IKKβ-Mediated Resistance to Skin Cancer Development Is Ink4a/Arf-Dependent

Cancer is a multifactorial disease, caused by alterations in oncogenes that activate growth-promoting signaling cascades or in tumor suppressor genes whose inactivating mutations lead to DNA reparation defects, genomic instability, or defective control of cell proliferation. Multiple studies indicate that deregulation of the activity of NF-κB can result in cancer development. NF-κB is a conserved family of ubiquitous transcription factors that regulates the expression of many genes involved in cell proliferation, survival, apoptosis, and other essential processes, which is usually overactivated in inflammatory diseases and in cancer (reviewed in ref. 1). NF-κB is usually bound to inhibitory proteins of the I-κB family, which upon activation are phosphorylated by the IKK complex and degraded by the proteasome; consequently, NF-κB is released and modifies the expression of a myriad of genes containing specific DNA-regulatory elements.

The multiprotein complex IKK is formed by the regulatory subunit IKKγ and the catalytic subunits IKKα and IKKβ. The first known function of IKKβ kinase was the regulation of NF-κB activation upon proinflammatory stimuli, but it also performs important NF-κB–independent functions, as IKKβ phosphorylates a plethora of target proteins, thus controlling their activity and stability. Important tumor suppressor genes (TSG), like p53 (2) and p16 (3), are targets of IKKβ; the same holds true for other relevant proteins in cancer (as MDM2, p63 family members, ATM, IRS-1, TSC-1, or FOXO3a; reviewed in refs. 4–6). Furthermore, studies performed in MCF-7 mammary gland cells revealed the astonishing variety of IKKβ substrates, as more than 4,000 proteins containing phosphorylation sites for IKKβ were found (7). In this way, IKKβ affects mTOR, insulin, and Wnt signaling, immune responses, autophagy, response to DNA damage, and cell transformation. In summary, IKKβ acts over many different cellular processes, including tumoral transformation, both dependently and independently of NF-κB.

Animal models with altered expression of members of the NF-κB signaling pathway highlight the importance of these proteins in skin cancer. For example, mice lacking p65, the main NF-κB subunit, are resistant to skin carcinogenesis (8); the expression in proliferative skin cells of a super repressor form of I-κBα leads to hyperplasia, to a strong inflammatory response and finally to the development of squamous cell carcinomas (SCC; ref. 9); IKKα also has a role in skin cancer, showing protumoral or tumor-suppressive activity depending on the experimental setting (10, 11).

Regarding the relationship between IKKβ and cancer, a tumor-promoting activity of IKKβ in intestinal, hepatic, lung, and pancreatic cancer has been reported (12–17). In myeloid cells, by contrast, IKKβ is implicated in orchestrating an antitumoral immune response against melanoma cells (18). So, although IKKβ promotes tumor development in general, its actual effect over tumor growth and malignancy is cell type specific. We have previously described that IKKβ promotes tumor development in oral epithelia (19), but its role in skin cancer has not been determined yet. In this work, we explore the effect of IKKβ in skin cancer and the underlying molecular pathways by using transgenic mice with increased levels of epidermal IKKβ and absence of p53 or p16/p19 tumor suppressor proteins. At difference to the generally reported protumoral role of IKKβ in epithelial carcinogenesis, IKKβ showed a strong tumor suppressor activity in skin tumorigenesis that is independent of p53; interestingly, it is mediated by the proteins coded by the Ink4a/Arf locus (p16 and/or p19). In the absence of p16 and p19, we found that IKKβ promotes the emergence of spindle SCCs, a rare variant of malignant skin tumor. These results have implications for the implementation of antitumoral or anti-inflammatory therapies in skin, as IKKβ is considered a potential target for the development of these types of interventions.

Mouse experimental procedures were performed according to European and Spanish regulations and were approved by the Ethics Committee for Animal Welfare of CIEMAT and by the legal authority (protocol codes BME02/10 and PROEX086/15). Transgenic mice used in this work have been described previously: line L1of K5-IKKβ mice in B6D2 hybrid background (20); TgAC (FVB/NTac-Tg(Hba-x-v-Ha-ras)TG.ACLed) mice (21), p53^EKO mice (22, 23), and Ink4a/Arf^KO mice (lacking p16 and p19 proteins in FVB background; ref. 24). K5-IKKβ mice are available at European Mouse Mutant Archive (code EM: 09179).

The mice used in this study were viable and lack any evident alteration. In the breedings, all the genotypes were obtained in the expected ratios.

For two-stage skin carcinogenesis experiments, back skins of mice were initiated two days after shaving by treatment with 100 μg of 7,12-Dimethylbenz[a]anthracene (DMBA, Sigma-Aldrich), and promoted with 5 μg of 12-O-tetradecanoylphorbol-13-acetate (TPA, Sigma-Aldrich) in 200 μL of acetone twice weekly during 12 weeks. In carcinogenesis experiments in TgAC background, skins were TPA-treated twice weekly for 12 weeks. Tumor number and size were recorded weekly. For bromodeoxyuridine (BrdUrd) labeling, mice were injected intraperitoneally with BrdUrd (0.1 mg/g weight, Sigma-Aldrich); 1.5 hours later, mice were sacrificed and the skin was harvested. For hyperplasia induction, mice back skin was shaved and topically treated twice with 5 μg of TPA or vehicle at days 3 and 5 after shaving.

Genotyping of K5-IKKβ and of p53 and Ink4a/Arf loci were performed by PCR analysis of tail DNA, as described previously (20, 23, 25).

Mouse tissues were dissected and fixed in 10% buffered formalin or 70% ethanol and embedded in paraffin. Five-micron thick sections were used for hematoxylin and eosin (H&E) staining or IHC preparations. Antibody references are provided in Supplementary Material and Methods. Immunoreactivity was revealed using an ABC avidin–biotin–peroxidase system and ABC substrate (Vector Laboratories), and counterstained slightly with hematoxylin.

For quantification of BrdUrd staining in Fig. 1E, the number of BrdUrd-positive cells in 600 to 800 basal cells in four epidermal hyperplasias from 4 TgAC/K5-IKKβ mice and in 6 papillomas from 4 TgAC mice were counted. For vessel density (Fig. 1F), the number of blood vessels were counted in 10 different fields (20× objective) in 5 different TgAC mice and 4 different TgAC/K5-IKKβ mice topically treated with TPA for 3 weeks.

For the flow cytometry analysis shown in Supplementary Fig. S2B and S2C, mean values were obtained from the analysis of five individual K5-IKKβ and wild-type (wt) mice. Similar results were obtained in three independent experiments. The images shown in Supplementary Fig. S2D and S2E are representative of the data obtained in the analysis of 3 wt and 3 K5-IKKβ 10-week-old mice. Data in Supplementary Fig. S2F–S2H comes from the analysis of 4 animals of each genotype.

Whole-cell protein extracts from mouse tumors and tissues were prepared using standard techniques. Protein content was determined by the Bradford colorimetric protein assay (Bio-Rad Laboratories). Antibody references are provided in Supplementary Material and Methods.

Primary keratinocytes were obtained from skins of 2-day-old mice and cultured as described previously (20). Plates with 3-day–cultured keratinocytes were scrapped-off, pelleted, and the protein extracts subjected to Western blotting.

Data are expressed as mean ± SD. Statistical significance of data was assessed using the unpaired, two-tailed Student t test or Fisher exact test. P values <0.05 were considered significant.

With the aim of determining the role of IKKβ in nonmelanoma skin cancer (NMSC), we performed DMBA-TPA carcinogenesis assays in transgenic mice overexpressing IKKβ in basal cells of skin (K5-IKKβ mice) and in nontransgenic littermates (n = 17 for each genotype). K5-IKKβ mice developed fewer tumors than wt mice (Fig. 1A). In addition, fewer K5-IKKβ mice generated tumors, and those tumors were smaller than wt tumors (not shown); all these data suggest an antitumoral activity of IKKβ in skin. Given the low tumor number obtained in this carcinogenesis protocol, we aimed to confirm these data in other genetic backgrounds more prone to skin tumorigenesis. We used TgAC mice, which express a mutated Ha-Ras oncogene in skin (26). We generated cohorts of 9 mice bearing and 11 mice without the K5-IKKβ transgene in TgAC background, which were subjected to topical treatment with TPA to induce papilloma formation. Although 100% of the control TgAC mice developed papillomas by week 7 of treatment, none of the TgAC/K5-IKKβ mice developed tumors at this time point (P < 0.0001), and only around half of them developed tumors by week 12 (not shown). Furthermore, the number of papillomas per mouse and their size was markedly diminished in TgAC/K5-IKKβ mice compared with TgAC mice (Fig. 1B and C). There were marked macroscopic and microscopic differences between both genotypes: while TgAC lesions were truly exophytic papillomas typical of this carcinogenesis protocol, the lesions obtained in TgAC/K5-IKKβ were much smaller papillomatous epidermal hyperplasias (Fig. 1C), which proliferated less and induced a weaker angiogenic response than TgAC papillomas (Fig. 1D and E). This weaker angiogenic response was also observed in TgAC/K5-IKKβ skin TPA-treated for 3 weeks (before the development of skin lesions) when compared with TgAC skin (Fig. 1F). These results indicate that IKKβ precludes skin tumor formation. Transgenic mice express IKKβ in a variety of tissues; the marked difference in IKKβ effect over tumorigenesis between skin and oral epithelia do not seem to be related with differences in the expression level of IKKβ between both tissues (19).

Inflammation by infiltrating immune cells facilitates the acquisition of the fundamental features of cancer (27) and deeply affects skin tumor development (28). One important function of IKKβ is the regulation of the inflammatory process by controlling NF-κB activity. So, the antitumoral activity of IKKβ in skin could be mediated by differences in skin resident immune cell populations. We studied by flow cytometry the number of different immune cells both in epidermis and in dermis of wt mice and K5-IKKβ littermates (Supplementary Fig. S1A and S1B). Interestingly, the percentage of CD45⁺ cells is roughly double in dermis than in epidermis of normal mice (9.3% vs. 4.3%); these cells are mainly T lymphocytes (CD3⁺ cells), both inflammatory (CD3⁺, Tβ⁺) and skin-resident T cells (CD3⁺, Tγδ⁺). We also detected other immune cell populations potentially relevant in carcinogenesis, as macrophages (CD11b⁺), dendritic cells (CD11c⁺), and natural killer cells (Supplementary Fig. S1A). In K5-IKKβ epidermis, we observed a modest, nonsignificant decrease in CD3⁺ cells (both in Tβ⁺ and in Tγδ⁺ subpopulations), a slight increase in macrophages, and a more prominent diminution in NK cells (Supplementary Fig. S1B). In dermis, all the populations studied were roughly equally abundant in Tg and in wt mice. From these results, we conclude that the minor changes observed in the frequency of immune cells (T lymphocytes and myeloid cells) do not seem responsible for the marked differences in skin cancer susceptibility between K5-IKKβ Tg and wt mice.

The decreased susceptibility to tumoral transformation in K5-IKKβ mice could be related to differences in the amount of cells able to originate a tumor. CD34 is a cell surface marker expressed by skin stem cells in the bulge region of the hair follicles (29), which is needed for papilloma formation in mice (30). CD34⁺ cells become more abundant during the process of malignization (31), and CD34⁺ cells isolated from primary tumors can originate new tumors if transplanted to secondary recipients (32). Therefore, CD34 is considered a marker of stem cells and of cancer stem cells in skin. To assess whether the lower frequency of skin cancer in IKKβ-overexpressing mice is related to a reduction in the number of skin CD34⁺ cells, we studied the abundance and cell-cycle distribution of these cells (Supplementary Fig. S2A–S2C). Surprisingly, K5-IKKβ mice have roughly double the number of CD34⁺ cells than their nontransgenic counterparts (Supplementary Fig. S2A and S2B). As expected, these CD34⁺ cells were located in the hair follicle bulge both in wt mice and in K5-IKKβ mice skin (Supplementary Fig. S2D and S2E), but they seem to cycle slightly less actively in Tg than in wt mice (statistically nonsignificant, Supplementary Fig. S2C). In BrdUrd label-retaining assays, we did not detect a reduction in the stem cell population in K5-IKKβ mice (Supplementary Fig. S2F–S2H). Therefore, we conclude that the decreased sensitivity to skin cancer found in K5-IKKβ mice is not due to the lack of stem cells able to generate tumors, or to differences in their location or proliferative state.

To understand the differences observed in tumoral predisposition between wt and K5-IKKβ transgenic skin, we studied the expression of several TSGs important in tumoral transformation. We analyzed p53 (encoded by the TSG most frequently mutated in cancer), p19 and p16 proteins (products of the transcription of the Ink4a/Arf locus, implicated in the functionality of p53 and Rb proteins, respectively), and p21 (a protein transcriptionally regulated by p53 that controls cell-cycle progression). As the analysis of these proteins in complete skin by Western blot analysis is not appropriate, due to the variety of cellular types present in skin, we studied cultured skin keratinocytes from K5-IKKβ and wt mice. We consistently found that transgenic keratinocytes express increased amounts of p53, p16, and p19 (Fig. 2A). We confirmed the result for p19 by IHC staining in TPA-treated skin sections (Fig. 2B and C: 62% of basal nuclei were intensely stained in Tg skin, at difference with the weak staining found in 24% of the nuclei in wt mice). In accordance with the data obtained in the Western blot analysis, we found, in qRT-PCR analysis of RNAs from Tg- and wt-cultured skin keratinocytes, a significant increase for p16, p19, and p53 in Tg keratinocytes, but not for p21 (Fig. 2D). Altogether, these results suggest that the skin tumor–protective activity of IKKβ could well be mediated by the increased amount of p16, p19, and p53 in skin keratinocytes.

We aimed to analyze the implication of p53 in the skin tumor–suppressive activity of IKKβ. To this end, we crossed K5-IKKβ mice with p53^fl/fl/K14-Cre mice (22, 33). These mice efficiently delete p53 exons 2 to 10 in the vast majority of epidermal cells (23), and so we named them p53^EKO mice; mice carrying both genetic modifications (i.e., p53^EKO/K5-IKKβ mice), simultaneously lack p53 and overexpress IKKβ in skin basal proliferative cells, where skin tumors arise. We performed a two-stage skin carcinogenesis experiment and compared the rate of tumor appearance, tumor size, and histologic classification between p53^EKO/K5-IKKβ and control p53^EKO mice (Fig. 3). Interestingly, p53^EKO/K5-IKKβ mice also showed a great decrease in the mean number of skin tumors per mice along all the experiment (Fig. 3A), similar to the results obtained in p53^wt background (Fig. 1A and B). Also, skin tumors in p53^EKO/K5-IKKβ mice were smaller than in p53^EKO mice, being the proportion of medium and large tumors (diameter of 2–5 mm, and >5 mm, respectively) greater in p53^EKO mice (Fig. 3B). Macroscopically, most tumors arising in p53^EKO/K5-IKKβ mice were small, exophytic, and papillomatous (although four of the animals of this genotype developed only one tumor each of faster growth and apparently more malignant, with ulceration and infiltration of the subcutaneous tissue, not shown). Most p53^EKO mice, in contrast, developed multiple sessile-infiltrating tumors, presumably malignant. We assessed histologically 51 p53^EKO and 26 p53^EKO/K5-IKKβ tumors collected at weeks 13–18 (Fig. 3C). We found several tumor types, and classified them as benign tumors (including squamous papillomas, trichoepitheliomas, adenosquamous tumors, and basosquamous tumors) or malignant tumors (including squamous papillomas with invasive microcarcinoma foci and more malignant undifferentiated SCCs). Interestingly, the percentage of malignant tumors was lower in p53^EKO/K5-IKKβ mice than in p53^EKO mice (Fig. 3C). Representative examples of these tumors are shown in Fig. 3D–G. Benign tumors keep a clear basal membrane that prevents invasion (arrowheads in Fig. 3D and E). Carcinomas were frequently accompanied by inflammatory cells, and showed highly dysplastic keratinocytes invading underlying dermis, with condensed nuclei (karyopyknosis), variable in size (anisokariosis), probably as a consequence of p53 absence (asterisks in Fig. 3F and G). The tendency to develop of less malignant tumors in p53^EKO/K5-IKKβ mice was further confirmed by the decreased expression of markers associated to malignant transformation, such as keratin K13 and P-Akt, even in the few SCCs obtained in this genotype (Supplementary Fig. S3). In summary, these results indicate that IKKβ overexpression leads to the development of fewer and less malignant tumors in mice lacking skin p53. Western blot analysis of SCCs from both genotypes confirmed that tumors originated from cells lacking p53, as expected (Fig. 3H). In addition, p53^EKO/K5-IKKβ carcinomas showed increased p19 and p16 levels, in accordance with the results shown in Fig. 2A for cultured keratinocytes, and also of p21. When we studied Stat-3 and Akt, two proliferative pathways frequently activated in carcinogenesis, we did not find marked differences between p53^EKO and p53^EKO/K5-IKKβ tumors. Altogether, these results indicate that the tumor-suppressive activity of IKKβ in skin is not mediated by p53.

In addition to the DMBA/TPA carcinogenesis protocol, we also studied the combined effect of IKKβ overexpression and p53 absence over spontaneous tumorigenesis. We generated additional cohorts of p53^EKO and p53^EKO/K5-IKKβ mice, followed tumor appearance over time, and classified histologically the tumors generated (Table 1A). Skin tumors were mainly SSCs of different degrees of differentiation. Interestingly, the percentage of mice with tumors was lower in p53^EKO/K5-IKKβ than in p53^EKO mice, and the mean number of tumors per mice with tumors was also lower in p53^EKO/K5-IKKβ mice (1.4 vs. 1.8). Furthermore, the percentage of skin tumors was roughly half in mice overexpressing IKKβ (38% vs. 78%; P < 0.001). Collectively, these results reinforce, in a different experimental in vivo situation, the protective effect of IKKβ in skin cancer and confirm its independence from p53.

We next wondered whether the tumor-suppressive activity of IKKβ could be mediated by p16 and p19, which are coded by the same Ink4a/Arf locus. Heterozygous mice for an Ink4a/Arf^KO allele (24) were crossed with K5-IKKβ mice and backcrossed with FVB mice for three generations; we obtained mice homozygous for the Ink4a/Arf^KO allele (lacking p16 and p19 in all their cells), and simultaneously bearing the K5-IKKβ transgene. We performed a DMBA/TPA skin carcinogenesis experiment in 11 Ink4a/Arf^KO mice and 12 Ink4a/Arf^KO/K5-IKKβ mice. This experiment could not be extended for more than 14 weeks after DMBA application due to humanitarian reasons, given the fast growth or the malignant aspect of some of the tumors. Interestingly, in the absence of the proteins coded by the locus Ink4a/Arf, there were no differences in tumor burden between animals carrying or lacking the K5-IKKβ transgene (Fig. 4A). The histopathologic classification of the tumors analyzed (121 in total) is shown in Fig. 4B. Of note, due to the absence of p16 and p19, and in spite of the relatively short period of time of the experiment, the majority of the tumors were SCCs of varying levels of differentiation. In contrast, K5-IKKβ mice bearing two wt copies of the Ink4a/Arf locus resulted in a marked lower amount of SCCs that arose at later times than SCCs in Ink4a/Arf^KO background. In addition, all the SCCs observed in Ink4a/Arf ^wt/wt were well differentiated SCCs (Supplementary Fig. S4A–S4C).

Interestingly, in contrast to the results obtained in wt and p53^EKO backgrounds (Figs. 1 and 3), malignancy was augmented in Ink4a/Arf^KO/K5-IKKβ tumors, as the percentage of undifferentiated SCCs and spindle cell SCCs (SpSCCs) was much higher in Ink4a/Arf^KO/K5-IKKβ mice than in Ink4a/Arf^KO mice (Fig. 4B; P < 0,0001). Representative examples of SCCs with increased levels of malignancy, from well-differentiated SCCs to highly undifferentiated cutaneous SpSCCs are shown in Fig. 4C–H. SpSCC is a rare variant of SCC in which tumor cells with spindle-shaped appearance infiltrate the dermis diffusely, without formation of epithelial nests or cords (34); this tumor type is so malignant and undifferentiated that determination of its origin is sometimes doubtful as spindle keratinocyte appearance is very similar to that of malignant cells from fibrosarcomas or melanomas. We have studied the expression patterns of keratin K5, vimentin, and protein S100 (epithelial, mesenchymal, and melanocytic cell markers, respectively), to ensure the epidermal origin of uncertain tumors containing spindle cells and in some other tumors with clear classification; we analyzed 19 tumors from 9 Ink4a/Arf^KO/K5-IKKβ mice and 8 tumors from 4 Ink4a/Arf^KO mice (Table 2). Typically, SpSCCs coexpress K5 and vimentin (Fig. 4I and J) with the absence of S100 (Fig. 4K; compare with the staining in a positive control shown in Supplementary Fig. S5A and S5B), which differentiate them from fibrosarcomas (positive for vimentin but negative for keratins) and spindle melanomas, tumors that have been described in Ink4a/Arf^KO mice. The low percentage of fibrosarcomas in comparison with the data published in ref. 24 could be due to differences in genetic background, latency time, or tumorigenesis treatment (DMBA/UV in ref. 24 and DMBA/TPA in this work).

It is worth noting that we have been able to determine the follicular origin of SpSCCs, as K5-positive epidermal follicular cells infiltrate the underlying dermis, showing highly dysplastic spindle shape cells and giving rise to a SpSCC (Supplementary Fig. S5C). We have not found consistent differences in the activity of Stat-3, Erk, or Akt pathways between tumor histologic types or genotypes (Supplementary Fig. S5D). Spontaneous tumorigenesis affected slightly more strongly the Ink4a/Arf^KO/K5-IKKβ than the control Ink4a/Arf^KO mice, as a higher percentage of mice developed tumors (Table 1B). In addition, although the number of tumors is low, a tendency to develop more malignant skin tumors was also observed (as 2 of 6 skin tumors in Ink4a/Arf^KO/K5-IKKβ were highly malignant SpSCCs, while this tumor type was not observed in Ink4a/Arf^KO mice), reinforcing the data obtained in two-stage skin carcinogenesis experiments.

Taken together, these results indicate that the antitumoral effect of IKKβ is mediated by p16 or p19. Even more, in the absence of these tumor suppressors, IKKβ is not able to protect against skin tumor appearance, but instead it seems to increase tumor malignancy, as described for other tissues.

IKKβ is a fundamental multitarget kinase (5), and mice lacking IKKβ in all their cells die during gestation due to excessive hepatic apoptosis (35, 36); the absence of IKKβ in epidermis and stratified epithelia results in death at the beginning of postnatal life by an inflammatory skin disease mediated by TNF (37). In humans, however, IKKβ deletion seems to be less harmful, as homozygous deletion of the IKBKB gene is not lethal, at least in some cases, but leads to immunodeficiency (38). In agreement with the multiple functions of IKKβ, its effect on tumoral transformation of the cells can be mediated by different pathways, such as NF-κB activation (39); adaptation to metabolic and oxidative stresses, increasing the ability of cancer cells to survive at low glutamine concentration (40); establishment of stemness properties by Wnt pathway regulation (41); modulation of T-cell–dependent antitumoral immune response (42); or inactivation of important cell-cycle controllers as p53 (2) and p16 (3) among others. As a result of these pleiotropic functions, IKKβ promotes cellular transformation and tumor growth in lung, pancreas, and oral epithelia, among others (16, 17, 19), but it also results in an antitumoral effect in myeloid cells during melanoma tumorigenesis (18). There are two recent reports that illustrate clearly this context-dependent pro- or antitumoral dual activity of IKKβ, as they report that IKKβ in mesenchymal cells has both tumor-promotor and tumor-suppressor roles in intestinal tumorigenesis, probably as a consequence of subtle differences in the subpopulations targeted in these experiments (43, 44). Here, we have found a surprising antitumoral role of epidermal IKKβ in nonmelanoma skin cancer. This activity does not seem to be exerted by modifications in the epidermal or dermal inflammatory milieus, nor by a diminution in the amount of stem cells. As skin-specific p65 knockout leads to downregulation of NF-κB and also protects against tumoral transformation (8), the effect of IKKβ overexpression over skin cancer is probably mediated mainly by NF-κB–independent pathways. When we searched for possible mediators of this skin tumor–protective function of IKKβ, we found that K5-IKKβ epidermal keratinocytes expressed high levels of several tumor suppressor proteins (p53, p16, and p19) that could explain the reduced sensitivity to tumoral transformation of IKKβ-overexpressing skin.

Experimental skin carcinogenesis in animals lacking epidermal p53 indicated that the antitumoral activity of IKKβ in skin is independent of p53; so, probably the increased p53 protein level found in K5-IKKβ keratinocytes is secondary and not directly caused by IKKβ. Interestingly, the antitumoral effect of IKKβ is even stronger in a p53^EKO background than in a p53^wt background (Figs. 1A and 3A); this could be explained by the mutual inhibitory activity described for IKKβ and p53, that would weaken the effect of IKKβ in the presence of p53 (2, 45).

IKKβ protection against skin cancer, on the contrary, acts through p16, p19, or both. Although some of the activities of p16 and p19 proteins are different, and even opposite (46), both of them have strong and nonredundant tumor suppressor activity. The lack of either p16 or p19 alone results in increased cancer susceptibility, and is less harmful than the combined absence of both proteins (47). Favoring the possible leading role of p19 over p16 as mediator of skin IKKβ antitumoral activity are the facts that p19 is expressed at higher level than p16 in skin keratinocytes in qRT-PCR experiments and that mice with one of these proteins knocked-out confirm the higher general importance of p19 (47). If the main actor in IKKβ-mediated tumor protection were p19, it would exert its antitumoral function by p53-independent functions, as p53 absence does not diminish the skin antitumor activity of IKKβ. In any case, the generation of K5-IKKβ mice lacking p16 or p19 individually and the study of their sensitivity to skin cancer would be required to clarify which of these proteins mediates IKKβ skin-suppressive cancer functions.

It is interesting to highlight that skin overexpressing IKKβ gives rise, in the absence of p16 and p19, to SpSCCs, both after DMBA/TPA treatment and spontaneously (Fig. 4B; Table 1B). Spindle-cell SCC is a rare type of skin tumor, more frequent in immunosuppressed individuals. It has been described, in two-stage chemical skin carcinogenesis, that SpSCC is associated to lower expression level or to deletion of p16 and p19 (28). Our results also suggest that this malignant carcinoma is favored by increased expression of IKKβ, as it is found preferentially in Ink4a/Arf^KO/K5-IKKβ.

The generation in recent years of a growing body of expression and genomic data from tumoral samples supports the idea that IKKβ is a truly important molecule in human cancer development. So, the data included in the catalogue of somatic mutations in cancer (COSMIC; http://cancer.sanger.ac.uk/cosmic) indicate that IKKβ is frequently overexpressed (sometimes in association with copy number gain) in several human cancers, mainly in esophagus (24.8% of tumors), large intestine (18.5%), and breast (9.6%). Interestingly, IKKβ is also overexpressed in more than 7% of the samples of malignant melanomas and of stomach, ovary, and lung cancers; unfortunately, there are not data available for NMSC. Data in The Cancer Genome Atlas (TCGA) indicate that IKKβ is altered (mostly amplified) in around 5% of head and neck squamous carcinomas; in 5%–10% of the cases of ovarian serous cystadenocarcinoma, colorectal and stomach adenocarcinomas, liver hepatocellular adenocarcinoma, uterine corpus endometrial carcinoma, lung adenocarcinoma, lung SCCs, and esophageal carcinoma; and over 10% of the cases in breast invasive carcinoma, and bladder urothelial carcinoma. Interestingly, TCGA data indicate that IKKβ is not altered in NMSC, although the number of samples analyzed is low (48, 49). Overall, these data confirm and expand the growing body of evidence obtained in genetically modified animal models, indicating that IKKβ overexpression favors tumor growth in many cells types, but not in epidermal keratinocytes, where IKKβ prevents tumor development. This represents a warning note against the proposed treatment of inflammatory and tumoral diseases with IKKβ inhibitors (50), as ubiquitous pharmacologic IKKβ inhibition could favor skin tumor appearance.

In summary, we have found and described for the first time a strong skin tumor–suppressive function of IKKβ. Mechanistically, this antitumoral activity is mediated by p16 or p19, but not by p53. In addition, IKKβ seems to cooperate with lack of INK4A/ARF in the genesis of SpSCCs. Overall, these results draw attention to the need of a careful evaluation of therapies aimed to IKKβ inhibition in the treatment of inflammatory and tumoral diseases.

No potential conflicts of interest were disclosed.

The results shown here are in part based upon data generated by the TCGA Research Network (http://cancergenome.nih.gov/).

Conception and design: A. Page, A. Ramirez

Development of methodology: A. Page, A. Bravo, C. Lorz, C. Segrelles, J.M. Paramio, A. Ramirez

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Page, A. Bravo, C. Suarez-Cabrera, J.P. Alameda, M.L. Casanova, J.C. Segovia, A. Ramirez

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Page, A. Bravo, C. Suarez-Cabrera, J.P. Alameda, M.L. Casanova, J.C. Segovia, J.M. Paramio, M. Navarro, A. Ramirez

Writing, review, and/or revision of the manuscript: A. Page, A. Bravo, J.C. Segovia, J.M. Paramio, M. Navarro, A. Ramirez

Study supervision: A. Ramirez

Other (histopathological study of the samples, diagnosis and classification of tumors, and obtaining microphotographs): A. Bravo

The authors would like to thank Rebeca Sanz, Berta Hernanz, Nerea Guijarro for their valuable technical help; Rebeca Sánchez-Domínguez and Omaira Alberquilla for their help with the flow cytometry studies; Federico Sánchez-Sierra and Pilar Hernández for excellent histologic processing of the samples; and Edilia de Almeida and the personnel of the CIEMAT Animal Unit for mice care and for their help with mice treatments. We also thank Manuel Serrano (Centro Nacional de Investigaciones Oncológicas, CNIO, Spain) for his generous gift of Ink4a/Arf^KO mice and Anton Berns (Netherlands Cancer Institute, NKI, the Netherlands) for supplying p53^EKO mice.

This research was supported partially by funds from Fondo Europeo de Desarrollo Regional (FEDER) and by grants from the Spanish government (PI14/01403, to A. Ramírez; PI16/00161, to M.L. Casanova; SAF2015-66015-R, ISCIII-RETIC RD12/0036/0009, PIE 15/00076, and CB/16/00228, to J.M. Paramio; and RD16/0011/0011, SAF2014-54885-R, and RTC2015-3393-1, to J.C. Segovia).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.