letter

410 nature genetics • volume 21 • april 1999

A common human skin tumour is caused by activatingmutations in β-catenin

Edward F. Chan1, Uri Gat1, Jennifer M. McNiff2 & Elaine Fuchs1

1Howard Hughes Medical Institute, Department of Molecular Genetics and Cell Biology, The University of Chicago, Chicago, Illinois 60637, USA.2Department of Dermatology, Yale University School of Medicine, New Haven, Connecticut 06520, USA. Correspondence should be addressed to E.F. (e-mail:lain@midway.uchicago.edu).

WNT signalling orchestrates a number of developmental pro-grams1–3. In response to this stimulus, cytoplasmic β-catenin(encoded by CTNNB1) is stabilized, enabling downstream trans-criptional activation by members of the LEF/TCF family4,5. Oneof the target genes for β-catenin/TCF encodes c-MYC, explain-ing why constitutive activation of the WNT pathway can leadto cancer, particularly in the colon6. Most colon cancers arisefrom mutations in the gene encoding adenomatous polyposiscoli (APC), a protein required for ubiquitin-mediated degrada-tion of β-catenin7, but a small percentage of colon and someother cancers harbour β-catenin–stabilizing mutations (refs8–17). Recently, we discovered that transgenic mice expressingan activated β-catenin are predisposed to developing skintumours resembling pilomatricomas18. Given that the skin ofthese adult mice also exhibits signs of de novo hair-follicle mor-phogenesis, we wondered whether human pilomatricomasmight originate from hair matrix cells and whether they mightpossess β-catenin–stabilizing mutations. Here, we explore thecell origin and aetiology of this common human skin tumour.We found nuclear LEF-1 in the dividing tumour cells, providingbiochemical evidence that pilomatricomas are derived from hairmatrix cells. At least 75% of these tumours possess mutationsaffecting the amino-terminal segment, normally involved inphosphorylation-dependent, ubiquitin-mediated degradationof the protein. This percentage of CTNNB1 mutations is greaterthan in all other human tumours examined thus far, anddirectly implicates β-catenin/LEF misregulation as the majorcause of hair matrix cell tumorigenesis in humans.Pilomatricomas are common human skin tumours of unknownorigin and aetiology. The histologic appearance of the tumour ischaracterized by an exterior zone of densely packed, smallbasophilic cells. These cells resemble those of the hair folliclematrix, a compartment of proliferating cells that differentiate intothe hair shaft and its surrounding sheaths (Fig. 1a,b). The tumoursalso contain a transitional zone of cells displaying a gradual loss ofnuclei and an inner zone of ‘shadow’ cells consisting of enucleated,eosinophilic cellular ghosts (Fig. 1c,d). Pilomatricomas are similarin appearance to the tumours that develop in transgenic miceexpressing a keratin promoter-driven, stable form of β-catenintruncated at its N terminus18 (Fig. 1e,f).

Although LEF1 mRNA has also been detected in cultured epider-mal keratinocytes19 and in some melanoma cell lines8, proliferatinghair matrix cells can be distinguished from other epithelial cells ofthe skin by their high level of LEF1 mRNA and by the presence ofnuclear LEF-1 (refs 18–20). To further explore the relation betweenpilomatricomas and hair matrix, we stained human tumour sec-tions with anti-LEF-1 antibodies. Anti-LEF-1 prominently stainedthe nuclei of the proliferating cells of pilomatricomas, but it was notdetected in the transitional or shadow cells of the tumour (Fig. 2).

Under the conditions used here, LEF-1 antibodies did not stain thenuclei of normal skin cells surrounding the tumour (no normalhair follicle matrix cells were present in the sections).

Fig. 1 The hair follicle and derived pilomatricomas. a, Schematic of human hairfollicle. The transiently dividing matrix cells are relatively undifferentiated, butperiodically they withdraw from the cell cycle and commit to terminal differen-tiation. As they move upward they form the cortex and hair shaft, and theinner and outer root sheaths of the follicle. The portion of the follicle abovethe putative stem-cell population of the follicle (bulge) is permanent and is notderived from the matrix. b, Haematoxylin and eosin (H&E)-stained human skinsection depicting the morphology of the hair matrix and dermal papilla (DP).c,d, H&E-stained sections of a human pilomatricoma. e,f, H&E-stained sectionsof a pilomatricoma from a transgenic mouse expressing ∆N87β-catenin, an N-terminally truncated, activating form of the protein18. Bar represents 44 µm in(b,c,e) and 22 µm in (d,f).

a b

fe

c d

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m

letter

nature genetics • volume 21 • april 1999 411

The nuclear expression of LEF-1 in proliferating pilomatri-coma cells provided biochemical data in support of morphologi-cal evidence that these common tumours are derived from hairmatrix cells. Moreover, the intensity of anti-Lef/Tcf staining inmice was several-fold higher in proliferating tumour cells com-pared with wild-type hair matrix cells (data not shown). Asnuclear LEF-1 is often a consequence of WNT signalling, ourfindings suggested that uncontrolled proliferation in thesetumours might arise from constitutive stabilization and activa-tion of β-catenin/LEF-1, similar to a complex recently shown tobe a transactivator for the MYC proto-oncogene6. Due to inher-ent difficulties in detecting endogenous nuclear β-catenin evenwhen a WNT signal is active, we turned to genetic methods todirectly test this hypothesis.

Given the predisposition of keratin promoter-driven ∆N87 β-catenin mice to pilomatricoma-like tumours18, we first focusedon whether CTNNB1 itself might be mutated. Normally, β-catenin that is not assembled into adherens junctions becomesphosphorylated in its N-terminal segment, thereby targeting theprotein for ubiquitin-mediated degradation21–23 . Most humancancers that involve CTNNB1 mutations possess changes inamino acid residues in the N-terminal segment 8–17. We thereforescreened human pilomatricomas for the presence of theseknown, dominant, activating mutations. To separate as much aspossible of the tumour tissue from normal skin, we firstmicrodissected formalin-fixed, paraffin-embedded sections ofpilomatricoma biopsies from 16 different human patients.Because pilomatricomas are often not well encapsulated, it wasdifficult to obtain completely pure tumour samples. Followingisolation of genomic DNA from these samples, we used PCR toamplify a 200-bp fragment from exon 3 encoding an N-terminalportion of β-catenin.

Direct sequencing of this segment revealed a single nucleotidesubstitution in the DNAs from 12 samples (Fig. 3, left). Sequenc-ing of sense and antisense strands confirmed that only a singlebase pair was altered in each of these cases. In all samples, a mix-ture of altered and wild-type nucleotide was present, as expectedgiven the dominant positive behaviour of activating mutations inCTNNB1 and the somatic nature of β-catenin–activating muta-tions found in human cancers. The altered nucleotide peak wasproportionally smaller than the wild-type peak, a feature thatprobably was due to the presence of some wild-type tissue in themicrodissected tumour samples. Sequencing of the DNA frag-ments amplified from the microdissected normal tissue sur-rounding the tumour revealed wild-type CTNNB1 sequence.This was also true for DNA fragments amplified from total tissue

sections, indicating that the samples had to be relatively pure todetect the altered nucleotide present in the tumour cells.

We used two different methods to confirm the presence ofpoint substitutions in the patient samples. For four tumoursamples, we subcloned PCR products and sequenced individualclones. As expected, we obtained two types of clones, those withwild-type sequence and mutant clones harbouring the alteredbase (Fig. 3, right). In most cases, the point substitutionsresulted in the ablation of one of two HinfI sites in the PCRfragment, enabling the use of restriction endonuclease analysisto verify the existence of a substitution (Fig. 4, lanes 1, 3, 5). Incontrast, a fragment missing one of the HinfI sites was notgenerated in the normal tissue samples microdissected from thesame tumour blocks (Fig. 4, lanes 2, 4, 6). This confirmed thesomatic nature of CTNNB1 alterations in these tumours.

Fig. 2 Nuclear LEF-1, a marker ofhair matrix cells, is expressed in theproliferating cells of human piloma-tricomas. a, Immunofluorescence mi-croscopy of human pilomatricoma 4showing nuclear localization of LEF-1antibody (green) in matrix cells butnot in transitional or shadow cells.No staining was seen in the epider-mis and dermis. b, Serial section fromthe same tissue block processed asfor (a) but without addition of pri-mary antibody. Non-specific fluores-cence stems from the presence of redblood cells in the tissue sections.c, DAPI staining (blue) of same sec-tion as shown in (a). Arrows denotecommon reference point in all threeframes. Bar in (c) represents 30 µm.

Fig. 3 Point substitutions in exon 3 of CTNNB1 in human pilomatricomas.Sequences in the left column were derived directly from PCR products ampli-fied from genomic DNA of microdissected tumour tissue. Sequences in theright column were derived from individual subclones of the PCR fragments.Arrows denote the altered nucleotide present in a subpopulation of DNAs inthe microdissected tissue. The sequence of subcloned DNA from pilomatricoma3 is from the antisense strand (asterisk).

Genomic Subcloned

cba

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m

letter

412 nature genetics • volume 21 • april 1999

CTNNB1 alterations found in the tumour samples appeared tobe bona fide mutations and not polymorphic variations, asjudged by the fact that all genomic blood DNAs from 56 normalhuman controls contained both Hinf1 sites (Fig. 4, lanes 7−11).

The 12 nucleotide alterations found in the different inci-dences of pilomatricomas all encode putative missense muta-tions in the N-terminal segment of β-catenin: D32Y(GAC→TAC), D32G (GAC→GGC), S33F (TCT→TTT; foundin two pilomatricomas), S33Y (TCT→TAT; found in two pilo-matricomas), G34E (GGA→GAA; found in three pilomatrico-mas), S37C (TCT→TGT), S37F (TCT→TTT) and T41I(ACC→ATC; Fig. 5a). Each of these mutations has been previ-ously associated with a variety of human carcinomas and celllines (Fig. 5a). Seven mutations represent alterations in serine33, serine 37 and/or threonine 41, essential for GSK-3β-depen-dent phosphorylation24–26. Five mutations are in the asparticacid or glycine residues flanking serine 33. The DSG sequence,along with serine 37 (Fig. 5a, underlined residues), has beencharacterized as a ubiquitination-targeting motif on the basis ofits conservation with IκBα, another protein that can be targetedfor degradation by ubiquitin ligases that recognize phosphory-lated sequence26,27. Mutations in these adjoining residues mayinterfere with ubiquitination, possibly through altering theGSK-3β kinase-recognition sites.

The difference emerging between CTNNB1 mutations in pilo-matricomas and those in previously described tumours lies notin their location but in their significantly higher frequency(Fig. 5b). In fact, of 16 pilomatricomas analysed, only 1 wasdemonstrated to be wild type, whereas 12 were mutant for β-catenin. For the other three, the quality and/or quantity of iso-lated tumour DNA from archived paraffin blocks precluded ourability to confirm CTNNB1 mutations by sequencing, althoughrestriction endonuclease digestion assays indicated that alter-ations existed in two of these samples. Thus, pilomatricomas rep-resent the first case in which a human tumour is commonlygenerated through mutations in CTNNB1.

Pilomatricomas now join colon cancers as the two tumourtypes that are known to arise primarily from genetic alterationsthat act by influencing β-catenin stabilization. These two tumourtypes, however, seem to favour different genetic mechanisms,CTNNB1 versus APC mutations, in achieving this outcome. Apriori, this may be due to the exposure of skin and intestine todifferent mutagens, namely ultraviolet light and dietary carcino-gens, respectively. Although this variable may be a contributingfactor, the two epithelial cell types may also respond differently tomutations in CTNNB1 versus APC, perhaps due to tissue-specificdifferences in either the levels of APC or other β-catenin or LEF-1 inhibitory factors. It is notable that mice expressing an N-ter-minally truncated form of β-catenin in their intestine did notdevelop colon tumours28, whereas those expressing a similartransgene in skin displayed abundant tumorigenesis18. Con-versely, some humans have germline mutations in APC anddevelop a complex form of familial adenomatous polyposis coli

Fig. 4 A restriction endonuclease assay to verify the existence of point substitu-tions in CTNNB1 of human pilomatricomas. HinfI digestion of a 200-bp PCRproduct encompassing exon 3 of CTNNB1. DNA samples are from human pilo-matricomas 1−3 (lanes 1, 3, 5, respectively), normal tissue from the respectivetissue blocks (lanes 2, 4, 6) and peripheral blood leukocytes from normal indi-viduals (lanes 7−11). The 62-bp band represents an altered allele lacking one ofthe two natural HinfI sites. The second site is located 7 base pairs away. Bothsites are present in the wild-type allele represented by the 55-bp band. The 7-bp band is not visible on the gel. Fragment sizes were determined using DNAstandards. The asterisk denotes a faint 62-bp band present in ‘wild-type’ tissuederived from pilomatricoma 1, most likely due to the presence of contaminat-ing tumour cells in the sample.

Fig. 5 β-catenin mutations in pilomatricomas. a, Schematic of β-catenin, depicting the location of the GSK3β-dependent phosphorylation sites in the N-terminalsegment (with phosphorylation sites in bold) and the Armadillo repeats characteristic of this family of proteins. Shown are the sequences of the wild-type, con-served N-terminal segment from human (Hu), mouse (Mu), Xenopus laevis (Xe) and Drosophila melanogaster (Drosoph), and the corresponding location of pilo-matricoma mutations and previously described mutations from other types of human tumours and cell lines. b, Incidence of activating mutations in CTNNB1found in various primary human tumour samples. Compiled from studies of colorectal carcinomas, hepatocellular carcinomas, medulloblastoma, ovarian carcino-mas, prostate carcinomas and endometrial carcinomas8–17. Of 16 skin samples, 13 contained pilomatricoma material of sufficient quality and amount to be con-firmed by both restriction endonuclease digestion and DNA sequencing. Twelve had the CTNNB1 mutations shown; one was wild type. The estimate of 75% isthus conservative, as the level in this group alone may be as high as 92%.

a

b

tumours and cell lines

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m

letter

nature genetics • volume 21 • april 1999 413

referred to as Gardner syndrome; although these patients developcolon cancer at a very high rate, they display only occasional pilo-matricomas29,30. The manifestations of APC and CTNNB1 muta-tions in hair matrix and colon epithelia appear to be distinct,suggesting that there may be differences in the ways in which β-catenin may be most effectively stabilized in different WNT-responsive tissues.

Our results identify for the first time a tumour in which mostcases (at least 75%) contain β-catenin–stabilizing mutations inCTNNB1. Our studies indicate that pilomatricomas are derivedfrom hair matrix cells and that they are typified by expression ofthe β-catenin partner protein, LEF-1. Our results further suggestthat acquisition of a β-catenin–stabilizing mutation in CTNNB1results in LEF-1 transactivation, leading to the development ofhuman tumours of hair matrix differentiation.

MethodsTumour samples. Formalin-fixed, paraffin-embedded specimens of pilo-matricomas (from sporadic incidences) were obtained from 13 patients atYale University Medical Center, 2 patients at Columbia University Schoolof Medicine and 1 patient at Rush Presbyterian-St. Luke’s Medical Center.Genomic DNAs isolated from blood samples of normal control patientswere generously provided by G. Bell.

DNA analysis. Sections (10 µm) of paraffin-embedded tumour tissue werestained with H&E to reveal regions containing the tumour. Using thisinformation, thicker (50 µm) unstained serial sections were then microdis-sected to separate the tissue into areas enriched for tumour versus normalcells. Genomic DNAs were then extracted from the tissue after proteinase Ktreatment (Qiagen tissue kit). Exon 3 of CTNNB1 was amplified from tis-sue genomic DNAs using PCR and specific oligonucleotide primers asdescribed10. Another round (25 cycles) of PCR with PFU polymerase(Stratagene) was performed on the 200-bp gel-purified PCR product to

1. Miller, J.R. & Moon, R.T. Signal transduction through β-catenin and specificationof cell fate during embryogenesis. Genes Dev. 10, 2527–2539 (1996).

2. Willert, K. & Nusse, R. β-catenin: a key mediator of Wnt signaling. Curr. Opin.Genet. Dev. 8, 95–102 (1998).

3. Barth, A.I., Nathke, I.S. & Nelson, W.J. Cadherins, catenins and APC protein:interplay between cytoskeletal complexes and signaling pathways. Curr. Opin.Cell Biol. 9, 683–690 (1997).

4. Behrens, J. et al. Functional interaction of β-catenin with the transcription factorLEF-1. Nature 382, 638–642 (1996).

5. Bienz, M. TCF: transcriptional activator or repressor? Curr. Opin. Cell Biol. 10,366–372 (1998).

6. He, T.C. et al. Identification of c-MYC as a target of the APC pathway. Science 281,1509–1512 (1998).

7. Kinzler, K.W. & Vogelstein, B. Lessons from hereditary colorectal cancer. Cell 87,159–170 (1996).

8. Rubinfeld, B. et al. Stabilization of β-catenin by genetic defects in melanoma celllines. Science 275, 1790–1792 (1997).

9. Morin, P.J. et al. Activation of β-catenin-Tcf signaling in colon cancer by mutationsin β-catenin or APC. Science 275, 1787–1790 (1997).

10. Palacios, J. & Gamallo, C. Mutations in the β-catenin gene (CTNNB1) inendometrioid ovarian carcinomas. Cancer Res. 58, 1344–1347 (1998).

11. Muller, O., Nimmrich, I., Finke, U., Friedl, W. & Hoffmann, I. A β-catenin mutationin a sporadic colorectal tumor of the RER phenotype and absence of β-cateningermline mutations in FAP patients. Genes Chromosomes Cancer 22, 37–41(1998).

12. Miyoshi, Y. et al. Activation of the β-catenin gene in primary hepatocellularcarcinomas by somatic alterations involving exon 3. Cancer Res. 58, 2524–2527(1998).

13. Zurawel, R.H., Chiappa, S.A., Allen, C. & Raffel, C. Sporadic medulloblastomascontain oncogenic β-catenin mutations. Cancer Res. 58, 896–899 (1998).

14. Voeller, H.J., Truica, C.I. & Gelmann, E.P. β-catenin mutations in human prostatecancer. Cancer Res. 58, 2520–2523 (1998).

15. Fukuchi, T. et al. β-catenin mutation in carcinoma of the uterine endometrium.Cancer Res. 58, 3526–3528 (1998).

16. Sparks, A.B., Morin, P.J., Vogelstein, B. & Kinzler, K.W. Mutational analysis of theAPC/β-catenin/Tcf pathway in colorectal cancer. Cancer Res. 58, 1130–1134 (1998).

17. de La Coste, A. et al. Somatic mutations of the β-catenin gene are frequent in

mouse and human hepatocellular carcinomas. Proc. Natl Acad. Sci. USA 95,8847–8851 (1998).

18. Gat, U., DasGupta, R., Degenstein, L. & Fuchs, E. De novo hair folliclemorphogenesis and hair tumors in mice expressing a truncated β-catenin in skin.Cell 95, 605–614 (1998).

19. Zhou, P., Byrne, C., Jacobs, J. & Fuchs, E. Lymphoid enhancer factor 1 directs hairfollicle patterning and epithelial cell fate. Genes Dev. 9, 570–583 (1995).

20. van Genderen, C., Okamura, R.M., Farinas, I., Quo, R.-G. & Parslow, T.G.Development of several organs that require inductive epithelial-mesenchymalinteractions is impaired in LEF-1-deficient mice. Genes Dev. 8, 2691–2703 (1994).

21. Munemitsu, S., Albert, I., Rubinfeld, B. & Polakis, P. Deletion of an amino-terminalsequence stabilizes β-catenin in vivo and promotes hyperphosporylation of theadenomatous polyposis coli tumor suppressor protein. Mol. Cell. Biol. 16,4088–4094 (1996).

22. Aberle, H., Bauer, A., Stappert, J., Kispert, A. & Kemler, R. β-catenin is a target forthe ubiquitin-proteasome pathway. EMBO J. 16, 3797–3805 (1997).

23. Cadigan, K.M. & Nusse, R. Wnt signaling: a common theme in animaldevelopment. Genes Dev. 11, 3286-3305 (1997)

24. Peifer, M., Pai, L.M. & Casey, M. Phosphorylation of the Drosophila adherensjunction protein Armadillo: roles for wingless signal and zeste-white 3 kinase.Dev. Biol. 166, 543–556 (1994).

25. Yost, C. et al. The axis-inducing activity, stability, and subcellular distribution of β-catenin is regulated in Xenopus embryos by glycogen synthase kinase 3. GenesDev. 10, 1443–1454 (1996).

26. Orford, K., Crockett, C., Jensen, J.P., Weissman, A.M. & Byers, S.W. Serinephosphorylation-regulated ubiquitination and degradation of β-catenin. J. Biol.Chem. 272, 24735–24738 (1997).

27. Chen, Z.J., Parent, L. & Maniatis, T. Site-specific phosphorylation of IκBα by anovel ubiquitination-dependent protein kinase activity. Cell 84, 853–862 (1996).

28. Wong, M.W., Rubinfeld, B. & Gordon, J.I. Effects of forced expression of a NH2-terminal truncated β-catenin on mouse intestinal epithelial homeostasis. J. CellBiol. 141, 765–777 (1998).

29. Foulkes, W.D. A tale of four syndromes: familial adenomatous polyposis, Gardnersyndrome, attenuated APC and Turcot syndrome. QJM 88, 853–863 (1995).

30. Pujol, R.M., Casanova, J.M., Egido, R., Pujol, J. & de Moragas, J.M. Multiplefamilial pilomatricomas: a cutaneous marker for Gardner’s syndrome? Pediatr.Dermatol. 12, 331–335 (1995).

increase the yield. PCR products were gel-purified (Qiagen gel-extractionkit) and sequenced with dRhodamine Terminator Cycle Sequencing ReadyReaction kit (Perkin-Elmer Applied Biosytems). Sequencing reactions wereperformed in both sense and antisense directions using the above primers.Data were collected and analysed using Applied Biosystems sequencinganalysis software and MacVector 6.0 (Oxford Molecular Group). Confir-mation of mutations was performed by subcloning of PCR products(TOPO-TA cloning kit, Invitrogen). Sequence analysis was performed asabove with M13 forward and reverse oligonucleotide primers. Wheneverpossible, mutations were verified by restriction endonuclease analysis offreshly prepared PCR product. Digested samples were subjected to elec-trophoresis through 5% Metaphor agarose gels (FMC Bioproducts).

Immunohistochemistry. Sections (5 µm) were cut from paraffin-embed-ded tissue specimens, which were then deparaffinized and rehydrated.Antigen retrieval was performed by autoclaving sections for 15 min insodium citrate buffer (10 mM, pH 6.0). Sections were blocked with 20%heat-inactivated normal goat serum (HINGS), 0.1% Triton X-100 in PBSat 25 °C for 1 h. Rabbit anti-LEF-1 polyclonal antibody (1:100) was addedin 3% HINGS, 0.1% Triton X-100 in PBS at 25 °C for 1 h. Fluorescence-conjugated secondary antibodies were obtained from Jackson Laborato-ries. In negative control reactions, the primary antibody was omitted.

AcknowledgementsWe thank A. Christiano, M. Tharp and R. Elenitsas for tissue samples; R.Grosschedl for LEF-1 antibody; and M. Medenica for providing H&E-stainedslides of normal human scalp skin. This work was supported in part by theNational Institutes of Health (NIH-RO1-AR31737 and NCI-P50DE/CA-11921). E.F. is an Investigator of the Howard Hughes Medical Institute.E.F.C. is supported by the Howard Hughes Medical Institute PostdoctoralResearch Fellowship for Physicians.

Received 2 February; accepted 5 March 1999.

© 1999 Nature America Inc. • http://genetics.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://g

enet

ics.

nat

ure

.co

m