Lamellar Bodies of Human Epidermis

Lamellar bodies (LBs) are tubulovesicular secretory organelles of epithelial cells related to lysosomes. In the epidermis, they play a crucial role in permeability barrier homeostasis, secreting their contents, lipids, a variety of hydrolases, protease inhibitors, and antimicrobial peptides, in the upper keratinocyte layers. The identification of proteins transported in epidermal LBs is still far from complete, and the way their secretion is controlled unknown. In this study, we describe the first proteomics characterization by nano-LC-MS/MS of a fraction enriched in epidermal LBs. We identified 984 proteins, including proteins known or thought to be secreted by LBs. Moreover 31 proteins corresponded to lysosomal components further suggesting that LBs are a new class of secretory lysosomes. Many of the newly found proteins could play a role in the epidermal barrier and desquamation (one acid ceramidase-like protein, apolipoproteins, glycosidases, protease inhibitors, and peptidases) and in LB trafficking (e.g. Rab, Arf, and motor complex proteins). We focus here on CLIP-170/restin, a protein that mediates interactions between organelles and microtubules. Western blotting confirmed the presence of CLIP-170 and its known effectors IQGAP1 and Cdc42 in the LB-enriched fraction. We showed, by confocal microscopy analysis of skin cryosections, that CLIP-170 was expressed in differentiated keratinocytes, first at the periphery of the nucleus then with a granular cytoplasmic labeling evocative of LBs. It was preferentially co-localized with Cdc42 and with the known LB protein cathepsin D. CLIP-170 was also largely co-localized with Rab7. This study strongly suggests a new function for CLIP-170, its involvement together with Cdc42 and/or Rab7 in the intracellular trafficking of LBs, and provides evidence that nano-LC-MS/MS combined with monodimensional electrophoresis separation constitutes a powerful method for identifying proteins in a complex mixture such as subcellular structures.

Terminal differentiation of keratinocytes is a program of gene expression oriented from the basal layer to the surface of the epidermis. Keratinocytes express specific proteins in a sequential manner and undergo a series of structural and metabolic modifications resulting in the formation of corneocytes. The accumulation of these cells forms the stratum corneum and establishes a protective permeability barrier (1). Secretory structures called lamellar bodies (LBs) 1 or keratinosomes are crucial to stratum corneum homeostasis and barrier function. As observed by classical electron microscopy, LBs are round or oblong membrane-delimited and lamellate organelles measuring 0.1-0.5 m in diameter. Recent cryotransmission electron microscopy studies have revealed that LBs are rather branched tubulovesicular structures derived from the trans-Golgi network (2). Moreover Norlen (2) has proposed that membranes of the trans-Golgi network and of LBs and the keratinocyte plasma membrane may well be part of one and the same continuous membrane structure. The observations that LBs contain lysosomal enzymes like cathepsin D (3,4) and vacuolar H ϩ -ATPase and have acidic contents (5) have led them to be considered as lysosomerelated organelles. They are produced in the keratinocytes of the upper stratum spinosum and are more abundant in the cytoplasm of the granular keratinocytes (6,7). They secrete their contents, lipids (glucosylceramides, sphingomyelin, phospholipids, and cholesterol) and lipid-processing enzymes (e.g. ␤-glucocerebrosidase also known as glucosylceramidase, acid sphingomyelinase, and secretory phospho-lipase A 2 ) that make the upper layers waterproof (8,9), structural proteins like corneodesmosin that optimizes corneocyte cohesion (10), proteases (kallikreins and cathepsins), glycosidases and protease inhibitors (e.g. elafin, secretory leucocyte protease inhibitor, lympho-epithelial kazal-type-related inhibitor, and ␣ 2 -macroglobulin-like) involved in the control of desquamation (11,12), and antimicrobial peptides (defensins and cathelicidin) (13,14), at the stratum granulosum/ stratum corneum interface. The secreted proteins seem to be delivered by independent trafficking of various cargoes (4). Recently Sando et al. (15) have shown an enrichment of caveolin-1 in an isolated LB fraction and have confirmed the location of caveolin-1 in LBs by immunoelectron microscopy, suggesting that caveolins may play a role in LB assembly, trafficking, and/or function.
LB secretion is often increased to correct an impaired epidermal barrier function, including recovery from adhesive stripping, repair of experimental wounds (16), and psoriasis (17). Defective delivery of LB content has been noted in atopic dermatitis, leading to a significant increase of their volume in the transition zone between the stratum granulosum and the stratum corneum (18). The LB lipid transporter ABCA12 is underexpressed in harlequin ichthyosis (OMIM 242500), resulting not only in reduced transfer of lipids to LBs but also in abnormal delivery of enzymes to the stratum corneum extracellular spaces. This induces profound abnormalities in the stratum corneum and epidermal barrier and a delay in desquamation (19,20).
Organelles morphologically similar to LBs have been observed in other tissues implicated in the establishment of a barrier between the organism and the environment: mucosa of the gastrointestinal tract, the vermilion border of the lip, tongue papillae, and occasionally in the epithelium of the cervix (21,22). LBs of pneumocytes are the organelles for surfactant secretion into the alveolar lumen of the lung, reducing surface tension at the air-liquid interface and preventing collapse of the organ during expiration (23). Interestingly mutations of ABCA3 are associated with respiratory distress syndrome (OMIM 267450) (24), suggesting common or similar pathways between epidermal and lung epithelium LBs.
The identification of proteins transported by epidermal LBs and of their constitutive components is still far from complete, and little is known about the molecular mechanisms involved in their trafficking. Rab11, a small GTPase involved in recycling endosomes, has been co-localized with known LB molecules, including glucosylceramides, corneodesmosin, and cathepsin D, suggesting that it could play a role in the intracellular trafficking of various types of LBs to the cell surface (25).
In this study, we analyzed a fraction enriched in epidermal LBs by LC-MS/MS, and then we focused on CLIP-170/restin, a protein described previously as mediating interactions between organelles and microtubules and being involved in vesicular trafficking (26). By confocal microscopy analysis, we showed that CLIP-170, together with one of its effectors, was localized on a subpopulation of LBs in normal human epidermis.

EXPERIMENTAL PROCEDURES
Human Tissues-All the studies described concerning human tissues were conducted in accordance with the Declaration of Helsinki and with the ethical guidelines of both the University of Toulouse III and the French Ministry of Research and Technology. Normal human breast skin fragments were obtained from patients without any history of skin diseases undergoing plastic surgery and after informed consent. They were provided by Prof. J.-P. Chavoin (Plastic Surgery Department, Rangueil Teaching Hospital, Toulouse, France).
Subcellular Fractionation-The method we used to prepare a fraction enriched in LBs (supplemental Fig. 1) closely followed that described previously (15). Dermoepidermal cleavage was performed by heat treatment. The subsequent steps were performed at 4°C. Epidermis from 100 cm 2 of skin was minced with scissors and disrupted by a loose fitting Teflon-glass Potter-Elvehjem homogenizer in 40 ml of homogenization buffer (20 mM sodium phosphate, pH 6.5, 0.15 M NaCl, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml pepstatin, 2 g/ml aprotinin, 2 g/ml leupeptin, and 10 mM 2-mercaptoethanol). The homogenate was centrifuged for 15 min at 800 ϫ g to remove cell debris, and the supernatant obtained (S1) was centrifuged for 30 min at 20,000 ϫ g. The 20,000 ϫ g supernatant was centrifuged again for 30 min at 100,000 ϫ g to obtain the S100 cytosolic fraction. The 20,000 ϫ g pellet was washed in 40 ml of homogenization buffer, centrifuged again for 30 min at 20,000 ϫ g, reconstituted in 1 ml of homogenization buffer, and layered onto a preformed linear density gradient of 20 -50% Optiprep (Sigma). After centrifugation for 2 h at 26,000 ϫ g, we observed three opaque bands at the top, middle, and bottom of the gradient. The upper band was collected and filtered through a 0.8-m-pore size polyethersulfone membrane filter (Millipore Corp.) to get the LB-enriched fraction.
Proteins of the LB-enriched fraction were diluted into 10 ml of homogenization buffer and either centrifuged for 15 min at 15,000 ϫ g for electron microscopy analysis or precipitated with 10% trichloroacetic acid for separation by SDS-PAGE. To obtain the Tris-EDTA-Nonidet P-40 epidermis extract, after dermoepidermal cleavage had been performed by heat treatment, the epidermis was extracted in a Tris-EDTA buffer containing 0.5% Nonidet P-40 as reported previously (27). Protein concentrations were measured using the Bio-Rad protein assay.
Antibodies-The characteristics of the antibodies used in the study are summarized in Table I.
Protein Electrophoresis and Immunoblotting-Proteins were separated by SDS-PAGE on 7.5 or 12.5% acrylamide gels (7 or 12 cm long) and electrotransferred to nitrocellulose membranes. The membranes were stained with Ponceau Red, then incubated for 30 min in blocking buffer (TBS (40 mM Tris-HCl, pH 8, and NaCl 0.9%) supplemented with 0.5% low fat dried milk and 0.05% Tween 20), and probed with antibodies as described previously (27). Immunoreactivities were revealed with an ECL TM Western blotting kit as recommended by the manufacturer (GE Healthcare).
Electron Microscopy-For ultrastructural studies by standard electron microscopy, whole skin or the pelleted LB-enriched fraction was fixed in 0.1 M cacodylate-buffered 3% formaldehyde, 3% glutaraldehyde overnight at 4°C and postfixed in 1% osmium tetroxide with 1.5% potassium ferrocyanate in 0.2 M cacodylate buffer at pH 6.8. After rinsing and dehydration through an ascending acetone series, the tissues were embedded in LR White resin (SPI Supplies, West Chester, PA). Thin sections were stained with uranyl acetate and examined in a Hitachi HU12A electron microscope.
Indirect Immunofluorescence-Indirect immunofluorescence (IIF) was performed on unfixed 4-m-thick cryosections of breast skin samples blocked with PBS containing 4% goat serum and 0.05% Tween 20 and then incubated with primary antibodies diluted in PBS containing 2% goat serum and 0.05% Tween 20 for 1 h at 37°C or overnight at 4°C. The sections were washed twice in PBS containing 0.05% Tween 20 and twice in PBS and then incubated for 1 h at room temperature with the secondary antibody (Alexa Fluor 488 goat antimouse IgG (H ϩ L), Alexa Fluor 555 goat anti-rabbit IgG (H ϩ L), or Alexa Fluor 488 donkey anti-goat IgG (H ϩ L), Molecular Probes, Eugene, OR) diluted to 1: 400 in PBS containing 2% goat serum and 0.05% Tween 20. After two washes in PBS containing 0.05% Tween 20 and two washes in PBS, nuclei were stained with Toto-3 iodide (1:1000; Molecular Probes) for 30 min at room temperature. The sections were washed in distilled water and eventually mounted in an antifading solution (Mowiol, Calbiochem) before observation under a confocal laser microscope (LSM 510, Carl Zeiss, Oberkochen, Germany). For negative controls, sections were incubated without primary antibodies. Results were analyzed using the NIS-Element AR 2.30 software (Nikon, Tokyo, Japan). In particular, co-localization was evaluated by a comparison of the emission spectra of the fluorochromes along a line crossing several stained structures. This was performed to analyze all, at least 100 (the exact number is indicated by "n"), of the stained structures on one focal plan.
Analysis by 1D Gel/Nano-LC-MS/MS-One hundred micrograms of proteins of the LB-enriched fraction were resuspended in 40 mM Tris-HCl, pH 7.5, and 6 M urea and incubated for 30 min at room temperature after addition of Laemmli buffer. Proteins of the S100 fraction were directly incubated after addition of Laemmli buffer. Proteins of both fractions were alkylated in 90 mM iodoacetamide for 30 min at room temperature, separated by SDS-PAGE on 12.5% acrylamide short gels (7 cm long), and detected by Coomassie Blue staining. The entire gels were cut into 16 homogeneous slices. Alternatively 200 g of proteins prepared in the same conditions were separated on 12-cm-long gels that were cut into 55 slices.
Gel slices were separately washed and then digested with modified sequencing grade trypsin (Promega, Madison, WI), and the resulting peptides were extracted following established protocols (28). The resulting peptides were analyzed by nano-LC-MS/MS using an LC Packings system (Dionex, Amsterdam, Netherlands) coupled to a QSTAR XL mass spectrometer (Applied Biosystems). Dried peptides were reconstituted in 12 l of solvent AЈ (5% acetonitrile and 0.05% trifluoroacetic acid in HPLC grade water), and 6 l were loaded onto a precolumn (300-m inner diameter ϫ 5 mm) using the Switchos unit of the LC Packings system, delivering a flow rate of 20 l/min of solvent AЈ. After desalting for 7 min, the precolumn was switched on line with the analytical column (75-m inner diameter ϫ 15 cm, PepMap C 18 ) equilibrated in 95% solvent A (5% acetonitrile and 0.1% formic acid in HPLC grade water) and 5% solvent B (95% acetonitrile and 0.1% formic acid in HPLC grade water). Peptides were eluted from the precolumn to the analytical column and then to the mass spectrometer with a gradient from 5 to 50% solvent B for 80 min at a flow rate of 200 nl/min delivered by the Ultimate pump. The QSTAR XL was operated in information-dependent acquisition mode with the Analyst QS 1.1 software. MS and MS/MS data were recorded continuously with a 5-s cycle time. Within each cycle, MS data were accumulated for 1 s over the mass range m/z 300 -2000 followed by two MS/MS acquisitions of 2 s each on the two most abundant ions over the mass range m/z 80 -2000. Dynamic exclusion was used within 60 s to prevent repetitive selection of the same ions. Collision energies were automatically adjusted according to the charge state and mass value of the precursor ions. The MS to MS/MS switch threshold was set to 10 cps.
Database Searching-The Mascot Daemon software (Mascot Daemon version 2.1.6, Matrix Science, London, UK) was used to automatically extract peak lists from Analyst .wiff files and to perform database searches in batch mode with all the .wiff files acquired on each gel slice. For peak list creation, the default charge state was set to 2ϩ, 3ϩ, and 4ϩ. MS and MS/MS centroid parameters were set to 50% height percentage and a merge distance of 0.1 amu. All peaks in MS/MS spectra were conserved (threshold intensity set to 0% of highest peak). For MS/MS grouping, the following averaging parameters were selected: reject spectra with less than five peaks or precursor ions with less than 5 cps or more than 10,000 cps; the precursor mass tolerance for grouping was set to 0.1 Da, the maximum number of cycles per group was set to 10, and the minimum number of cycles per group was set to 1  (Mascot Daemon version 2.1.6, Matrix Science). Carbamidomethylation of cysteines was set as a fixed modification, and oxidation of methionine and phosphorylation of serine and threonine were set as variable modifications for all Mascot searches. Specificity of trypsin digestion was set for cleavage after lysine or arginine except before proline, and two missed trypsin cleavage sites were allowed. Both the peptide MS and MS/MS tolerances were set to 0.5 Da.
Mascot results were parsed with the in-house developed software Mascot File Parsing and Quantification (MFPaQ) (29), which allows automatic validation of the proteins returned by Mascot according to user-defined criteria and also provides a convenient interface for performing manual validation. All the Mascot result files (.dat files) corresponding to a series of nano-LC-MS/MS analysis (fractions of a 1D gel) were extracted in batch, and protein hits were automatically validated if they satisfied one of the following criteria: identification with at least one top ranking peptide ("red" peptide) with a Mascot score of more than 51 (corresponding to a p value of 0.001, i.e. 99.9% identification confidence), or at least two top ranking peptides each with a Mascot score of more than 36 (corresponding to a p value of 0.0316), or at least three top ranking peptides each with a Mascot score of more than 31 (corresponding to a p value of 0.1). These two last validation criteria correspond to a final confidence of p ϭ 0.001 if each peptide identification is considered independent. To evaluate the false positive rate in these large scale experiments, we repeated the Mascot searches using identical search parameters against a decoy database and performed MFPaQ parsing on these results using the same criteria. The decoy database was the compilation of UniProt Swiss-Prot and UniProt TrEMBL Databases (same versions as mentioned above) in which the sequences had been reversed. The false positive rate was calculated for each fraction by dividing the number of validated identifications in the reversed database with the number of validated identifications in the forward database. A final percentage of false positives was obtained by averaging the false positive rates obtained for each fraction.
From all the validated result files corresponding to the fractions of a 1D gel lane, MFPaQ was used to generate a unique, non-redundant list of proteins. The software compares proteins or protein groups (composed of all the proteins matching the same set of peptides) and creates clusters from protein groups found in different gel slices if they have one common member. This feature allowed a global list to be edited of unique proteins (or clusters) representing the entire sample analyzed in the experiment.

RESULTS
Preparation of LB-enriched Fraction-An LB-containing fraction was obtained from a human epidermis homogenate with a powerful method using subcellular fractionation and density gradient sedimentation as described previously (15) but with an additional filtration step (supplemental Fig. 1). This step cleared out most of the contamination from large fragments of membranes and melanosomes as observed by electron microscopy (data not shown). The LB-containing fraction obtained was then analyzed by Western blotting, electron microscopy, and immunoelectron microscopy ( Fig. 1). Multilamellar vesicles were observed with morphology and size (0.1-0.5 m in diameter) very similar to those of LBs observed in the entire epidermis ( Fig. 1, a and b). A few vesicles were intact (Fig. 1b, black arrows), but most of them were damaged and more or less empty (Fig. 1b, open arrows). To study the efficacy of the enrichment in LBs, equal amounts of proteins of the crude (S1 fraction) and LB-containing fractions were immunodetected using antibodies directed against LB markers caveolin-1, kallikrein-8, and corneodesmosin (Fig. 1c). Much higher amounts of caveolin-1 (19-and 22-kDa isoforms) and kallikrein-8 (28 kDa) were detected in the LB-fraction, thus showing the enrichment. Corneodesmosin was also enriched but was largely degraded during the procedure despite the presence of several protease inhibitors. In contrast, high levels of the cytosoluble protein caspase-14 were detected in the S1 fraction at 28 and 17 kDa, the molecular mass of its proform and active form, respectively, but caspase-14 was hardly present in the LB-enriched fraction. In the LB-enriched fraction analyzed by immunoelectron microscopy, we were also able to detect kallikrein-7 (Fig. 1d) and dermokine (data not shown), two other proteins known to be secreted via LB (30,31). Taken together, these results show that the LB fraction was indeed enriched in LB, allowing further proteomics characterization. Herein it will be referred to as the LB-enriched fraction.
Identification of Proteins by LC-MS/MS Analysis of the LBenriched Fraction-The LB-enriched fractions prepared from the homogenates obtained from the epidermis of 15 different individuals (ϳ250 cm 2 ) were pooled. The cytosoluble (S100) fractions were also pooled, and 100 g of proteins of each pool were separated by monodimensional electrophoresis on two 7-cm-long acrylamide gels. The entire gel lanes were cut into 16 slices, and proteins were digested by trypsin and then identified by nano-LC-MS/MS analysis and Mascot database searches. Mascot result files were parsed automatically, and protein hits were validated using different criteria based on the number, rank, and score of the assigned peptides. Using these criteria, 570 proteins were identified from the S100 fraction, and 467 proteins were identified from the LB-enriched fraction. The lists of proteins corresponding to each sample are provided (supplemental Tables 1 and 2, respectively) together with the lists of peptides identified in each gel slice with their associated MS data (supplemental Tables 4  and 5). Protein identification with a single, top ranking peptide match was validated when the score of this single peptide was higher than 51 (p Ͻ 0.001), and for these identifications, the corresponding annotated MS/MS spectra are shown in supplemental Fig. 2. With the validation criteria used, the estimated rate of protein false positives from the reversed database search was closed to 0.3%. Among the 467 pro-teins, 241 were common to both proteomes, whereas 226 were specific to the LB-enriched fraction. A large number of the common proteins corresponded to components of the cytoskeleton and to abundant nuclear, ribosomal, and proteasomal proteins (data not shown). The latter probably correspond to contamination as observed in other high throughput MS proteomics approaches (32). Proteins were sorted according to their subcellular distribution either automatically ( Fig. 2, a and b) with the knowledge-based software developed by Ingenuity Systems Inc. that helps to interpret data sets in the context of biological process and molecular network or manually (data not shown) using several broad (PubMed, ExPASy, Human Protein Reference Database, and EBIMed) and specialized (MEROPS, Secreted Protein Database, and Protein Data Bank of Transmembrane Proteins) databases. Comparable results were obtained with both methods, showing the validity of data obtained with the Ingenuity software. For example, similar percentages of proteins from the LB-enriched fraction were classified as membranous (16 versus 17%) and extracellular (5% for both) or were not classified (unmapped and unknown; 8 versus 6%). The major discrepancies were noted for cytoplasmic (58 versus 62%) and nuclear (13 versus 10%) proteins. Confirming the enrichment, 16% of proteins identified in the LB-enriched fraction were automatically sorted as membrane proteins as compared with only 5% of the proteins identified in the S100 fraction.
We noted that two LB markers present in the enriched fraction as shown by Western blotting, corneodesmosin ( Fig.  1c) and kallikrein-7 (data not shown), were not identified dur-

FIG. 2. Cellular and functional distribution of the proteins identified by the LC-MS/MS analyses. a and b
, 570 proteins were identified from the S100 fraction, and 467 proteins were identified from the LB-enriched fraction during the first LC-MS/MS analysis (sample fractionated into 16 gel slices). They were sorted automatically with the knowledge-based software developed by Ingenuity Systems Inc. according to their subcellular distribution. c, during a second LC-MS/MS identification, 517 additional proteins were identified in the LBenriched fraction (sample fractionated into 55 gel slices). The 984 non-redundant proteins from the LB-enriched fraction were manually analyzed according to their cellular locations. ER, endoplasmic reticulum.
ing the LC-MS/MS approach. This suggested that other proteins had probably also escaped the MS identification. Complete proteomics analysis of a complex mixture is limited by the sequencing speed of the mass spectrometer as many peptides co-elute from the chromatographic column at any given time and cannot all be selected for MS/MS sequencing. This leads to preferential identification of high abundance components, whereas minor proteins remain undetected. To increase the power of the LC-MS/MS method, we separated twice the amount of proteins (200 g) of the LB-enriched fraction from another preparation (prepared from 28 different skin samples) on a longer gel and cut the migration lane into 55 slices. Most (89%) of the previously identified proteins were found again, showing the reproducibility of the purification method and of the proteomics analysis. Moreover extensive separation of the proteins of the LB-enriched fraction, resulting in longer analysis time (higher number of nano-LC-MS/MS runs performed on mixtures of lower complexity), allowed the number of MS/MS scans performed by the mass spectrometer and the number of proteins identified to be increased (supplemental Tables 3 and 6). As shown in supplemental Table 3, 517 additional proteins were identified from this experiment. Using the same validation criteria as before, the reversed database search used as a negative control revealed 0.86% false positives. The improvement in the method allowed us to identify more proteins of low abundance, such as enzymes (37% (191 of 517) as compared with 29% (135 of 467)), and of a hydrophobic nature like transmembrane transporters/ion channels (12% (62 of 517) versus 8% (37 of 467)). Altogether the two procedures identified 984 non-redundant proteins from the LB-enriched fraction. It is noteworthy that 91% of the proteins were identified by two peptides hits or more.
Classification of the Identified Proteins-The identified proteins were manually analyzed according to their cellular locations (Fig. 2c). Only a few proteins corresponded to endosomes (0.4%), melanosomes (0.2%), and peroxisomes (0.6%). The major classes corresponded to cytoplasmic (18.6%) and membrane (16.0%) proteins in accordance with the location of LBs in the cytoplasm of keratinocytes and with the fact that LBs are surrounded by a membrane. Other classes mainly represented include cytoskeleton (12.4%), nucleus (10.7%), and ribosomes (10.1%). When we tentatively classified the proteins in terms of relative abundance, as estimated by the number of MS/MS queries, we once again observed at the top of the list many cytoskeleton and ribosomal components. Even if this estimation was biased, e.g. hydrophobic proteins may be under-represented because of their biochemical properties, it shows the difficulty to obtain purified LBs and was clearly not sufficient to designate genuine constituents of LBs from possible contaminations. One possibility could be to quantitatively compare the distribution of the identified proteins all along the density gradient used to prepare the LB-enriched fraction using the protein correlation profiling approach (33). However, this approach in our case may be not completely differential: indeed if cytosoluble proteins are entrapped inside small size vesicles co-migrating with LBs during their enrichment (see "Discussion" for reasons of that) the contaminating proteins may not have a fractionation profile very different from that of true LB proteins. We also could test for the cellular localization of the identified proteins by immunohistology performed on skin sections with specific antibodies when available; a punctate labeling of the upper epidermal cells is suggestive of LB components. However, this is difficult to envisage with such a high number of proteins. We therefore chose a candidate approach based on the location and functional classification of the identified proteins with the help of both the literature knowledge and Ingenuity software. Some (331 of 984) turned out to be of particular interest (Table II and

CLIP-170, a Protein Known to Mediate Interactions between Vesicles and Microtubules, and Its Direct Effector
Cdc42 Are Associated with LBs Together with the Rab7 Protein-Among the proteins known to take part in vesicular and membrane trafficking, several direct and indirect effectors of a protein called CLIP-170/restin were highlighted using the Ingenuity software (Fig. 3). CLIP-170 is one of the cytoplasmic linker proteins, a class of proteins that facilitate the interaction between cellular organelles and microtubules. Because CLIP-170 itself was not present in the protein list obtained from nano-LC-MS/MS analysis, we went back to the Mascot database search results. CLIP-170 had indeed been identified but with a Mascot score too low to be validated. To confirm the presence of CLIP-170 in the LB-enriched fraction we immunodetected the protein sample with a monoclonal anti-CLIP-170 antibody specific for amino acids 1128 -1427 (clone F3). As expected, a 170-kDa protein was specifically detected (Fig. 4). In IIF confocal microscopy analysis on skin cryosections (Fig 5a), the anti-CLIP-170 antibody displayed a punctate labeling, starting in the upper spinous keratinocytes at the nucleus periphery, and more intense labeling in the cytoplasm LGALS1 Lectin, galactoside-binding, soluble, 1 (galectin 1) 9; 59% P17931 LGALS3 Lectin, galactoside-binding, soluble, 3 (galectin 3) 42; 44% Q6IB87 LGALS7 Lectin, galactoside-binding, soluble, 7 (galectin 7) 262; 88% Tripeptidyl-peptidase I 6; 17% Calcium-binding proteins S100 proteins Q3KRB9 S100A2 S100 calcium-binding protein A2 4; 22% P26447 S100A4 S100 calcium-binding protein A4 6; 37% Q5RHS4 S100A6 S100 calcium-binding protein A6 (calcyclin) 3; 24% P05109 S100A8 S100 calcium-binding protein A8 (calgranulin A) 6; 53% Q5T1C5 S100A10 S100 calcium-binding protein A10 15; 74% P31949 S100A11 S100 calcium-binding protein A11 (calgizzarin) 6; 50% Q9HCY8 S100A14 S100 calcium-binding protein A14 17; 71% Q5RHS6 S100A16 S100 calcium-binding protein A16 9; 50% Q9UDP3 None Putative S100 calcium-binding protein H_NH0456N16.     of the granular keratinocytes. This pattern, which can be clearly observed in a reconstructed three-dimensional view (supplemental Fig. 3), was highly suggestive of the labeling of vesicles that could be LBs. Indeed the stained structures showed a diameter of 240 Ϯ 120 nm (n ϭ 1107), which is in accordance with the previously described size of LBs. Because CLIP-170 is considered as a ubiquitously expressed microtubule-binding protein, we were surprised of this labeling that was mainly restricted to the more differentiated keratinocytes. We therefore performed two types of control experiments. First, we tested the anti-CLIP-170 monoclonal antibody F3 on HeLa cells by double staining IIF using a polyclonal anti-tubulin antibody. As expected, we observed a patchy distribution in the interphase cells often organized into linear arrays that correlate with microtubules. Moreover there was an accumulation of CLIP-170 at the peripheral ends of some but not all microtubules (supplemental Fig. 4a). Second, we analyzed the expression of CLIP-170 in human epidermis by IIF using another unrelated antibody, namely N18, directed against the amino terminus of the protein. Once again, we only detected CLIP-170 in the upper part of the epidermis with a punctate labeling of the granular keratinocytes (supplemental Fig. 4b). We then performed double stainings of normal skin cryosections with the anti-CLIP-170 monoclonal antibody F3 and with polyclonal rabbit antibodies directed against either corneodesmosin or cathepsin D, two LB markers (Fig. 5, b-d). Corneodesmosin was mainly detected in the granular keratinocytes as published previously (34). CLIP-170 and corneodesmosin showed few co-locations, and only a few anti-CLIP-170-labeled vesicles were stained by the anticorneodesmosin antibodies (9%, n ϭ 110). Cathepsin D was detected in the cytoplasm of all keratinocytes with a more intense labeling of the granular cells as described previously (4). In the granular keratinocytes, a large proportion (88%, n ϭ 124) of anti-CLIP-170-labeled vesicles were stained by the anti-cathepsin D antibodies (Fig. 5, c and d). These data suggest that CLIP-170 is associated with a subpopulation of LBs, at least those containing cathepsin D. CLIP-170 is known to form a tripartite complex with IQ motif-containing GTPase-activating protein 1 (IQGAP1) and Rac1/Cdc42 (35). Because Rac1 is only expressed in the basal layer of human epidermis (36), we were interested in a possible interaction in the upper epidermis between CLIP-170 and Cdc42. Rab proteins have been reported to recruit effectors to specific membrane compartments in their GTP-bound form and are involved in vesicular trafficking (37,38), and Rab11 has been shown to be associated with epidermal LB (25). In particular, Rab7 may regulate exocytosis of vesicles with endosomal/lysosomal characteristics, via the CLIP-170 tripartite complex, by binding Cdc42 (39). We therefore checked for a possible co-location of CLIP-170, IQGAP1, Cdc42, and Rab7 in the human epidermis.

CLIP-170 Association with Lamellar Bodies of Human Epidermis
First we confirmed the presence of both CLIP-170 partners (IQGAP1 (190 kDa) and Cdc42 (25 kDa)) and of Rab7 (23 kDa) in the LB-enriched fraction by Western blotting (Fig. 4). We then analyzed their expression on human normal skin cryosections by IIF and confocal microscopy (Fig. 6). IQGAP1 was detected in the cytoplasm of all keratinocytes with a more intense staining at the cell periphery as described previously (40). Anti-Cdc42 antibodies labeled the cytoplasm of keratinocytes in all layers of the epidermis with an increased staining intensity at the nucleus periphery. Double labeling did not reveal any particular co-location of CLIP-170 and IQGAP1 (Fig. 6a), whereas many anti-CLIP-170-labeled vesicles in the granular layer were also stained by the anti-Cdc42 antibodies (84%, n ϭ 115) (Fig. 6, b and  d). Confirming an interaction between CLIP-170 and Cdc42, we identified Cdc42 by MS as co-immunoprecipitated from  Table 3).
b The sequence coverage and number of matched peptides correspond to the MS/MS analysis of the proteins identified in the analysis of the short (7-cm-long) gel fractionated into 16 slices.
c Proteins known to either interact with caveolin or be a component of caveolae.
an epidermal extract (Tris-EDTA-Nonidet P-40 extract) by the anti-CLIP-170 antibodies but not by an isotypic negative control (data not shown). Anti-Rab7 antibodies mainly produced a cytoplasmic labeling of the granular keratinocytes (Fig 6c). Some of the CLIP-170 and Rab7 were also shown to be co-localized; some vesicles (43%, n ϭ 101) were labeled by both anti-CLIP-170 and anti-Rab7 antibodies (Fig. 6, c and e). Furthermore we observed a partial co-localization between Rab7 and corneodesmosin (Fig. 7, a and c). There was a clear, puzzling co-localization between corneodesmosin and Cdc42 in the granular keratinocytes (Fig. 7, b and d). This suggests that Cdc42 is associated with most if not all LBs, including those containing corneodesmosin, but not routed via CLIP-170 molecules.

Efficacy and Limits of the Preparation of the LB-enriched Fraction
We modified and improved a previously described protocol for the preparation of an LB-enriched fraction from human epidermis; in particular, we added a filtration step that largely removed membrane sheets and melanosome contamination. The final fraction obtained was mainly composed of vesicles with morphology and size similar to that of LBs. The presence and/or enrichment of known LB markers (caveolin-1, kallikrein-7, kallikrein-8, corneodesmosin, and dermokine) was evidenced by Western blotting and immunoelectron microscopy, confirming the efficacy of the method. However, the organelles seemed to be partly emptied of their genuine cargos as shown by electron microscopy observations and by the fact that the enrichment of the membrane protein caveolin-1 was more effective than that of secreted proteins (kallikreins and corneodesmosin) (Fig. 1). This could be related to the tubulovesicular nature of LBs and to their proposed intertwining with cell cytoskeletons (2), which would induce their damage during cell fractionation. As expected, when we compared the proteins of the LB-enriched fraction identified by LC-MS/MS with the proteins of the S100 fraction, we noted a clear increase in the number of known membrane and membrane-associated proteins (16 versus 5%) (Fig. 2, a and b). Further confirming the efficiency of the fractionation, only a few of the 984 proteins identified corresponded to melanosome (2 of 984) or peroxisome (5 of 984) intrinsic components (Fig. 2c). The main contaminants of our preparation seemed to be nuclearly and cytoplasmically abundant proteins, in particular mitochondrial, ribosomal, and proteasomal compo- ; GDI2, GDP dissociation inhibitor 2; HMGA1, high mobility group AT-hook 1; ITGB4, integrin ␤4; JUP, junction plakoglobin; MAPK1, mitogen-activated protein kinase 1; MAPRE1, microtubule-associated protein RP/EB family, member 1; MSN, moesin; PAFAH1B1, platelet-activating factor acetylhydrolase, isoform Ib subunit 1; PAFAH1B2, platelet-activating factor acetylhydrolase, isoform Ib subunit 2; PPIA, peptidylprolyl isomerase A; PPP2CA, protein phosphatase 2 catalytic subunit, ␣ isoform; PSMA7, proteasome subunit, ␣ type, 7; RAB1B, RAB1B, member RAS oncogene family; RAB6A, RAB6A, member RAS oncogene family; RAC1, ras-related C3 botulinum toxin substrate 1; RDX, radixin; RHOA, ras homolog gene family, member A; RSN, CAP-GLY domain-containing linker protein 1; S100A6, S100 calcium-binding protein A6; S100A11, S100 calcium-binding protein A11; S100B, S100 calcium-binding protein B; SAFB, scaffold attachment factor B; SNCA, synuclein ␣; SOD1, superoxide dismutase 1; TAF15, TAF15 RNA polymerase II; TLN1, talin 1; VCL, vinculin; VIL2, villin 2; WASF2, WAS protein family, member 2; YWHAZ, tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, polypeptide. The major interactions found between these proteins are indicated by different arrows. Solid thick line, binding; solid thick arrow, acts on; dashed thin line, indirect interaction; solid thin line, direct interaction; horizontally elongated diamond, peptidase; triangle, phosphatase; inverted triangle, kinase; vertically elongated diamond, other enzyme; trapezoid, transporter; horizontally elongated oval, transcription regulator; vertically elongated oval, transmembrane receptor; circle, others.

CLIP-170 Association with Lamellar Bodies of Human Epidermis
nents and glycolytic enzymes. This could be due to the great abundance of these proteins in the cells and/or to their physicochemical properties, which cause them to interact with membranes after cell fractionation, as reported in a similar proteomics analysis strategy (41). Moreover during the enrich-ment procedure, high mechanical forces must be applied to disrupt the skin tissues and keratinocytes that may damage the cytoplasmic organelles. LBs may open and close up, encapsulating cytoplasmic proteins. Also other types of damaged organelles as well as pieces of plasma membrane can close up in the same way and form small size vesicles full of cytoplasmically abundant proteins. They may subsequently co-migrate with LB vesicles on the density gradient, and their protein content would then be preferentially identified during the nano-LC-MS/MS analysis. Thus, it may be difficult to distinguish between genuine components of the LBs, proteins associated with the LB vesicles, or contaminating proteins. For example, many of the MS/MS-identified proteins in the LBenriched fraction corresponded to actin (76 of 984) cytoskeleton. This could either be because these proteins are abundant in the cells or, more likely, because cytoskeleton elements become attached to the vesicles that pass along them. In the following sections, we discuss some proteins identified in this study that seem relevant in light of the known function of LBs.
Moreover we chose a candidate approach to go further into the analysis of the present data and performed validation and functional experiments on some particular proteins to show their implication in the trafficking of LBs.

Analysis of the Epidermal LB Proteome
Overall 984 proteins were identified from the LB-enriched fraction. For most of them, this is the first evidence of their expression in human epidermis. Four hundred seventy-six have been shown to be expressed in the granular layer of the epidermis at the mRNA level (42).
Among the identified proteins, we found caveolin-1 and 21 proteins that have been previously localized in caveolae

CLIP-170 Association with Lamellar Bodies of Human Epidermis
and/or described to interact with caveolin-1 (Table II), confirming the suspected role of caveolae in the organization and/or function of LB (15). For example, apolipoproteins (A-I, A-II, A-IV, B, and H) may be involved in cholesterol and phospholipid transport in collaboration with ABCA receptors (43). Flotillins 1 and 2, considered as markers for lipid microdomains, have been reported to play a role in vesicle trafficking and cytoskeleton rearrangement (44,45). Interestingly flotillin 1 has been localized in endo/lysosomal vesicles in adipocytes (46) and identified in fractions enriched in lysosomes of natural killer T cells (47,48). Caveolae serve as assembly sites for signaling complexes (49), and we consistently identified 54 proteins known to play a role in signal transduction like kinases, phosphatases, and guanine nucleotide-binding proteins, including some already known to interact with caveolae: mitogen-activated protein kinase 1 and protein kinase A (50,51). The way these proteins might influence epidermal LB secretion and more widely the role of caveolae proteins in LB remain to be discovered.
Based on the identification of proteins that are or may be either transported into LBs or involved in LB trafficking, we propose a functional model of these secretory structures (Fig.   8). It is discussed in detail below. Clearly more investigations are required to determine whether these proteins are really localized at the level of LBs and, for some of them, in the stratum corneum extracellular spaces and to dissect what role they play in the upper epidermis, e.g. immunoelectron microscopy analysis with specific antibodies, or small interfering RNA inactivation in three-dimensional skin models (necessary to achieve sufficient differentiation of cultured keratinocytes to get LBs).
Secreted Proteins-Among the 984 proteins, five are already known to be secreted via LBs: two proteases, cathepsin D and kallikrein 7 (the latter protein did not satisfy the criteria used for automatic validation but was identified with a total Mascot score of 54.85 and two top ranking peptides for which the MS/MS spectra were verified manually and are shown in supplemental Fig. 2); the ␤-glucosylceramidase involved in the maturation of glucosylceramides to ceramides in the stratum corneum; and two proteins with unknown function, dermokine ␤ and suprabasin (52).
Five proteins have been previously reported to be expressed in the granular layer of the epidermis and present in the stratum corneum. They have been supposed to be com- ponents of but have not been localized in LBs before this work. These are the zinc-␣ 2 -glycoprotein (also known as desquamin), the function of which is unclear (53,54); the retroviral-like aspartic protease saspase (55); the antimicrobial protein RNase7 (56); the protease inhibitor cystatin A (57); and prosaposin, which is involved in transport and maturation of sphingolipids (58).
We detected several extracellular matrix proteins: osteoglycin, decorin, lumican, ladinin 1, and periostin. Nothing is known about the secretion of extracellular matrix elements in the stratum granulosum, although the extracellular matrix is increasingly being considered as important for the properties and function of the stratum corneum (59). This identification leads us to suppose that LBs may be involved in the secretion of extracellular matrix components in the upper epidermis.
Proteins Potentially Involved in Desquamation-Among the proteins identified in the LB-enriched fraction, we found three proteins (two proteases, namely cathepsin D and kallikrein 7, and one member of a family of cathepsin inhibitors, namely cystatin A) already known to play a role in desquamation and 19 others that could also be implicated in this complex process. This is the case for cystatin B. Indeed cystatin M/E, another member of the family, has been detected in the upper layers of human epidermis and shown to be secreted by LBs in the intercorneocyte spaces where it is assumed to be involved in the control of desquamation (60,61). This is also the case of six lysosomal glycosidases (a fucosidase, the ␣-glucosidase, ␤ 1 -galactosidase, ␤-glucuronidase, hexosaminidase B, and ␣-mannosidase). Secreted glycosidases are suspected to be involved in desquamation, and synergis-   Table II except for kallikrein 7 (two matched peptides; 10% sequence coverage; gene name, KLK7) and acid ceramidase-like (two matched peptides; 10% sequence coverage; gene name, ASAHL; protein name, N-acylethanolamine-hydrolyzing acid amidase). tic activities of proteases and glycosidases have been demonstrated in the degradation of corneodesmosomes in the upper stratum corneum. They could hydrolyze the sugar moieties of the extracellular parts of corneodesmosome proteins, facilitating their proteolytic degradation (62,63). However, the exact nature of the glycosidases present in the stratum corneum is not known. In addition, lysosomal peptidases identified in the LB proteome (tripeptidyl-peptidase I, dipeptidylpeptidase II, vitellogenic carboxypeptidase-like protein, and prolyl carboxypeptidase) may also be secreted and involved, like the above discussed proteases, in the control of desquamation. Finally we identified eight serpins (serpins A1, A3, A12, B5, C1, F1, G1, and H1), members of a family of serine protease inhibitors (64). Serpins B2 and B5 have already been shown to be expressed in the granular layer of the epidermis, and serpin B2 is cross-linked to the cornified envelope by transglutaminase 1 (65,66). Overexpression of serpin B2 has been observed in patients with Netherton syndrome (OMIM 256500), suggesting that it plays a role in desquamation. Therefore we propose that several serpins, including those we identified, are secreted by LBs and involved in the regulation of desquamation.
Proteins of Lysosomes-Epidermal LBs are considered to be lysosome-related organelles, so we paid attention to proteins previously localized in lysosomes. Confirming this relationship, we found lysosome luminal (17 in addition to cathepsin D and prosaposin) and membrane (11) proteins, including LAMP1 and LAMP2. In agreement with this evidence for their location in the epidermal LBs, LAMP1 and -2 have previously been described in the LBs of the lung (67). The acidic content of epidermal LBs has been highlighted previously (5), but the exact nature of H ϩ -ATPase involved in the maintenance of this acidic character had not been identified. In this study we distinguished seven subunits of the ATP6V1 proton pump that we propose as a good candidate for this function. We also found several glycosidases and peptidases that could take part in desquamation as discussed above. Also among the lysosomal proteins that we suggest are secreted by LBs, there are acid ceramidase 1 and a recently described acid ceramidase-like protein (again the latter protein was not automatically validated but was identified with a Mascot score of 46.8 and with MS/MS spectra shown in supplemental Fig. 2) that should take part in the catabolism of ceramides in the stratum corneum. Acidic ceramidase activity is preferentially localized within the lysosomal compartment in extracutaneous tissues (68). In the epidermis, an acidic ceramidase activity and acid ceramidase 1 have been described in the stratum corneum. Until now, based on the physicochemical properties of these enzymes, the ceramidase activity has been considered to be restricted to the extracellular space. Moreover the strong increase in acidic ceramidase activity observed during keratinocyte differentiation is not paralleled by a similar change in acid ceramidase 1-encoding mRNA (69). Although this discrepancy has been tentatively explained by post-transcriptional modifications, it may also result from the expression of another acid ceramidase, i.e. the acid ceramidase-like protein we identified. In addition, the galactosylceramidase present in the LB-enriched fraction has never been described in the epidermis and could play a role in the maturation of galactosylceramides in the stratum corneum (70).
Our results suggest that LBs could be considered as a specialized kind of secretory lysosomes. Lysosomes are the main degradative compartment within eukaryotic cells for misfolded trans-Golgi network-derived and cytosolic proteins. They also represent the end point of the endocytic pathway; the endocytosed material is either degraded or recycled back to the plasma membrane. However, some cell types contain lysosome-related organelles, also called secretory lysosomes, which secrete their contents into the extracellular spaces in response to external stimuli. These organelles share many characteristics with conventional lysosomes: an acidic luminal pH, many acid hydrolases, and the presence of LAMP proteins at their membrane but the absence of mannose 6-phosphate receptors. Various molecules are stored in the secretory lysosomes before their secretion according to the cell type: melanins in melanocytes, histamine in mast cells, perforin and granzymes in natural killer T cells, and clotting agents in platelets (71,72). Another example of this specialization of secretory lysosomes can be found in osteoclasts, which are the cells responsible for the degradation of the organic and inorganic bone matrix during bone remodeling. When bone resorption is induced, osteoclasts migrate to the site of resorption and secrete the contents of their lysosomerelated acidic vesicles. The mineral is dissolved by the acid, the collagen matrix is degraded by the proteinases, and then the degradation products are internalized for recycling (39).
Our results indicate that epidermal LBs are of this type of secretory lysosomes.
Proteins Potentially Implicated in the Transport and Secretion of LBs-Among the proteins identified from the LB-enriched fraction, a number are suspected to be involved in vesicular trafficking, in particular the Rab family and Rabassociated proteins. Rab proteins in their GTP-bound form ensure a linkage between vesicle membranes and trafficking complexes recruiting effectors with high specificity to membrane compartments (73). Among the identified Rab proteins, we found Rab11 that was recently detected in the upper spinous and granular layers of human epidermis where it is associated with LBs (25), Rab3D previously implicated in the exocytosis of a subpopulation of LBs in alveolar epithelial type II cells (74), and Rab27B that regulates the secretion of the platelet secretory lysosomes (75). In this study, we focused on Rab7, a protein known to be localized in endo/lysosomerelated vesicles (76,77) and shown to control, together with Rac1, the transport of secretory lysosomes to the apical zone of the plasma membrane in bone-resorbing osteoclasts (39). We found that Rab7 was mainly expressed in the cytoplasm of granular keratinocytes where it was partially co-localized with corneodesmosin by indirect immunofluorescence. Using immunoelectron microscopy, we also detected Rab7 in very close proximity with some LBs in human epidermis (data not shown). Rab7 could therefore be involved in some particular steps of LB secretion or in the secretion of a particular subpopulation of LBs. In agreement with this line of reasoning, in living A431 cells, the following of several Rabs tagged with green fluorescent protein and of transferrin as an endosomal cargo has shown that the Rab proteins are compartmentalized within the same continuous membrane of early endosomes, creating mosaics of different Rab domains through the recruitment of specific effector proteins (78).
We also identified proteins known to form intermediate complexes between vesicles and cytoskeletal elements such as the ADP-ribosylation factors (Arf1, -4, -5, and -6 and Arflike 8A and 8B) (79). Proteins of the Arf family are believed to function as regulators of membrane trafficking in the secretory pathway (80). For example, Arf6 seems to be required for endosome recycling and has been shown to regulate actin modeling through the activation of Rac1 (81). Arf-like 8A and 8B have been localized to lysosomes in mammalian cells. Their overexpression results in microtubule-dependent redistribution of lysosomes toward the cell periphery, suggesting a role in lysosome transport (82). Detailed studies would be necessary to prove the implication of the identified Arf proteins in LB trafficking.
Motor proteins and some of their receptors implicated in the vesicular movements along the microtubules or actin filaments (kinesin light chain 1, kinectin 1, dynein components, dynactin 2, and myosin-Id) are present in the proteome of the LB-enriched fraction. This together with the presence of actin and tubulin subunits and also actin-and microtubule-associated proteins (supplemental Tables 1 and 3 and Table II) strongly suggests that LBs are routed in keratinocytes on microtubules and maybe microfilaments like other vesicles in other cells (83).
According to the "SNARE hypothesis," a model explaining the process and specificity of docking and fusion of vesicles to their target membranes during vesicular transport, membrane proteins from the donor vesicle (v-SNAREs or VAMPs) recognize and interact with proteins of the acceptor membrane (t-SNAREs or syntaxins). Once the two SNAREs are bound to each other, they recruit the general elements of the fusion apparatus, namely the N-ethylmaleimide-sensitive factor (NSF) and soluble NSF attachment proteins (SNAPs), to the site of membrane fusion. A mutation in SNAP29 causes a neurocutaneous syndrome called the cerebral dysgenesis, neuropathy, ichthyosis, and palmoplantar keratoderma (CED-NIK) syndrome (OMIM 609528), particularly characterized by an abnormal maturation of LB and, in consequence, a mislocation of epidermal lipids and proteases. The resulting ichthyotic phenotype suggests a crucial role for membrane fusion proteins in the epidermis barrier homeostasis (84). In this study, we found a v-SNARE, namely VAMP8, which is known to be implicated in the secretion of platelet (85) and mast cell (86) secretory lysosomes, and a t-SNARE, namely syntaxin 4. We also identified components of the fusion apparatus, the proteins SNAP-␣, SNAP23, and palladin (87). Moreover we identified the SNAP23-associated transmembrane protein SCAMP2, which also plays a role in membrane fusion during exocytosis (88).
Proteins Potentially Implicated in the Sorting of LB Cargos-The sorting nexins constitute a large conserved family of hydrophilic molecules. Based on the functions of their yeast homologs, it has been suggested that mammalian sorting nexins ensure the proper delivery of organelle-specific proteins (89). Among the five sorting nexins we identified, sorting nexin-1 is particularly relevant for LB because it has been shown to play a role in targeting ligand-activated epidermal growth factor receptor to the lysosomes for its degradation (90). In yeast, the vacuolar protein sorting (Vps) proteins are involved in trafficking of endocytic and biosynthetic proteins to the vacuole, the fungal equivalent of the mammalian lysosomes. In association with sorting nexin-1 and -2 and other Vps proteins, human Vps35 is required to retrieve proteins from the endosomes to the late-Golgi network (91). Proteins of the sorting nexin and Vps families identified in the LBcontaining fraction may be implicated in the biosynthetic sorting of LB cargos, but their exact function remains to be established as is the case for the NIPSNAP3A protein (92).

Evidence for a Role of CLIP-170 and Its Associated Proteins in the Trafficking and/or Secretion of LBs
Among the identified proteins that had previously been implicated in vesicular trafficking, we focused on a protein called restin/CLIP-170, a member of the cytoplasmic linker protein (CLIP) family, and on some of its effectors. CLIP-170 is known to mediate the linkage of vesicles to microtubules. It specifically accumulates at the plus ends of growing microtubules where it recruits dynactin that can then regulate the motor proteins dynein and kinesin (26). CLIP-170 has been reported to form a tripartite complex with IQGAP1 and Rac1/ Cdc42 that appears to function as a linker between vesicles, microtubules, and cortical actin meshwork probably through Rab proteins (35,39,93). We were particularly interested in this protein for several reasons. (i) Microtubules have already been postulated to be involved in LB transport in epithelial pulmonary cells, and drug-or bacterial infection-induced changes in microtubules have been associated with changes in the metabolism of surfactant components (94). (ii) In cytotoxic T lymphocytes polarized after interaction with their target cells, secretory lysosomes move along microtubules toward the cell-cell contact sites. There their content is released into the so-called secretory cleft inducing the target cell death (95). (iii) Disruption of microtubules in human monocytes appears to impair lysosome exocytosis (96). We showed that, in human epidermis, CLIP-170 is expressed in the actual cells in which LBs are observed from the nuclear edge in the spinous keratinocytes to the cell periphery in the upper granular layer. It is not expressed, or was not detected, in the basal and lower suprabasal keratinocytes. Although CLIP-170 is usually considered as a ubiquitously expressed protein, to our knowledge no exhaustive studies explore its expression in every tissue and cell type. Because it has not been detected in myocytes of heart muscle and in some glial cells (Human Protein Atlas), CLIP-170 expression may depend on cell type and differentiation state. Moreover on skin cryosections analyzed by confocal microscopy, the anti-CLIP-170 antibodies produced a punctate staining highly evocative of vesicles with a diameter compatible with that of LBs. We also demonstrated that CLIP-170 is co-localized with cathepsin D, a protease secreted via LBs (Fig. 5). Unfortunately we were not able to localize CLIP-170 at the ultrastructural level probably because, in immunoelectron microscopy, masking of the epitope recognized by the anti-CLIP-170 antibody takes place. In agreement with the location of CLIP-170 on LBs, a large number (Ͼ50) of the proteins we identified in the LBenriched fractions, including Cdc42, Cdc42-binding protein kinase ␣, and BAI1-associated protein 2, are known to interact either directly or indirectly with CLIP-170 as highlighted using the knowledge-based software developed by Ingenuity Systems Inc. (Fig. 3). A majority of (84%), but not all, CLIP-170-positive vesicles were also stained by an anti-Cdc42 antibody (Fig. 6). Further confirming the interaction between CLIP-170 and Cdc42 in human epidermis, we observed that both proteins were co-immunoprecipitated from an epidermis extract by the anti-CLIP-170 antibody. Finally we found evidence that CLIP-170 was also partially co-localized with Rab7 (Fig. 6). These data strongly suggest that the three proteins form a complex at the membranes of cathepsin D-containing LBs and are probably involved in their trafficking and/or secretion. They may facilitate the transport of the vesicles along microtubules from the Golgi apparatus to the keratinocyte periphery. Because CLIP-170 seems not to be colocalized with corneodesmosin, another protein transported in LBs, it might be that CLIP-170 is implicated in the transport of a subpopulation of LBs. If true, this would confirm that LBs are a heterogeneous population of vesiculotubular structures. Indeed Ishida-Yamamoto et al. (4) have observed that the proteases and corneodesmosin are sorted independently as distinct aggregates. It has also been shown that the serine protease inhibitor lympho-epithelial kazal-related inhibitor is localized in LBs separated from its epidermis-specific target proteases KLK7 and KLK5 (97). In addition, the observation that a proportion of Cdc42 is co-localized with corneodesmosin and not with CLIP-170 suggests that all LBs are linked to Cdc42 independently of their binding by CLIP-170. It would be interesting to look for the presence of another member of the CLIP family on the CLIP-170-negative and corneodesmosin-positive LBs.
Surprisingly IQGAP1, which is expressed in the cytoplasm of all epidermal keratinocytes, was not found to be co-local-ized with CLIP-170, its described partner in other cell types (35) (Fig. 6a). However, anti-IQGAP1 antibody stained LBs both before and during secretion as observed by immunoelectron microscopy (data not shown). The potential implication of this multifunctional protein in LB secretion cannot be excluded but remains to be analyzed in more detail. In any case, it does not seem to be implicated in the CLIP-170⅐Cdc42 complex.
In conclusion, this study provides evidence that nano-LC-MS/MS combined with monodimensional separation constitutes a powerful method for identifying proteins in a complex mixture such as subcellular structures. It provides the first data on the complex molecular mechanisms involved in LB transport and secretion in the epidermis and highlights a new function for CLIP-170 in a complex with Cdc42 and/or Rab7. We also confirmed the existence of several subpopulations of epidermal LBs and suggest that different molecular complexes control their transport in the granular keratinocytes. Our data also indicate that epidermal LBs must be considered as secretory lysosomes. Further investigations on each protein identified will be necessary for a clear understanding of LB trafficking and secretion in the epidermis and more generally in other epithelia.