Hair follicle stem cells (HF-SCs) alternate between periods of quiescence and proliferation, to finally differentiate into all the cell types that constitute the hair follicle. Also, they have been recently identified as cells of origin in skin cancer. HF-SCs localize in a precise region of the hair follicle, the bulge, and molecular markers for this population have been established. Thus, HF-SCs are good model to study the potential role of oncogenic activations on SC physiology. Expression of a permanently active form of Akt (myrAkt) in basal cells leads to Akt hyperactivation specifically in the CD34+Itga6H population. This activation causes bulge stem cells to exit from quiescence increasing their response to proliferative stimuli and affecting some functions such as cell migration. HF-SC identity upon Akt activation is preserved; in this sense, increased proliferation does not result in stem cell exhaustion with age suggesting that Akt activation does not affect self-renewal an important aspect for normal tissue maintenance and cancer development. Genome-wide transcriptome analysis of HF-SC isolated from myrAkt and wild-type epidermis underscores changes in metabolic pathways characteristic of cancer cells. These differences manifest during a two-step carcinogenesis protocol in which Akt activation in HF-SCs results in increased tumor development and malignant transformation. Stem Cells2014;32:1917–1928
The skin is an essential organ that protects the individual from external aggressors. To prevent unwanted consequences due to accumulation of cell alterations caused by environmental injury, the epidermis (outermost skin layer) and its appendages have a rapid turnover rate. Newly formed keratinocytes will differentiate as they travel through the epidermal layers to be finally sloughed off, hair follicles are cyclically formed and destroyed, and sebocytes continually burst to shed their contents. In this scenario, distinct stem cell (SC) populations have been identified as responsible for the maintenance of the different compartments of the skin epidermis. Homeostasis in the interfollicular epidermis (IFE) is ensured by long-lived progenitors , sebaceous gland (SG) precursors reside in the hair follicle isthmus [2-4], and hair follicle (HF) regeneration is carried out by hair follicle stem cells (HF-SCs) localized in the HF bulge [5, 6]. HF-SCs are long-lived and endowed with an enduring proliferative potential, as they sustain formation of all HF layers in each of the hair cycles that take place throughout the life of the individual, interchanging episodes of proliferation, at the beginning of anagen (growth phase of the hair cycle), and periods of quiescence (during catagen and telogen phases of the hair cycle). Bulge stem cells have also been shown to be capable of acting as multipotent SCs when tissue homeostasis is disrupted. Thus, HF-SCs contribute to the newly formed IFE during initial wound healing [7, 8] and to the formation of SG in the context of sebaceous hypoplasia . Over the past decade, research in the field has identified and developed ways to detect and isolate bulge SCs. Bona fide HF-SC markers comprise bulge label-retaining cells (LRCs) [6, 9], the cell surface marker CD34 [10, 11], cytokeratin 15 (K15) [12, 13], and the G protein-coupled receptor 5 (Lgr5) . Also, lineage tracing models have made possible to follow the fate of the SC progeny during normal homeostasis and injury [7, 14]. This has unleashed a very welcomed insight into the biology of the epidermal stem cells in health and disease.
Normal tissue stem cells are likely candidates to be the cancer initiating cells in the cancer stem cell model, which considers that a limited number of stem cells capable of self-renewal are at the origin of tumor growth [15, 16]. The rapid turnover rate of the epidermis hinders the accumulation of the mutations needed for tumor initiation and development. Thus, in cutaneous squamous cell carcinoma (SCC) long-lived cells such as IFE progenitors or HF-SCs are the most likely candidates to initiate a carcinoma. This has been recently demonstrated using a genetically defined temporal model of tumorigenesis in situ [17, 18], where cells of the IFE were found to develop only benign papillomas while bulge SC and their immediate progeny were competent to produce both, papilloma and SCC. The study by White et al.  identified activation of the Akt pathway as a possible molecular mechanism underlying Ras-induced tumorigenesis from HF-SCs. In this context, it has been shown that Akt activity increases during SCC progression in mouse models of chemical carcinogenesis  and in human head and neck cancer . Also, Akt activation is sufficient to transform keratinocytes [20, 21] and to induce tumor formation in vivo [22, 23]. Importantly, activation of Akt in adult skin induces HFs to enter the growth phase (anagen) of the hair cycle  and has been associated with increased SC proliferation [24, 25] and with an expansion of the HF-SCs population .
We have used a conditional model in which expression of a permanently active form of Akt (myrAkt) is targeted to epidermal cells with proliferative capacities . Our studies show that, during hair telogen (quiescence), Akt is not active in HF-SCs in normal skin; however, transgene expression in K15, CD34-positive bulge SCs drives Akt activation in these cells. We have analyzed the impact of this activation on the normal homeostasis of HF-SCs and on their regenerating ability. We also explore the response of Akt-activated bulge SCs to proliferative stimuli and its consequences in the context of epidermal tumor formation.
Materials and Methods
All mice husbandry and experimental procedures were approved by the Animal Ethical Committee (CEEA) and conducted in compliance with Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas (CIEMAT) guidelines. Mice expressing a permanently active form of Akt in the basal layer of stratified epithelia, K5-myrAktL84 , and mice with ubiquitous expression of EGFP, beta-actin-EGFP , have been described. K5-myrAkt;beta-actin-EGFP animals were generated by mating and their genotypes determined by PCR.
In Vivo Regeneration Assay
Keratinocytes were harvested from the back skin of 9-week-old telogen mice as previously described . Keratinocytes of the indicated genotypes were mixed with wild-type (WT) dermal fibroblasts, resuspended in a final volume of 100 µl phosphate buffered saline (PBS), and injected into 7-mm inner diameter silicon chambers (Renner GmbH, Dannstadt, Germany) that had been previously implanted onto the back of anesthetized immunodeficient nu/nu mice (Janvier, Saint-Berthevin, France) as previously described .
Keratinocyte populations isolated from mice back skin were labeled with anti-CD34 (clone RAM34, eBioscience, San Diego, CA, http://www.ebioscience.com/) followed by staining with an anti-Itgα6 antibody (BD Pharmingen, San Agustin de Guadalix, Madrid, Spain) as previously described , or fixed and permeabilized and stained with either an anti p-Akt Ser473 (Cell Signalling, Danvers, MA, https://www.cellsignal.com/), HA (Hemagglutinin, clone 16B12, Covance, Princeton, NJ, http://www.covance.com/), or K5 (Covance) antibodies. For cell-cycle analysis, cells were incubated 1 hour with a solution containing propidium iodide, RNase, and NP-40 . Cells were analyzed using an EPICS XL (Beckman Coulter Electronics, L′Hospitalet de Llobregat, Barcelona, Spain) or a LSRFortessa (BD Biosciences) flow cytometer, and data were analyzed using CXP (Beckman Coulter) or FlowJo software (Tree Star, Inc.), respectively. Cell cycle was analyzed using Cylchred software (T. Hoy, University of Cardiff, UK). Dead cells and debris were excluded from the analysis. Cells were gated for single events and 100,000–500,000 events were collected from each sample.
Total RNA was prepared from CD34+ and CD34− keratinocytes obtained from pools of at least six 9-week-old telogen K5-myrAkt mice using magnetic cell separation (MACS) isolation procedure . A total of three isolation procedures were performed. Microarray data are accessible at NCBI Gene Expression Omnibus, http:www.ncbi.nlm.nih.gov/geo/, accession number GSE48522. Accession number for gene expression data from CD34+ and CD34− keratinocytes from WT mice is GSE19448 .
Akt Activation in Epidermis Confers Increased SC Potential
We and others [24, 25] have previously shown, using a colony-forming efficiency assay, that skin keratinocytes expressing myrAkt produce more colonies and with more undifferentiated morphology as compared to WT. This in vitro analysis suggests that Akt activation confers increased SC potential. To demonstrate this in vivo, we performed a competitive repopulation assay where we studied long-term regeneration capability of myrAkt versus WT skin keratinocytes. In this kind of regeneration assays, only long-term repopulating cells (SCs) persist in the newly formed epidermis after 7 weeks . Keratinocytes were isolated from the back skin of either beta-actin-EGFP (WT-GFP) or K5-myrAkt;beta-actin-EGFP (myrAkt-GFP) mice and combined with equal numbers of non-green fluorescent protein (GFP) competitor cells. The different cell combinations were implanted in silicone chambers inserted onto the dorsal fascia of immunodeficient mice (Fig. 1A) and regeneration was studied 8–10 weeks postengraftment. Our data show that myrAkt keratinocytes were more efficient at regenerating epidermis in this assay (Fig. 1B). This was quantified by flow cytometry analysis of WT-GFP + myrAkt grafts where transplanted cells were selected based on their HA (myrAkt) or GFP (WT) label (Fig. 1C). The relative contribution to the transplant of the myrAkt keratinocytes was fourfold compared to WT (79.6% ± 8.9% HA+; 19.1% ± 7.9% GFP+) (Fig. 1D). We used different cell combinations for our experiments to show that expression of GFP-actin does not affect WT or myrAkt keratinocyte capacity to regenerate a stratified epidermis (Fig. 1A, 1B, Supporting Information S1A, S1B). Moreover, WT and myrAkt keratinocytes showed similar adhesion kinetics (Supporting Information Fig. S1C) and sensitivity to cell death upon lack of attachment (anoikis) (Supporting Information Fig. S1D).
Hair follicle formation in this assay occurred infrequently compared to experiments where newborn keratinocytes or purified HF-SCs are used [11, 14, 31]. However, myrAkt keratinocytes seemed to have increased SC potential, as GFP-labeled HFs and SGs were only detected in regeneration experiments arising from GFP-expressing myrAkt keratinocytes (Supporting Information Fig. S1B, S1E). These in vivo data further demonstrate that permanent activation of Akt confers increased epidermal SC potential.
Transgenic Akt Activation in HF-SCs Does Not Affect SC Identity
Next, we studied Akt activation in HF-SC. The active form of Akt (p-Akt) was detected above the SG of telogen WT HFs and does not colocalize with the CD34 (Fig. 2A) and K15 (Fig. 2A′) bulge SC markers. In contrast, during anagen, p-Akt stained cells are found in the upper half of the HF below the outer root sheath, and in a subpopulation of cells of the lower bulge that coexpress CD34 and K15 (Supporting Information Fig. S2A). This subpopulation of bulge p-Akt cells are probably SCs beginning to migrate from the bulge . Flow cytometry analysis showed that the myrAkt transgene (HA+ cells) was expressed in CD34+ cells leading to Akt activation (p-Akt+) (Fig. 2C); consistently, p-Akt was detected in the bulge area colocalizing with CD34 and K15 staining (Fig. 2B, 2B′). Because not all CD34+ keratinocytes in myrAkt epidermis had active Akt (Fig. 2C), we performed flow cytometry analysis for cytokeratin 5 (K5) expression to rule out that this could be due to nonkeratinocyte contaminating cells. This analysis revealed that 99% of the CD34+ cells in our skin preparations were keratinocytes (Supporting Information Fig. S2B).
The CD34+ HF-SCs compartment comprises two cell subpopulations that differ in their surface level of the hemidesmosomal α6 integrin (Itgα6) and show different spatial distribution within the bulge . CD34-positive Itgα6 high-expressing cells (CD34+Itgα6H) are attached to the basal lamina while those expressing low levels of the integrin (CD34+Itgα6L) are suprabasal and thought to be derived from their basal counterparts . We found that, in transgenic mice, Akt was mainly activated in the CD34+Itgα6H population (Fig. 2D). This expression pattern could indicate that Akt activation in transgenic mice may affect bulge SC activation, as previously suggested [32, 33], and prompted us to study whether permanent Akt activation in HF-SCs could have an impact in their homeostasis.
We first explored whether Akt activation affected SC identity. To this end, we performed whole transcriptome analysis of purified CD34+ and CD34− populations from back skin myrAkt keratinocytes and studied the expression of a previously described 2,297-probests HF-SC signature in these samples and in WT CD34+ and CD34− populations . This HF-SC signature significantly overlaps with those previously published for human and mouse skin [5, 6, 34-36]. Unsupervised hierarchical cluster analysis for the expression of this signature genes revealed that CD34+ HF-SCs from myrAkt and WT are very similar, as they were mixed and clustered close together, as compared to CD34− cells from either genotype (Fig. 2E). Moreover, the expression of core bulge SC markers was conserved in CD34+ myrAkt keratinocytes (Supporting Information Fig. S2C). Therefore, transcriptome analysis revealed that Akt activation does not affect the expression of genes that characterize HF-SC identity.
Akt Activation Induces Proliferation of Quiescent HF-SCs But Does Not Affect Long-Term Maintenance of the Epidermal SC Pool
Previous studies have provided indirect evidence for a role of Akt in the activation of resting HF-SCs [24, 32, 33]. To address this issue, we studied HF-SC proliferation during the extended second telogen (resting) phase of the hair cycle (diagrams in Fig. 3A, 3C). First, we analyzed back skin keratinocytes with high expression of CD34 (CD34H) immediately after a 2-day BrdU pulse. Consistent with their quiescent state, WT bulge SCs incorporated little BrdU (0.04% ± 0.03%). In contrast, myrAkt showed increased BrdU labeling (0.21% ± 0.005%) (Fig. 3A). We also found increased BrdU incorporation in CD34 bulge SCs from aged myrAkt mice as compared to WT (Supporting Information Fig. S3A). Analogous results were observed when we studied the cell-cycle profile, telogen myrAkt HF-SCs contained a significantly higher percentage of S-G2M phase cells than WT (Fig. 3B and Supporting Information Fig. S3B). These experiments showed that myrAkt HF-SCs proliferated more than WT at the time of analysis. Analysis of BrdU incorporation immediately after a short pulse and cell-cycle profiles only provide information about the proliferative status of the SCs at a given time point. To examine HF-SC proliferation throughout the whole telogen phase, we took advantage of the BrdU label-retaining experiments where a 2-day BrdU pulse in perinatal anagen is chased until either the beginning of the second telogen or at its end (Fig. 3C). Label retention is a characteristic of bulge SCs, consequently the decrease in the number of bulge LRCs in WT mice is slight and not significant, underscoring the quiescent status of the telogen bulge. In contrast, the number of bulge LRCs was significantly reduced in myrAkt mice over this period, which supports the notion that Akt activation induces proliferation of HF-SCs.
To investigate whether myrAkt bulge SCs are also more responsive to a proliferative stimulus, we treated back skin epidermis with the phorbol ester 12-O-tetradecanoylphorbol 13-acetate (TPA) and pulsed the mice with BrdU for 1 hour before sacrifice. TPA treatment stimulates proliferation in different skin compartments such as the IFE and the infundibulum (where more proliferative cells reside ), and in the hair bulge (Fig. 3D). Bulge LRCs that are activated and mobilized upon TPA treatment colocalize with the HF-SC marker K15 [13, 38]. In agreement with the above-mentioned results, in the absence of TPA, myrAkt epidermis showed increased number of BrdU-incorporating cells within the bulge K15 compartment as compared to WT follicles that rarely showed any K15 proliferating cell (Fig. 3E and Supporting Information Fig. S3C). As expected, TPA treatment stimulated the proliferation of bulge K15-expressing cells and this effect was more pronounced in myrAkt epidermis (Fig. 3E and Supporting Information Fig. S3C). Overall our data provide evidence that Akt signaling induces activation of resting bulge SC and increases their response to an external proliferative stimulus.
Stem cells are endowed to maintain a cell lineage or a tissue for a lifetime. Thus, the balance between differentiation and self-renewal must be tightly regulated during their proliferation; any alteration of this balance can cause SC depletion or accumulation. We investigated whether increased proliferation of myrAkt HF-SCs affects the maintenance of this SC pool with age. For this purpose, we analyzed the expression profile of CD34 and Itgα6 in young and aged mice (Fig. 4A) and found that the number of HF-SCs was higher in myrAkt as compared to WT in 2- and 15-month-old mice (Fig. 4B and Supporting Information Fig. S3D). Also, SC abundance was not affected in myrAkt skin by aging (Fig. 4B and Supporting Information Fig. S3D). It is worth mentioning that the ratio of CD34H/Itgα6L versus CD34H/Itgα6H was previously reported to be augmented in myrAkt young mice as well as the clonogenicity [24, 25], and both were maintained with age (Fig. 4, Supporting Information Fig. S4A, S4B). This increase was not due to differences in the number of HFs in the transgenic mice (Supporting Information Fig. S4C). Also, similar to second telogen follicles (2-month-old mice), aged myrAkt CD34-expressing cells proliferate more actively than WT (Fig. 3 and Supporting Information Fig. S3). These data show that the proliferative effect of Akt signaling in HF-SCs leads to an increase in the SC pool that persists with age.
Wounds Do Not Re-epithelialize Earlier in myrAkt Mice and This Could Be Due to Impaired HF-SC Migration
Hair follicle SCs are implicated in the repair of damaged epidermis following injury, especially during the initial phase of wound-healing [4, 7, 8]. In particular, 8 days after wounding, the contribution of K15 follicular bulge SCs to the newly formed epithelium is 26% . Thus, based on the increased response of myrAkt HF-SCs to a chemical (TPA) proliferative stimulus (Fig. 3), we hypothesized that myrAkt skin would re-epithelialize faster than WT after a full thickness excisional wound. However, wounds in both genotypes closed at a similar rate (Fig. 5A) independently of the age of the mouse (Supporting Information Fig. S5A, S5B). Also, the length of the epidermal tongues and wound gaps studied 5 days postwounding were similar (Fig. 5B). Despite comparable healing kinetics, cell proliferation (as assessed by BrdU incorporation) was significantly higher in myrAkt epidermal tongues (Fig. 5C). These seemingly contradictory results led us to investigate migration, another major event that takes place during wound repair. It has been shown that keratinocyte migration, as opposed to keratinocyte proliferation alone, is critically important for timely skin re-epithelization . Wound closure in keratinocyte monolayers in culture was delayed in myrAkt keratinocytes (Fig. 5D), and we found that while WT keratinocytes migrated efficiently and directionally to close the wound, myrAkt keratinocytes moved randomly (Supporting Information Movies S1 and S2). In accordance with an anomalous migration, myrAkt keratinocytes showed altered organization of F-actin bundles and decreased activation of ERK and Rac1 (Supporting Information Fig. S5C–S5F′). Hair follicle growth also proceeds with HF-SC proliferation and migration. WT and myrAkt HFs show similar capacity to enter anagen 4 days after depilation; however, at day 7 myrAkt HFs had progressed less that WT (Supporting Information Fig. S5G) and at day 13 hair growth was clearly delayed (Supporting Information Fig. S5A). These data support the notion that defects in the migration ability could help to explain similar wound healing kinetic than WT despite increased myrAkt HF-SCs proliferation.
Akt Activation in HF-SCs Promotes Malignant Tumor Development
Sporadic tumors arise in aged myrAkt mice . These mice frequently develop oral cavity lesions (leukoplakias and erythroplakias) that do not progress into aggressive squamous cell carcinoma (SCC) due to p53-mediated senescence . Also, Akt activity is frequently increased during SCC progression in mice and humans [18, 19, 41]. Thus, at least in the oral epithelium, Akt activation acts as an oncogenic input, these data are in agreement with the role of the PI3K/Akt pathway in human head and neck SCC (HNSCC) [41, 42].
Recently, a signature of genes overexpressed in cancer stem cells (CSCs) has been established for skin SCC . We have used gene set enrichment analysis (GSEA)  to analyze the similarities between the gene expression of CD34+ bulge SC of WT and myrAkt mice and this CSC-gene signature. GSEA showed a significant enrichment of CSC-genes in myrAkt CD34+ HF-SCs, which is not present in WT CD34+ HF-SCs (Fig. 6A, 6B; Supporting Information Table S1). Functional analysis of CSC-CD34+ myrAkt genes revealed enrichment in processes deregulated in SCC and other cancers  (Fig. 6A). These data suggest that Akt activation in HF-SCs promotes a CSC gene expression profile. As described above (Fig. 2E), the expression pattern that characterizes SC identity is maintained upon Akt activation.
We used GSEA to search for signaling pathways enriched in myrAkt CD34+ cells as compared to WT. This analysis revealed that Akt activation has a great impact on the metabolic pathways of HF-SCs (Fig. 6C, 6D). Processes related to glucose, nucleic acid, lipid metabolism, and fat cell differentiation (PPAR signaling pathway) are overrepresented in CD34+ myrAkt HF-SC (Fig. 6C and Supporting Information Table S2). Transcripts for glucose transporter type 4 (SLC2A4/GLUT4), involved in glucose uptake, and for 7 of the 10 glycolysis enzymes (GPI1, PFKL, ALDOA, GAPDH, PGK1, ENO1, and PKM2) were increased. In particular PKM2, a key regulator of glycolysis that, by slowing the conversion of phosphoenolpyruvate to pyruvate, promotes the accumulation of upstream glycolytic intermediates and their shunting into anabolic pathways such as the pentose phosphate pathway (PPP). Upregulation of lactate dehydrogenase (LDHA, LDHB) suggests increased aerobic conversion of glucose to lactate in myrAkt SCs. This aerobic glycolysis is known as the Warburg effect and is characteristic of cancer cells (Fig. 6D). Other enzymes involved in lipid biosynthesis and acetyl-CoA metabolism (ACO1, IDH1, FASN, ACSS1, and ACSS2) are overexpressed in CD34+ myrAkt HF-SC (Fig. 6C, 6D and Supporting Information Table S2). Furthermore, when we used GSEA analysis to search for gene expression signatures that were enriched in myrAkt CD34+ HF-SCs deregulated genes as compared to WT, we found a significant overlap with gene sets relative to adipogenesis, skin cancer, and metabolism (Supporting Information Fig. S6A and Table S3). Similar results for glucose and lipid metabolism were obtained using gene ontology analysis of the biological processes for the genes overexpressed in these cells (p < .01) (Supporting Information Fig. S6B–S6D and Supporting Information Table S4).
To get a further insight into the transcriptional regulation of bulge SCs, we searched for transcription factors that were most likely to be responsible for the changes observed in gene expression between WT and myrAkt CD34+ HF-SCs. ChiP enrichment analysis (ChEA)  and TRANSFAC  revealed that a relevant number of the genes differentially expressed in CD34+ myrAkt HF-SCs show specific binding and/or transcription binding motif gene sets for E2F transcription factors, as well as LEF1, TCF3 and 4, and CTNNB1 (β-catenin) (Supporting Information Fig. S7) that are involved in bulge SC quiescence and activation [29, 47, 48], and regulation by PRDM1 (Blimp1) and PPARG (PPARγ) transcription factors, involved in commitment to SG progenitors and sebocyte differentiation, respectively (Supporting Information Fig. S7A–S7C). ChEA analysis showed that the Akt target HIF1A, that is overexpressed (×3.1-fold change) in myrAkt CD34+ HF-SCs, is the main factor regulating the expression of the genes involved in glucose catabolism in these cells (Fig. S7D).
There is strong evidence of a key role of long-lived SCs in SCC formation in multistage chemical carcinogenesis protocol, a well-established mouse model of skin SCC. This two-step model involves an initiation step with a low dose of the mutagen 7,12 dimethylbenz[α]anthracene (DMBA), followed by the continuous stimulation of epidermal proliferation. During this last promotion step benign tumors (papillomas) arise, and some of them progress to SCCs [49-51]. Because initiated skin retains the ability to form papillomas, even with long intervals between carcinogen exposure and tumor promotion, it is believed that the initial mutation arises in slow-cycling SCs . Furthermore, the proportion of cells expressing CD34 increases during the progression to SCC defining a CSC-like cell population with long-term self-renewal capacity , and the absence of CD34 impairs tumor development in two-stage carcinogenesis . Given that Akt activation led to proliferation and expansion of the CD34+ HF-SCs, we studied tumor susceptibility of myrAkt mice to the two-stage tumorigenesis protocol. Tumors appeared in all myrAkt mice (100% incidence) starting 3 weeks after DMBA treatment, whereas in WT incidence was 70% and tumors formed later (>10 weeks) (Fig. 7A, 7B). Tumor multiplicity (mean number of tumors per mice) was also higher in myrAkt mice throughout the experiment (Fig. 7A, 7B′). In this regard, myrAkt mice developed 11.7 tumors per mice at the end of the experiment (20 weeks after DMBA) while WT developed 1.6. In addition, myrAkt tumor size was bigger than WT (Fig. 7C).
The most common event during chemical carcinogenesis in mouse skin is the formation of benign tumors (papillomas), in some cases conversion occurs and papillomas evolve to SCCs. This conversion rate is influenced by the genetic background. In WT mice, we did not observe any papilloma conversion; however, more than 50% of the tumors found in myrAkt showed malignant conversion (Fig. 7D). Interestingly, we found that most of the malignant tumors in myrAkt mice have an important sebaceous component (sebaceous adenomas and adenosquamous carcinomas) (Fig. 7D, 7E); this is consistent with increased lipid metabolism found in myrAkt CD34+ HF-SCs (Fig. 6D, 6E). In agreement with previously described , these tumors show CD34 expression at the tumor-stroma interface colocalizing with the undifferentiated keratinocyte marker K5 (Supporting Information Fig. S8A). Tumor CD34-expressing cells also show Akt activation (Supporting Information Fig. S8B). Collectively, these observations show that Akt activation promotes proliferation of quiescent HF-SCs and changes in their metabolism that could account for the increased susceptibility to skin tumor development and malignant conversion.
Recently, HF-SCs and their immediate progeny have been identified at the origin of squamous skin tumors [18, 43, 55], and Akt pathway activation in the stem cell population has been related to the tumorigenesis process . In this context, our studies have important implications. First, they reveal that the expression of a permanently active form of Akt leads to HF-SCs activation, increasing their response to proliferative stimuli. Second, the cells maintain their HF-SC identity and functionality. Nevertheless, some biological processes such as migration and metabolism are altered. Finally, in the context of skin tumorigenesis, Akt activation increases tumor development and malignancy.
Stem cell transition from quiescence to activation into a proliferating progeny is a tightly controlled mechanism involving inhibition of bone morphogenetic proteins (BMP) signaling [32, 33, 56] and activation of the Wnt pathway [57-59]. This BMP/Wnt counterbalance is not exclusive to HF-SC [60, 61] and relies on the stabilization and nuclear translocation of β-catenin that acts as a transcriptional cofactor to promote SC activation. In this scenario, Wnt signaling per se is not sufficient to induce SC activation [59, 60] and blockade of the BMP signal, releasing Akt inhibition by phosphatase and tensin homolog (PTEN), would cooperate in the process of SC activation by helping control β-catenin nuclear localization [32, 33, 61]. Consistently, active Akt is not detected in the quiescent bulge, and becomes active in label retaining-SCs migrating from the bulge during early anagen . Activation of Akt signaling in the skin during the resting period of the hair cycle (telogen) induced HF growth  and an increase in the size and colony-forming efficiency of the CD34+ HF-SC population [24, 25]. Here, we show that Akt signaling is enough to induce HF-SCs to proliferate during the quiescent period of the hair cycle, and it also renders these cells more responsive to proliferative stimuli. Gene expression data showed diminished expression of known HF-SC growth inhibitor FGF18  and increased expression of growth stimulating factors such as TGFB2 (TGF-β2) . Diverse insulin-like growth factor binding proteins (IGFBPs) have been reported to be augmented in follicular SC as compared to epidermis . These proteins control cell growth by regulating insulin-like growth factor (IGF) bioavailability; however, it has been shown that IGFBPs can signal independently of IGF, exerting both growth-inhibitory and -potentiating effects. In fact, IGFBP4 acts as an inhibitor of the canonical Wnt pathway by directly interacting with the Wnt receptor Fzd8 to prevent Wnt3a binding . Our results show increased expression of IGFBP-3, −4, and −5 and downregulation of all IGF receptors (IGFR1 and IGFR2) in myrAkt HF-SCs as compared to WT. Importantly, Akt activation did not affect HF-SC identity, as revealed by the maintenance of a core gene signature.
Typically, HF-SCs fuel hair growth during a lifetime, and this potential relies on their capacity to self-renew. Also, bulge SCs require a degree of multipotency, as they might need to regenerate other epidermal compartments such as IFE, under situations of wound repair, or SG. We show that Akt activation in HF-SCs does not affect their self-renewal ability or multipotency, as aged mice retain their bulge SC pool and the ability to regenerate wounds after injury to the epidermis. Given that myrAkt keratinocytes have increased regeneration potential as compared to WT in competitive repopulation assays and the important input of the bulge SC cells to the newly forming epidermis during initial wound repair, one would expect that Akt-activated HF-SCs would contribute to a more rapid wound closure. However, despite increased cell proliferation in the epidermal tongues of myrAkt mice, wound closure was not accelerated as compared to WT. Similar findings came from hair regrowth experiments, where anagen entry did not seem to be affected but hair follicle growth was delayed in myrAkt mice. Our results suggest that impaired migration in myrAkt HF-SCs could be, at least in part, responsible for this effect. In addition to cell proliferation, migration of the HF-SCs out of the bulge is important for the processes of hair growth and wound healing. It has recently become clear that Akt activation influences cell motility in a cell-type dependant manner and a variety of mechanisms have been proposed [64, 65]. In particular, PI3K/Akt signaling is involved in the control of the actin cytoskeleton organization via small GTPases such as Rac1 or Cdc42 [66, 67], and can inhibit cell migration in epithelial cells through the disruption of F-actin bundles , destabilization of the TSC2 complex , suppression of ERK activity , and/or degradation of the transcription factor NFAT [67, 71]. Also, there is decreased expression of several extracellular matrix binding proteins, including ITGA6 (Itgα6), in CD34+ myrAkt HF-SCs as compared to WT. Altogether, these data suggest that Akt signaling could be influencing HF-SC migration which undoubtedly merits future investigation.
Our gene expression data show that myrAkt activation in epidermal stem cells, while maintaining the stemness of these cells, has a profound impact on the regulation of their metabolic status. In fact, we found that the myrAkt stem cells show increased expression of genes involved in glucose uptake, glycolysis, and conversion of phosphoenolpyruvate to lactate. This aerobic glycolysis, termed “the Warburg effect” , is a seemingly inefficient way of generating energy; however, it is used by most cancer cells and in general by proliferating cells [73, 74]. The advantage of this pathway is that facilitates the uptake and incorporation of nutrients into the biomass to sustain cell proliferation. In this sense, alterations in genes involved in cancer (growth) related pathways can promote or inhibit this proliferative metabolism, for instance, PTEN expression confers cancer resistance through a tumor suppressive metabolic state that involves inhibition of the PI3K/Akt activity [75, 76]. We find increased expression of Akt targets that are key regulators of glycolysis, such as HIF1A (HIF1α), PKM2, SLC2A4 (GLUT4), and PFKL, and upregulation of synthetic pathways such as the PPP, involved in nucleic acid synthesis, amino acid, and lipid synthesis (Fig. 6D). Array data have shown that glycolysis pathway genes are commonly overexpressed in different types of cancer .
Akt activation in the epidermis yields mainly spontaneous benign tumors [23, 25] and, in combination with p53 loss, leads to aggressive tumor development . Similarly, expression of oncogenic Ras leads to tumor formation ; however, when expressed at physiological levels, it drives formation of papillomas and a second hit is required for the development of SCCs [17, 18]. These last models show that HF-SCs or their immediate progeny are the cellular origin of the SCCs. Our data show that, in the context of oncogenic Ras [51, 79], Akt activation in bulge SCs induces cutaneous SCC development. An important number of the tumors that developed in our DMBA/TPA-treated myrAkt mice were sebaceous adenomas and adenosquamous carcinomas. In this regard, transgenic myrAkt mice develop SG hyperplasia ; this could be due to Akt activation in SG progenitors. However, recent findings suggest that constitutive activation of Ras oncogene in bulge SC can induce their migration toward the SG leading to gland hyperplasia [17, 18]. Our microarray analysis showed that genes differentially expressed in myrAkt CD34+ HF-SC are under the regulation of transcription factors important for commitment and differentiation into SG lineage, such as PRDM1 (Blimp1) and PPARG (PPARγ). Also, Akt activates the transcription of lipid metabolism genes in bulge SC.
Our data suggest that activation of Akt in epidermal stem cells promotes metabolic changes that could favor carcinogenesis. In addition, our study reveals that Akt activation in these cells affects the regulation of genes involved in SG lineage commitment and differentiation, thus providing a molecular base for spontaneous SG hyperplasia and for increased number of sebaceous tumors developed by myrAkt mice during induced carcinogenesis.
In summary, we show that Akt activation in HF-SC induces important changes that affect their activation status and makes them more responsive to proliferating stimuli while maintaining their SC identity. In addition, our results suggest that processes involved in their functionality, such as migration, could be altered. Genome-wide transcriptome analysis shows a deregulation of glucose metabolism in these cells, consistent with a Warburg phenotype, and suggests an early commitment into the sebaceous lineage. Accordingly, these mice develop more aggressive tumors upon skin carcinogenesis protocols and the tumors frequently display a sebaceous component.
We thank A. Holguín for technical help with silicone chamber implantation, R. Fernández-Senso for technical support, M.J. Escámez and L. Martínez-Santamaría for their advice on wound-healing assays, and the technical support by the personnel of the CIEMAT Animal Facility. The excellent support from J.C. Segovia and R. Sánchez from Lacisep (Laboratory of flow cytometry and cell separation, CIEMAT) is specially acknowledged. This work was supported by MINECO Grants SAF2011-26122-C02-01 and SAF2012-34378, CAM Oncocycle Program Grants S2010/BMD-2470, and ISCIII-RETIC Grants RD06/0020/0029 and RD12/0036/0009. J.F.A. is currently affiliated with the Departments of Medicine, Leon H. Charney Division of Cardiology, and Cell Biology and the Marc and Ruti Bell Vascular Biology and Disease Program, New York University School of Medicine, New York, NY. J.M.A. is currently affiliated with the Research Department of Cancer Biology, Cancer Institute, University College of London, London, U.K.
C.S.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; R.G.E.: data analysis and interpretation; M.I.G.: collection and assembly of data and data analysis and interpretation; J.F.A., J.M.A., and M.S.: collection and assembly of data, P.H.: histology work; J.M.P.: conception and design, manuscript writing, and financial support; C.L.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.