Differential Expression between Human Dermal Papilla Cells from Balding and Non-Balding Scalps Reveals New Candidate Genes for Androgenetic Alopecia

Open ArchivePublished:April 06, 2016DOI:https://doi.org/10.1016/j.jid.2016.03.032
      Androgenetic alopecia (AGA) is a common heritable and androgen-dependent hair loss condition in men. Twelve genetic risk loci are known to date, but it is unclear which genes at these loci are relevant for AGA. Dermal papilla cells (DPCs) located in the hair bulb are the main site of androgen activity in the hair follicle. Widely used monolayer-cultured primary DPCs in hair-related studies often lack dermal papilla characteristics. In contrast, immortalized DPCs have high resemblance to intact dermal papilla. We derived immortalized human DPC lines from balding (BAB) and non-balding (BAN) scalp. Both BAB and BAN retained high proportions of dermal papilla signature gene and versican protein expression. We performed expression analysis of BAB and BAN and annotated AGA risk loci with differentially expressed genes. We found evidence for AR but not EDA2R as the candidate gene at the AGA risk locus on chromosome X. Further, our data suggest TWIST1 (twist family basic helix-loop-helix transcription factor 1) and SSPN (sarcospan) to be the functionally relevant AGA genes at the 7p21.1 and 12p12.1 risk loci, respectively. Down-regulated genes in BAB compared to BAN were highly enriched for vasculature-related genes, suggesting that deficiency of DPC from balding scalps in fostering vascularization around the hair follicle may contribute to the development of AGA.

      Abbreviations:

      AGA (androgenetic alopecia), BAB (immortalized balding dermal papilla cell), BAN (immortalized non-balding dermal papilla cell), DHT (dihydrotestosterone), DP (dermal papilla), DPC (dermal papilla cell), kb (kilo-base pair)

      Introduction

      Androgenetic alopecia (AGA) is a prevalent hair loss condition that affects up to 80% of European men by age 80 years (
      • Hamilton J.B.
      Patterned loss of hair in man; types and incidence.
      ) and is characterized by bitemporal regression of hair growth in combination with hair loss at the vertex (
      • Trueb R.M.
      Molecular mechanisms of androgenetic alopecia.
      ). AGA development is attributed to androgen dependence (
      • Hamilton J.B.
      Male hormone stimulation is prerequisite and an incitant in common baldness.
      ) and genetic predisposition (
      • Nyholt D.R.
      • Gillespie N.A.
      • Heath A.C.
      • Martin N.G.
      Genetic basis of male pattern baldness.
      ,
      • Rexbye H.
      • Petersen I.
      • Iachina M.
      • Mortensen J.
      • McGue M.
      • Vaupel J.W.
      • et al.
      Hair loss among elderly men: etiology and impact on perceived age.
      ). The X-chromosomal androgen receptor (AR)/ectodysplasin A2 receptor (EDA2R) locus was the first described and replicated risk locus for AGA (
      • Ellis J.A.
      • Stebbing M.
      • Harrap S.B.
      Genetic analysis of male pattern baldness and the 5alpha-reductase genes.
      ,
      • Prodi D.A.
      • Pirastu N.
      • Maninchedda G.
      • Sassu A.
      • Picciau A.
      • Palmas M.A.
      • et al.
      EDA2R is associated with androgenetic alopecia.
      ). In recent years, a total of 12 AGA risk loci have been described (
      • Brockschmidt F.F.
      • Heilmann S.
      • Ellis J.A.
      • Eigelshoven S.
      • Hanneken S.
      • Herold C.
      • et al.
      Susceptibility variants on chromosome 7p21.1 suggest HDAC9 as a new candidate gene for male-pattern baldness.
      ,
      • Heilmann S.
      • Kiefer A.K.
      • Fricker N.
      • Drichel D.
      • Hillmer A.M.
      • Herold C.
      • et al.
      Androgenetic alopecia: identification of four genetic risk loci and evidence for the contribution of WNT signaling to its etiology.
      ,
      • Hillmer A.M.
      • Brockschmidt F.F.
      • Hanneken S.
      • Eigelshoven S.
      • Steffens M.
      • Flaquer A.
      • et al.
      Susceptibility variants for male-pattern baldness on chromosome 20p11.
      ,
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      ,
      • Richards J.B.
      • Yuan X.
      • Geller F.
      • Waterworth D.
      • Bataille V.
      • Glass D.
      • et al.
      Male-pattern baldness susceptibility locus at 20p11.
      ). However, the genes responsible for increased AGA risk at most risk loci are still unclear. Functional understanding of AGA biology is lacking, because it is not trivial or scalable to dissect the different compartments of hair follicles. Appropriate model systems for AGA are also unavailable. Medical treatments for AGA include minoxidil, which is hypothesized to promote hair growth by increasing vascular endothelial growth factor (VEGF) expression in dermal papilla cells (DPCs) during the growth (anagen) phase of the hair cycle. DPCs, located at the base of the hair follicle, are essential to the hair growth process. DPCs have the most consistent AR expression among all compartments of the hair follicle (
      • Hodgins M.B.
      • Choudhry R.
      • Parker G.
      • Oliver R.F.
      • Jahoda C.A.
      • Withers A.P.
      • et al.
      Androgen receptors in dermal papilla cells of scalp hair follicles in male pattern baldness.
      ) and thus are the postulated sites of androgen response, where AR activation by androgens affects gene expression regulating hair growth and cycle (
      • Itami S.
      • Sonoda T.
      • Kurata S.
      • Takayasu S.
      Mechanism of action of androgen in hair follicles.
      ,
      • Kitagawa T.
      • Matsuda K.
      • Inui S.
      • Takenaka H.
      • Katoh N.
      • Itami S.
      • et al.
      Keratinocyte growth inhibition through the modification of Wnt signaling by androgen in balding dermal papilla cells.
      ). To date, only a macroarray-based comparative study interrogating 1,185 genes has been conducted on primary DPCs from balding and non-balding scalps (
      • Midorikawa T.
      • Chikazawa T.
      • Yoshino T.
      • Takada K.
      • Arase S.
      Different gene expression profile observed in dermal papilla cells related to androgenic alopecia by DNA macroarray analysis.
      ). The limited life span of primary cell lines hampers complex and long-term experiments.
      In this study we present a robust and practical model system for AGA. We immortalized two human DPC lines, one derived from a balding vertex scalp area (BAB) and one derived from a non-balding occipital scalp area (BAN) and determined that both cell lines possess dermal papilla (DP) character. The utility of single cell lines is supported by their usage in discovery (
      • Edwards D.P.
      • Murthy S.R.
      • McGuire W.L.
      Effects of estrogen and antiestrogen on DNA polymerase in human breast cancer.
      ,
      • Elenbaas B.
      • Spirio L.
      • Koerner F.
      • Fleming M.D.
      • Zimonjic D.B.
      • Donaher J.L.
      • et al.
      Human breast cancer cells generated by oncogenic transformation of primary mammary epithelial cells.
      ) and functional studies (
      • Boukamp P.
      • Petrussevska R.T.
      • Breitkreutz D.
      • Hornung J.
      • Markham A.
      • Fusenig N.E.
      Normal keratinization in a spontaneously immortalized aneuploid human keratinocyte cell line.
      ) for various disease conditions. We analyzed BAB and BAN gene expression alterations in response to dihydrotestosterone (DHT) treatment by microarray technology interrogating 26,000 genes and used the expression information to interpret described AGA risk loci. Further, we identified differentially expressed genes between balding and non-balding DPCs, showing down-regulation of vasculature-related genes in DPCs of balding hair follicles.

      Results

       Immortalized balding and non-balding DPCs retain DP character

      Immortalized balding DPCs (BAB; Figure 1a) and non-balding DPCs (BAN; Figure 1b), established from biopsy samples of balding and non-balding scalps of male patients by human telomerase reverse transcriptase (hTERT) expression, were observed to have different cell morphologies. Both BAB and BAN retained DP characteristics after in vitro two-dimensional culturing, as indicated by the high expression of versican (VCAN) protein (Figure 1c and d) and the expression of human DP signature genes (
      • Ohyama M.
      • Kobayashi T.
      • Sasaki T.
      • Shimizu A.
      • Amagai M.
      Restoration of the intrinsic properties of human dermal papilla in vitro.
      ), with 83 out of 117 (70.9%, P < 0.001) and 75 out of 117 (64.1%, P < 0.05) DP signature genes expressed in BAN and BAB, respectively (Figure 1e and see Supplementary Figures S1 and S2 online and Supplementary Table S1a online). DP signature genes have previously been defined as significantly up-regulated genes in freshly dissected DP compared with dermal fibroblasts (
      • Ohyama M.
      • Kobayashi T.
      • Sasaki T.
      • Shimizu A.
      • Amagai M.
      Restoration of the intrinsic properties of human dermal papilla in vitro.
      ). We observed similar fractions of DP signature gene expression in primary non-balding DP (n = 2) and primary balding DP (n = 3) cells, which expressed 79 out of 116 (68.1%, P < 0.01) and 69 out of 116 (59.5%, P < 0.05) DP signature genes, respectively (see Supplementary Table S1b). Furthermore, we found that the DP signature gene expression profiles of both BAN and BAB were more similar to primary occipital human follicle dermal papilla cells (HFDPC) than to primary fibroblasts (L5-F and NHDF) (see Supplementary Materials, Supplementary Figure S3, and Supplementary Table S1a and c online).
      Figure 1
      Figure 1Characterization of immortalized balding DPC line (BAB) and non-balding DPC line (BAN). (a) BAB grows in a non-monolayer manner and does not cover the entire culture area at 100% confluency. (b) BAN grows in a monolayer and displays a whorl-like pattern. Scale bar for a and b = 1,000 μm. Both (c) BAB and (d) BAN express dermal papilla marker protein, versican. Cell nuclei marked by 4',6-diamidino-2-phenylindole staining (blue). Scale bar for c and d = 100 μm. (e) Out of 117, 83 (70.9%, P < 0.001) DP signature genes were expressed in BAN (red square); 75 (64.1%, P < 0.001) out of 117 DP signature genes were expressed in BAB (blue square); and 69 DP signature genes were expressed in both BAN and BAB. Twenty-eight DP signature genes not expressed in BAB and BAN are listed outside the blue and red squares.

       Differential gene expression analysis between BAB and BAN

      Both inherent and DHT-dependent gene expression differences were found between BAB and BAN (Figure 2a and see Supplementary Figure S4 online). Among the differentially expressed genes in BAB compared to BAN under 1 nmol/L and 10 nmol/L of DHT stimulation, 1,482 and 1,508 genes were found to be commonly up-regulated and down-regulated, respectively (Figure 2b–d and see Supplementary Tables S2–4 online). Subsequent analysis was carried out with commonly up- and down-regulated genes between both DHT concentrations. The trend of expression of quantitative PCR (qPCR)-validated genes (CAV1 [caveolin 1], EDNRA [endothelin receptor type A], IGFBP5 [insulin-like growth factor binding protein 5], and SCG2 [secretogranin II]) was consistent with that from the microarray data (Figure 2e–h and Supplementary Table S5 online).
      Figure 2
      Figure 2Differential gene expression in non-balding (BAN) compared to balding DPC (BAB). (a) Gene expression in BAN and BAB are inherent and DHT dependent. (b) Differential genes in BAB compared to BAN under 1 and 10 nmol/L DHT treatment, and commonly (c) up- and (d) down-regulated genes under both concentrations of DHT stimulation. Quantitative real-time reverse transcriptase–PCR validation of down-regulated genes (e) CAV1, (f) EDNRA, (gIGFBP5, and (h) SCG2 in BAB compared to BAN under 10 nmol/L DHT treatment. Expression fold decrease in BAB compared to BAN from microarray analysis is presented in the upper plots. GAPDH-normalized relative gene expression in BAB compared to BAN from quantitative real-time reverse transcriptase–PCR analysis (ddCp) is presented in the lower plots. A fold change threshold cutoff of 1.7 is indicated in blue. #Fold change ≤ 1.7 from microarray analysis. ddCp, delta delta crossing point; DHT, dihydrotestosterone; DP, dermal papilla; DPC, dermal papilla cell; h, hours; M, mol/L; min, minutes.

       Annotation of AGA risk loci with differentially expressed genes in BAB compared with BAN

      Genome-wide association studies have identified genomic loci that confer risk for developing AGA. However, it is largely unclear which genes at these loci are causal for the associations. Our tissue- and phenotype-specific expression data allowed us to infer which genes at the risk loci most likely contribute to AGA development. Twelve differentially expressed genes in BAB compared to BAN overlapped within 500 kilo-base pairs (kb) of eight lead single-nucleotide polymorphisms at established AGA risk loci (Table 1 and see Supplementary Table S6 online). On Xq12, the major AGA risk locus, AR was up-regulated in BAB compared to BAN, suggesting that AR rather than the centromeric EDA2R contributes to AGA. At the 2q37 risk locus, only PER2 (period circadian clock 2) and TWIST2 (twist family bHLH transcription factor 2) showed expression differences within 500 kb from the lead single-nucleotide polymorphism rs9287638 and therefore are more likely candidate genes at this locus than non-differentially expressed HDAC4 (histone deacetylase 4). At the 5q33.3 risk locus, down-regulated RNF145 (ring finger protein 145) is a potential candidate gene. Up-regulation of TWIST1 was found at the 7p21.1 risk locus. Furthermore, our finding of 7q11.22 candidate gene AUTS2 (autism susceptibility candidate 2) being down-regulated in BAB compared to BAN is in agreement with a previous study (
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      ) that suggested AUTS2 as the candidate gene causing the association of this locus with AGA. On 12p12.1, SSPN showed down-regulation and is the more likely candidate gene at this locus instead of non-differentially expressed ITPR2 (inositol 1,4,5-trisphosphate receptor type 2). On 17q21.31, down-regulated gene MAPT (microtubule associated protein tau) is the likely candidate gene instead of non-differentially expressed SPPL2C (signal peptide peptidase like 2C).
      Table 1Overlap of differentially expressed genes identified in balding DPCs compared to non-balding DPCs with AGA risk loci of genome-wide significance
      ChrPosition
      1 Chromosomal position in GRCh37/hg19 assembly.
      SNP
      2 References include first descriptions of associations of the respective locus that do not necessarily have the same single-nucleotide polymorphism.
      Genomic RegionGenes
      3 Candidate genes harboring single-nucleotide polymorphism or in close proximity to single-nucleotide polymorphism. Squared brackets indicate position of single-nucleotide polymorphism.
      500 kb100 kb50 kbMedian Fold Change
      Diff genes
      4 Differentially expressed genes in balding DPCs (i.e., BAB) compared to non-balding DPCs (i.e., BAN) that are within window stated to single-nucleotide polymorphism.
      Up/ down
      5 up indicates that differentially expressed gene is up-regulated in BAB compared to BAN; down indicates that differentially expressed gene is down-regulated in BAB compared to BAN.
      Diff genes
      4 Differentially expressed genes in balding DPCs (i.e., BAB) compared to non-balding DPCs (i.e., BAN) that are within window stated to single-nucleotide polymorphism.
      Up/ down
      5 up indicates that differentially expressed gene is up-regulated in BAB compared to BAN; down indicates that differentially expressed gene is down-regulated in BAB compared to BAN.
      Diff genes
      4 Differentially expressed genes in balding DPCs (i.e., BAB) compared to non-balding DPCs (i.e., BAN) that are within window stated to single-nucleotide polymorphism.
      Up/ down
      5 up indicates that differentially expressed gene is up-regulated in BAB compared to BAN; down indicates that differentially expressed gene is down-regulated in BAB compared to BAN.
      1 nmol/L DHT10 nmol/L DHT
      111,033,082rs12565727
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      1p36[]--TARDBPCASZ1
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      down2.0101.945
      EXOSC10
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      upEXOSC10
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      up1.9181.998
      FRAP1
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      up1.740
      SRMupSRMup2.2542.122
      UBIAD1
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      up2.1541.945
      2219,756,383rs7349332
      7
      • Heilmann S.
      • Kiefer A.K.
      • Fricker N.
      • Drichel D.
      • Hillmer A.M.
      • Herold C.
      • et al.
      Androgenetic alopecia: identification of four genetic risk loci and evidence for the contribution of WNT signaling to its etiology.
      .
      2q35[WNT10A]
      2239,694,631rs9287638
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      2q37[]--HDAC4PER2
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      down1.9331.909
      TWIST2
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      downTWIST2
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      down1.9492.305
      3151,653,368rs4679955
      7
      • Heilmann S.
      • Kiefer A.K.
      • Fricker N.
      • Drichel D.
      • Hillmer A.M.
      • Herold C.
      • et al.
      Androgenetic alopecia: identification of four genetic risk loci and evidence for the contribution of WNT signaling to its etiology.
      .
      3q25.1SUCNR1--[]---MBNL1
      5158,310,631rs929626
      7
      • Heilmann S.
      • Kiefer A.K.
      • Fricker N.
      • Drichel D.
      • Hillmer A.M.
      • Herold C.
      • et al.
      Androgenetic alopecia: identification of four genetic risk loci and evidence for the contribution of WNT signaling to its etiology.
      .
      5q33.3[EBF1]RNF145
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      down1.730
      718,877,874rs2073963
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      8
      • Brockschmidt F.F.
      • Heilmann S.
      • Ellis J.A.
      • Eigelshoven S.
      • Hanneken S.
      • Herold C.
      • et al.
      Susceptibility variants on chromosome 7p21.1 suggest HDAC9 as a new candidate gene for male-pattern baldness.
      .
      7p21.1[HDAC9]TWIST1
      15 Novel candidate genes not considered by previous publications, to our knowledge.
      up2.5482.244
      768,611,960rs6945541
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      7q11.22[]---AUTS2AUTS2down2.8012.868
      1226,426,420rs9668810
      7
      • Heilmann S.
      • Kiefer A.K.
      • Fricker N.
      • Drichel D.
      • Hillmer A.M.
      • Herold C.
      • et al.
      Androgenetic alopecia: identification of four genetic risk loci and evidence for the contribution of WNT signaling to its etiology.
      .
      12p12.1SSPN--[]--ITPR2SSPNdownSSPNdownSSPNdown2.1972.133
      1743,924,219rs12373124
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      17q21.31[SPPL2C]--MAPTMAPTdownMAPTdownMAPTdown1.9301.835
      1842,800,148rs10502861
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      18q21.1SETBP1---[]
      2022,037,575rs6047844
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      9
      • Hillmer A.M.
      • Brockschmidt F.F.
      • Hanneken S.
      • Eigelshoven S.
      • Steffens M.
      • Flaquer A.
      • et al.
      Susceptibility variants for male-pattern baldness on chromosome 20p11.
      .
      10
      • Richards J.B.
      • Yuan X.
      • Geller F.
      • Waterworth D.
      • Bataille V.
      • Glass D.
      • et al.
      Male-pattern baldness susceptibility locus at 20p11.
      .
      20p11PAX1--[]--FOXA2
      X66,563,018rs2497938
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      11
      • Ellis J.A.
      • Stebbing M.
      • Harrap S.B.
      Polymorphism of the androgen receptor gene is associated with male pattern baldness.
      .
      12
      • Hillmer A.M.
      • Hanneken S.
      • Ritzmann S.
      • Becker T.
      • Freudenberg J.
      • Brockschmidt F.F.
      • et al.
      Genetic variation in the human androgen receptor gene is the major determinant of common early-onset androgenetic alopecia.
      .
      13
      • Brockschmidt F.F.
      • Hillmer A.M.
      • Eigelshoven S.
      • Hanneken S.
      • Heilmann S.
      • Barth S.
      • et al.
      Fine mapping of the human AR/EDA2R locus in androgenetic alopecia.
      .
      14
      • Prodi D.A.
      • Pirastu N.
      • Maninchedda G.
      • Sassu A.
      • Picciau A.
      • Palmas M.A.
      • et al.
      EDA2R is associated with androgenetic alopecia.
      .
      Xq12EDA2R---[]--ARARup1.7511.737
      Abbreviations: DHT, dihydrotestosterone; diff, differentially expressed; DPC, dermal papilla cell; kb, kilo-base pairs; SNP, single-nucleotide polymorphism.
      1 Chromosomal position in GRCh37/hg19 assembly.
      2 References include first descriptions of associations of the respective locus that do not necessarily have the same single-nucleotide polymorphism.
      3 Candidate genes harboring single-nucleotide polymorphism or in close proximity to single-nucleotide polymorphism. Squared brackets indicate position of single-nucleotide polymorphism.
      4 Differentially expressed genes in balding DPCs (i.e., BAB) compared to non-balding DPCs (i.e., BAN) that are within window stated to single-nucleotide polymorphism.
      5 up indicates that differentially expressed gene is up-regulated in BAB compared to BAN; down indicates that differentially expressed gene is down-regulated in BAB compared to BAN.
      6
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      .
      7
      • Heilmann S.
      • Kiefer A.K.
      • Fricker N.
      • Drichel D.
      • Hillmer A.M.
      • Herold C.
      • et al.
      Androgenetic alopecia: identification of four genetic risk loci and evidence for the contribution of WNT signaling to its etiology.
      .
      8
      • Brockschmidt F.F.
      • Heilmann S.
      • Ellis J.A.
      • Eigelshoven S.
      • Hanneken S.
      • Herold C.
      • et al.
      Susceptibility variants on chromosome 7p21.1 suggest HDAC9 as a new candidate gene for male-pattern baldness.
      .
      9
      • Hillmer A.M.
      • Brockschmidt F.F.
      • Hanneken S.
      • Eigelshoven S.
      • Steffens M.
      • Flaquer A.
      • et al.
      Susceptibility variants for male-pattern baldness on chromosome 20p11.
      .
      10
      • Richards J.B.
      • Yuan X.
      • Geller F.
      • Waterworth D.
      • Bataille V.
      • Glass D.
      • et al.
      Male-pattern baldness susceptibility locus at 20p11.
      .
      11
      • Ellis J.A.
      • Stebbing M.
      • Harrap S.B.
      Polymorphism of the androgen receptor gene is associated with male pattern baldness.
      .
      12
      • Hillmer A.M.
      • Hanneken S.
      • Ritzmann S.
      • Becker T.
      • Freudenberg J.
      • Brockschmidt F.F.
      • et al.
      Genetic variation in the human androgen receptor gene is the major determinant of common early-onset androgenetic alopecia.
      .
      13
      • Brockschmidt F.F.
      • Hillmer A.M.
      • Eigelshoven S.
      • Hanneken S.
      • Heilmann S.
      • Barth S.
      • et al.
      Fine mapping of the human AR/EDA2R locus in androgenetic alopecia.
      .
      14
      • Prodi D.A.
      • Pirastu N.
      • Maninchedda G.
      • Sassu A.
      • Picciau A.
      • Palmas M.A.
      • et al.
      EDA2R is associated with androgenetic alopecia.
      .
      15 Novel candidate genes not considered by previous publications, to our knowledge.

       Gene ontology analysis of differentially expressed genes in BAB compared to BAN

      Down-regulated genes in BAB compared to BAN under DHT treatment were found to be most enriched in gene ontology clusters related to vasculature (maximum enrichment score = 10.516; 65 genes) and included the subterms of vasculature development, blood vessel development, and blood vessel morphogenesis (Figure 3a and Table 2, and see Supplementary Tables S7a and S8a online). We found support for vasculature-related differences between BAB and BAN by immunofluorescence (Figure 3c and d) and migration assay experiments (see Supplementary Figure S5 online and Supplementary Materials) and weaker expression signals for vasculature-related genes in fibroblasts (see Supplementary Figure S6 online). We tested the down-regulated vasculature-related genes CAV1 (caveolin 1), CYR61 (cysteine-rich, angiogenic inducer, 61), and MMP14 (matrix metallopeptidase 14) by immunofluorescence and observed protein down-regulation for MMP14 (P = 0.0004) and CAV1 (P = 0.0002) in BAB compared to BAN (Figure 3c and d). A significant number of the identified vasculature-related genes (Table 2 and see Supplementary Table S7a) were also found to be down-regulated in independent samples of primary balding (n = 3) compared to non-balding DPCs (n = 2, P < 0.001, 16 out of 65; see Supplementary Table S7c and Supplementary Materials). Up-regulated genes were most enriched in gene ontology clusters related to cell cycle and mitosis (maximum enrichment score = 38.551, 139 genes; Figure 3b and see Supplementary Table S8b).
      Figure 3
      Figure 3Gene ontology clustering analysis of differentially expressed genes in balding (BAB) compared to non-balding DPCs (BAN). (a) Down-regulated genes in BAB compared to BAN were highly enriched in vasculature-related gene ontology clusters (subterms: vasculature development, blood vessel development, and blood vessel morphogenesis) (), cell motion and cell migration, cell death, phosphate and phosphorous metabolism, and protein kinase (see a). (b) Gene ontology clustering of up-regulated genes in BAB compared to BAN (see b). Maximum enrichment score of each gene ontology cluster is indicated below respective term. (c, d) Expression levels of vasculature-related proteins CYR61, MMP14 (P = 0.0004), and CAV1 (P = 0.0002) were found to be decreased in BAB compared to BAN. ∗∗P < 0.001. Cell nuclei marked by 4',6-diamidino-2-phenylindole staining (blue). Scale bar = 100 μm. HepG2 was used as negative control for CAV1. DP, dermal papilla; DPC, dermal papilla cell.
      Table 2List of vasculature-related genes that are down-regulated in balding DPCs compared to non-balding DPCs
      Gene NameMedian Fold ChangeGene NameMedian Fold Change
      1 nmol/L DHT10 nmol/L DHT1 nmol/L DHT10 nmol/L DHT
      ACTC12.6282.532GNA131.885
      ADORA2A1.9241.752GUCY1A31.8321.846
      ADRA1B1.8331.749HEY12.0462.402
      ADRB22.1062.590HIF1A2.5572.670
      AGT2.0081.881HIF1A2.8062.850
      AGTR12.7882.643HIF1A2.8222.546
      AGTR12.8032.621HMOX11.8951.934
      AMOT1.8511.910HTATIP26.7646.038
      ANGPTL42.0542.220ITGA12.0081.956
      ANXA22.5002.211ITGA41.7841.748
      APOLD12.3712.661JUNB2.3252.002
      BGN2.3142.324KLF51.7381.802
      CAV12.8842.708LEPR2.0421.987
      CAV21.8571.954LOX3.1162.974
      CAV21.8191.775MMP142.0902.003
      CAV21.8271.741NOS32.1541.955
      CCBE12.2502.171NOTCH11.7131.813
      CDH131.8291.867P2RX41.8301.806
      CDH21.7341.782PDE5A2.8092.697
      CHD72.0472.187PDPN1.9311.957
      CITED22.9672.623PLAU1.7881.882
      COL1A15.0694.926PLCD31.7501.770
      COL1A21.8081.757PLXDC11.9921.846
      COL3A13.7233.615PPAP2B1.8791.870
      COL18A116.75015.675PPAP2B2.2352.225
      CTGF3.5463.587RECK2.4632.351
      CTGF3.9634.229RECK2.4582.314
      CXCL122.6352.631SCG25.3655.938
      CXCL122.3912.151SGPL11.9101.828
      CXCL122.5182.643SMAD71.8411.729
      CYR613.1673.023SMO1.7831.830
      DHCR723.46420.353SOD21.752
      DHCR717.99414.520SOD22.3062.355
      DICER11.8601.800SOD21.9692.367
      DICER11.8961.898TGFA3.2813.060
      EDNRA1.8311.978TGFA2.1462.328
      EREG1.9862.303THY12.8712.781
      FGF22.0872.018TNFAIP23.1802.792
      FGF21.9711.954ZC3H12A1.8902.237
      GCH12.9573.348ZFP36L13.2443.055
      GNA131.7841.755ZMIZ11.8351.746
      Abbreviations: DHT, dihydrotestosterone; DPC, dermal papilla cell.

       Motif analysis in promoters of differentially expressed genes in BAB compared to BAN

      We performed a transcription factor motif enrichment analysis in promoters of differentially regulated genes in BAB compared to BAN to investigate potential transcription factors implicated in characteristic differences between the cell types. The SIX1 (SIX homeobox 1) motif, which was enriched in promoters of down-regulated genes (see Supplementary Tables S2b and S9a online), is implicated in hair placode development (
      • Rhee H.
      • Polak L.
      • Fuchs E.
      Lhx2 maintains stem cell character in hair follicles.
      ). Enriched motifs in the promoters of up-regulated genes (see Supplementary Tables S2a and Table S9b) include E2F and CHR motifs, which correlate with biological function in cell cycle and mitosis (Figure 3b and see Supplementary Table S8b). We have also identified genes that could be synergistically regulated by multiple transcription factors because of the co-occurrence of E2F, NFY, and CHR sites (see Supplementary Table S9d and Supplementary Materials.) (
      • Tabach Y.
      • Milyavsky M.
      • Shats I.
      • Brosh R.
      • Zuk O.
      • Yitzhaky A.
      • et al.
      The promoters of human cell cycle genes integrate signals from two tumor suppressive pathways during cellular transformation.
      ).

      Discussion

      In this study, we have shown that both two-dimensional–cultured immortalized balding DPC and non-balding DPC maintained DP characteristics and therefore provide a model for studying AGA and hair biology. First, both BAB and BAN retained significant expression of most DP signature genes (BAB: 64.1%, P < 0.05; BAN: 70.9%, P < 0.001) (Figure 1e and see Supplementary Table S1a) defined from expression profiling of freshly dissected human DPs (
      • Ohyama M.
      • Kobayashi T.
      • Sasaki T.
      • Shimizu A.
      • Amagai M.
      Restoration of the intrinsic properties of human dermal papilla in vitro.
      ). DP signature genes interrogated for include classical DP signature genes such as ALPL, WIF1, LEF1, (
      • Kratochwil K.
      • Dull M.
      • Farinas I.
      • Galceran J.
      • Grosschedl R.
      Lef1 expression is activated by BMP-4 and regulates inductive tissue interactions in tooth and hair development.
      ) and VCAN (
      • Ohyama M.
      • Kobayashi T.
      • Sasaki T.
      • Shimizu A.
      • Amagai M.
      Restoration of the intrinsic properties of human dermal papilla in vitro.
      ,
      • Soma T.
      • Tajima M.
      • Kishimoto J.
      Hair cycle-specific expression of versican in human hair follicles.
      ), of which ALPL, LEF1, and VCAN were expressed in BAB and BAN (Figure 1e and see Supplementary Table S1a). Second, both DPC lines stained positively for VCAN protein, expression of which is correlated with DP inductivity and hair morphogenesis (
      • Kishimoto J.
      • Ehama R.
      • Wu L.
      • Jiang S.
      • Jiang N.
      • Burgeson R.E.
      Selective activation of the versican promoter by epithelial- mesenchymal interactions during hair follicle development.
      ,
      • Soma T.
      • Tajima M.
      • Kishimoto J.
      Hair cycle-specific expression of versican in human hair follicles.
      ). The BAB and BAN cell lines serve as a useful model system for AGA because they are simple to maintain in culture, requiring only fetal bovine serum-supplemented media. In contrast, primary DPCs, which are widely used in hair-related studies, are less ideal models because primary DPCs lose DP character upon two-dimensional–culturing (
      • Higgins C.A.
      • Chen J.C.
      • Cerise J.E.
      • Jahoda C.A.
      • Christiano A.M.
      Microenvironmental reprogramming by three-dimensional culture enables dermal papilla cells to induce de novo human hair-follicle growth.
      ) and are challenging to maintain in culture over extended periods. We were able to immortalize balding and non-balding DPCs only from non-matched patient biopsy samples, because primary balding DPCs were prone to senescence during the initial culturing process and had low success rates for immortalization. The difficulty of establishing immortalized DPC lines is further evidenced by the lack of such cell lines to date, to our knowledge. We would like to emphasize that single cell lines such as HaCaT (
      • Boukamp P.
      • Petrussevska R.T.
      • Breitkreutz D.
      • Hornung J.
      • Markham A.
      • Fusenig N.E.
      Normal keratinization in a spontaneously immortalized aneuploid human keratinocyte cell line.
      ), MCF7 (
      • Edwards D.P.
      • Murthy S.R.
      • McGuire W.L.
      Effects of estrogen and antiestrogen on DNA polymerase in human breast cancer.
      ), and HMEC (
      • Elenbaas B.
      • Spirio L.
      • Koerner F.
      • Fleming M.D.
      • Zimonjic D.B.
      • Donaher J.L.
      • et al.
      Human breast cancer cells generated by oncogenic transformation of primary mammary epithelial cells.
      ) have contributed significantly to our understanding of cell type-specific behavior and serve as disease models. It is desirable to establish more balding and non-balding DP cell lines to validate the transcriptomic signatures. Further, cell lines from vertex with full hair should be included to distinguish balding-specific from vertex-specific signatures.
      We annotated AGA risk loci with differentially expressed genes in BAB compared to BAN and found differential gene expression in eight of the 12 interrogated genomic regions (Table 1 and see Supplementary Table S6). Notably, we found AR to be the likely candidate gene at the major risk locus for AGA (Xq12, rs2497938) instead of EDA2R. This is consistent with previous observations of higher AR levels in balding scalps (
      • Hibberts N.A.
      • Howell A.E.
      • Randall V.A.
      Balding hair follicle dermal papilla cells contain higher levels of androgen receptors than those from non-balding scalp.
      ) and reinforced the importance of the androgen/AR interaction in AGA development. Furthermore, the identification of MAPT as the relevant AGA candidate gene at 17q21.31 instead of SPPL2C suggests that changes in MAPT function connect the risk for AGA with the risk for Parkinson’s disease (
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      ).
      Neither of the candidate genes at the 20p11 susceptibility locus (rs6047844), FOXA2 (forkhead box A2) and PAX1 (paired box 1), were differentially expressed between BAB and BAN. PAX1, but not FOXA2, was expressed in BAB and BAN. The lack of differentially expressed genes within 500 kb from rs6047844 suggest (i) that this susceptibility region may be involved in long-range chromatin interaction that results in the regulation of distant candidate genes, (ii) that DPCs are not the cell type that confers the AGA-related functional effect of this locus, or (iii) that the BAN/BAB model cell lines do not preserve this characteristic.
      At the 1p36 locus, non-differentially expressed candidate gene TARDBP (TAR DNA binding protein) is unlikely to be causative. Instead, SRM (spermidine synthase), previously mentioned as a potential candidate gene because of its proximity to rs12565727 (
      • Li R.
      • Brockschmidt F.F.
      • Kiefer A.K.
      • Stefansson H.
      • Nyholt D.R.
      • Song K.
      • et al.
      Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
      ), was up-regulated in BAB compared to BAN. SRM is involved in the synthesis of spermidine, which acts on matrix keratinocytes to promote hair elongation and prolong anagen (
      • Ramot Y.
      • Tiede S.
      • Biro T.
      • Abu Bakar M.H.
      • Sugawara K.
      • Philpott M.P.
      • et al.
      Spermidine promotes human hair growth and is a novel modulator of human epithelial stem cell functions.
      ). However, the effect of SRM activity and the resultant spermidine synthesized in DPC is unknown. Further, we found CASZ1 (castor zinc finger 1), EXOSC10 (exosome component 10), FRAP1 (also known as MTOR [mechanistic target of rapamycin]), and UBIAD1 (UbiA prenyltransferase domain containing 1) to be differentially expressed at this locus. They provide new potential candidate genes for hair loss/growth modulation.
      The candidate genes HDAC4, EBF1 (early B-cell factor 1), and HDAC9 (histone deacetylase 9) at susceptibility loci 2q37, 5q33.3, and 7p21.1, respectively, were not differentially expressed between BAB and BAN and thus are unlikely to be the causative genes at these loci. Instead, we found four other differentially expressed genes in BAB compared to BAN as potential candidates. RNF145, located at 5q33.3, is implicated in endoplasmic reticulum-associated protein degradation (
      • Kikkert M.
      • Doolman R.
      • Dai M.
      • Avner R.
      • Hassink G.
      • van Voorden S.
      • et al.
      Human HRD1 is an E3 ubiquitin ligase involved in degradation of proteins from the endoplasmic reticulum.
      ) and apoptosis (
      • Ho S.R.
      • Mahanic C.S.
      • Lee Y.J.
      • Lin W.C.
      RNF144A, an E3 ubiquitin ligase for DNA-PKcs, promotes apoptosis during DNA damage.
      ). At the 7p21.1 locus, we found TWIST1, a DP signature gene expressed in both BAB and BAN (Figure 1e and see Supplementary Table S1a), being up-regulated in BAB (Table 1). Twist1, a basic helix-loop-helix protein, is crucial for anagen-to-catagen transition during the hair growth cycle with Twist1 protein ablation in adult mice DP resulting in prolonged anagen (
      • Xu Y.
      • Liao L.
      • Zhou N.
      • Theissen S.M.
      • Liao X.H.
      • Nguyen H.
      • et al.
      Inducible knockout of Twist1 in young and adult mice prolongs hair growth cycle and has mild effects on general health, supporting Twist1 as a preferential cancer target.
      ). Hence, TWIST1 up-regulation in DPCs of balding scalps compared to non-balding scalps may result in accelerated transition from anagen to catagen and thus a shortened period of anagen during the hair cycle, a phenomenon in balding scalps that leads to the formation of short vellus hairs instead of long terminal hairs (
      • Paus R.
      • Cotsarelis G.
      The biology of hair follicles.
      ). Furthermore, basic helix-loop-helix proteins such as Twist1 bind to the consensus 5′-NCANNTGN-3′ E-box motif; suggesting that the down-regulation of genes in BAB compared to BAN with E-box motif in their promoter regions (see Supplementary Table S9a and Supplementary Materials) may be attributed to repression by TWIST1. Interestingly, we identified another TWIST protein gene, TWIST2, as a potential candidate gene at the 2q37 risk locus. TWIST2 has been implicated in mesenchymal cell lineage development (
      • Li L.
      • Cserjesi P.
      • Olson E.N.
      Dermo-1: a novel twist-related bHLH protein expressed in the developing dermis.
      ), but little is known about its function in the DP. The combinatorial binding and interaction of TWIST1 and/or TWIST2 with other basic helix-loop-helix proteins (
      • Franco H.L.
      • Casasnovas J.
      • Rodriguez-Medina J.R.
      • Cadilla C.L.
      Redundant or separate entities?—Roles of Twist1 and Twist2 as molecular switches during gene transcription.
      ) at the promoters of target genes may result in gene regulation in DPC. TWIST1 also interacts and binds to HDAC4 (
      • Danciu T.E.
      • Whitman M.
      Oxidative stress drives disulfide bond formation between basic helix-loop-helix transcription factors.
      ) to regulate gene expression (
      • Gong X.Q.
      • Li L.
      Dermo-1, a multifunctional basic helix-loop-helix protein, represses MyoD transactivation via the HLH domain, MEF2 interaction, and chromatin deacetylation.
      ,
      • Lee Y.S.
      • Lee H.H.
      • Park J.
      • Yoo E.J.
      • Glackin C.A.
      • Choi Y.I.
      • et al.
      Twist2, a novel ADD1/SREBP1c interacting protein, represses the transcriptional activity of ADD1/SREBP1c.
      ). In addition, binding of TWIST1 at E-boxes in the AR promoter region results in up-regulated AR expression (
      • Shiota M.
      • Yokomizo A.
      • Tada Y.
      • Inokuchi J.
      • Kashiwagi E.
      • Masubuchi D.
      • et al.
      Castration resistance of prostate cancer cells caused by castration-induced oxidative stress through Twist1 and androgen receptor overexpression.
      ). Our observation of both TWIST1 and AR up-regulation in BAB compared to BAN could be attributed to increased regulation of AR expression by the increased TWIST1 levels in BAB compared to BAN. This potential relationship between AR and TWIST1 in balding DPC, which to our knowledge has not been considered previously, provides support for these two candidate genes to be the causative AGA genes at the 7p21.1 and Xq12 susceptibility loci.
      Down-regulated genes in BAB compared to BAN were significantly enriched in vasculature-related gene ontology cluster (Figure 3a and Table 2, and see Supplementary Table S7a), with 41.5% (27 genes) of the vasculature-related genes (see Supplementary Table S7a) being down-regulated in BAB compared to BAN through DHT treatment (see Supplementary Table S7b and Supplementary Figure S7 online). Moreover, a significant number of vasculature-related genes identified (Table 2 and see Supplementary Table S7a) were also found to be down-regulated in independent samples of primary balding DPCs compared to primary non-balding DPCs (P < 0.001, 16 out of 65; see Supplementary Table S7c and Supplementary Materials). The Maz motif, which has been found in promoters of angiogenesis-related genes (
      • Yang R.
      • Amir J.
      • Liu H.
      • Chaqour B.
      Mechanical strain activates a program of genes functionally involved in paracrine signaling of angiogenesis.
      ), was enriched in down-regulated genes in BAB compared to BAN (see Supplementary Table S9a). Furthermore, a trend analysis suggested less coherent vasculature-related gene expression changes in BAB compared to BAN over time (see Supplementary Figure S8 online). Hair follicle vascularization affects follicle size and length (
      • Yano K.
      • Brown L.F.
      • Detmar M.
      Control of hair growth and follicle size by VEGF-mediated angiogenesis.
      ), with the arrangement and density of vasculature altering during anagen (
      • Durward A.
      • Rudall K.M.
      The vascularity and pattern of growth of hair follicles.
      ) and catagen (
      • Ellis R.A.
      • Moretti G.
      Vascular patterns associated with categen hair follicles in the human scalp.
      ). Reduced vasculature has also been observed around the DP of balding vertex scalps compared to non-balding scalps (
      • Cormia F.E.
      • Ernyey A.
      Circulatory changes in alopecia. Preliminary report, with a summary of the cutaneous circulation of the normal scalp.
      ). The ability to attract endothelial cells for follicular vasculature network regeneration is therefore closely tied to hair follicle cycling and growth.
      In summary, the immortalized DPC lines, BAB and BAN, introduced in this study provide a useful resource for AGA-related studies. We identified a set of vasculature-related genes that were down-regulated in balding DPCs under the influence of DHT. With the limitation that our finding is based on two individual cell lines only, we suggest that DPC in balding scalps may be deficient in fostering vasculature development compared to DPC in non-balding scalps. Further in-depth studies are required to validate this finding and to decipher the contribution of identified molecular elements in altering vasculature in the hair follicle and whether this underlies the development of AGA. Our annotation of AGA risk loci with (differentially) expressed genes in BAB compared to BAN has shown that it is insufficient to rely solely on the expression of proximal genes to lead association single-nucleotide polymorphisms for candidate gene discovery at susceptibility loci. Our findings point to AR as the candidate gene at the lead AGA risk locus on chromosome X. We also identified TWIST1 as a functionally relevant gene for AGA at the 7p21.1 susceptibility locus. In-depth studies are required to ascertain the causative role of these candidate genes in the development of AGA.

      Materials and Methods

       Isolation and immortalization of human balding and non-balding primary DPCs

      Between one and three DP were isolated from each matched 2-mm punch biopsy sample of balding (frontal) and non-balding (occipital) scalps of male AGA patients who were undergoing hair transplant surgery and not currently taking hair loss medications, as described previously (
      • Bahta A.W.
      • Farjo N.
      • Farjo B.
      • Philpott M.P.
      Premature senescence of balding dermal papilla cells in vitro is associated with p16(INK4a) expression.
      ,
      • Philpott M.P.
      • Sanders D.
      • Westgate G.E.
      • Kealey T.
      Human hair growth in vitro: a model for the study of hair follicle biology.
      ,
      • Upton J.H.
      • Hannen R.F.
      • Bahta A.W.
      • Farjo N.
      • Farjo B.
      • Philpott M.P.
      Oxidative stress-associated senescence in dermal papilla cells of men with androgenetic alopecia.
      ). Ethics approval was obtained from East London and City health authority (T/98/008), and all biopsies were performed with full patient written consent. All experiments adhered to the Declaration of Helsinki Principles. Isolated primary DPCs were cultured up to passage 3 (see Supplementary Materials and Methods) and immortalized with human telomerase reverse transcriptase using pBABE-hygro-hTERT (a gift from Bob Weinberg; Addgene plasmid # 1773; Addgene, Cambridge, MA) (
      • Counter C.M.
      • Hahn W.C.
      • Wei W.
      • Caddle S.D.
      • Beijersbergen R.L.
      • Lansdorp P.M.
      • et al.
      Dissociation among in vitro telomerase activity, telomere maintenance, and cellular immortalization.
      ). Hygromycin-resistant clones with stable human telomerase reverse transcriptase expression were then cultured as described in the Supplementary Materials and Methods. We were able to derive one immortalized balding (BAB) and one non-balding (BAN) cell line originating from two different male individuals due to limitations in tissue materials, difficulties in establishing pure primary DPC cultures, and low transformation efficiencies. Both BAB and BAN have been established from white men with Hamilton-Norwood grade IV AGA.

       Microarray and processing

      Gene expression in BAB, BAN, primary occipital human follicle dermal papilla cells (HFDPC) and primary fibroblasts (L5-F and normal human dermal fibroblast [NHDF]) were interrogated with HumanHT-12 v4 BeadChip arrays (Illumina, St. Diego, CA). Gene expression of three primary balding DPC samples (passage 2) and two primary non-balding DPC samples (passage 2) were interrogated with HG 133A GeneChips arrays (Affymetrix Inc., Santa Clara, CA) (see Supplementary Figure S1 and Supplementary Materials and Methods). All five primary DPC samples and the immortalized BAB and BAN were isolated from independent subjects.

       Characterization of BAB and BAN

      Immunostaining for versican was conducted on BAB and BAN to determine the expression of DP marker (see Supplementary Materials and Methods). Expression of 117 DP signature genes was interrogated in BAB, BAN, primary balding and non-balding DPCs from AGA individuals, primary occipital human follicle dermal papilla cells (HFDPC), and primary fibroblasts (L5-F and NHDF) (see Supplementary Materials and Methods). Gene expression of DP signature genes was also validated by quantitative real-time reverse transcriptase–PCR in BAB and BAN (see Supplementary Materials and Methods).

       Differential gene expression analysis between BAB and BAN

      BAB and BAN were conditioned in phenol red-free DMEM media for 24 hours and treated with 1 nmol/L or 10 nmol/L of DHT for 15 minutes, 30 minutes, 1 hour, 3 hours, 6 hours, 12 hours, 18 hours, 20 hours, 24 hours, 36 hours, and 48 hours before microarray analysis. Gene ontology analysis was then carried out on differentially expressed genes found in BAB compared to BAN under both 1- and 10-nmol/L DHT treatment (see Supplementary Figure S1 and Supplementary Materials and Methods). The concentration of DHT applied on DPCs was within the adult human male physiological DHT serum levels (
      • Cailleux-Bounacer A.
      • Rohmer V.
      • Lahlou N.
      • Lefebvre H.
      • Roger M.
      • Kuhn J.M.
      Impact level of dihydrotestosterone on the hypothalamic-pituitary-leydig cell axis in men.
      ,
      • Feldman H.A.
      • Longcope C.
      • Derby C.A.
      • Johannes C.B.
      • Araujo A.B.
      • Coviello A.D.
      • et al.
      Age trends in the level of serum testosterone and other hormones in middle-aged men: longitudinal results from the Massachusetts male aging study.
      ). Expression of differentially expressed genes CAV1, EDNRA, IGFBP5, and SCG2 were also validated by quantitative real-time reverse transcriptase–PCR (see Supplementary Materials and Methods). Immunostaining for CYR61, MMP14, and CAV1 was conducted to validate expression of differentially expressed genes (see Supplementary Materials and Methods).

       Annotation of AGA risk loci with differentially expressed genes in BAB compared to BAN

      Because differentially expressed genes at the risk loci were likely to be causative genes that contribute to AGA development, we overlapped differentially expressed genes with 12 lead single-nucleotide polymorphisms with genome-wide association significance using upstream and downstream windows of 500 kb, 100 kb, and 50 kb (see Supplementary Materials and Methods).

      ORCID

      Conflict of Interest

      The authors state no conflict of interest.

      Acknowledgments

      We are grateful for the patients who contributed hair follicles to the study. This work was supported by the Agency for Science, Technology and Research (A*STAR). EC is supported by the A*STAR Graduate Scholarship program. The data reported in this paper have been deposited in Gene Expression Omnibus, http://www.ncbi.nlm.nih.gov/geo (GEO Series accession number GSE66663 and GSE66664).

      Supplementary Material

      References

        • Bahta A.W.
        • Farjo N.
        • Farjo B.
        • Philpott M.P.
        Premature senescence of balding dermal papilla cells in vitro is associated with p16(INK4a) expression.
        J Invest Dermatol. 2008; 128: 1088-1094
        • Boukamp P.
        • Petrussevska R.T.
        • Breitkreutz D.
        • Hornung J.
        • Markham A.
        • Fusenig N.E.
        Normal keratinization in a spontaneously immortalized aneuploid human keratinocyte cell line.
        J Cell Biol. 1988; 106: 761-771
        • Brockschmidt F.F.
        • Heilmann S.
        • Ellis J.A.
        • Eigelshoven S.
        • Hanneken S.
        • Herold C.
        • et al.
        Susceptibility variants on chromosome 7p21.1 suggest HDAC9 as a new candidate gene for male-pattern baldness.
        Br J Dermatol. 2011; 165: 1293-1302
        • Brockschmidt F.F.
        • Hillmer A.M.
        • Eigelshoven S.
        • Hanneken S.
        • Heilmann S.
        • Barth S.
        • et al.
        Fine mapping of the human AR/EDA2R locus in androgenetic alopecia.
        Br J Dermatol. 2010; 162: 899-903
        • Cailleux-Bounacer A.
        • Rohmer V.
        • Lahlou N.
        • Lefebvre H.
        • Roger M.
        • Kuhn J.M.
        Impact level of dihydrotestosterone on the hypothalamic-pituitary-leydig cell axis in men.
        Int J Androl. 2009; 32: 57-65
        • Cormia F.E.
        • Ernyey A.
        Circulatory changes in alopecia. Preliminary report, with a summary of the cutaneous circulation of the normal scalp.
        Arch Dermatol. 1961; 84: 772-789
        • Counter C.M.
        • Hahn W.C.
        • Wei W.
        • Caddle S.D.
        • Beijersbergen R.L.
        • Lansdorp P.M.
        • et al.
        Dissociation among in vitro telomerase activity, telomere maintenance, and cellular immortalization.
        Proc Natl Acad Sci USA. 1998; 95: 14723-14728
        • Danciu T.E.
        • Whitman M.
        Oxidative stress drives disulfide bond formation between basic helix-loop-helix transcription factors.
        J Cell Biochem. 2009; 109: 417-424
        • Durward A.
        • Rudall K.M.
        The vascularity and pattern of growth of hair follicles.
        in: Montagna W. Ellis R.A. The biology of hair growth. Academic, New York1958: 189-217
        • Edwards D.P.
        • Murthy S.R.
        • McGuire W.L.
        Effects of estrogen and antiestrogen on DNA polymerase in human breast cancer.
        Cancer Res. 1980; 40: 1722-1726
        • Elenbaas B.
        • Spirio L.
        • Koerner F.
        • Fleming M.D.
        • Zimonjic D.B.
        • Donaher J.L.
        • et al.
        Human breast cancer cells generated by oncogenic transformation of primary mammary epithelial cells.
        Genes Dev. 2001; 15: 50-65
        • Ellis J.A.
        • Stebbing M.
        • Harrap S.B.
        Genetic analysis of male pattern baldness and the 5alpha-reductase genes.
        J Invest Dermatol. 1998; 110: 849-853
        • Ellis J.A.
        • Stebbing M.
        • Harrap S.B.
        Polymorphism of the androgen receptor gene is associated with male pattern baldness.
        J Invest Dermatol. 2001; 116: 452-455
        • Ellis R.A.
        • Moretti G.
        Vascular patterns associated with categen hair follicles in the human scalp.
        Ann N Y Acad Sci. 1959; 83: 448-457
        • Feldman H.A.
        • Longcope C.
        • Derby C.A.
        • Johannes C.B.
        • Araujo A.B.
        • Coviello A.D.
        • et al.
        Age trends in the level of serum testosterone and other hormones in middle-aged men: longitudinal results from the Massachusetts male aging study.
        J Clin Endocrinol Metab. 2002; 87: 589-598
        • Franco H.L.
        • Casasnovas J.
        • Rodriguez-Medina J.R.
        • Cadilla C.L.
        Redundant or separate entities?—Roles of Twist1 and Twist2 as molecular switches during gene transcription.
        Nucleic Acids Res. 2010; 39: 1177-1186
        • Gong X.Q.
        • Li L.
        Dermo-1, a multifunctional basic helix-loop-helix protein, represses MyoD transactivation via the HLH domain, MEF2 interaction, and chromatin deacetylation.
        J Biol Chem. 2002; 277: 12310-12317
        • Hamilton J.B.
        Male hormone stimulation is prerequisite and an incitant in common baldness.
        Am J Anat. 1942; 71: 451-480
        • Hamilton J.B.
        Patterned loss of hair in man; types and incidence.
        Ann N Y Acad Sci. 1951; 53: 708-728
        • Heilmann S.
        • Kiefer A.K.
        • Fricker N.
        • Drichel D.
        • Hillmer A.M.
        • Herold C.
        • et al.
        Androgenetic alopecia: identification of four genetic risk loci and evidence for the contribution of WNT signaling to its etiology.
        J Invest Dermatol. 2013; 133: 1489-1496
        • Hibberts N.A.
        • Howell A.E.
        • Randall V.A.
        Balding hair follicle dermal papilla cells contain higher levels of androgen receptors than those from non-balding scalp.
        J Endocrinol. 1998; 156: 59-65
        • Higgins C.A.
        • Chen J.C.
        • Cerise J.E.
        • Jahoda C.A.
        • Christiano A.M.
        Microenvironmental reprogramming by three-dimensional culture enables dermal papilla cells to induce de novo human hair-follicle growth.
        Proc Natl Acad Sci USA. 2013; 110: 19679-19688
        • Hillmer A.M.
        • Brockschmidt F.F.
        • Hanneken S.
        • Eigelshoven S.
        • Steffens M.
        • Flaquer A.
        • et al.
        Susceptibility variants for male-pattern baldness on chromosome 20p11.
        Nat Genet. 2008; 40: 1279-1281
        • Hillmer A.M.
        • Hanneken S.
        • Ritzmann S.
        • Becker T.
        • Freudenberg J.
        • Brockschmidt F.F.
        • et al.
        Genetic variation in the human androgen receptor gene is the major determinant of common early-onset androgenetic alopecia.
        Am J Hum Genet. 2005; 77: 140-148
        • Ho S.R.
        • Mahanic C.S.
        • Lee Y.J.
        • Lin W.C.
        RNF144A, an E3 ubiquitin ligase for DNA-PKcs, promotes apoptosis during DNA damage.
        Proc Natl Acad Sci USA. 2014; 111: E2646-E2655
        • Hodgins M.B.
        • Choudhry R.
        • Parker G.
        • Oliver R.F.
        • Jahoda C.A.
        • Withers A.P.
        • et al.
        Androgen receptors in dermal papilla cells of scalp hair follicles in male pattern baldness.
        Ann N Y Acad Sci. 1991; 642: 448-451
        • Itami S.
        • Sonoda T.
        • Kurata S.
        • Takayasu S.
        Mechanism of action of androgen in hair follicles.
        J Dermatol Sci. 1994; 7: S98-S103
        • Kikkert M.
        • Doolman R.
        • Dai M.
        • Avner R.
        • Hassink G.
        • van Voorden S.
        • et al.
        Human HRD1 is an E3 ubiquitin ligase involved in degradation of proteins from the endoplasmic reticulum.
        J Biol Chem. 2004; 279: 3525-3534
        • Kishimoto J.
        • Ehama R.
        • Wu L.
        • Jiang S.
        • Jiang N.
        • Burgeson R.E.
        Selective activation of the versican promoter by epithelial- mesenchymal interactions during hair follicle development.
        Proc Natl Acad Sci USA. 1999; 96: 7336-7341
        • Kitagawa T.
        • Matsuda K.
        • Inui S.
        • Takenaka H.
        • Katoh N.
        • Itami S.
        • et al.
        Keratinocyte growth inhibition through the modification of Wnt signaling by androgen in balding dermal papilla cells.
        J Clin Endocrinol Metab. 2009; 94: 1288-1294
        • Kratochwil K.
        • Dull M.
        • Farinas I.
        • Galceran J.
        • Grosschedl R.
        Lef1 expression is activated by BMP-4 and regulates inductive tissue interactions in tooth and hair development.
        Genes Dev. 1996; 10: 1382-1394
        • Lee Y.S.
        • Lee H.H.
        • Park J.
        • Yoo E.J.
        • Glackin C.A.
        • Choi Y.I.
        • et al.
        Twist2, a novel ADD1/SREBP1c interacting protein, represses the transcriptional activity of ADD1/SREBP1c.
        Nucleic Acids Res. 2003; 31: 7165-7174
        • Li L.
        • Cserjesi P.
        • Olson E.N.
        Dermo-1: a novel twist-related bHLH protein expressed in the developing dermis.
        Dev Biol. 1995; 172: 280-292
        • Li R.
        • Brockschmidt F.F.
        • Kiefer A.K.
        • Stefansson H.
        • Nyholt D.R.
        • Song K.
        • et al.
        Six novel susceptibility Loci for early-onset androgenetic alopecia and their unexpected association with common diseases.
        PLoS Genet. 2012; 8: e1002746
        • Midorikawa T.
        • Chikazawa T.
        • Yoshino T.
        • Takada K.
        • Arase S.
        Different gene expression profile observed in dermal papilla cells related to androgenic alopecia by DNA macroarray analysis.
        J Dermatol Sci. 2004; 36: 25-32
        • Nyholt D.R.
        • Gillespie N.A.
        • Heath A.C.
        • Martin N.G.
        Genetic basis of male pattern baldness.
        J Invest Dermatol. 2003; 121: 1561-1564
        • Ohyama M.
        • Kobayashi T.
        • Sasaki T.
        • Shimizu A.
        • Amagai M.
        Restoration of the intrinsic properties of human dermal papilla in vitro.
        J Cell Sci. 2012; 125: 4114-4125
        • Paus R.
        • Cotsarelis G.
        The biology of hair follicles.
        N Engl J Med. 1999; 341: 491-497
        • Philpott M.P.
        • Sanders D.
        • Westgate G.E.
        • Kealey T.
        Human hair growth in vitro: a model for the study of hair follicle biology.
        J Dermatol Sci. 1994; : S55-S72
        • Prodi D.A.
        • Pirastu N.
        • Maninchedda G.
        • Sassu A.
        • Picciau A.
        • Palmas M.A.
        • et al.
        EDA2R is associated with androgenetic alopecia.
        J Invest Dermatol. 2008; 128: 2268-2270
        • Ramot Y.
        • Tiede S.
        • Biro T.
        • Abu Bakar M.H.
        • Sugawara K.
        • Philpott M.P.
        • et al.
        Spermidine promotes human hair growth and is a novel modulator of human epithelial stem cell functions.
        PLoS One. 2011; 6: e22564
        • Rexbye H.
        • Petersen I.
        • Iachina M.
        • Mortensen J.
        • McGue M.
        • Vaupel J.W.
        • et al.
        Hair loss among elderly men: etiology and impact on perceived age.
        J Gerontol A Biol Sci Med Sci. 2005; 60: 1077-1082
        • Rhee H.
        • Polak L.
        • Fuchs E.
        Lhx2 maintains stem cell character in hair follicles.
        Science. 2006; 312: 1946-1949
        • Richards J.B.
        • Yuan X.
        • Geller F.
        • Waterworth D.
        • Bataille V.
        • Glass D.
        • et al.
        Male-pattern baldness susceptibility locus at 20p11.
        Nat Genet. 2008; 40: 1282-1284
        • Shiota M.
        • Yokomizo A.
        • Tada Y.
        • Inokuchi J.
        • Kashiwagi E.
        • Masubuchi D.
        • et al.
        Castration resistance of prostate cancer cells caused by castration-induced oxidative stress through Twist1 and androgen receptor overexpression.
        Oncogene. 2010; 29: 237-250
        • Soma T.
        • Tajima M.
        • Kishimoto J.
        Hair cycle-specific expression of versican in human hair follicles.
        J Dermatol Sci. 2005; 39: 147-154
        • Tabach Y.
        • Milyavsky M.
        • Shats I.
        • Brosh R.
        • Zuk O.
        • Yitzhaky A.
        • et al.
        The promoters of human cell cycle genes integrate signals from two tumor suppressive pathways during cellular transformation.
        Mol Syst Biol. 2005; 1 (2005.0022–2005.0022)
        • Trueb R.M.
        Molecular mechanisms of androgenetic alopecia.
        Exp Gerontol. 2002; 37: 981-990
        • Upton J.H.
        • Hannen R.F.
        • Bahta A.W.
        • Farjo N.
        • Farjo B.
        • Philpott M.P.
        Oxidative stress-associated senescence in dermal papilla cells of men with androgenetic alopecia.
        J Invest Dermatol. 2015; 135: 1244-1252
        • Xu Y.
        • Liao L.
        • Zhou N.
        • Theissen S.M.
        • Liao X.H.
        • Nguyen H.
        • et al.
        Inducible knockout of Twist1 in young and adult mice prolongs hair growth cycle and has mild effects on general health, supporting Twist1 as a preferential cancer target.
        Am J Pathol. 2013; 183: 1281-1292
        • Yang R.
        • Amir J.
        • Liu H.
        • Chaqour B.
        Mechanical strain activates a program of genes functionally involved in paracrine signaling of angiogenesis.
        Physiol Genomics. 2008; 36: 1-14
        • Yano K.
        • Brown L.F.
        • Detmar M.
        Control of hair growth and follicle size by VEGF-mediated angiogenesis.
        J Clin Invest. 2001; 107: 409-417