Tracking solvents in the skin through atomicallyresolved measurements of molecular mobility in intactstratum corneumQuoc Dat Phama, Daniel Topgaarda, and Emma Sparra,1

aDivision of Physical Chemistry, Chemistry Department, Lund University, 22100 Lund, Sweden

Edited by Jacob N. Israelachvili, University of California, Santa Barbara, CA, and approved December 1, 2016 (received for review July 5, 2016)

Solvents are commonly used in pharmaceutical and cosmeticformulations and sanitary products and cleansers. The uptake ofsolvent into the skin may change the molecular organization ofskin lipids and proteins, which may in turn alter the protective skinbarrier function. We herein examine the molecular effects of 10different solvents on the outermost layer of skin, the stratumcorneum (SC), using polarization transfer solid-state NMR onnatural abundance 13C in intact SC. With this approach it is possibleto characterize the molecular dynamics of solvent molecules whenpresent inside intact SC and to simultaneously monitor the effectscaused by the added solvent on SC lipids and protein components.All solvents investigated cause an increased fluidity of SC lipids,with the most prominent effects shown for the apolar hydrocar-bon solvents and 2-propanol. However, no solvent other thanwater shows the ability to fluidize amino acids in the keratin fila-ments. The solvent molecules themselves show reduced molecularmobility when incorporated in the SC matrix. Changes in the mo-lecular properties of the SC, and in particular alternation in thebalance between solid and fluid SC components, may have signif-icant influences on the macroscopic SC barrier properties as well asmechanical properties of the skin. Deepened understanding ofmolecular effects of foreign compounds in SC fluidity can there-fore have strong impact on the development of skin products inpharmaceutical, cosmetic, and sanitary applications.

keratin filaments | solid-state NMR | extracellular lipids | corneocytes |phase behavior

The human skin makes up a large interfacial barrier film thatprotects the body from desiccation and uptake of hazardous

chemicals. In many situations in everyday life, the skin is, con-sciously or unconsciously, exposed to solvents and other chem-icals. The uptake of solvents in the skin may lead to changes inthe molecular organization of skin lipids and proteins, which inturn may alter macroscopic properties of the skin, including skinflexibility, softness, and permeability (1–6). The exposure tosolvents may also cause irritation, corrosion, or irritant contactdermatitis (7). When the skin is exposed to solutions or creamswith complex composition—such as skin lotions, cleansers, washingagents, or pharmaceutical products—several molecular mecha-nisms operate simultaneously due to interactions between molec-ular components in the skin and molecules in the complex solution,in which a solvent is often the main component.The barrier function of the skin is assured by its outermost

layer, the stratum corneum (SC), which consists of dead keratin-filled cells, corneocytes, embedded in a multilamellar lipid ma-trix (Fig. 1A) (8, 9). SC differs from most other biologicalmembranes in that the main fraction of both lipids and proteinsis solid at ambient conditions (10, 11). Still, a minor fraction oflipids and protein components is mobile at physiological tem-peratures in conditions of hydrated SC, and after the addition ofso-called moisturizers or penetration enhancers (6, 11–15). Themacroscopic material properties of the SC barrier membranenaturally depend on the molecular properties of its buildingblocks. In particular, transport properties as well as mechanical

properties and water-holding capacity can be altered by changingthe balance between solid and fluid structures in the complex SCmaterial (13, 14, 16–19). Increasing fluidity is expected to lead tohigher solubility and a higher diffusion coefficient for mostadded compounds, and thus increased effective SC permeability(14, 18, 19). A clear correlation between SC fluidity/molecularmobility and permeability of small molecules in SC has beendemonstrated in previous studies (11–14, 16, 17, 20). The pres-ence of fluid domains in the SC matrix is likely also important forother biological functions, for example the enzyme activity in theSC intercellular space (21). There is a rather extensive literaturedescribing how solvents and excipients influence the extracellularsolid SC lipids based on studies using scattering, diffraction, andinfrared spectroscopy techniques (2, 3, 5, 22, 23). Calorimetricstudies have further shown that solvents may alter the thermalsolid–fluid transitions in SC (2, 3, 5), which typically occur atnonphysiological elevated temperatures. However, these studiesprovide very limited information on how solvents influence theminor fractions of SC lipid and protein components that are fluidat ambient temperatures.In this study, we use polarization transfer solid-state NMR

(PT ssNMR) (24) on natural abundance 13C to obtain atomi-cally resolved information on molecular dynamics in intact SC(11). We are able to track the solvent molecules when in-corporated into the SC complex matter and study their interac-tions with SC matrix via changes in their molecular dynamics.Simultaneously, we obtain information on the molecular effectscaused by the added solvent on SC lipid and protein componentsin the very same sample. The present NMR method is sufficiently

Significance

Our skin is regularly exposed to solvents in cosmetics, washingand sanitary agents, and drug formulations. The uptake ofsolvents into the skin may change essential properties of theskin, for example, its protective barrier function, as well as itsflexibility and softness. Herein different solvents relevant toskin formulations and sanitary products were added to sam-ples of intact stratum corneum (SC), which is the outer layer ofthe skin. The solvent molecules can be tracked inside SC,showing reduced mobility. Furthermore, the solvents inducefluidity in SC components. These changes depend on solventidentity and concentration and on SC hydration conditions.Changes in SC components can be related to changes in mac-roscopic properties of SC, including skin barrier function.

Author contributions: Q.D.P. and E.S. designed research; Q.D.P. performed research; Q.D.P.,D.T., and E.S. analyzed data; and Q.D.P., D.T., and E.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: emma.sparr@fkem1.lu.se.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1608739114/-/DCSupplemental.

E112–E121 | PNAS | Published online December 27, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1608739114

sensitive to detect changes in the minor fraction of mobile com-ponents, and it also gives information on the major fraction ofrigid components (11, 24), thereby providing information on SCmolecular dynamics with a molecular detail that has not beenreachable before. In addition, it is possible to distinguish severalcoexisting fractions of the same type of molecular component withdifferent molecular dynamics.We report the molecular effects of 10 different solvents rep-

resenting different molecular classes with respect to structureand hydrophobicity (Table 1). We use samples of intact porcineSC treated with solvents, and in the very same experiments westudy the molecular dynamics (“fluidity”) in solvent, as well asthe SC lipid and protein components. When added to the com-plex structure of SC the solvent molecules will partition betweenthe relatively hydrophilic domains in the keratin-filled corneo-cyte cells and hydrophobic extracellular lipid regions (Fig. 1A). Ithas previously been shown that substantial amounts of water canbe taken up into the corneocytes (25), which has a strong effecton the molecular mobility in the hydrophilic amino acids that areenriched in the terminal chains of keratin filaments (11). Theextracellular matrix contains hydrophobic lipids with long acylchains and very low water content (26, 27). Still, tiny amounts ofwater and other polar solvents can be present in the vicinity ofthe polar headgroups of the lipids (3, 11, 12). Apolar com-pounds, on the other hand, likely partition into the hydrocarbonlayer in the extracellular lipid matrix. In many cases there is aneffect of the treatment with the external chemicals on the fluidityin lipids and keratin filaments (2, 4, 12, 13, 28), which in turn mayalter the solvent partitioning balance.Most of the solvents investigated here are found in pharma-

ceutical, cosmetic, and sanitary applications. Methanol andethanol are common laboratory solvents. Ethanol is also fre-quently used in washing and sanitizing as well as in (trans)dermalpatches with long skin exposure times. Some of the polar andapolar solvents [e.g., octamethyltrisiloxane (OMTS) and 1-dec-anol] are typical (co)solvents in personal care, cosmetic, andtopical pharmaceutical formulations, where propylene glycol(PG) is also used as a penetration enhancer (6). Glycerol isnaturally present in SC as a component of the so-called naturalmoisturizing factor (29) and it is commonly used in skin careproducts as humectant and/or penetration enhancer (13, 15, 20).DMSO is used in industry and topical applications as a solventfor unsaturated and aromatic hydrocarbons. DMSO has alsobeen shown to enhance percutaneous absorption of certain

chemical agents (30) and it possesses some biological functionsincluding antiinflammatory activity (31).We aim at deepened understanding of interactions between

various solvents and SC lipid and protein components in well-defined controlled conditions in terms of hydration and solventconcentration. We investigate samples of intact SC using PTssNMR experiments, with the following objectives:

i) Characterize molecular dynamics and partitioning of thevarying classes of solvent molecules between different re-gions within the complex structure of SC.

ii) Elucidate how the different solvents influence the molecularmobility in different segments of SC lipids and proteins.

iii) Distinguish general trends of how different classes of sol-vents influence SC molecular components and relate thisto the properties of solvents.

iv) Investigate the consequences of washing SC in excesssolvent.

ResultsPT ssNMR to Monitor Molecular Conformation and Dynamics inMolecular Matter with Complex Composition. We aim at molecu-lar characterization of the interaction between solvent moleculesand lipids and proteins within intact SC. From the PT ssNMRmeasurements we observe changes in molecular dynamics of thesolvent molecules inside intact SC. Simultaneously, we detectchanges in the molecular mobility in SC lipid and protein com-ponents. We start by shortly introducing the PT ssNMR methodto study intact SC (Fig. 1) and a summary of the main findings(Fig. 2 and Table 1). This is followed by a more detailed pre-sentation of the results for each individual case.In a 13C NMR spectrum the molecular moieties of the SC

components and solvents can be differentiated based on theirchemical shifts, and the mobility of each segment is furtherobtained by using polarization transfer. The PT ssNMR experi-ments are sensitive to the minor fraction of mobile componentsin intact SC, and they simultaneously provide information on therigid/obstructed components in the very same sample. We candistinguish the same molecular segment that resonates at dif-ferent chemical shifts depending on molecular conformation anddynamics. Even at the same chemical shift, chemically identicalsegments with different molecular mobility can be discernedbased on the line shapes and intensities of the peaks. Further-more, for each molecular segment, it is possible to characterizethe timescale and anisotropy of C–H bond reorientation (Fig. 2A

A

C D E

B

Fig. 1. (A) The schematics illustrate the brick-and-mortar model of SC with corneocytes filled with keratin filaments, surrounded by a multilamellar lipidmatrix. (B) Chemical structures with numbered segments of cholesterol and relevant lipid carbons (here illustrated with ceramide nonhydroxy sphin-gosine) are also shown. (C–E ) 13C MAS NMR spectra (DP, gray; CP, blue; and INEPT, red) of SC at dry condition (C) and at 20 (D) and 40 (E ) wt % waterat 32 °C.

Pham et al. PNAS | Published online December 27, 2016 | E113

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

and SI Appendix, Fig. S1) by comparing the signal intensitiesobtained with the direct polarization (DP), cross-polarization(CP) (32), and insensitive nuclei enhanced by polarizationtransfer (INEPT) (33) acquisition schemes. Example spectra forhydrated intact SC are shown in Fig. 1E. The DP spectrumgenerally shows resonances from all carbons in the sample and ishere used as reference. Under the present experimental condi-tions, the DP signal is quantitative when the rotational correla-tion time τc < 10 ns and gradually disappears at longer τc (24).CP is efficient in boosting the signal of segments with slow (τc >0.1 ms) and/or anisotropic motions. The CP signal is not ob-servable for nearly isotropic reorientation (order parameterjSCHj < 0.01) with fast dynamics (τc < 10 ns). However, the signalfrom mobile segments (τc < 10 ns) is selectively enhanced in theINEPT spectra, whereas the INEPT signal is not visible forslower motion (τc ≥ 0.1 μs) and/or highly anisotropic reor-ientation (jSCHj > 0.5) (24). The dependence of CP and INEPTintensities on τc and jSCHj of the C–H bond reorientation isshown in SI Appendix, Fig. S1. In addition, the lineshape of theCP resonances also provides information on the molecular dy-namics in different segments. Under the experimental conditionsused here, maximum line broadening due to motion-decoupling

interference is expected at τc around 0.1 μs (24, 34), and thiseffect gradually disappears at both faster and slower motions.Furthermore, a segment with molecular reorientation on thetimescale of magic-angle spinning (MAS) (∼0.1 ms) would giverise to a broad CP signal due to spinning-motion interference,and this broadening effect also dissipates at both faster andslower motions (35). In fact, a segment with a range of molecularconformations and microenvironments with different chemicalshifts, as observed in the present system, would give rise to a broadsignal when τc > 0.1 ms because the motion is not fast enough toaverage the resonance frequencies (36). Taken together, the linenarrowing effect due to faster motion is expected to be observed atτc ∼0.1 μs.To facilitate the presentation of the data in this paper, we de-

fine “mobile” and “rigid” segments for SC components accordingto the rate and anisotropy of C–H bond reorientation (SI Ap-pendix, Fig. S1). The INEPT signal indicates a “mobile” segment,which can be “fast isotropic” or “fast anisotropic.” In the lattercase, the INEPT signal is accompanied by a sharp CP signal. Theterm “rigid” is defined by the conditions when only CP signal isdetected. The solvent molecules are characterized as isotropic,fast anisotropic, or “obstructed,” where the latter signifies solvent

Table 1. Summary of molecular effects of adding solvents to intact SC, including molecular dynamics of the solvent molecules inside SCand changes in molecular mobility of SC lipid and protein components induced by solvent

Here, the solvent was added at varying concentrations to dry SC with no additional water. The solvent concentration was varied to identify a concentrationregime where an obstructed/fast anisotropic fraction of solvent coexists with an isotropic fraction of solvent. The logarithm of octanol/water partitioncoefficient (log P) and the molecular weight (Mw, g/mol) of the solvents are also listed. log P is used as a facile reference value for the hydrophobicity ofthe solvents. It does not account for differences in solubility due to variations in the fluid and solid structures. The colors indicate the types of solvents: polarsolvents (blue), apolar solvents (yellow), and the solvent with log P ∼ 0 (green). In the case of water, the molecular dynamics of the solvent is not studied,because water does not contain any carbons. The effects of the solvents on SC lipid chain mobility are compared with the effects of water at the sameconcentration. The symbols should be read as weaker (<), stronger (>), and much stronger (>>) fluidizing effect compared with water. Abbreviations: fast ani,fast anisotropic; iso, isotropic; N.A., not applicable; N.D., not detected; and obs, obstructed. All experiments were performed at 32 °C.

E114 | www.pnas.org/cgi/doi/10.1073/pnas.1608739114 Pham et al.

segments that are detected in the CP and DP spectra but not inthe INEPT spectra and that have slower motion (τc ∼ 0.1 μs)compared with fast isotropic reorientation (Fig. 2A). For thepresent system with solvents incorporated in SC we can rule outthe interpretation of sharp CP signal of the solvents as a sign ofsolid crystals or the fast (τc < 10 ns) and highly anisotropic (jSCHj ≥0.5) motion that is expected for rotator- and liquid-orderedphases (37).

Characterizing Solvents Within the SC. We investigate interactionsbetween SC and 10 different solvents (Table 1) that are widelyused in skin products. NMR spectra obtained for SC with addedsolvents are shown in Figs. 3 and 4 and SI Appendix, Figs. S2–S4.As reference, we compare with dry SC (Fig. 1C). In cases wherethe solvent molecules contain one or more C–H bonds, the PTssNMR experiments provide detailed information on moleculardynamics of the solvent molecules. In neat solutions, the molec-ular motion of the solvents is isotropic and fast. When the solventmolecules are present within the SC they may interact with lipidand protein components, which can alter the molecular dynamicsof the solvent molecules.The comparison between results obtained for the different

systems (Table 1) reveals that all polar solvents are strongly af-fected by being incorporated into SC. At low concentrations, one

or several segments of the solvent molecules are obstructed withτc ∼ 0.1 μs, as inferred from the broad CP signals accompaniedby DP signals and no INEPT intensity at their resonances (Figs.3A and 4A and SI Appendix, Fig. S2 A, C, and E). From thesedata, it is shown that the correlation time for the solvents in SC isseveral orders of magnitudes lower compared with solvents inneat solutions [for comparison, τc ∼ 0.5 ps, 4 ps, and 0.6 ns inneat solutions of methanol, DMSO, and glycerol, respectively, at25 °C (38–40)]. The reduction in molecular mobility is a strongindication of interaction between the polar solvent moleculesand SC molecular components. When the amount of solvent inSC is increased (Fig. 3 B and C and SI Appendix, Fig. S2 B, D,and F), the motion of the polar solvents becomes faster, as im-plied from the sharper CP signals. At the highest solvent con-centrations investigated we also observe a coexisting fraction ofsolvent in the sample that has the characteristic of being fastisotropic (Fig. 3 B and C and SI Appendix, Fig. S2 B, D, and F).Similar behavior as shown for the polar solvents is also seen forthe slightly more apolar 2-propanol (Fig. 4B and SI Appendix,Fig. S3) in SC.The apolar solvents are less affected than the polar ones by

being incorporated into SC. As summarized in Table 1, at least afraction of the apolar solvents show the characteristics of beingfast isotropic also at the lowest concentrations investigated

Dry SC

Lipids

Keratin filament

+ 20 wt% waterB

+ + + +

A

+ more water

(up to 40 wt%)

Phase coexistence

+ solvent at low concentrations

C(except water)

Polar solvents

Apolar hydrocarbon solvents

OMTS OMTS

+ moresolvent

up to 40 wt %

/

+

+

faster

motion

Obstructed

Polar solvents

Apolar hydrocarbon solvents

+

SC+ 40 wt % water

D+ 5 wt%

glycerol / PG

Glycerol / PG

Dry SC

ERH 84 %

1. soaked in solvent2. dried and remove solvent3. RH 84%

Fig. 2. Schematic summary of all results obtained at 32 °C. (A) Interpretation of rotational correlation time (τc) and C–H bond order parameter (jSCHj) ofsolvents inside SC as obtained from the intensity and the lineshape of CP (blue) and INEPT (red) signals in PT ssNMR experiments. The shapes of the symbolsillustrate the anisotropy, isotropic (circle) and anisotropic (ellipsoid), and the colors represent the dynamical regimes: τc < 10 ns (red), τc ∼ 0.1 μs (purple), and τc >0.1 μs (blue). The double arrows illustrate a range of anisotropy between the symbols. †The broad CP signal is assigned to a motion with τc ∼ 0.1 μs when it isaccompanied by a DP signal. (B–E) Schematic illustration of the molecular dynamics of solvents in SC, as well as of SC lipids and proteins for different ex-perimental situations, including addition of water to dry SC (B), addition of other solvents to dry SC (C), addition of glycerol or PG to the hydrated SC (D), andSC after washing in excess solvent and thereafter hydrated at RH 84% (E). Blue represents rigid SC molecular segments [as seen from the (blue) CP spectra]and red represents mobile SC molecular segments [as seen from the (red) INEPT spectra].

Pham et al. PNAS | Published online December 27, 2016 | E115

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

(Table 1). For the hydrocarbon solvents (Figs. 3 D and E and 4Cand SI Appendix, Fig. S4A) we detect two coexisting populationsof solvent molecules at all solvent concentrations. One populationof the solvent exhibits fast isotropic motion, and the other pop-ulation shows fast anisotropic motion (Fig. 2C and Table 1). Finally,the apolar siloxane solvent molecules (OMTS) only show fast iso-tropic motion when added to SC (SI Appendix, Fig. S4B), similar toa neat isotropic solution of OMTS. From the present analysis, weare not able to pinpoint the location of fast isotropic solvent mol-ecules, and we cannot distinguish between cases when the solvent ispresent in isotropic fluid domains within SC (11, 41–44) and when itis present in an excess solution. However, we combine the studies ofthe molecular dynamics of the solvent molecules with an analysis ofhow the added solvent influences the molecular dynamics in SClipid and protein components. Based on the combination of in-formation on molecular dynamics in the solvent as well as in SClipid and protein components we are able to draw conclusions onthe uptake of solvents into the SC.

PT ssNMR to Study Molecular Effects on SC Molecular Components. Inprevious studies, we managed to assign the chemical shifts of allmajor peaks in the crowded 13C spectra to carbons in lipids andamino acids in intact SC (11). The assignment was based onsamples of intact SC together with reference systems of isolatedSC lipids and corneocytes as well as simplified model systems. Inthe analysis of the spectra, we here focus on representative SClipid and protein molecular segments (SI Appendix, Table S1), as

illustrated in Fig. 1E. The Gly Cα, Ser Cα, and Ser Cβ peaks areused as signature peaks for the terminal domains of the keratinfilaments that are rich in glycine and serine (45) (UniProt IDcodes P04264 and P13645). The core of the keratin filament isenriched in leucine and lysine (UniProt ID codes P04264 andP13645), which are examined at the resonances of Leu Cβand Lys Ce. For the SC lipids, the acyl chains of fatty acids andceramides can be probed at the resonances of the terminalmethyl/methylene carbons [ωCH3, (ω-1)CH2, and (ω-2)CH2] andof αCH2. In addition, the majority of the lipid acyl chain (CH2)nresonates within a range of 30–34 ppm, in which the all-transconformation (AT) predominating in solid phases is visualized at33.4 ppm in the CP spectrum, whereas the liquid-like distributionof trans/gauche conformations (TG) are probed at 31 ppm in theINEPT spectrum (46). Different segments of cholesterol (CHOL4, 9, 10, 20, 22, and 24) and ceramide headgroup (Cer C1 andC2) can also be distinguished. The strategy when analyzing thecrowded spectra from SC is to look for consistency in the ob-served changes. In other words, we draw conclusions aboutchanges in molecular mobility for a certain molecule if we ob-serve changes in more than one segment in the same molecule(if applicable).A general conclusion from the combination of data presented

here is that all spectra obtained for SC with added solvents at32 °C (Figs. 1 C–E and 3–5 and SI Appendix, Figs. S2–S7) aredominated by the CP signal for most of the spectral range. Thisimplies that the majority of SC lipid and protein components are

(13C) / ppm1015202530354045505560657075

ωC

H3

(ω-1

)CH

2

TG

AT

(ω-2

)CH

2

αC

H2

(CH2)n

OH

CHH3C CH3

12

1

2

*

*

2

SC + 20 wt% 2-propanol SC + 20 wt% PG SC + 20 wt% 1-decanol

(13C) / ppm1015202530354045505560657075

CH

OL

12,2

4

CH

OL

4

CH

OL

10,2

0,22

2 5* * * 3

123456

1 * * * *H2CH3C H2C CH2( )CH2 4 CH2 CH2 OH

7

6 74

(13C) / ppm1015202530354045505560657075

OH

CH2CHCH3

OH*1,2

*33 12

A B C

Fig. 4. Different solvents in SC at the same concentration of 20 wt % solvent. 13C MAS NMR spectra (DP, gray; CP, blue; and INEPT, red) of the SC at 20 wt %PG (A), 2-propanol (B), and 1-decanol (C) at 32 °C. Resonances from solvent carbons are marked with asterisks.

A

D E F

B C

Fig. 3. Concentration dependence of different solvents in SC. 13C MAS NMR spectra (DP, gray; CP, blue; and INEPT, red) of the SC with 10 (A), 20 (B), and 40(C) wt % glycerol, 5 (D) and 20 (E) wt % hexadecane, and 20 wt % (F) deuterated hexadecane at 32 °C. Resonances from solvent carbons are marked withasterisks.

E116 | www.pnas.org/cgi/doi/10.1073/pnas.1608739114 Pham et al.

rigid under all conditions investigated. The solid lipids are pro-bed at the sharp and prominent CP (CH2)n AT peak (46) and therigid Cα resonances from all amino acid residues (except glycineCα) are observed as the large and broad CP peak around 57 ppm.

Changes in SC Molecular Dynamics in the Presence of Solvents. Weinvestigated how various solvents influence the molecular dy-namics in different segments of SC lipid and protein compo-nents, and the findings are summarized in Fig. 2 B and C andTable 1. In dry SC, the keratin filaments are completely rigid, asapparent from the total absence of INEPT signal from proteincarbons (Fig. 1C). For the lipid components, however, there is aminute fraction that is mobile, which gives rise to the low-amplitudeINEPT peaks of the terminal methyl carbons (ωCH3) and thetrans/gauche conformation (CH2)n TG (Fig. 1C). Upon hydra-tion, the fraction of mobile SC lipids increases (Fig. 1 D and E).The protein components remain solid up to water content of atleast 24 wt % (Fig. 1D) (11), whereas at higher water contents(40 wt %, Fig. 1E) the amino acids in the terminal segments ofthe keratin filaments become mobile.We then replace water with another solvent. For all systems

investigated, it is shown that the addition of solvent leads toincreased mobility of the SC lipid acyl chains. The magnitude ofthis fluidizing effect varies with the identity and the concentra-tion of the solvents. A comparison between different solvents atthe same concentration shows the strongest effect for the apolarsolvents (1-decanol, hexadecane, and siloxane), as well as for2-propanol (Table 1). Among the polar solvents investigated, onlywater shows detectable effects on cholesterol and ceramideheadgroups at very high concentration (40 wt % in SC, Fig. 1E),whereas the apolar solvent 1-decanol influences the molecularsegments of cholesterol already at lower concentration (20 wt % inSC, Fig. 4C). It is notable that the protein components remainrigid for all solvents investigated at concentrations up to 40 wt %in SC, with water as the only exception (Fig. 1E).

Molecular Mobility of Polar Solvents Within Intact SC. One strikingobservation from the present data is that the polar solventmolecules themselves seem obstructed when present inside SC atlow concentrations (Table 1, Figs. 3A and 4A, and SI Appendix,Fig. S2 A, C, and E). Further examination of the lineshape of thepeaks reveals more detailed information on the molecular dy-namics of these solvent molecules. The CP resonances of carbonsbound to hydroxyl group of glycerol (Fig. 3A), PG (Fig. 4A),methanol and ethanol (SI Appendix, Fig. S2 A and C), and thecarbon of the –CH3 group of DMSO (SI Appendix, Fig. S2E) arebroad and accompanied by DP resonances, indicating that thesecarbons are obstructed with τc ∼ 0.1 μs (Fig. 2A). However, the–CH3 groups of PG and ethanol have faster motion as seen bytheir sharp CP peaks. Taken together, the obstructed character ofthe solvents strongly implies interactions between the solvent

molecules and SC components, likely involving the hydroxylgroups in the solvent molecules. At higher solvent concentrations,the average motion of the obstructed solvent molecules becomesfaster, and another coexisting fraction of solvent molecules withisotropic motion can be detected. Among the solvents in-vestigated, the lowest concentration limit where the coexistingisotropic fraction appears is shown for glycerol (Fig. 3 A–C).All polar solvents investigated induce mobility in the SC lipid

acyl chains as concluded from the increased INEPT intensity andthe concomitant decrease in the CP intensity of (CH2)n AT seg-ments, although the magnitude of this effect varies between thedifferent solvents. A minor increase in lipid mobility is observedwhen adding PG, methanol, ethanol, and glycerol to SC. Thefluidizing effect of all these solvents is smaller compared withwater at the same concentration of 20 wt %. However, the addi-tion of DMSO leads to a slightly more pronounced effect on lipidmobility compared with water. When the concentration of solventsincreases from 20 wt % to 40 wt %, the mobility of SC acyl chainsfurther increases (Figs. 3 A–C and 4A and SI Appendix, Fig. S2A–D and F).To further investigate the interaction between DMSO and

protein filaments in the corneocytes, we performed experimentswhere DMSO was added to isolated corneocytes at differentconcentrations (20 and 40 wt %) (SI Appendix, Fig. S5). Again,we confirm that the addition of DMSO has no effect on themobility in any amino acid segments of the keratin filament. Still,there is a dramatic effect on the DMSO mobility by being in-corporated into the corneocyte sample. Similar to the observa-tion when DMSO was added to intact SC, it is observed that thesolvent molecules show the characteristics of being obstructedwhen present in the isolated corneocytes at low concentrations.At higher DMSO concentrations, the DMSO molecule fractiongains faster motion and coexists with a fraction of isotropic fluidsolvent in the corneocyte sample.

Molecular Mobility of Apolar Solvents Within Intact SC. Among the10 solvents investigated, the apolar ones are clearly more mobilecompared with the polar ones when present in SC (Table 1). Thedifference is most pronounced at the lowest concentrations andthe apolar hydrocarbon solvents give rise to both sharp CP andstrong INEPT signals (Figs. 3 D and E and 4C and SI Appendix,Fig. S4A) even when present at 5 wt % in SC (Fig. 3D and SIAppendix, Fig. S4A). One major complication in the analysis ofthe NMR data for the hydrocarbon solvents is that most of theresonances from the solvent hydrocarbon chains overlap with thepeaks from the SC lipid acyl chains. For 1-decanol, the carbonatom bound to hydroxyl group (C1) can be uniquely identified at62.5 ppm, and based on the intensities of this peak (Fig. 4C) weconclude coexistence of two populations of 1-decanol that differin their molecular mobility: one fast isotropic fraction and oneslower/anisotropic fraction. To enable detailed characterization

(13C) / ppm1015202530354045505560657075

CH

3

(-1

)CH

2

TG

AT

(-2

)CH

2

CH

OL

4G

ly C

Cer

C1

Ser

CCer

C2

Ser

C

CH

OL

9 CH

2

(CH2)n

CH

OL

14, 1

7

(13C) / ppm1015202530354045505560657075

CH

3

(-1

)CH

2

TG

AT

(-2

)CH

2

CH

OL

4G

ly C

Cer

C1Ser

C

Cer

C2

Ser

C

CH

OL

9 CH

2

(CH2)n

CH

OL

14, 1

7

6263 6263 6263 5657 5657 5657 4344 4344 4344

Gly CSer CSer C

Wat

er

Wat

er +

PG

Wat

er +

gly

cero

l

Wat

er Wat

er +

PG

Wat

er +

gly

cero

l

Wat

er Wat

er +

PG

Wat

er +

gly

cero

l

(13C) / ppm

SC + 5 wt% glycerol + 40 wt% water SC + 5 wt% PG + 40 wt% water312 * ** **1 2

A B C

Fig. 5. 13C MAS NMR spectra (DP, gray; CP, blue; and INEPT, red) of the SC with 40 wt % water and added cosolvents 5 wt % PG (A) or glycerol (B) at 32 °C.Resonances from solvent carbons are marked with asterisks. The close-up figures (C) show the relation between INEPT, CP, and DP signals for amino acidsegments Ser Cβ, Ser Cα, and Gly Cα for different solvent conditions. The intensities in C are normalized against the CP AT signal. The lipid CP AT signal ishigher upon adding these cosolvents (compare with Fig. 1E). The dramatic increase in the mobility of the proteins leads to a decrease in the broad CPbackground formed by the rigid proteins as seen in SI Appendix, Fig. S5.

Pham et al. PNAS | Published online December 27, 2016 | E117

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

of hexadecane in SC, we performed additional experiments usingthe deuterated version of the same solvent. The uniformly2H-labeled hexadecane (Fig. 3F) does not contribute to 13Cspectra acquired with 1H→13C polarization transfer. A compari-son of the spectra from SC with hydrogenated and deuteratedhexadecane (Fig. 3 E and F) shows that the sharp CP signals at theoverlapping resonances (Fig. 3E) indeed originate from the hex-adecane molecules. Based on these data, we conclude that hex-adecane behaves similar to 1-decanol when added to intact SC,and that it is present as two populations.A comparison between all solvents investigated (Table 1)

shows that the apolar solvents and 2-propanol are more efficientin melting of SC lipids compared with the polar solvents, in-cluding water (Figs. 3 E and F and 4 B and C and SI Appendix,Fig. S4B) at the same concentration. For the hydrocarbon sol-vents, the effects on the lipids are indicated by the decrease ofthe AT peak in the CP spectra, compared with the dry SC sample(Fig. 1C). The effect is most clear for 1-decanol, which also leadsto increased mobility of cholesterol in the SC lipid matrix. Theexperiment in which we use deuterated hexadecane (Fig. 3F)makes it possible to resolve the effects of the solvent on SC lipidchain mobility, and it was clearly shown that this solvent hasstrong fluidizing effect on the SC lipids. The apolar siloxane sol-vent OMTS also fluidizes SC lipids, as is obvious from the INEPTspectra, although the effect is weaker compared with 2-propanoland the apolar hydrocarbon solvents.

Comparison Between PG and Glycerol in Hydrated Skin.Glycerol andPG are small polar molecules with similar structure motifs thatare commonly used in skin formulations. As described above,both solvents induce mobility in SC lipids but have no effect onSC proteins in the absence of water (Table 1). To get a moredetailed understanding of how these polar compounds influenceSC in varying hydration conditions, they were also added to SCas cosolvents (5 wt %) together with water (40 wt %) (Fig. 5).Both PG and glycerol show the characteristics of being fast andisotropic when added to hydrated SC, because their resonancesare only seen in the DP and INEPT spectra. This observation isin clear contrast to the situation when the same solvents wereadded to dry SC (Figs. 3 A–C and 4A and SI Appendix, Fig. S2F).The most prominent effect of PG and glycerol in hydrated skinis that they induce mobility in the keratin filaments, which isrevealed by a close inspection of the spectral region of, for ex-ample, Ser and Gly (Fig. 5C). In particular, PG has a strong flu-idizing effect in the terminal segments in the protein filaments(Fig. 5C). With respect to the SC lipid fraction in the hydrated SC,glycerol and PG show no or minor effects on the lipid acyl chainmobility, although they both cause increased mobility of choles-terol and ceramide headgroups. In summary, the comparisonsbetween experiments in Fig. 5 and Figs. 3 A–C and 4A and SIAppendix, Fig. S2F show that the effects of glycerol and PG areclearly different when added to dry and hydrated SC.

Molecular Mobility of SC Lipid and Protein Components After Washingin Excess Solvents.Washing of skin in excess solvent likely leads toincorporation of solvent molecules into SC. It may also causeextraction of SC components to an excess solvent solution. Torelate to practical situations where skin is exposed to solvents, weinvestigated changes in SC molecular mobility in samples thathad been washed in excess solutions of methanol, ethanol, orhexane (Fig. 2E and SI Appendix, Fig. S6). After the washingstep, the solvent was evaporated, and thereafter the SC sampleswere equilibrated at 84% relative humidity (RH). As a referencesample, we use SC equilibrated at the same RH without priorwashing in solvent (SI Appendix, Fig. S6A).The main conclusion from these experiments is that washing

SC in solvent strongly influences the SC INEPT spectra, dem-onstrating decreased SC molecular dynamics in all cases (Fig.

2E). For the reference SC sample at 84% RH, there is a minorfraction of SC lipid acyl chains that are fluid, whereas for the SCsamples that had been washed in excess solvent and equilibratedat the same RH no fluid components were detected (SI Appendix,Fig. S6 A–D). In addition, washing in excess ethanol or methanolalso leads to a reduction in the signal from the rigid SC lipid acylchains (SI Appendix, Fig. S6 B and C), implying that also solidlipids are lost in the washing step. For the sample that was washedin hexane (SI Appendix, Fig. S6D) we observe an increased in-tensity in the CP spectra at resonances corresponding to the ATlipid acyl chains, indicating reduced mobility in this fraction. Noeffects on SC protein components are detected after washing inany of these solvents.

DiscussionTracking Solvents in a Complex SC Matrix. We use PT ssNMR toinvestigate the molecular mobility of solvent molecules withinthe complex biological material of SC. Simultaneously, we studychanges in molecular dynamics of the SC lipids and proteinscaused by the presence of the solvent. This combination of in-formation enables deepened understanding of molecular inter-actions between external chemicals and the outer layer of theskin. The PT ssNMR method is highly sensitive for observingsmall changes in the minor fluid fraction (τc < 10 ns) of SCcomponents, and it simultaneously detects solid components.Thanks to the high sensitivity to variations in molecular dynamicswith close to atomic resolution, the method is a powerful tool tostudy fluidity in intact SC and to complement the extensive lit-erature that is mainly focused on characterization of solid SCcomponents (2, 3, 5). The method is not quantitative in the sensethat it does not provide a value on the fraction of mobile to rigid13C for a certain segment, mainly due to the nonlinear responsein INEPT and CP signals to changes in molecular dynamics andanisotropy (SI Appendix, Fig. S1), but also because of the diffi-culties of resolving overlapping peaks in the crowded spectrafrom intact SC.A variety of solvent molecules differing in size, hydrophobicity,

and chemical structure (Table 1) were used, and it is concludedthat all of them are partly or completely incorporated in SC. Theexperiments reveal interactions between solvent molecules andSC components, as demonstrated by the reduced molecularmobility of the solvent molecules when present within SC. Themost prominent effects are shown for polar solvents, which ap-pear obstructed when incorporated into SC even when present inrather high amounts (Table 1). From the present data it is notpossible to draw any conclusions about the molecular mobility inwater inside SC because it is not visible in the 13C spectra. Polarmolecules such as water, ethanol, methanol, PG, DMSO, andglycerol as well as 2-propanol can be present in narrow inter-lamellar layers [measured to ca. 2 Å at 60 wt % water (10)] in theextracellular lipid matrix, as well as in between the solid keratinrods inside the corneocytes. The polar solvents are then expectedto interact with the ceramide lipid headgroups or polar aminoacid side chains, for example, serine. Confinement and interac-tions with polar groups can explain the obstructed character ofthese small polar molecules in SC. The apolar hydrocarbon sol-vents also show reduced mobility compared with isotropic solu-tions. The most likely explanation for this is that the apolarhydrocarbon solvents partition into the hydrophobic layers of theSC lipid lamellae and fluidize the lipid hydrocarbon chains, whichis also consistent with previous studies on 1-decanol in modelsystems and human skin (4, 47). Similar behavior has previouslybeen suggested for, for example, isopropyl myristate in SC (3).The apolar hydrocarbon solvent molecules in the lipid layer showfast anisotropic reorientation rather than obstructed motion.At higher solvent concentrations a coexisting fraction of isotropic

solvent is detected together with the obstructed/fast anisotropicpolar and apolar solvent molecules (Table 1). The concentration

E118 | www.pnas.org/cgi/doi/10.1073/pnas.1608739114 Pham et al.

where the coexisting isotropic fraction occurs varies between thedifferent solvents, and it is generally lower for the apolar solventscompared with the polar ones. Adding more polar solvents to SCgradually leads to swelling of the corneocytes, and the solvent dy-namics approaches that of the neat liquid. The addition of apolarsolvents leads to an increased amount of fluid SC lipids, whichalso shows the characteristics of an isotropic fluid (11, 41–43). Athigh concentrations SC will be saturated with the solvent. Fromthe present data we cannot determine the location of the iso-tropic solvent. However, it is notable that increasing the amountof isotropic solvent in most cases is also associated with an in-creased fluidity in SC components (Table 1), clearly indicatingthe presence of solvents within the SC matrix. We note that themolecular dynamics of the polar solvents and 2-propanol in SCincreases with increasing concentration and increased fluidity inSC lipids. This indicates not only a higher fraction of fluid SClipids but also a faster dynamics of these fluid domains where thesolvents are located. It is finally noted that the solvent of in-termediate polarity, 2-propanol, shows characteristics in betweenthe polar and apolar solvents in terms of its molecular mobilitywhen incorporated into SC and the concentration where thecoexisting isotropic fraction of solvent is detected.

Water Is Critical for SC Protein Mobility. Water is the only solventamong those investigated that shows any detectable effect on theSC protein components when added to dry SC (Fig. 1 C and E).The other polar solvents do not affect SC protein mobility unlessthey are added together with water. Still, we expect the polarsolvents to be present between the solid keratin rods inside thecorneocytes. It was confirmed that relatively high amounts ofDMSO are taken up in isolated corneocytes and intact SC (SIAppendix, Figs. S2E and S5). The DMSO molecules then havecharacteristics of being obstructed, which can be explained byconfinement in polar microdomains in the solid keratin filamentstructures. Previous infrared microspectroscopic imaging studieshave also demonstrated colocalization of DMSO and PG withproteins in intact porcine skin (48). Similar behavior is also seenfor the polar solvents ethanol, methanol, PG, and glycerol whenincorporated into SC.The outstanding effect of water on the terminal segments in

the keratin filaments can be related to the fact that water is abetter solvent for these polar amino acids than any of the otherstudied solvents. Previous studies have shown that the interac-tions between, for example, DMSO and the polar unchargedamino acid glycine is highly unfavorable, which correlates withprecipitation of proteins by DMSO (49). Methyl, ethyl, andpropyl alcohols are also often used to precipitate water-solubleproteins (50), which can be explained by reduced solubility due tothe lower dielectric constant of the medium. In hydrated condi-tions the keratin filaments can be seen as solid rods covered bymobile terminal segments enriched in hydrophilic unchargedamino acids (UniProt ID codes P04264 and P13645) (11). Theincreased mobility in the terminal segments gives rise to long-range steric entropic repulsive forces between the filaments,which may explain the strong swelling capacity of corneocytes athigh water contents (41). This behavior is analogous to polymerbrushes that can expand or collapse depending on interactionswith the solvent (51). Because the swelling capacity of the cor-neocytes can be related to SC flexibility and strength (17), onecan also expect these mechanical properties of the skin to changewhen water is replaced by another solvent. It has indeed beenshown that the ability of corneocytes to take up water decreaseswith increasing concentration of ethanol, and that the uptake ofneat ethanol is negligible compared with that of neat water (52).

PG and Glycerol Show Different Effects on SC Lipids and ProteinsWhen Added As Cosolvents Together with Water. PG and glycerolare small polar molecules with two or three hydroxyl groups,

respectively. They are widely used in skin-care products as sol-vents, penetration enhancers, and humectants (6). The presentdata imply that the interactions of these compounds with SCcomponents clearly depend on the water content in SC (Fig. 5),which likely also has an impact on their penetration-enhancingeffects (20). A larger number of SC molecular segments are af-fected when PG or glycerol is added to SC as a cosolvent to-gether with water compared with when they are added to the drySC. Similar response to changes in SC hydration has been alsopreviously shown for other polar and apolar chemical penetra-tion enhancers (13). One remarkable observation from the datain Fig. 5 is that PG has a stronger effect on inducing mobility inthe amino acids in the terminal domains of keratin filamentcompared with a large number of other polar and apolar com-pounds investigated with this method (13, 14). The addition ofwater to SC leads to increased mobility in both lipid and proteindomains (Fig. 1 C–E). The added cosolvent can therefore moreeasily dissolve in the SC matrix and further fluidize these struc-tures. Previous studies have shown disturbance of lipid chainpacking (3, 53) and a decrease in lipid transition and keratindenaturation temperatures in hydrated SC pretreated withglycerol and PG (3). The effects on the thermal transitions havebeen shown to be more prominent in skin treated with PGcompared with glycerol (3, 5).

Extraction of Mobile Lipids from SC Due to Exposure to ExcessSolvent. The experimental condition where SC is soaked in ex-cess solutions of neat solvents is relevant to practical situationswith accidental and occupational exposure, for example, handwashing with ethanol. The solvents investigated are also com-monly used to dissolve hydrophobic active compounds in diffu-sion-cell in vitro studies of (trans)dermal delivery, where the skinis typically exposed to the excess solvent for several hours (13).The present experiments show that the exposure to excess amountof methanol and ethanol leads to extraction of lipids from SC (SIAppendix, Fig. S6). In particular, the amount of fluid lipids in SC isreduced by the washing treatment. Previous infrared studies alsogive evidence for SC lipid extraction after being soaked in meth-anol, ethanol (1, 7, 54–56), and n-hexane (7). Extraction of lipidcomponents with lower melting points may also shift the balancebetween the solid and fluid lipid phases in the remaining SC lipids,which causes stabilization of solid lipid structures. Such a shift ofthe phase balance might explain the observation of reduced lipidmobility in SC that was washed in hexane.

Implication for the Use of Solvent in Skin Applications. In this studyit is illustrated how different types of solvents influence SCmolecular components. All solvents investigated induce mobilityin SC components. The changes in the proportion between fluidSC and solid lipids and proteins can have a large impact on themacroscopic properties of the SC, including transport barrierfunction, flexibility, and softness (2, 6, 13, 16, 17). The observedeffects vary with the polarity of the solvent, its concentration, andSC hydration conditions.The findings can be directly related to practical situations where

solvents are used in pharmaceutical, cosmetic, and cleansing skinformulations. When the skin is exposed to a formulation the SCmolecular and macroscopic properties may be altered due to in-teractions between SC components and solvents, active sub-stances, humectants, penetration enhancers, and other excipients(12, 13, 20). The effects of the formulation also depend on skinhydration, the state of the skin, and the occurrence of skin diseases(13, 57–60). The exposure time also varies between different ap-plications in, for example, dermal patches, creams, salves, andwashing formulations. During the application period, the amountof solvent in the formulation will change due to evaporation andskin absorption for small molecules such as those studied here [thelimit for SC penetration has been reported to be 500 Da (61)]. The

Pham et al. PNAS | Published online December 27, 2016 | E119

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US

penetration of the solvent molecules into the skin will depend onthe skin’s permeability, which is in turn altered by molecular in-teractions between the diffusing molecules and SC components(13). For all systems investigated here the solvents induce fluid SClipid structures, and are therefore expected to act as penetrationenhancers. In most practical situations the distribution of foreignmolecules in SC is not considered to be uniform. Upon exposureto a formulation the thermodynamic activity of the foreign mol-ecules will be highest in the upper layer of SC, which is generallyalso the layer with the lowest water activity, which motivates theinvestigations of solvents in dry SC. The herein observed effects inhomogenous samples of SC and neat solvent can be used to makepredictions of the effects due to penetration of molecules fromcomplex formulations.In many experimental studies of skin permeation and molecular

organization, samples of skin or isolated SC are washed in solvents,and compounds or probes of interests are added to SC from asolvent solution (62). In the analysis of experimental data and in theevaluation of the effects of skin formulations it is crucial to distin-guish the effects caused by the solvents and by other components.This analysis may be complicated by synergistic effects due to in-teractions between solvents and other molecules (5, 63). For mostapplications, the solvent molecules are generally in large excess toother components in the formulation, and the effects observed inSC–neat solvent systems are likely also present in situation whenskin is treated in more complex solutions. In applications where theSC is exposed to excess solution of solvents one also needs toconsider the possibility that some lipid components are extractedfrom the SC as clearly demonstrated herein, which may also alterSC integrity and consequently SC permeability (1, 7).

Materials and MethodsMaterials. DMSO, 1-decanol, OMTS, NaCl, Na2HPO40.2H2O, and KH2PO4 werepurchased from Sigma-Aldrich. Methanol, ethanol, 2-propanol, and hexanewere obtained from Merck, PG was from Acros Organics, and hexadecanewas from Alfa Aesar. Deuterated hexadecane was purchased from LarodanFine Chemicals AB. PBS contained 130.9 mM NaCl, 5.1 mM Na2HPO4, and1.5 mM KH2PO4, pH 7.4. All solutions were prepared using Milli-Q water.

Preparation of Dermal Skin and SC. Porcine ears were obtained from a localabattoir and stored at –80 °C until use. Hair was removed by a trimmer, andthe skin from the inner ear was dermatomed (TCM 3000 BL; Nouvag) to athickness of ∼500 μm. To separate SC from tissue, the dermatomed skinstrips were placed on filter paper soaked in PBS solution with 0.2 wt %trypsin at 4 °C overnight. Sheets of SC were removed by forceps, washedwith PBS five times, and further dried under vacuum. The dry SC sheets werethen pulverized with the use of a mortar and pestle to facilitate the mixingand equilibration. Pulverized SC was dried again in vacuum and stored in afreezer until further use. To reduce complications related to the biologicalvariation between the different individuals, batches consisting of pulverizedSC from several individuals (∼15) were prepared, and all internal compari-sons were performed with samples from the same batches. Two differentbatches were used for the present experiments, and control experimentswere performed with both batches in same conditions to confirm consis-tency between the results (SI Appendix, Fig. S7). Corneocytes were prepared

by extracting lipids from SC as described in ref. 12. The NMR experiments fordifferent selected samples with the same composition were reproduced.

To study the effects of pure solvents on SC, ∼30 mg of dry SC powder wasmixed with different amount of solvent (5, 10, 15, 20, and 40 wt %, based onthe total weight of SC and solvent). For the effects of solvent on SC at hy-drated conditions, ∼25 mg of dry SC powder was mixed with 5 wt % solvent(based on the total weight of SC and solvent) and 40 wt % water (based onthe total weight of SC and water). These samples were then transferred intotight ssNMR inserts and incubated at 32 °C for 1–2 d before the NMRmeasurements. We have not observed significant difference between thesamples incubated for 1 and 2 d. The study design aims at controlled con-ditions in terms of hydration and concentration, and not to mimic a certainpractical situation.

To examine the effects of excess exposure to solvents, ∼30 mg of dry SCpowder was immersed and stirred in 20 mL of solvents for 2 h. The SC powderwas then filtered and washed with the pure solvent. The solvents were re-moved under vacuum and the SC samples were then incubated at RH 84% for2 d at 32 °C and then transferred into tight ssNMR inserts.

In a previous study we showed that there is no detectable differences inthe PT ssNMR spectra between pulverized SC and intact SC sheet in samehydration conditions (11), indicating that these samples do not differ on themolecular scale. It was here also confirmed that there is no significant dif-ferences in the PT ssNMR spectra between pulverized SC and intact sheets ofSC treated with methanol (SI Appendix, Fig. S6 B and E).

Solid-State NMR Experiments. All NMR experiments were performed on aBruker Avance II 500-MHz NMR spectrometer equipped with a Bruker Efree4-mmMAS probe andwere carried out at a spinning frequency of 5 kHz. In PTssNMR experiments, the following setup was used: spectral width of 248.5ppm, acquisition time of 0.05 s, 2,048 scans per experiment, recycle delay of5 s, and 1H and 13C hard pulse at ω1

H/C/2π = 80.6 kHz. The power of 1H wasramped up from 72 to 88 kHz during the contact time of 1 ms in the CPexperiment. The INEPT experiments were performed with delay times τ of1.8 ms and τ′ of 1.2 ms. All experiments were recorded under 68-kHz two-pulse phase modulation 1H decoupling (64). The 13C spectra were externallyreferenced to the methylene signal of solid α-glycine at 43.7 ppm (65). Thetemperature was calibrated by methanol (66). The data were processed witha line broadening of 20 Hz, zero-filling from 1,024 to 8,192 time-domainpoints, Fourier transform, and phase and baseline correction by in-houseMATLAB code partially from matNMR (67).

To estimate the detection limit of a fast isotropic segment, the signal-to-noise ratio of the INEPT peak from carbon C3 of PG in a sample of SC with 5 wt% PG and 40 wt % water (Fig. 5A) was calculated to be ∼100 based on theratio of the intensity of the INEPT peak and the SD of the intensities of thenoise in the INEPT spectrum. Assuming that the signal is directly proportionalto the concentration and the noise remains constant, the detection limit cor-responding to the concentration where the signal-to-noise is above the value 3is ∼0.15 wt % PG. If the amount of PG in the NMR insert is 1.32 mg, corre-sponding to 25 mg SC powder, the detection limit of the fast isotropic seg-ment of PG is about 2 μg.

ACKNOWLEDGMENTS. We thank Jenny Andersson for providing the iso-lated corneocytes and Håkan Wennerström and Olle Söderman for fruitfuldiscussions. This work was supported by the Swedish Research Councilthrough regular grants and the Linnaeus Center of Excellence “Organizingmolecular matter” (E.S. and D.T.), the Swedish Foundation for Strategic Re-search (E.S.), the Crafoord Foundation (E.S.), and the Royal PhysiographicSociety of Lund (Q.D.P.).

1. Levang AK, Zhao K, Singh J (1999) Effect of ethanol/propylene glycol on the in vitro

percutaneous absorption of aspirin, biophysical changes and macroscopic barrier

properties of the skin. Int J Pharm 181(2):255–263.2. Brinkmann I, Müller-Goymann CC (2003) Role of isopropyl myristate, isopropyl alcohol

and a combination of both in hydrocortisone permeation across the human stratum

corneum. Skin Pharmacol Appl Skin Physiol 16(6):393–404.3. Brinkmann I, Müller-Goymann CC (2005) An attempt to clarify the influence of

glycerol, propylene glycol, isopropyl myristate and a combination of propylene glycol

and isopropyl myristate on human stratum corneum. Pharmazie 60(3):215–220.4. Dias M, Naik A, Guy RH, Hadgraft J, Lane ME (2008) In vivo infrared spectroscopy

studies of alkanol effects on human skin. Eur J Pharm Biopharm 69(3):1171–1175.5. Cornwell PA, et al. (1996) Modes of action of terpene penetration enhancers in human

skin differential scanning calorimetry, small-angle X-ray diffraction and enhancer uptake

studies. Int J Pharm 127(1):9–26.6. Williams AC, Barry BW (2004) Penetration enhancers. Adv Drug Deliv Rev 56(5):

603–618.

7. Goldsmith LB, Friberg SE, Wahlberg JE (1988) The effect of solvent extraction on thelipids of the stratum corneum in relation to observed immediate whitening of theskin. Contact Dermat 19(5):348–350.

8. Elias PM, Friend DS (1975) The permeability barrier in mammalian epidermis. J CellBiol 65(1):180–191.

9. Scheuplein RJ (1976) Permeability of skin—Review of major concepts and some newdevelopments. J Invest Dermatol 67(5):672–676.

10. Bouwstra JA, Gooris GS, Bras W, Downing DT (1995) Lipid organization in pig stratumcorneum. J Lipid Res 36(4):685–695.

11. Björklund S, Nowacka A, Bouwstra JA, Sparr E, Topgaard D (2013) Characterization ofstratum corneum molecular dynamics by natural-abundance 13C solid-state NMR.PLoS One 8(4):e61889.

12. Björklund S, et al. (2014) Stratum corneum molecular mobility in the presence ofnatural moisturizers. Soft Matter 10(25):4535–4546.

13. Pham QD, Björklund S, Engblom J, Topgaard D, Sparr E (2016) Chemical penetrationenhancers in stratum corneum - Relation between molecular effects and barrierfunction. J Control Release 232:175–187.

E120 | www.pnas.org/cgi/doi/10.1073/pnas.1608739114 Pham et al.

14. Björklund S, et al. (2016) The effects of polar excipients transcutol and dexpanthenolon molecular mobility, permeability, and electrical impedance of the skin barrier.J Colloid Interface Sci 479:207–220.

15. Harding CR, Watkinson A, Rawlings AV, Scott IR (2000) Dry skin, moisturization andcorneodesmolysis. Int J Cosmet Sci 22(1):21–52.

16. Björklund S, Engblom J, Thuresson K, Sparr E (2010) A water gradient can be used toregulate drug transport across skin. J Control Release 143(2):191–200.

17. Alonso A, Meirelles NC, Yushmanov VE, Tabak M (1996) Water increases the fluidityof intercellular membranes of stratum corneum: Correlation with water permeability,elastic, and electrical resistance properties. J Invest Dermatol 106(5):1058–1063.

18. Xiang TX, Anderson BD (1998) Phase structures of binary lipid bilayers as revealed bypermeability of small molecules. Biochim Biophys Acta 1370(1):64–76.

19. Carruthers A, Melchior DL (1983) Study of the relationship between bilayer waterpermeability and bilayer physical state. Biochemistry 22(25):5797–5807.

20. Björklund S, Engblom J, Thuresson K, Sparr E (2013) Glycerol and urea can be used toincrease skin permeability in reduced hydration conditions. Eur J Pharm Sci 50(5):638–645.

21. Rawlings AV (2003) Trends in stratum corneum research and the management of dryskin conditions. Int J Cosmet Sci 25(1-2):63–95.

22. Potts RO, Francoeur ML (1990) Lipid biophysics of water loss through the skin. ProcNatl Acad Sci USA 87(10):3871–3873.

23. Bouwstra JA, et al. (1991) Thermodynamic and structural aspects of the skin barrier.J Control Release 15(3):209–220.

24. Nowacka A, Bongartz NA, Ollila OH, Nylander T, Topgaard D (2013) Signal intensitiesin 1H-13C CP and INEPT MAS NMR of liquid crystals. J Magn Reson 230:165–175.

25. Bouwstra JA, et al. (2003) Water distribution and related morphology in humanstratum corneum at different hydration levels. J Invest Dermatol 120(5):750–758.

26. Iwai I, et al. (2012) The human skin barrier is organized as stacked bilayers of fullyextended ceramides with cholesterol molecules associated with the ceramide sphin-goid moiety. J Invest Dermatol 132(9):2215–2225.

27. Schaefer H, Redelmeier TE (1996) Composition and structure of the stratum corneum.Skin Barrier: Principles of Percutaneous Absorption (Karger, Basel), pp 43–86.

28. Anigbogu ANC, et al. (1995) Fourier-transform Raman-spectroscopy of interactionsbetween the penetration enhancer dimethyl-sulfoxide and human stratum-corneum.Int J Pharm 125(2):265–282.

29. Caspers PJ, Lucassen GW, Carter EA, Bruining HA, Puppels GJ (2001) In vivo confocalRaman microspectroscopy of the skin: Noninvasive determination of molecular con-centration profiles. J Invest Dermatol 116(3):434–442.

30. Stoughton RB, Fritsch W (1964) Influence of dimethylsulfoxide (DMSO) on humanpercutaneous absorption. Arch Dermatol 90(5):512–517.

31. Pope DC, Oliver WT (1966) Dimethyl sulfoxide (DMSO). Can J Comp Med Vet Sci 30(1):3–8.

32. Pines A, Gibby MG, Waugh JS (1972) Proton‐enhanced nuclear induction spectros-copy. A method for high resolution NMR of dilute spins in solids. J Chem Phys 56(4):1776–1777.

33. Morris GA, Freeman R (1979) Enhancement of nuclear magnetic-resonance signals bypolarization transfer. J Am Chem Soc 101(3):760–762.

34. Rothwell WP, Waugh JS (1981) Transverse relaxation of dipolar coupled spin systemsunder rf irradiation—Detecting motions in solids. J Chem Phys 74(5):2721–2732.

35. Suwelack D, Rothwell WP, Waugh JS (1980) Slow molecular-motion detected in theNMR spectra of rotating solids. J Chem Phys 73(6):2559–2569.

36. Vanderhart DL, Earl WL, Garroway AN (1981) Resolution in 13C NMR of organic solidsusing high-power proton decoupling and magic-angle sample spinning. J MagnReson 44(2):361–401.

37. Ferreira TM, et al. (2013) Cholesterol and POPC segmental order parameters in lipidmembranes: Solid state 1H-13C NMR and MD simulation studies. Phys Chem ChemPhys 15(6):1976–1989.

38. Mussig S, Franck EU, Holz M (2000) 1H NMR relaxation of methanol CH3OD up to highpressures of 250 MPa and to 523 K. Z Phys Chem 214(7):957–968.

39. Wolfe M, Jonas J (1979) Reorientational motions in compressed viscous fluids—Selectively deuterated glycerol. J Chem Phys 71(8):3252–3262.

40. Liu HY, Mullerplathe F, Vangunsteren WF (1995) A force field for liquid dimethylsulfoxide and physical properties of liquid dimethyl sulfoxide calculated using mo-lecular dynamics simulation. J Am Chem Soc 117(15):4363–4366.

41. Silva CL, et al. (2007) Stratum corneum hydration: Phase transformations and mobilityin stratum corneum, extracted lipids and isolated corneocytes. Biochim Biophys Acta1768(11):2647–2659.

42. Chen X, et al. (2007) Fatty acids influence “solid” phase formation in models ofstratum corneum intercellular membranes. Langmuir 23(10):5548–5556.

43. Brief E, et al. (2009) Phase behavior of an equimolar mixture of N-palmitoyl-D-erythro-sphingosine, cholesterol, and palmitic acid, a mixture with optimized hydrophobicmatching. Langmuir 25(13):7523–7532.

44. Nowacka A, et al. (2012) Small polar molecules like glycerol and urea can preserve thefluidity of lipid bilayers under dry conditions. Soft Matter 8(5):1482–1491.

45. Candi E, Schmidt R, Melino G (2005) The cornified envelope: A model of cell death inthe skin. Nat Rev Mol Cell Biol 6(4):328–340.

46. Earl WL, VanderHart DL (1979) Observations in solid polyethylenes by carbon-13nuclear magnetic resonance with magic angle sample spinning. Macromolecules12(4):762–767.

47. Thewalt JL, Tulloch AP, Cushley RJ (1986) A deuterium NMR study of labelledn-alkanol anesthetics in a model membrane. Chem Phys Lipids 39(1-2):93–107.

48. Mendelsohn R, Chen HC, Rerek ME, Moore DJ (2003) Infrared microspectroscopicimaging maps the spatial distribution of exogenous molecules in skin. J Biomed Opt8(2):185–190.

49. Arakawa T, Kita Y, Timasheff SN (2007) Protein precipitation and denaturation bydimethyl sulfoxide. Biophys Chem 131(1-3):62–70.

50. Merrill MH, Fleisher MS (1932) Factors involved in the use of organic solvents asprecipitating and drying agents of immune sera. J Gen Physiol 16(2):243–256.

51. Israelachvili JN (2011) Steric (polymer-mediated) and thermal fluctuation forces.Intermolecular and Surface Forces (Academic, San Diego), 3rd Ed, pp 381–413.

52. Berner B, Juang RH, Mazzenga GC (1989) Ethanol and water sorption into stratumcorneum and model systems. J Pharm Sci 78(6):472–476.

53. Cornwell PA, Barry BW, Stoddart CP, Bouwstra JA (1994) Wide-angle X-ray diffractionof human stratum corneum: Effects of hydration and terpene enhancer treatment.J Pharm Pharmacol 46(12):938–950.

54. Bommannan D, Potts RO, Guy RH (1991) Examination of the effect of ethanol onhuman stratum corneum in vivo using infrared spectroscopy. J Control Release 16(3):299–304.

55. Goates CY, Knutson K (1994) Enhanced permeation of polar compounds throughhuman epidermis. I. Permeability and membrane structural changes in the presenceof short chain alcohols. Biochim Biophys Acta Biomembranes 1195(1):169–179.

56. Kai T, et al. (1990) Mechanism of percutaneous penetration enhancement: Effect ofn-alkanols on the permeability barrier of hairless mouse skin. J Control Release 12(2):103–112.

57. Lodén M (2003) Role of topical emollients and moisturizers in the treatment of dryskin barrier disorders. Am J Clin Dermatol 4(11):771–788.

58. Sparr E, Wennerström H (2001) Responding phospholipid membranes–interplay be-tween hydration and permeability. Biophys J 81(2):1014–1028.

59. Roger K, Liebi M, Heimdal J, Pham QD, Sparr E (2016) Controlling water evaporationthrough self-assembly. Proc Natl Acad Sci USA 113(37):10275–10280.

60. Sparr E, et al. (2012) Controlling the hydration of the skin though the application ofoccluding barrier creams. J R Soc Interface 10(80):20120788.

61. Bos JD, Meinardi MMHM (2000) The 500 Dalton rule for the skin penetration ofchemical compounds and drugs. Exp Dermatol 9(3):165–169.

62. Karande P, Jain A, Ergun K, Kispersky V, Mitragotri S (2005) Design principles ofchemical penetration enhancers for transdermal drug delivery. Proc Natl Acad Sci USA102(13):4688–4693.

63. Barry BW (1987) Mode of action of penetration enhancers in human skin. J ControlRelease 6(1):85–97.

64. Bennett AE, et al. (1995) Heteronuclear decoupling in rotating solids. J Chem Phys103(16):6951–6958.

65. Hayashi S, Hayamizu K (1991) Chemical-shift standards in high-resolution solid-stateNMR (1) C-13, Si-29 and H-1 nuclei. Bull Chem Soc Jpn 64(2):685–687.

66. Van Geet AL (1970) Calibration of methanol nuclear magnetic resonance thermom-eter at low temperature. Anal Chem 42(6):679–680.

67. van Beek JD (2007) matNMR: A flexible toolbox for processing, analyzing and visu-alizing magnetic resonance data in Matlab. J Magn Reson 187(1):19–26.

Pham et al. PNAS | Published online December 27, 2016 | E121

CHEM

ISTR

YBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

PNASPL

US