Permeability and microstructure of model stratum corneum lipid membranes containing ceramides with long (C16) and very long (C24) acyl chains

Accepted Manuscript Permeability and microstructure of model stratum corneum lipid membranes containing ceramides with long (C16) and very long (C24) acyl chains

Petra Pullmannová, Ludmila Pavlíková, Andrej Kováčik, Michaela Sochorová, Barbora Školová, Petr Slepička, Jaroslav Maixner, Jarmila Zbytovská, Kateřina Vávrová PII: DOI: Reference:

S0301-4622(16)30463-X doi: 10.1016/j.bpc.2017.03.004 BIOCHE 5971

To appear in:

Biophysical Chemistry

Received date: Revised date: Accepted date:

12 December 2016 9 February 2017 19 March 2017

Please cite this article as: Petra Pullmannová, Ludmila Pavlíková, Andrej Kováčik, Michaela Sochorová, Barbora Školová, Petr Slepička, Jaroslav Maixner, Jarmila Zbytovská, Kateřina Vávrová , Permeability and microstructure of model stratum corneum lipid membranes containing ceramides with long (C16) and very long (C24) acyl chains. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Bioche(2017), doi: 10.1016/j.bpc.2017.03.004

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Permeability and microstructure of model stratum corneum lipid membranes containing ceramides with long (C16) and very long (C24) acyl chains

Petra Pullmannová1, Ludmila Pavlíková1, Andrej Kováčik1, Michaela Sochorová1, Barbora Školová1,2, Petr Slepička3, Jaroslav Maixner3, Jarmila Zbytovská2,3, Kateřina Vávrová1 1

PT

Skin Barrier Research Group, Charles University, Faculty of Pharmacy in Hradec

Králové, Heyrovského 1203, 500 05 Hradec Králové, Czech Republic

RI

2

Charles University, Faculty of Pharmacy in Hradec Králové, Department of

SC

pharmaceutical technology, Heyrovského 1203, 500 05 Hradec Králové, Czech Republic 3

NU

University of Chemistry and Technology Prague, Technická 5,166 28 Prague, Czech

MA

Republic

Corresponding author:

D

Petra Pullmannová, Skin Barrier Research Group, Department of Inorganic and

PT E

Organic Chemistry, Charles University, Faculty of Pharmacy in Hradec Králové, Heyrovského 1203, 500 05 Hradec Králové, Czech Republic,

AC

CE

[email protected], Tel.: +420495067348, Fax: +420495067166

-1-

ACCEPTED MANUSCRIPT Abstract The Stratum corneum (SC) prevents water loss from the body and absorption of chemicals. SC intercellular spaces contain ceramides (Cer), free fatty acids (FFA), cholesterol (Chol) and cholesteryl sulfate (CholS). Cer with “very long” acyl chains (for example, N-lignoceroyl-sphingosine, CerNS24) are important for skin barrier function, whereas increased levels of “long” acyl Cer (for example, N-palmitoylsphingosine, CerNS16) occur in patients suffering from atopic eczema or psoriasis.

PT

We studied the impact of the replacement of CerNS24 by CerNS16 on the barrier properties and microstructure of model SC lipid membranes composed of

RI

Cer/FFA/Chol/CholS. Membranes containing the long CerNS16 were significantly

SC

more permeable to water (by 38-53%), theophylline (by 50-55%) and indomethacin (by 83-120%) than those containing the very long CerNS24 (either with lignoceric

NU

acid or a mixture of long to very long chain FFA). Langmuir monolayers with CerNS24 were more condensed than with CerNS16 and atomic force microscopy showed differences in domain formation. X-ray powder diffraction revealed that

MA

CerNS24-based membranes formed one lamellar phase and separated Chol, whereas the CerNS16-based membranes formed up to three phases and Chol.

D

These results suggest that replacement of CerNS24 by CerNS16 has a direct

AC

Abbreviations

CE

PT E

negative impact on membrane structure and permeability.

Atomic force microscopy (AFM); ceramide/s (Cer); cholesterol (Chol); free fatty acid/s (FFA); mixture of free fatty acids with 16, 18, 20, 22 and 24 carbons (FFA(16-24)); Nhexadecanoyl-D-erythro-sphingosine (CerNS16); indomethacin (IND); lignoceric acid (LIG); sodium cholesteryl sulfate (CholS); relative humidity (RH); stratum corneum (SC); N-tetracosanoyl-D-erythro-sphingosine (CerNS24); theophylline (TH); X-ray powder diffraction (XRPD)

-2-

ACCEPTED MANUSCRIPT 1. Introduction The outermost layer of the skin of terrestrial mammals, including humans – the stratum corneum (SC) – protects the body from both desiccation and the entry of substances from the environment. The SC consists of cornified cells, corneocytes, embedded in an extracellular matrix of highly ordered lipids. These lipids consist of ceramides (Cer), free fatty acids (FFA) and cholesterol (Chol) in an approximately

PT

equimolar ratio and a minor fraction of cholesteryl sulfate (CholS) [1]. Skin Cer (i.e., N-acylsphingosines) are a heterogeneous class of sphingolipids that are

RI

indispensable for the epidermal homeostasis. To date, 15 classes of free Cer have been identified in the human SC; these classes are based on five different sphingoid

SC

bases and three types of N-acyl chains, including the ultra-long acyl Cer (EO-class Cer), which contain 30-34C acyl chains with a linoleic acid ester linked to an ω-

NU

hydroxyl [2,3]. In addition to the common double chain Cer, two recently identified Cer classes contain a third chain at position 1 (1-O-acyl Cer) [4].

MA

The human skin barrier Cer are characterized by having very long (20-26 carbons) or ultra-long (>28 carbons) acyl chains [2–4]1. The biosynthesis of such very long or ultra-long acyl chains requires the elongation of palmitoyl coenzyme A by a series of

AC

CE

PT E

D

reactions catalyzed by a family of elongases (ELOVL) (Fig. 1).

Fig. 1: Schematic representation of the biosynthesis of Cer with a very long acyl (CerNS24) and a long acyl (CerNS16). CoA, Coenzyme A; FAS, fatty acid synthase; ELOVL, elongation of very long chain fatty acid protein (fatty acid elongase); CerS, ceramide synthase; Des1, sphingolipid ∆4-desaturase.

1

We use the nomenclature defined in Rabionet et al., in which long chains are defined as C14–C19, very long chains are defined as C20–C26 and ultra-long chains are those over 26C.

-3-

ACCEPTED MANUSCRIPT The acyl chains are then attached to dihydrosphingosine by one of six Cer synthases, which have distinct preferences in terms of acyl structure, leading to the formation of various Cer classes (Fig. 1). The Cer synthase 2 generates ceramides with the “very long” acyl chain length of 22 – 24 carbons. Intracellular Cer are further processed to their more polar precursors, i.e., sphingomyelin and glucosylceramide,and stored in lamellar bodies. At the stratum granulosum/SC interface, these fluid-phase precursors are enzymatically processed to yield Cer and

PT

other barrier lipids. The transition from fluid to crystalline lipids is likely further aided by the corneocyte lipid envelope (Cer covalently bonded to the cell surface proteins)

RI

that is supposed to have a templating effect on the orientation of the free lipids (for a

SC

review, see Ref. [4]).

Pewzner-Jung et al. demonstrated that the lipid profile and biophysical

NU

properties of liver lipid extract were altered in Cer synthase 2 null mouse, which lost the activity of this enzyme. Cer in Cer synthase 2 null mouse were devoid of very

MA

long acyl chains (C22 – C24), but the total Cer level was unaltered due to an increased amount of “long” acyl chains Cer with 16 carbons [5]. The length of the Cer acyl chain appears to be important for numerous physiological and

D

pathophysiological properties of Cer [4–6] and also for the effects of Cer on the

PT E

lateral organization of lipid membranes [7–9]. Several chronic skin diseases, such as psoriasis and atopic eczema, are accompanied by an altered skin lipid profile [10–12]. Recent works showed that

CE

atopic dermatitis patients had decreased levels of sphingosine-based C24 Cer with “very long” acyl chains (for example N-lignoceroyl-sphingosine, CerNS24) and an

AC

increased proportion of C16 Cer with “long” acyl chains (for example N-palmitoylsphingosine, CerNS16) [11,13]. This acyl chain length shift correlated with the aberrant lipid organization and increased transepidermal water loss observed in these patients [13]. Similar “shortening” of Cer acyl chains was also found in psoriasis, cultured human keratinocytes treated with interferon gamma [14], and a murine atopic dermatitis model [15]. Thus, it appears that the biosynthesis of the very long Cer (in addition to the ultra-long acyl Cer) is an important step in skin barrier development and that Cer with a palmitoyl chain (e.g., CerNS16) cannot substitute the very long chain Cer (e.g.,

-4-

ACCEPTED MANUSCRIPT CerNS24). However, it is difficult to link this particular change in Cer chain length to higher skin permeability because of the multifactorial nature of atopic dermatitis. Additional mechanistic insight into the role of Cer chain length in the epidermis can be provided by a study of model lipid membranes with a well-defined lipid composition. One of the first investigation of the artificial lipid membranes mimicking the human SC was published by Landman in 1984 [16]. Later on, the model lipid

PT

membranes have become widely used in the SC lipid barrier research [17–19] Several recent studies investigated biophysical aspects of model lipid membranes as

RI

a function of the Cer acyl chain length or the FFA chain length and reported that chain shortening led to a less tight lipid arrangement and a phase separation [20–29].

SC

Only two studies focused directly on CerNS16 in comparison to CerNS24. The first reported that CerNS16 in the fully hydrated SC model was mostly found in the gel

NU

phase, while CerNS24 was crystalline [28].The second study showed that CerNS24 preferred an extended (splayed-chain) conformation in which the fatty acid was

MA

associated with the ceramide acyl chain. In contrast, the shorter CerNS16 and fatty acids were mostly phase separated [29].

The Cer with very short (C4-C6) acyl chains showed negative effects on the

D

permeability of porcine skin and model membranes; however, Cer with C12 and

PT E

C18:1 acyl did not significantly increase the permeability relative to very long acyl CerNS24 [26,27]. These findings cast doubt over the hypothesis that Cer with C16 acyl chains instead of very long C24 acyl chains would have a negative effect on the

CE

barrier properties of SC lipid membranes. In light of the correlation between the impaired skin barrier function and the

AC

levels of Cer with C16 acyl in diseased skin, we decided to directly compare the effects of the long chain CerNS16 relative to the very long chain CerNS24 on the permeability and microstructure of model SC lipid membranes. The model lipid membranes were composed of an equimolar mixture of the studied Cer/FFA/Chol with 5 wt% CholS. Further variation was achieved using either lignoceric acid (LIG) or a mixture of FFA with C16, C18, C20, C22 and C24 acyl chains (FFA(16-24)), which is close to the FFA profile of the native SC [30]. In fact, skin lipids are structurally heterogeneous, in the case of FFA the differences are in the chain length. Therefore we decided to introduce a partial heterogeneity and compared the model membranes with the different FFA component. The membrane permeability was evaluated using

-5-

ACCEPTED MANUSCRIPT three permeability markers: the water loss through the membranes, the steady-state flux of theophylline (TH) and the steady-state flux of indomethacin (IND). X-ray powder diffraction (XRPD) was used to reveal how the studied Cer influenced the periodical arrangement of the lipid membranes. The area per lipid in the monolayer at the air-water interface was studied by the Langmuir monolayers and the domain formation was visualized in the supported Langmuir-Blodgett monolayers by atomic

RI

PT

force microscopy (AFM).

2. Material and Methods

SC

2.1. Chemicals and material

NU

N-tetracosanoyl-D-erythro-sphingosine (CerNS24) and N-hexadecanoyl-Derythro-sphingosine (CerNS16) were purchased from Avanti Polar Lipids (Alabaster, AL, USA). Cholesterol from lanolin (Chol), sodium cholesteryl sulfate (CholS),

MA

hexadecanoic acid, octadecanoic acid, eicosanoic acid, docosanoic acid, tetracosanoic acid (LIG), theophylline, (TH), indomethacin (IND), gentamicin sulfate from Micromonospora purpurea, sodium phosphate dibasic dodecahydrate,

D

propylene glycol and solvents were purchased from Sigma-Aldrich Chemie Gmbh

PT E

(Schnelldorf, Germany). All solvents were of analytical or HPLC grade. Sodium hydroxide, sodium chloride, potassium chloride and sodium phosphate monobasic dihydrate were supplied from Lachema (Neratovice, Czech Republic). The chemicals

CE

were of analytical grade and were used without further purification. Nuclepore™ track-etched polycarbonate membranes with a 0.015 µm pore size were from

AC

Whatman (Kent, Maidstone, United Kingdom). The aqueous solutions were prepared using Millipore water. 2.2. Preparation of model lipid membranes Free fatty acids (FFA(16-24)) were mixed in a molar ratio that corresponds to the native composition of human skin FFA: 1.8% hexadecanoic acid, 4.0% octadecanoic acid, 7.6% eicosanoic acid, 47.8% docosanoic acid and 38.8% LIG [30]. The model SC lipid membranes were prepared as mixtures of CerNS24 (or CerNS16), Chol and LIG (or FFA(16-24)) at the molar ratio 1:1:1 with the addition of 5 wt%CholS. The equimolar lipid fractions used for the membrane preparation

-6-

ACCEPTED MANUSCRIPT mimicked the lipid composition of native SC [31]. The lipid mixtures were dissolved in 2:1 hexane/96% ethanol (v/v) at 4.5 mg/mL. Nuclepore polycarbonate filters were washed in 2:1 hexane/96% ethanol, dried and mounted in a steel holder with a hole of 1 cm diameter, which exposed 0.79 cm2 of the polycarbonate filter. The lipid solutions (3 × 100 µL per 1 cm 2) were sprayed on the polycarbonate filters under a stream of nitrogen at a flow rate of 10.2 μL/min onto a 10 × 10 mm square using a Linomat V (Camag, Muttenz, Switzerland)

PT

equipped for additional y-axis movement [30]. The amount of lipids used per 1 cm2 was 1.35 mg. The lipid layer thickness was measured previously and found to be

RI

approximately 11 µm [17,26]. The prepared lipid membranes were dried in vacuum

SC

over P4O10 and solid paraffin in a desiccator and they were stored in refrigerator at 26 °C. Before use, the lipid membranes were heated to 90 °C, equilibrated at this

NU

temperature for 10 min and slowly (~3 h) cooled to 32 °C. The model membranes undergo the main phase transition at approximately 55 – 70 °C. The main phase transition converts the lipid mixture to melted-chain phase with predominant gauche

MA

conformation of hydrophobic chains. The transition is accompanied by a decrease in the rigidity and packing efficiency of the hydrophobic chains. Thus, at 90 °C the

D

hydrophobic chains are in a disordered state [29]. Afterward, the membranes were

PT E

equilibrated at 32 °C for at least 24 h.

2.3. Permeation Experiments

CE

The permeability of the model lipid membranes was evaluated using Franztype diffusion cells with an available diffusion area of 0.5 cm2 and an acceptor

AC

compartment with a volume of approximately 6.5 mL. The membranes equilibrated for at least 24 h at 32 °C were mounted into the Franz-type diffusion cells with the lipid film facing the donor compartment using Teflon holders. The acceptor compartment of the cell was filled with phosphate buffered saline (PBS) solution at pH 7.4 containing 50 mg/L gentamicin. The acceptor phase was stirred by a magnetic bar in a thermostated bath at 32 °C; the precise volume was measured for each cell individually and was included in the calculation. After a 12-h equilibration of the completed cells at 32 °C, the water loss (see below) was measured. On the next day, 100 µL of the first donor sample – either 5% theophylline (TH) or 2% indomethacin (IND) suspensions in 60% propylene glycol – -7-

ACCEPTED MANUSCRIPT was applied to the membrane. This setup ensured sink conditions for the selected drugs. Samples of the acceptor phase (300 µL) were withdrawn every 2 h over 10 h and were replaced by the same volume of PBS. During this period, a steady-state situation was reached. After the first 10 h-long permeation experiment, the applied donor suspension was carefully washed with PBS at pH 7.4, and the residual PBS was removed using

PT

cellulose swabs. The acceptor phase was removed and replaced with fresh PBS at pH 7.4. The membranes were allowed to equilibrate for 12 h.

RI

Afterward, 100 µL of the second donor sample (TH or IND) was applied. The permeation experiment was repeated as described above. To verify that the first

SC

experiment did not alter the outcome of the second experiment, we changed the experiment order in some membranes. The permeation profiles did not differ. Thus,

NU

this setup was used to obtain additional data from each membrane. For further validation and comments, see our previous work [26].

MA

2.4. Measurements of the water loss through the membranes The water loss through the model membranes was measured using equipment

D

for transepidermal water loss measurement [mg/cm2/h]: the Tewameter® TM 300

PT E

probe and Multi Probe Adapter Cutometer® MPA 580 (CK electronic GmbH, Köln, Germany) with a hollow cylinder (height of 2 cm, diameter of 1 cm). To measure the water loss through the model membranes, the upper part of the Franz-type diffusion cell was removed, and the probe was placed on the Teflon membrane holder

CE

containing a cylindrical hole with area of 0.5 cm2 (height of 0.6, diameter of 0.8 cm). The measuring time was usually between 80 and 100 s, and the average steady-

AC

state value was recorded. The environmental conditions were comparable during all measurements: ambient air temperature of 25-26 °C and relative air humidity of 29 46%. Because the usage of the membrane holder affects the measured values, we performed calibration measurements using an empty Nuclepore filter with and without the Teflon membrane holder. We measured the water loss over the various water/propylene glycol mixtures and plotted the values obtained with the holder against the values measured without the holder for the same conditions (see Fig. S1, supplementary material). The function obtained by a linear regression (R2 = 0.9301) served as a calibration line to correct the water loss values measured on the lipid membranes mounted in the Teflon cell holders in the Franz diffusion cells. The -8-

ACCEPTED MANUSCRIPT measurements were taken for each cell before the application of the donor samples. For details on the water loss measurements, see our previous work [32]. 2.5. High Performance Liquid Chromatography (HPLC) of Model Permeants The TH- and IND-containing samples were measured using isocratic reversephase HPLC using a Shimadzu Prominence instrument (Shimadzu, Kyoto, Japan) consisting of LC-20AD pumps with a DGU-20A3 degasser, SIL-20A HT autosampler,

PT

CTO-20AC column oven, SPD-M20A diode array detector, and CBM-20A communication module. Data were analyzed using the LCsolutions 1.22 software.

RI

The reverse-phase separation of TH was achieved on a LiChroCART 250-4 column (LiChrospher 100 RP-18, 5 µm, Merck, Darmstadt, Germany) at 35 °C using 4:6

SC

methanol/0.1 M NaH2PO4 (v/v) as a mobile phase at a flow rate of 1.2 mL/min. The acceptor phase sample (20 μL) was injected into the column, and the effluent UV

NU

absorbance was measured at 272 nm, with a bandwidth of 4 nm. The retention time of TH was 3.2±0.1 min. The IND samples were assayed on a LiChroCART 250-4

MA

column (LiChrospher 100 RP-18, 5 µm, Merck, Darmstadt, Germany) using a mobile phase containing 90:60:5 acetonitrile/water/acetic acid (v/v/v) at a flow rate of 2 mL/min. Next, 100 μL of the acceptor phase sample was injected into the column,

D

which was maintained at 40 °C. The UV absorbance was monitored at a wavelength

PT E

of 260 nm, the bandwidth was 4 nm, and the retention time of IND was 3.1±0.1 min. Both methods were previously validated according to EMA European Medicines Agency, Guideline on bioanalytical method validation (2011). [33].

CE

2.6. Permeation data evaluation

Using the measured concentrations of TH and IND and the volume of the

AC

Franz-type diffusion cell, the cumulative amounts of TH and IND that penetrated across the lipid membrane were calculated (Fig. 2A-C). The cumulative amounts were corrected for the acceptor phase replacement. For example, the cumulative amount at the time t = 6 h was calculated according to the following equation: m 6 h = (c6 h×V + c4 h×0.3 + c2 h×0.3)/A, where mi h is the cumulative amount [µg/cm2] at the time i h, ci h is the concentration [µg/ml] at the time i h, V is the acceptor phase volume [ml], and A is the diffusion area [cm2]. The cumulative amounts were plotted against time, and the steady-state flux of TH or IND [µg/cm2/h] was calculated as the slope of the linear regression function obtained by fitting the linear region of the plot in Microsoft Excel (Fig. 2A). The data are presented as the means±SEM, and the -9-

ACCEPTED MANUSCRIPT number of replicates is given in the pertinent figure. A parametric unpaired t-test with Welch's correction was used for the statistical analysis. We chose the t-test because we compared pertinent two populations of data separately. Parametric t-test assumes a Gaussian distribution of data. Unpaired t-test with Welch’s correction does not assume equal standard deviations of two data groups. 2.7. X-ray powder diffraction

PT

The lipid mixtures for the X-ray powder diffraction (XRPD) measurements were prepared in the same manner as those for permeation experiment, but the lipids were

RI

sprayed onto a 22 × 22 mm supporting cover glass instead of the supporting filters. Prior to the measurement, the samples were heated to 90 °C, equilibrated at this

SC

temperature for 10 min and allowed to cool down for approximately 3 h. Afterwards, the samples were equilibrated at 32 °C for at least 24 h The XRPD data were

NU

collected at ambient temperature with an X'Pert PRO θ-θ powder diffractometer (PANalytical B.V., Almelo, Netherlands) with parafocusing Bragg-Brentano geometry

MA

using CuKα radiation (λ = 1.5418 Å, U = 40 kV, I = 30 mA) in modified sample holders over the angular range of 0.6-30° (2θ). Data were scanned with an ultrafast positionsensitive linear (1D) X'Celerator detector with a step size of 0.0167° (2θ) and a

D

counting time of 20.32 s step-1. The data were evaluated using X'Pert Data Viewer

PT E

software (PANalytical B.V., Almelo, Netherlands) and Jandel Scientific Peakfit 2.01 software (AISN Software). The peaks in the area of short-range arrangement were fitted by Lorentzian function above a linear background. The XRPD diffractograms

CE

show the scattered intensity as a function of the scattering vector Q [nm -1], which is proportional to the scattering angle 2θ according to the equation: Q = 4π sinθ/λ (λ =

AC

0.15418 nm is a wavelength of the X-rays). The repeat distance d [nm] characterizes the spacings between adjacent scattering planes in real space for a lamellar lipid phase (L) of parallel lipid layers arranged in a one-dimensional lattice. The diffractograms of the lamellar phases exhibit a set of Bragg reflections whose reciprocal spacings are in the characteristic ratios of Qh = 2πh/d (Miller index h = 1, 2, 3…). The repeat distance d was obtained from the slope a of a linear regression of the dependence Qh = a.h, according to the equation d = 2π/a. 2.8. Langmuir Monolayers Next, we studied the lipid monolayers at the air−water interface to compare their area per molecule and compressibility. Surface pressure−area isotherms were - 10 -

ACCEPTED MANUSCRIPT measured using a small Langmuir−Blodgett trough (KSV-NIMA, Espoo, Finland). Lipids (1 mg/mL, 10 μL) were mixed in chloroform/methanol 3:1 and spread at the air−water interface, and the solvents were allowed to evaporate for 20 min. The lipid film was compressed at 20 mm/min at 23 °C, and the surface pressure recorded using a platinum plate at least in triplicate. The molecular areas per lipid Al were calculated at 1.5 mN/m (onset) and 20 mN/m. Maximum surface compression moduli were calculated according to Cs−1 = −Al(∂π/∂A), where Al is the area per lipid and π is

PT

the surface pressure.

RI

2.9. Atomic Force Microscopy (AFM)

To visualize domains in the lipid monolayers, we used AFM. The lipid

SC

monolayer at the air−water interface was compressed to 20 mN/m surface pressure and then transferred onto freshly cleaved mica (12 × 15 mm 2) (SPI Supplies, West

NU

Chester, Pennsylvania, USA) by raising the mica support vertically through the air−water interface at 2 mm/min. The surface morphology of the samples was

MA

examined by the AFM technique using a VEECO CP II device (Bruker Corp., Billerica, Massachusetts, USA) in tapping mode with an RTESPA-CP Si probe (Bruker Corp., Billerica, Massachusetts, USA) with a spring constant of 20-80 N m-1.

D

The domain areas and surface roughness were determined using Veeco DI SPMLab

PT E

NT 6.0.2 software (Bruker Corp., Billerica, Massachusetts, USA).

CE

3. Results and discussion

3.1. Effects of ceramide acyl chain length and fatty acid heterogeneity on the

AC

permeability of the model membranes To test the hypothesis that the shortening of the acyl chain in CerNS24 by 8 carbons to CerNS16 will result in increased permeability, we prepared model lipid membranes composed of an equimolar mixture of the CerNS24 or CerNS16/FFA/Chol with 5 wt% CholS. Further variation was achieved using either lignoceric acid (LIG) or a heterogeneous mixture of FFA(16-24), which more closely models the heterogeneous FFA profile of the native SC [30]. For the construction of these membranes, we selected CerNS because changes in the acyl chain length in this particular Cer class were reported in all the above mentioned studies [11,13–15].

- 11 -

ACCEPTED MANUSCRIPT Water loss [mg/cm2/h] is a common in vivo measure of the SC’s capability of preventing water loss from the organism and is increased in skin diseases [13]. This parameter was also used in vitro on both skin and model membranes [13,14]. The water loss (Fig. 2D) is a measure of the water barrier in the inside-to-outside direction. Thus, to also probe the outside-to-inside barrier to exogenous chemicals, the steady-state flux (henceforth referred to as flux) values of two model compounds (TH and IND) were measured (Fig. 2E, F). TH and IND were selected because of

PT

their different physicochemical properties: TH, a small molecule with balanced lipophilicity (M = 180.2 g/mol, log P ~ 0), and IND (M = 357.8 g/mol, log P ~ 4.3),

RI

which is larger and more hydrophobic than TH. Model permeants from the same

SC

groups and with similar log P values (caffeine and diclofenac sodium) were used previously to characterize model SC membranes [34]. TH and IND most likely have

NU

different skin penetration pathways [35]. Both TH and IND were employed as skin permeability markers previously: the permeability to TH was 2 and 2.5 times higher in non-lesional and lesional atopic skin, respectively, than in healthy controls [36]; the

MA

skin absorption of IND was found to be susceptible to the skin condition [37]. The membranes containing very long Cer and LIG (CerNS24/LIG/Chol/CholS)

D

attained the lowest permeability to all the measured permeability markers (water loss:

PT E

0.13±0.01 mg/cm2/h, flux of TH: 0.09±0.02 µg/cm2/h, flux of IND: 0.06±0.01

AC

CE

µg/cm2/h).

- 12 -

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Fig. 2: Permeation profile of a single membrane (A) and of all the membranes (n = 8-

CE

20) (B, C): the cumulative amount [µg/cm2] of theophylline (TH) or indomethacin (IND) that penetrated across the lipid membrane as a function of time [h]. Effects of

AC

Cer acyl chain length (CerNS24 vs. CerNS16) and FFA heterogeneity (LIG vs. FFA(16-24)) on the permeability of the Cer/FFA/Chol/CholS membranes: Water loss (D); Steady-state flux of TH (E); Steady-state flux of IND (F); n = 8-20. Data are presented as the means±standard error of mean (SEM); n is the number of pertinent samples. Statistical significance against control: * p < 0.05; ** p < 0.01.

When the shorter CerNS16 was present instead of CerNS24, the membrane CerNS16/LIG/Chol/CholS showed 38% higher permeability to water (0.18±0.02 mg/cm2/h), 55% higher permeability to TH (0.14±0.02 µg/cm 2/h) and almost twice

- 13 -

ACCEPTED MANUSCRIPT higher permeability to IND (0.11±0.02 µg/cm2/h) relative to the CerNS24/LIG/Chol/CholS membrane. When we used CerNS24 and replaced LIG by a heterogeneous FFA(16-24) mixture (CerNS24/FFA(16-24)/Chol/CholS), the water loss increased by 46% (0.19±0.02 mg/cm2/h), the flux of TH by 100% (0.18±0.02 µg/cm2/h) and the flux of IND by 66% (0.10±0.01 µg/cm2/h) compared with those for CerNS24/LIG/Chol/CholS.

PT

The weakest barrier was created by the membrane that contained CerNS16 with a shorter acyl chain and the heterogeneous FFA mixture CerNS16/FFA(16-

RI

24)/Chol/CholS (water loss: 0.29±0.02 mg/cm2/h, flux of TH: 0.27±0.04 µg/cm2/h, flux

SC

of IND: 0.22±0.03 µg/cm2/h).

Both the shortening of the Cer acyl chain and the increase in FFA chain

NU

heterogeneity raised the model membrane permeability to water and model permeants. The magnitude of this effect of CerNS16 and/or FFA(16-24) on water loss are comparable to the effect of the partial lipid extraction of human skin by organic

MA

solvents (55% higher water loss) or sodium lauryl sulfate treatment (57% higher water loss compared with untreated skin) [38]. Although the water loss in the studied model membranes cannot be directly compared to the values reported in the skin

D

disease patients due to many other contributing factors in vivo, our results confirmed

PT E

that an inadequate Cer acyl chain length has an effect comparable to the effect from the standard methods of experimental skin barrier impairment.

CE

Interestingly, this effect of CerNS16 on the membrane permeability to IND was nearly twice greater than its effect on the membrane permeability to the smaller and less lipophilic TH or water. In contrast, the effects of the heterogeneous FFA(16-24)

AC

mixture on the membrane permeability to both model drugs, IND and TH, were similar (up to twice higher permeability of the FFA(16-24)-based membranes compared with the corresponding LIG membranes). The results are consistent with the expectation based on the abovementioned in vivo findings that the permeability of model membranes should rise with the shortening of the Cer acyl chain [11,13–15]. However, these findings are in contrast with the trend observed using the model membranes prepared as equimolar mixtures of CerNS with a short acyl chain of 2, 4, 6, 8, 12 C or oleoyl (18:1) chain, Chol, and LIG with the addition of 5 wt % CholS, in which membranes with CerNS12 and CerNS18:1 showed similar permeability to

- 14 -

ACCEPTED MANUSCRIPT those with CerNS24 [26]. The effects of CerNS16 on permeability do not fit into this trend, which may be consistent with the chain symmetry of this lipid. The heterogeneous mixture of FFA(16-24) mimicked more closely the composition of the SC FFA fraction than single LIG and excluded the possibility that the observed increased permeability upon Cer acyl shortening is LIG-dependent. The higher permeability of the membranes (either with CerNS16 or CerNS24) induced by

PT

FFA(16-24) is consistent with the results found by Lee et al. They reported that the water permeability of the membranes containing mixed FFA (either a mixture with the

RI

most abundant very long FFA or with long FFA) was higher than that of the membranes containing a single FFA with the chain length ranging from 14 to 24

SC

carbons [24].

NU

3.2. Langmuir monolayers

The representative pressure-area isotherms of the CerNS24/LIG/Chol/CholS

MA

and CerNS16/LIG/Chol/CholS lipid mixtures are presented in Fig. 3. At the onset (1.5 mN/m), we found the area per lipid Al = 0.42 ± 0.01 nm2 for the CerNS16/LIG/Chol/CholS mixture, which was significantly larger than that of the

D

CerNS24/LIG/Chol/CholS system (Al = 0.36 ± 0.02 nm2), suggesting more tightly

PT E

packed lipid chains in the latter. Similar differences were observed over the entire range of surface pressures. The shape of the isotherm indicated that the system with the CerNS16 was less condensed (the maximum compression modulus was 318

CE

mN/m) than that with the CerNS24 (the maximum compression modulus was 508 mN/m) and underwent a phase transition at approximately 20 mN/m. A similar phase

AC

transition was detected in the ternary system LIG/Chol/CholS and was identified as a liquid condensed−solid transition [26]. The area per lipid in the CerNS16/LIG/Chol/CholS corresponded to the values obtained for the lipid mixture with Cer composed of extremely short acyl chain (CerNS4 and CerNS6) and exceeded the area per lipid in CerNS12-based mixtures in our previous experiment [26].

- 15 -

SC

RI

PT

ACCEPTED MANUSCRIPT

3.3. Atomic force microscopy (AFM)

NU

Fig. 3: Pressure−area isotherms of lipid monolayers at the air−water interface.

MA

The monolayers of CerNS24/LIG/Chol/CholS and CerNS16/LIG/Chol/CholS at the air-water interface were transferred onto a mica support at the surface pressure of 20 mN/m, which was previously used to characterize SC model systems [39], and

D

studied by AFM. The AFM topographical images of the supported monolayers (Fig. 4)

PT E

showed separated domains in both mixtures. Light areas represent high domains. The high domains covered approximately 52% of the surface in CerNS24/LIG/Chol/CholS and only 32% in CerNS16/LIG/Chol/CholS. An analogous

CE

reduction in the high domains in supported monolayers was reported for Cer/LIG/Chol/CholS mixtures with short acyl chains and was suggested to correlate

AC

with the higher permeability of these mixtures [26]. The AFM of the CerNS24 mixture is typical of a membrane with lipid domains, but that of CerNS16 seems more that of a conventional phase separation where the line tension is not competing with aggregation (because of the domain shape and the existence of small fluid pools inside the rigid regions). The estimated height difference of 1.5 ± 0.3 nm between the high and low domains in CerNS16/LIG/Chol/CholS corresponds to the predicted difference between the LIG chain in the all-trans conformation (3.1 nm) and Chol (1.6 nm). The high domains most likely formed by the LIG or LIG-rich phase, which is characteristic for well-ordered very long hydrophobic chains. The low domains contained the major - 16 -

ACCEPTED MANUSCRIPT fraction of Chol and likely also CerNS16, which is consistent with the larger area of the lower domain compared to the higher domain. In the CerNS24/LIG/Chol/CholS mixture, the difference in thickness between the high and low domains was 1.2 ± 0.2 nm and is comparable to previously reported values [26]. The high domains mostly contained very long hydrophobic chains of CerNS24 and LIG, and the low domains were a Chol-rich phase likely containing a small amount of CerNS24. This inclusion of CerNS24 in the higher domain is consistent with the 52% area of this domain and

PT

the slightly lower height difference compared to the CerNS16 membrane.

RI

Thus, the CerNS16 and LIG-containing lipid mixture formed a smaller fraction of high domains and showed a larger area per lipid molecule in monolayers than the

SC

CerNS24 and LIG-based lipid mixture. These results are consistent with the increased permeability of the CerNS16 and LIG-based membranes relative to that of

AC

CE

PT E

D

MA

NU

the CerNS24 and LIG-based membranes.

Fig. 4: AFM images of the lipid monolayers composed of Cer/LIG/Chol/CholS.

- 17 -

ACCEPTED MANUSCRIPT 3.4. Long-range arrangement of the model membranes revealed by X-ray powder diffraction (XRPD) 3.4.1. Membranes with the very long chain CerNS24 To shed more light on the permeability of model lipid membranes, we studied the biophysical behavior of the lipid membranes. XRPD patterns of the model membranes kept at room air humidity are summarized in Fig. 5. We measured the

PT

membranes also in hydrated state (after the equilibration of the membrane at 100% humidity for 24 h) and found no changes of the lamellar structure relative to the

RI

membranes measured at ambient humidity (not shown). The patterns contain a region of long-range arrangement at the scattering vector Q ~ 0.5 – 9 nm-1 and a

SC

region of short-range arrangement at Q ~ 14 – 18 nm-1. In the region of long-range arrangement, the diffractogram of the CerNS24/LIG/Chol/CholS membrane (Fig 5,

NU

black line) showed a set of peaks (see Tab. 1), which are marked by roman numerals. This set of peaks was ascribed to the lamellar phase L1 created by the

MA

parallel stacked lipid layers. Its repeat distance d = 5.39 nm was determined from the slope of the dependence of the position of the peak maximum in Q (nm -1) vs. Miller index h; this relationship passed through the origin (zero).

D

The additional two peaks of the CerNS24/LIG/Chol/CholS membrane marked

PT E

by asterisks gave a repeat distance d ~ 3.4 nm and were assigned to the separated Chol or Chol/CholS mixture, further referred to as separated Chol. Chol, as well as Chol monohydrate, forms a triclinic lattice with a bilayer arrangement of Chol

CE

molecules and a lattice parameter b of ~ 3.42 – 3.44 nm [40,41]. The reason for the Chol separation may be that lipid mixtures containing a high relative percentage of

AC

Chol are prone to lipid demixing during organic solvent removal [42]. However, separated Chol was also detected in other model SC lipid membranes [26,43–45] and in the human SC [46]. Chol molecules are fully incorporated in the lamella only up to a 1:1:0.5 molar ratio of Cer/FFA/Chol, and they phase separate at higher Chol proportions (E.H. Mojumdar, G.S. Gooris, J.A. Bouwstra, unpublished data, cited in [47]). Thus, it appears that the Chol separation is driven by its high concentration in the studied membranes. When we replaced a single LIG with a heterogeneous FFA(16-24) mixture, the diffractogram of the CerNS24/FFA(16-24)/Chol/CholS membrane (Fig. 5, blue line) showed a structure similar to that of the CerNS24/LIG/Chol/CholS membrane. We

- 18 -

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

MA

Fig. 5: XRPD diffractograms of Cer/FFA/Chol/CholS membranes in the region of long- (left) and short-range arrangement (right). Roman numerals indicate the L1 phase; Arabic numerals indicate the L2 phase; letters indicate the L3 phase; and

D

asterisks indicate the crystalline cholesterol (Chol). The inset shows the scaled

PT E

intensity of the XRPD pattern of the pertinent model membrane. The intensity is

CE

shown in arbitrary units.

identified in addition to the separated Chol peaks a set of peaks that belonged to the

AC

parallel stacked lipid lamellae of the L1 phase. This lamellar phase L1 had a repeat distance d = 5.33 nm, what is comparable to the d = 5.39 nm value for CerNS24/LIG/Chol/CholS. 3.4.2. Membranes with the long chain CerNS16 The structure of the CerNS16/LIG/Chol/CholS membranes (Fig. 5, green line) differed substantially from the structure of CerNS24/LIG/Chol/CholS membranes. The dominant L1 lamellar phase (Tab. 1) with a repeat distance of d = 4.07 nm was significantly shorter than the repeat distance of the membranes with CerNS24.

- 19 -

ACCEPTED MANUSCRIPT Table 1: Lamellar phases and crystalline cholesterol (Chol); the corresponding peaks’ maxima positions in Q [nm-1] and their order are shown in brackets.

Q [nm-1] Lamellar phase L1

Lamellar phase L2

Lamellar phase L3

Chol

1.16 (1), 2.33 (2),

1.84,

LIG/

3.50 (3), 4.66 (4),

3.68

3.55 (3), 4.72 (4),

Chol/CholS

7.07 (6), 8.25 (7)

CerNS16/

1.45 (1), 3.10 (2),

LIG/

4.64 (3), 6.19 (4), 7.72 (5)

Chol/CholS CerNS16/

1.56 (1), 3.11 (2),

D

4.61 (3), 7.67 (5)

SC

FFA(16-24)/

NU

1.19 (1), 2.37 (2),

1.70 (1), 5.04 (3),

1.24 (1), 2.43 (2),

1.87,

8.41 (5)

4.82 (4), 6.01 (5)

3.62

1.68 (1), 3.30 (2),

1.86,

5.04 (3), 8.48 (5)

3.71

PT E

Chol/CholS

1.85, 3.68

MA

CerNS24/

RI

6.98 (6)

Chol/CholS

FFA(16-24)/

PT

CerNS24/

This shortening appears to be consistent with previous studies that found that the

CE

repeat distances of the lamellar phases found in CerNS/LIG/Chol/CholS model membranes prepared with a series of short acyl CerNS (from 2 to 18 C) strongly

AC

depended on the Cer acyl chain length [26]. In the CerNS/LIG/Chol/CholS membranes containing CerNS with 12 C or oleoyl (18:1) chains, the repeat distances were 4.02 nm and 4.24 nm, respectively. Thus, the d of this CerNS16-containing lipid mixture fits well into the d vs. Cer acyl chain length dependence. It appears that the Cer acyl chain length dictates the periodicity of the main lamellar phase: the shorter the Cer acyl chain, the thinner the lamellae it forms. This finding suggests that L1 is a Cer-rich phase. Most likely, L1 is not a pure CerNS16 phase because the repeat distances of pure CerNS16 phase and our L1 phase slightly differ: the repeat distances of pure anhydrous CerNS16 at 26 °C and pure CerNS16 in excess water at 22 °C and at 26 °C were 4.21 nm [48], 4.46 nm [49] and 4.69 nm [48], respectively. - 20 -

ACCEPTED MANUSCRIPT Beside separated Chol peaks, a minor L2 lamellar phase was detected in the membrane and showed a repeat distance of d = 3.74 nm. Similar CerNS16 behavior was reported previously by Souza et al. [49], who added 5-55 molar % Chol to CerNS16 in excess water. Upon the addition of Chol, a new peak appeared at the spacing d = 3.5 nm, and this peak coexisted with the pure or almost pure CerNS16 phase. Another report described a phase with an even shorter repeat distance of 3.44 nm in a mixture of CerNP/Chol; this phase was also ascribed to a Chol-rich

PT

phase. However, this phase disappeared upon the addition of an FFA(16-26) mixture [43]. Because the above-mentioned publications confirmed that CerNS16 and Chol

RI

mix well, the L2 phase in our samples can most likely be attributed to a

SC

Chol/CerNS16-rich domain.

In addition, a third lamellar phase (L3) that gave a repeat distance d = 5.21 nm

NU

was detected. A lamellar phase with a similar repeat distance to L3 (together with separated Chol) was found in the sample that contained Chol/LIG/CholS without any

MA

Cer [26]. Thus, we suppose that the observed L3 phase corresponds to the separated and regularly arranged LIG or LIG-rich phase. The phase separation of LIG observed in the CerNS16/LIG/Chol/CholS membrane could originate from the mismatch

D

between the acyl chain length of LIG and CerNS16. A similar separation of LIG was

PT E

reported for the whole series of short acyl Cer [26]. Furthermore, long chain FFA formed a separate FFA-rich phase when mixed with CerAP18, Chol and CholS [50] and also when mixed with bovine brain Cer type III [51]. The assumption that

CE

CerNS16 did not mix completely with LIG is supported also by the results from infrared studies [29]. Contrary to CerNS16, CerNS24 mixed well with the

AC

hydrophobically matched LIG in the model lipid membranes. From this point of view, the hydrophobic matching of the lipid chain lengths seems to be decisive for their proper miscibility. However, the hydrophobically matched CerNS16 and palmitic acid in a ternary mixture with Chol also spontaneously separate into crystalline Cer-enriched and palmitic acid-enriched domains [52]. Thus, there must be other factors that influence the lipid miscibility/separation in addition to the matching of chain lengths. Some authors emphasized the importance of lipid chain length heterogeneity for their miscibility in binary or ternary mixtures, for instance, for FFA miscibility with Cer [43]. Brief et al. [52] suggested that the acyl chain heterogeneity of Cer is a more important factor

- 21 -

ACCEPTED MANUSCRIPT than hydrophobic chain matching for enhancing lipid miscibility. Oguri et al. investigated the effect of FFA chain length and chain length heterogeneity on the mixing properties and thermotropic behavior of SC model membranes. The use of palmitic acid (C16) in the CerNS24/FFA/Chol membrane caused a phase separation between FFA and Cer. When mixed FFA (or LIG) was used, FFA and Cer participated in the same crystals [21]. Indeed, in the sample with mixed FFA and CerNS16, i.e. the

PT

CerNS16/FFA(16-24)/Chol/CholS membrane (Fig. 5, red line), only two lamellar phases L1 with d = 4.09 nm and L2 with d = 3.75 nm ) and separated Chol were

RI

identified (Tab. 1). However, we also have to consider a key feature of XRPD: the

SC

ability to detect only periodically organized structures. If separated FFA-rich domains do not form a periodically stacked long-range arrangement, they are not detectable

NU

by XRPD. In fact, we previously detected an incomplete mixing of deuterated FFA(16-24) and CerNS16 by infrared spectroscopy [29]. A similar observation was reported by Mojumdar et al. [23]. They showed using infrared spectroscopy that a

MA

fraction of deuterated FFA in an isolated pig Cer/FFA/Chol mixture (molar ratio of 1:2:1) phase separates. However, no additional lamellar phase of separated FFA was

D

detected by X-ray diffraction. These phase separated domains of FFA detectable by infrared spectroscopy were most likely only laterally segregated within the same lipid

PT E

lamellae.

The presence of several lamellar phases in the diffractograms indicated that

CE

some lipid species were separated and formed an organized structure. Thus, the CerNS16-based membranes either with LIG or with FFA(16-24) were not able to create one homogeneous phase oversaturated with Chol; instead, we observed

AC

phase-separated fractions, which created organized multiple arrangements in the XRPD patterns.

3.5. Short-range arrangement of the model membranes revealed by XRPD In the region of short-range arrangement of all the studied membranes, two peaks at Q ~ 15.2 and 16.8 nm-1 were found, corresponding to distances between diffracting planes ~ 0.37 nm and ~ 0.41 nm, respectively. Previous works reported that two peaks at these positions are typical for the lateral orthorhombic arrangement of the long saturated polymethylene chains in the SC [53]. The observed peaks at ~ - 22 -

ACCEPTED MANUSCRIPT 0.41 nm (with higher intensity) and ~ 0.37 nm were assigned to Miller indices and

, respectively [54]. The X-ray diffraction pattern with typical

two peaks corresponding to the orthorhombic arrangement is well known, however, we decided to look at the peaks’ positions in detail, because they do not occur at the same Q. We wondered, if the variances in their position are random, or if they can indicate something meaningful. According to Marsh, the lateral chain packing can be treated as a centered rectangular 2-D lattice and one can calculate its parameters a,

PT

b according to the equation (1) [55]:

RI

(1)

SC

where shk is the spacing between adjacent scattering planes in real space [nm]; a, b are the lattice parameters of the centered rectangular 2-D lattice, and h and k are

NU

Miller indices. The area per lipid chain arranged on a centered rectangular lattice Ach can be calculated according to the equation:

MA

(2)

The calculated values of the parameters a, b and Ach are summarized in Tab.

D

2. The lattice parameters and area per lipid chain Ach of CerNS24/FFA(16-

PT E

24)/Chol/CholS (Tab. 2) are very similar to those of the CerNS24/LIG/Chol/CholS membranes. The area per lipid chain organized in the orthorhombic lattice in CerNS24/LIG or FFA(16-24)/Chol/CholS (Ach ~ 0.186 nm2) resembled an area per

CE

lipid chain reported for a pure skin Cer analog: crystallographic data on N-(2D,3Ddihydroxyoctadecanoyl)-phytosphingosine showed that it organized a single layer arrangement with extended hydrocarbon chains and became packed in an

AC

orthorhombic subcell with Ach = 0.186 nm2 [56]. N-tetracosanoyl-phytosphingosine showed an area per lipid chain Ach = 0.188 nm2 [57]. Thus, the hydrophobic chains of CerNS24-based model membranes packed very densely in the orthorhombic lattice, similar to the packing of pure Cer. The tight orthorhombic arrangement of the hydrophobic chains most likely resulted exclusively from mixed Cer acyl chains and FFA. Therefore, the model of asymmetrically organized SC membrane can be applied here as was proposed by several authors [58–60]. In this model, the very long

- 23 -

ACCEPTED MANUSCRIPT Table 2: The lattice parameters of the orthorhombic lateral chain packing a and b [nm] and the area per lipid chain Ach [nm2]; n is the number of pertinent samples. The data are presented as the means±SEM.

a [nm]

Ach [nm2] n

0.748±0.001 0.498±0.001 0.186±0.001 6

PT

CerNS24/LIG/Chol/CholS

b [nm]

CerNS24/FFA(16-24)/Chol/CholS 0.748±0.001 0.498±0.001 0.186±0.001 6 0.735±0.001 0.495±0.001 0.182±0.001 3

RI

CerNS16/LIG/Chol/CholS

SC

CerNS16/FFA(16-24)/Chol/CholS 0.743±0.001 0.495±0.001 0.184±0.001 3

NU

Cer prefers an extended conformation, and its chains are distributed asymmetrically such that the sphingosine chain associates with Chol and the acyl chain associates

MA

with FFA [60]. This arrangement was also suggested in our infrared study with deuterated CerNS24, both with LIG and FFA(16-24) [29]. Surprisingly, the membranes based on CerNS16 showed an even tighter

D

packing of the hydrophobic chains relative to that of the CerNS24-based membranes.

PT E

However, the presence of various domains in the CerNS16 membranes indicate that the tight orthorhombic arrangement could be present in only one domain and that not all of the membrane components were involved in this domain (and contributed to

CE

this orthorhombic packing). In fact, the area per lipid chain in the orthorhombic lattice of CerNS16/LIG/Chol/CholS membranes corresponded to the value of Ach of the

AC

orthorhombic chain lattice reported for the stable polymorphic C-form of noctadecanoic acid [61] or for 1-octanol phase-separated in phospholipid bilayer [62]. It is well known that various crystal forms of long and very long chain FFA organize their hydrophobic chains in an orthorhombic lattice [63–65]; thus, we assume that the observed orthorhombic packing in the CerNS16-based membranes resulted from the FFA-rich domain (either FFA(16-24) or LIG). As can be deduced from the XRPD data, both membranes with CerNS24 formed one well-ordered lamellar lipid phase that was oversaturated with Chol and became organized with orthorhombic chain packing, in agreement with our previous studies [26,32]. All the membrane components except the excess Chol were mixed, - 24 -

ACCEPTED MANUSCRIPT as proven using infrared spectroscopy [29]. The membranes based on CerNS16 also showed a tight orthorhombic packing of the hydrophobic chains. However, the presence of various domains in the CerNS16 membranes indicate that the tight orthorhombic arrangement could be present in only one domain and that not all of the membrane components were involved in this domain (and contributed to this orthorhombic packing). The driving force for the miscibility of Cer with other skin lipids can be the formation of the asymmetric arrangement of the SC lipid matrix [58–60].

PT

This unique arrangement would explain not only the miscibility of the very long acyl chain Cer with other skin barrier lipids but also the low permeability of such

SC

RI

membranes.

4. Discussion of the lipid models and the techniques used

NU

Several technics were employed with the aim to reveal the permeability and physico-chemical behavior of the complex lipid mixtures CerNS/FFA/Chol/CholS. The

MA

composition of the lipid mixtures mimicked the content and ratio of the SC lipid species, i.e. Cer, FFA, Chol and CholS, but not their structural heterogeneity. A partial heterogeneity was introduced in the lipid mixture by using FFA(16-24) instead

D

of LIG in some samples. The model lipid membranes used in the permeation

PT E

experiment and in XRPD measurements had a multilamellar character similar to the SC lipid barrier, but they lacked the specific long periodicity (~13 nm) arrangement of the SC lipids. The Langmuir monolayers were even more simplified compared to the

CE

structure of SC, but were used in this work to show some aspects of intermolecular interactions between the lipid species.

AC

The experimental conditions, i.e., temperature, hydration, pH and equilibration time, were controlled in the permeation experiment so that they approximated the in vivo situation as far as possible. The time of equilibration (the time between the sample annealing procedure and experiment) was at least 24 h. In the Franz-type diffusion cells kept at 32 °C, the lipid layer faced the unsealed empty donor compartment, while the porous filters faced the phosphate buffered saline (PBS) in the acceptor compartment. This setup enabled the water diffusion through the lipid layer in the direction of water concentration gradient. Thus, lipids were allowed to hydrate in the manner similar to physiological conditions of stratum corneum (SC). Björklund et al. showed that a water gradient across the barrier (in our case across

- 25 -

ACCEPTED MANUSCRIPT the model membranes) strongly affects the rate of drug permeation [66]. In the experiment, we created the water gradient between PBS in the acceptor compartment and the drug formulation containing 60 % propylene glycol in the donor compartment. These conditions respect the in vivo situation, i.e., an application of a topical formulation on the skin. The XRPD was performed at room air humidity for technical reasons. The SC

PT

lipids do not extensively hydrate. Under normal conditions, SC is a relatively dry tissue. The SC water content amounts about 20% w/w and this is located primarily in

RI

cells [67]. At low relative humidity (RH) up to ~ 80 %, corneocytes take up more water than SC lipids do. Only at RH over 80%, hydration of lipids is more pronounced

SC

than that of corneocytes [68]. In the SC model membranes, 2 water molecules bind per one lipid after 24 h equilibration at 100 % RH [69]. To evaluate the effects of

NU

hydration, we performed several control XRPD measurements of the annealed model lipid membranes. The studied model lipid membranes showed neither membrane

MA

swelling nor substantial structure modification after hydration at ~ 100 RH for 24 h (not shown).

The results obtained on Langmuir monolayers should be compared with those

D

acquired by other experimental technics cautiously because of the fundamental

PT E

differences between the monolayers and multilamellar systems. The Langmuir monolayers were fully hydrated by ultrapure water in a subphase. Considering the effect of the subphase pH on FFA ionization, the apparent pKa values of FFA in

CE

monolayers are not known and are likely to be higher that the pKas in molecular solutions. Within hydrated (30% w/w) model SC lipid mixtures, the pKa values of FFA

AC

were in the range of 6.2-7.3. These are some 1.5-2.5 pH units higher than the pKas of fatty acids in molecular solution [70]. These pKa values indicate that majority of FFA in monolayers will be unionized. However, this is only a speculation because the pKa values of FFA in Langmuir monolayers could have differed from those in multilamellar system or in solution. Langmuir monolayers enable to shed more light on the basic intermolecular interactions of Cer, FFA and Chol with CholS. Another question is: What time is required to attain a stable polymorphic structure or to form orthorhombic domains? Mendelsohn et al. studied the phase separation kinetics in ternary mixtures of Cer/perdeuterated acid/Chol. Bovine brain Cer (type NS)/ stearic acid-d35 (SA-d35)/Chol mixture was heated above the - 26 -

ACCEPTED MANUSCRIPT temperature of the main phase transition, quenched to 31 °C and measured by kinetic IR spectroscopy. Ordered ceramide chains with some orthorhombically packed structures appeared after 0.5–8 h; phase separation of large orthorhombic domains of SA-d35 after 4–10 h [71]. The Mendelsohn’s group also found the kinetics to be sensitive to the difference in chain length between the Cer acyl chain and fatty acid. In CerNS24/perdeuterated palmitic acid (PA-d)/Chol the orthorhombic subcell of separated PA-d domain formed in the time interval 10 -100 h [22]. We observed

PT

similar phase separation linked to the chain length mismatch between CerNS16 and LIG or FFA(16-24) by XRPD. However, some authors referred even longer times

RI

(days) were required for the formation of orthorhombic phase. Fenske et al. showed

SC

that bovine brain Cer (type III)/PA-d31/Chol mixture required in the reestablishment of a solid phase from a fluid phase time periods of a week or more [72]. The

NU

equilibration of our multilamellar model lipid membranes (at least 24 h), which followed the annealing procedure, provide a period sufficient for the formation of

MA

orthorhombic domains in accordance with the publications of Mendelsohn’s group. Different times were attained in the Langmuir monolayers. Hartel et al. showed that the sphingomyelin-Cer conversion driven by sphingomyelinase led to the Cer-

D

enriched domain formation in the time-period of 40 s. The subsequent morphological

PT E

transition of Cer-enriched domains proceeded in hundreds of seconds [73]. Therefore, the time applied in the preparation of Langmuir monolayers (20 min) was adequate for solvent evaporation and lipid film equilibration.

CE

Even if there are technical restrictions in the effort to set the equal experimental conditions in various experimental techniques, each method provide a

AC

piece of the puzzle in the basic understanding of the skin lipid barrier.

5. Conclusion

The long chain CerNS16-based model membranes provided a less effective barrier to water and two model permeants than did the very long chain CerNS24based membranes. This higher permeability associated with the shorter ceramide was observed in model membranes that contained either a single fatty acid (lignoceric acid) or a heterogeneous mixture of fatty acids with 16 to 24 carbons. In addition, higher permeability was also detected in the membranes containing mixed

- 27 -

ACCEPTED MANUSCRIPT free fatty acids than in the membranes containing lignoceric acid with either the long or very long chain ceramide. The very long CerNS24- and the long CerNS16-based membranes showed differences in their microstructure: CerNS16 led to the formation of lamellae with a shorter repeat distance and was less miscible with fatty acids, whereas CerNS24 was well miscible with the studied fatty acids. At the air-water interface, the lipid mixture with CerNS16 formed a more expanded lipid monolayer with a larger area per lipid and larger fraction of low domain than did the mixture with

PT

CerNS24. Balanced interactions between Cer and Chol and Cer and FFA are most likely needed for the appropriate organization of the SC lipid lamellae. Cer with 16C

RI

acyl chains do not satisfy this requirement; therefore, they create several phases with

SC

a more advantageous composition than the initial mixture. The altered structure of the long chain CerNS16 membranes most likely contributes to their lower barrier

NU

function. Despite the obvious differences between the composition and structure of the studied system and in vivo SC lipid matrix, these findings suggest that the replacement of a very long chain ceramide by a long chain ceramide, i.e., shortening

MA

of the ceramide acyl chain by 8 carbons, can directly contribute to a less effective

PT E

Supporting Information

D

lipid barrier.

The calibration curve for the water loss measurements (Fig. S1), the validation of lipid homogeneity in the membrane (Fig. S2) and the effect of 60% propylene glycol on

CE

the membrane (Fig. S3).

AC

Acknowledgments

The study is co-financed by the European Social Fund and the state budget of the Czech Republic, project no. CZ.1.07/2.3.00/30.0061, by the Czech Science Foundation (13-23891S) and Charles University in Prague (SVV 260 291).

- 28 -

ACCEPTED MANUSCRIPT

References

[3]

[4]

[7]

[11]

[12]

[13]

CE

[10]

AC

[9]

PT E

D

[8]

MA

[6]

NU

SC

[5]

PT

[2]

P.M. Elias, Epidermal Lipids, Membranes, and Keratinization, Int. J. Dermatol. 20 (1981) 1–19. doi:10.1111/j.1365-4362.1981.tb05278.x. J. van Smeden, L. Hoppel, R. van der Heijden, T. Hankemeier, R.J. Vreeken, J.A. Bouwstra, LC/MS analysis of stratum corneum lipids: ceramide profiling and discovery, J. Lipid Res. 52 (2011) 1211–1221. doi:10.1194/jlr.M014456. B. Breiden, K. Sandhoff, The role of sphingolipid metabolism in cutaneous permeabilitybarrier formation, Biochim. Biophys. Acta BBA - Mol. Cell Biol. Lipids. 1841 (2014) 441–452. doi:10.1016/j.bbalip.2013.08.010. M. Rabionet, K. Gorgas, R. Sandhoff, Ceramide synthesis in the epidermis, Biochim. Biophys. Acta BBA - Mol. Cell Biol. Lipids. 1841 (2014) 422–434. doi:10.1016/j.bbalip.2013.08.011. Y. Pewzner-Jung, H. Park, E.L. Laviad, L.C. Silva, S. Lahiri, J. Stiban, R. ErezRoman, B. Brugger, T. Sachsenheimer, F. Wieland, M. Prieto, A.H. Merrill, A.H. Futerman, A Critical Role for Ceramide Synthase 2 in Liver Homeostasis: I. Alterations in Lipid Metabolic Pathways, J. Biol. Chem. 285 (2010) 10902– 10910. doi:10.1074/jbc.M109.077594. S. Grösch, S. Schiffmann, G. Geisslinger, Chain length-specific properties of ceramides, Prog. Lipid Res. 51 (2012) 50–62. doi:10.1016/j.plipres.2011.11.001. S. Chiantia, N. Kahya, P. Schwille, Raft Domain Reorganization Driven by Short- and Long-Chain Ceramide:  A Combined AFM and FCS Study, Langmuir. 23 (2007) 7659–7665. doi:10.1021/la7010919. S. Nybond, Y.J.E. Björkqvist, B. Ramstedt, J.P. Slotte, Acyl chain length affects ceramide action on sterol/sphingomyelin-rich domains, Biochim. Biophys. Acta BBA - Biomembr. 1718 (2005) 61–66. doi:10.1016/j.bbamem.2005.10.009. J. Sot, F.J. Aranda, M.-I. Collado, F.M. Goñi, A. Alonso, Different Effects of Long- and Short-Chain Ceramides on the Gel-Fluid and Lamellar-Hexagonal Transitions of Phospholipids: A Calorimetric, NMR, and X-Ray Diffraction Study, Biophys. J. 88 (2005) 3368–3380. doi:10.1529/biophysj.104.057851. W.M. Holleran, Y. Takagi, Y. Uchida, Epidermal sphingolipids: metabolism, function, and roles in skin disorders, FEBS Lett. 580 (2006) 5456–5466. doi:10.1016/j.febslet.2006.08.039. J. Ishikawa, H. Narita, N. Kondo, M. Hotta, Y. Takagi, Y. Masukawa, T. Kitahara, Y. Takema, S. Koyano, S. Yamazaki, A. Hatamochi, Changes in the Ceramide Profile of Atopic Dermatitis Patients, J. Invest. Dermatol. 130 (2010) 2511–2514. doi:10.1038/jid.2010.161. M.J. Choi, D.H.I. Maibach, Role of Ceramides in Barrier Function of Healthy and Diseased Skin, Am. J. Clin. Dermatol. 6 (2005) 215–223. doi:10.2165/00128071-200506040-00002. M. Janssens, J. van Smeden, G.S. Gooris, W. Bras, G. Portale, P.J. Caspers, R.J. Vreeken, T. Hankemeier, S. Kezic, R. Wolterbeek, A.P. Lavrijsen, J.A. Bouwstra, Increase in short-chain ceramides correlates with an altered lipid organization and decreased barrier function in atopic eczema patients, J. Lipid Res. 53 (2012) 2755–2766. doi:10.1194/jlr.P030338.

RI

[1]

- 29 -

ACCEPTED MANUSCRIPT

[17]

[18]

[20]

[24]

[25]

[26]

[27]

CE

[23]

AC

[22]

PT E

D

[21]

MA

NU

[19]

PT

[16]

RI

[15]

C. Tawada, H. Kanoh, M. Nakamura, Y. Mizutani, T. Fujisawa, Y. Banno, M. Seishima, Interferon-γ Decreases Ceramides with Long-Chain Fatty Acids: Possible Involvement in Atopic Dermatitis and Psoriasis, J. Invest. Dermatol. 134 (2014) 712–718. doi:10.1038/jid.2013.364. Y.-H. Park, W.-H. Jang, J.A. Seo, M. Park, T.R. Lee, Y.-H. Park, D.K. Kim, K.M. Lim, Decrease of Ceramides with Very Long–Chain Fatty Acids and Downregulation of Elongases in a Murine Atopic Dermatitis Model, J. Invest. Dermatol. 132 (2012) 476–479. doi:10.1038/jid.2011.333. L. Landmann, The epidermal permeability barrier. Comparison between in vivo and in vitro lipid structures, Eur. J. Cell Biol. 33 (1984) 258–264. M. de Jager, W. Groenink, J. van der Spek, C. Janmaat, G. Gooris, M. Ponec, J. Bouwstra, Preparation and characterization of a stratum corneum substitute for in vitro percutaneous penetration studies, Biochim. Biophys. Acta BBA Biomembr. 1758 (2006) 636–644. doi:10.1016/j.bbamem.2006.04.001. M. Ochalek, S. Heissler, J. Wohlrab, R.H.H. Neubert, Characterization of lipid model membranes designed for studying impact of ceramide species on drug diffusion and penetration, Eur. J. Pharm. Biopharm. 81 (2012) 113–120. doi:10.1016/j.ejpb.2012.02.002. M. Ochalek, H. Podhaisky, H.-H. Ruettinger, J. Wohlrab, R.H.H. Neubert, SC lipid model membranes designed for studying impact of ceramide species on drug diffusion and permeation – Part II: Diffusion and permeation of model drugs, Eur. J. Pharm. Biopharm. 82 (2012) 360–366. doi:10.1016/j.ejpb.2012.06.008. A. Paz Ramos, M. Lafleur, Chain Length of Free Fatty Acids Influences the Phase Behavior of Stratum Corneum Model Membranes, Langmuir. 31 (2015) 11621–11629. doi:10.1021/acs.langmuir.5b03271. M. Oguri, G.S. Gooris, K. Bito, J.A. Bouwstra, The effect of the chain length distribution of free fatty acids on the mixing properties of stratum corneum model membranes, Biochim. Biophys. Acta BBA - Biomembr. 1838 (2014) 1851–1861. doi:10.1016/j.bbamem.2014.02.009. R. Mendelsohn, E. Rabie, R.M. Walters, C.R. Flach, Fatty Acid Chain Length Dependence of Phase Separation Kinetics in Stratum Corneum Models by IR Spectroscopy, J. Phys. Chem. B. 119 (2015) 9740–9750. doi:10.1021/acs.jpcb.5b03045. E.H. Mojumdar, Z. Kariman, L. van Kerckhove, G.S. Gooris, J.A. Bouwstra, The role of ceramide chain length distribution on the barrier properties of the skin lipid membranes, Biochim. Biophys. Acta BBA - Biomembr. 1838 (2014) 2473–2483. doi:10.1016/j.bbamem.2014.05.023. M.H. Lee, B. Lim, J.W. Kim, E.J. An, D. Lee, Effect of composition on water permeability of model stratum corneum lipid membranes, Soft Matter. 8 (2012) 1539–1546. doi:10.1039/C1SM06613G. F. Dupuy, B. Maggio, The hydrophobic mismatch determines the miscibility of ceramides in lipid monolayers, Chem. Phys. Lipids. 165 (2012) 615–629. doi:10.1016/j.chemphyslip.2012.06.008. B. Školová, B. Janůšová, J. Zbytovská, G. Gooris, J. Bouwstra, P. Slepička, P. Berka, J. Roh, K. Palát, A. Hrabálek, K. Vávrová, Ceramides in the Skin Lipid Membranes: Length Matters, Langmuir. 29 (2013) 15624–15633. doi:10.1021/la4037474. B. Janůšová, J. Zbytovská, P. Lorenc, H. Vavrysová, K. Palát, A. Hrabálek, K. Vávrová, Effect of ceramide acyl chain length on skin permeability and

SC

[14]

- 30 -

ACCEPTED MANUSCRIPT

[33]

[34]

[35]

[36]

[37]

[38]

[39]

PT

RI

SC

NU

MA

[32]

D

[31]

PT E

[30]

CE

[29]

AC

[28]

thermotropic phase behavior of model stratum corneum lipid membranes, Biochim. Biophys. Acta BBA - Mol. Cell Biol. Lipids. 1811 (2011) 129–137. doi:10.1016/j.bbalip.2010.12.003. S. Stahlberg, B. Školová, P.K. Madhu, A. Vogel, K. Vávrová, D. Huster, Probing the Role of the Ceramide Acyl Chain Length and Sphingosine Unsaturation in Model Skin Barrier Lipid Mixtures by 2 H Solid-State NMR Spectroscopy, Langmuir. 31 (2015) 4906–4915. doi:10.1021/acs.langmuir.5b00751. B. Školová, K. Hudská, P. Pullmannová, A. Kováčik, K. Palát, J. Roh, J. Fleddermann, I. Estrela-Lopis, K. Vávrová, Different phase behavior and packing of ceramides with long (C16) and very long (C24) acyls in model membranes: infrared spectroscopy using deuterated lipids, J. Phys. Chem. B. 118 (2014) 10460–10470. doi:10.1021/jp506407r. D. Groen, G.S. Gooris, J.A. Bouwstra, Model Membranes Prepared with Ceramide EOS, Cholesterol and Free Fatty Acids Form a Unique Lamellar Phase, Langmuir. 26 (2010) 4168–4175. doi:10.1021/la9047038. W. Abraham, D.T. Downing, Preparation of Model Membranes for Skin Permeability Studies Using Stratum Corneum Lipids, J. Invest. Dermatol. 93 (1989) 809–813. doi:10.1111/1523-1747.ep12284431. P. Pullmannová, K. Staňková, M. Pospíšilová, B. Školová, J. Zbytovská, K. Vávrová, Effects of sphingomyelin/ceramide ratio on the permeability and microstructure of model stratum corneum lipid membranes, Biochim. Biophys. Acta BBA - Biomembr. 1838 (2014) 2115–2126. doi:10.1016/j.bbamem.2014.05.001. B. Školová, A. Kováčik, O. Tesař, L. Opálka, K. Vávrová, Phytosphingosine, sphingosine and dihydrosphingosine ceramides in model skin lipid membranes: permeability and biophysics, Biochim. Biophys. Acta BBA Biomembr. (2017) In Press. doi:10.1016/j.bbamem.2017.01.019. M. Ochalek, H. Podhaisky, H.-H. Ruettinger, R.H.H. Neubert, J. Wohlrab, SC lipid model membranes designed for studying impact of ceramide species on drug diffusion and permeation, Part III: Influence of penetration enhancer on diffusion and permeation of model drugs, Int. J. Pharm. 436 (2012) 206–213. doi:10.1016/j.ijpharm.2012.06.044. S. Mitragotri, Modeling skin permeability to hydrophilic and hydrophobic solutes based on four permeation pathways, J. Controlled Release. 86 (2003) 69–92. doi:10.1016/S0168-3659(02)00321-8. T. Yoshiike, Y. Aikawa, J. Sindhvananda, H. Suto, K. Nishimura, T. Kawamoto, H. Ogawa, Skin barrier defect in atopic dermatitis: increased permeability of the stratum corneum using dimethyl sulfoxide and theophylline, J. Dermatol. Sci. 5 (1993) 92–96. doi:10.1016/0923-1811(93)90076-2. K.-P. Wilhelm, C. Surber, H.I. Maibach, Effect of Sodium Lauryl Sulfate— Induced Skin Irritation on In Vivo Percutaneous Penetration of Four Drugs, J. Invest. Dermatol. 97 (1991) 927–932. doi:10.1111/1523-1747.ep12491710. K. Vávrová, A. Hrabálek, S. Mac-Mary, P. Humbert, P. Muret, Ceramide analogue 14S24 selectively recovers perturbed human skin barrier, Br. J. Dermatol. 157 (2007) 704–712. doi:10.1111/j.1365-2133.2007.08113.x. M. Eeman, G. Francius, Y.F. Dufrêne, K. Nott, M. Paquot, M. Deleu, Effect of cholesterol and fatty acids on the molecular interactions of fengycin with Stratum corneum mimicking lipid monolayers, Langmuir ACS J. Surf. Colloids. 25 (2009) 3029–3039. doi:10.1021/la803439n.

- 31 -

ACCEPTED MANUSCRIPT

[47]

[48]

[49]

[50]

[51]

[52]

[53]

[54]

PT

RI

SC

[46]

NU

[45]

MA

[44]

D

[43]

PT E

[42]

CE

[41]

B.M. Craven, Pseudosymmetry in cholesterol monohydrate, Acta Crystallogr. B. 35 (1979) 1123–1128. doi:10.1107/S0567740879005719. H.S. Shieh, L.G. Hoard, C.E. Nordman, The structure of cholesterol, Acta Crystallogr. B. 37 (1981) 1538–1543. doi:10.1107/S0567740881006523. J. Huang, J.T. Buboltz, G.W. Feigenson, Maximum solubility of cholesterol in phosphatidylcholine and phosphatidylethanolamine bilayers, Biochim. Biophys. Acta BBA - Biomembr. 1417 (1999) 89–100. doi:10.1016/S00052736(98)00260-0. M.W. de Jager, G.S. Gooris, I.P. Dolbnya, W. Bras, M. Ponec, J.A. Bouwstra, The phase behaviour of skin lipid mixtures based on synthetic ceramides, Chem. Phys. Lipids. 124 (2003) 123–134. doi:10.1016/S0009-3084(03)000501. J.A. Bouwstra, G.S. Gooris, F.E.R. Dubbelaar, A.M. Weerheim, A.P. IJzerman, M. Ponec, Role of ceramide 1 in the molecular organization of the stratum corneum lipids, J. Lipid Res. 39 (1998) 186–196. J.A. Bouwstra, G.S. Gooris, K. Cheng, A. Weerheim, W. Bras, M. Ponec, Phase behavior of isolated skin lipids, J. Lipid Res. 37 (1996) 999–1011. J.A. Bouwstra, G.S. Gooris, M.A.S. Vries, J.A. van der Spek, W. Bras, Structure of human stratum corneum as a function of temperature and hydration: A wide-angle X-ray diffraction study, Int. J. Pharm. 84 (1992) 205– 216. doi:10.1016/0378-5173(92)90158-X. E.H. Mojumdar, D. Groen, G.S. Gooris, D.J. Barlow, M.J. Lawrence, B. Demé, J.A. Bouwstra, Localization of Cholesterol and Fatty Acid in a Model Lipid Membrane: A Neutron Diffraction Approach, Biophys. J. 105 (2013) 911–918. doi:10.1016/j.bpj.2013.07.003. J. Shah, J.M. Atienza, R.I. Duclos, A.V. Rawlings, Z. Dong, G.G. Shipley, Structural and thermotropic properties of synthetic C16:0 (palmitoyl) ceramide: effect of hydration., J. Lipid Res. 36 (1995) 1936–1944. S.L. Souza, M. José Capitán, J. Álvarez, S.S. Funari, M.H. Lameiro, E. Melo, Phase Behavior of Aqueous Dispersions of Mixtures of N-Palmitoyl Ceramide and Cholesterol: A Lipid System with Ceramide−Cholesterol Crystalline Lamellar Phases, J. Phys. Chem. B. 113 (2009) 1367–1375. doi:10.1021/jp803331k. A. Ruettinger, M.A. Kiselev, T. Hauss, S. Dante, A.M. Balagurov, R.H.H. Neubert, Fatty acid interdigitation in stratum corneum model membranes: a neutron diffraction study, Eur. Biophys. J. 37 (2008) 759–771. doi:10.1007/s00249-008-0258-3. X. Chen, S. Kwak, M. Lafleur, M. Bloom, N. Kitson, J. Thewalt, Fatty Acids Influence “Solid” Phase Formation in Models of Stratum Corneum Intercellular Membranes, Langmuir. 23 (2007) 5548–5556. doi:10.1021/la063640+. E. Brief, S. Kwak, J.T.J. Cheng, N. Kitson, J. Thewalt, M. Lafleur, Phase behavior of an equimolar mixture of N-palmitoyl-D-erythro-sphingosine, cholesterol, and palmitic acid, a mixture with optimized hydrophobic matching, Langmuir ACS J. Surf. Colloids. 25 (2009) 7523–7532. doi:10.1021/la9003643. J.A. Bouwstra, M. Ponec, The skin barrier in healthy and diseased state, Biochim. Biophys. Acta BBA - Biomembr. 1758 (2006) 2080–2095. doi:10.1016/j.bbamem.2006.06.021. D. Marsh, Handbook of Lipid Bilayers, Second Edition, CRC Press, 2013.

AC

[40]

- 32 -

ACCEPTED MANUSCRIPT

[57]

[58] [59]

[65] [66]

[67]

[68]

[69]

[70]

MA

D

[64]

PT E

[63]

CE

[62]

AC

[61]

NU

SC

[60]

PT

[56]

D. Marsh, Lateral order in gel, subgel and crystalline phases of lipid membranes: Wide-angle X-ray scattering, Chem. Phys. Lipids. 165 (2012) 59– 76. doi:10.1016/j.chemphyslip.2011.11.001. I. Pascher, S. Sundell, Molecular arrangements in sphingolipids: crystal structure of the ceramide N-(2d,3d-dihydroxyoctadecanoyl)-phytosphingosine, Chem. Phys. Lipids. 62 (1992) 79–86. doi:10.1016/0009-3084(92)90056-U. S. Abrahamsson, B. Dahlen, H. Löfgren, I. Pascher, Lateral packing of hydrocarbon chains, Prog. Chem. Fats Other Lipids. 16 (1978) 125–143. doi:10.1016/0079-6832(78)90039-3. L. Norlén, Skin barrier structure and function: the single gel phase model, J. Invest. Dermatol. 117 (2001) 830–836. doi:10.1038/jid.2001.1. T.J. McIntosh, Organization of Skin Stratum Corneum Extracellular Lamellae: Diffraction Evidence for Asymmetric Distribution of Cholesterol, Biophys. J. 85 (2003) 1675–1681. doi:10.1016/S0006-3495(03)74597-4. I. Iwai, H. Han, L. den Hollander, S. Svensson, L.-G. Öfverstedt, J. Anwar, J. Brewer, M. Bloksgaard, A. Laloeuf, D. Nosek, S. Masich, L.A. Bagatolli, U. Skoglund, L. Norlén, The Human Skin Barrier Is Organized as Stacked Bilayers of Fully Extended Ceramides with Cholesterol Molecules Associated with the Ceramide Sphingoid Moiety, J. Invest. Dermatol. 132 (2012) 2215– 2225. doi:10.1038/jid.2012.43. G. Degerman, E. von Sydow, The Thermal Expansion of the Crystalline B-and C-Forms of Stearic Acid, Acta Chem Scand. 13 (1959) 984–988. M. Klacsová, J. Karlovská, D. Uhríková, S.S. Funari, P. Balgavý, Phase behavior of the DOPE + DOPC + alkanol system, Soft Matter. 10 (2014) 5842. doi:10.1039/C4SM00530A. E. von Sydow, On the structure of the crystal form B of stearic acid, Acta Crystallogr. 8 (1955) 557–560. doi:10.1107/S0365110X55001746. E. von Sydow, On the structure of the crystal form B’ of n-pentadecanoic acid, Acta Crystallogr. 7 (1954) 823–826. doi:10.1107/S0365110X54002538. E. von Sydow, On the structure of the crystal form C’ of n-hendecanoic acid, Acta Crystallogr. 8 (1955) 810–813. doi:10.1107/S0365110X55002442. S. Björklund, J. Engblom, K. Thuresson, E. Sparr, A water gradient can be used to regulate drug transport across skin, J. Controlled Release. 143 (2010) 191–200. doi:10.1016/j.jconrel.2010.01.005. J.A. Bouwstra, P.L. Honeywell-Nguyen, G.S. Gooris, M. Ponec, Structure of the skin barrier and its modulation by vesicular formulations, Prog. Lipid Res. 42 (2003) 1–36. C.L. Silva, D. Topgaard, V. Kocherbitov, J.J.S. Sousa, A.A.C.C. Pais, E. Sparr, Stratum corneum hydration: Phase transformations and mobility in stratum corneum, extracted lipids and isolated corneocytes, Biochim. Biophys. Acta BBA - Biomembr. 1768 (2007) 2647–2659. doi:10.1016/j.bbamem.2007.05.028. D. Groen, G.S. Gooris, D.J. Barlow, M.J. Lawrence, J.B. van Mechelen, B. Demé, J.A. Bouwstra, Disposition of Ceramide in Model Lipid Membranes Determined by Neutron Diffraction, Biophys. J. 100 (2011) 1481–1489. doi:10.1016/j.bpj.2011.02.001. R. Lieckfeldt, J. Villalaín, J.C. Gómez-Fernández, G. Lee, Apparent pKa of the fatty acids within ordered mixtures of model human stratum corneum lipids, Pharm. Res. 12 (1995) 1614–1617.

RI

[55]

- 33 -

ACCEPTED MANUSCRIPT

PT

RI SC NU MA D PT E

[73]

CE

[72]

R. Mendelsohn, I. Selevany, D.J. Moore, M.C. Mack Correa, G. Mao, R.M. Walters, C.R. Flach, Kinetic Evidence Suggests Spinodal Phase Separation in Stratum Corneum Models by IR Spectroscopy, J. Phys. Chem. B. 118 (2014) 4378–4387. doi:10.1021/jp501003c. D.B. Fenske, J.L. Thewalt, M. Bloom, N. Kitson, Models of stratum corneum intercellular membranes: 2H NMR of macroscopically oriented multilayers, Biophys. J. 67 (1994) 1562–1573. doi:10.1016/S0006-3495(94)80629-0. S. Härtel, M.L. Fanani, B. Maggio, Shape Transitions and Lattice Structuring of Ceramide-Enriched Domains Generated by Sphingomyelinase in Lipid Monolayers, Biophys. J. 88 (2005) 287–304. doi:10.1529/biophysj.104.048959.

AC

[71]

- 34 -

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

NU

SC

RI

PT

Graphical abstract

- 35 -

ACCEPTED MANUSCRIPT Highlights Shortening of ceramide chains from 24 to 16 carbons increased membrane permeability. C16 acyl ceramide does not mix well with other skin barrier lipids. A higher area per lipid in C16 ceramide monolayers than in C24 ceramide

AC

CE

PT E

D

MA

NU

SC

RI

PT

monolayers.

- 36 -