Systematic Identification and Characterization of Novel Human Skin-Associated Genes Encoding Membrane and Secreted Proteins

Through bioinformatics analyses of a human gene expression database representing 105 different tissues and cell types, we identified 687 skin-associated genes that are selectively and highly expressed in human skin. Over 50 of these represent uncharacterized genes not previously associated with skin and include a subset that encode novel secreted and plasma membrane proteins. The high levels of skin-associated expression for eight of these novel therapeutic target genes were confirmed by semi-quantitative real time PCR, western blot and immunohistochemical analyses of normal skin and skin-derived cell lines. Four of these are expressed specifically by epidermal keratinocytes; two that encode G-protein-coupled receptors (GPR87 and GPR115), and two that encode secreted proteins (WFDC5 and SERPINB7). Further analyses using cytokine-activated and terminally differentiated human primary keratinocytes or a panel of common inflammatory, autoimmune or malignant skin diseases revealed distinct patterns of regulation as well as disease associations that point to important roles in cutaneous homeostasis and disease. Some of these novel uncharacterized skin genes may represent potential biomarkers or drug targets for the development of future diagnostics or therapeutics.


Introduction
DNA microarray technology has been used in numerous expression profiling experiments on mammalian skin designed to understand normal and pathological skin biology but also to characterize the various cell types present in skin at the molecular level [1,2,3,4,5,6]. However, there have been no comparative studies on gene expression in normal human skin relative to other tissues. As a consequence, skin-associated genes are poorly represented in sequence databases [5].
We have used DNA microarray technology to measure the expression levels of more than 47,000 unique transcripts in 105 different adult human tissue and cell types. The results provide a comprehensive genome-wide perspective of each tissue's expression signature. We have called this database the Body Index of human Gene Expression (BIGE) [7,8]. In the present study, we have used the BIGE database to identify skinassociated genes (SAGs). To this end, we compared the expression value of a given gene in skin to its mean relative expression value in all other tissues and cell types (n = 104) and selected genes that are highly expressed in the skin and little or no expression elsewhere. The resulting set of 687 genes was classified into functional classes utilizing both manual and automated annotation to better understand the molecular composition of skin and to identify significant pathways. Furthermore, we have focused on the top 100 genes preferentially expressed in normal human skin for further characterization. These include a subset of eight poorly characterized genes that are selectively and strongly expressed in human skin which have not been previously associated with skin and encode either secreted or membrane proteins. The skin expression of these novel SAGs was confirmed by semi-quantitative PCR (qPCR) and expression of the proteins encoded by a subset of SAGs was confirmed by immunohistochemistry (IHC) and Western blot. In addition, we analyzed the expression of these genes in cultured primary human keratinocytes and a wide panel of inflammatory, autoimmune or malignant skin diseases and found several important associations. Our results indicate that we have identified several novel skin-associated genes that may represent potential biomarkers and/or drug targets.

Identification of Skin-associated Genes
We used the BIGE database [7,8] to identify genes preferentially expressed in human skin. The five skin samples were from two sources; three were collected from the lumbar region of individual donors (two female, one male) and two were total skin RNAs from commercial sources. Selection of genes with a mean skin to mean body (the remaining 104 tissue/cell types) expression ratio $2.0 fold resulted in a list of SAGs containing 687 genes (see Table S1).
This systematic approach leverages two powerful attributes of the database: the inclusion of gene expression data from a large spectrum of tissues and cell types, and the use of genome-wide microarray platform (the Affymetrix human genome U133 plus 2.0 array) that provides a comprehensive, ''global'' profile of gene expression in each sample [7]. By including 104 tissue and cell types (listed in Table S2), our analysis excluded broadly expressed ''housekeeping'' genes and focused instead on genes that participate in functions and pathways restricted to skin. The use of genome-wide arrays provided data for genes not included in many prior studies, including ,20% human genes that encode proteins with unknown function [9]. Importantly, a subset of SAGs identified in this study represents uncharacterized genes, as described below.

Functional Overview
To understand the biological significance of the full list of 687 SAGs we used manual annotation that utilized gene ontology (GO) term descriptors, published data and protein family (Pfam) sequence alignments to assign SAGs to broad functional classes, and the ''Database for annotation, visualization and integrated discovery'' (DAVID) tool that systematically identifies functional classes enriched in gene lists [10]. Results from both analyses are summarized in Figure 1A, (see Table S3 for the complete DAVID output).
Metabolism, signaling, gene expression and protein metabolism/trafficking constitute the largest gene classes by manual annotation, representing 35% of our SAGs, indicating that normal skin is a highly active tissue. Many of the remaining functional classes reflect the barrier function of the skin. Our data and mining strategy were validated by the identification of large numbers of genes in the structural, cell adhesion and keratins functional groups and the highly significant enrichment of pathways associated with ectoderm and epidermis development apparent in the DAVID output. Many cell adhesion genes encode desmosome and tight junction components [11] associated with cytoskeletal filaments that form the keratinized squamous cell layer of the skin. We also detected a large number of genes (n = 21) clustered in the epidermal differentiation complex (EDC) locus (located in human chromosomal band 1q21) that encode components of the cornified cell envelope, intermediate filamentassociated proteins and calcium binding proteins [12,13] (see Table S4) also associated with barrier function, and some of which have been implicated in several common skin disorders [14].
Expression of many known SAGs is restricted to hair follicles. Previously, other groups have studied gene expression in hair follicles using laser capture microscopy (LCM), but did not use genome-wide microarrays [15,16]. 133 of the 687 SAGs we identified were described by Ohyama et al. as expressed in hair follicle bulge cells (see Table S5). The unusually high number of genes encoding enzymes involved in lipid metabolism (38 of 82 in the metabolism category, Figure 1A) may be explained by the presence of sebaceous glands, which secrete lipid-rich sebum.
We identified 37 genes associated with skin defenses, including many genes previously linked to both innate or adaptive immune responses as well as several genes encoding antimicrobial peptides (AMPs). Expression of these genes in normal skin suggests the presence of very active mechanisms of microbiome control that operate in the normal skin to prevent bacterial or fungal infections. Table 1 lists the top 100 genes that exhibit the highest specific expression in human skin. The most abundant skin-associated gene, dermcidin (DCD), is a well characterized AMP produced by eccrine sweat glands [19]. Interestingly, the DCD gene encodes a multifunctional precursor protein, which is cleaved to generate a microbicidal C-terminal 47 amino acid peptide and an N-terminal peptide that acts as a pleiotropic soluble factor associated with cell survival and cachexia [20]. The second most abundantly expressed gene in skin, keratin 2 (KRT2, formerly KRT1B), is the highest expressed of eleven keratin genes in the top 100 list. Transcripts for the recently described filaggrin family member 2 (FLG2) [21] and desmocollin 1 (DSC1), a desmosomal cadherin, are also highly enriched in skin. Both of these proteins participate in skin barrier formation [21,22]. Two members of the secretoglobin gene superfamily, SCGB2A2, (also known as mammaglobin 1) and SCGB1D2, (also known as lipophilin B) are also highly expressed in skin. Expression of both secretoglobins in mammary gland has been reported [17,23,24,25,26,27]. SCGB2A2 has also been shown to be expressed in human sweat glands and down-regulated in sweat gland tumors [17]. While the normal function of secretoglobins in healthy tissues remains unclear, other members of this superfamily are anti-inflammatory [28,29,30]. Significantly, SCGB1D2 is also present at high levels in tears [31,32] suggesting a protective role in either defense or barrier function as a mucinlike protein [31]. Both SCGB2A2 and DCD exhibit similar BIGE expression profiles with high expression levels in skin, but relatively low expression levels elsewhere (in contrast to other highly expressed SAGs). As expression of both genes is restricted to eccrine sweat glands, these data may point toward significant differences in the presence and makeup of sweat glands at different sites in the human body with wide distribution of eccrine sweat glands versus limited, site-specific distribution of apocrine glands.

Individual Gene Expression
Taken together, our data indicate that nine out of the top 10 genes we have identified qualify as highly specific skin-markers, showing exclusive expression in skin or in other skin-containing organs (i.e. nipple and penis). SCGB2A2 is an exception, given it is also expressed in the mammary gland ( Figure 2 and see Table S7).
Some immune system-associated genes are highly expressed in normal skin. Indeed, sixteen genes encoding cytokines, chemokines and their receptors are present in the full list of 687 SAGs including four in the top 100: CCL27, IL-1F7/IL-37, IL-20RB and TNFRSF19. CCL27, the highest expressed immune associated gene we identified, is a skin-associated chemokine expressed by epidermal basal keratinocytes, which chemoattracts skin-homing cutaneous lymphocyte-associated antigen bearing (CLA + ) T cells [33] and has been associated with allergic and irritant reactions [34] as well as atopic dermatitis [35]. Another homeostatic chemokine, CXCL14 (rank 154), is also highly expressed in normal skin and other mucosal sites [36,37] and exhibits direct microbicidal activity [38,39] suggesting that it may play an active role in the control of the skin microbiome. Two members of the IL-1 family, IL-37 and IL-36G, are prominently expressed in skin and may play antagonistic roles in skin immunity. Pro-inflamma-tory IL-36G induces anti-microbial peptide expression by keratinocytes in psoriasis [40], whereas, IL-37 acts as a repressor of innate immunity [41,42]. The balance of these two cytokines may contribute to the control of inflammation in the skin.
IL-37 represents the second most skin-specific cytokine, after CCL27 (see Figure S2 and Table S7). The IL-37 gene appears late in evolution and does not exist in the mouse. The latter observation may explain why it has not been previously associated with the skin. Results from a recent study linking IL-37 and psoriatic arthritis suggest that this gene may also play a role in autoimmunity [43].
According to information currently available in PubMed and GenBank, 20 of the top 100 SAGs identified in our study have not previously been associated with the skin. Of these, 14 represent known genes while 6 represent uncharacterized transcripts ( Figure 1B). A similar trend was observed for the top 300 genes (see Figure S1 and Table S6).
Notably, our complete analysis identified .50 genes with no informative annotation or functional association. This represents one of the most significant collections of novel skin-associated genes. Given that only about 18% of previous microarray studies on skin gene expression used genome-wide arrays and none utilized gene expression in other tissues as a filter may explain these results (http://www.ncbi.nlm.nih.gov/gds/). Of greatest interest are the poorly characterized genes in the top 300 list that encode highly expressed secreted or transmembrane proteins as these represent potential targets for diagnosis and/or treatment of skin disease (see Table S6). We therefore undertook a more detailed expression analysis of eight of these genes (see Table S8).
We confirmed high levels of expression in normal skin for all eight uncharacterized and two established SAGs (KRT1; rank 15, and DCD; rank 1) using qPCR ( Figure 3) in eleven independent normal human skin samples. These SAGs exhibited little or no expression in other human tissue samples tested. Measurement of the expression of aldolase B (ALDOB), which is expressed by kidney and liver, confirmed our analysis showing a marked expression in these organs but no expression in spleen, brain or skin [44].
We also used qPCR to identify the cell type(s) that express each selected SAG using cultured primary human skin-derived cells ( Figure 4). Four genes (GPR87, GPR115, SERPINB7 and WFDC5)  Table S1 for functional annotation for all 687 SAGs. doi:10.1371/journal.pone.0063949.g001  or MUCL1 in the skin cells tested. As DCD expression is known to be restricted to eccrine sweat glands, these data suggest that both C5orf46 and MUCL1 may also be secreted by eccrine sweat or by seborrheic glands [19]. MUCL1 (also known as SBEM) is a small secreted sialomucin (similar to SCGB1D2 and SCGB2A2) that has been proposed as a breast cancer biomarker [45]. Accordingly, MUCL1 is significantly expressed in normal mammary gland (a pattern shared by other genes including keratins 5, 14, 15 and 17), and also desmoplakin, a tight junction protein [46,47,48]. This suggests common features shared between the skin and mammary gland proteomes. Indeed, myoepithelium, which is found in glandular structures present in sweat, mammary, lacrimal and salivary glands, is composed of keratin-rich epithelial cells and may be the source of these commonly expressed transcripts [49]. Interestingly, both the MUCL1 and DCD genes are in the same locus on chromosome 12 (12q13) and adjacent to another small secreted glycoprotein, lacritin, which is expressed in the lacrimal and mammary gland tissue as well as in breast cancer [50,51].
Cadherin-related family member 1 (CDHR1; formerly known as protocadherin 21, PCDH21) is a calcium-dependent adhesion protein and belongs to the e-group of non-clustered protocadherins [52]. CDHR1 is a proposed photoreceptor-specific gene and CDHR1 null-mutations have been shown to cause retinitis pigmentosa [52,53].
WAP four-disulfide core domain 5 (WFDC5) is one of an eighteen-member family of small secreted proteins originally reported as protease inhibitors but now known to also participate in inflammation and host defense [54,55,56]. However, the function(s) and sites of expression of WFDC5 are currently unknown. Our data strongly suggests that WFDC5 expression plays a role in inflammatory skin diseases and host defense.
TMEM45A is a highly conserved protein with strong sequence conservation in vertebrates (47% identity with its Zebrafish ortholog) and limited identity to OAS1, a protein involved in anti-viral responses [57]. TMEM45A contains the Pfam domain DUF716 (domain of unknown function 716) associated with viral infection in plants (InterPro IPR006904). Thus, this gene and its protein product may play a role in immunity in the skin. A recent study reported that TMEM45A is overexpressed in breast cancer and is associated with poor prognosis in these patients [58]. Conversely, TMEM45A suppression promotes the progression of ductal carcinoma in situ to invasive breast cancer [59].
Human clade B serpins belong to a cohort of evolutionarily dispersed proteinase inhibitor clades that protect cells from promiscuous proteolysis [60]. SERPINB7 (Megsin) has been shown to be specifically expressed in kidney and in epithelial cells of the uterus [61,62]. Over expression of human SERPINB7 in transgenic mice led to increased mesangial cell proliferation and extracellular matrix (ECM) deposition [61] and, in an in vitro cell invasion/ migration assay, markedly suppressed cell motility and invasion [63]. Thus, this gene product appears to participate in tissue integrity by maintaining ECM homeostasis and loss of expression may lead to loss of cell adhesion and tissue integrity.
GPR115 is one of 33 human adhesion-GPCRs characterized by long extracellular N-terminal domains [64]. It was identified in a systematic search for glucocorticosteroid-response genes and may mediate the effects of this hormone in the skin [65]. GPR87, a rhodopsin-type (class A) GPCR, was identified by sequence data mining [66,67] and was later shown to be overexpressed in squamous cell carcinoma [68] and act as a pro-survival, p53induced gene [69]. Taken together, these observations suggest that GPR87 may play a role in skin cancers.
We confirmed the expression of three SAGs (WFDC5, TMEM45A and GPR115) at the protein level using IHC of both PHKs and healthy skin sections ( Figure 5). WFDC5 staining was significant in the cytoplasm of keratinocytes throughout the stratum basale (SB), stratum spinosum (SS) and stratum granulosum (SG) of the epidermis. In the stratum corneum (SC) WFDC5 expression was confined to the most apical layer. In addition, WFDC5 staining was also apparent in the small vessel endothelia of the dermis possibly due to the protein accumulating on the luminal surface of these vessels. TMEM45A was also expressed in the epidermis and exhibited an increase in intensity from the SB to the SC. As expected, TMEM45 was also detected in dermal fibroblasts. GPR115 shows weak expression in the SB and a marked expression in SG and SC. Increases in expression from basal to apical layers of the epidermis for both TMEM45A and GPR115 suggest that their expression increases upon differentiation. The keratinocyte-specific expression at the protein level was confirmed for all three genes by Western blot.
Next, we sought to analyze the functional regulation of our selected SAGs in primary human keratinocytes and also in common human skin diseases. As C5orf46 did not show expression in primary keratinocytes and is potentially expressed by another skin-resident cell subset (Figure 4), this gene was excluded from the keratinocyte analyses. Primary human keratinocytes (PHKs) were incubated with medium controls or a panel of cytokines, including GM-CSF, TNF-a/IL1-b, IFN-c, IL-4, IL-13, IL-31, IL-17 and IL-22. The relative expression of selected SAGs was subsequently measured by qPCR ( Figure 6). The following skin diseases were selected for our survey: psoriasis vulgaris (PSO), atopic dermatitis (AD), prurigo nodularis (PRU), lupus erythematosus (LE), lichen planus (LP), basal cell carcinoma (BCC), actinic keratosis (AK) and squamous cell carcinoma (SCC), (Figure 7). As none of our selected SAGs showed any expression in melanocytic nevi or malignant melanoma, these entities were excluded from our disease panel (data not shown). We found distinct patterns of regulation for each gene. MUCL1 expression was significantly inhibited by IL-31 ( Figure 6A). IL-31 is a recently discovered T-cell derived cytokine that is overexpressed in pruritic skin inflammation [70]. In contrast to the in vitro results, MUCL1 was only significantly upregulated in LE and not conditions associated with pruritis ( Figure 7A).
WFDC5 showed a marked and significant induction by TNF-a/ IL1-b ( Figure 6B). Notably, WFDC5 was upregulated significantly in PSO, followed by AK and AD ( Figure 7B). The latter observation, along with the fact that the WFDC protein family has been proposed to play an important role in inflammation [54,55,56], strongly suggests that WFDC5 is a proinflammatory protein. PSO represents one of the major inflammatory skin disorders and TNF-a is one of the predominant cytokines involved in the pathogenesis of this disease. Notably, anti-TNF-a therapy has evolved as an effective strategy in the management of PSO [71,72]. Moreover, other WFDC proteins have been reported to exhibit antimicrobial activity [54]; this leads us to postulate that  WFDC5 may also have antimicrobial activity. This hypothesis is supported by the clinical observation that PSO patients seldom develop cutaneous infections. Furthermore, PSO scales are a good source of skin-derived antimicrobial proteins [73]. We conclude that WFDC5 deserves further study to clarify its potential role in the pathogenesis of PSO.
The expression of TMEM45A was significantly suppressed by IFN-c while GPR115 expression was impaired by TNF-a/IL1-b, IFN-c, IL-4, IL-13 and IL-31 ( Figures 6 C and D). GPR87 was markedly induced by IL-4 and IL-22, whereas TNF-a/IL1-b stimulation resulted in significant suppression of expression ( Figure 6G). Disease-association analyses revealed a significant upregulation of TMEM45A in PSO and in AK ( Figure 7C). Similar regulation was observed for both skin-associated GPCRs, GPR115 and GPR87 ( Figure 7D and G). However, despite their induction in AK (which defines common precursors to sun-related SCC), their expression in SCC was reduced. Interestingly, TMEM45A was recently reported to suppress the progression of ductal carcinoma in situ to invasive breast cancer [59]. Additional studies are needed to address whether TMEM45A may play a similar role in the progression from AK to SCC. GRP87 is induced by the p53 tumor suppressor gene [69] and knockdown of GPR87 sensitizes breast and colon cancer cells to DNA damage-induced growth suppression in a p53-dependent manner. Our observations suggest that GPR87, GPR115 and/or TMEM45A may have similar functions in skin cancers. Interestingly, TMEM45A expression in primary human keratinocytes is induced upon differentiation [58]. These findings are in line with the results of our immunohistochemical analyses, showing increased TMEM45A staining from the basal layers of the epidermis to the stratum corneum (Figure 5c). Similar staining patterns were observed for GPR115 and, to a lesser extent, for WFDC5, indicating that these proteins are induced upon differentiation ( Figure 5A and E), with maximal expression in the stratum granulosum. Double-immunofluorescence analyses with filaggrin, a protein known to be expressed in the upper layers of the epidermis, demonstrated co-localization with TMEM45A and GPR115 in the most apical layers of the epidermis and in particular in the stratum corneum (see Figure S3).
To characterize differentiation-dependent regulation in more detail, we followed the expression levels of the three SAGs and four control genes in PHKs induced to terminally differentiate by confluence [74] or by varying Calcium ion levels [75]. qPCR analysis demonstrated induction of expression for all four differentiation marker genes as well as all three novel SAGs upon differentiation (see Figures S4, S5). These findings also correlate in the upregulation of all three genes in PSO, as the histomorphologic characteristics of PSO include excessive hyperkeratosis and acanthosis with abundant keratinocytes and dysregulated differentiation. This relationship is especially significant for WFDC5 which exhibits the largest increase in expression in both in vitro differentiation models and in PSO. The additional significant induction of GPR115 in AD points to a potential role for this receptor in the pathogenesis of inflammatory skin disorders and may be associated with the therapeutic effects of glucocorticosteroids in these diseases [65]. Figure 5. Detection of selected poorly characterized SAG encoded proteins in normal human skin and primary keratinocytes. Panels A-F; formalin-fixed paraffin embedded normal human skin sections were stained for DNA with DAPI and also with either SAG protein-specific (A, C, E), or isotype control antibodies (B, D, F) and visualized by immunofluorescent microscopy, 406 magnification. WFDC5, TMEM45A and GPR115 protein expression compared to beta actin was also detected in primary cultured human keratinocyte lysates using Western blotting (panels G, H and I respectively). doi:10.1371/journal.pone.0063949.g005 CDHR1 expression was suppressed by IFN-c, whereas we observed a marked induction by IL-31 and IL-22 ( Figure 6E). CDHR1 expression was suppressed in all skin diseases included in our screen, except PRU (Fig. 7E). CDHR1 is a photoreceptorspecific gene and its expression and function in skin is currently unknown [52,53]. CDHR1 expression in healthy skin covers a broad range from nearly zero to .4000 relative units. The physiological significance of this observation is unknown.
The proteinase inhibitor SERPINB7 also exhibits substantial variation of expression in normal skin and while our data reveal significant downregulation in PSO and SCC, and the role of SERPINB7 in these diseases is uncertain. Shiiba et al. demonstrated that SERPINB7 and other genes of the SERPIN gene cluster on chromosome 18q21 were significantly downregulated in cell lines derived from oral SCC [76] and concluded that these SERPIN members may represent tumor suppressor genes. Our data are consistent with this hypothesis. Finally, C5orf46 was downregulated in PSO and AD. Further studies are needed to identify the cellular origin of this protein and to gain further insights into its regulation and potential role in skin disorders.
Our study represents the most comprehensive comparative molecular profiling of normal human skin to date. The overall strategy of our analytical approach was validated by our results: among the top 100 SAGs, we identified 14 known genes not yet known to be expressed in skin, as well as another 6 skin-expressed uncharacterized transcripts. Further analyses of selected SAGs indicate that several of these genes exhibit altered expression in a several human skin diseases. Our observations strongly suggest that some of these genes participate in the pathogenesis of these human skin diseases.

Tissue Samples
All human biopsy skin samples were collected with written consent from the donors and with the approval of the Medical Faculty of the University of Düsseldorf IRB. All samples received from Zoion Diagnostics were approved by an IRB in previous studies by the authors [7,8].

Gene Expression Data Acquisition and Analysis
Generation of the Body Index of Gene Expression database (BIGE) has been described [7]. Briefly, flash-frozen human tissue samples were obtained from multiple male and female human donors, between 3-5 hours post-mortem (Zoion Diagnostics, Hawthorne, NY), (see Table S2). Gene expression data was obtained using Affymetrix Human Genome U133 Plus 2.0 gene arrays and standard protocols (Affymetrix, Santa Clara, CA). Background adjustment, signal normalization, and summarization were done using the Robust Multi-array Average (RMA) algorithm [77] in ArrayAssist software (Iobion Labs, La Jolla, CA). To identify genes expressed exclusively or predominantly in the skin [skin-associated genes (SAGs)], we calculated the ratio between the means of the skin samples (n = 5) and the samples representing the remaining body tissues and cell types (n = 435) and selected genes with skin to body expression ratios $2.0.
As control samples for human spleen, kidney, brain and liver total RNA were obtained from Clontech, Mountain View, CA.

Immunohistochemistry and Immunofluorescence
Sections from human skin were routinely fixed with formalin and embedded in paraffin. After demasking with Protease XXIV (Sigma-Aldrich, St. Louis, MO) in 20 mM Tris pH 7.9 the skin sections were incubated with antibodies directed against human WFDC5 (mouse polyclonal; Abcam, Cambridge, UK), TMEM45A (rabbit polyclonal; Sigma-Aldrich), GPR115 (rabbit polyclonal; Abcam) and filaggrin (mouse monoclonal; Abcam) or appropriate isotype control antibodies. Primary antibody binding was detected using the secondary antibodies donkey anti-mouse or rabbit IgG AviAlexa 594 (Invitrogen, Carlsbad, CA) and donkey anti-mouse IgG AviAlexa 488 (Invitrogen). Sections were fixed with Fluoromount G. Immunoreactions were detected by use of a microscope (Axiovert 200 M) (Zeiss, Jena, Germany) using Axiovision 4.7 (Zeiss).

Western Blot Analysis of WFDC5, TMEM45A and GPR115 Expressed in Human Keratinocytes
Human keratinocytes were harvested and resuspended in lysis buffer (50 mM HEPES [pH 7.4], 150 mM NaCl, 1% NP40, 0.1% SDS, 1% Triton X-100 and protease inhibitor cocktail). The lysates were centrifuged at 5,000 g for 5 min to remove cellular debris, resolved by polyacrylamide electrophoresis and analyzed by Western blotting using the appropriate antibodies according to the manufacturers protocol. Protein blots were developed using Western blot detection solution and hyperfilm (Amersham ECL detection reagents; GE Healthcare, Pittsburgh, PA).  Figure S3 Detection of (A) transmembrane protein 45A (TMEM45A) and (B) G protein-coupled receptor 115 (GPR115) in normal human skin sections co-stained with flagillin (FLG), a stratum granulosum-associated protein. Formalin-fixed paraffin embedded normal human skin sections from the same donor (tissue block) were stained with DAPI+ isotype control antibody to determine background staining and visualize cell nuclei and also with either TMEM45A or GPR115 and FLG-specific antibodies and visualized by immunofluorescent microscopy, 406 magnification. Merging of the TMEM45A or GPR115 images with the FLG image shows co-localization of both novel proteins with the known stratum granulosum-associated protein.

Supporting Information
(PPTX) Figure S4 Expression of known skin differentiation marker genes and selected SAGs increases during keratinocyte differentiation. Semi-quantitative PCR analysis of gene expression (mean 6 SD) of four known skin differentiation marker genes (top row) and three selected SAGs (WFDC5, WAP four-disulfide core Table S6 Top 300 skin-associated genes (SAGs) ranked by fold change of expression compared to the mean of the remaining 104 adult human tissue and cell types in the BIGE. Classes: k/k, known skin-associated gene; k/n, known gene not previously associated with skin; n/n, uncharacterized (novel) gene. Functional classification as reported in the literature or, by inference, from bioinformatics analysis, marked with an asterisk. (XLSX)