The stratum corneum comprises three layers with distinct metal-ion barrier properties

The stratum corneum (SC), the outermost barrier of mammalian bodies, consists of layers of cornified keratinocytes with intercellular spaces sealed with lipids. The insolubility of the SC has hampered in-depth analysis, and the SC has been considered a homogeneous barrier. Here, we applied time-of-flight secondary ion mass spectrometry to demonstrate that the SC consists of three layers with distinct properties. Arginine, a major component of filaggrin-derived natural moisturizing factors, was concentrated in the middle layer, suggesting that this layer functions in skin hydration. Topical application of metal ions revealed that the outer layer allowed their passive influx and efflux, while the middle and lower layers exhibited distinct barrier properties, depending on the metal tested. Notably, filaggrin deficiency abrogated the lower layer barrier, allowing specific metal ions to permeate viable layers. These findings elucidate the multi-layered barrier function of the SC and its defects in filaggrin-deficient atopic disease patients.


Results
TOF-SIMS imaging of freeze-dried murine tissue sections. We prepared sections of flash-frozen mouse tails and freeze-dried the sections without sample thawing to minimize delocalization of non-fixable molecules (see Methods). The sections were directly analyzed by TOF-SIMS to visualize the distribution of natural substances without applying a staining procedure and were subsequently processed for immunofluorescence (Fig. 1). The obtained fluorescent images were superimposed on the corresponding TOF-SIMS images in silico. Muscles were visualized as actin high by immunofluorescence and potassium (K) high sodium (Na) low in TOF-SIMS, and connective tissues were visualized as collagen I high in immunofluorescence and K low Na high in TOF-SIMS (Fig. 2a-c), suggesting that cell-rich areas are K high Na low and that matrix-rich areas are K low Na high , probably due to the exchange of Na and K ions on the cell surface 16,17 . Bone was visualized as calcium (Ca) high areas (Fig. 2a). Along the outside of the filaggrin-positive granular layer 18 , a positive-ion peak of putative ceramide fragments (m/z 5 264.3) 19 was exclusively seen in the SC (Fig. 2a-c). Positive-mode TOF-SIMS mass spectra of each ion are presented in Fig. 2d. Identification of the SC in skin sections by TOF-SIMS. Next, we analyzed 100 mm square areas of skin in mouse tail sections by TOF-SIMS followed by immunofluorescence imaging. In the immunofluorescent images, the viable keratinocyte layer and the SC were clearly distinguished by the staining of desmoplakin, loricrin, and nuclei ( Supplementary Fig. 1a). We superimposed the TOF-SIMS images on the corresponding immunofluorescent images in silico using hair follicles and debris as reference points and determined the border between the SC and viable cell layer ( Fig. 3 and Supplementary Fig. 1b). There were some methodological limitations to superimposing images gained by two different methods (see Methods). Nonetheless, the border appeared clear on the image of Na/K distribution, where the Na low K high viable layer transitioned to Na high K low in the SC (Fig. 3a, b and Supplementary Fig. 2), probably due to the termination of ATP-dependent Na-K exchange on cell membranes undergoing cornification. In the upper layer of the SC, (b) Immunofluorescence images of the same mouse tail section analyzed in (a). Actin staining shows six muscles of the mouse tail, and filaggrin staining shows the granular layer of the epidermis. (c) In silico superimposition of the TOF-SIMS images of the indicated m/z peaks (purple) on the immunofluorescence images of the indicated proteins (green). (d) Positive-mode TOF-SIMS mass spectra in the indicated m/z range from a mouse tail section. The spatial distribution of the molecules of the indicated peaks (arrows) is presented in (a). Each image is representative of three mice, for each of which two tail sections were investigated. Scale bars, 1 mm (a-c).
www.nature.com/scientificreports SCIENTIFIC REPORTS | 3 : 1731 | DOI: 10.1038/srep01731 the concentration of K ions became high again (Fig. 3b). This K likely originated from the external environment, as described below.
Positive fragment ions of choline (m/z 5 86.1), which is derived from intracellular membranes 17 , were highly detected in the viable layer but not in the SC, demonstrating the total disappearance of cytoplasmic organelles under cornification 20 ( Fig. 3b and Supplementary Fig. 2). Several peaks within the m/z range of 260-300 were specifically detected in the SC (Fig. 3c, d). As one peak (m/z 5 264.3) has been reported as resulting from a fragment of ceramide 19 , we analyzed purified ceramide 6, one of the SC ceramides, by TOF-SIMS. Four major peaks of m/z 5 264.3, 268.3, 282.3, and 300.3 were detected, all of which were specifically detected in the SC (Fig. 3c, d and Supplementary Fig. 3), indicating that TOF-SIMS successfully detected a significant portion of SC ceramide. Integrating these results comprehensively, we defined the SC as Na high ceramide high choline low and the viable cell layer as Na low ceramide low choline high .
Three layers in the SC with an arginine-rich layer in the middle. Filaggrin deficiency is the major predisposing factor for atopic dermatitis. Mature filaggrin proteins are produced from proteasedependent processing of profilaggrin under cornification and align keratin filaments into highly ordered and condensed arrays in the lower layers of the SC 2,21-23 . Filaggrin is degraded into so-called natural moisturizing factors (NMFs) within corneocytes and functions in skin hydration.
First, we sought to determine the spatial distribution of NMFs within the SC by comparing the ion peaks detected in the SC from wild-type (WT) versus filaggrin-knockout (KO) mice. We identified an ion peak of m/z 5 175.1, which was specifically reduced in filaggrin-KO SC (Fig. 4a-d and Supplementary Fig. 4). The m/z value highly suggests that this peak is due to a free arginine molecule, one of the major NMFs reduced in filaggrin-KO SC 18 , which was confirmed by TOF-SIMS analysis of purified arginine ( Supplementary Fig. 4).  respectively, overlaid on the images of arginine/K. The average signal count of each area is shown to the right. The SC in WT skin is striped with the K high arginine low upper SC (purple), K low arginine high mid SC (green), and K low arginine low lower SC (yellow) layers. In filaggrin-KO mice, the mid and lower SC layers were indistinguishable and are both illustrated in gray. Each image in a-d is representative of three mice, for each of which at least three tail sections were investigated. (e) The average arginine and K signal counts of each area for nine section views from three mice. When the average K signal counts were compared between neighboring layers, significant differences were detected between the upper SC (WT; 5609   We noticed that the spatial distribution of K and arginine reproducibly stratifies the SC into three layers (Fig. 4a, c). Most of the arginine was specifically concentrated in the middle layer of the SC in WT mice (Fig. 4e), which markedly decreased in filaggrin-KO mice, suggesting that this layer functions in skin hydration. The border of the lower arginine low layer and the arginine high middle layer appeared sharp in the WT SC (Fig. 4a, c), indicating that the production of arginine, and probably other NMFs, from filaggrin is tightly controlled. In the K high upper layer of the WT SC, the amount of arginine markedly decreased, suggesting its further modification (Fig. 4a, c, e). Thus, we divide the SC into three layers according to the distribution of K and arginine: an upper K high arginine low layer (upper SC), a middle K low arginine high layer (mid SC), and a thin innermost K low arginine low layer (lower SC), as shown in Fig. 4a, c and e. In filaggrin-KO mice, weak arginine signals were still detected in the lower K low layer, suggesting the existence of minor NMF resources other than filaggrin (Fig. 4b, d, e). When the lower K low layer was provisionally divided into two sub-layers (putative mid SC and lower SC), no significant difference between the average arginine signal counts of the two layers was detected (Fig. 4d, e).
The upper SC allows the passive influx and efflux of exogenous ions. To investigate the properties of these three distinct SC layers with regard to outside-in barrier function, we performed soaking assays. First, tails of live mice were soaked in water for 15 min and processed for TOF-SIMS imaging. K in the upper SC markedly reduced, resulting in the entire SC becoming equally K-negative (Fig. 5a, c). Na also decreased in the upper SC, but not in the mid or lower SC layers. The distribution of ceramide and arginine appeared unchanged (Fig. 5a, c). When the water-soaked tails were then subjected to soaking in 0.1 M KCl solution for 15 min, K reappeared in the upper SC (Fig. 5b, d).
To further investigate the permeability of the upper SC, we soaked mouse tails in solutions containing several different metal ions. The metal ions used were chosen based on their ability to be detectable by TOF-SIMS and distinguishable from other ion peaks of naturally existing molecules in the skin; thus, we selected chromium (Cr) ions. We soaked the tails in 0.3 M K 2 Cr 2 O 7 water solution to investigate the influx of hexavalent Cr (Cr(VI)). Cr(VI) ions were detected exclusively in the upper SC after soaking for 15, 45, and 90 min (Supplementary Fig. 5 and data not shown; Cr ion peaks are shown in Supplementary Fig. 6). The distribution of K, ceramide, and arginine appeared unchanged, while Na in the upper SC, but not in the mid SC, decreased ( Supplementary Fig. 5). The disappearance of intra-corneocyte Na from the upper SC strongly suggests that the solution directly penetrated corneocytes and washed out the intracorneocyte Na. Thus, the upper SC is suggested to be a specific layer that allows the passive influx and efflux of exogenous ions.
The mid SC works as a first line of defense. These soaking experiments highlighted the mid SC as a barrier against metal ions. To confirm the relevance of TOF-SIMS imaging in the soaking assay, we soaked mouse tails in 0.03 M fluorescein/0.3 M K 2 Cr 2 O 7 water solutions to compare the distribution of fluorescein as detected by TOF-SIMS versus fluorescence microscopy. A specific peak of fluorescein (m/z 5 333.1) was detected from the desiccated fluorescein solution and from the SC of soaked tails, but not from control tails (Supplementary Fig. 7). The fluorescein soaked into the upper SC but not into the mid SC, as did K and Cr(VI) according to TOF-SIMS images (Fig. 5e, f). Sequential analysis of the same specimen by fluorescence microscopy confirmed the spatial distribution of fluorescein observed by TOF-SIMS ( Supplementary  Fig. 8). These observations indicated that the mid SC functions as a barrier against K and Cr(VI) ions as well as against fluorescein.
The lower SC works as a second line of defense. Next, we investigated the difference between the mid SC and lower SC with regard to barrier function. While the mid SC blocked K and Cr(VI) ions, we found that trivalent Cr (Cr(III)) ions soaked into deeper layers of the SC when tails of live mice were soaked in 0.3 M CrCl 3 solution for 45 min (Fig. 6a, c). In Cr(III)-diffused areas, the arginine signals of the mid SC became mostly undetectable, and the Na signals of the upper and mid SC layers markedly decreased. Beneath the Cr(III)-diffused area, a thin SC layer of Na high Cr(III) negative remained, suggesting that Cr(III) soaked into the corneocytes of the upper and mid SC but not into the lower SC (Fig. 6a, c). The blocking of Cr(III) influx into the lower SC remained unchanged when the concentration of CrCl 3 was elevated to 1 M or the soaking time was extended to 90 min (data not shown). These results indicate that the lower SC functions as a barrier that differs from that of the mid SC.
Filaggrin deficiency affects the lower SC barrier against Cr(III). To investigate whether filaggrin deficiency affected either of the two distinct barriers observed here, soaking assays were performed in filaggrin-KO mice. No apparent difference was observed in the barrier function of the mid SC against the influx of fluorescein, K, or Cr(VI) in filaggrin-KO versus WT mice (data not shown). These results are consistent with our previous observations that calcein water solution only penetrated the upper layer of the SC both in WT and filaggrin-KO SC 18 . In contrast, Cr(III) was observed to focally permeate into the viable layer of the epidermis in spots, in filaggrin-KO mice (Fig. 6b, d). Statistical analysis confirmed the significant increase of Cr(III) detected in the viable layer of filaggrin-KO mice compared with WT or control mice (Fig. 6e). Note that even in the areas where Cr(III) soaked into the viable layers, Na signals from the lower SC mostly remained (Fig. 6b), suggesting that the intra-corneocyte space of the lower SC corneocytes is not directly saturated with Cr(III) ions, although the integrity of the lower SC, consisting of the corneocyte and intercorneocyte lipids, was affected. Therefore, among the three SC layers with functionally distinct properties, filaggrin deficiency appeared to have no impact on barrier function of the mid SC, where arginine and probably other NMFs markedly decreased, but specifically affected the integrity of the lower SC layer (Fig. 7).

Discussion
In this study, we applied TOF-SIMS to investigate the function of the SC. TOF-SIMS imaging and matrix-assisted laser desorption ionization (MALDI)-MS imaging are similar techniques that have several advantages, including the lack of the need for fixation or chemical labeling, which avoids delocalization of unfixable molecules, and the ability to simultaneously image a variety of biological compounds in a single run 24,25 . The lateral resolution of MALDI-MS imaging depends on the laser spot size, which is greater than 10 mm 26 , and the use of a matrix further lowers the lateral resolution. In contrast, the lateral resolution of TOF-SIMS is hundreds of nanometers 25 , which allowed us to use TOF-SIMS for the SC analysis. The imaging capabilities of TOF-SIMS of biological samples are generally limited to low-mass ions (,2,000 daltons), and the variety of identifiable molecules is much lower than that for MALDI-MS. This has limited the biological applications of TOF-SIMS imaging 24,25 . In this study, we used bismuth cluster ions as a primary ion source 25 , which enabled us to detect fragments with relatively high masses (i.e., arginine, ceramide fragments, and fluorescein) for analysis of the SC barrier. MALDI-MS is not suitable for the detection of amino acids or low-mass ions such as Cr observed in this study. Furthermore, TOF-SIMS only sputters the surface molecules (to about 1 nm in depth) of samples in contrast to MALDI-MS, which samples a depth greater than 1 mm, allowing analysis of the same location on a sample sequentially by histological examination after the MS analysis.
Using TOF-SIMS, we found that the SC consists of three distinct layers that likely correspond to the metabolic processing of filaggrin.  The arginine low lower SC, which functions as an outside-in barrier, is consistent with the innermost SC harboring interlaced keratin patterns bundled by mature filaggrin that physically stabilize the corneocyte keratin framework 2,20,21,27 , which is lost under filaggrindeficient conditions 18 . Increased susceptibility to mechanical stress may induce focal barrier breakage in the lower SC of filaggrin-KO skin. In the analysis of filaggrin-KO mice, we identified premature detachment of corneocytes rather than direct destruction of corneocytes 18 . Together with the observation that intra-corneocyte Na is mostly preserved in the lower SC, even in the Cr(III)-permeated area, the barrier integrity of the inter-corneocyte space rather than the corneocyte itself might be affected in the filaggrin-KO lower SC (i.e., easy deformation of filaggrin-deficient corneocytes subjected to mechanical stress or weakening of inter-corneocyte corneodesmosome junctions breaks the barrier of the inter-corneocyte lipid lamellar structure). Although remnants of tight junction proteins were detected in the SC by immunoelectron microscopy 28,29 , their contribution to the water-repellant barrier of the SC remains in doubt because claudin-1-KO mice showed no apparent barrier defect in the para-corneocyte pathway of the SC 30 .
The mid SC, which also functions as an outside-in barrier, is rich in arginine and probably other NMFs produced from filaggrin, suggesting that it acts as a hydration layer 31 . The arginine level of the mid SC markedly decreased in filaggrin-KO mice, which made it difficult to distinguish the mid and the lower SC by arginine imaging in KO mice. Small amounts of arginine were still detected in the SC of filaggrin-KO mice, and likely originated from other minor sources of NMFs, such as filaggrin 2 [32][33][34] . The SC hydration of filaggrin-KO mice showed no significant decrease at 22-26uC and 40-60% humidity 18 . Further investigations including estimation of environmental effects, e.g., low humidity, on SC hydration, are needed to explore the contribution of NMFs to SC hydration.
The upper SC works like a ''sponge,'' where solutes (e.g., ions contained in sweat or antimicrobial molecules) flow in and some are retained. The low arginine levels in the upper SC are probably due to further metabolic modification, i.e., citrullination 22 , or direct loss to the external environment, and the K in the upper SC likely originates from the external environment, e.g., smears of urine. As the intra-corneocyte Na was washed out from the upper SC in soaking experiments, externally applied molecules would directly infiltrate the corneocytes of the upper SC. A candidate route for this trans-corneocyte pathway is permeation through degenerated corneodesmosomes, as has been demonstrated for mercury chloride permeation 35 . This pathway will likely facilitate an understanding of the trans-corneocyte infiltration of aqueous solutions specifically observed in the upper SC because corneodesmosomes are only limitedly degraded in the upper layers of the SC [36][37][38][39] .
Small metal ions have been shown to penetrate the skin 11,12,[40][41][42][43][44][45][46] . Our study is the first to show that fine structures within the SC absorb or repel small metal ions. Cr(III), but not Cr(VI), has binding activity with various molecules, including amino acids and proteins 47 , suggesting that this highly reactive property of Cr(III) facilitates deeper infiltration. Nickel allergy may be associated with filaggrin mutations due to increased risk of the metal's penetration through the SC 5-7 . Although we failed to visualize the infiltration of nickel ions because of the presence of molecules with the same m/z value within mouse skin (data not shown), our data suggest that the SC barrier could be weakened against penetration of particular small metal ions, e.g. Cr(III), by filaggrin deficiency.
We do not know which kinds of molecules are able to penetrate the epidermis and enhance percutaneous cellular and humoral immune responses in filaggrin-deficient people, which must be an early and important step in the pathogenesis of atopic dermatitis 1,3 . Further evaluation of the molecular aspects of the three SC layers, as well as assessment of the permeation of various molecules including haptens, will provide us with a better understanding of the SC functions and corresponding deficiencies that contribute to the development of allergic diseases. TOF-SIMS sample preparation. Solutions of ceramide 6, arginine, K 2 Cr 2 O 7 , CrCl 3 , and fluorescein were desiccated on a silicon wafer for the TOF-SIMS analysis. For the mouse tissue analysis, the tails of 6-8-day-old mice were rapidly frozen in crushed dry ice and cut into 1-cm-long sections at -24uC. A small mound of OCT compound (Sakura Finetek, Tokyo, Japan) was mounted on the prechilled cryostat base, and the tail was stuck into the mound. Thus, we omitted sample embedding for sectioning. The upper part of the tail, the lower part of which was fixed by the mound, was sectioned into 12-mm-thick pieces by cryostat (Leica Microsystems, Wetzlar, Germany). All procedures were done at 224uC. To avoid thawing and delocalization of molecules, the sections were directly pasted on prechilled electrically conductive carbon tapes (Nisshin EM, Tokyo, Japan) on glass coverslips at 224uC, freeze-dried for more than 10 h within the cryostat, and preserved at 280uC in airtight plastic tubes in the presence of silica gel until analysis by TOF-SIMS. The sample preparation and analysis procedure is schematically illustrated in Fig. 1. Sample analysis. TOF-SIMS analyses were performed with PHI-TRIFT-IV (ULVAC-PHI, Chigasaki, Japan), using a 60-keV Bi 3 21 primary ion beam with a typical beam size of less than 0.4 mm 48 . The SIMS spectra were generated with a pulsed ion beam that had a duration of approximately 700 ps. Under computer control, the pulsed ion beam was rastered over the desired area to produce MS images. The mass analyzer of the PHI-TRIFT-IV is a triple focusing time-of-flight (TRIFT) mass analyzer based on three 90u electrostatic sectors and a pulse-counting secondary ion detector. The TRIFT analyzer records the mass spectrum at each X-Y pixel of the image generated by the rastered primary ion beam with a typical mass resolution . 15,000 M/Dm [full width at half maximum (FWHM)] and a mass range of over 10,000 m/z, where M is the m/z value of the peak of interest and Dm is the m/z full width at one half peak intensity of the peak of interest. The three electrostatic sectors in the TRIFT analyzer eliminate almost all metastable ions from the background in the mass spectra, which provided spectra with extremely high signal/background ratios for the organic samples used in this study. A pulsed low energy electron beam (, , 10 eV) was used to provide charge neutralization for the insulating samples. The obtained mass spectra data for each X-Y point were processed and integrated using the WinCadenceN software (ULVAC-PHI) to visualize virtually the X-Y distribution of secondary ions of each m/z peak. Positive secondary ion spectra were obtained from 150 3 150 mm or 100 3 100 mm square areas. All the spectra were calibrated using C 2 H 5 , C 3 H 5 , and C 4 H 7 peaks before data analysis. In low magnification analysis, TOF-SIMS images over a surface area of 2250 3 2250 mm were obtained at 256 3 256 pixel density after integration of 15 3 15 tiles, each having 256 3 256 pixel density, using WinCadenceN software (ULVAC-PHI). In high magnification analysis, approximately 5 3 10 7 counts of secondary ions were obtained from a 100 3 100 mm square area, and images of each ion peak were generated at 256 3 256 pixel density. The representative images show typical regions chosen for the analysis in each case. To visualize the distribution of fluorescein via fluorescence microscopy, the samples were mounted in SCMM-R2 (Kawamoto's film method kit 49 ; Leica Microsystems) and rapidly polymerized by UV irradiation to minimize delocalization of fluorescein. In some cases, the specimens were immunostained after TOF-SIMS analysis. Freeze-dried samples were immersed and blocked in phosphate-buffered saline (PBS) containing 10% fetal bovine serum (FBS) and 5% goat serum (Dako, Tokyo, Japan) for 30 min at room temperature and processed for immunostaining. The samples were incubated with primary antibodies in the blocking solution at 4uC for overnight, washed three times with PBS, and incubated with secondary antibodies in the blocking solution at room temperature for 1 h. Samples were washed with PBS, mounted in Mowiol (Merck, Darmstadt, Germany). The reagents and antibodies used were Hoechst 33342, Alexa 647conjugated phalloidin (Invitrogen, Carlsbad, CA), anti-filaggrin (Covance, Berkeley, CA), anti-loricrin (Abcam, Cambridge, MA), anti-desmoplakin (DP-2.15 1 DP-2.17 1 DP-2.20 antibody cocktail; Progen, Heidelberg, Germany), anti-collagen type I (Abcam), and Alexa-conjugated secondary antibodies (Invitrogen).
Superimposition in silico. The area analyzed by TOF-SIMS was subsequently analyzed by fluorescence microscopy. The area was imaged by laser confocal microscopy (TCS SP5; Leica Microsystems). TOF-SIMS images were superimposed on the immunofluorescent images using Photoshop CS4 software (Adobe, San Jose, CA). A methodological limitation exists to superimposing the images gained by two different methods. As the freeze-dried sample analyzed with TOF-SIMS was immersed in PBS for immunostaining and mounted with aqueous mounting medium, the sample swelled and became slightly deformed. Another challenge was that TOF-SIMS scans the surface of the samples, but confocal microscopy scans a cross section of the sample, which causes some positional distortions of the visualized structure depending on the z-axis. Thus, the two images were not completely superimposed.
Signal quantification and statistical analysis. Line scans were generated from the 100 3 100 mm acquired images using WinCadenceN software. The line scans across the image were positioned perpendicular to the imaged layers of interest. The line scan intensity of the desired m/z peaks is the average signal intensity based on a number of pixels parallel to any point along the line scan and within a symmetrical width of 50 mm. In the Cr permeation assay, the total counts of Cr ions detected from the 40 3 20 mm areas of the viable cell layers were compared. Multiple comparison tests were performed by Dunn's multiple comparison procedure using PRISM v6 software (GraphPad Software, La Jolla, CA).