Meis1 Regulates Epidermal Stem Cells and Is Required for Skin Tumorigenesis

Previous studies have shown that Meis1 plays an important role in blood development and vascular homeostasis, and can induce blood cancers, such as leukemia. However, its role in epithelia remains largely unknown. Here, we uncover two roles for Meis1 in the epidermis: as a critical regulator of epidermal homeostasis in normal tissues and as a proto-oncogenic factor in neoplastic tissues. In normal epidermis, we show that Meis1 is predominantly expressed in the bulge region of the hair follicles where multipotent adult stem cells reside, and that the number of these stem cells is reduced when Meis1 is deleted in the epidermal tissue of mice. Mice with epidermal deletion of Meis1 developed significantly fewer DMBA/TPA-induced benign and malignant tumors compared with wild-type mice, suggesting that Meis1 plays a role in both tumor development and malignant progression. This is consistent with the observation that Meis1 expression increases as tumors progress from benign papillomas to malignant carcinomas. Interestingly, we found that Meis1 localization was altered to neoplasia development. Instead of being localized to the stem cell region, Meis1 is localized to more differentiated cells in tumor tissues. These findings suggest that, during the transformation from normal to neoplastic tissues, a functional switch occurs in Meis1.


Introduction
Homeobox genes such as Meis1 (myeloid ecotropic insertion site 1) are known to play a crucial role in normal development and tumor development. Meis1 was first identified as a common viral integration site in myeloid leukemic cells of BXH-2 mice [1]. Meis1 expression is frequently up-regulated in primary acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) [2]. Germline targeted knockout of Meis1 results in embryonic lethality at day 14.5 as a result of multiple hematopoietic and vascular defects [3,4].
Meis1 encodes a TALE family homeodomain transcription factor that forms a heterodimeric DNA binding complex with Pbx. The interaction with Pbx1 enables Meis1 to interact with additional Hox transcription factors such as HOX-9 and HOX-10 paralog proteins. These interactions in effect functionally incorporate Meis1 into a range of Hox-dependent developmental programs [5], including vertebrate hindbrain development and limb morphogenesis [6,7], maintenance of an undifferentiated state, and expansion of retinal progenitor cells [8,9], and olfactory and thymic epithelial cells [10,11].
While a number of studies have suggested that Meis1 has a functional role in epithelial tissues, its functions in the epidermis and in skin carcinogenesis remain poorly understood. Studies of Meis1 in epithelial tumor development have been limited to correlative studies based on gene expression and clinical outcome. As in leukemia, gene expression studies in lung adenocarcinomas [12], neuroblastomas [13,14,15], ovarian carcinomas [16], and nephroblastomas [17] have shown that the expression of Meis1 is elevated in tumor tissues, suggestive of an oncogenic role. In contrast, gene expression studies in prostate cancer have shown that decreased expression of Meis1 is correlated with poor prognosis, suggesting that it may have tumor suppression activity in prostate cancer development [18].
To gain insight into the role of Meis1 in the epidermis, we used a tamoxifen-inducible, epithelial-specific Meis1 knockout model in combination with a Meis1-EGFP reporter and cell marker colocalization studies to analyze its functions in both normal epidermal development and in skin carcinogenesis, using a twostage carcinogenesis mouse model.
The pathology of the two-stage chemically induced skin carcinogenesis mouse model is almost identical to the development of human skin cancers and thus offers an ideal model to study skin cancer initiation and growth [19,20]. In the first step of the chemically induced carcinogenesis protocol, mice are treated with a low dose of the mutagen 7,12-dimethylbenz(a)anthracene (DMBA) to initiate tumor development. This first chemical treatment step leads to ''tumor initiation''. In the second step, mice are treated continuously with 12-O-tetradecanoylphorbol-13acetate (TPA) to stimulate epidermal tumor proliferation. This second chemical treatment step influences ''tumor promotion''. During tumor promotion, benign tumors, known as papillomas, are thought to arise by additional mutations caused by the TPA chemical treatment. After prolonged treatment (,20 weeks), some of the papillomas will progress into carcinogenic tumors, such as squamous cell carcinomas (SCC). The role of various genes and cell-signaling pathways involved in skin tumor development can be explored in this two-stage skin carcinogenesis model by the use of genetically engineered mouse models [21][22][23][24][25][26][27][28][29][30][31].
In the present study, we reveal for the first time Meis1's crucial function in maintaining the epidermal stem cells that act to maintain homeostasis in the epidermis. Furthermore, we present findings that demonstrate Meis1's oncogenic role in epithelial tumor development. Specifically, we present findings that suggest its role in tumor development and in malignant conversion. Furthermore, our marker studies collectively indicate that Meis1 has distinct molecular mechanisms in normal and tumorigenic tissues. Finally, we present a model for Meis1's function in normal and neoplastic epidermis.

Meis1 is expressed in stem cells of the hair follicle bulge region in normal epidermis
The epidermis is comprised of a stratified squamous epithelium and an underlying dermis consisting of matrix-rich connective tissue. Furthermore, epidermal stem cells are found at the basal layer of the epidermis and are involved in maintaining proper epidermal architecture and function throughout an organism's lifespan. To investigate Meis1's role in the epidermis, we first examined Meis1 expression in normal skin tissue. We used a Meis1-EGFP reporter [32] in normal wild-type mice to determine where, if at all, Meis1 is expressed in the epidermis. The Meis1-EGFP reporter was made from an artificial chromosome (BAC) transgene (RP23-306E8) corresponding to ,80 kb upstream and ,30 kb downstream of the mouse Meis1 gene into which an EGFP cDNA was inserted just 59 of the Meis1 translation start site, thereby ensuring that no additional exogenous expression of Meis1 was introduced from expression of the transgene reporter. Immunofluorescence analysis of the epidermis of 8-week-old Meis1-EGFP mice showed that Meis1-EGFP was localized to the bulge region of hair follicles (Fig. 1A-1D). The bulge region, where multipotent stem cells are activated upon wounding, provides progenitor cells for regeneration and repair of the epidermis or at the start of a new cycle of hair development. To investigate further Meis1's role in the bulge region, we carried out co-immunostaining with Meis1-EGFP and epidermal stem cell markers: CD34, keratin 15 (Krt15/K15), and the basal cell marker b4-integrin. We found that Meis1-EGFP partially overlapped with the epidermal stem cell markers CD34 and K15 in the bulge region ( Fig. 1I-1P) and that Meis1-EGFP was not co-expressed with b4-integrin in the basal layer cells of the epidermis (Fig. 1Q-1T).
To investigate further Meis1 expression in the bulge region, we performed BrdU-chase experiments to label quiescent stem cells of the epidermis in Meis1-EGFP mice. We found that Meis1-EGFP signals were co-expressed with BrdU-LRCs (Label Retaining Cells) (Fig. S2C-S2F). We also carried out co-immunostaining with Meis1-EGFP and a differentiation marker, keratin 10 (Krt10/K10) (Fig. 1E-1H). We did not detect Meis1-EGFP signal overlapping with K10 in the epidermis. Taken together, these results demonstrate that Meis1 is predominantly expressed in the epidermal stem cells of the hair follicle bulge region and suggest that Meis1 is involved in stem cell function or maintenance in the epidermis.

Meis1 is required for epidermal stem cell maintenance
To investigate further the hypothesis that Meis1 plays a role in stem cell maintenance, we generated mice harboring conditional alleles of Meis1 (Meis1 fl ), in which exon 8 of the Meis1 gene encoding the Meis1 homeodomain (DNA-binding site) is flanked by loxP sites. Next, we crossed the Meis1 fl conditional-knockout mice with mice carrying a K14CreER allele [33], which is specifically expressed in the epidermis, to generate K14Cre ER -Meis1 fl/fl progeny in which Meis1 can be rendered non-functional in the skin upon induction with tamoxifen. One subcutaneous injection of tamoxifen into K14Cre ER -Meis1 fl/fl mice was sufficient to disrupt the floxed Meis1 locus in dorsal back skin and tail (Fig.  S1A). After confirming that the knockout had indeed been induced in our mouse model, we analyzed the dorsal back skin of K14Cre ER -Meis1 fl/fl mice (n = 5) at one week after tamoxifen induction and compared it with the dorsal back skin of treated Meis1 fl/fl (n = 3) littermates by HE (hematoxylin and eosin) analysis. We observed that K14Cre ER -Meis1 fl/fl mice had massively thickened epithelium in comparison to their Meis1 fl/fl littermates ( Fig. 2A and 2B), suggesting that, in the absence of Meis1, K14Cre ER -Meis1 fl/fl mice have increased cell proliferation. Indeed, immunofluorescence analysis with the cell proliferation marker Ki-67 showed an increased number of Ki-67-positive cells in K14Cre ER -Meis1 fl/fl mice, mainly in the basal layer of the epidermis (P = 0.0004, t-test) (Fig. 2C, 2D, and 2G), indicating that Meis1 deficiency induces hyperproliferation of the epidermis.
To investigate further the requirement for Meis1 in quiescent epidermal stem cells, we again performed BrdU-chase experiments in K14Cre ER -Meis1 fl/fl mice. BrdU-LRCs were almost undetectable in the bulge of K14Cre ER -Meis1 fl/fl mice (n = 16), compared with the case in the Meis1 fl/fl control mice (n = 10) (P = 7.939610 210, t-test) (Fig. 2E, 2F, and 2H). These results demonstrate that the disruption of Meis1 function results in a reduction in the number of quiescent stem cells and a concomitant increase of differentiation into epidermal cells, as evidenced by the increased thickening of the epidermis of K14Cre ER -Meis1 fl/fl mice.

Meis1 is required for benign tumor development
To investigate the role of the Meis1 gene in skin carcinogenesis, we subjected 32 K14Cre ER -Meis1 fl/fl mice and 43 Meis1 fl/fl control mice to the DMBA/TPA chemical carcinogenesis protocol and monitored their tumor development for a period of 35 weeks. In the standard DMBA/TPA protocol, mice are treated once with DMBA, followed by repeated application of TPA. However, since the K14Cre ER -Meis1 fl/fl mice were in a C57BL6/J background, which is known to be resistant to tumor induction by DMBA/TPA chemical treatment [20], we modified the standard DMBA/TPA protocol and included four additional rounds of DMBA-TPA treatment to overcome the chemical resistance of C57BL6/J mice (Fig. 3A, Methods, Okumura K et al. unpublished data). The histology of these papillomas was confirmed to be the same as that induced by the standard DMBA/TPA protocol ( Figure 4A, B). We also checked Hras mutation by restriction digestion (Fig.  S1B). Hras is known to be mutated in more than 80 percent of  Papillomas are visually identified by their cauliflower-like projections that may appear white or normal-colored and may be pedunculated or sessile. Carcinomas are distinguished from  papillomas by their flattened appearance. For each mouse, we documented the number of papillomas present at 20 weeks after tumor initiation as well as the number of carcinomas present at 35 weeks after tumor promotion. In addition, the size of each papilloma was measured and recorded. In the control mice, Meis1 fl/fl (before-and after-DMBA combined) mice developed 5.0564.27 papillomas at 20 weeks after initiation (Fig. 3B, 3C). In contrast, before-DMBA-K14Cre ER -Meis1 fl/fl and after-DMBA-K14Cre ER -Meis1 fl/fl mice developed 2.3663.68 and 2.8964.79 papillomas, respectively (Fig. 3B, 3C), indicating that the total number of papillomas at 20 weeks was significantly lower than in control Meis1 fl/fl mice (Meis1 fl/fl vs. before-DMBA, P = 0.014; Meis1 fl/fl vs. after-DMBA, P = 0.027; all t-tests) (Fig. 3B, 3C). These results suggest that Meis1 functions in papilloma development.

Meis1 functions to maintain cell proliferation in benign tumors
To investigate the mechanism by which Meis1 supports papilloma development, we analyzed the papillomas in K14Cre ER -Meis1 fl/fl and Meis1 fl/fl mice by standard histology methods. HE staining of papillomas from knockout K14Cre ER -Meis1 fl/fl and control Meis1 fl/fl showed no significant morphological differences (Fig. 4A, 4B). Next, we conducted immunostaining with keratin 10 (K10/Krt10), which is a cytoplasmic intermediate filament protein expressed in the suprabasal layer of the normal epidermis. It is also a marker for differentiated tumor cells in papillomas. As a result, we observed almost no K10positive cells in papillomas of K14Cre ER -Meis1 fl/fl mice (n = 5) compared with the case in papillomas of the control Meis1 fl/fl mice (n = 6) (Fig. 4C, 4D). These results suggest that Meis is required for maintaining more differentiated tumor cells in papillomas. We then performed immunostaining with the cell proliferation marker Ki-67 to investigate the cell proliferation capacities of papillomas. We observed fewer Ki-67-positive cells in papillomas of K14Cre ER -Meis1 fl/fl mice (n = 5) than in papillomas of the control Meis1 fl/fl mice (n = 6) (P = 0.0001, t-test) (Fig. 4E-4G). These observations indicate that papillomas from Meis1-deficient mice grow more slowly when Meis1 is present. Taken together, our findings suggest that Meis1 supports both the development and the growth of papillomas by maintaining cell proliferation.

Meis1 is required for malignant conversion
To investigate the effect of Meis1 deletion on carcinoma development, we monitored 32 chemically treated K14Cre ER -Meis1 fl/fl mice and 43 chemically treated Meis1 fl/fl mice up to 35 weeks after tumor initiation. We observed that K14Cre ER -Meis1 fl/ fl mice have a lower incidence of carcinoma development as well as a later onset of carcinoma development compared with control Meis1 fl/fl mice (Fig. 5).
We observed that the K14Cre ER -Meis1 fl/fl (after-DMBA) mice developed the first carcinoma at 26 weeks after initiation, while control Meis1 fl/fl mice developed the first carcinoma at 19 weeks. In addition, we observed that less than 20 percent of K14Cre ER -Meis1 fl/fl (after-DMBA) mice had developed carcinoma at 35 weeks, whereas 53 percent of control Meis1 fl/fl mice had done so. This difference between K14Cre ER -Meis1 fl/fl (after-DMBA) and control Meis1 fl/fl mice was statistically significant (P = 4.86610 28 , Fisher's test) (Fig. 5).
In the K14Cre ER -Meis1 fl/fl (before-DMBA), we observed that the first carcinoma developed at 20 weeks, while control Meis1 fl/fl mice developed the first carcinoma at 19 weeks. Furthermore, 33 percent of K14Cre ER -Meis1 fl/fl (before-DMBA) mice had developed carcinoma at 35 weeks, whereas 53 percent of control Meis1 fl/fl mice had done so. Interestingly, we found that the difference in the number of carcinomas between K14Cre ER -Meis1 fl/fl (after-DMBA) and K14Cre ER -Meis1 fl/fl (before-DMBA) was statistically significant (P = 0.006, Fisher's test), as was the difference between K14Cre ER -Meis1 fl/fl (before-DMBA) and control Meis1 fl/fl mice (P = 0.004, Fisher's test) (Fig. 5). Our investigations of Meis1 in tumorigenesis indicate that, in addition to the role in papilloma development, Meis1 also functions to support the malignant conversion of benign papillomas into malignant tumors.

Meis1 expression successively increases with skin tumor progression
To assess further the role of Meis1 in skin carcinogenesis, we examined the expression of Meis1 in chemically induced papillomas, carcinomas in situ, and metastatic carcinomas from wild-type FVB/N mice by semi-quantitative RT-PCR and RNA sequencing (Aoto et al., unpublished data). The semi-quantitative RT-PCR study showed that Meis1 mRNA expression level increased during skin tumor development and progression (Fig. 6A). The expression of S100a6, one of the S100 family members, was checked in parallel as a malignancy marker [34], (Fig. 6A). For the RNA sequencing assay, the fragment number of cDNA containing the Meis1 gene was counted and plotted as FPKM (fragments per kilobase of exon per million mapped fragments, Fig. 6B). Data from RNA sequencing were similar to those detected by RT-PCR analysis (Fig. 6A, 6B). Specifically, both RNA sequencing and RT-PCR analysis indicated that Meis1 expression is successively increased from papillomas to carcinomas in situ to metastatic carcinomas. Collectively, these results indicate a proto-oncogenic role for Meis1.
We next investigated Meis1 expression in chemically induced papillomas, carcinomas in situ, and metastatic carcinomas of Meis1-EGFP mice by immunofluorescence (Fig. 6C-6J). Tumors were harvested at 11, 18, and 35 weeks after initiation. We costained the tumors for Meis1-EGFP and keratin 14 (K14/Krt14) expression. K14 is a cytoplasmic intermediate filament protein expressed widely in the normal epidermis and skin tumors. We observed co-expression of Meis1-EGFP and K14 in epidermal cell layers of early papillomas at 11 weeks after initiation ( Fig. 6C and  6D). In the papillomas harvested at 18 weeks, we observed that the range of Meis1-EGFP expression expanded beyond the basal layer cells of the epidermis (Fig. 6E and 6F). In carcinomas in situ, we found that Meis1-EGFP was co-expressed with K14 ( Fig. 6G and  6H). In metastatic carcinomas, Meis1-EGFP was expressed throughout the metastatic carcinoma ( Fig. 6I and 6J). Our observations indicate that Meis1-EGFP expression successively increases from the basal cell layer to the entire tumor as skin tumor development progresses.
Epithelial-to-mesenchymal transition (EMT) is a biological process that occurs during tumor development that transforms the existing epithelial tissue of cancer cells into mesenchymal tissue as a benign tumor, such as a papilloma developing into a malignant tumor such as a carcinoma. To determine whether or not the observed increase of Meis1 expression is associated with the development of increased mesenchymal tissue owing to EMT, we analyzed carcinomas in which the expansion of mesenchymal tissue is known to occur. Immunofluorescence analysis was performed to determine whether Meis1-EGFP is co-expressed with epithelial or mesenchymal tissue markers. In both carcinomas in situ and metastatic carcinomas, we observed that Meis1-EGFP was co-expressed with the epithelial basal marker K14, but not with the mesenchymal marker vimentin (Fig. S3A-S3D). Taken together, these results indicate that the successive increase of Meis1 expression during skin tumor development is not due to the increase in mesenchymal tissue associated with EMT, but rather indicates oncogenic potential of Meis1 during tumorigenesis.

Meis1 is localized to a non-stem cell region in benign tumors
To investigate further the role of Meis1 in cancer tissues, we analyzed Meis1 expression in chemically induced papillomas from Meis1-EGFP reporter mice by immunostaining. We co-stained for Meis1-EGFP with K14, K10, and cancer stem cell markers such as b4 integrin, CD34, and K15 ( Fig. 7A-7O). In contrast to Meis1 expression in normal epidermis (Fig. 1), we found that Meis1-EGFP was not co-expressed with b4 integrin, CD34, or K15 in the tumor basal layers, where epidermal cancer stem cells are thought to reside in papillomas (Fig. 7G-7O). Interestingly, in papillomas, we observed that instead of localizing to a stem cell compartment, Meis1-EGFP was co-expressed with K10 ( Fig. 7D-7F), which is a marker for differentiated tumor cells in papillomas. The change in Meis1 localization in malignant tissues suggests that a functional switch in Meis1 roles occurs when the epidermis is transformed from normal into hyperproliferative tissue.

Discussion
Meis1's functions and mechanisms in the epithelium remain poorly understood. Here, we present knockout and cell marker studies revealing Meis1's roles in both normal and tumor development. Our investigations of normal tissue indicate that Meis1 functions to maintain epidermal adult stem cells in the bulge region of the epidermis to maintain homeostasis of the epidermis. In contrast, our studies in the two-stage skin carcinogenesis mouse model indicate that Meis1 function changes to a more pro-tumorigenic role, supporting tumor development and malignant conversion. Collectively, our findings suggest a multi-function model for Meis1 in the epidermis (Fig. 8).
Here, we show that Meis1 expression in normal skin is predominantly associated with stem cells of the hair follicle bulge (Fig. 1). The hair follicle bulge is one of the stem cell niches of the epidermis [35]. However, the expression of Meis1 is not restricted to the bulge (Fig. 1). It is also expressed mainly in temporary portions passing through hair cycles within the hair follicles. Several lines of evidence have suggested the existence of non-LRC-type stem cells, such as Lgr (Leucine-rich repeat-containing G-protein-coupled Receptor)-positive stem cells in the same region within the hair follicles [36]. Further marker staining, such as of Lgr proteins, and mouse cross experiments with LGR-Cre mice will be required for future study. Meis1 is also expressed in the hair bulbs and dermal papillae (Fig. 1). It could be involved not only in stem cell maintenance but also in hair growth or maintenance. Nevertheless, our data suggest that disruption of Meis1 in the epidermis causes reduction of epidermal stem cells in the bulge and then thickened epithelium (Fig. 2). This is consistent with the observation that Meis1 expression is predominantly associated with the stem cells of the hair follicle bulge. Our findings of Meis1 function in the epidermis are similar to Meis1 studies in the bone marrow that showed that hematopoietic stem cell (HSC) compartments were severely affected and potentials for colony formation as well as repopulation of lethally irradiated recipient mice were profoundly impaired in Meis1-deficient cells, suggesting a critical role of Meis1 in HSC maintenance [37]. On the basis of these findings, we propose a model in which Meis1 is essential for maintaining quiescent epidermal stem cells in the hair follicle bulge of the epidermis (Fig. 8).
Additionally, we present RT-PCR and RNA sequencing data that show that Meis1 expression is successively enhanced with skin tumor progression (Fig. 6). The observed successive increase in expression of Meis1 during carcinogenesis, independent of EMT, suggests an oncogenic role for Meis1 in the two-stage skin carcinogenesis mouse model. Consistent with this conclusion, elevated Meis1 expression has also been observed in several tumor types including acute myeloid leukemia, lung adenocarcinoma tumors, neuroblastomas, ovarian carcinomas, and nephroblastomas. We propose a model in which Meis1 has two distinct functions influencing the promotion and progression of skin carcinogenesis (Figure 8).
The balance between self-renewal and differentiation of adult stem and progenitor cells maintains tissue homeostasis. Loss of Meis1 causes reduction of epidermal stem cells in the bulge, hyperproliferative epidermal cells, and then thickened epidermis (Fig. 2), indicating that Meis1 drives self-renewal in order to maintain quiescence of epidermal stem cells. However, in the absence of Meis1, the balancing mechanisms are disrupted, and the balance is shifted to differentiation. There are a lot of reports of knockout mice showing abnormal epidermal thickness. In the case of mice with the knockout of genes driving the self-renewal of stem cells, like Meis1, they tend to show thicker epidermis [38,39,40]. In particular, aPKCl (atypical PKCl) knockout mice show loss of quiescent hair follicle bulge stem cells, a temporary increase in proliferating epidermal cells, and thicker epidermis, which were all seen in our Meis1 study. These two genes might thus have very closely related functions. On the other hand, when the gene expression is restricted to progenitor cells in the epidermis, like for DNMT1 [41], disruption of the gene does not necessarily result in thicker epidermis. This is probably because of the difference between quiescent stem cells and more rapidly cycling progenitor cells. Because stem cells have higher differentiation potential than progenitor cells, deficiency of a balancing factor driving self-renewal in stem cells causes drastic differentiation into epidermal cells. Furthermore, our investigations of the role of Meis1 in tumorigenesis revealed an unexpected switch of Meis1 localization from a stem cell region in normal epidermal tissues to a more differentiated region in tumor tissues (Fig. 7). Interestingly, a few recent studies have shown that carcinoma stem cell-like cells were capable of forming welldifferentiated tumor tissue. Previous studies indicated that epidermal cells of all types including cancer cells were derived from keratinocytes of the bulge region [35,42]. Another recent study showed that CD34-expressing (a carcinoma stem cell (CSC) marker) cells, residing in DMBA/TPA-induced skin tumors, formed secondary tumors upon transplantation into immunodeficient mice [43]. However, another study showed that well-organized, differentiated papillomas containing CD34, b4-integrin, and K15-expressing cells (CSC markers) gave rise to layers of differentiated tissues [44]. Although tumor-forming potential appears to be maintained in carcinoma stem cells of skin tumors, such an explanation for the role of Meis1 in tumorigenesis is inconsistent with our findings. The cell marker expression studies in papillomas from Meis1 knockout mice indicate that Meis1 is not essential for maintaining CSCs in papillomas, as Meis1 expression in papillomas did not overlap with CSC-like markers (Fig. 7). While Meis1 function is required for maintaining epidermal stem cells in normal skin, our results suggest that Meis1 loses its stem cell maintenance function and switches to a mechanistically distinct role in tumorigenesis.
It has been known for many years that, in mouse skin carcinogenesis, initiated cells persist throughout the lifetime of the animal, suggesting that the incipient tumor cell is long-lived, a quality of stem cells [45]. LRCs of hair follicles are slow-cycling and long-lived, thus presenting characteristics of stem cells. In carcinogen-treated mouse skin, carcinogen-DNA adducts persist in these cells [46], and even after ablation of cycling cells in the epidermis with a chemotherapeutic drug prior to DMBA treatment, the rate of carcinoma formation is unchanged, indicating that tumor initiation occurs in quiescent stem cells rather than rapidly proliferating (TA) cells [47]. Meis1 was predominantly expressed in quiescent epidermal stem cells and essential for maintaining these cells. In addition, when the knockout of Meis1 was induced before DMBA tumor initiation, the number of papillomas significantly decreased. These results clearly suggest that tumor initiation occurs in quiescent stem cells and Meis1 is indispensable for maintaining these tumor-initiating (quiescent stem) cells. To investigate further Meis1 functions in  skin carcinogenesis, we subjected tamoxifen-inducible, conditional knockout K14Cre ER -Meis1 fl/fl mice to DMBA/TPA chemical carcinogenesis using two different treatment schemes. One set of mice were treated with tamoxifen, such that the knockout of Meis1 was induced one week before DMBA tumor initiation (referred to as ''before-DMBA'') and the other set of mice were treated with tamoxifen, such that the knockout of Meis1 was induced nine weeks after DMBA tumor initiation (referred to as ''after-DMBA'') (see Fig. 3A). Both the ''before-DMBA'' and the ''after-DMBA'' K14Cre ER -Meis1 fl/fl treated mice showed significant decreases in the number of papillomas compared with control Meis1 fl/fl mice (Fig. 3B). However, the mechanisms driving the observed papilloma reduction in the ''before-DMBA'' mice are likely to differ from those in the ''after-DMBA'' mice. Given the absence of epidermal stem cells in K14Cre ER -Meis1 fl/fl mice (Fig. 2E, 2F), taken together, this suggests that the decrease in papilloma in the ''before-DMBA'' mice is probably due to the decrease of epidermal stem cells. Meanwhile, the observed increase in Meis1 expression in early and late papillomas (Fig. 6A, 6B) suggests that the decrease in papilloma in the ''after-DMBA'' mice was due to Meis1's function in papilloma maintenance.
When we examined the effects on malignant conversion in the ''before-DMBA'' and ''after-DMBA'' K14Cre ER -Meis1 fl/fl mice, we observed that both of these groups showed significant decreases in the number of SCCs compared with control Meis1 fl/fl mice (Fig. 5). In addition, we found that the ''after-DMBA'' mice showed significantly lower incidence of SCC than the ''before-DMBA'' mice at 35 weeks (Fig. 5). Collectively, our results show that Meis1 has an oncogenic role in the epidermis, where it supports papilloma development and malignant conversion of skin tumors. Furthermore, the results from our ''before-DMBA'' and ''after-DMBA'' treatment studies provide further insight on the fundamental question of when malignant conversion is determined in the DMBA/TPA carcinogenesis mouse model. Our results suggest that malignant conversion is determined in this model no later than nine weeks after tumor initiation.
Our results also suggest that targeting Meis1 in skin tumors could be a powerful strategy for the treatment of skin cancer as well as other epithelial tumors in which Meis1 appears to have an oncogenic role, such as endometrial and ovarian cancers.

Mice
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Ministry of Education, Culture, Sports, Science, and Technology of Japan. The protocol was approved by the Committee on the Ethics of Animal Experiments of Chiba Cancer Center (Permit Number: [13][14][15][16][17][18]. All efforts were made to minimize suffering. Details on the generation of mice carrying the floxed allele of the Meis1 gene (Meis1 fl/fl ) are described elsewhere [11,37]. Meis1 fl/fl mice were mated to K14-Cre ER transgenic mice [48]. K14-Cre ER transgenic mice were obtained from The Jackson Laboratory (JaX 005107). Meis1-EGFP BAC-transgenic reporter mice, generated by the GENSAT BAC transgenic project [32], were obtained from the Mutant Mouse Regional Resource Center. All mice had a C57BL/6J background. Mice carrying alleles or transgenes were maintained and genotyped as previously described. To disrupt Meis1, 8-week-old K14Cre ER -Meis1 fl/fl and Meis1 fl/fl mice were injected subcutaneously 1 time with 0.5 mg of tamoxifen (final concentration 5 mg/ml, Sigma-Aldrich) that had been prepared following the manufacturer's instructions by completely dissolving tamoxifen into 250 ml of 100% ethanol at 55uC, adding 4750 ml of sunflower oil, and mixing them well by vortexing.

Skin carcinogenesis
7,12-Dimethylbenz(a)anthracene (DMBA) was purchased from Sigma Japan, and 12-O-tetradecanoylphorbol-13-acetate (TPA) was purchased from Calbiochem. DMBA is used as a carcinogen and TPA as a promoter. A total of 43 Mes1 fl/fl mice and 32 K14Cre ER -Meis1 fl/fl mice were treated according to the modified two-stage carcinogenesis protocol. At 8-10 weeks of age, the backs of the mice were carefully shaved with an electric clipper. Two days after shaving, DMBA (25 mg per mouse in 200 ml of acetone) was applied to shaved dorsal back skin. Three days after the first DMBA treatment, TPA (10 mg per mouse in 200 ml of acetone) was applied. After four rounds of this single DMBA and TPA treatment, the mice were treated with TPA twice weekly for 20 weeks. Papilloma number and size (mm in diameter) of each papilloma were recorded from 10 weeks up to 20 weeks, and carcinoma development was monitored up to 35 weeks post-TPA treatment.

Immunofluorescence
The dorsal back skin or tumors were fixed in 4% paraformaldehyde at 4uC overnight. Frozen skin and tumor embedded in OCT compound (Sakura Finetek) were cut into 10 mm sections. In contrast, dehydrated samples were embedded in paraffin and sectioned as 10 mm slices, which were stained with hematoxylin and eosin. The endogenous peroxidase activity in the specimens was blocked by treatment with 0.3% H 2 O 2 and samples were then rinsed with PBS. Sections were incubated with primary antibodies diluted in blocking buffer overnight at 4uC. The following primary antibodies were used:

Classification between papillomas and carcinomas
The majority of papillomas were determined by visual inspection. Papillomas appeared as outgrowths on the dorsal skin of mice. Some of the papillomas began to convert to carcinomas by becoming flatter on the skin and penetrating deeper into the dermis. They were normally easily distinguishable from one another. However, in some intermediate-type tumors, we prepared a paraffin section and confirmed the histology.

BrdU chase experiments
For chase experiments, BrdU (Sigma-Aldrich) was administered by peritoneal injection of postnatal day 24 (P24) mice (40 mg per gram of body weight) on three consecutive days. Mice were then chased for five weeks. Meanwhile, P44 mice were treated with tamoxifen. After the five weeks of chase, skin tissues were fixed with 4% paraformaldehyde and stained with rat anti-BrdU (Abcam) antibody.

RT-PCR
Total RNA was isolated from normal skin, papillomas, carcinomas, and metastatic carcinomas of mice after 40 weeks of DMBA/TPA treatment using TRIzol (Invitrogen) following the manufacturer's protocol. cDNA was generated with the iScript Select cDNA Synthesis Kit (Bio-Rad) using 0.1 mg of DNasepretreated total RNA. The cDNA was amplified for 40 cycles (94uC for 30 sec, 50uC for 30 sec, 72uC for 30 sec) using Prime Taq DNA Polymerase Kit (Genet Bio). Primer sets used for amplification of the Meis1 gene were 59-GCA AAG TAT GCC AGG GGA GTA-39 and 59-TCC TGT GTT AAG AAC CGA GGG-39. Primer sets used for amplification of the S100a6 gene were 59-CAG TGA TCA GTC ATG GCA TGC CCT C-39 and 59-CAT TTT ATT TCA GAG CTT CAT TGT AG-39. The products were subjected to agarose gel electrophoresis. cDNA integrity was confirmed using b-actin.

Gene expression analysis by RNA-Seq
Gene expression levels were estimated using FPKM (expected fragments per kilobase of transcript per million fragments sequenced) with Cufflinks [49]. RNA-Seq reads were aligned to the mouse genome sequence (GRCm38-release71) with a gene annotation file using TopHat2 [50], and the FPKM value for each gene was estimated based on the mapping results obtained using Cufflinks.
Hras mutation analysis DNA was prepared from papillomas of Meis1 fl/fl and K14Cre ER -Meis1 fl/fl mice. Exon 2 of the Hras gene was amplified as previously described [51]. Amplified 207-bp fragments were digested with a restriction enzyme, XbaI, at 37uC for 4 h and electrophoresed in a 4% Nusieve 3:1 agarose gel.

Statistical analysis
Statistical significance was calculated by the unpaired two-tailed Student's t-test or 262 Chi square test (Fisher's test). A P-value, 0.05 was considered statistically significant and a P-value,0.01 was considered highly statistically significant.

Supporting Information
Figure S1 Meis1 deletion and Hras mutation analysis. (A) Confirmation of Meis1 deletion in K14Cre ER -Meis1 fl/fl mice. Skin DNA prepared from K14Cre ER -Meis1 fl/fl mice that were either treated (+) or untreated (2) with TAM (tamoxifen) was subjected to PCR analysis using primer pairs described elsewhere [37]. DNA from Meis1 fl/fl mice was used as a control. FA denotes floxed allele. KOA denotes knockout allele. (B) XbaI digests of PCR products in the Hras gene. XbaI was used to detect the point mutation at codon 61 of the Hras gene induced by DMBA. After amplification of Hras exon 2 as previously described [51], PCR products from the mutant Hras allele produce 116-and 91-bp fragments after XbaI single digestion.