A Systems Biology Approach to Anatomic Diversity

8/13/2019 A Systems Biology Approach to Anatomic Diversity

1/7

A Systems Biology Approach to Anatomic Diversity

of SkinJohn L. Rinn1, Jordon K. Wang1, Helen Liu1, Kelli Montgomery2, Matt van de Rijn2 and Howard Y. Chang1

Human skin exhibits exquisite site-specific morphologies and functions. How are these site-specific differencesspecified during development, maintained in adult homeostasis, and potentially perturbed by diseaseprocesses? Here, we review progress in understanding the anatomic patterning of fibroblasts, a majorconstituent cell type of the dermis and key participant in epithelialmesenchymal interactions. The geneexpression programs of human fibroblasts largely reflect the superimposition of three gene expression profilesthat demarcate the fibroblasts position relative to three developmental axes. The HOX family of homeodomaintranscription factors is implicated in specifying site-specific transcriptional programs. The use of gene, tiling, andtissue microarrays together gives a comprehensive view of the gene regulation involved in patterning the skin.

Journal of Investigative Dermatology (2008)128, 776782; doi:10.1038/sj.jid.5700986

IntroductionThe human skin shows remarkablediversity in its structure and functionacross anatomic sites. Scalp skin iseasily recognizable by the numerousterminal hair follicles; in contrast pal-moplantar skin possesses no hair folli-cles but is characterized by increasednumber of eccrine glands and compacthyperkeratosis of the stratum corneum.The anatomic diversity of human skin is

also reflected in the site specificity ofmany skin diseases and their responseto treatment; the distribution of skin

lesions is indeed often one of the keyclues to the correct diagnosis. From abasic science perspective, the anatomicdiversity of skin raises many intriguingquestions about how cells acquire andmaintain their positional identities in acomplex, self-renewing tissue. Here, wereview recent progress in understandingsite-specific differentiation of cell typesin the skin, focusing on emerging

systems biology approaches that arebeginning to provide comprehensivedescriptions of this fascinating biology.

Studies of skin development haveshown that site-specific differentiationof epithelia critically depends onepithelialmesenchymal interactions.Hair and other skin appendagesdevelop through a complex series ofreciprocal interactions between epi-dermal cells and fibroblasts, beginningwith a signal from dermal fibroblasts tothe overlying epidermis to proliferate,

forming the placodes that are progeni-tors of hair follicles (Millar, 2002).Classic heterotopic recombination

PERSPECTIVE

776 Journal of Investigative Dermatology (2008), Volume 128 &2007 The Society for Investigative Dermatology

Editors Note

In 1950, Dr William Montagna, a biologist at BrownUniversity, began a symposium that was designed to bringtogether basic scientists and clinically trained dermato-logists to further our understanding of skin (see Origin ofthe annual symposium on the biology of skin. Kligman AM(2002) J Invest Dermatol Symp Proc7:1). This month we

celebrate the 56th Annual Montagna Symposium on theBiology of Skin, which focused on the development anddiseases of skin appendages, with two Perspectives articles.Dr Montagna would be pleased with such a topic, as muchof his investigative interest was focused on skin appen-dages, including hair and sebaceous glands, and on the

development of skin, from mice to primates. In the firstarticle, Rinn and co-authors highlight how the emergingscience of systems biology can be used to study andunderstand the wide diversity that is present in skin. In thesecond article, Wang and co-authors describe howsignaling in the skin through Smads can impact skin

appendage development. These articles highlight how theapplication of new technology and new signaling pathwayscan be applied to skin biology to help us understand thecomplexity of skin.

Russell P. Hall III,Deputy Editor

Received 15 February 2007; revised 10 April 2007; accepted 5 May 2007

1Program in Epithelial Biology, Department of Dermatology, Stanford University, Stanford, California, USA and 2Program in Epithelial Biology, Department ofPathology, Stanford University, Stanford, California, USA

Correspondence: Professor Howard Y. Chang, CCSR2155c, 269 Campus Drive, Stanford, California 94305, USA. E-mail: [email protected]

Abbreviations: EC, endothelial cell; nt, nucleotide


2/7

experiments showed that a primarydermal signal dictates the overlyingepithelial fate; for instance, grafting ofwing epidermis to foot dermis in chickled to the development of scales ratherthan feathers (Dhouailly, 1973, 1984).

Specialized fibroblasts from the dermalpapilla of hair follicles (Jahoda et al.,1984), but not dermal fibroblasts a fewhundred microns away, are able toinduce de novo hair follicle develop-ment when transplanted into nave skinwith epidermal stem cells (Blanpainet al., 2004). Moreover, the type of hairvaries throughout the body, and thispositional information also is dictatedby the local fibroblasts. For example,dermal papilla fibroblasts derived fromwhiskers induce long, thick whisker-

like hairs when transplanted into reci-pient animals at heterotopic sites(Jahoda, 1992). These intimate andspecific epithelialmesenchymal inter-actions are not confined only to skin,but appear to be a major theme in thedevelopment and homeostasis of allepithelial organs: local fibroblast-likecells in the urogenital sinus inducedifferentiation of prostatic epithelialprecursors to form the prostate gland(Cunha, 1994), and local fibroblastsare also responsible for metanephric

induction and pruning of nephrons inthe kidney (Schedl and Hastie, 2000;Levinson and Mendelsohn, 2003).Branching morphogenesis of the lungsimilarly depends critically on recipro-cal interactions between bud epithelialcells and surrounding fibroblasts (Shan-non and Hyatt, 2004). Because theepidermis is continually shed andreplaced by newly developed keratino-cytes (every 28 days in humans),it stands to reason that the site-specificinductive capacity of fibroblasts must

persist into adulthood, perhaps throughthe entire lifetime. For instance, cellcell contact between adult palmoplan-tar fibroblasts with trunk keratinocytesreprograms these keratinocytes to ex-press palmoplantar keratin genes (Ya-maguchiet al., 1999).

Fibroblasts are the principal celltype in the dermis and stroma ofinternal organs that synthesize connec-tive tissue proteins such as collagen(Dunphy, 1963; Gabbiani and Rung-ger-Brindle, 1981). Fibroblasts were

first described histologically in 1847by Schwann as spindle-shaped orlongish corpuscles which are thickestin the middle and gradually elongatedin both extremities into minute fibres(Schwann, 1847). In practice, fibro-

blasts are usually identified by theirspindle-shaped morphology, the abilityto adhere to plastic tissue culturevessels, and the absence of markersfor other cell lineages (Dunphy, 1963;Gabbiani and Rungger-Brindle, 1981;Normand and Karasek, 1995). Giventhese rather non-selective criteria, it isnot surprising that cells we currentlyconsider fibroblasts may comprise adiverse group of cells, each withdistinct patterns of synthetic activitiesand functions.

Site-specific differentiation of fibroblastsA first test of this idea came with thecomprehensive gene expression analy-sis of primary human dermal fibroblastsfrom distinct anatomic sites (Changet al., 2002). Using cDNA array tech-nology, we analyzed the expression ofB21,000 genes in 50 primary humanfibroblasts culture from 10 anatomicsites. The results revealed that althoughall fibroblasts are morphologicallysimilar, the gene expression patterns

of cultured fibroblasts from differentsites are strikingly distinct. Approxi-mately 8% of all genes transcribed infibroblasts are differentially expressedin a site-specific manner. The variationand magnitude of gene expressiondifferences among fibroblasts are com-parable to the variation observedamong different types of white bloodcells. When their gene expressionpatterns were grouped by similarityusing a technique called hierarchicalclustering (Eisen et al., 1998), fibro-

blasts from the same topographic sitesof the skin were consistently groupedtogether, and the distinctiveness oftopographic gene expression was notobscured among different donors, bypassage in tissue culture, or by envir-onmental changes such as serum star-vation. In four cases where fibroblastswere derived from multiple sites of thesame individual, fibroblasts from thesame site of different individuals werefar more similar to each other thanfibroblasts from different sites of the

same individual. These results demon-strate that in fact there are manydifferent cell types that go under thetraditional heading of fibroblasts.The main rule of differentiation amongdermal fibroblasts appears to be dic-

tated by the anatomic site of origin.The genes expressed by fibroblasts

in a site-specific manner demonstratedistinct choreographed programs inextracellular matrix synthesis, lipidmetabolism, and signaling pathwayscontrolling cell migration and cell fatespecification (Chang et al., 2002). Thegene expression patterns particular todistinct types of fibroblasts were bio-logically consistent with their anatomicalorigin. The site-specific expression ofmany cell growth and differentiation

molecules such as members of TGF-b,Wnt, receptor tyrosine kinase andphosphatase families indicates thatfibroblasts make important contri-butions in mesenchymal induction ofepithelia. Moreover, the expressiondomains of genes underlying severalgenetic diseases affecting skin ormusculoskeletal connective tissue cor-related closely with the phenotypicdefects. For example, by comparinggenes that were induced in dermalfibroblasts compared to lung fibro-

blasts, we were able to identify thegenes involved in 6 out of 10 types ofEhlersDanlos syndrome, a congenitaldisease characterized by skin fragilityand joint laxity. Similarly, we observedthat HOXA13 is induced in toe andforeskin fibroblasts, and mutation ofHOXA13 in humans leads to hand-footgenital syndrome, a disease char-acterized by syndactyly, hypospadias,and malformations of the urogenitalsystem. These results indicate thatmany important aspects of site-specific

differentiation in fibroblasts are pre-served in vitro and thus amenable todissection by molecular and functionalgenomic approaches.

Many epithelialmesenchymal inter-actions that specify epidermal appen-dages occur within local signalingenvironments of even finer specializationof fibroblasts. Overlaid on the regionalfibroblast specialization, recent genomicprofiling experiments have indeed iden-tified gene expression signatures of suchspecialized cells, such as that of the

www.jidonline.org 7

JL Rinnet al.Systems Biology Approach to Anatomic Diversity of Skin


3/7

dermal papilla cells (Rendlet al., 2005).Similarly, we and others have identifiedunique gene expression signatures ofstromal fibroblasts of basal-cell carcino-mas, a cancer of hair follicle origin, thatdistinguished such cells from local fibro-

blasts of the face (Sneddon et al., 2006).In contrast, genome-wide expression

profiling of purified endothelial cells(ECs) revealed that unlike fibroblasts, ECdifferentiation is primarily dictated bythe vessel size of origin. ECs derivedfrom large vessels from various ana-tomic locations are more similar to eachother than ECs from microvasculature.In particular, dermal microvascular ECsfrom different anatomic sites were notparticularly distinct and were similar tomicrovascular ECs from intestine and

lung (Chi et al., 2003). These resultsreveal the contrasting regulatory logic ofdifferent stromal cell types and reinforcethe uniqueness and likely develop-mental importance of the site-specificdifferentiation of fibroblasts.

Patterning of adult fibroblasts alongthree developmental axesIn his influential treatise on develop-ment Lewis Wolpert (1969), famouslypostulated two potential mechanismsof achieving pattern formation: spatial

organization of cellular differentiationmay be achieved by unique specifica-tion of each cell type; alternatively,organization may arise by cells inter-preting their position relative to refer-ence points, and adopting specificdifferentiation programs based on theirpositional identity within a coordinatesystem. A particularly attractive anddistinctive feature of the positionalidentity model is the parsimonious useof molecular entities to construct thesystem, leading to universality of the

coordinate system. For example, thesame reference points can be used tospecify the proximaldistal axis of theupper and lower limbs even though thelimbs are spatially distant from eachother. Similarly, the same referencepoints can pattern the anteriorposter-ior axis of the trunk and also distinguishthe upper and the lower limbs.

We hypothesized that these modelsof pattern formation can be distin-guished by comparing the global geneexpression profiles of cells distributed

throughout the body. Pattern formationby unique specification predicts that thesimilarity of gene expression profiles ofcells would be sporadic and not relatedby distance from their sites of origin.Conversely, if the positional model of

specification is at work, expressionprofiles will exhibit spatial relationshipsbetween fibroblasts in similar positionalquadrants of the body.

To distinguish between these mod-els, we analyzed the genome-wide geneexpression profiles of 47 primary fibro-blast cultures from 43 unique anatomi-cal sites spanning the entire humanbody (Figure 1a) (Rinnet al., 2006). Weconfirmed at a much finer level thatfibroblasts exhibited large-scale differ-ences in their gene expression programs

in a site-specific fashion. Moreover,comparison of gene expression differ-ences among cells from local vicinitiesversus those far away revealed thatthese gene expression differences areglobally related to three recurringsegmental patterns: proximaldistal,anteriorposterior (rostralcaudal), and

internalexternal (non-dermal vs der-mal) (Figure 1b). For instance, fibro-blasts originating from distal upper limb(hand, fingers) had a distinct geneexpression signature compared to thosefrom more proximal sites of the arm;

fibroblasts from the leg also showeddistinct gene expression signatures thatportioned them to those above andbelow the ankle. Remarkably, compar-ison of the distal arm and distal leg genesignatures showed that these signatureswere largely the same, suggesting thatthis was a signature that reflected distalposition along a developmental axis,regardless of the actual position onupper or lower limb. Similarly, the geneexpression data that automatically dis-tinguished fibroblasts originating from

the top half of the body (anterior orrostral) had a distinctive gene expres-sion profile than from those from thebottom half of the body (posterior orcaudal). A third gene expression signa-ture distinguished dermal fibroblastsfrom fibroblasts that originate frominternal organs. These findings suggest

Ds

a

c

b

Ds

D HOXD HOXD

HOXB

HOXA13 HOXA13 Internal Chest

Upr leg Foot

Frsk Hand

Upr arm HOXB9

A

P

V

Px

Figure 1. The embryonic pattern of HOX gene expression is retained in adult human fibroblasts.

(a) A total of 47 primary cultures representing 43 unique anatomic sites (gray circles) that finely map the

human body were profiled by gene expression microarrays. (b) Model of fibroblast differentiation by

overlapping positional patterns of gene expression: proximal (yellow), distal (orange), anterior (blue),

posterior (green) and internal organs (red). (c) A decision tree of HOX expression that distinguish unique

anatomic positions. Right: red indicates higher-than-average and green lower-than-average expression of

a given HOX gene, relative to the average expression level across 47 cultures. Each anatomic site can be

correctly identified by monitoring the expression level of three HOX genes or less. Left: the

developmental axes that demarcate site-specific gene expression of fibroblasts: anteriorposterior,

proximaldistal and dermalnon-dermal. A third developmental axis, dorsalventral, did not correlate

with large-scale site-specific gene expression in fibroblasts.

78 Journal of Investigative Dermatology (2008), Volume 128



4/7

that site-specific variations in fibroblastgene expression programs are not idio-syncratic, but rather are systematicallyrelated to their positional identitiesrelative to major anatomic axes. Muchlike a global positioning system, the

site-specific gene expression program ofa fibroblast reflects the superimpositionof three positional coordinates, poten-tially providing critical cues for devel-opment, trafficking, and homeostasis ofsurrounding cells in the skin.

The positional identities of adultfibroblasts raise the question ofwhether their cognate coordinate sys-tem was established during embryonicdevelopment. During embryogenesis,expression of specific HOX genesdemarcates distinct positional identities

that lead to site-specific cellular differ-entiation and tissue morphogenesis.The HOX family of homeodomaintranscription factors comprise 39 genesthat are clustered on four chromosomalloci; their physical order on the chro-mosomes reflects their spatial pattern ofexpression along the anteriorposteriorand proximaldistal axes of the em-bryoa property called colinearity.We found that a very similar patternof HOX gene expression was alsoretained in fibroblasts both in vitro

andin vivo (Rinnet al., 2006). In fact,HOX gene expression alone was pre-dictive of the anatomic origin of a givenfibroblast culture (Figure 1c). Forexample, both dermal and non-dermalfibroblasts from the trunk expressHOXB genes. However, HOXD geneexpression is limited to dermal fibro-blasts and is not found in non-dermalfibroblasts. Moreover, distal fibroblastscan be distinguished from proximalfibroblasts by HOXA13 gene expres-sion. Thus, using a very simple decision

tree of HOX gene expression, theanatomic position of a fibroblast canbe predicted (Figure 1c). HOX genesmay also be globally expressed in skinin select circumstances, such as theHOXC13 gene during terminal hairfollicle development (Godwin andCapecchi, 1998); however, we mainlyobserved canonical, site-specific ex-pression of HOX genes in fibroblasts.The salient description of fibroblastanatomic origin by HOX gene expres-sion suggests that this family of

transcription factors may have a rolein the establishment of site-specificfibroblast gene expression programs.

Revealing the epigenetic landscapeby tiling microarrays

Major features of the embryonic patternof HOX gene expression are retained inadult human fibroblasts from youngand old donors alike, and these site-specific patterns of HOX gene expres-sion are also preserved after extensiveex vivo cell divisions (J.L.R. andH.Y.C., unpublished observation). Thestability of the site-specific transcrip-tional patterns suggests the possibilityof epigenetic mechanisms in theirmaintenance. Classic genetic studiesin Drosophila established that HOX

gene expression patterns are main-tained by two opposing histone modifi-cation activities. The Polycomb groupproteins mediate histone H3 lysine 27methylation and are required for themaintenance of HOX gene repression,whereas the Trithorax proteins mediatehistone H3 lysine 4 methylation andare required for maintenance of HOX

gene activation (Ringrose and Paro,2004). In addition, intergenic transcrip-tion of long noncoding RNAs alsoplays important roles in the epigeneticmaintenance of HOX transcriptionalpatterns (Schmitt and Paro, 2006).

While some of these components areknown in model organisms, the tran-scriptional and epigenetic landscapesin human cells are much less under-stood. A comprehensive view of tran-scription and chromatin structure ofthe HOX loci is needed to decipherthe regulatory mechanisms that resultin the spatial and temporal patterningof HOX gene expression.

Primary cells from distinct anatomicsites of human skin offer a uniqueresource of purified cells of specific

positional identities that may be used tointerrogate the epigenetic mechanismsof site-specific HOX expression pro-grams. We and others have used atechnology called tiling microarrays togain a comprehensive view of the HOXloci (Bernstein et al., 2005; Guentheret al., 2005; Squazzo et al., 2006)(Figure 2). Unlike conventional DNA

50 nt

Silent chromation domain Active chromation domain

5 nt

5

5 3

+40

+4

Log2

(ChlP/input)

+4

H3-K27me3

Suz12

H3-K4me2

Pol II

RNA

0

+40

70,000

A13 A11 A10 A9 A7 A6 A5 A4 A3 A1

0

3

HOXN+1 HOXN

a

b

Figure 2. The epigenetic regulation of HOX expression patterns in fibroblasts. (a) Schematic

representation of the DNA tiling design consisting ofB400,000 fifty-nucleotide probes that overlap by

45 nt such that each probe interrogates a unique 5 nt sequence. The tiling array covers all four human

HOX loci and 2 Mb of control regions. Brown boxes represent specific HOX genes. (b) Opposing histone

modifications demarcate transcriptional accessibility of HOXA locus. The top four rows show occupancy

for the named protein (polycomb group protein Suz12 or RNA polymerase II (PolII)) or histone

modification (H3K27me3 and H3K4me2) across B100kb of the HOXA locus (xaxis), as measured by

ChIP-chip. Theyaxis plots the ratio of hybridization signals of chromatin immunoprecipitation over input

genomic DNA in log 2 space. The bottom row shows RNA hybridization signal over the same genomic

region in linear scale (070,000 intensity units). Chromatin immunoprecipitation experiments were

performed as described (Squazzo et al., 2006).

www.jidonline.org 7



5/7

microarrays where one probe is placedfor each gene of interest, the tilingmicroarray employs many more probesto tile across genomic regions ofinterest with large overlap betweenadjacent probes (Figure 2a). We

constructed a HOX tiling array thatinterrogates all four human HOX loci(HOXA, B, C, D) using 50-mer oligo-nucleotide probes that overlap by 45nucleotides (nt). Thus, each 5 nt ofDNA sequence is uniquely interro-gated. Hybridization of RNA or chro-matin immunoprecipitation on tilingmicroarrays (so-called ChIP-chip ana-lysis) then reveals all of the transcriberegions or sites of chromatin modifica-tion(s) in an unbiased manner. Con-ventional ChIP experiments typically

yield DNA fragments of 5001,000base pairs (bp) and hence can beinterrogated by much lower resolutionpromoter or tiling arrays. However,interrogation of nucleosome position-ing (B140 bp per nucleosome) andnovel noncoding RNA (including smallRNAso100 nt) require substantiallyhigher resolution for their demarcation.

We and others have found evidenceof broad domains of HOX chromatinmodifications that demarcate activeand silent regions of the HOX loci

(Bernstein et al ., 2005; Guentheret al., 2005; Squazzo et al., 2006).Actively transcribed domains of HOXare marked by RNA polymerase IIoccupancy and histone H3 lysine 4dimethylation in the gene-coding andintergenic regions whereas transcrip-

tionally silent domains are marked byhistone H3 lysine 27 trimethylation andPolycomb group protein occupancy,such as the Polycomb group proteinsubunit Suz12 (Figure 2b). These chro-matin domains are regulated in a site-

specific manner. Moreover, a numberof noncoding RNAs are transcribed in asite-specific manner (Sessa et al .,2007).

In addition to normal homeostasis, asimilar genomic approach may beemployed to investigate diseases ofthe skin. The HOX loci are alsoimportant for epidermal appendagedevelopment and regulation of cellgrowth. HOXC13 is required for hairoutgrowth (Godwin and Capecchi,1998). Translocation of the human

Trithorax protein gene MLL, leading toactivation of the HOX loci, is afrequent etiology of leukemias(Guenther et al., 2005), and the Poly-comb group protein gene EZH2 isfrequently amplified in human epithe-lial tumors and in melanoma (Brackenet al., 2003). For instance, studies ofgenome-wide occupancy sites of MLLin leukemic cells have suggested that itfunctions as transcriptional start site-specific histone methyltransferase(Guenther et al., 2005). With the easy

accessibility of skin disease tissues andcells, interrogation with tiling micro-arrays is a promising tool to enhanceour understanding of the epigeneticregulation of regulatory genes in skinand how these epigenetic codes maybe disrupted during disease.

Tissue microarray: genomics and proteo-mics in the context of skin architectureGene and tiling microarrays are usefulfor elucidating the genetic and epige-netic elements that differentiate celltypes across the body; however, it is

equally important to determine the invivoanatomic localization of genes inthe three-dimensional context of theskin. To address this challenge, wehave constructed a skin diversitytissue microarray, where multipleskin sections are placed on a singleslide to be used for in situhybridizationand immunohistochemistry. Our tissuemicroarray is comprised of 42, two-millimeter formalin-fixed, paraffin-em-bedded cores of skin from diverseanatomic sites and 8 internal organs

such as cervix, intestine, lung, liver andbone (Figure 3a). Immunohistoche-mistry or RNA in situ hybridizationcan be performed on all 50 tissuesin parallel, allowing unbiasedand high-throughput comparison ofprotein or gene expression levelsand localization. A potential limitationof this technology is that proteinsand mRNAs present in low levels maybe better visualized in frozen sectionsthan formalin-fixed tissues, and condi-tions for antigen retrieval or signal

amplification may need to be devel-oped to visualize low-abundance geneproducts.

To illustrate the use of such a skindiversity tissue microarray, we per-formed RNA in situ hybridizations forkeratin 14 and keratin 9. As expected,

Plantar foot

Dorsal hand

Keratin 14 Keratin 9

2 3

4

57

6

14

15

16

17

18

19

20

21

9

1

8

10

11

12

13

a b

Figure 3. A human skin diversity tissue microarray. (a) Positional map of tissue microarray (left) comprised of 42, two-millimeter longitudinal cross-sections

of skin and 8 sections of other organs, in total representing 21 unique sites of skin (blue circles). Hematoxylin and eosin staining of the tissue microarray

(middle) and zoom in of cores representing plantar (top) and dorsal hand skin (bottom). Bar 250mm. (b) Skin section from plantar skin (top) and dorsal hand

(bottom) following in situhybridization with probes complimentary to either keratin 14 or keratin 9 mRNA. Bar 50mm.




6/7

we observed expression of keratin 14 inthe basal layer of epidermis in skinfrom all anatomic sites. Keratin 9 is asuprabasal keratin of palmoplantarskin, and indeed only palmoplantarskin on our tissue microarray showed

strong keratin 9 signal (Figure 3). Theskin diversity tissue microarray shouldbe useful for the discovery and valida-tion of novel site-specific genes orsignaling pathways. Again, this ap-proach can be extended to includetissues of skin diseases to monitor theinvivo expression of genes perturbed indisease. The combined power of geneexpression, tiling and tissue micro-arrays will greatly facilitate our under-standing of the genes that are importantin skin patterning and their roles in skin

disease.

Conclusion and future challengesThe use of multiple systems biologyapproaches has started to paint apicture of the diversity of humandermal fibroblasts and the anatomicpatterning of skin at the molecularlevel. This area of investigation is stillat an early stage and much remains tobe learned. Four challenges are likelyto engross investigators in the nearfuture. First, the transcriptional network

of HOX genes in adult skin needs tobe clarified. The mammalian targetsof human HOX genes have eludeddetection for many years owing to lackof human material and the embryoniclethality of most HOX genes in mice(Svingen and Tonissen, 2006). Primaryhuman fibroblasts may be a tractablesystem to study the transcriptionalnetwork of HOX genes. Second, theepigenetic mechanisms that maintainsite-specific gene expression programsremains incompletely understood. The

identification of specific chromatindomains, their specific histone modifi-cations, and associated noncodingRNAs are providing a list of candidatefactors that may play a role in thisregulatory program. Third, how thesemechanisms of positional identity innormal skin may relate to the patho-genesis of many skin diseases withsite-specific manifestations is largelyunknown. Nonetheless, the historyof investigative dermatology providesmany examples of developmental

pathways that become subverted anddrive skin diseases, including diseasesof epidermal adhesion, inflammation,and skin cancers. Fourth, site-specificdifferentiation is a dynamic process.For instance, hair cycling in many

species occur in a seasonal fashion,and in mice hair cycling progressesgradually in an anteriorposterior fash-ion (Stenn and Paus, 2001). The use ofconditional genetic approaches thatcan synchronize dynamic developmen-tal programs by inducible arrest andrelease may aid to capture the temporalregulation of site-specific differentia-tion (Hutchin et al., 2005; Sarin et al.,2005; Estrach et al., 2006). It is likelythat a multifaceted approach incorpor-ating emerging technologies will pro-

vide a wealth of information about themolecular cues involved in skin pat-terning and their dysregulation in skindiseases.

CONFLICT OF INTERESTThe authors state no conflict of interest.

ACKNOWLEDGMENTSWe thank members of Chang lab and the StanfordProgram in Epithelial Biology for discussions. Thisstudy was supported by NIAMS K08-AR050007(H.Y.C.), Damon Runyon Cancer ResearchFoundation Postdoctoral Fellowship (J.L.R.),

National Science Foundation PredoctoralFellowship (J.K.W.), and Stanford Med ScholarsProgram and American Skin Association (H.L.).H.Y.C. is the Kenneth G. and Elaine A. LangoneScholar of the Damon Runyon Cancer ResearchFoundation.

REFERENCES

Bernstein BE, Kamal M, Lindblad-Toh K, Bekir-anov S, Bailey DK, Huebert DJ et al. (2005)Genomic maps and comparative analysis ofhistone modifications in human and mouse.Cell120:16981

Blanpain C, Lowry WE, Geoghegan A, Polak L,

Fuchs E (2004) Self-renewal, multipotency,and the existence of two cell populationswithin an epithelial stem cell niche. Cell118:63548

Bracken AP, Pasini D, Capra M, Prosperini E,Colli E, Helin K (2003) EZH2 is downstreamof the pRB-E2F pathway, essential for pro-liferation and amplified in cancer. EMBO J22:532335

Chang HY, Chi J-T, Dudoit S, Bondre C, van deRijn M, Botstein D et al. (2002) Diversity,topographic differentiation, and positionalmemory in human fibroblasts. Proc NatlAcad Sci USA 99:1287782

Chi JT, Chang HY, Haraldsen G, TroyanskayaOG, Chang DS, Wang Z et al. (2003)

Endothelial diversity revealed by globalexpression profiling. Proc Natl Acad SciUSA100:106238

Cunha GR (1994) Role of mesenchymalepithelialinteractions in normal and abnormal devel-opment of the mammary gland and prostate.Cancer74:103044

Dhouailly D (1973) Dermo-epidermal interactionsbetween birds and mammals: differentiationof cutaneous appendages. J Embryol ExpMorphol30:587603

Dhouailly D (1984) Specification of feather andscale patterns. In: Malincinski G, Bryant S(eds).Pattern formation. New York: Macmil-lan Pub. Co, 581601

Dunphy JE (1963) The fibroblasta ubiquitousfriend of the surgeon. N Engl J Med 268:136777

Eisen MB, Spellman PT, Brown PO, Botstein D(1998) Cluster analysis and display ofgenome-wide expression patterns. Proc NatlAcad Sci USA 95:148638

Estrach S, Ambler CA, Lo Celso C, Hozumi K,Watt FM (2006) Jagged 1 is a beta-catenintarget gene required for ectopic hair follicleformation in adult epidermis. Development133:442738

Gabbiani G, Rungger-Brindle E (1981) Thefibroblast.In: Glynn LE (ed). Handbook ofinflammation tissue repair and regeneration.Amsterdam: Elsevier/North Holland Bio-medical Press, 150

Godwin AR, Capecchi MR (1998) Hoxc13 mutantmice lack external hair.Genes Dev12:1120

Guenther MG, Jenner RG, Chevalier B, NakamuraT, Croce CM, Canaani Eet al.(2005) Globaland Hox-specific roles for the MLL1 methyl-

transferase. Proc Natl Acad Sci USA 102:86038

Hutchin ME, Kariapper MS, Grachtchouk M,Wang A, Wei L, Cummings D et al. (2005)Sustained Hedgehog signaling is required forbasal cell carcinoma proliferation and survi-val: conditional skin tumorigenesis recapitu-lates the hair growth cycle. Genes Dev19:21423

Jahoda CA, Horne KA, Oliver RF (1984)Induction of hair growth by implantation ofcultured dermal papilla cells. Nature 311:5602

Jahoda CA (1992) Induction of follicle formationand hair growth by vibrissa dermal papillaeimplanted into rat ear wounds: vibrissa-typefibres are specified. Development 115:11039

Levinson R, Mendelsohn C (2003) Stromal pro-genitors are important for patterning epithe-lial and mesenchymal cell types in theembryonic kidney Semin. Cell Dev Biol14:22531

Millar SE (2002) Molecular mechanisms regula-ting hair follicle development. J InvestDermatol118:21625

Normand J, Karasek MA (1995) A method for theisolation and serial propagation of keratino-cytes, endothelial cells, and fibroblasts froma single punch biopsy of human skin. In vitroCell Dev Biol Anim31:44755

www.jidonline.org 7



7/7

Rendl M, Lewis L, Fuchs E (2005) Mole-

cular dissection of mesenchymalepithelial

interactions in the hair follicle. PLoS Biol

3:e331

Ringrose L, Paro R (2004) Epigenetic regulation of

cellular memory by the Polycomb and

Trithorax group proteins. Annu Rev Genet

38:41343

Rinn JL, Bondre C, Gladstone HB, Brown PO,

Chang HY (2006) Anatomic demarcation by

positional variation in fibroblast gene ex-

pression programs. PLoS Genet2:e119

Sarin KY, Cheung P, Gilison D, Lee E, Tennen RI,

Wang Eet al.(2005) Conditional telomerase

induction causes proliferation of hair follicle

stem cells. Nature436:104852

Schedl A, Hastie ND (2000) Cross-talk in kidney

development. Curr Opin Genet Dev 10:

5439

Schmitt S, Paro R (2006) RNA at the steeringwheel. Genome Biol7:218

Schwann T (1847) Microscopical researches intothe accordance in the structure and growthof animals and plants. London: The Syden-ham Society, 1268

Sessa L, Breiling A, Lavorgna G, Silvestri L,

Casari G, Orlando V (2007) NoncodingRNA synthesis and loss of Polycomb grouprepression accompanies the colinear activa-tion of the human HOXA cluster. RNA 13:22339

Shannon JM, Hyatt BA (2004) Epithelialmesenchy-mal interactions in the developing lung.Annu Rev Physiol66:62545

Sneddon JB, Zhen HH, Montgomery K, van deRijn M, Tward AD, West R, et al. (2006)Bone morphogenetic protein antagonistgremlin 1 is widely expressed by cancer-associated stromal cells and can promote

tumor cell proliferation. Proc Natl Acad SciUSA103:148427

Squazzo SL, OGeen H, Komashko VM, Krig SR,Jin VX, Jang SWet al. (2006) Suz12 binds tosilenced regions of the genome in a cell-type-specific manner.Genome Res16:890900

Stenn KS, Paus R (2001) Controls of hair follicle

cycling.Physiol Rev81:44994

Svingen T, Tonissen KF (2006) Hox transcriptionfactors and their elusive mammalian genetargets.Heredity 97:8896

Wolpert L (1969) Positional information andthe spatial pattern of cellular differentiation.

J Theor Biol25:147

Yamaguchi Y, Itami S, Tarutani M, Hosokawa K,Miura H, Yoshikawa K (1999) Regulation ofkeratin 9 in nonpalmoplantar keratino-cytes by palmoplantar fibroblasts throughepithelialmesenchymal interactions.J InvestDermatol112:4838



Documents

A Systems Biology Approach to Anatomic Diversity