behavior and lipid composition have never been reported in the same animal source. However, Kawana et al. (see Table 1) reported that mouse SC contains around 16.9 m/m% CER EO in comparison with human SC, which contains around 6.3 m/m% CER EO [17]. Therefore, the SC lipid model systems show a high similarity in phase behavior with the SC lipid matrix in ex vivo skin, in vitro cultured skin and clinical studies. However, despite similarities in the formation of the LPP, lipid model systems seem to require a higher fraction of CER EO (30 m/m%) to solely form the LPP than in native skin or in in vitro cultured human skin. Other factors thus may play a role here. A possible explanation would be a slight difference in the structure of CER EO in mouse and human SC compared to CER EO used in the in vitro cultured skin. For example, the molecular structure of CER EO is slightly different in in vitro cultured human skin because more of the CER EO is esterified to an oleate, instead of a linoleate [2, 38]. However, based on studies in model systems, this is not expected to have a large influence on the formation of the LPP [198]. Possibly other lipids, so far not studied, may facilitate the formation of the LPP in native skin. A candidate is the group of cholesterol esters with an unsaturated linoleate/oleate. When these cholesterol esters are incorporated at the position of CHOL in the unit cell (see Figure 9E), these lipids may enhance the linkage of different layers by having the linoleate in the same position as the linoleate of CER EOS in the unit cell of the LPP protruding across the inner head group regions.

Concerning the lateral packing there are also several similarities between observations in the lipid mixtures and the SC in cultured in vitro human skin, ex vivo skin and clinical studies. In lipid mixtures long chain FFAs are crucial for the formation of the orthorhombic packing. This has been shown for mixtures with isolated human and porcine CERs, and for human skin equivalents [178, 179, 199, 268]. In clinical studies of AD as well as lamellar ichthyosis a reduced FFA fraction as well as a decrease in average FFA chain length could be related to the observed reduction in lipids forming the orthorhombic packing [109, 269]. In the in vitro cultured human skin models, the reported shorter chain length in CERs and FFAs, as well as a higher fraction of muCERs and muFFA, and a lower FFA/CER ratio may all contribute to the observed formation of primarily a hexagonal packing. [38, 219]

Importantly when comparing the LPP in lipid model systems with the LPP in the lipid matrix in SC, a similar intensity distribution of the diffraction peaks attributed to the LPP has been observed: the 1^st and 3^rd order diffraction peaks have a lower peak intensity compared to the 2^nd order [72, 212, 266]. Also, if the SC sample is heated to 120 °C, which is above the lipid matrix melting temperature range, in a process that is similar to that performed on the lipid model systems, the recrystallized lipid matrix in the SC forms an LPP with a similar intensity distribution of the LPP in the X-ray pattern, indicating that indeed this procedure results in the formation of the LPP.

In conclusion, many similarities have been observed in the phase behavior between model systems and the lipid phase behavior in the SC in clinical studies, in ex vivo skin and in SC of in human skin equivalents. Therefore, lipid model systems are a valuable tool to obtain more details on the interaction between SC lipids.

4.4. The presence of a liquid phase in the stratum corneum

As the lipid lamellae should follow as closely as possible the corneocyte shape, flexibility in the shape of the lipid domains is required, as e.g. dry corneocytes have entirely different shapes than hydrated swollen corneocytes [270]. This was already postulated by Forslind in his domain mosaic model presented in 1994 [271]. A possible source for this flexibility is the esterified linoleate and oleate of CER EO. Solid state NMR, Raman spectroscopy as well as FTIR all showed that these lipid moieties are in an isotropic phase in the LPP [210, 213, 214]. This is very remarkable because all the remaining lipid components are in a hexagonal or orthorhombic phase. The studies by Pham et al. even showed that except for the linoleate, the remaining CER EOS is in a highly immobile crystalline state. When extrapolating this to the SC, one would expect also to see isotropic domains in SC. This is indeed observed using solid state NMR: small liquid domains are observed in a crystalline environment [272]. These observations emphasize again the similarity between the lipid organization in model systems and the SC.

4.5. Lipids esterified to the cornified envelope in relation to the barrier

Lipids bound to the cornified envelope form an intimate contact between the hydrophobic intercellular lipids and the hydrophilic corneocytes. Furthermore, they are important for the alignment of the lipid lamellae that run approximately parallel to the corneocyte surfaces. It is also interesting that the bound lipids contain a higher level of unsaturated CER O and the chain length of the bound CER O are shorter than those of the unbound CER EO (esterified unsaturated chain is removed during the linking process) [56]. This could strengthen the skin barrier if CER EO with shorter chains is incorporated into the bound lipids leaving a narrower chain length distribution of the CER EO in the unbound lipids.

4.6. Similarities between the arrangement in the various molecular models.

When comparing the various molecular models for the LPP (Figure 9) and SPP (Figure 10), several similarities are observed. In all of the most recent models a linear arrangement (i.e. extended as opposed to hairpin) of the CERs is the preferred conformation. In these models CHOL is located next to the sphingoid base of the CERs, while the FFAs are located next to the acyl chain of the CERs. This is observed in the SPP arrangement as well as in the LPP arrangement [88, 97, 245]. In the LPP, the linear conformation of CERs between the central and outer layers together with the CER EOS, offers a tight connection between neighboring lipid layers. A linear arrangement of the CERs may reduce permeability and hinder swelling within the LPP and SPP upon hydration [214]. An extended configuration also reduces the polar head group cross section, providing a higher lipid packing density and reducing packing strain [89]. When focusing on the difference in permeability between the LPP and SPP, the tight packing region caused by the acyl chains of CERs and the FFA in the central layer is broader than the width of the acyl chain and FFA region in the SPP. In the former, the FFAs and CER acyl chains are interdigitating, in which head groups are located on both sides of this layer, while in the SPP the CER acyl chains and FFAs are located on one side of the repeating unit. This may contribute to a more effective barrier of the LPP. In addition, the fluid lipid domains in the LPP may entrap molecules also increasing the barrier further. Finally, the presence of the very long acyl chains of CER EO in the regions where CHOL and the sphingosine chain are primarily located may also increase the van der Waals forces and strengthen the barrier.

5. PERSPECTIVES IN SKIN BARRIER REPAIR

How can we combine the knowledge collected in clinical studies and in vitro measurements for new insights to repair effectively the barrier function of inflammatory skin diseases?

When focusing on these skin diseases similar changes in lipid composition have been observed in the SC of AD, NTS and psoriasis patients, see Table 6 for a comparison. The most important modulations in lipid composition that are noticed in SC of these inflammatory skin diseases are:

1. A change in CER subclass composition, in which the CER AS increased, while CER NP and CER NH are reduced. In some studies CER NS does not change and in other studies it increases significantly.

2. A decrease in CER EO subclasses

3. An average decrease in total chain length of CERs

4. An increase in specifically CERs with a total chain length of 34 carbon atoms, which is most abundantly present in the CER NS subclass

5. A decrease in the average FFA chain length (not studied in psoriasis)

6. An increase in the unsaturated FFA fraction (not studied in psoriasis)

Table 6.

Summary of changes in stratum corneum (SC) organization and SC lipid composition observed in atopic dermatitis (AD), Netherton syndrome (NTS) and psoriasis compared with control. In addition, synthetic lipid systems are provided, in which changes in lipid composition (similar to those observed in AD, NTS and psoriasis) are related to change in lipid organization that could possibly modulate the lipid barrier. No systems are included that result in clear phase separation other than phase separated CHOL. For abbreviations CER subclasses, see Figure 2 and page 2.

Skin diseases: organization and lipid composition changes relative to control
Disease	Organization changes	Changes in composition	Ref.
AD	a. Changes in lamellar phasesa b. Orthorhombic packing↓ c. Conformational disordering chains↑	FFA chain length:b↓ FFA unsaturation:↑ CER chain length:c ↓ CER-C34:d↑ CER AS↑, NSe, AHe↑ CER NP, NdS, NH, AH, EOS, EOH, EOP↓	[49, 101–109, 113, 119]
NTS	a. Dramatic changes in lamellar phases b. Conformational disordering↑	FFA chain length:f FFA unsaturation: ↑ 3% to 28.7% CER chain length ↓ CER-C34: ↑ 8.5 to 27.2% CER subclasses: CER NS/AS ↑ EOS, EOP, EOdS, EOH, NP↓	[154, 155]
Psoriasis	a. Changes in lamellar phases	CER chain length↓ CER NS↑ CER EOS, NP, AH↓	[18, 135]
Model systems with synthetic CER compositions
Model with lamellar phases	Change in organization, ordering and/or flux	Change in composition
SPP**** porcine CER or NS model	Short chain FFA increase conformational ordering Ethyl-PABA*flux: 7x↑	FFA C24 to FFAmix:g FFA chain length ↓	[246]
SPP**** CER NS model	TH flux**: 3x↑ IND flux***: 3.8x↑ TEWL: 2.3x↑	FFA C24 to FFAmix:g FFA chain length ↓	[243]
LPP/SPP**** human model	Orthorhombic packing ↓ Ethyl-PABA flux*: 2.4x↑ TEWL: 2.8x↑	FFA chain length: h↓	[220]
LPP/SPP**** porcine model	Orthorhombic packing↓ Conformational ordering: ↓ LPP and SPP not affected Hydrocortisone flux: 3x↑ TEWL: 4.5x ↑	FFA MUFA:j↑	[219]
LPP/SPP**** porcine model	Orthorhombic packing↓ Conformational ordering↓ Hydrocortison flux:13x↑	Synth porcine CER uniform chain length to isolated porcine CERs	[44]
LPP/SPP**** human model	No changes in LPP, SPP and lateral packing, Conformational disordering CER NS C34: ↑ Ethyl-PABA*: flux 1.4↑	CER NS-C34+CER AS-C34↑:k 0 to 13%	[112]
LPP****: NS/NP model	No change in lipid organization; TEWL↑	CER NS/NP ratio varied from 1: 2 to 2:1	[274]
LPP/SPP**** porcine model	LPP ↓ Ethyl-PABA* flux: 1.9x ↑	CER EO: 15% to 0%	[211]
LPP/SPP**** human model	LPP↓ TH flux**: no change IND flux***: around 2x↑ TEWL: no change	CER EOmixl: 30% to 0% CER EOdS, EOS, EOP also reported: 30% to 0	[206]
SPP**** NS model	LPP not formed Orthorhombic packing: ↑ TH flux**: 3x↓ IND flux***:2.5↓ TEWL: no change	CER EOmixl: from 10% to 0%	[217]
LPP/SPP**** human model	LPP↓ Orthorhombic packing:↓ TH flux**: no change IND flux***: 2x↑ TEWL: ↑trend	CER EOmixl: 10% to 0% CER EOS, EOP and EOdS also reported	[217]
SPP**** NS model	Slight phase separation NdS system TH flux**: no change IND flux***: no change	CER NS replaced by CER NdS	[247]
SPP**** porcine CER model	No changes lipid organization Ethyl-PABA flux*: no change	Porcine CERs replaced by CER NS	[246]
LPP**** NS model	No changes in lipid organization Ethyl-PABA flux*: no change	CER NS replaced by CER AS	[222]

^*Ethyl-PABA=ethyl-para-amino benzoic acid;

^**TH= theophylline;

^***IND=indomethacin

^****LPP= long periodicity phase, SPP= short periodicity phase

^a= reduction in LPP repeat distance and reduction in shoulder of the complex peak. This shoulder is attributed to the LPP.

^b= Average chain length of FFA reduced from 21.4 to 18.2 carbon atoms [109].

^c= Average chain length of CERs reduced from 47.0 to 44.3 total carbon atoms [109].

^d= CER NS-C34 increased from 3 to 11.4 m/m% [119]; 7 to 16 m/m%. [147]; 1 to 7 m/m% [49].

^e= Some papers report AH↑, NS↑, others not.

^f= Average chain length of FFA reduced: from 24.5 to 23.4 (C16/C18 not included) carbon atoms [154].

^g= Average chain length of FFA reduced from 24 to 22.4 carbon atoms.

^h= Average chain length of FFA reduced from 22.4 to 20.0 carbon atoms.

^j= Saturated FFA C22 (42.6%) or FFA C24 (34.7%) replaced by its unsaturated counterpart.

^k= CER NS-C34 and CER AS-C34 have a acyl chain of 16 carbon atoms, spingoid base is 18 carbon atoms

^l= EO_mix is composed of CER EOS, CER EOP and CER EOdS (approximately 78, 16,4 and 5.6 w/w%, respectively).

It has been demonstrated that the skin barrier in patients with these three inflammatory skin diseases is impaired; In AD, an increased TEWL value correlates with a reduced hydrocarbon chain length of FFA and CERs [107–109, 120]. Furthermore, changes in CER subclass composition, in which the ratios of CER NS/CER NP and CER AS/CER NP increase together with decreased levels of CER EO, strongly correlate with the impaired skin barrier as monitored by TEWL for psoriasis as well as AD [107, 108, 113]. However, from clinical studies it is difficult to select the changes in lipid composition that have the most pronounced impact on the skin barrier. This knowledge is central to designing methods to repair the skin barrier. Some of the above-mentioned compositional changes have been studied using synthetic lipid model membranes by measuring the lipid organization, conformational disordering and permeability of the lipid barrier.

As discussed in section 4.4, lipid model membrane systems mimic many aspects of the lipid organization in SC. This allows us to use these model systems for more detailed evaluation. In this review the model systems selected for evaluation have known flux data and do not show phase separation (other than phase separated CHOL). We included the non-phase separation criterion because phase separation can influence the permeability substantially and clinical studies show no evidence of phase separated crystalline domains with high order (sharp diffraction peaks), except for CHOL.

When inspecting Table 6, the observation that differences in CER subclass composition cause changes in fluxes by less than 2 fold is striking. Even when comparing compositions with and without CER EO, and thus the absence of the LPP, changes in flux of model drugs and increase in TEWL are unremarkable. In sharp contrast, a large permeability increase was observed when FFA C24 in the equimolar mixture of CER NS:CHOL:FFA was replaced by FFA_mix, which contains only 11 m/m% short FFA chains (a chain length of either 20, 18 or 16 carbon atoms) in a mixture of FFAs with an average chain length of 22.4 carbon atoms. Specifically, the fluxes of ethyl-PABA, indomethacin and theophylline increased by factors of around 7, 3.3 and 4, respectively [243, 246]. Almost no changes in the lipid organization could be detected when FFA_mix replaced FFA C24, although an increase in conformational ordering of the short chain FFA C16/C18 was noticed. This indicates that a wider distribution of FFA chain lengths can have a considerably greater impact on the flux of model compounds than changes in CER subclass composition.

An even larger (factor 14) increase in ethyl-PABA flux was observed in the LPP only version using a similar mixture (CER EOS 40%:CER NS 60%):CHOL:FFA C24, when FFA_mix replaced FFA C24 [112, 222]. This demonstrates that a wider distribution of FFA chain lengths can dramatically decrease the barrier of lipid membranes, also in the presence of the LPP. It is important to notice that these data are from two different studies. Although the permeability studies were performed in the same way using the same equipment; they were part of separate studies and were not run in parallel. Therefore, the comparison of these studies has not been reported in Table 6.

Another example is the difference in permeability of membranes prepared (i) with a synthetic porcine CER subclass composition (Table 5) that has well-defined chain lengths of the sphingoid (18 carbon atoms) and acyl chains (16, 24 or 30 carbon atoms) and (ii) with isolated porcine CERs with almost the same CER subclass composition, but a wide distribution of chain length. Here the hydrocortisone flux through the lipid membrane with a wider chain length distribution was at least 13-fold higher: calculation of the exact factor is uncertain because the curve decreased from its maximum probably due to depletion in the donor phase of the diffusion cell [44]. (iii) Finally, in a human CER model membrane, FFA chain length was also reduced, keeping the same FFA components, but increasing the short chain FFAs (16, 18 and 20 carbon atoms) at the expense of long chain FFAs (22 and 24 carbon atoms). Although the 3-fold increase in TEWL and the 2.4 fold increase in ethyl-PABA for the short chain FFAs was less dramatic then when FFA_mix replaced FFA C24, these were larger changes than were observed for all other compositional changes examined in that study [220].

How do these changes in FFA and CER chain length distribution in model systems relate to the changes in chain length distribution encountered in AD, psoriasis and NTS skin? In SC of these patients, the shorter chains of FFAs and CERs (only CERs have been examined in psoriasis skin) are much more prominently present, while the fraction of longer chains are reduced. So the balance between short and long chains has been shifted. For example in AD SC compared to control SC, the fraction of CERs with a total chain length ≤ 40 carbon atoms is 7.1 m/m% in non-lesional SC and 18.6 m/m% in lesional SC being much higher than the 5.4 m/m% in control skin [49]. Although this change in chain length distribution is different than in the in vitro lipid model systems, in both cases the fraction of short chain CERs (and/or FFA) increased with a corresponding permeability increase. Thus, changes in chain length distribution that increases the fraction of short chains or make the chain length distribution wider can be crucial for skin barrier dysfunction. Furthermore, when the chain lengths of FFAs in particular, but also CERs, are reduced to such an extent that the fraction of lipids forming a hexagonal packing increases, this will further reduce the skin barrier [178, 179].

Besides, the CER subclass composition and chain length composition a higher degree of unsaturation is encountered in AD skin and NTS skin [109, 154]. Unsaturation in AD skin is around 5 m/m% to 10 m/m%, which according to model systems should not impact the barrier function drastically. However, when unsaturation increases to around 30 m/m%, as observed in NTS, this can also have a large impact on skin barrier function [219].

These observations, although somewhat speculative, may have consequences for developing methods to repair the skin barrier. In fact, all these studies show that formulations containing one or two CER subclasses that can be delivered into the skin are unlikely to have a high impact on the skin barrier repair. A more promising approach would be to tackle those enzymes that can repair the chain length distribution, such as activating the elongases being responsible for the elongation of FFAs or the acyl chains in CERs. Of course, lipid mixtures can always be designed to form an additional layer on the skin surface to support the skin barrier, but this is certainly not the final goal in repairing the skin barrier.

As mentioned above, all inflammatory skin diseases show the same changes in SC lipid composition. Several studies showed that cytokines may induce changes in lipid synthesis in the skin, indirectly or in a direct way [134, 135, 147, 149]. Interestingly, even in human skin equivalents several of the observed changes in lipid composition are very similar to those observed in inflammatory skin diseases [38]. This indicates that inflammation is not an absolute requirement (that is not causative), changes may have also been induced by other factors such as a too high activation of epidermal keratinocytes that often occurs in human skin equivalents.

In conclusion, a robust integration between the knowledge from in vitro lipid model studies to select those compositional changes that most dramatically change the lipid barrier and clinical studies that show the deviation in lipid composition and organization is needed to develop methods that effectively repair the skin barrier.

1. Abbreviations:

AD

atopic dermatitis

aSMASE

sphingomyelinase

ATR-FTIR

attenuated total reflection Fourier transform infrared spectroscopy

CER

ceramide

CER XY Cn

ceramide

X = A

alpha hydroxy fatty acid

B

beta hydroxy fatty acid

EO

esterified omega hydroxy

N

non-hydroxy

O

omega hydroxy

Y = dS

dihydrosphingosine

H

6-hydroxysphingosine

P

phytosphingosine

S

sphingosine

SD

4,14-sphingadiene

T

dihydroxy dihydrosphingosine, CXX, number of atoms in acyl chain or fatty acid

CER NS-C34

CER NS with a total chain length of 34 carbon atoms

CerS3

ceramide synthase 3

CerS4

ceramide synthase 4

CHOL

cholesterol

ELOVL1

elongase 1

ELVOL3

elongase 3

ELOVL6

elongase 6

FA2H

fatty-acid 2 hydroxylase

FAR2

Acyl-CoA Reductase 2

FFA

free fatty acid

FFA_mix

mixture of FFA with primarily chain length of 22 carbon atoms and longer

GBA

β-glucoseribrosidase (contains different GBA enzymes)

GBA-1

glucocerebrosidase-1

²H

deuterium

HPTLC

high performance thin layer chromatography

IL

interleukine

INF

interferon

LC/MS

liquid chromatography combined with mass spectroscopy

LEKTI

lymphoepithelial kazal-type related inhibitor

LPP

long periodicity phase

muFFA

monounsaturated fatty acid

NMR

nuclear magnetic resonance

NTS

Netherton syndrome

SC

stratum corneum

SCD⁻¹

steaoryl CoA-desaturase

SLD

scattering length density profile

SPP

short periodicity phase

TEWL

trans epidermal water loss

TLC

thin layer chromatography

Th1

T-helper 1, Th2, T-helper 2

TNF-α

tumor necrosis factor α