Defining the origins of Ras/p53-mediatedsquamous cell carcinomaAndrew C. Whitea,b, Kathy Trana, Joan Khuua, Christine Danga, Yongyan Cuia, Scott W. Binderc, andWilliam E. Lowrya,b,d,1

aDepartment of Molecular, Cell, and Developmental Biology, bEli and Edythe Broad Center of Regenerative Medicine and Stem Cell Research, cDepartment ofPathology, Dermatopathology, David Geffen School of Medicine, and dJonsson Comprehensive Cancer Center, University of California, Los Angeles, CA 90095

Edited by Brigid L. M. Hogan, Duke University Medical Center, Durham, NC, and approved March 29, 2011 (received for review September 9, 2010)

Theprecise identityof cancer cellsoforiginand themolecular eventsof tumor initiation in epidermal squamous cell carcinoma (SCC) areunknown.Herewe show thatmalignancy potential is related to thedevelopmental capacity of the initiating cancer cell in a geneticallydefined, intact, and inducible in vivo model. Specifically, these datademonstrate that SCCs can originate from inside the hair folliclestem cell (SC) niche or from immediate progenitors, whereas moredevelopmentally restricted progeny, the transit amplifying (TA)cells, areunable togenerateevenbenign tumors in thesamegeneticcontext. Using a temporal model of tumorigenesis in situ, we high-light the phenotypes of cancer progression from the hair follicleSC niche, including hyperplasia, epithelial to mesenchymal transi-tion, and SCC formation. Furthermore, we provide insights into theinability of hair follicle TA cells to respond to tumorigenic stimuli.

mouse models of cancer | tumor initiating cells | epidermal stem cells

Squamous cell carcinoma (SCC), a common type of non-melanoma skin cancer, harbors significant risk of metastasis

that can eventually lead to death (1). Permanent treatment andprevention of SCC requires an improved understanding of theunique cell types capable of initiating these malignancies andthe mechanisms that underlie their transformation. Mutations inthe Ras gene family are found in 30% of all human cancers andare often found in human cases of SCC (2–4). Numerous animalstudies have extended the understanding of SCC by demon-strating a causative role for the Ras family in the development ofSCC. In these mouse models, Ras signaling has been shown to besufficient to drive epidermal tumorigenesis through either over-expression or targeted expression of an oncogenic form of Hrasor Kras. These studies have targeted broad epidermal cell pop-ulations that include stem cells (SCs), as well as cells of the outerroot sheath (ORS) and interfollicular epidermis (IFE) (4–9).Among these, a landmark article indicated that the SCC cells oforigin reside in the hair follicle rather than the IFE (5). How-ever, the keratin promoter used to drive oncogenic Ras expres-sion did not allow for a direct determination of the cancer cell oforigin amongst the different cell types in the hair follicle. Thecandidates for an SCC cell of origin therefore remained follic-ular SCs, ORS cells, transit amplifying (TA) cells in the matrix,and the differentiated cells of the inner root sheath (IRS) andhair shaft. In addition, other studies have shown that lineage-restricted or differentiated cells of the IFE are sufficient to serveas cancer cells of origin when subjected to supraphysiologicallevels of Ras activity at sites of mechanical stress (10). Thesehigh levels of Ras expression, however, are not normally found inhuman cases of SCC (10, 11). Between follicular SCs and TAcells, it is not clear whether one or both are capable of initiatingtumorigenesis. Thus, we sought to directly compare the malig-nancy potential between follicular SCs and TA cells.Recent evidence suggests that a range of cell types populate

the SC niche of the hair follicle, each capable of reconstitutingdifferent portions of the hair follicle. These subpopulations eachexhibit SC characteristics and have at least one unique expres-sion marker (12–16). Lgr5 marks the bottom of the bulge during

telogen (resting phase of the hair cycle) and the entire ORSduring anagen (growth phase) (13). Lgr6 marks the top of thebulge, exclusive of the LGR5 population, and represents a slowerdividing subset of cells during telogen (14). A fragment of thekeratin 15 promoter has been shown to be active across the entirebulge in telogen and anagen and in hair germs during telogen(15, 17). Because the K15CrePR allele seems to be specific to thebulge during anagen, whereas the LGR5-Cre-eGfp allele marksthe entire ORS during anagen, we used the K15CrePR allele tomanipulate gene expression specifically in the bulge at any pointin the hair cycle.The inducible mouse model used here allows for a direct spatial

and temporal comparison of tumorigenic potential between SCsand TA cells of the hair follicle. This model also enables a precisecharacterization of the earliest steps of oncogenic Ras and Ras/p53 (oncogenic Ras expression with p53 gene deletion) inducedtumorigenesis, including stages of hyperplasia, epithelial to mes-enchymal transition (EMT), and dedifferentiation. Finally, weanalyze downstream signaling pathways of Ras in different targetcells to highlight the role these pathways play in Ras and Ras/p53-induced tumorigenesis.

ResultsTo determine the cellular requirements for SCC, we generatedan inducible, in vivo model system that allows for expression ofan oncogene and deletion of a tumor suppressor in either the SCor the TA cell populations of the intact hair follicle. To target thehair follicle SC compartment, an allele was used that comprisesa fragment of the keratin 15 gene regulatory unit driving to theCrePR gene. This allele is active in the hair follicle bulge, whereboth LGR5+ and LGR6+ cells reside, and is known to be theprimary site for multipotent follicular SCs (Fig. 1 A and B andFig. S1A) (17, 18). To target the TA cell compartment, theShhCreER knockin allele was used, because it has been shown tobe active in a subset of hair follicle matrix cells and not in otherlineages of the epidermis (Fig. 1 A and B and Fig. S1C) (19, 20).ShhCreER+ matrix cells are both rapidly dividing and producethe differentiated IRS of the follicle and thus meet the definitionof TA cells.The K15CrePR and ShhCreER inducible alleles were used in

combination with the Lox-Stop-Lox-KrasG12D (LSL-KrasG12D)knockin and a floxed p53 allele to induce tumorigenic trans-formation in a cell type-specific manner during murine adult-hood (21–24). Importantly, this oncogenic form of the Kras gene

Author contributions: A.C.W. and W.E.L. designed research; A.C.W., K.T., J.K., C.D., Y.C.,and W.E.L. performed research; A.C.W., K.T., J.K., S.W.B., and W.E.L. analyzed data; andA.C.W. and W.E.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed: E-mail: blowry@ucla.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1012670108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1012670108 PNAS | May 3, 2011 | vol. 108 | no. 18 | 7425–7430

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 20

, 202

0

is driven by the endogenous locus and therefore does not resultin supraphysiological expression of a transgene. Because murineand human SCC have been linked to activating mutations in Rasfamily members (including both Hras and Kras) and mutations in

p53 (2, 4, 5, 25–31), the LSL-KrasG12D allele alone or with floxedp53 alleles was used in conjunction with either K15CrePR orShhCreER to determine the relative capacity of hair follicle SCsvs. lineage-restricted TA cells to serve as epidermal cancer cellsof origin in SCC. Although this same LSL-KrasG12D allele hasbeen used to induce tumorigenesis in the epidermis previously,a direct comparison of SC vs. TA was not performed (6, 9).First, to determine the effect of targeting oncogenic KrasG12D

expression specifically to either the hair follicle SC niche or amore restricted TA population, the mating strategy outlined inFig. 1B was used. Hair follicles undergo cycles of growth (ana-gen), rest (telogen), and degeneration (catagen) throughout thelife of the animal. Animals were treated with either tamoxifen ormifepristone at the onset of anagen, as judged by regrowth aftershaving. Six to 10 weeks after treatment with mifepristone(K15CrePR) or tamoxifen (ShhCreER) by i.p. injection, the epi-dermis of K15CrePR; LSL-KrasG12D (K15CrePR; KrasG12D) miceconsistently exhibited structures that appeared as keratin-filledcysts and displayed widespread hyperplasia of the hair follicleand sebaceous gland (n = 18 of 18 animals; Fig. 1C). Further-more, affected follicles never returned to telogen. These phe-notypes were found in both dorsal and ventral epidermis,although cysts appeared to be much more prevalent in the ven-tral epidermis (Fig. 1D). Some of these phenotypes were con-sistent with a recent report that also targeted this populationwith the same tumorigenic Kras allele (32).In contrast, induction of the same oncogenic stimulus in a

portion of the TA cells of the hair follicle matrix during anagenin ShhCreER; LSL-KrasG12D (ShhCreER; KrasG12D) mice resul-ted in no abnormal epithelial structures, and follicles returned totelogen within a 10-wk timeframe (n = 9 of 9; Fig. 1C). Thesedata suggest that when the KrasG12D oncogene is expressed in thehair follicle SC niche, benign tumor formation ensued, whereasthe same stimulus delivered to Shh-expressing matrix TA cellswas unable to generate any signs of tumorigenesis. These dataindicate that bulge cells or their immediate descendants, beforebecoming lineage-restricted TA cells, can serve as cells of originfor epidermal tumorigenesis.To confirm that Cre activity independent of mifepristone or

tamoxifen administration did not confound this model system,untreated controls containing the appropriate tumorigenicalleles were examined. Many untreated K15CrePR; KrasG12D

animals did eventually develop hair follicle hyperplasias andfollicular cysts, but these were less severe and occurred only attimepoints well beyond those used for the analyses described inthis report (Fig. 1D). This effect was caused by uninducedtranslocation of CrePR to the nucleus, not because of expressionfrom the K15 promoter at sites other than the follicular bulge, asdetermined by lineage tracing (see below; Fig. S1 A and B).To verify the fidelity of transgene expression specifically to

bulge SCs and matrix TA populations, K15CrePR and ShhCreERmice were crossed to LSL-YFP mice to perform lineage tracing(33). Consistent with previously published work, these two meth-ods generated YFP expression specifically in either the hairfollicle SC niche and hair germ or in matrix TA cells after a briefpulse with mifepristone or tamoxifen (Fig. S1 A and C) (20).ShhCreER; LSL-YFP animals further showed lineage tracing ofYFP from the TA compartment to the precortex, the IRS, rarelyto the cortex and cuticle, and never to the companion layer (Fig.S2 A and B).To determine the rate of recombination induced by Cre in

K15CrePR and ShhCreER skin, we used the YFP allele to cal-culate recombination efficiency on a per cell per hair folliclebasis. A similar level of recombination was found in these twomouse lines: K15CrePR skin exhibited 17.4 ± 2.6 YFP+ cells perbulge (34.0 ± 4.7%, across 17 follicles), and ShhCreER skinshowed 18.2 ± 1.4 YFP+ cells per Shh-expressing matrix (36.7 ±2.1%, across 18 follicles). To confirm that the KrasG12D allele was

Fig. 1. Phenotypes resulting from KrasG12D expression in SC and TA cellpopulations of the hair follicle. (A) Model depicting the location and pro-moter specificity of SCs (K15+) and TA cells (Shh+) of the hair follicle duringthe growing phase of the hair cycle (anagen). In addition, a number ofrelevant cell types are shown (cp, companion layer; cx, hair cortex; ors, outerroot sheath; dp, dermal papillae; mx, matrix; ge, hair germ). (B) Model of thetargeting strategy to express oncogenic KrasG12D in either bulge SCs ormatrix TA cells. (C) Histopathology of phenotypes induced by KrasG12D ex-pression in the follicular bulge, including hyperplastic sebaceous glands (i),hyperplastic hair follicles (ii), and follicular cysts (iii). No apparent abnormalhair follicle phenotypes were detected in ShhCreER; KrasG12D skin. (Magni-fication, 10×.) (D) Quantification of phenotypes in treated and untreatedmice. The indicated phenotypes were quantified on a per-follicle basison the indicated genotypes with or without mifepristone treatment. Thetimecourse indicates the number of weeks after anagen, as measured byregrowth after shaving. The number of phenotypes found in treated animalswas significantly increased over those found in untreated animals, even attimepoints far later than those used for phenotypic analysis in treated ani-mals. *P ≤ 0.049).

7426 | www.pnas.org/cgi/doi/10.1073/pnas.1012670108 White et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 20

, 202

0

expressed in Cre-activated YFP+ populations, we performed RT-PCR on YFP+ and YFP− sorted cells to demonstrate that theHindIII restriction site introduced into theKras allele by theG12Dmutation was present (34). This KrasG12D-specific HindIII re-striction site was enriched in the YFP+ population but not foundin the KrasG12D YFP− population or in wild-type Kras controls(Fig. S2C). Together, these analyses indicated that biased re-combination frequency or efficiency could not account for drasticdifferences in tumorigenic potential of K15+ and SHH+ cells.Lineage tracing performed in other studies with this K15CrePR

allele and a Rosa-LSL-LacZ reporter, or with K15-GFP mice,indicated that rare labeled cells can be found in the IFE (17, 18).We were rarely able to detect YFP+ cells (0.2%, n = 5,772 IFEcells) in the IFE of treated K15CrePR; LSL-YFP animals, evenat extended timepoints (Fig. S1B). Furthermore, the phenotypesfound in treated K15CrePR; KrasG12D and K15CrePR; KrasG12D;p53ff animals during initiation and progression distinctly arosefrom the hair follicle within the dermis (see below) and occurredat a much higher frequency relative to the leaky expression in theIFE. These results indicated that although the K15 promoterused here may have minor off-target expression in the IFE, thesecells did not factor into the interpretation of these experiments.Immunostaining for total Kras protein in both follicular

SC (CD34+) and matrix populations also indicated very similarlevels of expression in both control and induced animals (Fig. S3Aand B). This suggests that the difference observed in the tumori-genic sensitivity between SC and TA compartments is likely notdue to a significantly higher level of oncogenic Kras expression inthe SC population. Furthermore, mRNA analysis for all Rasfamily members indicated that whereas Kras was expressed ata similar level in hair follicle bulge and matrix cells, Hras was sig-nificantly different between these populations (Fig. S3C); thus, theuse of Kras provided a fair comparison by expression levels.To confirm that tumorigenesis was initiated from the follicular

bulge and to better describe the phenotype generated by ex-pression of KrasG12D in hair follicle bulges, broad sections oftissue were isolated and analyzed with a variety of markers of cellidentity and proliferation. This approach enabled the visualiza-tion of both phenotypic initiation and progression. Temporalhistopathological examination revealed that the first event intumor formation is an expansion of the bulge and hair germ atthe onset of telogen to anagen transition (Fig. 2 and Figs. S3Band S4B). Within these expanded bulges, most cells were CD34+

and Sox9+, confirming their bulge origin (Figs. S3A, S4B, andS5E). Furthermore, these hyperplastic CD34+ bulges also ex-hibited YFP fluorescence in many cells, representing allelic re-combination from activated Cre expression (Fig. S4B). In fullanagen, hyperplasia extended downward and was most apparentin the middle ORS (Fig. S4D). At this point in phenotype pro-gression, CD34 was found to be aberrantly expressed throughoutall of the cells of the hyperplastic ORS (Fig. S4D), consistentwith previous reports suggesting that CD34 is up-regulated in theprogression of SCC (35). These observations are significant be-cause they suggest that SCs do not first travel to the matrix andbecome TA cells before generating hyperplasia, but that in factthe first phenotype observed is bulge expansion.Expansion of the bulge region and then the ORS led to either

a hyperplastic follicle or a keratin-filled cyst. Bulge and ORSexpansion correlated with expression of markers typical ofhyperproliferative epidermis such as keratin 6 (Krt6) and Ki67(Fig. S5 C and D). CAATT displacement protein (CDP),a marker of matrix TA cells (36), excluded a matrix origin forthese expanded hair follicles and keratin-filled cysts (Fig. S5F).Furthermore, none of the phenotypes generated seemed to beassociated with the IFE across 18 experiments and hundreds offollicles, suggesting that KrasG12D expression and initiation ofhyperplasias was confined to the SC niche of the hair follicle inK15CrePR; KrasG12D animals.

To determine the minimum cellular and molecular require-ment for SCC in the epidermis, K15CrePR; KrasG12D and Shh-CreER; KrasG12D mice were crossed to mice bearing floxed allelesof the p53 tumor suppressor (6, 23). K15CrePR; KrasG12D; p53ff

and ShhCreER; KrasG12D; p53ff mice were treated with mife-pristone or tamoxifen as described for the “one-hit” (KrasG12D-only) model described above (Fig. 3A). In addition to the hy-perplastic hair follicles and keratin-filled cysts found in one-hitanimals, introduction of these two genetic hits into the SCcompartment generated between one and six SCCs per animal(n = 9 of 9; Fig. 3 B and C). This occurred over a similar timecourse as the benign tumorigenesis observed in K15CrePR;KrasG12D mice. Notably, the IFE in these two-hit animals ap-peared normal. YFP was detected in the spindle-shaped cancercells invading the dermis (Fig. 3D), indicating that the SCC wasindeed generated by cells receiving the indicated genetic hits andnot by genetically wild-type cells stimulated by the surroundingenvironment. To confirm that the second hit had occurred asexpected, YFP+ cells isolated from SCCs showed a completelack of p53 expression (Fig. S6 A and B). This further demon-strated that the Cre allele used is highly efficient and that lineagetracing with the YFP allele is accurate for not only KrasG12D butalso for p53 gene deletion.To determine whether expression of KrasG12D and ablation of

p53 in the TA compartment of the hair follicle could also initiatetumorigenesis, ShhCreER; KrasG12D; p53ff mice were generatedand treated. No epidermal tumors of any kind were generatedwith this allelic combination (n = 6 of 6) (Fig. 3C). Additionally,to determine whether any changes in the developmental programof the hair follicle could be found in ShhCreER; KrasG12D; p53ff

mice, markers for proliferation and differentiation of the matrix

Fig. 2. Bulge expansion resulting from KrasG12D expression. (A and D)During telogen, control and KrasG12D-induced follicles appear similar. (B andE) At the onset of anagen, before full hair follicle formation but after SC exitfrom quiescence, bulge expansion is evident in KrasG12D-induced animals.(C and F) After formation of a full hair follicle, notable ORS hyperplasia canbe detected in most follicles.

White et al. PNAS | May 3, 2011 | vol. 108 | no. 18 | 7427

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 20

, 202

0

and its descendants were examined after a brief pulse of ta-moxifen. No difference was found in Ki67 (proliferation), CDP(matrix and IRS), Krt31 (precortex, cortex, cuticle), or Krt6(companion layer) staining (Fig. S6 C–G). Comparison of thehistopathology of K15CrePR; KrasG12D; p53ff and ShhCreER;KrasG12D; p53ff skin demonstrated that bulge cells or their im-mediate descendants have a high malignancy potential and canserve as cancer cells of origin, whereas their lineage-restrictedprogeny, the Shh-expressing TA cells, cannot (Fig. 3C).Examination of the epidermis outside of SCC formations in

K15CrePR; KrasG12D; p53ff and in regions of K15CrePR; KrasG12D

skin showed the presence of hair shafts within the dermis sur-rounded by spindle-like cells, instead of a distinct ORS (Fig. 3Cand Fig. S7 A and B). EMTs are commonly found in developingepithelial cancers and are thought to be essential for invasion ofthe epithelial cells into an underlying stroma and development ofcarcinoma (37). We examined a number of markers for EMT todetermine whether these spindle-like cells surrounding the dis-placed hair shafts could be epithelial cells undergoing EMT.Examination of the surrounding area around these cells dem-onstrated high Tenascin-C (TnC) (Fig. S7 A and B), an extra-

cellular matrix protein deposited during EMT (38–40). TnC isnot normally found in the dermis of the skin but was found ad-jacent to the ORS of anagen follicles and adjacent to the bulgeand infundibulum in telogen follicles (Fig. S7A).EMTs also exhibit decreased expression of epithelial-specific

markers, such as keratin 5 (Krt5) and E-cadherin (E-cad) (37, 41,42). Cells exhibiting high TnC also demonstrated very low or ab-sent Krt5 and E-cad, consistent with the suggestion that these cellsare undergoing EMT (Fig. S7A). Next, markers known to be up-regulated during EMT were examined. Vimentin (Vim), one suchmarker found at low levels in the dermal cells of the skin, wasfound at very high levels in cells exhibiting high TnC (Fig. S7B). Asimilar pattern of expression was detected in definable SCCsduring cancer progression (Fig. S8). Additionally, keratin 8 (Krt8),amarker of simple epithelium and of spindle cells of SCCs (43, 44),was only detected in cells of SCCs (Figs. S6B, S7B, and S8D). Thisresult indicates that Krt8 protein expression is found relatively lateduring SCC progression. To determine whether cells expressingEMT markers were also proliferating, Ki67 staining was used. Wedid find colocalization of Ki67 andmarkers of EMT in some cases,suggesting the possibility that these cells might be capable ofdriving formation of SCC.As expected, proliferation was abundantin SCCs (Figs. S7B and S8E). In summary, these data indicatedthat hair follicle SCs of K15CrePR; KrasG12D or K15-CrePR;KrasG12D; p53ff mice generate cells that undergo hyperpla-sia and EMT. Only K15CrePR; KrasG12D; p53ff transformedcells, however, were able to form SCCs.To determine the molecular mechanisms underlying the ability

of hair follicle SCs to act as cancer cells of origin in SCC, po-tential downstream signaling pathways were probed for activa-tion by immunostaining. The Mek/Erk, Akt, and p38 pathwayshave been shown to be directly downstream of Ras activation ina variety of settings, including oncogenic transformation (45, 46).In control skin, phosphorylated Erk (p-Erk) was found in basalcells of the IFE and in the bulges and ORS of hair follicles (Fig.4A and Fig. S9A). p-Erk was weakly expressed in bulges duringthe initial stages of hyperplasia, very strongly in hyperplasticanagen follicles, only weakly in the basal cells of cysts, and notwithin cells of the SCC (Fig. 4A and Fig. S9A). Aberrant ex-pression of p-Erk was not found in ShhCreER; KrasG12D; p53ff

induced matrix or IRS cells (Fig. S10B).Akt activity is sufficient to induce skin tumors in vivo (47–51),

and Akt activity is increased during SCC progression initiatedby chemical carcinogenesis (54). Akt pathway activation wasdetected by p-Akt in the suprabasal cells of epithelial cysts and

Fig. 3. Expression of oncogenic KrasG12D and p53 ablation in the SC nichegenerates SCC, whereas the same stimulus in TA cells does not. (A) Modeldepicting the strategy for KrasG12D expression combined with deletion of thetumor suppressor p53. (B) Macroscopic phenotype of an SCC derived fromthe hair follicle bulge in K15CrePR; KrasG12D; p53ff mice. Tumors can bedetected on the animal within 10 wk after mifepristone treatment. (C) K15-CrePR; KrasG12D; p53ff skin demonstrates hyperplastic sebaceous glands,hyperplastic hair follicles, epidermal cysts, transformed spindle cells invadingand populating the dermis, and large exophytic SCCs. No skin abnormalitieswere detected in ShhCreER; KrasG12D; p53ff animals. (D) Cancer cells invadingthe dermis were derived from YFP-labeled bulge K15CrePR; KrasG12D; p53ff;LSL-YFP cells. Asterisks indicate Krt5+ remnant hair follicles.

Fig. 4. Activation of signaling pathways downstream of Ras signalingduring tumorigenesis arising from the bulge. (A) Erk activity (p-Erk)appeared in hyperplastic follicles but not in follicular cysts or in SCCs. (B) Aktactivity (p-Akt) was detected in bulges, in non-ORS cells of hyperplastic fol-licles, cysts, and in remnant potions of hair follicles within SCCs (Inset). (C)Phosphorylated S6 (p-S6) was found in hyperplastic follicles, cysts and, ata lower level, throughout SCCs. (Magnification, 20×.)

7428 | www.pnas.org/cgi/doi/10.1073/pnas.1012670108 White et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 20

, 202

0

was found in remnant follicular structures within SCCs (Fig. 4B).Control skin showed p-Akt in only a few cells in the telogenbulge and in a subset of non-ORS cells near the bulge duringanagen (Fig. 4B and Fig. S9B). This p-Akt population was ex-panded in hyperplastic bulges but was not found in early anagenORS hyperplasias (Fig. S9B). Furthermore, p-Akt was neverdetected in the hair follicle TA population, in any context (Fig.S10B). To determine the status of further downstream signaling,a number of phosphorylated cell signaling proteins were exam-ined for activity. In control skin, p-S6 was detected in the pre-cortex and IRS of hair follicles but not in the SC or TA niche(Fig. 4C and Figs. S9C and S10B). p-S6 was also present in hy-perplastic cysts and at low levels in hyperplastic bulges, the ORSof hyperplastic anagen follicles, and SCCs of K15CrePR;KrasG12D; p53ff mice (Fig. 4C and Fig. S9C). p-Ikkα/β and p-Nfκb, downstream of Akt, were also detected in hyperplastic hairfollicles and cysts but were not found in control hair follicles(Fig. S10C). Taken together, it seems that mutant Kras expres-sion activates Erk during follicular hyperplasia and Akt/S6 sig-naling during cyst formation. The fact that the hair follicle TApopulation seemed refractory to Ras signaling pathway stimu-lation upon induction suggests a mechanism used by TA cells toprevent tumorigenesis in this system.

DiscussionThe results of this study have several important implications.First, these data reinforce previous studies suggesting that SCCcan arise from the hair follicle (3) and demonstrate that SCCsprobably arise from follicular SCs more often than from anyother type of cell. Second, the data shown here demonstrate thatcells of hair follicle bulge SC niche, or cells immediately exitingthe bulge, can serve as the origin of SCC. Third, Shh-expressingTA cells of the hair follicle cannot serve as SCC cells of origin, atleast not with the tumorigenic load provided by this model sys-tem. Fourth, although physiological expression of oncogenic Rascan drive formation of hyperplasia and EMT, a second hit seemsto be required for the development of SCC. Fifth, this studyprovides evidence that both the Erk and Akt pathways are ac-tivated during Ras-induced tumorigenesis from hair follicle SCs.Finally, hair follicle TA cells seem to be refractory to stimulationby the Ras pathway, suggesting a molecular explanation for theirinability to drive tumor formation in this model.The data presented here directly compare the relative suscep-

tibility of bulge cells and their more restricted progeny to serve ascancer cells of origin in SCC; these results are consistent with datafrom cancers of other tissues, such as the intestine, blood and brain(51–53). Similar to these other studies, SCs aremore likely to serveas cancer cells of origin than their more restricted progeny.Because TA cells are only found transiently, it is also possible

that they are not present for a sufficient time to initiate hyper-plasia before undergoing hair cycle-mediated apoptosis. Thiscould indicate that TA cells in general do not persist long enoughto serve as cancer cells of origin and would provide a simple ex-planation for the discrepancy between TA cells and SCs as tumorinitiators. However, we cannot detect any effect of oncogenicKrasexpression in the TA cells immediately after anagen initiation, atwhich time bulge cells have already initiated hyperplasia (Fig. 2and Figs. S3 and S4) and some weak activation of downstreamkinase pathways (Fig. S9). Additionally, Shh-expressing TA cellsare only found on one side of the matrix (Figs. S1C and S6A) andcontribute to the IRS but not to the companion cell layer and onlyrarely to the cortex and cuticle layers (Figs. S2 and S6). Therefore,

we cannot exclude the possibility that non–Shh-expressing matrixcells can act as cancer cells of origin.These data also do not exclude the possibility that Shh-

expressing TA cells of the hair follicle or other cells within theIFE could serve as cancer cells of origin in other contexts.Lapouge et al.(50) demonstrated that targeting of the same Krasallele to the IFE is capable of driving benign papilloma forma-tion. It is possible that hair follicle TA cells could generatecancer upon introduction of different genetic lesions or the ad-dition of environmental insults, such as inflammation. Alongthese lines, previous studies showed that transgenic expression ofoncogenic Hras in the differentiated compartment of the IFEwas sufficient to drive tumorigenesis in areas of high mechanicalstress or wounding (10, 54). Additionally, a recent study high-lighted this tumorigenic mechanism in a model of pancreaticcancer. In that model, certain cell populations were refractory totransformation until exposed to an inflammatory environmentgenerated by wounding (55). This raises the question of whethera tumorigenic load exists that can drive cancer formation in theShh-expressing hair follicle TA population and represents a di-rection for future experiments.

Materials and MethodsAnimals. Animals were acquired from Jackson Labs (K15CrePR and ShhCreER)or the National Cancer Institute Mouse Models of Human Cancers Consor-tium repository (LSL-KrasG12D and p53ff) and maintained under conditionsset forth by the Institutional Animal Care and Use Committee and the Ani-mal Research Committee (ARC) (University of California, Los Angeles).K15CrePR; KrasG12D and K15CrePR; KrasG12D; p53ff animals were treated byi.p. injections of mifepristone (10 mg/mL dissolved in sunflower seed oil,2 mg per day) immediately before the onset of the second adult hair cycle(roughly 10 wk postnatal, as measured by regrowth after shaving), for 3–5 d.ShhCreER; KrasG12D and ShhCreER ;KrasG12D; p53ff animals were treated withtamoxifen for 3–5 d (10 mg/mL, 2 mg per day) at the beginning of thesecond adult hair cycle (anagen), as measured by regrowth after shaving.Phenotypes shown from treated animals were produced 6–10 wk afteranagen. Because of Shh expression in tissues other than the skin, soft-tissuesarcomas did develop over longer periods of time and were similar to thosedescribed previously (62).

Immunostaining. Fresh frozen sections were cut at 7 μM for H&E and im-munofluorescence, except for those assayed for YFP with a GFP antibody,which were fixed in formalin overnight before embedding in optimal cut-ting temperature (OCT)embedding medium. Immunostaining was carriedout on frozen sections as previously described (63), except when assayed forYFP with a GFP antibody, which required antigen retrieval with citratebuffer for 30 min at 90 °C. The following antibodies were used: Krt5, Krt6,Krt14, Vim (Covance), Itgα6 (Becton Dickinson), Sox9 (Millipore), CDP, Kras2B(Santa Cruz), Krt8/18 (Troma-I, Developmental Studies Hybridoma Bank),TnC, Ki67, GFP (Abcam), CD34 (eBioscience), Krt31 (Progen), p-Ikkα/β, and p-Nfκb and E-cad (Cell Signaling). Immunohistochemistry was performed onformalin-fixed tissue as previously described (64) with the following anti-bodies: p-Erk, p-Akt, p-S6, p-mTor (Cell Signaling), and p14Arf (Bethyl).

ACKNOWLEDGMENTS. We acknowledge the technical support of OtarenAimiuwu and Hung Trinh. We thank the Division of Laboratory AnimalMedicine, University of California, Los Angeles (UCLA) and the TranslationalPathology Core Laboratory (Department of Pathology, UCLA) for theirsupport. We also thank K. Plath, S. Kurdistani, R. Kopan, and members ofthe Lowry lab for their input during preparation of the manuscript. A.C.W.was supported by training grants from California Institute for RegenerativeMedicine (CIRM) (TG2-01169) and National Institutes of Health (NIH) T32(CA009056). W.E.L. holds the Maria Rowena Ross Term Chair in Cell Biologyand Biochemistry, and the University of California Cancer Research Co-ordinating Committee (CRCC), Jonsson Cancer Center Foundation (JCCF),American Cancer Society (ACS), and the National Institutes of Health(NIAMS) (5R01AR057409) supported this work.

1. Alam M, Ratner D (2001) Cutaneous squamous-cell carcinoma. N Engl J Med 344:

975–983.2. van der Schroeff JG, Evers LM, Boot AJ, Bos JL (1990) Ras oncogene mutations in basal

cell carcinomas and squamous cell carcinomas of human skin. J Invest Dermatol 94:

423–425.

3. Spencer JM, Kahn SM, Jiang W, DeLeo VA, Weinstein IB (1995) Activated ras genes

occur in human actinic keratoses, premalignant precursors to squamous cell

carcinomas. Arch Dermatol 131:796–800.4. Corominas M, et al. (1991) ras activation in human tumors and in animal model

systems. Environ Health Perspect 93:19–25.

White et al. PNAS | May 3, 2011 | vol. 108 | no. 18 | 7429

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 20

, 202

0

5. Brown K, Strathdee D, Bryson S, Lambie W, Balmain A (1998) The malignant capacityof skin tumours induced by expression of a mutant H-ras transgene depends on thecell type targeted. Curr Biol 8:516–524.

6. Caulin C, et al. (2007) An inducible mouse model for skin cancer reveals distinct rolesfor gain- and loss-of-function p53 mutations. J Clin Invest 117:1893–1901.

7. Wang XJ, Greenhalgh DA, Roop DR (2000) Transgenic coexpression of v-Ha-ras andtransforming growth factor alpha increases epidermal hyperproliferation andtumorigenesis and predisposes to malignant conversion via endogenous c-Ha-rasactivation. Mol Carcinog 27:200–209.

8. Roop DR, et al. (1986) An activated Harvey ras oncogene produces benign tumours onmouse epidermal tissue. Nature 323:822–824.

9. Vitale-Cross L, Amornphimoltham P, Fisher G, Molinolo AA, Gutkind JS (2004)Conditional expression of K-ras in an epithelial compartment that includes the stemcells is sufficient to promote squamous cell carcinogenesis. Cancer Res 64:8804–8807.

10. Bailleul B, et al. (1990) Skin hyperkeratosis and papilloma formation in transgenicmice expressing a ras oncogene from a suprabasal keratin promoter. Cell 62:697–708.

11. Greenhalgh DA, et al. (1993) Induction of epidermal hyperplasia, hyperkeratosis, andpapillomas in transgenic mice by a targeted v-Ha-ras oncogene. Mol Carcinog 7:99–110.

12. Jensen KB, et al. (2009) Lrig1 expression defines a distinct multipotent stem cellpopulation in mammalian epidermis. Cell Stem Cell 4:427–439.

13. Jaks V, et al. (2008) Lgr5 marks cycling, yet long-lived, hair follicle stem cells. NatGenet 40:1291–1299.

14. Snippert HJ, et al. (2010) Lgr6 marks stem cells in the hair follicle that generate all celllineages of the skin. Science 327:1385–1389.

15. Greco V, et al. (2009) A two-step mechanism for stem cell activation during hairregeneration. Cell Stem Cell 4:155–169.

16. Horsley V, et al. (2006) Blimp1 defines a progenitor population that governs cellularinput to the sebaceous gland. Cell 126:597–609.

17. Morris RJ, et al. (2004) Capturing and profiling adult hair follicle stem cells. NatBiotechnol 22:411–417.

18. Ito M, et al. (2005) Stem cells in the hair follicle bulge contribute to wound repair butnot to homeostasis of the epidermis. Nat Med 11:1351–1354.

19. Harfe BD, et al. (2004) Evidence for an expansion-based temporal Shh gradient inspecifying vertebrate digit identities. Cell 118:517–528.

20. Levy V, Lindon C, Zheng Y, Harfe BD, Morgan BA (2007) Epidermal stem cells arisefrom the hair follicle after wounding. FASEB J 21:1358–1366.

21. Tuveson DA, et al. (2004) Endogenous oncogenic K-ras(G12D) stimulates proliferationand widespread neoplastic and developmental defects. Cancer Cell 5:375–387.

22. Marino S, Vooijs M, van Der Gulden H, Jonkers J, Berns A (2000) Induction ofmedulloblastomas in p53-null mutant mice by somatic inactivation of Rb in theexternal granular layer cells of the cerebellum. Genes Dev 14:994–1004.

23. Jonkers J, et al. (2001) Synergistic tumor suppressor activity of BRCA2 and p53 ina conditional mouse model for breast cancer. Nat Genet 29:418–425.

24. Jackson EL, et al. (2001) Analysis of lung tumor initiation and progression usingconditional expression of oncogenic K-ras. Genes Dev 15:3243–3248.

25. Wakabayashi Y, Mao JH, Brown K, Girardi M, Balmain A (2007) Promotion of Hras-induced squamous carcinomas by a polymorphic variant of the Patched gene in FVBmice. Nature 445:761–765.

26. Ise K, et al. (2000) Targeted deletion of the H-ras gene decreases tumor formation inmouse skin carcinogenesis. Oncogene 19:2951–2956.

27. Pierceall WE, Mukhopadhyay T, Goldberg LH, Ananthaswamy HN (1991) Mutations inthe p53 tumor suppressor gene in human cutaneous squamous cell carcinomas. MolCarcinog 4:445–449.

28. Lara MF, Paramio JM (2007) The Rb family connects with the Tp53 family in skincarcinogenesis. Mol Carcinog 46:618–623.

29. Martínez-Cruz AB, et al. (2008) Spontaneous squamous cell carcinoma induced by thesomatic inactivation of retinoblastoma and Trp53 tumor suppressors. Cancer Res 68:683–692.

30. Sutter C, et al. (1993) ras gene activation and aberrant expression of keratin K13 inultraviolet B radiation-induced epidermal neoplasias of mouse skin. Mol Carcinog 8:13–19.

31. Ziegler A, et al. (1994) Sunburn and p53 in the onset of skin cancer. Nature 372:773–776.

32. Mukhopadhyay A, Krishnaswami SR, Yu BD (2011) Activated Kras alters epidermalhomeostasis of mouse skin, resulting in redundant skin and defective hair cycling.J Invest Dermatol 131:311–319.

33. Srinivas S, et al. (2001) Cre reporter strains produced by targeted insertion of EYFPand ECFP into the ROSA26 locus. BMC Dev Biol 1:4.

34. Caulin C, et al. (2004) Inducible activation of oncogenic K-ras results in tumorformation in the oral cavity. Cancer Res 64:5054–5058.

35. Malanchi I, et al. (2008) Cutaneous cancer stem cell maintenance is dependent onbeta-catenin signalling. Nature 452:650–653.

36. Silva-Vargas V, et al. (2005) Beta-catenin and Hedgehog signal strength can specifynumber and location of hair follicles in adult epidermis without recruitment of bulgestem cells. Dev Cell 9:121–131.

37. Kalluri R, Weinberg RA (2009) The basics of epithelial-mesenchymal transition. J ClinInvest 119:1420–1428.

38. Dandachi N, et al. (2001) Co-expression of tenascin-C and vimentin in human breastcancer cells indicates phenotypic transdifferentiation during tumour progression:Correlation with histopathological parameters, hormone receptors, and onco-proteins. J Pathol 193:181–189.

39. Maschler S, Grunert S, Danielopol A, Beug H, Wirl G (2004) Enhanced tenascin-Cexpression and matrix deposition during Ras/TGF-beta-induced progression ofmammary tumor cells. Oncogene 23:3622–3633.

40. Ramos DM, Dang D, Sadler S (2009) The role of the integrin alpha v beta6 inregulating the epithelial to mesenchymal transition in oral cancer. Anticancer Res 29:125–130.

41. Zeisberg M, Neilson EG (2009) Biomarkers for epithelial-mesenchymal transitions.J Clin Invest 119:1429–1437.

42. Acloque H, Adams MS, Fishwick K, Bronner-Fraser M, Nieto MA (2009) Epithelial-mesenchymal transitions: The importance of changing cell state in development anddisease. J Clin Invest 119:1438–1449.

43. Larcher F, et al. (1992) Aberrant expression of the simple epithelial type II keratin 8 bymouse skin carcinomas but not papillomas. Mol Carcinog 6:112–121.

44. Diaz-Guerra M, et al. (1992) Expression of simple epithelial cytokeratins in mouseepidermal keratinocytes harboring Harvey ras gene alterations. Cancer Res 52:680–687.

45. Scholl FA, et al. (2007) Mek1/2 MAPK kinases are essential for Mammaliandevelopment, homeostasis, and Raf-induced hyperplasia. Dev Cell 12:615–629.

46. Scholl FA, et al. (2009) Selective role for Mek1 but not Mek2 in the induction ofepidermal neoplasia. Cancer Res 69:3772–3778.

47. Segrelles C, et al. (2006) Molecular determinants of Akt-induced keratinocytetransformation. Oncogene 25:1174–1185.

48. Segrelles C, et al. (2008) Constitutively active Akt induces ectodermal defects andimpaired bone morphogenetic protein signaling. Mol Biol Cell 19:137–149.

49. Segrelles C, et al. (2007) Deregulated activity of Akt in epithelial basal cells inducesspontaneous tumors and heightened sensitivity to skin carcinogenesis. Cancer Res 67:10879–10888.

50. Lapouge G, et al. Identifying the cellular origin of squamous skin tumors. Proc NatlAcad Sci USA, 10.1073/pnas.1012720108.

51. Barker N, et al. (2009) Crypt stem cells as the cells-of-origin of intestinal cancer.Nature 457:608–611.

52. Alcantara Llaguno S, et al. (2009) Malignant astrocytomas originate from neural stem/progenitor cells in a somatic tumor suppressor mouse model. Cancer Cell 15:45–56.

53. Sabnis AJ, et al. (2009) Oncogenic Kras initiates leukemia in hematopoietic stem cells.PLoS Biol 7:e59.

54. Arwert EN, et al. (2010) Tumor formation initiated by nondividing epidermal cells viaan inflammatory infiltrate. Proc Natl Acad Sci USA 107:19903–19908.

55. Gidekel Friedlander SY, et al. (2009) Context-dependent transformation of adultpancreatic cells by oncogenic K-Ras. Cancer Cell 16:379–389.

56. Kirsch DG, et al. (2007) A spatially and temporally restricted mouse model of softtissue sarcoma. Nat Med 13:992–997.

57. Blanpain C, Lowry WE, Geoghegan A, Polak L, Fuchs E (2004) Self-renewal,multipotency, and the existence of two cell populations within an epithelial stem cellniche. Cell 118:635–648.

58. Lowry WE, et al. (2005) Defining the impact of beta-catenin/Tcf transactivation onepithelial stem cells. Genes Dev 19:1596–1611.

7430 | www.pnas.org/cgi/doi/10.1073/pnas.1012670108 White et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 20

, 202

0