Identifying the cellular origin of squamous skin tumors

Abstract

Squamous cell carcinoma (SCC) is the second most frequent skin cancer. The cellular origin of SCC remains controversial. Here, we used mouse genetics to determine the epidermal cell lineages at the origin of SCC. Using mice conditionally expressing a constitutively active KRas mutant (G12D) and an inducible CRE recombinase in different epidermal lineages, we activated Ras signaling in different cellular compartments of the skin epidermis and determined from which epidermal compartments Ras activation induces squamous tumor formation. Expression of mutant KRas in hair follicle bulge stem cells (SCs) and their immediate progeny (hair germ and outer root sheath), but not in their transient amplifying matrix cells, led to benign squamous skin tumor (papilloma). Expression of KRas^G12D in interfollicular epidermis also led to papilloma formation, demonstrating that squamous tumor initiation is not restricted to the hair follicle lineages. Whereas no malignant tumor was observed after KRas^G12D expression alone, expression of KRas^G12D combined with the loss of p53 induced invasive SCC. Our studies demonstrate that different epidermal lineages including bulge SC are competent to initiate papilloma formation and that multiple genetic hits in the context of oncogenic KRas are required for the development of invasive SCC.

Squamous cell carcinoma (SCC) is one of two most frequent skin cancers and occurs most frequently in the sun-exposed regions of the skin and in immunocompromised patients. Approximately 250,000 patients per year develop skin SCC in the United States. Although skin SCC is often cured by surgical excision, ≈8% of patients with skin SCC relapse, and 5% present metastasis within 5 y. In patients with metastatic SCC, the prognosis is very poor, with only a 10–20% survival rate over 10 years (1). A better understanding of the early steps of SCC initiation is thus warranted.

The most extensively used mouse cancer model for skin SCC is a multistage chemically induced carcinogenesis (2). In the first step, called initiation, mice are treated with a low dose of the mutagen 9,10-dimethyl-1,2-benzanthracene (DMBA). In the second step, called promotion, mice are treated continuously with a drug that stimulates epidermal proliferation, such as 12-O-tetradecanoyl phorbol-13-acetate (TPA). During promotion, benign tumors (papillomas) arise, probably as a consequence of additional mutations, some of which will progress into invasive SCC. Interestingly, using this protocol, papillomas contain activating mutations in HRas gene, suggesting that this mutation confers a selective advantage to epithelial cells (3). However, not all papillomas contain mutations in the HRas gene, and some papillomas and SCC in both mouse and human present mutations in the KRas gene instead (4–6). Whereas Ras mutations seem to be an early step in skin cancer initiation, p53 mutations are associated with malignant progression (7, 8), although a recent report suggests that gain of p53 function increases tumor initiation and progression compared with p53 loss of function (9).

Conflicting results have been reported regarding the determination of which epidermal lineages give rise to squamous cutaneous cancers. The skin epidermis contains different compartments, including the hair follicle (HF) and the sebaceous gland (SG), as well as the interfollicular epidermis (IFE), which are maintained during homeostasis by different types of stem cells (SCs) (10). In the absence of injury, bulge SCs mediate HF regeneration (11–14), whereas the isthmus SCs mediate SG homeostasis (15–17). The maintenance of the IFE is ensured by the juxtaposition of many small epidermal proliferative units containing long-lived progenitors (18, 19). Upon wounding, bulge and isthmus SCs migrate upward and contribute to the repair of the IFE (11, 13, 17, 20). SCCs often present signs of squamous differentiation, as illustrated by the keratin pearls found in the center of the tumors, suggesting that SCCs may originate from the IFE (8). TPA administration stimulates papilloma formation even when TPA administration is started 1 y after the last DMBA treatment, suggesting that the initial mutation arises in long-lived SCs (21). There is some evidence that cells targeted by the benzo(a)pyrene or DMBA in the skin can be slow-cycling cells residing in the HF but also along the IFE (21–23). Dermabrasion techniques can selectively remove the IFE while leaving intact HFs, which then contribute to the repair of wounded skin. When TPA is applied after the repair of the epidermis, papilloma formation is reduced but the formation of SCC is not abolished, suggesting that initiating cells may reside in the IFE as well as in HFs (24). However, some studies suggested that oncogenic mutations could also occur in differentiated cells of the epidermis. For example, suprabasal expression of α6β4 integrins can increase the frequency of papillomas after DMBA-TPA treatment (25). In addition, transgenic mice expressing a mutated form of HRas under the promoter of K10, a marker of the committed suprabasal cells of the IFE, develop papillomas at sites of wounding (26). Transgenic mice expressing the same oncogene under the control of a truncated form of the K5 promoter, which is expressed preferentially in the HFs of adult mice, develop papillomas and SCCs (27). In addition, these studies used transgenic mice expressing Ras mutant constitutively from embryonic development to adulthood and at supraphysiological level, leaving open the question of which cells of the skin epidermis are competent to give rise to SCCs upon mutated Ras expression at a physiological level in adult mice.

To determine which epidermal compartments are competent to initiate squamous tumors, we used mouse genetics to assess the incidence of skin cancer after the expression of oncogenic KRas at a physiological level in different epidermal compartments. To avoid the drawback of transgenic approaches that result in the random integration of several copies of the oncogene into differently accessible chromatin regions, we used a genetic mouse model to specifically activate a constitutively active mutant of KRas (G12D) knocked in into its own locus (KRas^LSL-G12D). Expression of this mutant KRas induced a broad variety of Ras-induced neoplasia (28), including oral (29) and skin (9) squamous tumors. The oncogene is only expressed when the Lox-STOP-Lox (LSL) cassette is removed by a CRE recombinase. A short course of tamoxifen (TAM) administration in mice expressing CREER leads to a mosaic expression of KRas^G12D in the epidermal compartments expressing the CREER, mimicking the natural occurrence of sporadic tumors.

Results

KRas^G12D Expression in Bulge SCs and Their HF Progeny Induces Papilloma Formation.

We first induced the expression of KRas^G12D in bulge SCs and their progeny using mice expressing inducible CRE under the regulatory region of HF-specific promoter [K15-CREPR (12) and K19CREER (30)]. We did observe some leakiness in the K15CREPR/RosaYFP mice, visible by the presence of patches of YFP-expressing cells in the absence of RU486 (Fig. S1 A and B), whereas no leakiness was detected in the K19CREER/RosaYFP mice (Fig. 1 A and B) (31). Administration of RU486 in K15CREPR/Rosa-YFP mice induced the expression of YFP mostly in bulge SCs, hair germ (HG), and outer root sheath (ORS) cells but also to a lesser extent in the isthmus, infundibulum (the region that connects the bulge to the IFE) and the IFE (Fig. S1 C–G), as previously described (12). TAM administration in K19CREER/RosaYFP mice resulted in the labeling of bulge SCs and their HF progeny (Fig. 1 C–G). After TAM administration in K19CREER/KRas^G12D mice, recombination of KRas^G12D allele was detected in bulge SCs (α6HCD34H) but not in a6HCD34⁻ cells (Fig. S2A), suggesting that the recombination of KRas allele is restricted to the same cells labeled with the RosaYFP. In the K19CREER/KRas^LSL-G12D, the expression of the KRas^G12D oncogene in bulge SCs and their progeny led to a transient increase in bulge SC proliferation (Fig. S2B) and the enlargement of the majority of SGs 1 mo after TAM administration (Fig. S2C). These SGs became hyperproliferative, as shown by the increased number of Ki67-positive cells and the expression of K6 (Figs. S1–S3), although they were still able to undergo terminal differentiation, as shown by the presence of sebum (Figs. S1 and S3). The defect in SG homeostasis also led to the formation of cystic structures (Figs. S1 and S3). The SG hyperplasia observed here is reminiscent of the phenotype observed after c-Myc (32) and ΔN-Lef1 overexpression (33) or Blimp1 deletion (34) in the skin epidermis. Occasionally, the IFE was also abnormal after KRas^G12D expression, as shown by the hyperthickening of the IFE located at the orifice of some abnormal HF with hypertrophic SGs (Figs. S1 and S3). In addition to sebaceous cysts, KRas^G12D expression in bulge SCs and their progeny also led to the enlargement and abnormal differentiation of HF and to cystic structures negative for Oil-Red-O staining (Figs. S1 and S3). Altogether, these data show that KRas^G12D expression in bulge SCs and their HF progeny led to frequent histological abnormalities, indicating that the majority of HFs expressing mutant KRas developed pathologic features including SG hyperplasia and dysplasia, as well as hyperplasia and aberrant differentiation of the lower HF, some of which degenerated into cystic structures.

Fig. 1.

Targeting KRas^G12D expression in bulge SCs and their progeny induces papilloma formation. (A) Bulge SCs and their HF progeny targeted by K19 promoter are highlighted in dark blue. (B) YFP immunostaining in the absence of TAM administration. (C–G) Lineage tracings of K19CREER/RYFP mice are shown in the bulge (C), HG (D), and ORS (E and F) of the back skin and in the bulge of the lips (G) 1 wk after 10 mg TAM administration. (H–M) Papilloma formation in K19CREER/KRas^LSL-G12D mice. (H and K) Macroscopic pictures of tumors seen in the K19CREER/KRas^LSL-G12D lips (H), face, and back skin (K) 4 mo after TAM administration. Expression of K5 (I and L) and of K1 (J and M) in papilloma from lips and back skin arising from K19CREER/KRas^LSL-G12D mice. Bu, bulge; IRS, inner root sheath; Mx, matrix. (Scale bars, 50 μm.)

To investigate why KRas expression in HF cells induces SG alterations, we performed lineage tracing together with KRas expression (K19CREER/KRas^LSL-G12D/RosaYFP mice). Already at 2 wk after TAM administration, more YFP-positive cells were seen in the SG progenitors compared with wild-type mice (Fig. S4A). At 4 wk after TAM administration, some YFP-positive cells were detected in hyperplastic SG and sometimes even in the IFE of KRas/YFP mice, whereas no YFP-positive were cells were observed in the SG or the IFE of control mice (K19CREER/RosaYFP mice) (Fig. S4B). These findings suggested that KRas^G12D expression induces migration of bulge SCs toward the SG and the IFE. However, because there is not a perfect correlation between the hyperplastic phenotype and the YFP expression, our data cannot rule out that KRas expression in HF cells promotes epidermal proliferation by a noncellular autonomous mechanism, and thereby promotes tumorigenesis as it occurs after Notch1 deletion in the skin epidermis (35).

Tumors arose in the lips (57% of the animals treated), face (35%), and back skin (14%) of K15CREPR/KRas^LSL-G12D mice (n = 14) 2–4 mo after RU486 administration (Fig. S1). Similarly, tumors also arose with similar latency and frequency in the lips (78% of the mice treated) and back skin (33%) of K19CREER/KRas^LSL-G12D mice (n = 9) after TAM administration (Fig. 1 H and K). Consistent with the leakiness observed in K15CREPR/RosaYFP mice, all untreated K15CREPR/KRas^LSL-G12D mice eventually developed some papillomas, although the latency was increased in untreated mice (>6 mo) (Fig. S1S). These tumors possessed all of the histological characteristics of benign papillomas, as demonstrated by the expression of K5 in the basal layers and K1 in the suprabasal layers (Fig. 1 I–M and Fig. S1 M–R), indicative of squamous differentiation. No sign of malignant transformation was observed in these mice up to 4 mo after KRas^G12D expression. Altogether, these data indicate that bulge SCs and their HF progeny are competent to develop benign papillomas upon KRas^G12D expression, but other genetic events are required to induce the development of invasive carcinomas.

HF Transit Amplifying Cells Are Not Competent to Initiate Papilloma Formation upon KRas^G12D Expression.

During HF regeneration, bulge SCs are activated, proliferate, and give rise to transit-amplifying (TA) cells, which differentiate into the hair shaft and its envelope (36). To determine whether KRas^G12D expression in HF TA compartment is sufficient to transform these cells and induce squamous tumor formation, we administered TAM to ShhCREER/KRas^LSL-G12D mice at day 28 when HFs were in full anagen (Fig. 2A). No leakiness was detected in ShhCREER/RosaYFP mice (Fig. 2B). Administration of TAM to ShhCREER/RosaYFP mice preferentially labeled one side of the matrix. These labeled cells constitute a subset of matrix cells that differentiate into the different HF lineages (Fig. 2 C–G), showing that ShhCREER (37) activity is indeed restricted to the HF TA cells, as we previously described in the HF of the tail (31). Few marked YFP cells were still observed in the companion layer of the club hair 1 mo after TAM administration (Fig. S5A). RT-PCR analysis of FACS-isolated YFP cells from ShhCREER/RosaYFP demonstrated that KRas is expressed at a similar level in Shh-derived cells, in bulge SC, and in cells of the IFE and the infundibulum (Fig. S5B). Despite the similar level of KRas expression in matrix cells, no macroscopic tumors or microscopic HF or IFE defects (Fig. 2 H–J and Fig. S5 C and D) were observed in ShhCREER/KRas^LSL-G12D mice 4 mo after TAM administration (n = 7), suggesting that KRas^G12D expression in HF TA cells is not sufficient to induce long-term renewal potential and initiate tumor formation.

Fig. 2.

Transit amplifying HF matrix cells do not give rise to papilloma upon KRas^G12D expression. (A) Matrix cells and their progeny targeted by Shh promoter are highlighted in blue. (B) YFP immunostaining in the absence of TAM administration. (C–G) Lineage tracing in ShhCREER/RYFP mice showed the presence of YFP cells in the matrix cells of the lips (C), HF (D), matrix (Mx), and hair shaft (HS) (E and F); inner root sheath (IRS) of the back skin (G) shows that ShhCREER marked TA matrix cells and their HF progeny 1 wk after 2.5 mg TAM administration. (D–H) Comparison between wild-type and ShhCREER/KRas^LSL-G12D back skin 4 mo after 2.5 mg TAM administration. Hematoxylin-eosin staining (H) and expression of K5 (I) and K6 (J), (H) Oil-red-O staining in back skin, show no epidermal defect or tumor formation in mice expressing KRas^G12D in matrix cells. Bu, bulge; Mx, matrix. (Scale bars, 50 μm.)

KRas^G12D Expression in the IFE Induces Papilloma Formation.

Transgenic mice expressing a mutated form of HRas under the promoter of K10, a keratin expressed in the suprabasal cells of the IFE, induced hyperkeratosis and papilloma formation in response to mechanical stress (26). To determine whether physiological levels of KRas^G12D expression in the IFE can induce benign tumor formation during adult homeostasis, we generated transgenic mice expressing the TAM-inducible CREER in the suprabasal cells of the IFE. To this end, we generated transgenic mice expressing the CREERT2 under the human involucrin promoter (38) (InvCREER) (Fig. 3A). In the absence of TAM administration, no YFP cells was observed in InvCREER/RosaYFP, demonstrating the absence of leakiness of the InvCREER (Fig. 3B). Administration of TAM to InvCREER/RosaYFP mice resulted in the labeling of suprabasal cells of the IFE of the back skin and the lips (Fig. 3 C–E), although some basal cells (Fig. 3D) were also labeled in all three founders, consistent with the previous demonstration that ≈10% of basal cells already express markers of suprabasal cells (39). InvCREER also labeled some cells of the companion layer of the HF, as well as suprabasal cells of the infundibulum (Fig. S6A). After TAM administration in InvCREER/KRas^LSL-G12D mice, recombination of KRas^G12D allele was detected in IFE cells (α6HCD34⁻) but not in bulge SCs (α6HCD34H) (Fig. S6B). TAM administration in InvCREER/KRas^LSL-G12D induced hyperthickening of the IFE (Fig. S6 C–E), with an increase in the number of layers expressing K5 (Fig. S6E), an increase in the proportion of cells expressing Ki67, and the expression of K6 (Fig. S6C). These data indicate that KRas^G12D expression in the IFE induced hyperproliferation and hyperplasia of the IFE. InvCREER/KRas^LSL-G12D mice treated with TAM developed tumors in the lips (100% of the mice), face (40%), and back skin (80%) 2–4 mo after TAM administration (n = 5) (Fig. 3 F and I). Microscopic examination of these tumors demonstrated that they presented the hallmarks of benign papilloma with signs of squamous differentiation, as demonstrated by the expression of K1 and K5 (Fig. 3 G–K), and these tumors were indistinguishable from the papillomas obtained in K15CREPR/KRas^LSL-G12D and K19CREER/KRas^LSL-G12D mice. It has been recently shown that benign papillomas and invasive SCCs contained cancer SCs expressing CD34, a marker of bulge SCs, suggesting that squamous tumors may arise from bulge SCs (40). Interestingly, papillomas arising from the IFE of InvCREER/KRas^LSL-G12D mice also express CD34 and K17, another follicular marker (Fig. 3 L–O), indicating that expression of HF markers by tumor cells does not necessarily reflect their cellular origin. None of these mice developed invasive carcinoma within the 4 mo after TAM administration. These results demonstrate that different cell lineages of the skin epidermis, including HF SCs and their progeny, as well as cells of the IFE, are competent to develop benign papillomas upon physiological KRas^G12D expression with similar latency and without the need of stimulation caused by injury.

Fig. 3.

KRas^G12D expression in the IFE induces papilloma formation. (A) Interfollicular suprabasal cells and rare basal cells targeted by involucrin promoter are highlighted in dark pink. (B) YFP immunostaining in the absence of TAM administration. (C–E) Lineage tracing of InvCREER/RosaYFP back skin mice 1 wk after 5 mg TAM administration. Costaining of YFP and β4, K5, and K1 shows that InvCREER mostly targets the suprabasal differentiated cells of the IFE, as well as some basal cells (D, arrow). (F–K) Papilloma formation in InvCREER/KRas^LSL-G12D mice. (F and I) Macroscopic picture of tumors in the lips (F) and back skin (I) of InvCREER/KRas^LSL-G12D mice 4 mo after TAM administration. Expression of K5 (G and J) and K1 (H and K) in papilloma from the back skin and lips of InvCREER/KRas^LSL-G12D mice. (L–O) Follicular marker expression in papilloma arising from InvCREER/KRas^LSL-G12D mice. Expression of K17 (L and M), a follicular marker, and CD34 (N and O), a bulge SC marker in wild-type epidermis (L and N) and papilloma (M and O) from InvCREER/KRas^LSL-G12D mice, indicating that tumor differentiation does not reflect their cellular origin. (Scale bars, 50 μm.)

Combined Expression of KRas^G12D and p53 Deletion in Epidermal Cells Is Necessary to Initiate SCC.

Because no malignant tumor was observed in mice expressing KRas^G12D in any of the epidermal compartments tested and because p53 gene dosage has been reported to control malignant progression in DMBA/TPA-induced skin tumors (41), we tested whether p53 deletion can promote the formation of invasive SCC in skin expressing KRas^G12D. To this end, we first determined whether KRas^G12D expression and p53 deletion in the epidermis are sufficient to initiate invasive SCC. We administered 1 mg TAM to K14CREER/KRas^LSL-G12D/p53^fl/fl mice (Fig. 4A). At this dose, TAM administration in K14CREER/RosaYFP induced the expression of YFP in a mosaic manner in all epidermal compartments of the skin epidermis (31). TAM administration induced KRas^G12D expression and p53 deletion in K14CREER/KRas^LSL-G12D/p53^fl/fl mice (Fig. S7). Within less than 2 mo they developed rapidly growing ulcerative lesions in the back skin (Fig. 4B) (n = 12). These ulcerative lesions represented invasive SCCs characterized by a fibroblast-like shape, expressed K5 but not K1 (Fig. 4 C–G), displayed disruption of basal lamina (Fig. 4H), and expressed markers of epithelial to mesenchymal transition (Fig. 4I). In addition to the macroscopic visible ulcerative lesions, many microscopic and small invasive SCCs were also observed in the dermis of the back skin (Fig. 4 E–I). These results demonstrate that, in the context of oncogenic KRas, at least two genetic hits (gain of Ras and loss of p53 function) are required for the development of invasive carcinoma.

Fig. 4.

Combined KRas^G12D expression and p53 deletion in the skin epidermis induces SCC. (A) IFE and HF basal cells targeted by K14 promoter at clonal induction are highlighted in dark green. (B) Macroscopic picture of ulcerative back skin lesions from K14CREER/KRas^LSL-G12D/p53^fl/fl mouse 1 mo after TAM administration. (C and D) Immunostaining of K5 (C) and K1 (D) in K14CREER/KRas^LSL-G12D/p53^fl/fl back skin tumors show the absence of K1 expression in these tumors. (E–I) Comparison between wild-type back skin and SCC from K14CREER/KRas^LSL-G12D/p53^fl/fl mice. Hematoxylin-eosin staining (E), expression of K5 (F), K1 (G), laminin5, a basal lamina marker (H), and vimentin, a mesenchymal marker (I), showing that these tumors present all histological and biochemical characteristics of invasive SCC. (Scale bars, 50 μm.)

We then determined more specifically whether the expression of KRas^G12D combined with the p53 deletion in HF bulge SCs and their progeny also led to invasive SCC. To this end, we administered TAM to K19CREER/KRas^LSL-G12D/p53^fl/fl mice, which specifically induce KRas^G12D expression and p53 deletion in bulge SCs and their progeny (Fig. S7). However, these treated mice died within 2 mo after TAM administration from intestinal cancers before developing skin tumors (n = 8). To circumvent this issue, we grafted the back skin of K19CREER/KRas^LSL-G12D/p53^fl/fl mice onto the back skin of nude mice one month after TAM administration. All mice grafted with the skin of TAM-treated K19CREER/KRas^LSL-G12D/p53^fl/fl mice developed many skin tumors, including ulcerative lesions, within 2 mo after TAM administration (Fig. 5A) (n = 10 grafts). Microscopic examination revealed that the majority of these tumors were differentiated benign papillomas, as shown by K1 expression (Fig. 5 C and F), but some of them presented all of the characteristics of invasive SCC (Fig. 5 D–H), which were indistinguishable from the SCC observed in K14CREER/KRas^LSL-G12D/p53^fl/fl mice. To determine whether the expression of KRas^G12D and the loss of p53 in matrix TA cells can induce skin tumor formation, we administered TAM to ShhCREER/KRas^LSL-G12D/p53^fl/fl mice. As in ShhCREER/KRas^LSL-G12D mice, no macroscopic or microscopic tumors (Fig. S8) were observed in ShhCREER/KRas^LSL-G12D/p53^fl/fl treated mice up to 4 mo after TAM administration (n = 10). These data clearly showed that the combined expression of KRas^G12D and p53 deletion in bulge SCs and/or their HF progeny are required to induce full-blown invasive skin SCC, whereas TA matrix cells were not competent to initiate benign or malignant tumors.

Fig. 5.

KRas^G12D expression and p53 deletion in bulge SCs and their HF progeny induce SCC. (A) Macroscopic picture of tumors arising in K19CREER/KRas^LSL-G12D/p53^fl/fl back skin grafted onto nude mouse. Immunostaining of K5 (B) and K1 (C) in K19CREER/KRas^LSL-G12D/p53^fl/fl carcinoma. (D–H) Comparison between wild-type back skin, papilloma, and carcinoma arising from grafted K19CREER/KRas^LSL-G12D/p53^fl/fl back skin. Hematoxylin-eosin staining (D) and expression of K5 (E), K1 (F), laminin5 (G), and vimentin (H) in wild-type back skin and in papilloma and SCC from K19CREER/KRas^LSL-G12D/p53^fl/fl mice. (Scale bars, 50 μm.)

Discussion

For most cancers, the precise identification of tumor-initiating cells remains elusive. We have recently demonstrated that upon oncogenic smoothened expression, basal cell carcinoma (BCC) arose most frequently from the IFE, whereas bulge SCs and their HF progeny were not competent to initiate BCC formation upon smoothened expression (31). In sharp contrast, KRas^G12D oncogene expression in bulge SCs and their HF progeny gave rise to benign papillomas, and KRas^G12D expression combined with p53 deletion in these cells led to invasive SCC. Because K15 and K19 inducible CRE lines also targeted cells located within the HG and the lower ORS, we cannot exclude the possibility that some short-lived HF progenitors could be as potent as bulge SCs in initiating squamous tumors. In addition to bulge SCs, cells of the IFE are also competent to give rise to benign squamous tumors upon physiological expression of KRas^G12D oncogene without the need of exogenous inflammation or stimulation of tissue turnover by chemical abrasion or wounding. These data show that in contrast to the restricted epidermal lineages competent to initiate BCC upon smoothened expression (31), multiple epidermal compartments of the epidermis, but not matrix HF TA cells nor their hair shaft progeny, are competent to initiate squamous tumors upon KRas^G12D expression. The absence of tumor formation upon KRas^G12D expression in HF TA cells could be related to the transient period during which matrix cells actively proliferate. Indeed, it takes approximately 1 mo to detect the first microscopic lesions after KRas^G12D expression in HF or IFE progenitors, and after 1 mo the lower part of the HF has degenerated during catagen stage, and only rare Shh-derived cells are still present within the HF, as terminally differentiated cells.

Papillomas arising from bulge SCs targeted by KRas^G12D expressed K1, whereas bulge SCs and their HF progeny do not express K1 during physiological conditions. Similarly, papilloma arising from KRas^G12D expression in the IFE expressed CD34, like papillomas from DMBA/TPA-treated mice (40), although in normal skin CD34 is specifically expressed by bulge SCs but not IFE cells (42, 43). These results indicated that the expression of differentiation markers by tumor cells does not necessarily reflect the cellular origin of squamous tumors. Similarly, oncogenic smoothened expression induced the expression of follicular markers in interfollicular progenitors during BCC initiation (31).

Whereas oncogenic smoothened expression drove the majority of targeted cells to invasive BCC (31), only a minor fraction of KRas^G12D targeted cells formed squamous tumors, suggesting that papilloma initiation could require other additional genetic and/or epigenetic modifications in addition to oncogenic Ras expression. Interestingly, simultaneous deletion of p53 and KRas^G12D expression can initiate the formation of invasive SCC without first forming papilloma, suggesting that p53 loss might control SCC initiation, in addition to promoting tumor progression (41).

The mouse model of invasive SCC resulting from the expression of oncogenic KRas and p53 deletion in HF bulge SCs is likely to be relevant for human SCC because the same oncogenes (Ras and p53) are also implicated in the pathogenesis of human SCC (7, 8), and the histology of SCC found in our mouse model recapitulates the histology of human SCC. In addition, our findings, showing that squamous tumors arise from HF bulge SCs as well as from nonhairy epidermis, recapitulate the spectrum of locations of squamous tumors found in humans, such as hairy and nonhairy skin, oral cavity, head and neck, and esophagus.

Methods

Mice.

K14CREER (44), ShhCREER (37), K15CREPR (12), K19CREER (30), Rosa-YFP (45), KRas^LSL-G12D (28), and p53^fl/fl mice (46) have been previously described. Generation of InvCREER mice and induction of CRE-mediated recombination are described in SI Methods.

Histology, Immunostaining, and Imaging.

These were performed as previously described (31) and are detailed in SI Methods.

Skin Graft Protocol.

Mice were anesthetized with a mixture of 5% xylazine and 10% ketamine in PBS. Back skin of 2-mo-old TAM-treated mice were removed and grafted onto the back skin of nude mice. Skin grafts were monitored for carcinoma formation.

Isolation of Keratinocytes, DNA and RNA Extraction, Real-Time RT-PCR, and PCR Analysis of Cre-Mediated Recombination.

These are described in SI Methods.