Advertisement

Repigmentation of Human Vitiligo Skin by NBUVB Is Controlled by Transcription of GLI1 and Activation of the β-Catenin Pathway in the Hair Follicle Bulge Stem Cells

Open ArchivePublished:October 17, 2017DOI:https://doi.org/10.1016/j.jid.2017.09.040
      Vitiligo repigmentation is a complex process in which the melanocyte-depleted interfollicular epidermis is repopulated by melanocyte precursors from hair follicle bulge that proliferate, migrate, and differentiate into mature melanocytes on their way to the epidermis. The strongest stimulus for vitiligo repigmentation is narrow-band UVB (NBUVB), but how the hair follicle melanocyte precursors are activated by UV light has not been extensively studied. To better understand this process, we developed an application that combined laser capture microdissection and subsequent whole transcriptome RNA sequencing of hair follicle bulge melanocyte precursors and compared their gene signatures to that of regenerated mature epidermal melanocytes from NBUVB-treated vitiligo skin. Using this strategy, we found up-regulation of TNC, GJB6, and THBS1 in the hair follicle bulge melanocytes and of TYR in the epidermal melanocytes of the NBUVB-treated vitiligo skin. We validated these results by quantitative real-time–PCR using NBUVB-treated vitiligo skin and untreated normal skin. We also identified that GLI1, a candidate stem cell-associated gene, is significantly up-regulated in the melanocytes captured from NBUVB-treated vitiligo bulge compared with untreated vitiligo bulge. These signals are potential key players in the activation of bulge melanocyte precursors during vitiligo repigmentation.

      Abbreviations:

      bp (base pair), F-LCM (fluorescent laser capture microdissection), FC (fold change), HF (hair follicle), IE (interfollicular epidermis), NBUVB (narrow-band UVB), qRT-PCR (quantitative real-time–PCR), RNA-seq (RNA-sequencing)

      Introduction

      Vitiligo is a common depigmentation disorder, affecting 0.3–0.5% of the population worldwide. It is characterized by white patches of the skin, due to autoimmune destruction of epidermal melanocytes by melanocyte-reactive cytotoxic T cells (
      • Eby J.M.
      • Kang H.K.
      • Klarquist J.
      • Chatterjee S.
      • Mosenson J.A.
      • Nishimura M.I.
      • et al.
      Immune responses in a mouse model of vitiligo with spontaneous epidermal de- and repigmentation.
      ,
      • Rashighi M.
      • Harris J.E.
      Vitiligo pathogenesis and emerging treatments.
      ,
      • Spritz R.A.
      Modern vitiligo genetics sheds new light on an ancient disease.
      ). Vitiligo repigmentation is characterized by repopulation of melanocyte-depleted interfollicular epidermis (IE) with melanocyte precursors, mostly originating from the hair follicle (HF) (
      • Birlea S.A.
      • Costin G.E.
      • Roop D.R.
      • Norris D.A.
      Trends in regenerative medicine: repigmentation in vitiligo through melanocyte stem cell mobilization.
      ). The strongest stimulus for the activation of melanocyte precursors is narrow-band UVB (NBUVB), but this process has not yet been adequately studied in the HF. Our recent immunostaining study, using key functional markers, showed that the HF bulge of depigmented vitiligo skin is inhabited by melanocyte stem cells (DCT(+)/C-KIT(–)) and melanoblasts (DCT(+)/C-KIT(+)), which proliferate, migrate, and differentiate during NBUVB treatment on their way to epidermis (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Spoelstra N.S.
      • Kechris K.J.
      • Robinson S.E.
      • et al.
      Narrow band ultraviolet B treatment for human vitiligo is associated with proliferation, migration, and differentiation of melanocyte precursors.
      ). The proportion of melanocyte precursor populations slightly increased in the HF bulge after NBUVB treatment, indicating melanocyte activation. To more thoroughly characterize the repigmentation process, we developed an application that, to our knowledge, was previously unreported that combined laser capture microdissection of skin cells isolated from vitiligo patients with RNA-sequencing (RNA-seq). With this application, we tested the gene signature of the HF bulge (the source of melanocyte precursors) and of the IE (the repopulated site) in NBUVB-treated vitiligo skin. Our strategy consisted of a rapid immunostaining protocol, followed by fluorescent laser capture microdissection (F-LCM) of melanocytes and, separately, of keratinocytes from the HF bulge and epidermal basal layer. This was followed by RNA isolation, whole-transcriptome RNA-seq, and gene expression analysis. Our RNA-seq analysis identified the integrin pathway as the top pathway activated in the bulge melanocytes of NBUVB-treated vitiligo skin, compared with NBUVB-treated vitiligo epidermis, and β-catenin as the top upstream transcription regulator. We also found up-regulation of TNC, GJB6, and THBS1 in the bulge melanocytes and of TYR in the epidermal melanocytes of NBUVB-treated vitiligo skin, results validated by quantitative real-time–PCR (qRT-PCR). We also identified that GLI1, a candidate stem cell-associated gene, is significantly up-regulated in melanocytes captured from the NBUVB-treated vitiligo bulge compared with the untreated vitiligo bulge. These signals and pathway may have regulatory roles in the activation of bulge melanocyte precursors during the vitiligo repigmentation process.

      Results

      We used frozen biopsy samples from untreated and NBUVB-treated vitiligo patients to perform bulge mapping, rapid immunostaining, F-LCM of specific skin cells, and RNA-seq analysis.

      Alignment and mapping of RNA-seq reads

      Whole-transcriptome RNA-seq, performed with a HiSeq 2000 sequencer (Illumina Inc., San Diego, CA), generated on average 3.2 and 3.6 million single-read sequences per sample from each studied site (HF bulge and IE, respectively), with an average length of 250 base pairs (bp) without adaptors and of 325 bp with adaptors. The median total raw reads for the bulge and IE samples were approximately 5.3 million and 5.9 million, respectively, and the median passing rates for the bulge and epidermis were 84.6% and 85.6%, respectively, indicating good data quality. We discarded samples that did not meet the control criteria because of quality, presence of contaminants formed by adapter-adapter ligation, and presence of reads without insert tags.

      RNA-seq gene expression profiles show cell-specific and site-specific F-LCM capture differences

      We isolated RNA from both melanocyte and keratinocyte samples. The purpose of including the keratinocytes was to show the specificity of our capture. After the validation step, we focused this study on the melanocyte material of the NBUVB-treated vitiligo skin. The total RNA from melanocyte samples and separately from keratinocyte samples was isolated by F-LCM from six NBUVB-treated vitiligo patients and subjected to RNA-seq followed by gene expression analysis. From the principal component analysis (PCA), we observed a good segregation of gene expression values between (i) the melanocytes captured from the NBUVB-treated vitiligo bulge versus the melanocytes from the NBUVB-treated vitiligo IE (see Supplementary Figure S1a online), (ii) the keratinocytes captured from the NBUVB-treated vitiligo bulge versus keratinocytes from the treated vitiligo IE (see Supplementary Figure S1b), (iii) the melanocytes versus keratinocytes captured from the NBUVB-treated vitiligo IE (see Supplementary Figure S1c), (iv) the melanocytes versus keratinocytes captured from the NBUVB-treated vitiligo bulge (see Supplementary Figure S1d). Additionally, we examined the expression of melanocyte-specific genes and keratinocyte-specific genes in our samples to determine cell specificity of capture (Figure 1a–d). We found significant enrichment of the melanocyte-specific genes in melanocytes compared with keratinocytes and significant enrichment of the keratinocyte-specific genes in keratinocytes compared with melanocytes in both NBUVB-treated vitiligo epidermis (Figure 1c) and NBUVB-treated vitiligo bulge (Figure 1d). Based on the results of principal component analysis and cell-type–specific gene expression analysis, we concluded that our F-LCM method is effective and accurate for selectively isolating RNA material from distinct populations of melanocytes and adjacent keratinocytes.
      Figure 1
      Figure 1Heatmaps of melanocyte (MC)- and keratinocyte (KC)-specific genes in the RNA-sequencing data. Heatmaps of RNA expression levels of melanocyte-specific and keratinocyte-specific genes in the vitiligo skin (a) of NBUVB-treated IE and (b) of NBUVB-treated bulge (n = 6). We found up-regulation of melanocyte-specific genes in melanocyte samples (green squares, right-sided upper corner) and keratinocyte-specific genes in keratinocyte samples (green squares, left-sided lower corner). The red squares represent down-regulated melanocyte-specific genes. Bar plots represent the gene enrichment in the (c) interfollicular epidermis and (d) bulge ± standard error of the mean (n = 6). In both NBUVB-treated IE and NBUVB-treated bulge, we found a significant enrichment of melanocyte-specific genes in the melanocyte samples (upper panels of c [green bars] and of d [yellow bars], with expression values set to 1-fold in keratinocytes) and a significant enrichment of keratinocyte-specific genes in the keratinocytes samples (lower figures of c [white bars], and d [red bars], with expression values set to 1-fold in melanocytes). (P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001). Each bar represents mean and standard deviation (n = 3). IE, interfollicular epidermis; NBUVB, narrow-band UVB.

      RNA-seq shows differentially expressed genes in the melanocyte samples of the HF bulge compared with epidermis of the NBUVB-treated vitiligo skin

      Next, we focused our analysis on characterization of melanocyte samples captured from different anatomic regions (IE and HF bulge) of six NBUVB-treated vitiligo patients. Of the total 54,009 genes assessed by RNA-seq, 39 (0.07%) were differentially expressed in the melanocyte samples from NBUVB-treated vitiligo bulge compared with the patient-matched melanocyte samples from NBUVB-treated vitiligo IE, including 27 significantly up-regulated genes (0.05%) and 12 significantly down-regulated genes (0.02%) (false discovery rate-adjusted Q-value ≤ 0.05). The top 10 differentially expressed protein-coding genes in the melanocyte samples are summarized in Table 1. We selected for qRT-PCR validation the top five differentially expressed genes, among which TNC (Q = 2.0 × 10–2, fold change [FC] = 22.7), THBS1 (Q = 2.7× 10–2, FC = 20.2), TM9SF3 (Q = 2.7 × 10–2, FC = 2.5), and GJB6 (Q = 2.9 × 10–5, FC = 27.7) were up-regulated in the melanocytes of the NBUVB-treated vitiligo bulge, whereas TYR (Q = 2.7 × 10–2, FC = 9.1) was up-regulated in the melanocytes of the NBUVB-treated vitiligo IE. We confirmed that TNC (Padjusted = 7.1× 10–3, FC = 65.8), THBS1 (Padjusted = 1.7 × 10–2; only amplified in melanocyte samples from the bulge), and GJB6 (Padjusted = 4.9 × 10–2; FC = 185.3) were up-regulated in bulge melanocytes of NBUVB-treated vitiligo skin and that TYR (Padjusted = 2.0× 10–3, FC = 65.0) was up-regulated in epidermal melanocytes of NBUVB-treated vitiligo skin (Figure 2a). In the IE melanocytes, we could not detect THBS1 transcript amplification in four NBUVB-treated patients, or GJB6 transcript amplification in two of the four NBUVB-treated patients.
      Table 1Top genes differentially expressed in the RNA-sequencing study
      Top 10 protein encoding genes differentially expressed in the melanocyte samples from the NBUVB-treated vitiligo bulge compared with melanocyte samples from the NBUVB-treated interfollicular epidermis. Fold changes, P-values, false discovery rate-adjusted P-values (Q-values), and gene functions are provided.
      GeneFold Change

      Bulge
      P-ValueQ-ValueKnown FunctionReference
      TNC22.74.0 × 10–60.020Promotes β-catenin–mediated transcription in the presence of Wnt3a

      Highly expressed during embryonic development of humans, mice
      • Hendaoui I.
      • Tucker R.P.
      • Zingg D.
      • Bichet S.
      • Schittny J.
      • Chiquet-Ehrismann R.
      Tenascin-C is required for normal Wnt/β-catenin signaling in the whisker follicle stem cell niche.


      • Imanaka-Yoshida K.
      • Aoki H.
      Tenascin-C and mechanotransduction in the development and diseases of cardiovascular system.
      THBS120.21.4 × 10–50.027Downstream effector of p53, which coordinates UV-induced pigmentation; highly expressed in embryonic stem cells
      • Sundaram P.
      • Hultine S.
      • Smith L.M.
      • Dews M.
      • Fox J.L.
      • Biyashev D.
      • et al.
      p53-responsive miR-194 inhibits thrombospondin-1 and promotes angiogenesis in colon cancers.


      • Liu Y.
      • Shin S.
      • Zeng X.
      • Zhan M.
      • Gonzalez R.
      • Mueller F.J.
      • et al.
      Genome wide profiling of human embryonic stem cells (hESCs), their derivatives and embryonal carcinoma cells to develop base profiles of U.S. Federal government approved hESC lines.
      TYR–9.11.7 × 10–50.027Involved in melanin biosynthesis; highly expressed by regenerated epidermal melanocytes in the NBUVB-treated vitiligo
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Spoelstra N.S.
      • Kechris K.J.
      • Robinson S.E.
      • et al.
      Narrow band ultraviolet B treatment for human vitiligo is associated with proliferation, migration, and differentiation of melanocyte precursors.
      TM9SF32.52.2 × 10–50.027Marker of tumor invasion
      • Oo H.Z.
      • Sentani K.
      • Sakamoto N.
      • Anami K.
      • Naito Y.
      • Oshima T.
      • et al.
      Identification of novel transmembrane proteins in scirrhous-type gastric cancer by the Escherichia coli ampicillin secretion trap (CAST) method: TM9SF3 participates in tumor invasion and serves as a prognostic factor.
      GJB627.73.3 × 10–50.029Mutated in Clouston syndrome, which is associated with impairment of hair growth; expressed in human skin and mouse embryo
      • Lamartine J.
      • Munhoz Essenfelder G.
      • Kibar Z.
      • Lanneluc I.
      • Callouet E.
      • Laoudj D.
      • et al.
      Mutations in GJB6 cause hidrotic ectodermal dysplasia.


      • Fujimoto A.
      • Kurban M.
      • Nakamura M.
      • Farooq M.
      • Fujikawa H.
      • Kibbi A.G.
      • et al.
      GJB6, of which mutations underlie Clouston syndrome, is a potential direct target gene of p63.
      SAMD572.63.7 × 10–50.029Stem cell gene in the mouse bulge
      • Kadaja M.
      • Keyes B.E.
      • Lin M.
      • Pasolli H.A.
      • Genander M.
      • Polak L.
      • et al.
      SOX9: a stem cell transcriptional regulator of secreted niche signaling factors.
      CTNND2188.15.6 × 10–50.029Promotes disruption of adherens junction by E-cadherin, inducing cell migration
      • Zhang Y.V.
      • Cheong J.
      • Ciapurin
      • McDermitt D.J.
      • Tumbar T.
      Distinct self-renewal and differentiation phases in the niche of infrequently dividing hair follicle stem cells.
      TRPS112.45.8 × 10–50.029Regulates epithelial proliferation in the mouse embryo
      • Fantauzzo K.A.
      • Kurban M.
      • Levy B.
      • Christiano A.M.
      Trps1 and its target gene Sox9 regulate epithelial proliferation in the developing hair follicle and are associated with hypertrichosis.
      LAMB4-26.96.1 × 10–50.029Identified overexpressed in malignant melanoma
      • Liu W.
      • Peng Y.
      • Tobin D.J.
      A new 12-gene diagnostic biomarker signature of melanoma revealed by integrated microarray analysis.
      SOX918.96.2 × 10–50.029Stem cell marker detected in >80% of malignant melanomas

      Stem cell gene in the human bulge
      • Jo A.
      • Denduluri S.
      • Zhang B.
      • Wang Z.
      • Yin L.
      • Yan Z.
      The versatile functions of Sox9 in development, stem cells, and human diseases.


      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      Abbreviations: NVUVB, narrow-band UVB.
      1 Top 10 protein encoding genes differentially expressed in the melanocyte samples from the NBUVB-treated vitiligo bulge compared with melanocyte samples from the NBUVB-treated interfollicular epidermis. Fold changes, P-values, false discovery rate-adjusted P-values (Q-values), and gene functions are provided.
      Figure 2
      Figure 2Expression of top genes differentially expressed in the NBUVB-treated vitiligo bulge melanocytes and NBUVB-treated epidermis melanocytes. (a) qRT-PCR confirmation of the top differentially expressed genes resulted from the RNA-sequencing study: TNC, GJB6, and THBS1 were significantly up-regulated in the melanocyte samples from the NBUVB-treated bulge, whereas TYR was significantly up-regulated in the melanocyte samples from the NBUVB-treated IE (paired t-test, Padjusted < 0.05) (n = 4 NBUVB-treated vitiligo patients). Each bar represents mean and standard deviation (n = 3). Expression values were set to 1-fold in the interfollicular epidermis for TNC and THBS1 and in the bulge for TYR and GJB6. (b) qRT-PCR gene expression analysis comparing the expression values of TNC, GJB6, THBS1, and TYR in the bulge melanocyte samples collected from NBUVB-treated vitiligo patients (n = 6), untreated vitiligo patients (n = 4), and normal control subjects (n = 6). The expression values of the four genes did not vary significantly between the groups tested (one-way analysis of variance, Tukey post hoc test: adjusted P > 0.05 for all pairwise comparisons). Expression values were set to 1-fold in the untreated vitiligo samples. Each bar represents mean and standard deviation (n = 3). NBUVB, narrow-band UVB; qRT-PCR, quantitative real-time–PCR.

      qRT-PCR shows differentially expressed genes in the melanocyte samples from the bulge versus epidermis of normal untreated control skin

      We tested the expression of top genes (that were found by the RNA-seq study and were validated by qRT-PCR) in melanocytes captured from normal untreated skin of six healthy control individuals (primer sequences listed in Supplementary Table S1 online). We identified a similar expression trend with that observed in the NBUVB-treated vitiligo skin: up-regulation in bulge melanocytes of TNC (Padjusted = 2.4 × 10–2, FC = 11.8), GJB6 (Padjusted = 1.0× 10–3, FC = 30.5), and THBS1 (Padjusted = 0.15, only amplified in 4/6 bulge samples) and up-regulation in epidermal melanocytes of TYR (Padjusted = 7.0× 10–4, FC = 13.4) (See Supplementary Figure S2 online).
      Next, we tested whether the expression of TNC, GJB6, THBS1, and TYR varied in the bulge melanocytes captured from untreated vitiligo skin (n = 4), NBUVB-treated vitiligo skin (n = 6), and normal skin (n = 6). We did not find significant variation either among the three groups tested or among any of the paired comparisons (Figure 2b).

      Pathway analysis, upstream regulators, and disease functions

      To better understand the biology and functional relationship among differentially expressed genes in our RNA-seq data, we performed pathway analysis using the Ingenuity Pathway Analysis tool (Qiagen Inc., Germantown, MD; available at http://www.ingenuity.com/products/ipa) and a combined melanocyte and keratinocyte data set of 1,873 differentially expressed genes (P < 0.05), comparing gene expression values between samples captured from the NBUVB-treated vitiligo bulge and samples captured from the NBUVB-treated vitiligo IE. We found 15 canonical pathways activated and one down-regulated in the bulge compared with the IE (Table 2). The top activated pathway in the bulge (with the highest Z-score = 3.16 and the lowest P-value = 1.0 × 10–11) was Integrin signaling, and its component genes that were differentially expressed (P < 0.05) are provided in Supplementary Table S2 online. The top putative upstream regulator in the bulge was CTNNB1 encoding β-catenin (Z-score = 3.61; P-value = 2.0 × 10–16) (see Supplementary Table S3 online), and the top activated cellular function was cellular movement (Z-score = 3.30; P-value = 1.4 × 10–42) (see Supplementary Table S4 online).
      Table 2Top canonical pathways in the RNA-sequencing study
      Top canonical pathways generated by combined datasets of melanocyte and keratinocyte samples from the NBUVB-treated vitiligo bulge versus NBUVB-treated vitiligo interfollicular epidermis (differentially expressed genes, P < 0.05). P-values, Z-scores, the number of up-regulated or down-regulated genes in each pathway, gene functions, and reference articles are provided.
      Ingenuity Canonical PathwaysP-ValueZ-ScoreDown-Regulated, n (%)Up-Regulated, n (%)FunctionReference
      Integrin signaling1.0 × 10–103.1628/201 (4)37/201 (18)Melanoma progression and metastasis
      • Kuphal S.
      • Bauer R.
      • Bosserhoff A.K.
      Integrin signaling in malignant melanoma.
      Paxillin signaling2.1 × 10–92.6004/101 (4)24/101 (24)Decreased paxillin impairs adequate generation of pro-melanoma signals
      • Velasco-Velázquez M.A.
      • Salinas-Jazmín N.
      • Mendoza-Patiño N.
      • Mandoki J.J.
      Reduced paxillin expression contributes to the antimetastatic effect of 4-hydroxycoumarin on B16-F10 melanoma cells.
      ILK signaling3.6 × 10–83.5697/186 (4)31/186 (17)Pro-proliferative roles on several types of cancers, including melanoma
      • Dai D.L.
      • Makretsov N.
      • Campos E.I.
      • Huang C.
      • Zhou Y.
      • Huntsman D.
      • et al.
      Increased expression of integrin-linked kinase is correlated with melanoma progression and poor patient survival.
      Mouse embryonic stem cell pluripotency2.0 × 10–72.8584/95 (4)20/95 (21)Pro-proliferative roles
      • Niwa H.
      How is pluripotency determined and maintained?.
      Regulation of eIF4 and p70S6K signaling2.2 × 10–72.1384/145 (3)27/145 (19)Pro-proliferative roles
      • Flynn A.
      • Proud C.G.
      The role of eIF4 in cell proliferation.
      Rac signaling3.2 × 10–72.4497/104 (7)18/104 (17)Controls melanocyte dendricity
      • Scott G.
      • Leopardi S.
      The cAMP signaling pathway has opposing effects on Rac and Rho in B16F10 cells: implications for dendrite formation in melanocytic cells.
      Signaling by rho family GTPases2.2 × 10–62.66710/234 (4)30/234 (13)Controls melanocyte dendricity
      • Scott G.
      • Leopardi S.
      The cAMP signaling pathway has opposing effects on Rac and Rho in B16F10 cells: implications for dendrite formation in melanocytic cells.
      Calcium signaling2.7 × 10–62.1327/178 (4)26/178 (15)Regulator of keratinocyte differentiation
      • Bikle D.D.
      • Xie Z.
      • Tu C.L.
      Calcium regulation of keratinocyte differentiation.
      EIF2 signaling5.9 × 10–63.6382/184 (1)31/184 (17)Key component of the translation initiation system in living cells
      • Stolboushkina E.A.
      • Garber M.B.
      Eukaryotic type translation initiation factor 2: structure- functional aspects.
      Role of NANOG in mammalian embryonic stem cell pluripotency1.3 × 10–52.6466/111 (5)17/111 (15)NANOG functions as trigger for the reprogramming process in human cells
      • Silva J.
      • Nichols J.
      • Theunissen T.W.
      • Guo G.
      • van Oosten A.L.
      • Barrandon O.
      • et al.
      Nanog is the gateway to the pluripotent ground state.
      Melanocyte development and pigmentation signaling2.0 × 10–52.0655/84 (6)14/84 (17)Melanocyte proliferation and differentiation
      • Costin G.E.
      • Hearing V.J.
      Human skin pigmentation: melanocytes modulate skin color in response to stress.
      PAK signaling4.8 × 10–52.3573/89 (3)16/89 (18)Promotes α-MSH/UVB-induced melanogenesis
      • Yun C.Y.
      • You S.T.
      • Kim J.H.
      • Chung J.H.
      • Han S.B.
      • Shin E.Y.
      • et al.
      p21-activated kinase 4 critically regulates melanogenesis via activation of the CREB/MITF and β- catenin/MITF pathways.
      Wnt/Ca+ pathway7.8 × 10–52.8872/56 (4)12/56 (21)Putatively involved in the invasion and metastasis of melanoma
      • Yang Y.
      • Qian Q.
      Wnt5a/Ca (2+) /calcineurin/nuclear factor of activated T signaling pathway as a potential marker of pediatric melanoma.
      IGF-1 signaling1.6 × 10–42.4965/97 (5)14/97 (14)Regulates keratinocyte shape and migration
      • Haase I.
      • Evans R.
      • Pofahl R.
      • Watt F.M.
      Regulation of keratinocyte shape, migration and wound epithelialization by IGF-1- and EGF-dependent signalling pathways.
      FGF signaling8.1 × 10–42.5003/85 (4)13/85 (15)Regulates cellular proliferation, survival, migration, and differentiation
      • Turner N.
      • Grose R.
      Fibroblast growth factor signalling: from development to cancer.
      RhoGDI signaling1.1 × 10–5–2.1919/173 (5)22/173 (13)Rho family GTPase inhibitor; suppresses melanoma cell growth
      • Wang P.
      • Xu S.
      • Wang Y.
      • Wu P.
      • Zhang J.
      • Sato T.
      • et al.
      GM3 suppresses anchorage-independent growth via Rho GDP dissociation inhibitor beta in melanoma B16 cells.
      Abbreviation: MSH, melanocyte stimulating hormone.
      1 Top canonical pathways generated by combined datasets of melanocyte and keratinocyte samples from the NBUVB-treated vitiligo bulge versus NBUVB-treated vitiligo interfollicular epidermis (differentially expressed genes, P < 0.05). P-values, Z-scores, the number of up-regulated or down-regulated genes in each pathway, gene functions, and reference articles are provided.

      Candidate stem cell-associated gene expression analysis

      To describe the stem cell gene signature in the bulge, we examined the RNA-seq expression of 189 genes from a published stem cell differentiation panel (nCounter Virtual Stem Cell Gene Set, NanoString Technologies, Seattle, WA). The top differentially expressed genes in melanocytes from the NBUVB-treated bulge compared with the NBUVB-treated epidermis were FZD7 (P = 4.1 × 10–4, Q-value = 0.05, FC = 46.9) and GLI1 (P = 7.7 × 10–4; Q-value = 0.06; FC = 49.6) (Figure 3a). We further validated these results by qRT-PCR using new laser capture rounds of melanocytes from NBUVB-treated vitiligo patients (n = 7) (FZD7: Padjusted = 4.0 × 10–3, FC = 39.5 and GLI1: Padjusted = 9.7 × 10–4; only amplified in bulge samples)] (Figure 3b). Next, we examined whether NBUVB modulates the expression of FZD7 and GLI1 transcripts in the bulge melanocytes isolated from NBUVB-treated vitiligo skin (n = 7), untreated vitiligo skin (n = 6), or control skin (n = 6). We found that GLI1 was significantly up-regulated in the bulge melanocytes of NBUVB-treated vitiligo skin compared with untreated vitiligo skin (Padjusted = 2.7 × 10–3, FC = 9.2) but did not significantly vary in other comparisons (Figure 3c). In addition, FZD7 transcript was not significantly modulated by NBUVB after multiple testing correction (Padjusted = 0.5, FC = 1.5) (Figure 3c).
      Figure 3
      Figure 3Candidate stem cell-associated gene expression analysis in the RNA-sequencing data and GLI1 confirmation in the hair follicle bulge of vitiligo skin. (a) Candidate stem cell-associated gene expression analysis of the RNA-sequencing data: FZD7 and GLI1 were the top candidate stem cell-associated genes differentially expressed in the melanocyte samples from the bulge versus interfollicular epidermis collected from NBUVB-treated vitiligo patients (n = 6). (b) qRT-PCR confirmation study of FZD7 and GLI1 after the RNA-sequencing study: both genes were significantly up-regulated in the melanocyte samples from the bulge compared with the IE of NBUVB-treated vitiligo patients (n = 6) (∗∗Padjusted < 0.01; ∗∗∗Padjusted < 0.001). RNA analyzed was extracted from melanocyte samples after new rounds of laser capture microdissection. Expression values were set to 1-fold in the IE for FZD7 and in the bulge for GLI1. (c) Analysis of NBUVB effects on FZD7 and GLI1 in the human HF bulge of vitiligo skin: expression analysis of FZD7 and GLI1 in melanocyte samples from the bulge of NBUVB-treated vitiligo skin (n = 7), untreated vitiligo skin (n = 6), and normal control skin (n = 6). GLI1 expression showed significant up-regulation in the bulge of NBUVB-treated versus untreated vitiligo samples (∗∗Padjusted < 0.01) and did not show significant variation in the other comparisons. Expression values were set to 1-fold in the untreated vitiligo samples. Each bar represents mean and standard deviation (n = 3). (d–f) Immunostaining analysis of GLI1 localization in the hair follicle bulge. Double fluorescent immunostaining of formalin-fixed paraffin-embedded transverse sections using an anti-GLI1 antibody (red) in combination with anti-DCT antibody (green, labels melanocytes) in the (d) untreated vitiligo bulge and (e) NBUVB-treated vitiligo bulge. White scale bars = 50 μm. Red arrows indicate DCT+/GLI1+ cells; green arrows indicate DCT+/GLI1 cells. White dotted lines indicate areas of higher magnification (inset). GLI1 is also expressed in the bulge keratinocytes, which are DCT/GLI1+ cells. (f) Signal intensity analysis for the anti-GLI1 antibody. The average intensity of the anti-GLI1 antibody signal is significantly higher in the bulge melanocytes of the NBUVB-treated vitiligo skin (n = 7) compared with untreated vitiligo skin (n = 7) (∗∗∗∗P = 4.4 × 10–4, +1.5-fold). IE, interfollicular epidermis; NBUVB, narrow-band UVB; qRT-PCR, quantitative real-time reverse transcriptase–PCR.
      Next, to validate the expression of GLI1 in the bulge melanocytes in response to NBUVB, we performed immunohistochemistry using an anti-GLI1 antibody combined with anti-DCT antibody (melanocyte specific) using skin from new NBUVB-treated vitiligo patients (n = 6) and untreated patients (n = 6). GLI1 protein was expressed in the bulge melanocytes of untreated vitiligo skin, carrying the DCT(+)/GLI1(+) phenotype, and in the bulge keratinocytes of untreated vitiligo skin, carrying the DCT(–)/GLI1(+) phenotype (Figure 3d); GLI1 was also expressed in the bulge melanocytes and keratinocytes of NBUVB-treated vitiligo skin (Figure 3e) and in the melanocytes and keratinocytes of NBUVB-treated and untreated epidermis (data not shown). We observed that GLI1(–) melanocytes, carrying the DCT(+)/GLI1(–) phenotype, and GLI1(–) keratinocytes, carrying the DCT(–)/GLI1(–) phenotype, were more numerous in the untreated bulge (Figure 3d), whereas GLI1(+) melanocytes DCT(+)/GLI1(+) and GLI1(+) keratinocytes DCT(–)/GLI1(+) were more numerous in the NBUVB-treated bulge (Figure 3e). Intensity analysis of microscopic images showed a stronger anti-GLI1 antibody signal in the bulge melanocytes of NBUVB-treated vitiligo skin compared with the signal in untreated vitiligo skin (P = 4.4 × 10–4, FC = 1.5) (Figure 3f).
      We further examined in the RNA-seq data the expression of two gene targets that directly interact with GLI1: SOX9 (encoding SOX9) (
      • Deng W.
      • Vanderbilt D.B.
      • Lin C.C.
      • Martin K.H.
      • Brundage K.M.
      • Ruppert J.M.
      SOX9 inhibits β-TrCP-mediated protein degradation to promote nuclear GLI1 expression and cancer stem cell properties.
      ) and CTNNB1 (encoding β-catenin) (
      • Liao X.
      • Siu M.K.
      • Au C.W.
      • Chan Q.K.
      • Chan H.Y.
      • Wong E.S.
      Aberrant activation of hedgehog signaling pathway contributes to endometrial carcinogenesis through beta-catenin.
      ), together with other key components of the Wnt/β-catenin pathway (FZD8, SFRP1, WIF1, and FZD8) (see Supplementary Figure S3 online). We found that the expression values of SOX9 transcript (Q = 2.0 × 10–2, P = 6.2 × 10–5, FC = 18.9) and SFRP1 transcript (Q = 3.0 × 10–2, P = 8.3 × 10–5, FC = 43.8) were significantly up-regulated in the melanocyte samples of NBUVB-treated vitiligo bulge compared with those of NBUVB-treated vitiligo epidermis, whereas expression differences of CTNNB1, FZD8, and WIF1 did not surpass the false discovery rate adjustment threshold (which was Q ≤ 0.05) (see Supplementary Figure S3).

      Discussion

      In this study, we report an application (that, to our knowledge, is previously unreported) consisting of rapid immunostaining combined with F-LCM (to isolate RNA from specific cells [melanocytes], located at specific sites [bulge and epidermis]) and with RNA-seq. In a previous study, we showed that the RNA captured from the HF bulge and epidermal melanocytes of the NBUVB-treated vitiligo and normal skin is of satisfactory quality for qRT-PCR analysis (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      ). Our current application offers a better characterization of melanocyte populations in the regenerated epidermis and the bulge of the NBUVB-treated vitiligo skin, because it uses RNA-seq technology that has deeper coverage and higher sensitivity than qRT-PCR. We showed that the resulting RNA from the enriched melanocyte or keratinocyte samples is of satisfactory quality and reliability for RNA-seq based on (i) the high rate of passed sequences in all RNA-seq samples; (ii) significant enrichment of melanocyte-specific genes in the melanocyte samples and of keratinocyte-specific genes in the keratinocyte samples, in both epidermis and bulge of NBUVB-treated vitiligo skin (Figure 1a–d); (iii) good segregation of gene signatures between the mature epidermal reservoir and the bulge stem cell reservoir of the NBUVB-treated vitiligo skin, in both melanocyte and keratinocyte samples (see Supplementary Figure S1a and b); (iv) good segregation of gene signatures between the melanocyte and keratinocyte samples, in both epidermis and bulge of the NBUVB-treated skin (see Supplementary Figure S1c and d).
      We found by RNA-seq analysis and confirmed by qRT-PCR that TNC, GJB6, and THBS1 were significantly up-regulated in the melanocyte samples from the NBUVB-treated vitiligo bulge, whereas TYR was significantly up-regulated in the melanocyte samples of the NBUVB-treated epidermis (Figure 2a). We observed similar expression trends in the samples of untreated normal skin (see Supplementary Figure S2). All of these data indicate that we have successfully isolated RNA from stem-like melanocytic cells in the bulge that are TNC(+), GJB6(+), and THBS1(+) and from differentiated melanocytes that are TYR(+) in the IE. TNC encodes an extracellular matrix protein that promotes CTNNB1 transcription in the presence of Wnt3a, which is required for Wnt/β-catenin signaling in the mouse HF bulge (
      • Hendaoui I.
      • Tucker R.P.
      • Zingg D.
      • Bichet S.
      • Schittny J.
      • Chiquet-Ehrismann R.
      Tenascin-C is required for normal Wnt/β-catenin signaling in the whisker follicle stem cell niche.
      ). GJB6 encodes a gap junction protein responsible for Clouston syndrome, which is associated with hair growth impairment (
      • Lamartine J.
      • Munhoz Essenfelder G.
      • Kibar Z.
      • Lanneluc I.
      • Callouet E.
      • Laoudj D.
      • et al.
      Mutations in GJB6 cause hidrotic ectodermal dysplasia.
      ). THBS1 is a downstream effector of p53 (
      • Sundaram P.
      • Hultine S.
      • Smith L.M.
      • Dews M.
      • Fox J.L.
      • Biyashev D.
      • et al.
      p53-responsive miR-194 inhibits thrombospondin-1 and promotes angiogenesis in colon cancers.
      ), which controls pigmentation in response to UV (
      • Birlea S.A.
      • Costin G.E.
      • Roop D.R.
      • Norris D.A.
      Trends in regenerative medicine: repigmentation in vitiligo through melanocyte stem cell mobilization.
      ). TNC, GJB6, and THBS1 in the bulge were not up-regulated by the NBUVB treatment. It is possible that NBUVB can exert short-lived effects on these genes, which should be the subject of more comprehensive future study at different time points.
      We found that TYR was the top gene significantly up-regulated in the epidermal melanocytes of NBUVB-treated vitiligo skin, providing further validation of our method. TYR encodes tyrosinase, the major enzyme of melanin biosynthesis, and is highly expressed in melanocytes undergoing differentiation (
      • Cichorek M.
      • Wachulska M.
      • Stasiewicz A.
      • Tymińska A.
      Skin melanocytes: biology and development.
      ), such as epidermal melanocytes. Using immunostaining and qRT-PCR, we previously identified TYR up-regulation in the epidermal melanocytes of NBUVB-treated vitiligo skin and unremarkable TYR expression in the bulge melanocyte precursors (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Spoelstra N.S.
      • Kechris K.J.
      • Robinson S.E.
      • et al.
      Narrow band ultraviolet B treatment for human vitiligo is associated with proliferation, migration, and differentiation of melanocyte precursors.
      ,
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      ). Indeed, our RNA-seq study confirmed the low expression of TYR in the NBUVB-treated vitiligo bulge melanocytes (Figure 2a), with no significant increase in response to NBUVB (Figure 2b) compared with untreated vitiligo. Although the epidermal melanocytes in this study are presumed to be fully differentiated in contrast to those in the bulge, it is likely that a subpopulation of epidermal melanocytes in treated skin could be only partially differentiated and thus share expression of some precursor genes with the bulge melanocytes. Thus, it is possible that some melanocyte stem cell markers are present in both populations. In future studies, we might compare the expression values from epidermal melanocytes in NBUVB-treated skin with values in the NBUVB-treated nonlesional vitiligo skin, assuming that these melanocytes are enriched in genes associated with regeneration process, having pro-migratory, pro-proliferative, and pro-differentiation roles. Another useful comparison is with the expression values in the epidermal melanocytes of healthy normal skin, assuming that these melanocytes are fully differentiated.
      In addition, our application identified the integrin pathway as the top pathway up-regulated in the NBUVB-treated bulge melanocytes, along with 14 other significant pathways (Table 1), many of which have been previously associated with either melanoma or with melanocyte/keratinocyte proliferation, migration, differentiation, and stemness control.
      We found by RNA-seq (Figure 3a) and confirmed by qRT-PCR (Figure 3b) that GLI1, a candidate stem cell-associated gene, was significantly up-regulated in the melanocyte precursors of the NBUVB-treated vitiligo bulge compared with the melanocytes of NBUVB-treated vitiligo epidermis. Also using qRT-PCR (Figure 3c) and immunostaining (Figure 3d), we identified that GLI1 was significantly modulated by NBUVB in the bulge, with higher transcript and protein expression observed in the melanocyte samples from the NBUVB-treated vitiligo bulge compared with melanocyte samples from the untreated vitiligo bulge. Our immunostaining study showed that the GLI1 was up-regulated in both melanocytes and keratinocytes of NBUVB-treated and untreated vitiligo bulge, indicating that this may be an important bulge-associated gene involved in the response to the NBUVB in the HF niche. GLI1, an effector of the Shh pathway, is required for melanocyte proliferation and for melanoma growth and metastasis (
      • Barakat B.
      • Yu L.
      • Lo C.
      • Vu D.
      • De Luca E.
      • Cain J.E.
      • et al.
      Interaction of smoothened with integrin-linked kinase in primary cilia mediates Hedgehog signalling.
      ,
      • Santiago-Walker A.
      • Herlyn M.
      The ups and downs of transcription factors in melanoma.
      ). The Shh and Wnt/β-catenin pathways can interact through GLI1’s regulation of both nuclear localization and transcriptional activity of β-catenin (
      • Liao X.
      • Siu M.K.
      • Au C.W.
      • Chan Q.K.
      • Chan H.Y.
      • Wong E.S.
      Aberrant activation of hedgehog signaling pathway contributes to endometrial carcinogenesis through beta-catenin.
      ). In this study, β-catenin was identified as the top upstream regulator in the melanocyte precursors of NBUVB-treated vitiligo bulge (see Supplementary Table S3). This suggests that in vitiligo patients, NBUVB can act as an essential co-activator of CTNNB1 transcription in bulge melanocytes, which interact with GLI1 to induce melanocyte proliferation, migration, and differentiation (Figure 4a–c).
      Figure 4
      Figure 4Hypothetical model of the effects of NBUVB on the bulge and epidermis of human vitiligo skin during repigmentation. (a, b) Scheme of melanocyte-induced proliferation, migration, and differentiation through GLI1 activation. In human vitiligo skin, bulge-specific genes GJB6, THBS1, TNC, and FZD7 (blue shapes) are expressed in the melanocyte precursors in the bulge at similar levels (a) before and (b) after NBUVB treatment. These genes are proposed to be involved in cellular adhesion and in maintaining stemness in the bulge melanocyte precursors through β-catenin signaling. Epidermal melanocytes are absent in the interfollicular epidermis (IE) of untreated vitiligo skin, illustrated by the decreased height of all triangles, and their absence in the IE in a. In contrast, TYR (brown triangles) is expressed at low levels in both untreated and NBUVB-treated bulge melanocyte precursors and is expressed at significantly higher levels in the regenerated epidermal melanocytes of NBUVB-treated vitiligo skin, indicating active melanogenesis. GLI1 expression (red triangles) is significantly higher in the NBUVB-treated bulge melanocytes of vitiligo skin compared with untreated vitiligo skin, suggesting a role for GLI1 in the repigmentation process. We identified SOX9 constitutively expressed by the melanocyte precursors in the human bulge of normal and vitiligo skin (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      ). Activation of FZD7 (a Wnt receptor) was associated with reduced phosphorylation of β-catenin and its nuclear accumulation (
      • Merle P.
      • Kim M.
      • Herrmann M.
      • Gupte A.
      • Lefrançois L.
      • Califano S.
      • et al.
      Oncogenic role of the frizzled-7/beta-catenin pathway in hepatocellular carcinoma.
      ). Gli1 can induce the accumulation of transcriptionally active β-catenin in cell nuclei (
      • Li X.
      • Deng W.
      • Lobo-Ruppert S.M.
      • Ruppert J.M.
      Gli1 acts through Snail and E-cadherin topromote nuclear signaling by β-catenin.
      ). β-catenin can enhance the transcriptional activity of GLI1 (
      • Maeda O.
      • Kondo M.
      • Fujita T.
      • Usami N.
      • Fukui T.
      • Shimokata K.
      • et al.
      Enhancement of GLI1-transcriptional activity by beta-catenin in human cancer cells.
      ,
      • Song L.
      • Li Z.Y.
      • Liu W.P.
      • Zhao M.R.
      Crosstalk between Wnt/β-catenin and Hedgehog/Gli signaling pathways in colon cancer and implications for therapy.
      ,
      • Varnat F.
      • Siegl-Cachedenier I.
      • Malerba M.
      • Gervaz P.
      Ruiz i Altaba A. Loss of WNT TCF addiction and enhancement of HH GLI1 signalling define the metastatic transition of human colon carcinomas.
      ). GLI1 induction in the bulge melanocyte and keratinocyte precursors by NBUVB is supposedly influenced by nuclear translocation of β-catenin, and by interaction with SOX9, with a shift in the balance of maintaining stemness to greater differentiation, proliferation, and migration. (c) Cross-talk activation between β-catenin and GLI1 under NBUVB. β-catenin can enhance the transcriptional activity of GLI1 (
      • Maeda O.
      • Kondo M.
      • Fujita T.
      • Usami N.
      • Fukui T.
      • Shimokata K.
      • et al.
      Enhancement of GLI1-transcriptional activity by beta-catenin in human cancer cells.
      ), likely through regulation of the β-catenin key downstream targets IGF2BP1 (CRD-BP) and myc (
      • Noubissi F.K.
      • Elcheva I.
      • Bhatia N.
      • Shakoori A.
      • Ougolkov A.
      • Liu J.
      • et al.
      CRD-BP mediates stabilization of betaTrCP1 and c-myc mRNA in response to beta-catenin signalling.
      ). Once activated by β-catenin, IGF2BP1 and myc can activate GLI1, likely by binding to the GLI1 mRNA coding region (
      • Noubissi F.K.
      • Goswami S.
      • Sanek N.A.
      • Kawakami K.
      • Minamoto T.
      • Moser A.
      • Grinblat Y.
      • Spiegelman V.S.
      Wnt signaling stimulates transcriptional outcome of the Hedgehog pathway by stabilizing GLI1 mRNA.
      ,
      • Varnat F.
      • Siegl-Cachedenier I.
      • Malerba M.
      • Gervaz P.
      Ruiz i Altaba A. Loss of WNT TCF addiction and enhancement of HH GLI1 signalling define the metastatic transition of human colon carcinomas.
      ). GLI1 also regulates β-catenin at the transcription level, through its targets SNAIL, WNT, and SFRP1 (
      • Song L.
      • Li Z.Y.
      • Liu W.P.
      • Zhao M.R.
      Crosstalk between Wnt/β-catenin and Hedgehog/Gli signaling pathways in colon cancer and implications for therapy.
      ). GLI1 can induce SNAIL expression, which then interacts with β-catenin and stimulates its expression activity at the transcription level (
      • Song L.
      • Li Z.Y.
      • Liu W.P.
      • Zhao M.R.
      Crosstalk between Wnt/β-catenin and Hedgehog/Gli signaling pathways in colon cancer and implications for therapy.
      ) and at the protein level (
      • Li X.
      • Deng W.
      • Lobo-Ruppert S.M.
      • Ruppert J.M.
      Gli1 acts through Snail and E-cadherin topromote nuclear signaling by β-catenin.
      ). SFRP1 and Wnt have opposite effects on β-catenin signaling, as a negative feedback or as a control mechanism after GLI1 induction, to prevent overactivation of the Wnt/β-catenin pathway. SFRP1 is a melanocyte stem gene that we identified up-regulated in the bulge of NBVUB-treated vitiligo skin and normal skin (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      ). Experimental evidence used for this figure was generated in either human models; in Gli1 transgenic mice; or on in vitro experiments performed on rat kidney cells, murine LLC-11 hepatocellular carcinoma cells, and human stomach, colon, and lung cancer cells. ##, cytoplasmic β-catenin; ◊◊, nuclear β-catenin; +, baseline stimulation in the untreated vitiligo bulge; ++, increased stimulation in the NBUVB-treated vitiligo bulge; ∗, significant upregulation of GLI1 in the NBUVB-treated vitiligo bulge vs. untreated vitiligo bulge; MC, melanocyte; NBUVB, narrow-band UVB.
      Moreover, our RNA-seq analysis shows that the increased expression of GLI1 was associated with increased expression of SOX9 (see Supplementary Figure S3) in the melanocyte samples from the NBUVB-treated vitiligo bulge compared with NBUVB-treated vitiligo epidermis. This supports our previous qRT-PCR study that identified SOX9 as a stem cell gene in the bulge of untreated normal skin and NBUVB-treated vitiligo skin (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      ). These findings suggest that GLI1 and SOX9 may interact in the melanocyte precursors of the human HF bulge to influence proliferation and/or stemness. A SOX9-GLI1 functional relationship has been previously reported: SOX9 was involved in the maintenance of GLI1 in RK3E cells, and repression of SOX9 in pancreatic ductal adenocarcinoma cells resulted in the reduction of endogenous GLI1 levels. Furthermore, mRNA analysis of Panc-1 cells transfected with SOX9-small interfering RNA showed synchronous loss of both SOX9 and GLI1 signals (
      • Deng W.
      • Vanderbilt D.B.
      • Lin C.C.
      • Martin K.H.
      • Brundage K.M.
      • Ruppert J.M.
      SOX9 inhibits β-TrCP-mediated protein degradation to promote nuclear GLI1 expression and cancer stem cell properties.
      ), and ectopic Sox9 expression induced the expression of Gli1 in maturing chondrocytes of chick embryos (
      • Zeng L.
      • Kempf H.
      • Murtaugh L.C.
      • Sato M.E.
      • Lassar A.B.
      Shh establishes an Nkx3.2/Sox9 autoregulatory loop that is maintained by BMP signals to induce somatic chondrogenesis.
      ).
      In summary, we present a working model on the effects of NBUVB in the human vitiligo bulge (Figure 4a and b). Under NBUVB, the intercellular adhesion molecules (TNC, GJB6, and THBS1) and FZD7 work together with GLI1, through β-catenin nuclear translocation and SOX9 constitutive expression in the bulge, to modulate the balance between stemness and activation of melanocyte precursors in the bulge of vitiligo skin. Furthermore, the GLI1 activation process is followed by melanocyte proliferation, migration, and differentiation (Figure 4b). Our previous comprehensive reviews have summarized essential signals and pathways that promote epidermal melanocyte regeneration by follicular melanocytes and that regulate the balance between stemness and differentiation states of melanocytes and keratinocytes (
      • Birlea S.A.
      • Costin G.E.
      • Roop D.R.
      • Norris D.A.
      Trends in regenerative medicine: repigmentation in vitiligo through melanocyte stem cell mobilization.
      ,
      • Birlea S.A.
      • Goldstein N.B.
      • Norris D.A.
      Repigmentation through melanocyte regeneration in vitiligo.
      ), including in vitro studies and studies on animal and human models of repigmentation. These genes and pathways include p53 and Wnt/β-catenin pathways; integrins, cadherins, tetraspanins, and metalloproteinases; and transforming growth factor-β (TGF-β) and its effector PAX3. Future functional studies focused on the bulge stem cell genes identified in this study (Figure 4) will examine their molecular interaction with the pathways and signals already known to be involved in the repigmentation process.
      Our long-term goal is to improve the treatment outcome for vitiligo through identification of molecules that activate melanocyte precursors in the HF bulge. Because existing F-LCM techniques were not adequate for these studies (
      • Amoh Y.
      • Aki R.
      • Hamada Y.
      • Niiyama S.
      • Eshima K.
      • Kawahara K.
      • et al.
      Nestin-positive hair follicle pluripotent stem cells can promote regeneration of impinged peripheral nerve injury.
      ,
      • Ohyama M.
      • Terunuma A.
      • Tock C.L.
      • Radonovich M.F.
      • Pise-Masison C.A.
      • Hopping S.B.
      • et al.
      Characterization and isolation of stem cell-enriched human hair follicle bulge cells.
      ,
      • Xu X.
      • Lyle S.
      • Liu Y.
      • Solky B.
      • Cotsarelis G.
      Differential expression of cyclin D1 in the human hair follicle.
      ), we developed this application (which, to our knowledge, is previously unreported) to isolate and characterize melanocyte populations in the human HF and epidermis. There are several important characteristics of this model: (i) the ability to perform a rapid immunostaining protocol that minimizes RNA degradation instead of the usual overnight antibody incubation, (ii) the ability to capture RNA from specific cells located in specific anatomic sites, and (iii) the ability to capture the cells from their natural microenvironment. Even so, there are limitations of our method (modest RNA contamination from neighbor cells, small number of cells captured, and RNA degradation), which we have tried to minimize. Contamination with RNA material of neighboring cells was minimized by using a small-diameter laser pulse (of ∼16 μm). The small number of cells harvested was not a major liability in previous studies that successfully analyzed RNA after F-LCM (
      • Dolter K.E.
      • Braman J.C.
      Small-sample total RNA purification: laser capture microdissection and cultured cell applications.
      ,
      • Nakamura N.
      • Ruebel K.
      • Jin L.
      • Qian X.
      • Zhang H.
      • Lloyd R.V.
      Laser capture microdissection for analysis of single cells.
      ,
      • Vandewoestyne M.
      • Goossens K.
      • Burvenich C.
      • Goossens K.
      • Burvenich C.
      • Van Soom A.
      • et al.
      Laser capture microdissection: should an ultraviolet or infrared laser be used?.
      ). We decreased RNA degradation by optimizing the rapid immunostaininig protocol, using the Illumina HiSeq 2000 sequencer, which processed only short RNA fragments (150–350 bp), and designing short amplicons (≤150 bp in our case) for qRT-PCR confirmation (
      • Vandewoestyne M.
      • Goossens K.
      • Burvenich C.
      • Goossens K.
      • Burvenich C.
      • Van Soom A.
      • et al.
      Laser capture microdissection: should an ultraviolet or infrared laser be used?.
      ).
      Taken together, our data suggest that our F-LCM samples are enriched in target cells that are anatomically distinct and that we can successfully use RNA-seq to analyze changes of melanocyte signals associated with NBUVB-induced repigmentation in human vitiligo. The HF bulge signals and pathways activated in the melanocyte precursors presented here (Figure 4a and b) are, to our knowledge, previously unreported in the HF bulge of human vitiligo or in untreated normal skin. Most of them have been associated with cell proliferation and migration in different malignancies or in organ formation in utero, suggesting that repigmentation initiation in the bulge shares common aspects with carcinogenesis and embryonic development.

      Materials and Methods

      Tissue sample collection and processing for RNA studies

      Skin biopsy samples from the untreated depigmented skin of six vitiligo patients and from the repigmented skin of 10 vitiligo patients treated with NBUVB were collected in the Dermatology Clinic at University of Colorado Hospital (see Supplementary Table S5 online). The study was approved by the Human Subjects Committee at the University of Colorado, and written informed consent was obtained from all subjects. Untreated normal human skin from six control subjects was either purchased from the National Disease Resource Interchange (Philadelphia, PA) or obtained from the Skin Cancer Biorepository at the University of Colorado (Aurora, CO).

      HF bulge mapping

      To locate the HF bulge in our samples, frozen or formalin-fixed paraffin-embedded transverse sections were immunostained with a combination of anti-keratin 15 and anti-desmin antibodies every 10th slide (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      ).

      Rapid fluorescent immunostaining and F-LCM

      Frozen sections were immunostained using the NKI-beteb antibody. F-LCM was performed with an infrared laser microdissection system, under direct fluorescent microscopic visualization (
      • Goldstein N.B.
      • Koster M.I.
      • Hoaglin L.G.
      • Wright M.J.
      • Robinson S.E.
      • Robinson W.A.
      • et al.
      Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
      ). From each sample, we captured 100 NKI-beteb+ melanocytes from the epidermal basal layer, 100 NKI-beteb+ cells from the bulge-outer root sheath, and 100 adjacent keratinocytes (NKI-beteb–) from each of these two regions.

      RNA extraction, amplification, whole-transcriptome RNA-seq, and data analysis

      We analyzed the RNA isolated from melanocyte and, separately, keratinocyte samples laser captured from the repigmented skin of six vitiligo patients treated with NBUVB for 3–4 months. Total RNA extracted was amplified. Good RNA quality required an RNA integrity number of 6.5 and a concentration of greater than 500 pg/μl. Approximately more than 1 ng total RNA was used for the final libraries and was sequenced with Illumina HiSeq 2000. The RNA-seq reads from each library were aligned to the human genome (hg19). Transcripts were assembled from the aligned reads, and gene expression levels were estimated. From the global transcription, we tested the accuracy of our data by principal component analysis, which assessed the overall similarity between samples. We used a paired t test in R to analyze the genes differentially expressed in the melanocytes captured from the NBUVB-treated bulge compared with the NBUVB-treated IE. To correct for multiple testing, we used the false discovery rate with a significance threshold of Q ≤ 0.05. From the differential gene expression, we identified the significantly modulated canonical pathways, the upstream regulators, and the gene functions that were critical in the NBUVB-treated bulge compared with the NBUVB-treated IE using the Ingenuity Pathway Analysis tool.

      qRT-PCR validation of RNA-seq results

      To confirm the top genes differentially expressed in the melanocytes captured from NBUVB-treated bulge and NBUVB-treated IE, we used biopsy samples from four NBUVB-treated vitiligo patients and performed new F-LCM sessions of melanocytes, followed by total RNA extraction and amplification (as described in the Supplementary Materials and Methods online). The resulting cDNA was subjected to qRT-PCR, using primers that we designed (see Supplementary Table S1). For each sample, expression values were calculated using the standard ΔCt method normalized to ACTB expression. Paired t tests were used to compare the average ΔCt values for NBUVB-treated samples collected from IE versus bulge (P < 0.05). Differences in gene expression in the bulge melanocyte capture from normal skin, untreated vitiligo skin, and NBUVB-treated vitiligo skin were analyzed with one-way analysis of variance followed by Tukey test for multiple comparisons (Padjusted < 0.05).

      Immunohistochemistry

      Vitiligo patients’ demographics and collection methods used for immunostaining are presented in Supplementary Table S6 online.
      The antibodies used for this study are listed in the Supplementary Materials and Methods. Standard procedures were used for immunohistochemistry. To quantify the GLI1 marker’s expression, we measured the average signal intensity of the anti-GLI1 antibody in melanocytes for each patient sample. The differences in signal intensity between the bulge-outer root sheath of NBUVB-treated vitiligo skin and the bulge-outer root sheath of untreated vitiligo skin were compared with a one-tailed unpaired t test.

      Conflict of Interest

      The authors state no conflict of interest.

      Acknowledgments

      This work was supported by a Research Scholar Award in Vitiligo from the American Skin Association to Stanca A. Birlea and by University of Colorado School of Medicine F-LCM Shared Resource funded by the National Institutes of Health/National Center for Advancing Translational Sciences Colorado Clinical and Translational Science Awards grant UL1 TR001082. We are grateful to our team in the Dermatology Clinic at University of Colorado Hospital (UCH) for sample collection, Kathleen Ryan-Morgan and Susan Chalmers. We thank Adriaan van Bokhoven, Zachary Grasmick, and Nicole Spoelstra from the Laser Capture Microdissection Core for their valuable support with the laser capture machine. We are grateful to Manabu Ohyama, Randall Cohrs, and Mark Burgoon for valuable advice on laser capture microdissection technique and quantitative real-time–PCR. We acknowledge the National Disease Research Interchange for providing human skin samples.

      Supplementary Material

      References

        • Amoh Y.
        • Aki R.
        • Hamada Y.
        • Niiyama S.
        • Eshima K.
        • Kawahara K.
        • et al.
        Nestin-positive hair follicle pluripotent stem cells can promote regeneration of impinged peripheral nerve injury.
        J Dermatol. 2012; 39: 33-38
        • Barakat B.
        • Yu L.
        • Lo C.
        • Vu D.
        • De Luca E.
        • Cain J.E.
        • et al.
        Interaction of smoothened with integrin-linked kinase in primary cilia mediates Hedgehog signalling.
        EMBO Rep. 2013; 14: 837-844
        • Bikle D.D.
        • Xie Z.
        • Tu C.L.
        Calcium regulation of keratinocyte differentiation.
        Expert Rev Endocrinol Metab. 2012; 7: 461-472
        • Birlea S.A.
        • Costin G.E.
        • Roop D.R.
        • Norris D.A.
        Trends in regenerative medicine: repigmentation in vitiligo through melanocyte stem cell mobilization.
        Med Res Rev. 2017; 37: 907-935
        • Birlea S.A.
        • Goldstein N.B.
        • Norris D.A.
        Repigmentation through melanocyte regeneration in vitiligo.
        Dermatol Clin. 2017; 35: 205-218
        • Cichorek M.
        • Wachulska M.
        • Stasiewicz A.
        • Tymińska A.
        Skin melanocytes: biology and development.
        Postepy Dermatol Alergol. 2013; 30: 30-41
        • Costin G.E.
        • Hearing V.J.
        Human skin pigmentation: melanocytes modulate skin color in response to stress.
        FASEB J. 2007; 21: 976-994
        • Dai D.L.
        • Makretsov N.
        • Campos E.I.
        • Huang C.
        • Zhou Y.
        • Huntsman D.
        • et al.
        Increased expression of integrin-linked kinase is correlated with melanoma progression and poor patient survival.
        Clin Cancer Res. 2003; 9: 4409-4414
        • Deng W.
        • Vanderbilt D.B.
        • Lin C.C.
        • Martin K.H.
        • Brundage K.M.
        • Ruppert J.M.
        SOX9 inhibits β-TrCP-mediated protein degradation to promote nuclear GLI1 expression and cancer stem cell properties.
        J Cell Sci. 2015; 128: 1123-1138
        • Dolter K.E.
        • Braman J.C.
        Small-sample total RNA purification: laser capture microdissection and cultured cell applications.
        Biotechniques. 2001; 30: 1358-1361
        • Eby J.M.
        • Kang H.K.
        • Klarquist J.
        • Chatterjee S.
        • Mosenson J.A.
        • Nishimura M.I.
        • et al.
        Immune responses in a mouse model of vitiligo with spontaneous epidermal de- and repigmentation.
        Pigment Cell Melanoma Res. 2014; 27: 1075-1085
        • Fantauzzo K.A.
        • Kurban M.
        • Levy B.
        • Christiano A.M.
        Trps1 and its target gene Sox9 regulate epithelial proliferation in the developing hair follicle and are associated with hypertrichosis.
        PLoS Genet. 2012; 8: e1003002
        • Flynn A.
        • Proud C.G.
        The role of eIF4 in cell proliferation.
        Cancer Surv. 1996; 27: 293-310
        • Fujimoto A.
        • Kurban M.
        • Nakamura M.
        • Farooq M.
        • Fujikawa H.
        • Kibbi A.G.
        • et al.
        GJB6, of which mutations underlie Clouston syndrome, is a potential direct target gene of p63.
        J Dermatol Sci. 2013; 69: 159-166
        • Goldstein N.B.
        • Koster M.I.
        • Hoaglin L.G.
        • Spoelstra N.S.
        • Kechris K.J.
        • Robinson S.E.
        • et al.
        Narrow band ultraviolet B treatment for human vitiligo is associated with proliferation, migration, and differentiation of melanocyte precursors.
        J Invest Dermatol. 2015; 135: 2068-2076
        • Goldstein N.B.
        • Koster M.I.
        • Hoaglin L.G.
        • Wright M.J.
        • Robinson S.E.
        • Robinson W.A.
        • et al.
        Isolating RNA from precursor and mature melanocytes from human vitiligo and normal skin using laser capture microdissection.
        Exp Dermatol. 2016; 25: 805-811
        • Haase I.
        • Evans R.
        • Pofahl R.
        • Watt F.M.
        Regulation of keratinocyte shape, migration and wound epithelialization by IGF-1- and EGF-dependent signalling pathways.
        J Cell Sci. 2003; 116: 3227-3238
        • Hendaoui I.
        • Tucker R.P.
        • Zingg D.
        • Bichet S.
        • Schittny J.
        • Chiquet-Ehrismann R.
        Tenascin-C is required for normal Wnt/β-catenin signaling in the whisker follicle stem cell niche.
        Matrix Biol. 2014; 40: 46-53
        • Imanaka-Yoshida K.
        • Aoki H.
        Tenascin-C and mechanotransduction in the development and diseases of cardiovascular system.
        Front Physiol. 2014; 5: 283
        • Jo A.
        • Denduluri S.
        • Zhang B.
        • Wang Z.
        • Yin L.
        • Yan Z.
        The versatile functions of Sox9 in development, stem cells, and human diseases.
        Genes Dis. 2014; 1: 149-161
        • Kadaja M.
        • Keyes B.E.
        • Lin M.
        • Pasolli H.A.
        • Genander M.
        • Polak L.
        • et al.
        SOX9: a stem cell transcriptional regulator of secreted niche signaling factors.
        Genes Dev. 2014; 28: 328-341
        • Kuphal S.
        • Bauer R.
        • Bosserhoff A.K.
        Integrin signaling in malignant melanoma.
        Cancer Metastasis Rev. 2005; 24: 195-222
        • Lamartine J.
        • Munhoz Essenfelder G.
        • Kibar Z.
        • Lanneluc I.
        • Callouet E.
        • Laoudj D.
        • et al.
        Mutations in GJB6 cause hidrotic ectodermal dysplasia.
        Nat Genet. 2000; 26: 142-144
        • Li X.
        • Deng W.
        • Lobo-Ruppert S.M.
        • Ruppert J.M.
        Gli1 acts through Snail and E-cadherin topromote nuclear signaling by β-catenin.
        Oncogene. 2007; 26: 4489-4498
        • Liao X.
        • Siu M.K.
        • Au C.W.
        • Chan Q.K.
        • Chan H.Y.
        • Wong E.S.
        Aberrant activation of hedgehog signaling pathway contributes to endometrial carcinogenesis through beta-catenin.
        Mod Pathol. 2009; 22: 839-847
        • Liu Y.
        • Shin S.
        • Zeng X.
        • Zhan M.
        • Gonzalez R.
        • Mueller F.J.
        • et al.
        Genome wide profiling of human embryonic stem cells (hESCs), their derivatives and embryonal carcinoma cells to develop base profiles of U.S. Federal government approved hESC lines.
        BMC Dev Biol. 2006; 6: 20
        • Liu W.
        • Peng Y.
        • Tobin D.J.
        A new 12-gene diagnostic biomarker signature of melanoma revealed by integrated microarray analysis.
        PeerJ. 2013; 1: e49
        • Maeda O.
        • Kondo M.
        • Fujita T.
        • Usami N.
        • Fukui T.
        • Shimokata K.
        • et al.
        Enhancement of GLI1-transcriptional activity by beta-catenin in human cancer cells.
        Oncol Rep. 2006; 16: 91-96
        • Merle P.
        • Kim M.
        • Herrmann M.
        • Gupte A.
        • Lefrançois L.
        • Califano S.
        • et al.
        Oncogenic role of the frizzled-7/beta-catenin pathway in hepatocellular carcinoma.
        J Hepatol. 2005; 43: 854-862
        • Nakamura N.
        • Ruebel K.
        • Jin L.
        • Qian X.
        • Zhang H.
        • Lloyd R.V.
        Laser capture microdissection for analysis of single cells.
        Methods Mol Med. 2007; 132: 11-18
        • Niwa H.
        How is pluripotency determined and maintained?.
        Development. 2007; 134: 635-646
        • Noubissi F.K.
        • Elcheva I.
        • Bhatia N.
        • Shakoori A.
        • Ougolkov A.
        • Liu J.
        • et al.
        CRD-BP mediates stabilization of betaTrCP1 and c-myc mRNA in response to beta-catenin signalling.
        Nature. 2006; 441: 898-901
        • Noubissi F.K.
        • Goswami S.
        • Sanek N.A.
        • Kawakami K.
        • Minamoto T.
        • Moser A.
        • Grinblat Y.
        • Spiegelman V.S.
        Wnt signaling stimulates transcriptional outcome of the Hedgehog pathway by stabilizing GLI1 mRNA.
        Cancer Res. 2009; 69: 8572-8578
        • Ohyama M.
        • Terunuma A.
        • Tock C.L.
        • Radonovich M.F.
        • Pise-Masison C.A.
        • Hopping S.B.
        • et al.
        Characterization and isolation of stem cell-enriched human hair follicle bulge cells.
        J Clin Invest. 2006; 116: 249-260
        • Oo H.Z.
        • Sentani K.
        • Sakamoto N.
        • Anami K.
        • Naito Y.
        • Oshima T.
        • et al.
        Identification of novel transmembrane proteins in scirrhous-type gastric cancer by the Escherichia coli ampicillin secretion trap (CAST) method: TM9SF3 participates in tumor invasion and serves as a prognostic factor.
        Pathobiology. 2014; 81: 138-148
        • Rashighi M.
        • Harris J.E.
        Vitiligo pathogenesis and emerging treatments.
        Dermatol Clin. 2017; 35: 257-265
        • Santiago-Walker A.
        • Herlyn M.
        The ups and downs of transcription factors in melanoma.
        J Natl Cancer Inst. 2010; 102: 1103-1104
        • Scott G.
        • Leopardi S.
        The cAMP signaling pathway has opposing effects on Rac and Rho in B16F10 cells: implications for dendrite formation in melanocytic cells.
        Pigment Cell Res. 2003; 16: 139-148
        • Silva J.
        • Nichols J.
        • Theunissen T.W.
        • Guo G.
        • van Oosten A.L.
        • Barrandon O.
        • et al.
        Nanog is the gateway to the pluripotent ground state.
        Cell. 2009; 138: 722-737
        • Song L.
        • Li Z.Y.
        • Liu W.P.
        • Zhao M.R.
        Crosstalk between Wnt/β-catenin and Hedgehog/Gli signaling pathways in colon cancer and implications for therapy.
        Cancer Biol Ther. 2015; 16: 1-7
        • Spritz R.A.
        Modern vitiligo genetics sheds new light on an ancient disease.
        J Dermatol. 2013; 40: 310-318
        • Stolboushkina E.A.
        • Garber M.B.
        Eukaryotic type translation initiation factor 2: structure- functional aspects.
        Biochemistry (Mosc). 2011; 76: 283-294
        • Sundaram P.
        • Hultine S.
        • Smith L.M.
        • Dews M.
        • Fox J.L.
        • Biyashev D.
        • et al.
        p53-responsive miR-194 inhibits thrombospondin-1 and promotes angiogenesis in colon cancers.
        Cancer Res. 2011; 71: 7490-7501
        • Turner N.
        • Grose R.
        Fibroblast growth factor signalling: from development to cancer.
        Nat Rev Cancer. 2010; 10: 116-129
        • Vandewoestyne M.
        • Goossens K.
        • Burvenich C.
        • Goossens K.
        • Burvenich C.
        • Van Soom A.
        • et al.
        Laser capture microdissection: should an ultraviolet or infrared laser be used?.
        Anal Biochem. 2013; 439: 88-98
        • Varnat F.
        • Siegl-Cachedenier I.
        • Malerba M.
        • Gervaz P.
        Ruiz i Altaba A. Loss of WNT TCF addiction and enhancement of HH GLI1 signalling define the metastatic transition of human colon carcinomas.
        EMBO Mol Med. 2010; 2: 440-457
        • Velasco-Velázquez M.A.
        • Salinas-Jazmín N.
        • Mendoza-Patiño N.
        • Mandoki J.J.
        Reduced paxillin expression contributes to the antimetastatic effect of 4-hydroxycoumarin on B16-F10 melanoma cells.
        Cancer Cell Int. 2008; 8: 8
        • Wang P.
        • Xu S.
        • Wang Y.
        • Wu P.
        • Zhang J.
        • Sato T.
        • et al.
        GM3 suppresses anchorage-independent growth via Rho GDP dissociation inhibitor beta in melanoma B16 cells.
        Cancer Sci. 2011; 102: 1476-1485
        • Xu X.
        • Lyle S.
        • Liu Y.
        • Solky B.
        • Cotsarelis G.
        Differential expression of cyclin D1 in the human hair follicle.
        Am J Pathol. 2003; 163: 969-978
        • Yang Y.
        • Qian Q.
        Wnt5a/Ca (2+) /calcineurin/nuclear factor of activated T signaling pathway as a potential marker of pediatric melanoma.
        J Cancer Res Ther. 2014; 10: C83-C88
        • Yun C.Y.
        • You S.T.
        • Kim J.H.
        • Chung J.H.
        • Han S.B.
        • Shin E.Y.
        • et al.
        p21-activated kinase 4 critically regulates melanogenesis via activation of the CREB/MITF and β- catenin/MITF pathways.
        J Invest Dermatol. 2015; 135: 1385-1394
        • Zeng L.
        • Kempf H.
        • Murtaugh L.C.
        • Sato M.E.
        • Lassar A.B.
        Shh establishes an Nkx3.2/Sox9 autoregulatory loop that is maintained by BMP signals to induce somatic chondrogenesis.
        Genes Dev. 2002; 16: 1990-2005
        • Zhang Y.V.
        • Cheong J.
        • Ciapurin
        • McDermitt D.J.
        • Tumbar T.
        Distinct self-renewal and differentiation phases in the niche of infrequently dividing hair follicle stem cells.
        Cell Stem Cell. 2009; 5: 267-278