Activator Protein1 Activity Regulates Epithelial Tumor Cell Identity

2006;66:7578-7588. Cancer Res Michael J. Gerdes, Maxim Myakishev, Nicholas A. Frost, et al. IdentityActivator Protein-1 Activity Regulates Epithelial Tumor Cell

Updated version

http://cancerres.aacrjournals.org/content/66/15/7578

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2006/08/02/66.15.7578.DC1.html

Access the most recent supplemental material at:

Cited Articles

http://cancerres.aacrjournals.org/content/66/15/7578.full.html#ref-list-1

This article cites by 54 articles, 25 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/66/15/7578.full.html#related-urls

This article has been cited by 10 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

[email protected] at

To request permission to re-use all or part of this article, contact the AACR Publications

Research. on December 7, 2013. © 2006 American Association for Cancercancerres.aacrjournals.org Downloaded from










http://cancerres.aacrjournals.org/cgi/alerts

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]




http://cancerres.aacrjournals.org/




Activator Protein-1 Activity Regulates Epithelial Tumor Cell Identity

Michael J. Gerdes,1Maxim Myakishev,

2Nicholas A. Frost,

1Vikas Rishi,

2Jaideep Moitra,

2

Asha Acharya,2Michelle R. Levy,

1Sang-won Park,

2Adam Glick,

1

Stuart H. Yuspa,1and Charles Vinson

2

1Laboratory of Cellular Carcinogenesis and Tumor Promotion and 2Laboratory of Metabolism, National Cancer Institute, Center for CancerResearch, Bethesda, Maryland

Abstract

To examine the consequences of inhibiting activator protein-1(AP-1) transcription factors in skin, transgenic mice weregenerated, which use the tetracycline system to conditionallyexpress A-FOS, a dominant negative that inhibits AP-1 DNAbinding. Older mice develop mild alopecia and hyperplasia ofsebaceous glands, particularly around the eyes. When A-FOSwas expressed during chemical-induced skin carcinogenesis,mice do not develop characteristic benign and malignantsquamous lesions but instead develop benign sebaceousadenomas containing a signature mutation in the H-rasproto-oncogene. Inhibiting AP-1 activity after tumor forma-tion caused squamous tumors to transdifferentiate intosebaceous tumors. Furthermore, reactivating AP-1 in seba-ceous tumors results in a reciprocal transdifferentiation intosquamous tumors. In both cases of transdifferentiation,individual cells express molecular markers for both cell types,indicating individual tumor cells have the capacity to expressmultiple lineages. Molecular characterization of culturedkeratinocytes and tumor material indicates that AP-1 regu-lates the balance between the wnt/B-catenin and hedgehogsignaling pathways that determine squamous and sebaceouslineages, respectively. Chromatin immunoprecipitation anal-ysis indicates that c-Jun binds several wnt promoters, whichare misregulated by A-FOS expression, suggesting thatmembers of the wnt pathway can be a primary targets ofAP-1 transcriptional regulation. Thus, AP-1 activity regulatestumor cell lineage and is essential to maintain the squamoustumor cell identity. (Cancer Res 2006; 66(15): 7578-88)

Introduction

Activator protein-1 (AP-1) is a heterodimeric transcription factorcomplex composed of a Jun family member and a FOS familymember that binds the TRE DNA sequence (5¶-TGAGTCA-3¶) and isinvolved in a variety of cellular processes, including growth,apoptosis, and differentiation (1). The contribution of individualfamily members to mouse development has been examined formost genes using the ‘‘knockout’’ technology, with phenotypesvarying from embryonic lethality for c-Jun, JunB, and Fra-1 tonormal development and a nurturing defect for FosB (reviewed in

ref. 2). Tissue-specific ablation in the skin of AP-1 family memberswith embryonic phenotypes indicates that c-Jun is needed forproper eyelid closure during development (3, 4). Transgenicexpression of different AP-1 family members has identified adultphenotypes but not in the skin (2). Skin epidermis expression of anAP-1 dominant negative consisting of c-Jun without a trans-activation domain did not produce a skin phenotype (5).

The mouse epidermal multistage carcinogenesis model provides awell-defined system for examining the evolution of squamousepithelial cells into benign squamous papillomas and their subse-quent progression into squamous cell carcinomas (6). Themajority ofthese tumors contain mutations in the H-ras oncogene. A functionfor AP-1 activity in skin tumorigenesis has been identified usingknockout and transgenic mice that modulate AP-1 components. Forexample, papillomas in c-Fos knockout mice did not convert intomalignant carcinomas (7). Deletion of c-Jun in skin produces micethat develop smaller papillomas (3). Expression of a c-Jun dominantnegative in skin dramatically reduced papilloma formation inseveral protocols of cutaneous chemical carcinogenesis (5, 8) andreduced squamous cell cancer induction by UVB exposure (9).

Furthermore, upstream activators of the AP-1 complex alsomodulate tumor formation (10). AP-1 activity in the skin isregulated through a cascade of signaling events involving proteinkinase C (PKC), Ras, and the p38 family of mitogen-activatedprotein kinases (11). Additionally, activation of AP-1 by c-Jun NH2-terminal kinase (12), PKC family members (13, 14), and tumornecrosis factor-a (15) and constitutive activation of Ras promotetumorigenesis (16–18).

Although previous studies have shown that specific AP-1 factorsmodify tumor incidence or progression (1, 7), studies to date havebeen unable to examine the effects of AP-1 modulation after tumordevelopment. We have used the two-transgene tetracycline system(19) to regulate expression of a high-affinity AP-1 dominant-negative protein (A-FOS). A-FOS is a chimeric protein containingthe c-FOS leucine zipper and a designed acidic protein sequencethat replaces the DNA-binding region. A-FOS heterodimerizes withc-FOS dimerization partners, which are primarily the JUN family ofproteins (c-Jun, JunB, and JunD; ref. 20) to prevent AP-1 DNAbinding (21). By establishing a mouse model whereby AP-1 activityin the skin is ‘‘on’’ or ‘‘off,’’ we can document the influence of AP-1transcriptional activity on specific stages of tumor development.Here, we report the powerful influence of AP-1 on determiningtumor cell lineage and premalignant progression in the mouse skinmodel of multistage carcinogenesis.

Materials and Methods

Transgenic mice. To make the 7X-tetOp A-FOS transgene, the plasmidpUHD 10-3 (19) was linearized with SacII and BamHI and ligated to a 62-bp

polylinker: top strand, 5¶-GGCCACCATGGCGTATCCCTACGACGTG-CCCGATTATGCCCGATTATGCCCATATGCAGGAATTCAAGCTTG-3¶ that

Note: Supplementary data for this article are available at Cancer Research Online(http://cancerres.aacrjournals.org/).

Requests for reprints: Stuart H. Yuspa, Laboratory of Cellular Carcinogenesis andTumor Promotion, National Cancer Institute, Center for Cancer Research, Bldg. 37 Rm.4068A, Bethesda, MD 20892. Phone: 301-496-2162; Fax: 301-496-8709; E-mail: [email protected] or Charles Vinson, Laboratory of Metabolism, National Cancer Institute,Center for Cancer Research, Bldg. 37, Rm 3128, Bethesda, MD 20892. Phone: 301-496-8753, Fax: 301-496-8419; E-mail: [email protected].

I2006 American Association for Cancer Research.doi:10.1158/0008-5472.CAN-06-1247

Cancer Res 2006; 66: (15). August 1, 2006 7578 www.aacrjournals.org

Research Article




encodes a Kozak consensus and a Hemagglutinin tag (YPYDVPDYA). AnSV40 poly(A) fragment containing the small T-antigen intron that preceded

the poly(A) site was obtained from the plasmid pRSVNeo as a 1,026-bp

BamHI-SmaI fragment. This fragment, when ligated to the BamHI-NaeI–

digested 2.7-kb vector (pTRE 850/85-ds) backbone, produced a plasmid withthe SV40 poly(A) in reverse orientation (pTRE850/851 ds reversed). To

obtain a plasmid with the SV40 poly(A) in the correct orientation,

‘‘pTRE850/851-ds-reversed’’ was digested with HindIII. The 2,704-bp vector

and the 1,067-bp insert were then religated, and recombinants were selectedin which the insert fragment [representing the SV40 poly(A)] is reversed

when compared with the parental plasmid. This plasmid was digested with

HindIII (partial) and NdeI and ligated to the A-FOS cDNA digested with the

same enzymes from the CMV566 plasmid (22) to produce the construct ‘‘tetA-FOS.’’ Tet-A-FOS was linearized with AatII and AflIII, and the 2.4-kb insert

(bearing the 7X tetOp, hemagglutinin-tagged A-FOS, and the SV40poly A

(with intron) was injected into mouse pronuclei by standard procedures.The founder pups on the FVB/N background were screened initially by

Southern hybridization of tail DNA preparations using the complete AatII-

AflII transgene fragment as probe and subsequently by PCR.

Mice from the keratin 5-tet-transactivator (K5-tTA) FVB/N line (23) were

crossed with homozygous mice from the tetA-FOS line to produce double

transgenic mice (K5/A-FOS) with tetracycline-regulated expression of A-FOS

restricted to K5-expressing cells. For convenience, the tetA-FOS pups from

this cross are referred to as wild type (WT). During mating and gestation,

mothers were kept on doxycycline feed (20 mg/kg of feed; Bio-serve, Baltimore,

MD) to repress expression of A-FOS and switched to a normal diet at weaning,

which induced A-FOS expression. The TRE-Luciferase (TRE-LUC) mouse was

originally described by Rincon et al. (24). This strain has 12-O-tetradecanoyl-

phorbol-13-acetate/phorbol 12-myristate 13-acetate (PMA)–inducible lucifer-

ase expression dependent on the TRE (AP-1) site (24). K5/A-FOS mice were

crossed with TRE-LUC mice to generate triple transgenic mice. Mice were

sacrificed according to protocols approved by the National Cancer Institute

Animal Care and Use Committee. Genotypes were determined by PCR

analysis of DNA from tail biopsies. The following forward and reverse

primer pairs were used: tetA-FOS, 5¶-CCACGCTGTTTTGACCTCCATAG-3¶ and5 ¶-ATTCCACCACTGCTCCCATTC-3 ¶; K5- tTA, 5 ¶-AACAACCCG-

TAAACTCGCCC-3¶ and 5¶-GCAACCTAAAGTAAAATGCCCCAC-3¶; TRE-LUC,5¶-GCGGAATACTTCGAAATGTC-3¶, 5¶-TTAGGTAACCCAGTAGATCCCC-3¶.

Electrophoretic mobility shift assay. Six B-ZIP dimers, including the

FOS|JUND heterodimer, CAAT/enhancer binding protein (C/EBP), cyclic

AMP–responsive element binding protein (CREB), protease-activated

receptor (PAR), activating transcription factor 6 (ATF6), and interleukin-3-regulated nuclear factor (NFIL3) homodimers, were mixed with 0, 1, 10, or

100 molar equivalents of A-FOS in a gel shift reaction buffer [25 mmol/L

Tris-HCl (pH 8), 50 mmol/L KCl, 0.5 mmol/L EDTA, 2.5 mmol/L DTT, 1 Ag/ALbovine serum albumin (BSA), 10% glycerol] to a final volume of 20 AL and

incubated at 65jC for 20 minutes. The protein mix was cooled to room

temperature, and 7 pg of double-stranded 32P-radiolabeled 28-mer

oligonucleotide was added and incubated at 37jC for 20 minutes. Thesequences of the 28-mer DNA probes are GTCAGTCAGAATGACTCA-TATCGGTCAG (AP-1) for FOS/JUND heterodimer, GTCAGTCAGATTGCG-CAATATCGGTCAG for C/EBP, GTCAGTCAGATTACGTAATATCGGTCAGfor PAR, and GTCAGTCAGATGACGTCATATCGGTCAG for CREB, ATF6,and NFIL3. Samples were separated in 7.5% acrylamide gel; the gel was

dried and exposed for autoradiograghy.

To confirm the inhibition of DNA binding by A-FOS expression intransgenic mice, K5/A-FOS or WT primary keratinocytes were cultured in

0.05 mmol/L Ca2+ medium and PMA treated for 6 hours in the presence or

absence of doxycycline. Nuclear extracts were prepared using the NE-PER

reagent (Pierce, Rockford, IL); 10,000 cpm of labeled oligonucleotide wasadded to 2 Ag of protein sample and then incubated in binding buffer

(10 mmol/L HEPES, 80 mmol/L KCl, 0.05 mmol/L EDTA, 6% glycerol,

1 mmol/L DTT, and 1 mmol/L MgCl2) at 37jC for 20 minutes. Samples were

separated on a 5% PAGE gel in 0.25� Tris-borate EDTA at 150 V for 2 hours.Oligonucleotides used contained DNA binding sites for AP-1 and CRE (25).

To verify specificity of DNA binding, 50-fold unlabeled probe was added in a

competition assay (26). To test the direct effect of recombinant A-FOS

protein on the electrophoretic mobility shift assay (EMSA) results, WT FVB/N primary keratinocytes were cultured in 0.05 mmol/L Ca2+ media. Nuclear

extracts were prepared identically except that the nuclei from one 150-mm

dish (1.2 � 107 cells) were lysed in the presence of 2.5 ng pure recombinant

A-FOS protein and incubated for 2 hours at 38jC before EMSA. Thesenuclear extracts were incubated with oligonucleotides containing binding

sites for AP-1, CRE, C/EBP, and SP1 as described above.

Determination of luciferase activity. TRE-LUC and K5/A-FOS/TRE

mice were placed on normal feed for 2 weeks before treatment with 5 AgPMA in 25 AL acetone on one ear, whereas the other ear was treated with

acetone alone. Extracts were prepared directly in lysis buffer (Promega,

Madison, WI) after 6 hours. Luminescence was quantified from 25 AL of

extract in 200 AL luciferase reagent in triplicate from three mice for each

group with error bars representing SE. Primary keratinocytes were cultured

in 0.05 mmol/L calcium-containing media in the presence of doxycycline

(15 ng/mL). Medium was changed to doxycycline-free medium 24 hours

before the addition of PMA (100 ng/mL). After 4 hours, lysates were

prepared (Promega), and luciferase activity was determined. For the TOP-

FLASH (Upstate Biotechnology, Charlottesville, VA) h-catenin reporter

assays, cells were plated in 24-well plates and transfected with either TOP�

or the mutant FOP� reporter at 0.8 Ag plasmid per well with Lipofect-

AMINE 2000 transfection reagent. After 6 hours, medium was changed to

0.05 mmol/L calcium growth medium overnight. LiCl (10 mmol/L) was

added 6 hours before harvest, and assays were done as above.

Tumor initiation/promotion studies. Mice at 8 to 10 weeks of age were

initiated with 7,12-dimethylbenz(a)anthracene (DMBA; 100 Ag/200 ALacetone) on shaved dorsal skin. PMA (5 Ag/200 AL acetone 1�/wk)promotion was started 1 week later and continued for 20 weeks, and tumor

number was recorded weekly. Protocols used for dietary doxycycline are

described in the Results and figure legends. Study groups included equal

numbers of male and female mice for all groups. In Fig. 2C , animal numberswere WT [+Dox (n = 17), �Dox (n = 9)] and K5/A-Fos [+Dox (n = 14), �Dox(n = 10)]. In Fig. 2D , animal numbers were WT [+Dox (n = 16), �Dox(n = 15)] and K5/A-FOS [+Dox (n = 11), �Dox (n = 10)]. Some animals weresacrificed at time points after 27 weeks to examine tumor histology.

Ras mutation analysis. Genomic DNA was isolated from normal skin,

three squamous papillomas, and a squamous cell carcinoma excised frommice

treated with DMBA/PMA and expressing AP-1 activity and 11 sebaceousadenomas produced from the same DMBA/PMA treatment during continual

A-FOS expression. DNA was amplified by nested PCR for a region spanning

codon 61 of the H-ras gene, before digestion with XbaI, which specifically

recognizes the activatingA-Tmutation, and subjected togel electrophoresis (27).Immunohistochemistry. Tissue sections were deparaffinized, rehy-

drated through a graded series of ethanol followed by PBS, and blocked with

10% normal goat serum/3% BSA/PBS for 1 hour. Primary antibodies wereapplied: rabbit anti-keratin 5 or 6-FITC conjugated (Covance, Princeton, NJ),

guinea pig anti-ADRP (Research Diagnostics, Flanders, NJ), rabbit anti-

h-catenin (Sigma, St. Louis, MO), goat anti-Indian Hedgehog (IHH; Santa

Cruz Biotechnology, Santa Cruz, CA), and goat anti-FRP3 (R&D Systems,Minneapolis, MN) in a solution of 3% BSA/PBS. Slides were washed in PBS

before application of respective FITC- or Cy3-conjugated donkey secondary

antibodies (Jackson Immunologicals, West Grove, PA). Slides were stained

with 4¶,6-diamidino-2-phenylindole for visualization of nuclei beforefluorescence microscopy.

Immunoblot analysis. Total protein lysates were prepared by extracting

primary keratinocyte cultures with M-PER reagent (Pierce). Protein extracts

were loaded into NuPage gels (Invitrogen, Carlsbad, CA) and run accordingto manufacturer’s protocols. Blotted extracts were blocked in TBS-Tween +

3% BSA. Primary antibodies diluted in TBS-Tween + 3% BSA were incubated

at 4jC overnight. Secondary antibodies conjugated to horseradishperoxidase were applied for 30 minutes at room temperature diluted in

TBS-Tween + 3% BSA. Immunoblots were developed using enhanced

chemiluminescence reagents from Pierce. Bound antibodies were detected

using the Pierce Pico substrate and exposing to BioMax film (Kodak,Rochester, NY). Membranes were reprobed for actin protein to confirm

equivalent loading between the samples. Immunoblot detection of A-FOS

protein in cultured K5/A-FOS primary keratinocytes grown in the presence

AP-1 and Epithelial Tumor Cell Identity

www.aacrjournals.org 7579 Cancer Res 2006; 66: (15). August 1, 2006




or absence of doxycycline for the times indicated. Cellular lysates werecollected and analyzed on a 12.5% SDS-PAGE gel. A-FOS protein was detected

with a c-FOS antibody (Santa Cruz Biotechnology) that recognizes the

conserved region between A-FOS and c-FOS. Keratinocyte lysates from single

transgenic tetA-FOS mice or 0.2 pg recombinant A-FOS protein (rA-FOS) wereincluded as negative and positive controls respectively (data not shown).

Reverse transcription-PCR analysis. RNA from primary keratinocytes

was isolated with Trizol (Invitrogen) and DNase treated (Ambion, Austin,

TX). Initial-strand cDNA was first prepared (Invitrogen) and used forreverse transcription-PCR with equal amounts of synthesized cDNA

determined by equal amplification of h-actin. For PCR amplification, the

forward and reverse primer pairs are listed in the Supplementary Data.

Oil Red-O analysis. WT and K5/A-FOS skin keratinocytes were culturedin 0.05 mmol/L calcium growth media for 48 hours, and then differentiation

was induced in 0.12 mmol/L calcium medium for 48 hours before fixation in

4% paraformaldehyde and staining with Oil Red-O. To quantify the number

of Oil Red-O cells, 25 fields were counted for each genotype using a �10objective, and mean F SE was determined for each genotype.

Chromatin immunoprecipitation: ChIP on chip. Binding of phospho-

c-JUN, CREB, and C/EBPh to the promoter of 13 genes shown on Fig. 6A

was examined using ChIP-on-chip assay with mouse promoter microarraysfrom NimbleGen (Madison, WI). Primary mouse keratinocytes were treated

with 1% formaldehyde for 8 minutes, and the reaction was stopped by

glycine. The chromatin immunoprecipitation (ChIP) was done with Upstate

ChIP reagents and protease inhibitors (Roche, Indianapolis, IN). The cellswere washed, collected in cell lysis buffer, sheared by ultrasound producing

DNA fragments of 700 bp average size. The lysate was centrifuged, diluted

20 times, precleared with BSA- and yeast tRNA-saturated protein A/Gagarose mix (Invitrogen), and immunoprecipitated with antibodies against

phospho-c-JUN (Santa Cruz Biotechnology), CREB (Upstate Biotechnology),

and custom C/EBPh antibody. The ChIP with nonspecific IgG was used as

negative control. The ChIP eluate was incubated with NaCl at 65jCovernight, at 95jC for 1 hour, and digested with Proteinase K and extracted

with phenol. The ChIP DNA was purified using Qiagen (Valencia, CA)

MinElute kit; 5 ng of ChIP DNA were randomly amplified using ‘‘RoundA/B’’

protocol (28) with Cy5-labeled primers in 400 AL total PCR volume. ThePCR reaction consisted of 42 cycles and produced 20 Ag of product. The

total genomic DNA (‘‘input’’) from an aliquot of the sonicated cell lysate was

extracted, amplified with Cy3 primers in parallel, and used as a reference inhybridization. The PCR product was cleaned by ethanol precipitation

and hybridized in presence of yeast tRNA, COT1 DNA, and poly-dATP

(10 Ag each) overnight in MAUI hybridization station at 42jC to mouse

promoter microarrays (Nimblegen). Each array contained promoters of21,815 unique genes. Each promoter was represented by 15 unique 50-mer

oligonucleotides located from 1,300 bp upstream to 200 bp downstream

of transcription start. The arrays were washed with 2� SSC, 0.1 % SDS for

2 minutes, 1� SSC for 1 minute, and with 0.2 � SSC for 15 seconds; scannedwith Axon 4000B scanner; and analyzed with NimbleScan software.

Results

A-FOS inhibits AP-1 binding to TRE DNA and transcrip-tional activity. To express A-FOS, an inhibitor of AP-1 binding tothe TRE sequence in mouse epidermis, we crossed K5-tTA mice

Figure 1. A, A-FOS protein was detected by immunostaining of tail skin ofK5/A-FOS mice with an antibody that recognizes the hemagglutinin antigen atthe NH2 terminus of the A-FOS protein. A-FOS protein expression is in basalepidermis and overlaps the expression seen with a keratin 5 antibody. Nostaining was seen in non–A-FOS-expressing skin. B, to examine doxycycline(Dox )–dependent A-FOS expression, RNA was isolated from skin of K5/A-FOSmice fed doxycycline or normal food from birth and analyzed by reversetranscription-PCR. A-FOS mRNA is strongly induced in the absence ofdoxycycline. C, immunoblot detection of A-FOS protein in cultured K5/A-FOSprimary keratinocytes grown in the presence or absence of doxycycline. A-FOSprotein was detected with a c-FOS antibody (Santa Cruz Biotechnology) thatrecognizes the c-FOS leucine zipper, a conserved region between A-FOS andc-FOS. D and E, AP-1-dependent luciferase activity in skin from single(TRE-LUC) or triple transgenic (K5/A-FOS/TRE) mice or primary skinkeratinocytes from each genotype was determined before and after PMAexposure. Units are arbitrary. F, A-FOS expression inhibits TRE DNA binding.K5/A-FOS or WT primary keratinocytes were cultured in 0.05 mmol/L Ca2+

medium and PMA treated for 6 hours in the presence or absence of doxycycline.G, specificity of A-FOS inhibition of DNA binding of AP-1 factors was shown inEMSAs by incubating lysates of WT keratinocytes with recombinant A-FOSprotein and labeled oligonucleotides for AP-1, CRE, and SP1. Samples wereseparated with 5% PAGE. H, inhibition of DNA binding of different B-ZIPs byA-FOS. Six B-ZIP dimers, including the FOS|JUND heterodimer, C/EBP, CREB,PAR, ATF6, and NFIL3 homodimers, were mixed with 0, 1, 10, or 100 molarequivalents of A-FOS. The protein mix was then incubated with double-stranded32P-radiolabeled 28-mer oligonucleotide for each protein dimer pair. Sampleswere separated with a 7.5% PAGE gel.

Cancer Research





that use the bovine keratin 5 promoter to express the tetracyclinetransactivator protein (tTA) in basal keratinocytes and outer rootsheath cells of the skin with tetA-FOS mice that express A-FOSunder the regulated control of seven tetracycline responsiveelements to produce double transgenic mice (K5-tTA/tetA-FOSreferred to as K5/A-FOS). Immunohistochemical analysis indicatesthat the K5/A-FOS mice express the A-FOS protein in the epidermisand hair follicles (Fig. 1A). A-FOS mRNA expression is inhibited inthe skin in the presence of doxycycline, a tetracycline analogue,because the tetracycline transactivator when bound to doxycyclineis inhibited from binding DNA (Fig. 1B). A similar doxycyclinedependence of A-FOS expression is observed in primary cultures ofnewborn keratinocytes from double transgenic mice (Fig. 1C).

To determine if A-FOS expression in skin epidermis inhibitsTRE-DNA transcriptional activity, triple transgenic mice (K5/A-FOS/TRE) containing K5-tTA, tetA-FOS, and the TRE reportertransgene TRE-Luciferase (TRE-LUC; ref. 24) were treated with thephorbol ester PMA to stimulate TRE activity, which increased f7-fold in WT skin. However, in the presence of A-FOS expression,basal TRE activity was unaffected, whereas PMA-induced TREactivity was nearly eliminated (Fig. 1D). In primary keratinocytecultures produced from triple transgenic newborn pups, A-FOSexpression suppressed both basal and induced activity. Thedifference in A-FOS inhibition of basal TRE activity in vivo and

in vitro most likely reflects the regulation of the K5 promoter that isdown-regulated in differentiating epidermis but not culturedkeratinocytes. Thus, A-FOS inhibits TRE-dependent transcriptionalactivity in both skin and cultured keratinocytes, presumably byinhibiting AP-1 binding to the TRE sequence. Previously, we haveshown that the AP-1 complex in murine epidermis encompassesc-Jun, Jun B, Jun D, c-Fos, Fra 1, and Fra 2, whereas c-Fos is notdetected in cultured keratinocytes (26).

A series of EMSA were used to evaluate the specificity of A-FOSat inhibiting the binding of nuclear extracts to different DNAsequences. Nuclear extracts isolated from primary newbornkeratinocytes expressing A-FOS for 24 hours after the removal ofdoxycycline do not bind either in the basal or PMA-stimulated caseto a TRE-containing oligonucleotide (5¶-TGAGTCA-3¶). However,these nuclear extracts do bind to the closely related CRE-containing oligonucleotides (Fig. 1F). To evaluate if inhibition ofTRE binding was a direct consequence of the A-FOS protein,recombinant A-FOS protein was added to nuclear extractsprepared from WT keratinocytes. The EMSA indicates thatexogenous A-FOS inhibited DNA binding to TRE-containing DNAoligonucleotide but not three additional oligonucleotides contain-ing consensus-binding sites for CREB, C/EBP, or SP1 (Fig. 1G).Furthermore, recombinant A-FOS protein specifically inhibits DNAbinding of AP-1 dimers (c-FOS|JunD) but not five other B-ZIP

Figure 2. Skin phenotype in A-FOS-expressing mice.A, alopecia and ulcerative dermatitis in 9-month-oldK5/A-FOS mouse expressing A-FOS. B, histologyof H&E-stained sections showing sebaceousgland hyperplasia (sq, squamous epithelium;sb, hyperplastic sebaceous glands). C, ADRP (red)specifically localizes to developing sebaceous gland inneonatal skin, whereas K14 (green ) localized to thebasal epidermis. Blue, 4¶,6-diamidino-2-phenylindolestaining of the nuclei. D, tumor induction in K5/A-FOSor tetA-FOS mice fed either a doxycycline or normaldiet. Mice were treated once with DMBA followed byweekly treatments with PMA for 20 weeks. At week 27,all mice were fed a normal diet through the remainderof the study. E, K5/A-FOS and tetA-FOS mice weretreated as in (C ), but at week 27, food for all groupswas switched. F, papillomas developed in controlgroups and K5/A-FOS mice on doxycycline diet(tetA-FOS shown at week 25). G, numerous smallyellowish tumors develop on K5/A-FOS mice fednormal diet when A-FOS is expressed (tumors atweek 25). H, mutations in the H-ras gene in genomicDNA isolated from normal skin, squamous papillomas,and a squamous carcinoma (SCC ) excised frommice treated with DMBA/PMA and expressing AP-1activity, and 11 sebaceous adenomas produced fromthe same DMBA/PMA treatment during continualA-FOS expression. Alternate lanes contain undigestedor Xba I-digested PCR products. Arrows, migration ofthe restriction susceptible allele containing the A-Tmutation in codon 61. The less robust amplification ofmutant bands in sebaceous tumors is due to extensiveoverlying and adjacent normal tissue excised alongwith the small sebaceous tumors.






dimers in EMSA assays at equimolar equivalents (Fig. 1H).Interference with C/EBP binding is detected only at 10- to 100-fold molar excess. Thus, A-FOS specifically and directly inhibitsDNA binding and transcriptional activity of the TRE.

A-FOS expression in skin causes hyperplasia of sebaceousglands. Expression of A-FOS in mouse epidermis did not producean obvious phenotype until late in life when mice often developedmild hair loss (alopecia), sebaceous gland hyperplasia (particularlybut not exclusively on the eyelids), and focal skin erosions (Fig. 2Aand B). In the focal erosions, the hyperplastic sebaceous glandsbecame independent of the adjacent hairs follicles that weredegenerating.

cDNA microarrays were used to further characterize theconsequences of expressing A-FOS in the epidermis. One of theup-regulated genes is adipocyte differentiation–related protein(ADRP), which functions in lipid sequestration (29). Immunohis-tochemical localization of ADRP indicates that it is a novel markerfor sebocytes in developing sebaceous glands from newborn skin(Fig. 3C, inset, red) and mature sebaceous glands (data not shown).These data further support the observation that A-FOS expressionshifts the balance toward the sebaceous lineage.

Inhibiting AP-1 during skin carcinogenesis produces seba-ceous adenomas. To examine the consequence of inhibiting AP-1activity in a skin carcinogenesis protocol, K5/A-FOS mice andsingle transgenic littermates (tetA-FOS referred to as WT) that donot express A-FOS were initiated at 8 to 10 weeks with a singletopical application of DMBA followed by weekly applications ofthe phorbol ester PMA for 20 weeks to induce skin tumors. Mice ofboth genotypes (K5/A-FOS and WT) were fed two different diets,a doxycycline-containing diet or normal diet 2 weeks beforeinitiation with DMBA. In this double transgenic system, a normaldiet results in A-FOS expression and the suppression of the TREactivity in K5/A-FOS mice, whereas mice on a doxycycline dietretain AP-1 activity regardless of genotype.

The three groups of mice with normal AP-1 activity (WTmice ondoxycycline, WT mice off doxycycline, and K5/A-FOS mice ondoxycycline) developed two to three tumors per mouse during thecourse of two independent carcinogenesis protocols (Fig. 2D andE). These tumors were large (100-300 mm3) glistening lesionstypical of squamous papillomas (Fig. 2F), and this diagnosis wasconfirmed with histologic examination of the tumors (Fig. 3A). Asexpected, these tumors had the A-to-T mutation in codon 61 of the

Figure 3. Modulating AP-1 activity during tumorigenesis alters tumor cell lineage. H&E-stained section of a (A ) squamous papilloma and (B ) sebaceous adenoma fromK5/A-FOS mice fed a normal diet respectively. C, immunofluorescent staining for ADRP (red) and K6 (green ) of sebaceous adenoma. D and E, H&E stain of a mixedsquamous-sebaceous lesion from K5/A-FOS mice where diets are switched. Double-immunofluorescent staining of ADRP (F ; red ) or IHH (H ; red) and K6 (green )with strong colocalization in individual cells (arrows ). G, papilloma where A-FOS expression was initiated at 27 weeks with total transdifferentiation into sebaceousadenoma. J, IHH is not detected on immunostaining of squamous papillomas compared with (K) K6 that is abundant in all squamous tumors. I, Oil Red-O staining of asebaceous adenoma.

Cancer Research





H-ras gene characteristic of DMBA-initiated papillomas (ref. 30;Fig. 2H). In contrast, the K5/A-FOS mice continually expressingA-FOS developed approximately five times more tumors per mouse(Fig. 2D and E) that were small (2-8 mm3) and pale yellow (Fig. 2G).Histologic examination of these tumors identified 91% well-differentiated sebaceous adenomas (Fig. 3B ; Table 1), an uncom-mon tumor type that arises from the sebaceous gland of the hairfollicle. These tumors also contained the characteristic H-rasmutation found in control papillomas (Fig. 2H). Tumors from allgroups were dependent on DMBA mutagenesis, as PMA treatmentalone for 20 weeks did not produce tumors (Table 1). One third ofthe squamous papillomas from control (tetA-FOS) mice convertedto carcinomas, whereas none of the tumors on the K5/A-FOS miceexpressing A-FOS underwent malignant conversion (summarizedin Table 1).

The expression of sebocyte specific molecular markers was usedto further characterize the differentiation status of the sebaceousadenomas observed histologically. ADRP is overexpressed inA-FOS-expressing skin, is expressed in normal sebaceous glands,and is also expressed in sebaceous adenomas (Fig. 3C, red).Additionally, the established sebaceous gland molecular markerIHH (31) also localized to sebaceous tumor cells (Fig. 3H, red). Thesedata indicate that the histologically observed sebaceous adenomasare expressing molecular markers of sebaceous cells. However,neither ADRP nor IHH was detected in squamous tumors (Fig. 3J).In contrast, the hyperproliferative keratinocyte cell marker, keratin6 (K6), was abundantly expressed in papillomas (Fig. 3K ; ref. 32) andnot expressed in the sebocytes or sebaceous adenomas (Fig. 3C,green). Expression of these molecular markers coupled with strongstaining of the adenomas with Oil Red-O (Fig. 3I) indicates thetumors are functionally well-differentiated sebaceous adenomas.

Inhibiting AP-1 activity in squamous tumors inducessebaceous adenomas. We used the inducible capacity of thetetracycline system to induce A-FOS expression in K5/A-FOS miceafter squamous papillomas formed. The doxycycline diet was

switched to a normal diet 27 weeks after the initiation of thecarcinogenesis protocol when squamous papillomas had formedand tumor growth continued for an additional 22 weeks.Expression of A-FOS results in an 80% reduction in malignantconversion, indicating that inhibition of AP-1 activity afterpapilloma formation is sufficient to block or delay carcinomaformation (Table 1). Provocatively, approximately half of thetumors showed a profound alteration of the standard squamouspapilloma histology not observed in control mice that do notexpress A-FOS: 26% of the tumors became mixed sebaceous/squamous lesions, 14% converted to pure sebaceous adenomas,and 53% remained squamous papillomas (see Table 1 for summary;Fig. 3D and E). These results show that continuous AP-1 activity isnecessary to maintain the squamous phenotype, and after A-FOSexpression, there is a conversion from a squamous to sebaceousphenotype.

Activating AP-1 activity in sebaceous adenomas inducessquamous tumors. To determine if restoring AP-1 activity insebaceous adenomas could produce a reciprocal conversion ofsebaceous tumors into squamous tumors, A-FOS expression wasrepressed in K5/A-FOS mice with existing sebaceous adenomasto reactivate AP-1 activity. Twenty-seven weeks after the initiationof the carcinogenesis protocol, K5/A-FOS mice were switched to adoxycycline diet to restore AP-1 activity, and tumors were collectedat multiple time points. Twenty percent of sebaceous adenomasconverted to papillomas; 24% were mixed squamous/sebaceoustumors; and 56% remained pure sebaceous adenomas when tumorswere analyzed for all time points (Table 1). Mixed tumors weredetected after only 5 weeks of AP-1 activity, whereas a completechange in tumor phenotype into squamous tumors was notobserved until 22 weeks.

In mixed tumors arising from either sebaceous adenomas orsquamous papillomas, the topology was similar; the squamousportion was at the periphery; and the sebocytes were in a centralportion of the tumor mass (Fig. 3D and E). In addition to

Table 1. Summary of tumor phenotype from carcinogenesis

Group R Squamous

papillomas

Sebaceous

adenomas

Mix squamous/

sebaceous tumors

Squamous

carcinomas

Sebaceous

carcinomas

% Conversion

Experiment 1

WT (n = 17/9), +Dox/�Dox 35 23 0 0 12 0 34%K5/A-FOS (n = 10), �Dox 11 1 10 0 0 0 0

K5/A-FOS (n = 14), +Dox (weeks 1-27),

�Dox (weeks 27-50)

17 8 0 6 3 0 17%

Experiment 2

WT (n = 16/16), +Dox/�Dox 18 9 0 0 9 0 50%

WT (n = 12), +Dox/�Dox (no DMBA) 0 0 0 0 0 0 0

K5/A-FOS (n = 11), +Dox (weeks 1-27),�Dox (weeks 27-50)

18 10 5 3 0 0 0

K5/A-FOS (n = 10), �Dox (weeks 1-27),

+Dox (weeks 27-50)

43 8 23 10 0 2 5%

K5/A-FOS (n = 10), +Dox/�Dox (no DMBA) 0 0 0 0 0 0 0

NOTE: Summary of tumors analyzed by histologic examination from both tumorigenesis studies. The genotype of each experimental group, the diet

regime, and number of animals (n) is shown in the first column. Also shown are the total no. tumors examined per group (R), no. papillomas, no.

sebaceous adenomas, no. mixed squamous sebaceous tumors, no. squamous cell carcinomas, no. sebaceous carcinomas, and % malignant conversion(no. carcinomas / total no. tumors � 100%).

Abbreviation: Dox, doxycycline.






modulating tumor lineage identity, continual suppression of AP-1activity inhibited malignant progression of sebaceous tumors(Table 1). However, when AP-1 activity was restored to sebaceousadenomas, 5% of the sebaceous tumors converted to sebaceouscarcinomas. These data are similar to the inhibition of malignantconversion of squamous papillomas expressing A-FOS, indicatingthat AP-1 activity is necessary for malignant conversion of bothsquamous and sebaceous tumors of the skin.

The cellular composition of the mixed tumors was complex.Although certain cells clearly seemed to be either squamous orsebaceous by histologic criteria, many cells seemed to havemorphologic similarity to sebocytes with condensed central nuclei,while maintaining a basophilic staining akin to squamouskeratinocytes (Fig. 3E). Immunohistochemical analysis of mixedtumors derived by expressing A-FOS in squamous papillomasidentified individual cells that express both K6 and ADRP (Fig. 3F),suggesting transdifferentiation of squamous cells into sebaceouscells. However, in some tumors where AP-1 activity has beeninhibited after papilloma formation, cells are observed that expressADRP but not K6, suggesting a complete transdifferentiation fromsquamous cells to sebaceous cells (Fig. 3G). Likewise, in mixedtumors derived through activating AP-1 in sebaceous adenomas, asimilar colocalization of squamous and sebaceous markers in thesame cells is observed (Fig. 3H , IHH and K6 are shown).

No differences in proliferation or cell death are observed in themixed tumors compared with either the papillomas or sebaceousadenomas. Tumor regression was not detected after A-FOSinduction (based on consistent tumor number and size), suggestingthat sebaceous conversion was not due to death of squamous cellsand selection of cells of the sebaceous lineage. All tumor typesshowed equivalent levels of cellular proliferation based on nuclearproliferating cell nuclear antigen staining (data not shown).Furthermore, apoptotic cells were not detected by terminaldeoxynucleotide transferase–mediated nick-end labeling or activecaspase-3 staining (data not shown) of papillomas, sebaceousadenomas, or mixed tumors.

Recent reports indicate that the sebaceous cell fate results fromdestabilization of the h-catenin protein (33). To determine if A-FOSexpression affects h-catenin protein abundance and distribution,papillomas (Fig. 4A), sebaceous adenomas (Fig. 4B), and mixedtumors (Fig. 4C and D) were immunostained for h-catenin.Papillomas stain strongly for h-catenin, with a membrane anddiffuse cytoplasmic localization, whereas sebaceous adenomas arenegative. In mixed tumors, intense h-catenin staining is onlyevident in the peripheral region of the tumor acquiring a squamousphenotype (Fig. 4D). After restoring AP-1 activity in sebaceousadenomas, some mixed tumors had h-catenin staining in regionswith sebaceous cells, further supporting the notion that trans-differentiation is occurring. The staining pattern of sFRP-3(previously known as Frizzled B), a secreted protein known toinhibit wnt signaling resulting in h-catenin protein destabilization(34), has the opposite staining pattern to that seen for h-catenin.sFRP-3 is abundant in a mixed tumor expressing A-FOS, localizingto cells with either a sebaceous or squamous phenotype (Fig. 4E)but was not detected in squamous papillomas (Fig. 4F).

Inhibiting AP-1 in primary keratinocytes induces sebaceousmarkers. We are able to recapitulate the induction of thesebaceous phenotype in cultured transgenic newborn primarykeratinocytes by expressing A-FOS. Keratinocyte cultures express-ing A-FOS have a 12-fold increase in cells staining with Oil Red-O(Fig. 5A-C), a marker for lipid accumulation as occurs in

differentiating sebaceous cells (Fig. 3I). Genetic approaches fromseveral laboratories have provided evidence that the balancebetween wnt and IHH signaling pathways determines commitmentto the squamous or sebaceous lineages (31, 35). Consistent with theinduction of a sebaceous phenotype in cultured keratinocytes,A-FOS alters the expression of genes in the wnt and frizzledfamilies. A-FOS up-regulates mRNA transcripts for the wntantagonists sFRP-3 and DKK3 (36) and down-regulates the wnttarget gene c-Myc (37) when keratinocytes are grown in either aproliferative medium (0.05 mmol/L calcium; Fig. 5D) or adifferentiation medium (0.12 mmol/L calcium; data not shown).Figure 5E confirms these changes at the protein level. Furthermore,up-regulation of the ADRP protein by A-FOS in keratinocytessupports a sebaceous phenotype (Fig. 5E ). Supporting thefunctional inhibition of AP-1 activity in these cultures is thedown-regulation of cyclooxygenase-2, a known AP-1-regulated gene(ref. 10; Fig. 5E).

The levels of h-catenin in keratinocyte cultures did notchange (Fig. 5D and E), suggesting that A-FOS acts downstreamto inhibit h-catenin function. To test this possibility, we examined ah-catenin-dependent luciferase reporter (TOP-FLASH), whichcontains TCF/LEF DNA-binding sites. When introduced into WTor K5/A-FOS keratinocytes and activated by lithium chloride,A-FOS inhibits the reporter activity, indicating that h-catenintranscriptional activity in this assay is dependent on active AP-1

Figure 4. h-Catenin protein is suppressed by A-FOS expression.A, immunofluorescence localization of h-catenin (red ) in a papilloma orepidermis surrounding a sebaceous adenoma (B) but not in the sebaceousadenoma. C, in a mixed tumor with reexpressed AP-1 activity (outlined withdashed line ), h-catenin (red) stains a subpopulation of transdifferentiatingsebaceous cells and surrounding squamous epidermis but is not present in theresidual sebaceous component. D, a mixed tumor reexpressing AP-1 activityshowing a h-catenin-positive peripheral region assuming the squamousphenotype. E, sFRP-3 is detected throughout tumors expressing A-FOS (shownis a mixed tumor from K5/A-FOS mice). F, sFRP-3 is not detected in papillomas.

Cancer Research





(Fig. 5F). Previous studies have shown physical cooperationbetween c-Jun, an AP-1 component and the TCF/LEF transcriptionfactors (38).

To evaluate the role for AP-1 in wnt signaling, we examined themRNA expression of canonical and noncanonical wnt family andpathway transcripts in primary keratinocytes by expressing A-FOS.Figure 6A indicates a broad regulation of wnt family transcripts byAP-1 activity with both increased and decreased expression in acomplex pattern, suggesting that an intimate crosstalk existsbetween AP-1 and wnt pathways. This relationship is supported byseveral recent studies using alternative approaches (39–41).

c-JUN binds to the promoters of Wnt family members. Todetermine if the wnt and frizzled family members whose mRNAconcentrations changed after A-FOS expression are direct targets ofAP-1 transcription factors in tissue culture, we examined ifphospho-c-Jun, a member of the AP-1 complex, is bound to theirpromoter regions using the ChIP assay. In brief, the ChIP assay wasdone with four different antibodies: phospho-c-JUN (specific tophosphorylated Ser73), CREB, C/EBP, and IgG. We used theNimblegen mouse promoter array to examine binding to 21,815promoters, each represented by 15 oligonucleotides. Phospho-c-JUN was enriched over 4-fold ( for z3 of the 15 oligos per promoter)in 1,060 promoters, CREB was enriched in 451 promoters andC/EBPh in 95 promoters. Previously identified known targets ofphospho-c-JUN were also enriched (42). The genes containingmultiple TRE motifs were enriched in phospho-c-JUN ChIPs.

The promoters of 31% (4 of 13) of the wnt and frizzled genesmentioned in Fig. 6A were enriched over 4-fold ( for three or moreoligos per promoter) for phospho-c-JUN, whereas on average, only5% (1,060 of 21,815) of promoters are enriched for phospho-c-JUN.This suggests that A-FOS misregulation of these genes is becausethey are direct targets of the AP-1 complex containing phospho-c-JUN. Figure 6B presents the binding of these three antibodiesto the promoter region for two genes. Phospho-c-JUN binds in thepromoter region of Wnt4 and Wnt5a , but C/EBPh and CREB do not.

Discussion

We describe a two-transgene tetracycline regulated mousesystem that allows for the reversible expression in the skin ofA-FOS, a dominant negative to the AP-1 transcription factorcomplex. The A-FOS dominant negative heterodimerizes withc-FOS dimerization partners, and these heterodimers can not bindto DNA, thus inhibiting AP-1 transcriptional activity. A genome-wide analysis of FOS dimerization partners indicates that the threeJun family members are the preferred FOS dimerization partners(20). A-FOS expression in skin results in hyperplasia of sebaceousglands in older mice. Expression of A-FOS during skin carcinogen-esis reveals a remarkable plasticity of tumor phenotype; squamoustumors form when AP-1 is active, and sebaceous adenomas formwhen AP-1 is not active. By reversing AP-1 activity after tumorsform, we observe each tumor type transdifferentiating into theother tumor type, schematically represented in Fig. 6C . Further-more, we observe an absolute requirement for functional AP-1activity for premalignant progression and malignant conversion inexperimental skin carcinogenesis. This requirement is found notonly in squamous tumors but also in sebaceous adenomas.

The AP-1 transcriptional complex is composed of multiplecombinations of Jun and Fos family heterodimers. A knockinexperiment that placed the JunB protein coding region into thec-Jun locus, creating a JunB protein under the regulation of c-Junpromoter, has been able to rescue the embryonic lethalityassociated with the c-Jun deletion (43). This shows that theseproteins have redundant functions during embryogenesis andsuggests that the difference in the phenotypes of c-Jun and JunBknockout mice is a consequence of their different spatial andtemporal expression properties (43). The redundant properties ofrelated Jun family members makes it difficult to identify thefunction of the AP-1 complex when individual members of the AP-1family are genetically deleted from the mouse genome.

Several knockout and transgenic mice have been generated thathave revealed the function of AP-1 in the skin. The tissue-specific

Figure 5. Development of Oil Red-O–positive cells intissue culture following A-FOS expression. WT (A ) andK5/A-FOS (B) skin keratinocytes were induced todifferentiate with 0.12 mmol/L calcium for 48 hours beforestaining with Oil Red-O. The differentiating K5/A-FOSkeratinocytes in 0.12 mmol/L calcium medium developedlarge intracellular deposits of lipid, whereas WT cellswere Oil Red-O negative. C, to quantify the number OilRed-O cells, 25 fields were counted for each genotypeusing a �10 objective. Columns, mean for each genotype;bars, SE. D, gene expression determined by reversetranscription-PCR from 0.05 mmol/L calcium cultures ofWT and K5/A-FOS keratinocytes for A-FOS, sFRP-3, IHH,DKK, c-Myc, h-catenin (b-cat ), and actin control.E, immunoblot analysis of protein expression from WTand K5/A-FOS keratinocytes cultured in 0.12 mmol/Lcalcium for 48 hours for COX-2, h-catenin, c-Myc, ADRP,IHH, and sFRP3. F, plasmid DNAs for the h-catenin-dependent reporter gene (TOP-FLASH) or control(FOP-FLASH) were transfected into WT FVB orK5/A-FOS cultured keratinocytes, and luciferase activitywas determined after treatment for 6 hours with 10 mmol/LLiCl to activate h-catenin.






deletion of c-Jun produces an eye-open-at-birth phenotype (EOB;refs. 3, 4). The double ablation of JunB and c-Jun in the skinproduces a hyperplastic, inflamed phenotype observed in ears andpaws of mice, which disappears when the mice are givenantibiotics (44). A transgenic mouse that expresses an NH2-terminal truncated c-Jun protein without a transactivation domainin the skin does not produce a basal phenotype but does diminishpapilloma and carcinoma formation (5, 45). This construct servesas a dominant negative outside of the DNA-binding complex andinteracts with multiple pathways, including AP-1, nuclear factor-nB, and the cAMP response element (8). The unique phenotype ofthe A-FOS mice suggests that this transgene inhibits more of theAP-1 complex than the single or double knockout mice alreadydescribed. Thus, it will be interesting to determine the phenotypeof the triple deletion of all three Jun family members in the skin todetermine if a sebaceous hyperplasia is observed. One conclusion isthat A-FOS is doing something distinct from inhibiting only c-Junas we do not observe an EOB phenotype.

Several additional possibilities could explain these differences inphenotypes produced by deleting Jun family members andexpressing A-FOS. One possibility is the incomplete inhibition ofAP-1 activity in the differentiated layers of the epidermis. As thenormal epidermis differentiates, the K5 promoter is silenced; thus,

A-FOS is not expressed in the differentiated strata. This is incontrast to the knockout experiment that irreversibly eliminatesactivity in all cells that are produced after basal layer differenti-ation. Furthermore, K5 expression is up-regulated and expandedinto the differentiating strata of skin tumors (46); thus, A-FOSwould be even more highly expressed in multiple differentiatingstates of tumor tissue. Thus, the hyperplastic sebaceous glands inaging K5/A-FOS mice could reflect a modification of K5 or AP-1expression in aging skin. The mixed mouse strains used in thetissue-specific knockout experiments and subsequent crosses couldalso influence the phenotype relative to the pure FVB/N strain wehave used. Transgenic mice expressing dominant-negative A-CREBor A-C/EBP do not produce the sebaceous adenoma phenotype insimilar carcinogenesis studies3, indicating that the A-FOS targetsdo not overlap with the specificity of A-CREB or A-C/EBP.

These results with tumor induction in K5/A-FOS mouse skinmay explain why only squamous tumors develop in the classicmouse skin carcinogenesis assay induced by DMBA, whereastumors of other phenotypes are not observed. In chemically

3 Vinson et al., unpublished results.

Figure 6. Analysis of mRNA levels ofWnt pathway family members in thepresence of A-FOS and c-Jun binding toWnt pathway promoter sequences.A, K5/A-FOS and WT FVB keratinocyteswere cultured in 0.05 mmol/L calciumgrowth medium for 5 days before RNAisolation. cDNA was prepared andamplified with primers specific for individualWnt pathway components. Also analyzedand found to not be altered by A-FOSexpression are APC, Frp-1, Disheveled,van Gogh, tcf1, tcf3, and tcf4 (data notshown). The + sign next to each geneindicates that phospho-c-JUN binds to thepromoter as defined by a 4-fold enrichmentin three or more oligonucleotides used tointerrogate the promoter according to aChIP-on-chip assay. Frzd5 and Prl1promoters (marked by dots) were not onthe Nimblegen microarray. B, ChIP withantibodies to phospho-c-JUN, C/EBPh,and CREB was used to isolate DNA boundby these transcription factors. This DNAwas amplified and hybridized to Nimblegenmouse promoter arrays. The results for theWnt4 and Wnt5a promoters are presented.Top, schematic for these promoters withthe transcriptional start site marked by anarrow. Bottom, location of potential AP-1binding sites with canonical bases in bold.The fold enrichment of the ChIP DNArelative to the reference genomic DNAshown for the oligonucleotides used tointerrogate the promoter is plottedat the location of the middle of theoligonucleotide. C, schematicrepresentation showing that modulation ofAP-1 activity causes alterations in wnt andhedgehog family members, shifting thebalance between squamous andsebaceous lineages, respectively.

Cancer Research





induced squamous tumors of the skin, H-ras mutations arefrequent and result in elevated AP-1 activity (17). This favors thewnt/h-catenin pathway and directs cells to the squamous lineage,resulting in papillomas that progress to squamous cell carcinomas(31). This signal could be further amplified by a positive feedbackloop of wnt signaling activating AP-1, as h-catenin overexpressionup-regulates the expression and activity of c-Jun and Fra-1 (47).

In contrast, tumor models that prevent signaling through h-catenin (31) may also up-regulate hedgehog signaling through theinhibition of AP-1 (this report). This would favor sebaceous tumorsthat rarely, if ever, progress to cancer in the absence of AP-1signaling. The elevation of IHH by AP-1 inhibition may alsocontribute to antagonizing wnt signaling as shown in colonicepithelium (48). Furthermore, the up-regulation of sFRP-3 by AP-1inhibition may contribute to suppression of wnt signaling andincreased h-catenin turnover, a finding that may have significancein other tumors of mixed phenotype. The involvement of activatedh-catenin in inducing squamous cells in glandular and secretoryepithelia argues for a general mechanism that epithelia use todifferentiate various cell types (49). The results from the mixedtumors and cultured keratinocytes support a dynamic link betweenAP-1 activity, wnt and hedgehog signaling, h-catenin stability, anddiscrimination between the squamous and sebaceous cell types.The ChIP experiment that shows that phospho-c-Jun is bound tothe promoters in the wnt and frizzled families suggests that the linkbetween AP-1 and the choice between the wnt and hedgehogsignaling pathways is a direct transcriptional link. A requirementfor AP-1 signaling in wnt/h-catenin–mediated intestinal cancerdevelopment has been shown in the APCmin mouse model,consistent with a broader role for AP-1/wnt signaling in cancerdevelopment (39). It will be interesting to determine if A-FOSexpression can switch cell identity in other epithelial systems.

Recently, attention has focused on the potential role of asubpopulation of stem-like cells in cancers that account for theinfinite replicative capacity and frequent phenotypic diversity ofthe tumor mass (50–53). Changes in the tumor microenvironmentcould regulate transcriptional pathways, including AP-1, in multi-potential cells to produce focal lineage alterations. In the K5/A-FOS tumor model described here, we detect cells that expressmarkers of two distinct lineages as the tumor transdifferentiatesfrom squamous to sebaceous or vice versa. Such biology suggeststhat tumor cells of defined terminal lineage are in fact plasticand can change lineage commitment between squamous orsebaceous end points as a result of changes in AP-1 transcrip-tional activity. Whether such cells also represent a population ofstem cells with infinite proliferative capacity remains to bedetermined. However, it was recently reported that terminallydifferentiated pancreatic islet cells can be induced to replicate(54). Our studies reveal a previously unknown role of AP-1transcriptional activity in determination and maintenance ofcellular lineage. The reversible activation of AP-1 transcriptionalpotential reveals the plastic nature of tumor cell differentiationand a mechanism for balancing signaling pathways that controlcell lineage determination.

Acknowledgments

Received 4/5/2006; revised 5/19/2006; accepted 5/26/2006.Grant support: Intramural Research Program of the NIH, National Cancer

Institute, Center for Cancer Research.The costs of publication of this article were defrayed in part by the payment of page

charges. This article must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Erin Kennedy and Susana Walters for their excellent assistance withcarcinogenesis studies and animal husbandry, Drs. Matthew Young and Nancy Colburn(National Cancer Institute) for providing TRE-Luciferase (TRE-LUC) mice, andDrs. Jeffery Rubin and Melinda Larsen for critical reading of the article.



References1. Eferl R, Wagner EF. AP-1: a double-edged sword intumorigenesis. Nat Rev Cancer 2003;3:859–68.

2. Jochum W, Passegue E, Wagner EF. AP-1 in mousedevelopment and tumorigenesis. Oncogene 2001;20:2401–12.

3. Zenz R, Scheuch H, Martin P, et al. c-Jun regulateseyelid closure and skin tumor development throughEGFR signaling. Dev Cell 2003;4:879–89.

4. Li G, Gustafson-Brown C, Hanks SK, et al. c-Jun isessential for organization of the epidermal leading edge.Dev Cell 2003;4:865–77.

5. Young MR, Li JJ, Rincon M, et al. Transgenic micedemonstrate AP-1 (activator protein-1) transactivationis required for tumor promotion. Proc Natl Acad SciU S A 1999;96:9827–32.

6. Dlugosz A, Merlino G, Yuspa SH. Progress incutaneous cancer research. J Investig Dermatol SympProc 2002;7:17–26.

7. Saez E, Rutberg SE, Mueller E, et al. c-fos is requiredfor malignant progression of skin tumors. Cell 1995;82:721–32.

8. Thompson EJ, Gupta A, Stratton MS, Bowden GT.Mechanism of action of a dominant negative c-junmutant in inhibiting activator protein-1 activation. MolCarcinog 2002;35:157–62.

9. Cooper SJ, MacGowan J, Ranger-Moore J, Young MR,Colburn NH, Bowden GT. Expression of dominantnegative c-jun inhibits ultraviolet B-induced squamouscell carcinoma number and size in an SKH-1 hairlessmouse model. Mol Cancer Res 2003;1:848–54.

10. Bachelor MA, Cooper SJ, Sikorski ET, Bowden GT.Inhibition of p38 mitogen-activated protein kinase and

phosphatidylinositol 3-kinase decreases UVB-inducedactivator protein-1 and cyclooxygenase-2 in a SKH-1hairless mouse model. Mol Cancer Res 2005;3:90–9.

11. Eckert RL, Efimova T, Balasubramanian S, Crish JF,Bone F, Dashti S. p38 Mitogen-activated protein kinaseson the body surface: a function for p38delta. J InvestDermatol 2003;120:823–8.

12. Chen N, Nomura M, She QB, et al. Suppression ofskin tumorigenesis in c-Jun NH(2)-terminal kinase-2-deficient mice. Cancer Res 2001;61:3908–12.

13. Jansen AP, Verwiebe EG, Dreckschmidt NE, WheelerDL, Oberley TD, Verma AK. Protein kinase C-epsilontransgenic mice: a unique model for metastaticsquamous cell carcinoma. Cancer Res 2001;61:808–12.

14. Wheeler DL, Ness KJ, Oberley TD, Verma AK.Inhibition of the development of metastatic squamouscell carcinoma in protein kinase C epsilon transgenicmice by a-difluoromethylornithine accompanied bymarked hair follicle degeneration and hair loss. CancerRes 2003;63:3037–42.

15. Arnott CH, Scott KA, Moore RJ, et al. Tumournecrosis factor-a mediates tumour promotion via a PKCa- and AP-1-dependent pathway. Oncogene 2002;21:4728–38.

16. Bode AM, Dong Z. Signal transduction pathways:targets for chemoprevention of skin cancer. LancetOncol 2000;1:181–8.

17. Rutberg SE, Adams TL, Glick A, Bonovich MT, VinsonC, Yuspa SH. Activator protein 1 transcription factorsare fundamental to v-rasHa-induced changes in geneexpression in neoplastic keratinocytes. Cancer Res 2000;60:6332–8.

18. Casanova ML, Larcher F, Casanova B, et al. A criticalrole for ras-mediated, epidermal growth factor receptor-

dependent angiogenesis in mouse skin carcinogenesis.Cancer Res 2002;62:3402–7.

19. Resnitzky D, Gossen M, Bujard H, Reed SI. Acceler-ation of the G1/S phase transition by expression ofcyclins D1 and E with an inducible system. Mol Cell Biol1994;14:1669–79.

20. Newman JR, Keating AE. Comprehensive identifica-tion of human bZIP interactions with coiled-coil arrays.Science 2003;300:2097–101.

21. Vinson C, Myakishev M, Acharya A, Mir AA, Moll JR,Bonovich M. Classification of human B-ZIP proteinsbased on dimerization properties. Mol Cell Biol 2002;22:6321–35.

22. Olive M, Krylov D, Echlin DR, Gardner K, TaparowskyE, Vinson C. A dominant negative to activation protein-1 (AP1) that abolishes DNA binding and inhibitsoncogenesis. J Biol Chem 1997;272:18586–94.

23. Diamond I, Owolabi T, Marco M, Lam C, Glick A.Conditional gene expression in the epidermis oftransgenic mice using the tetracycline-regulated trans-activators tTA and rTA linked to the keratin 5 promoter.J Invest Dermatol 2000;115:788–94.

24. Rincon M, Flavell RA. AP-1 transcriptional activityrequires both T-cell receptor-mediated and co-stimula-tory signals in primary T lymphocytes. EMBO J 1994;13:4370–81.

25. Moll JR, Acharya A, Gal J, Mir AA, Vinson C.Magnesium is required for specific DNA binding ofthe CREB B-ZIP domain. Nucleic Acids Res 2002;30:1240–6.

26. Rutberg SE, Saez E, Glick A, Dlugosz AA, SpiegelmanBM, Yuspa SH. Differentiation of mouse keratinocytes isaccompanied by PKC-dependent changes in AP-1proteins. Oncogene 1996;13:167–76.




Cancer Research


27. Rehman I, Lowry DT, Adams C, et al. Frequent codon12 Ki-ras mutations in mouse skin tumors initiated byN -methyl-N ¶-nitro-N -nitrosoguanidine and promotedby mezerein. Mol Carcinog 2000;27:298–307.

28. Bohlander SK, Espinosa R III, Le Beau MM, RowleyJD, Diaz MO. A method for the rapid sequence-independent amplification of microdissected chromo-somal material. Genomics 1992;13:1322–4.

29. Brasaemle DL, Barber T, Wolins NE, Serrero G,Blanchette-Mackie EJ, Londos C. Adipose differentia-tion-related protein is an ubiquitously expressed lipidstorage droplet-associated protein. J Lipid Res 1997;38:2249–63.

30. Quintanilla M, Brown K, Ramsden M, Balmain A.Carcinogen-specific mutation and amplification of Ha-ras during mouse skin carcinogenesis. Nature 1986;322:78–80.

31. Niemann C, Unden AB, Lyle S, Zouboulis Ch C,Toftgard R, Watt FM. Indian hedgehog and h-cateninsignaling: role in the sebaceous lineage of normal andneoplastic mammalian epidermis. Proc Natl Acad SciU S A 2003;100 Suppl 1:11873–80.

32. Wilson JB, Weinberg W, Johnson R, Yuspa S, Levine AJ.Expression of the BNLF-1 oncogene of Epstein-Barr virusin the skin of transgenic mice induces hyperplasia andaberrant expression of keratin 6. Cell 1990;61:1315–27.

33. Niemann C, Owens DM, Hulsken J, Birchmeier W,Watt FM. Expression of DeltaNLef1 in mouse epidermisresults in differentiation of hair follicles into squamousepidermal cysts and formation of skin tumours.Development 2002;129:95–109.

34. Jones SE, Jomary C. Secreted Frizzled-related pro-teins: searching for relationships and patterns. Bio-essays 2002;24:811–20.

35. Alonso L, Fuchs E. Stem cells in the skin: waste not,Wnt not. Genes Dev 2003;17:1189–200.

36. Weidinger G, Moon RT. When Wnts antagonizeWnts. J Cell Biol 2003;162:753–5.

37. He TC, Sparks AB, Rago C, et al. Identification ofc-MYC as a target of the APC pathway. Science 1998;281:1509–12.

38. Rivat C, Le Floch N, Sabbah M, et al. Synergisticcooperation between the AP-1 and LEF-1 transcriptionfactors in activation of the matrilysin promoter by thesrc oncogene: implications in cellular invasion. FASEB J2003;17:1721–3.

39. Nateri AS, Spencer-Dene B, Behrens A. Interaction ofphosphorylated c-Jun with TCF4 regulates intestinalcancer development. Nature 2005;437:281–5.

40. Hwang SG, Yu SS, Lee SW, Chun JS. Wnt-3a regulateschondrocyte differentiation via c-Jun/AP-1 pathway.FEBS Lett 2005;579:4837–42.

41. Le Floch N, Rivat C, De Wever O, et al. Theproinvasive activity of Wnt-2 is mediated through anoncanonical Wnt pathway coupled to GSK-3h andc-Jun/AP-1 signaling. FASEB J 2005;19:144–6.

42. Hayakawa J, Mittal S, Wang Y, et al. Identification ofpromoters bound by c-Jun/ATF2 during rapid large-scale gene activation following genotoxic stress. MolCell 2004;16:521–35.

43. Passegue E, Jochum W, Behrens A, Ricci R, WagnerEF. JunB can substitute for Jun in mouse developmentand cell proliferation. Nat Genet 2002;30:158–66.

44. Zenz R, Eferl R, Kenner L, et al. Psoriasis-like skindisease and arthritis caused by inducible epidermaldeletion of Jun proteins. Nature 2005;437:369–75.

45. Thompson EJ, MacGowan J, Young MR, Colburn N,Bowden GT. A dominant negative c-jun specifically

blocks okadaic acid-induced skin tumor promotion.Cancer Res 2002;62:3044–7.

46. Sundberg JP, Erickson AA, Roop DR, Binder RL.Ornithine decarboxylase expression in cutaneous papillo-mas in SENCAR mice is associated with altered ex-pression of keratins 1 and 10. Cancer Res 1994;54:1344–51.

47. Mann B, Gelos M, Siedow A, et al. Target genes ofh-catenin-T cell-factor/lymphoid-enhancer-factor sig-naling in human colorectal carcinomas. Proc Natl AcadSci U S A 1999;96:1603–8.

48. van den Brink GR, Bleuming SA, Hardwick JC, et al.Indian Hedgehog is an antagonist of Wnt signaling incolonic epithelial cell differentiation. Nat Genet 2004;36:277–82.

49. Li W, Qiao W, Chen L, et al. Squamous cell carcinomaand mammary abscess formation through squamousmetaplasia in Smad4/Dpc4 conditional knockout mice.Development 2003;130:6143–53.

50. Hope KJ, Jin L, Dick JE. Acute myeloid leukemiaoriginates from a hierarchy of leukemic stem cell classesthat differ in self-renewal capacity. Nat Immunol 2004;5:738–43.

51. Al-Hajj M, Becker MW, Wicha M, Weissman I, ClarkeMF. Therapeutic implications of cancer stem cells. CurrOpin Genet Dev 2004;14:43–7.

52. Dontu G, Al-Hajj M, Abdallah WM, Clarke MF, WichaMS. Stem cells in normal breast development and breastcancer. Cell Prolif 2003;36 Suppl 1:59–72.

53. Gerdes M, Yuspa S. The contribution of epidermalstem cells to skin cancer. Stem Cell Reviews 2005;1:225–32.

54. Dor Y, Brown J, Martinez OI, Melton DA. Adultpancreatic h-cells are formed by self-duplication ratherthan stem-cell differentiation. Nature 2004;429:41–6.




Documents

Activator Protein1 Activity Regulates Epithelial Tumor Cell Identity