Characterization of a new pathway that activates lumisterol in vivo to biologically active hydroxylumisterols

Using LC/qTOF-MS we detected lumisterol, 20-hydroxylumisterol, 22-hydroxylumisterol, 24-hydroxylumisterol, 20,22-dihydroxylumisterol, pregnalumisterol, 17-hydroxypregnalumisterol and 17,20-dihydroxypregnalumisterol in human serum and epidermis, and the porcine adrenal gland. The hydroxylumisterols inhibited proliferation of human skin cells in a cell type-dependent fashion with predominant effects on epidermal keratinocytes. They also inhibited melanoma proliferation in both monolayer and soft agar. 20-Hydroxylumisterol stimulated the expression of several genes, including those associated with keratinocyte differentiation and antioxidative responses, while inhibiting the expression of others including RORA and RORC. Molecular modeling and studies on VDRE-transcriptional activity excludes action through the genomic site of the VDR. However, their favorable interactions with the A-pocket in conjunction with VDR translocation studies suggest they may act on this non-genomic VDR site. Inhibition of RORα and RORγ transactivation activities in a Tet-on CHO cell reporter system, RORα co-activator assays and inhibition of (RORE)-LUC reporter activity in skin cells, in conjunction with molecular modeling, identified RORα and RORγ as excellent receptor candidates for the hydroxylumisterols. Thus, we have discovered a new biologically relevant, lumisterogenic pathway, the metabolites of which display biological activity. This opens a new area of endocrine research on the effects of the hydroxylumisterols on different pathways in different cells and the mechanisms involved.

on the temperature and UVB dose. T3 is the most photoreactive product and undergoes UVB-driven conversion to L3 via pre-D3, making L3 the major photoisomer generated with prolonged UVB exposure 6,7 .
The current view is that D3 is the only important biological regulator derived from photolysis of 7DHC. After its activation to 1,25(OH) 2 D3, D3 not only regulates calcium homeostasis, but displays anticancer activities and also has important pleiotropic effects which include regulation of proliferation, differentiation, apoptosis, and immune and endocrine activities 5 In contrast, it has been assumed that L3 affects neither calcium metabolism nor has any other significant biological activity. Its formation has been used to explain why UVB-induced cutaneous production of pre-D3 does not lead to systemic D3 intoxication with prolonged UVB exposure 5, 8, 9 . Until recently, it was believed that vitamin D activation only involved the sequential hydroxylations at C25 and C1: D3 → 25(OH)D3 → 1,25(OH) 2 D3 7, 10,11 . Surprisingly, the finding that CYP11A1 (the first enzyme of steroidogenesis 12 ) can hydroxylate the D3 side chain at C17, C20, C22 and C23 [13][14][15][16] and the D2 side chain at C20, C17 and C24 17,18 , has revealed new pathways of D activation. These pathways operate in vivo 13,19,20 with the major intermediates and products being detectable in human serum and epidermis 19 . The intermediates/products are biologically active 21,22 , acting as partial agonists on the vitamin D receptor (VDR) 22,23 and as inverse agonists on retinoic acid orphan receptors (ROR)α and γ 24 . RORs are expressed in normal and pathological skin; therefore, binding of these novel secosteroids 19,25 to RORs is likely relevant to the regulation of biological functions in this organ 26,27 .
Our surprising recent finding that purified and reconstituted CYP11A1 can hydroxylate L3, producing 20(OH)L3, 22(OH)L3, 20,22(OH) 2 L3 and pL 40 , has formed the basis for the current study on the production of L3 metabolites, in vivo, in the human body and for the testing of their phenotypic activity in skin cells, and for an effort to identify candidate receptors.

Results and Discussion
In vivo detection of lumisterol metabolites. Lumisterol (L3). Analysis of extracts of human epidermis and serum, and pig adrenals, by LC/qTOF-MS alongside the corresponding standards, demonstrated the presence of 7DHC, D3 and L3 (Fig. 1). Epidermal samples were obtained from 13 patients including 6 African-Americans (AA) and 7 Caucasians (C), and sera were from a separate group of 13 individuals (12 C and 1 Hispanic). These were analyzed by LC/qTOF-MS to determine concentrations of 7DHC and L3 ( Table 1). The concentration of 7DHC in human serum (~55 nM) is comparable to that reported for mouse serum 41 . In the epidermis, the level of 7DHC was 92-times higher than that of L3 but their concentrations in the serum were almost equal ( Table 1). The serum concentration of L3 is 10-times higher than that previously reported for D3 19 and the level in the epidermis is 13-times higher than that reported for D3 19 . The content of L3 and of its precursor 7DHC, in the epidermis or serum, showed no significant differences in relation to age, gender and race for these small sample sets (supplemental Figure 1). This is the first evidence that lumisterol formed in the skin can circulate in the serum and potentially accumulate in steroidogenic tissues such as the adrenal gland. Hydroxylumisterols. Lumisterol derivatives in extracts from the human epidermis and serum were first separated on a C18 column (25 cm long) with an acetonitrile in water gradient, as detailed in the materials and methods. The fractions with retention times (RT) corresponding to authentic standards of hydroxyderivatives of lumisterol or 20(OH)7DHC were then collected. The individual fractions were analyzed by UPLC on an Agilent Zorbax Eclipse Plus C18 column connected to a Xevo ™ G2-S qTOF, with a methanol gradient as described by us previously 19 . Thus, identification of hydroxymetabolites that had identical masses was based on their RT compared to standards in two different solvent systems. From analysis of the extracted ion chromatograms (EIC) (see legend to Fig. 2 for monitored ions), we identified monohydroxy-metabolites of 7DHC and lumisterol with RT corresponding to chemically or enzymatically synthesized 20(OH)7DHC, 20(OH)L3, 22(OH)L3 and 24(OH)L3 standards ( Fig. 2A). A dihydroxylumisterol was also identified with a RT corresponding to 20,22(OH) 2 L3 in the EIC of extracts of epidermis and serum (Fig. 2B). We also detected an ion at m/z = 367.3 (M + H-2H 2 O) + with a RT corresponding to 20(OH)Chol in extracts of human epidermis and serum (Fig. 2) Figure 3).

Continued
Based on previous enzymatic studies 40 , the expression of CYP11A1 in the skin 28 and our current data, we conclude that epidermal 20(OH)L3, 22(OH)L3, 24(OH)L3 and 20,22(OH) 2 L3 must be products of cutaneous CYP11A1-mediated metabolism of L3. Serum levels of these hydroxylumisterols may not only depend on their production rate in the skin, but also on their production rate from circulating L3 by the adrenal gland, the organ with the highest CYP11A1 concentration in the body 12 . Since the skin is intermittently exposed to UVB 42 , and the absorption of its energy by the unsaturated B ring of 7DHC 6-8 or its hydroxyderivatives 43 will ultimately lead to their transformation to compounds with the D3 or L3 configuration, UVB-induced transformation of locally produced 20(OH)7DHC, 22(OH)7DHC, or 20,22(OH) 2 7DHC 30, 31 could represent an additional source of the detected hydroxylumisterol compounds.
Pregna-lumisterol (pL) and its hydroxyl-pL metabolites. We recently reported that pL is produced from L3 by purified CYP11A1 and fragments of adrenals, by cleavage of the lumisterol side chain 40 . Analysis of extracts of human epidermis and serum, and pig adrenals showed species corresponding to the retention times of standards of pL, 17(OH)pL, and 17,20(OH) 2 Fig. 4), indicating that pL is metabolized by steroidogenic enzymes within the adrenal. Epidermal production of pL and its subsequent hydroxylation is consistent with the steroidogenic activity of the skin, as discussed recently 44 . An additional source of pL, 17(OH) pL and 17,20(OH) 2 pL in the epidermis could be the UVB-induced photochemical transformation of 7DHP, 17(OH)7DHP, or 17,20(OH) 2 7DHC, respectively 28,32,34 , since these 5,7-dienes can be produced in the skin 30,31 .
Tissue and serum concentrations of L3 metabolites. In the epidermis, the CYP11A1-derived mono-hydroxylumis-terols were present at significantly higher (p < 0.01) levels than the parental L3; however, the reverse was observed in serum ( Table 1). The 22(OH)L3 concentration was comparable to the 20(OH)L3 level in epidermis but lower than 20(OH)L3 in serum, possibly reflecting different rates of clearance. The concentration of 20,22(OH) 2 L3 was significantly lower (p < 0.01) than either 22(OH)L3 or 20(OH)L3. pL showed the lowest concentration of the CYP11A1-derived metabolites analyzed in both epidermis and serum, consistent with it being only a minor product of CYP11A1 action on L3 40 . Full statistical analysis is provided in supplemental Table 1. Analyses of levels of L3 metabolites for gender, age and racial group showed no statistical difference in epidermal or serum concentrations (not shown). The serum concentrations of 20(OH)L3 is 9 times higher than that previously reported for 20(OH)D3 while the 22(OH)L3 concentration is similar to that reported for 22(OH)D3 19 . Epidermal levels of 20(OH)L3 and 22(OH)L3 are 20-30 higher than those reported for 20(OH)D3, consistent with the more efficient metabolism of lumisterol than D3 by human CYP11A1 13,40 .

Groups Genes
Gene expression (fold change)  Table 2. Cyclophilin B was used as internal control. Data represent mean ± SD; n = 3. Significance was analyzed using student-t test, *P < 0.05; **P < 0.01; ***p The expression of a panel of genes by cultured human epidermal keratinocytes exposed to 20(OH)L3 was examined at the mRNA level (Table 2). Significantly, 20(OH)L3 stimulated the expression of genes encoding differentiation program markers (INL, LOR, FLG, TGM1, KRT1, KRT5, KRT10, and KRT14) and antioxidative enzymes (CAT, GPX1, GSR, GSTP1, SOD1, SOD2, GCS, TXNRD1, and TRN) ( Table 2). Potential pro-differentiation and anti-oxidative effects were further confirmed by stimulation of INL and SOD2 (Mn-SOD) protein expression (supplemental Figure 5). Of additional interest is the increased expression of BNIP3 (which is implicated in differentiation and maintenance of epidermal keratinocytes 46 ), increased expression of ICAM (which plays a role in immune surveillance in basal cell carcinoma 47,48 and in wound healing 49 ), and increased expression of gelsolin (which is implicated in apoptosis 50 , cancer, inflammation, infection and aging 51 ). 20(OH) , Inhibition of proliferation in monolayer assed by MTS assay. After 24 h of culture, the cells were exposed to graded concentrations of hydroxylumisterols suspended in Ham's F10 plus 10% charcoaltreated FBS. After 48 h, the plates were used for MTS assay performed at 490 nm. (B), Inhibition of growth in soft agar (anchorage independent growth). Melanoma cells were suspended in medium containing 0.4% agarose (American Bioanalytical, Natick, MA) and 5% charcoal-treated FBS, and seeded at 1,000 cells/well in a 0.8% agar layer in 24-well plates and treated with the graded concentrations of the listed compounds which were freshly added every 72 h over 13 days 54 . The colonies stained with MTT reagent (Promega, Madison, WI) were analyzed using the Cytation 5 Cell Imaging Multi-Mode Reader in three different z-planes and scored using Gen5 software 54 . Data represent means ± SE (n ≥ 3) where *p < 0.05, **p < 0.01 and ***p < 0.001 by the student t-test, and general ANOVA tests are shown.
L3 also enhanced the expression of CRH and URN, which control skin responses to stress 52 , and enhanced the expression of CYP1B1 which is involved in detoxification. Upregulation of the expression of the genes listed above indicates a role for hydroxylumisterols in the protective functions of the epidermis. With respect to immunomodulation and growth factors, regulation of these functions can be complex (Table 2) Since melanoma still represents a clinically challenging problem 53 , we evaluated the anti-melanoma activity of the hydroxylumisterols (Fig. 5A,B). 20(OH)L3, 22(OH)L3, 24(OH)L3, and 20,22(OH) 2 L3 markedly inhibited cell proliferation (Fig. 5A) with the structurally related 20(OH)Chol having no significant effect (not shown). At the same time point (48 h of incubation) hydroxylumisterols had no effect on the proliferation of normal melanocytes, with moderate inhibitory effects seen at 72 h, but only for 20(OH)L3 and 22(OH)L3 (supplemental Figure 6). All hydroxylumisterols inhibited the anchorage-independent melanoma growth in soft agar (Fig. 5B), indicating their antitumorogenic potential. They did not affect melanin production by melanoma cells (not shown). The effects on proliferation are similar to those described for CYP11A1-derived hydroxyvitamin D3-derivatives, which also showed strong anti-melanoma effects with weak or absent effects on normal melanocytes 55,56 . In addition, our previous studies demonstrated in vitro anti-melanoma activity of mono and dihydroxy pL compounds [33][34][35][36] . Thus, the novel hydroxyderivatives of lumisterol are good candidates for further testing of To measure the transactivation the cells were treated with graded concentrations of the hydroxylumisterols listed, and the RORE-mediated activation of the luciferase reporter activity was assayed with a Luciferase Assay Substrate kit (Promega) as described previously 24 . Assays were performed in triplicate. (B), RORα coactivator assay using LanthaScreen TR-FRET RORα Coactivator kit assay. RORα-LBD was added to graded concentrations of hydroxylumisterols followed by the addition of a mixture of peptide (TRAP220/DRIP2) and antibody (Tb-anti-GST). The reaction mixture was incubated at room temperature for 2 h and the TR-FRET ratio was calculated by dividing the fluorescein emission at 520 nm by the Terbium emission at 495 nm using Synergy neo2 (BioTek Instruments, Inc., Winooski, VT). Data represent means ± SE (n ≥ 3) where * p < 0.05, ** p < 0.01 and *** p < 0.001 student t-test; # p < 0.05, ## p < 0.01, ### p < 0.001 and #### p < 0.0001 by one-way ANOVA and general ANOVA tests are shown. (C), RORE luciferase assay in HaCaT keratinocytes. The cells were cotransfected with the reporter plasmids pGL4.27-(RORE) 5 and phRL-TK (Promega) using Lipofectamine (Invitrogen, Carlsbad, CA) following the manufacturer's protocol. After transfection, the cells were treated with hydroxylumisterols for 48 h. Luciferase reporter activity was measured using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI). Firefly and Renilla signals were read using Cytation 5 (BioTek Instruments, Inc., Winooski, VT), and the ratios were calculated. their therapeutic utilities using animal models of melanoma and patient-derived orthotopic xenografts (PDOX) models [57][58][59] .

Hydroxylumisterols can interact with RORα and RORγ. RORγ and RORα-mediated transactivation
assays. RORα and RORγ are expressed in the human skin 24,25 , while L3 analogs are structurally very similar to sterols that are examples of native ligands for RORs 26,60,61 . Therefore, we tested their effects on RORs using cell based and in vitro assays. First, using a previously described Tet-on CHO cell reporter system for analysis of RORγ and RORα-mediated transactivation 24 , we compared the inverse agonist activity of 20(OH)L3 with that of its structurally related D3 and sterol derivatives. Supplemental Figure 9A shows that 20(OH)L3 was the most potent inhibitor of RORγ-induced transcriptional activity, being less active on RORα. Additional tests on skin derived cells transfected with the RORE-LUC reporter showed a dose-dependent inhibition of luciferase activity by 20(OH)L3 (Supplemental Figure 9B). Interestingly, 20(OH)7DHC, a potential precursor to 20(OH)L3, was less potent (Supplemental Figure 9b).
Docking results for RORα and γ. The hydroxylumisterols and related compounds listed in Table 3 were docked into crystal structures of RORα and RORγ using the Glide docking in extra-precision mode (XP) (Schrödinger package). Preparation of structures and the docking protocol are described in the Supplemental information. Docking results are presented below.
RORα. The RORα ligand binding site with docked L3 is illustrated in Fig. 7a. The binding site of RORα is largely hydrophobic and shows structural complementarity with L3. Desolvation and formation of favorable non-polar interactions is the most significant contributor to the binding of L3. Similarly to L3, docked poses of (OH) n L3 compounds listed in Table 3 form non-polar contacts that are analogous to those present in the crystal structure between cholesterol and RORα residues. As illustrated in Fig. 7b,c, docked poses of cholesterol analogs and the (OH) n L3 series (Table 3) are approximately overlapping with the cholesterol co-crystallized in the RORα binding site. The residues displayed are predicted to contribute polar contacts with the ligands. L3 forms a hydrogen bonding interaction between its 3-hydroxyl group and the carbonyl backbone of Tyr380. Polar  interactions of cholesterol analogs involve hydrogen bonding through a tightly bound crystal water, as shown in Fig. 7b. In the (OH) n L3 series, all analogs hydrogen bond with the backbone carbonyl of Tyr380, an interaction shared among lumisterol analogs. Docked poses suggest that Cys323 and Cys396 may contribute to polar interactions as hydrogen bond donors to 22-hydroxyl and 24-hydroxyl groups, respectively. PL has a markedly lower docking score for RORα due to lack of a side chain, which in other lumisterol analogs contributes to binding through non-polar interactions. For the hydroxylated pL analogs, 17(OH)pL and 17,20(OH) 2 pL, two possible poses were obtained: one that is similar to the (OH) n L3 series and a second pose adopting the opposite or 'flipped' orientation, as illustrated in Fig. 7d. Compared to L3, these compounds are shorter, more polar and are capable of forming hydrogen-bonding interactions in either orientation. In the case of 17,20(OH) 2 pL, the 'flipped' pose may be more likely since it predicts an additional hydrogen bonding interaction via the 20-hydroxyl group (Fig. 7d). Interestingly, the theoretically deduced flipped positions for hydroxyl-pL compounds are further substantiated by their apparent agonistic activity on RORα as shown in supplemental Figure 10. The more favorable hydrogen bonding contribution as reflected in the improved Glide XP scores of 17,20(OH) 2 pL compared to pL (Table 3) is also supported by its higher potency in comparison to pL (supplemental Figure 10). Overall, docking results predict favorable binding of the (OH) n L3 series and hydroxylated pL analogs to RORα. RORγ. Analogously to docking results obtained at RORα, docked poses of cholesterol analogs in RORγ are overlapping closely with the co-crystallized 20-hydroxycholesterol. Similarly to the co-crystallized ligand, the 3-hydroxyl group in all docked cholesterol analogs participates in hydrogen bonding with the side chain of Gln286 and forms a water-bridged hydrogen bond with Arg367. As in RORα, the active site of RORγ is predominantly hydrophobic and ligand binding is primarily driven by non-polar interactions and desolvation of non-polar groups. Docking of the (OH) n L3 series (Table 3)   additional hydrogen bonding interactions. Thus, favorable binding of the (OH) n L3 series and hydroxylated pL is also predicted for RORγ, similarly to docking results obtained for the same series for RORα.
Co-crystallized inverse agonists in RORγ X-ray structures show disruption of a key hydrogen bonding interaction between H479 and Y502 while this interaction in agonist co-crystal structures is undisturbed or further stabilized by agonist ligands. W317 has also been identified as a key residue accessible to ligands that may contribute to propagating the effects of inverse agonists. Conformational changes triggered by the binding of inverse agonists are destabilization of helices 11, 11, 12, which leads to lack of coactivator recruitment. We have not considered using these inverse agonist bound RORγ structures for lumisterol series docking since these ligands The analog of this ligand lacking the one-carbon linker marked with a red arrow is a RORγ agonist. (c) RORγ crystal structures with inverse agonists are aligned onto the structure with PDB code 3KYT. The three key residues (as in Fig. 8b) are also shown, along with co-crystallized ligands from two structures: PDB codes 3KYT and 4WQP. Font colors of PDB codes listed correspond to the coloring of secondary structures and carbon atoms of co-crystallized ligands shown.
are significantly different in terms of shape and polarity from the lumisterol scaffold and ligand induced effects are important to reproduce optimal non-polar contacts in the RORγ binding site. Figure 8b illustrates the conformation of key residues for functional activity from three RORγ crystal structures with distinct inverse agonists bound (PDB codes 3B0W, 4NB6, 4WQP), based on their alignment onto the structure with PDB code 3KYT. Docked poses of the (OH) n L3 series are also displayed in comparison with the docked cholesterol (which is shown with thick bonds). In the docked pose of 24(OH)L3 the 24-hydroxyl group is close enough for hydrogen bonding interaction with H479. However, as shown in Fig. 8b, distinct conformations of the H479 side chain are possible in close proximity to both 22-and 24-hydroxyl groups of the (OH) n L3 series (while Y502 in helix 12 is disordered). Further, the hydroxylated aliphatic chain in the lumisterol series is overlapping with a functionally important region of the inverse agonist shown in the inset of Fig. 8b. This ligand induces flipping of the W317 side chain, which leads to displacement of Y502. Removal of a one-carbon linker group (marked with arrow) converts this inverse agonist into an RORγ agonist that shows no steric clashes with W317 (PDB: 4WPF). Alignment of the RORγ structures with inverse agonists onto RORγ with co-crystallized 20(OH)Chol (PDB: 3KYT) shows close overlap except for helices 11, 11' and 12 (Fig. 8c). Residues affected by these conformational differences that are accessible to docked lumisterol analogs are in the region of the key residues W317, H479, H502.
Considering the close proximity of hydroxyl group substituents in the (OH) n L3 series to the functionally important H479 we hypothesize that a possible mechanism of inverse agonism of these ligands is through hydrogen bonding with H479 and disruption of the H479 -Y502 interaction. Vitamin D receptor. Functional testing of binding to the VDR has shown that the hydroxylumisterols lack any effect on VDRE-transcriptional activity in HaCaT cells (Supplemental Figure 11) and do not bind to the genomic LBD of the VDR using the LanthaScreen TR-FRET competition kit (not shown). Docking of hydroxylumisterols into the genomic (G)-pocket of the VDR gave significantly poorer scores than those for 1,25(OH) 2 D3 and 1,25(OH) 2 L3 (Table 3). Therefore, we conclude that CYP11A1-derived hydroxylumisterols are not involved in the regulation of genomic VDR activity.
Surprisingly, we have found that the hydroxylumisterols stimulated translocation of the VDR from the cytoplasm to the nucleus but required relatively high concentrations (Supplemental Figure 12). However, these effects were significantly lower in comparison to 1,25(OH) 2 D3 and other hydroxyderivatives of vitamin D 23,62 . Therefore, we performed molecular modeling to predict whether the hydroxylumisterols can bind to the non-genomic site (A-pocket) of the VDR, a binding site for 1,25(OH) 2 L3 63, 64 , using docking simulations (see description in the Supplemental information). The resulting docked poses show favorable interactions in the A-pocket with Glide XP scores being comparable to or better than those for 1,25(OH) 2 L3 and 1,25(OH) 2 D3 (Supplemental Figures 13 and 14) (Table 3). Therefore, it is possible that the CYP11A1-derived hydroxylumisterols might act on the non-genomic A-pocket of the VDR. This exciting possibility deserves future investigation.

Conclusions
Lumisterol was previously considered to be a metabolically inactive end product of 7DHC exposure to high UVB energy, providing an explanation of why UVB-induced production of pre-D3 does not lead to systemic D3 intoxication 5,7 . The current study shows that this traditional view must be revised, since lumisterol not only enters the systemic circulation (having serum levels at ~5 × 10 The CYP11A1-derived hydroxylumisterols inhibit skin cell proliferation in a cell-type dependent fashion with pronounced effects on keratinocytes, and show anti-melanoma activity as well. 20(OH)L3, as a representative hydroxylumisterol, also stimulates expression of genes associated with keratinocyte differentiation and anti-oxidative programs. The functional data presented on the hydroxylumisterols, in conjunction with the previously described biological effects of pL analogs 34,36,39,65 , suggest that the novel lumisterogenic pathway might be involved in the regulation of cutaneous homeostasis. Their discovery also opens up exciting new areas for future research such as studies on the role of lumisterol derivatives in barrier function, photoprotection, skin cancer and studies on their possible therapeutic or adjuvant utility in the management of melanoma.
The cell based and in vitro analyses of activities of RORs, supported by molecular modeling, demonstrate that the hydroxylumisterols can act as ligands on RORα and RORγ. However, it is unlikely that they interact with the genomic site of the VDR.
In summary, this study reveals that a CYP11A1-mediated pathway of lumisterol metabolism occurs in vivo, the products of which have phenotypic/biological activities determined by their structure and cellular target.

Materials and Methods
Source of lumisterol derivatives. Lumisterol Use of tissues and serum samples. Collection of human or pig tissue and serum samples was approved by the Institutional Review Board (IRB) (Human Subject Assurance Number 00002301) and the Institutional Animal Care and Use Committee (IACUC) (Animal Welfare Assurance Number A3325-01) at the University of Tennessee Health Science Center (UTHSC) with details of protocols and collection of the samples previously described 19 . Pig adrenals were obtained from a female Landrace cross Large White pig, 2 years old. All methods were performed in accordance with the relevant guidelines and regulations, see below. Tissues or sera were extracted with organic solvents and stored at −80 °C 19 , prior to aliqots being taken for LC/MS analyses. These same samples have been used previously for vitamin D metabolism studies, as detailed in ref. 19.
Human skin samples (n = 13) were collected in Memphis during 2013 and 2014 from 7 males and 6 females that comprised 6 African-Americans (AA) and 7 Caucasians (C) whose age ranged from 30 to 90 years 66 . Human sera were collected on March 28, 2014 in Memphis from 13 volunteers (3 males and 10 females) comprising 12 C and 1 Hispanic who were 25-61 years old. The use of human skin was approved by the IRB at the UTHSC as an exempt protocol #4 (Dr. A. Slominski, P.I.). This protocol was classified for exempt status under 45CFR46.102 (f) in that it does not involve "human subjects" as defined therein 19 . Collection of human serum was approved by IRB protocol #7526 (Dr. A. Postlethwaite, P.I.). Informed consent was obtained from all subjects involved in this study and the samples were deidentified as previously described 19 .
Skin (foreskins) from AA that would normally be discarded were used to establish primary cultures of keratinocytes, melanocytes and dermal fibroblasts, and was approved by the IRB at the University of Alabama Birmingham. This protocol was identified as not subject to FDA regulation and not Human Subject Research (IRB protocol N150915001 -Endocrine Functions of the Skin -revised version).
Detection of lumisterol derivatives. Liquid chromatography and mass spectrometry (LC-MS) analyses followed protocols described previously 19 . For identification of lumisterol derivatives in extracted samples, we first separated the expected CYP11A1-derived hydroxylumisterols by HPLC using a Waters C18 column (250 × 4.6 mm, 5 μm particle size). The mobil)e phase used was a gradient of acetonitrile in water (40-100%) at a flow rate of 0.5 ml/min for 15 min followed by isocratic 100% acetonitrile for 30 min at a flow rate of 0.5 ml/min and then a flow rate of 1.5 ml/min for 20 min. Fractions with RT corresponding to the chemically or enzymatically synthesized standards (see section: Source of lumisterol derivatives) were collected and then subjected to UPLC [(Waters ACQUITY I-Class UPLC (ultra-performance liquid chromatography) system (Waters, Milford, USA)] on an Agilent Zorbax Eclipse Plus C18 column (2.1 × 50 mm, 1.8 µm particle size), connected to a Xevo ™ G2-S qTOF (quadrupole hybrid with orthogonal acceleration time-of-flight) tandem mass spectrometer (Waters, Milford, USA) as detailed previously 19 . The mobile phase for UPLC comprised a gradient of methanol in water containing 0.1% formic acid (20-60% for 3 min then 60-100% for 1 min), followed by isocratic 99.9% methanol plus 0.1% formic acid for 2.1 min, all at flow rate of 0.3 ml/min.
For quantification, the concentrations of L3 and related compounds were directly analyzed by LC-MS using two different conditions of LC, as described in the Figure legends. For L3, D3 and 7DHC, a Waters Atlantis dC18 column (100 × 4.6 mm, 5 μm particle size) was used with a gradient of methanol in water (85-100%) containing 0.1% formic acid for 20 min followed by 99.9% methanol and 0.1% formic acid for 10 min, at a flow rate of 0. Cell Culture. Normal human epidermal keratinocytes (NHEK) and melanocytes (NHEM) were grown in either keratinocyte media (Lonza Walkersville Inc., Walkersville, MD) or in melanocyte growth media (MGM) supplemented with either KGF or MGF (Lonza), respectively, while dermal fibroblasts were cultured in DMEM medium containing antibiotics and 10% charcoal-treated fetal bovine serum (ctFBS) as previously detailed 55,56,67,68 . Cells in the third passage were used for experiments. HaCaT immortalized keratinocytes were cultured in DMEM plus 5 or 10% FBS, while SKMel-188 melanoma cells were grown in Ham's F10 and 5 or 10% FBS as described before 55,68 . For experimental treatments, ctFBS was used as indicated.
Real-time polymerase chain reaction (qPCR). Briefly, total RNA was isolated from cultured normal keratinocytes and reverse transcribed into cDNA. qPCR data were generated as detailed previously 45 and described in the table legend.
Immunofluorescence in situ studies. Protein expression was measured by immunofluorescence (IF) following protocols previously described 45 . Briefly, HEM plated onto 96-well plates (see above) were further treated