Reduction in IκB Kinase α Expression Promotes the Development of Skin Papillomas and Carcinomas

We reported recently a marked reduction in IκB kinase α (IKKα) expression in a large proportion of human poorly differentiated squamous cell carcinomas (SCC) and the occurrence of Ikkα mutations in human SCCs. In addition, overexpression of IKKα in the epidermis inhibited the development of skin carcinomas and metastases in mice. However, whether a reduction in IKKα expression promotes skin tumor development is currently unknown. Here, we assessed the susceptibility of Ikkα hemizygotes to chemical carcinogen-induced skin carcinogenesis. Ikkα^+/− mice developed 2 times more papillomas and 11 times more carcinomas than did Ikkα^+/+ mice. The tumors were larger in Ikkα^+/− than in Ikkα^+/+ mice, but tumor latency was shorter in Ikkα^+/− than in Ikkα^+/+ mice. Some of the Ikkα^+/− papillomas and most Ikkα^+/− carcinomas lost the remaining Ikkα wild-type allele. Somatic Ikkα mutations were detected in carcinomas and papillomas. The chemical carcinogen-induced H-Ras mutations were detected in all the tumors. The phorbol ester tumor promoter induced higher mitogenic and angiogenic activities in Ikkα^+/− than in Ikkα^+/+ skin. These elevated activities were intrinsic to keratinocytes, suggesting that a reduction in IKKα expression provided a selective growth advantage, which cooperated with H-Ras mutations to promote papilloma formation. Furthermore, excessive extracellular signal-regulated kinase and IKK kinase activities were observed in carcinomas compared with those in papillomas. Thus, the combined mitogenic, angiogenic, and IKK activities might contribute to malignant conversion. Our findings provide evidence that a reduction in IKKα expression promotes the development of papillomas and carcinomas and that the integrity of the Ikkα gene is required for suppressing skin carcinogenesis. [Cancer Res 2007;67(19):9158–68]

We recently reported somatic IκB kinase α (Ikkα) mutations in human squamous cell carcinomas (SCC) and a marked reduction in IKKα expression in poorly differentiated human and mouse cutaneous SCCs (1), which highlights the importance of IKKα in human skin cancer. However, the natural role for IKKα in skin tumor development is unclear. The animal model provides an appropriate tool to address these questions.

IKKα is one subunit of the IKK complex, which is central for nuclear factor-κB (NF-κB) activation (2). Its involvement in the development of lymphoid organs and innate immunity requires IKK/NF-κB activity (3, 4). IKKα also plays an essential role in the formation of the epidermis during embryonic development in mice (5–7). Ikkα^−/− mice develop a striking hyperplastic epidermis that lacks terminal differentiation and these mice die at birth (5–7). Ikkα^−/− keratinocytes and skin preserve IKK/NF-κB activity (5, 8). Reintroduction of IKKα or kinase-inactive IKKα induced terminal differentiation in Ikkα^−/− keratinocytes, but reintroduction of IKKβ, RelA p65, or IκBα (an inhibitor for NF-κB) failed to do so (8). Furthermore, Sil et al. (9) reported that IKKα or kinase-inactive IKKα transgene driven by the keratin 14 promoter rescued the skin phenotype of Ikkα^−/− mice. These results suggest that the function of IKKα in determining the epidermal formation is IKK/NF-κB independent.

We reported recently that Lori.IKKα transgenic mice developed normal skin and that the elevated IKKα in the suprabasal epidermis enhanced terminal differentiation markers (1). These transgenic mice developed significantly fewer malignant carcinomas and metastases than did wild-type (WT) mice when they were challenged with the chemical carcinogen 7,12-dimethylbenz(a)anthracene (DMBA) and tumor promoter 12-O-tetradecanoylphorbol-13-acetate (TPA). The mitogenic and angiogenic activities were reduced in the DMBA/TPA–treated skin of Lori.IKKα mice relative to those in the DMBA/TPA–treated skin of WT mice. These results suggest that elevated IKKα expression antagonizes chemical carcinogen-induced mitogenesis and angiogenesis, thereby repressing the development of malignant carcinomas and metastases. However, the effect of reduced IKKα on skin tumor development is unknown.

Ras plays a prominent role in the development of human SCCs and H-Ras mutations are frequently found in human SCCs (10, 11). Ras activation is required for chemical carcinogen-induced skin carcinogenesis in mice. The carcinogen DMBA causes activating H-Ras mutations, and the tumor promoter TPA expands the population of Ras-initiated cells (12, 13). Most papillomas eventually regress, and only a few become carcinomas in mice with a C57BL6 or a C56BL/129/Sv background (14, 15). Mice lacking H-Ras developed significantly fewer skin tumors than did WT mice induced by DMBA/TPA (16). In the present study, we tested the susceptibility of Ikkα hemizygotes to DMBA/TPA–induced skin carcinogenesis. Ikkα^+/− mice developed 2 times more benign tumors (papillomas) and 11 times more malignant carcinomas than did WT mice. Furthermore, we found that most Ikkα^+/− carcinomas and some Ikkα^+/− papillomas lost the remaining WT Ikkα allele. These findings show that reduction in IKKα expression promotes papilloma formation and malignant conversion.

Mice and skin carcinogenesis. All the mice used in this study were cared for in accordance with the guidelines of our institution's Animal Care and Use Committee (animal protocol 04-01-05732). DMBA (Sigma-Aldrich, Inc.) and TPA (LC Laboratories) were used to treat mice. Six- to 8-week-old female Ikkα^+/+ and Ikkα^+/− mice with a C57BL6 background (C57BL6/129/Sv mice were backcrossed with C57BL6 mice thrice) were topically treated with 100 μg DMBA in 200 μL acetone; 2 weeks later, the mice were treated with 2.5 μg TPA in 200 μL acetone five times weekly for 28 weeks. For controls, 15 Ikkα^+/+ and 15 Ikkα^+/− mice were treated with 100 μg DMBA in acetone (no TPA) and 15 Ikkα^+/+ and 15 Ikkα^+/− mice were treated with 2.5 μg TPA in acetone (no DMBA) five times weekly for 28 weeks. For bromodeoxyuridine (BrdUrd; Sigma-Aldrich) incorporation experiment, mice were treated with 2.5 μg TPA in 200 μL acetone or 200 μL acetone alone as a control five times weekly for 4 weeks or with 100 μg DMBA in 200 μL acetone once. BrdUrd (0.05 mg/g body weight) was administered i.p. to mice 1 h before they were killed. Skin specimens were embedded in paraffin and then sectioned, and BrdUrd labeling was counted in 1,000 cells in the basal epidermis. Immunohistochemical BrdUrd detection was done in our Histology and Tissue Core.¹

Paraffin-embedded sections of tumor tissue and skin were prepared, and routine H&E and immunohistochemical staining of these sections with antibodies against proliferating cell nuclear antigen (PCNA) and IKKα were done in our Histology and Tissue Core.

Detection of H-Ras (V61 and V12) mutations. Analyses for H-Ras mutations were done in our Molecular Biology Core.²

PCR primers for H-Ras codon 61 mutation detection were 5′-ggtgtaggctggttctgtggattctc-3′ and 5′-gcacacggaaccttcctcac-3′, which generated a 329-bp band that was gel purified and used for a second PCR round. PCR primers for the second round were 5′-tgtggattctctggtctgaggagag-3′ and 5′-cataggtggctcacctgtactgatg-3′, which produced a 269-bp fragment. The PCR products were digested with XhaI. PCR primers for H-Ras codon 12 mutation detection were 5′-cctggattggcagccgctgt-3′ and 5′-tcatactcgtccacaaagtg-3′, which generated a 125-bp fragment. The fragments were digested with MnII.

Real-time PCR. Total RNA was isolated from skin, tumors, and keratinocytes by TRI reagent (Molecular Research Center, Inc.), and cDNA was synthesized by a RETROscript kit (Ambion, Inc.). The PCR primers and probes were purchased from Applied Biosystems [tumor necrosis factor α (TNFα), Mm00443258_m1; interleukin-1α (IL-1α), Mm00439620_m1; transforming growth factor α (TGFα), Mm00446231_m1; vascular endothelial growth factor-A (VEGF-A), Mm00437304_m1; epidermal growth factor (EGF), Mm00438696_m1; amphiregulin, Mm00437583_m1; heparin-binding EGF (HB-EGF), Mm00439307_m1; fibroblast growth factor 2 (FGF2), Mm00433287_m1; and FGF13, Mm00438910_m1]. The reactions were done according to the manufacturer's instructions. Each cDNA sample was analyzed in triplicate by an ABI 7900 sequence detector (Applied Biosystems). Thermal cycling was done as follows: 1 cycle at 50°C for 2 min and 95°C for 10 min and 40 cycles at 95°C for 15 s and 60°C for 1 min. Data were analyzed by the Prism Dissociation Curve software program (SDS 2. 2. 2, Applied Biosystems). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control.

Southern blot analysis. DNA was isolated from tumors and normal skin by an extraction kit (Promega). DNA (10 μg) was digested with BamHI overnight, applied to a 0.9% agarose gel, transferred to blotting membranes (Zeta-Probe GT, Bio-Rad), and fixed by UV light. The DNA blotting membranes were hybridized with an NH₂-terminal IKKα cDNA (nucleotides 1–575) probe and a 400-bp NH₂-terminal GAPDH cDNA probe as a control. Southern blotting was done according to the manufacturer's instructions (Bio-Rad).

Western blotting. A cell lysate (40 μg) was applied to SDS gel, and specific protein levels were measured by Western blotting as described previously (8) with the following antibodies against IKKα (Imgenex), phosphorylated extracellular signal-regulated kinase (p-ERK; Cell Signaling Technology), ERK1 (Santa Cruz Biotechnology), ERK2 (Santa Cruz Biotechnology), and β-actin (Sigma).

Keratinocyte culture. Mouse primary keratinocytes were isolated and cultured as described previously (8). Briefly, skin specimens isolated from newborn mice or E19 embryos were treated with 0.25% trypsin (Life Technologies) for 8 to 10 h at 4°C; the epidermis was separated from the dermis. Isolated keratinocytes were plated on 60-cm cell dishes containing keratinocyte serum-free medium (Life Technologies).

Ikkα mutation analysis. Total RNA was isolated from skin or tumors with TRI reagent, and cDNA from these samples was synthesized by a RETROscript kit. The PCR fragments were generated by an expanded high-fidelity^plus PCR system (Roche Diagnostics GmbH) with primers (IKKα, 5′-ccattcactattctgaggttggtgtc-3′ and 5′-tactggaggggttactgtgccttc-3′). The PCR products were subcloned into pGEM-T vectors (Promega) and sequenced (1). The sequences were compared with the National Center for Biotechnology Information, gi 6680941.

Chromatin immunoprecipitation assay. Chromatin immunoprecipitation (ChIP) assay was done following the manufacturer's instructions (Upstate). Keratinocytes were starved for 24 h (no growth factors) and cells were treated with EGF (10 ng/mL) for 24 h or with TPA (48 nmol/L) for 2 h. After cells were fixed, washed, and collected, their DNA was sheared by sonication in 200 μL lysis buffer. Then, 20 μL sonicated DNA from each sample was purified and used as an input DNA control. The rest of the DNA was precleared. The chromatin DNA was precipitated with an anti-IKKα antibody and protein A/G beads, and the DNA was eluted from the beads. The precipitated DNA was analyzed by PCR for 1 cycle at 95°C for 5 min; 35 cycles at 95°C for 20 s, 55°C for 20 s, and 72°C for 30 s; and 1 cycle at 72°C for 10 min. The PCR primers for proximal VEGF-A promoters (−903 to −1,233 bp) were 5′-gtgttcctgagcccagtttgaag-3′ and 5′-agtccgctgaatagtctgccttg-3′ and the primers for distal VEGF-A promoter (−2,414 to −2,065 bp) were 5′-tgggttagaggtgggggttttg-3′ and 5′-aactgaagccagggtgccaatg-3′. PCR primers for analyzing the mRNA level of VEGF-A (491 bp) were 5′-gcacccacgacagaaggagagcaga-3′ and 5′-cgccttggcttgtcacatctgcaa-3′.

Ikkα^+/− mice develop many more and larger skin tumors earlier than do Ikkα^+/+ mice. To determine the relationship between IKKα expression and skin carcinogenesis, we topically treated Ikkα^+/+ and Ikkα^+/− mice with a single dose of DMBA followed by repeated treatment with TPA. At week 28 of TPA treatment, we collected and weighed the tumors. Ikkα^+/− mice developed significantly more tumors than did Ikkα^+/+ mice (P < 0.0005, linear regression model; Fig. 1A; Supplementary Table S1A). Incidence of tumors in the Ikkα^+/+ mice reached 100% by 19 weeks and in the Ikkα^+/− mice by 14 weeks (P = 0.0014, log-rank test; Fig. 1B). Papillomas and carcinomas in Ikkα^+/− mice were significantly larger than those in Ikkα^+/+ mice (P < 0.0005, two-sided Fisher's exact test; Fig. 1C and D). Control mice treated with DMBA once or TPA (five times weekly) alone did not develop any tumors at week 28. The repeated experiment 2 showed similar results (Supplementary Fig. S1A and B; Supplementary Table S1B). To determine the rate of carcinoma conversion, we histologically examined all the tumors that weighed >0.03 g. Four of the 125 (3.2%) tumors that developed in the 15 Ikkα^+/+ mice and 47 of the 285 (16.5%) tumors that developed in the 15 Ikkα^+/− mice were carcinomas (P < 0.005, Fisher's exact test; Fig. 1C). These results suggested that Ikkα^+/− mice were far more susceptible to skin carcinogenesis than Ikkα^+/+ mice. Genetic background affects the susceptibility of mice to chemical carcinogen-induced skin carcinogenesis (15). Although the mice used for this study were not 100% C57BL6, Ikkα^+/− and Ikkα^+/+ mice were sisters. Mice with a mixed C57BL6 genetic background have been used for the studies of chemical carcinogen-induced skin carcinogenesis (14, 17). In addition, the differences in the tumor incidence and the tumor multiplicity between Ikkα^+/+ group and Ikkα^+/− group were dramatically significant. Thus, it was likely that the genetic background was not a major factor that caused these differences. We also examined apoptotic cells in tumors from Ikkα^+/+ and Ikkα^+/− mice by immunostaining analysis and did not observe a significant difference for this end point between Ikkα^+/+ and Ikkα^+/− tumors (data not shown).

H-Ras mutations are detected in papillomas and carcinomas. The papillomas in the Ikkα^+/+ and Ikkα^+/− mice displayed well-differentiated patterns with a thick surface layer of keratin. The carcinomas were less differentiated or undifferentiated in the Ikkα^+/+ and Ikkα^+/− mice (Fig. 2A). The carcinomas expressed PCNA in the entire tumor, but the papillomas expressed PCNA only in the hyperproliferating basal cells (Fig. 2A).

To determine whether these skin tumors were initiated by DMBA-mediated H-Ras mutations, we examined V61 mutations (CAA→CTA) of H-Ras in randomly selected papillomas and carcinomas from Ikkα^+/+ and Ikkα^+/− mice by using PCR. The H-Ras (V61) mutations were detected in 19 of 23 (82.6%) Ikkα^+/+ tumors and in 24 of 28 (85.7%) Ikkα^+/− tumors (Fig. 2B). Moreover, we found V12 mutations (GGA→GGC) in the tumors that were negative for V61 mutations of H-Ras (Fig. 2C). We did not detect any H-Ras mutations in seven untreated skin specimens of Ikkα^+/+ and Ikkα^+/− mice, respectively, and in eight DMBA/TPA–treated skin specimens of Ikkα^+/+ and Ikkα^+/− mice, respectively. These results suggested that all the skin tumors were initiated by DMBA-mediated H-Ras mutations.

Levels of IKKα proteins are markedly reduced in most Ikkα^+/− carcinomas and some Ikkα^+/− papillomas. We further evaluated the IKKα level in papillomas and carcinomas. Western blotting analysis showed higher levels of IKKα in 10 papillomas than in skin specimens from Ikkα^+/+ mice (Fig. 3A). The increased IKKα could be induced by long-term treatment with TPA, as suggested by Saleem et al. (18). We reported previously that poorly differentiated SCCs expressed markedly reduced IKKα (1). Reduction in IKKα was also detected in the 8 poorly differentiated carcinomas of the 15 Ikkα^+/+ carcinomas (Fig. 3A; Supplementary S2A and B). Furthermore, some of the 17 Ikkα^+/− papillomas expressed elevated IKKα, but the remainder expressed reduced IKKα (Fig. 3A). All 18 Ikkα^+/− carcinomas expressed markedly reduced IKKα. Comparison of IKKα relative levels suggested that the IKKα levels were inversely correlated with the increased numbers of carcinomas (Fig. 3B). Immunohistochemical staining showed a reduction in IKKα expression only in carcinomas, not in the skin that surrounded the tumors (Fig. 3C). Collectively, these results suggested that a reduction in IKKα expression contributed to the formation of papillomas; a further reduction greatly enhanced the development of carcinomas, although the mechanism by which IKKα was reduced might be different in Ikkα^+/+ and Ikkα^+/− tumors.

Genetic alterations of Ikkα occur in skin tumors. We reported previously that the poorly differentiated carcinomas expressed reduced IKKα proteins; however, IKKα mRNA levels were not reduced in these carcinomas obtained from WT mice (1), and we observed similar results in this study (Supplementary Fig. S2B). In addition, we observed that papillomas and carcinomas from Ikkα^+/− mice tended to lose IKKα (Fig. 3A), suggesting that loss of heterozygosity (LOH) of Ikkα might be responsible for loss of IKKα protein in Ikkα^+/− tumors. To test this hypothesis, we examined the BamHI-digested Ikkα gene in papillomas and carcinomas by Southern blotting with an NH₂-terminal IKKα cDNA probe covering exons 1 to 6 of Ikkα (Fig. 4A). This probe hybridized with three fragments, including a 9.3-kb fragment from exon 2 to upstream of the Ikkα gene, a 3.6-kb fragment from exons 3 to 4, and a 5.2-kb fragment from exons 5 to 9 in Ikkα^+/+ cells (Fig. 4B). In the Ikkα^−/− cells, the 5.2-kb fragment disappeared because exon 7 in the Ikkα allele included a neo gene that contained two additional BamHI sites at its 5′ and 3′ ends (Fig. 4A and B; ref. 5). After BamHI digestion, this probe only recognized a 1.3-kb fragment from exons 5 to 7 but not the fragment from exons 7 to 9. Thus, three DNA fragments (9.2, 3.6, and 1.3 kb) in Ikkα^−/− cells and four DNA fragments (9.3, 3.6, 5.2, and 1.3 kb) in Ikkα^+/− cells were detected. When Ikkα LOH occurred in Ikkα^+/− cells, only 9.3, 3.6, and 1.3 kb were detected. We found LOH in 4 of the 9 (44%) papillomas and 22 of the 23 (95%) carcinomas (Fig. 4B; Supplementary Table S2). The faint WT bands (5.2 kb) in Ikkα^+/− carcinomas were presumably caused by the presence of contaminating non-carcinoma cells in the tumor samples. We confirmed LOH in the Ikkα^+/− carcinomas by PCR (Supplementary Fig. S3A; ref. 5). Reverse transcription-PCR (RT-PCR) confirmed that Ikkα^+/− carcinomas retained low levels of IKKα mRNA (Supplementary Fig. S3B). Collectively, these results indicated that LOH of Ikkα might be largely responsible for loss of IKKα protein in Ikkα^+/− carcinomas and papillomas.

We then examined the Ikkα gene in 11 papillomas and 7 carcinomas from Ikkα^+/+ mice by Southern blotting with the IKKα cDNA probe and found no alterations in Ikkα (Fig. 4C; Supplementary Table S3). Ikkα mutations in human SCCs were reported previously (1). To identify mutations in the Ikkα murine gene, we used RT-PCR to amplify an IKKα transcript (nucleotides 1606–2090; amino acids 518–679 of IKKα; NM_007700) from either acetone-treated control skins or Ikkα^+/+ carcinomas. Amplified PCR products were subcloned into the pGEM vectors and sequenced. We sequenced 33 clones from two acetone-treated skin samples, which were used as PCR background controls (5 and 6 in Supplementary Table S4). Many Ikkα mutations, including insertions and deletions, were detected in 45 clones from four Ikkα^+/+ carcinomas (Fig. 4D; Supplementary Table S4). Ikkα mutations were also detected in Ikkα^+/− papillomas and Ikkα^+/− carcinomas (Supplementary Table S5). Interestingly, we detected many Ikkα mutations in the IKKα transcript of one Ikkα^+/− carcinoma (Supplementary Table S5). Ikkα LOH was not detected in all the Ikkα^+/− carcinomas (Supplementary Table S2). Thus, Ikkα mutations might also contribute to loss of IKKα proteins in carcinomas if some cells in carcinomas retained the remaining WT Ikkα allele. Apparently, there were more Ikkα mutations in carcinomas than in papillomas. These transition and transversion nucleotide substitutions caused missense and nonsense mutations (Fig. 4D). Although we detected some repeated mutations in the same tumors, the frequency of the repeated mutations was low, suggesting that these somatic mutations were not amplified, which was consistent with our previous results (1). In addition, sequencing results confirmed that the mutations were not homologous (Supplementary Fig. S3C). Collectively, Ikkα mutations were found in both human and mouse SCCs.

Reduction in IKKα expression enhances mitogenic activity in skin and carcinomas. A previous study suggested that ERK activity was a signature for Ras-initiated skin tumors in two-stage chemical carcinogenesis settings (17). We found H-Ras mutations in the tested papillomas and carcinomas. Because control mice treated with DMBA alone did not produce any tumors at 28 weeks, we investigated the effect of reduced IKKα on the promotion of TPA-induced tumors. Immunohistochemical staining showed a higher fraction of BrdUrd-positive cells in the epidermis of Ikkα^+/− mice than in the epidermis of Ikkα^+/+ mice treated with TPA (P = 0.009, Student's t test; Fig. 5A), suggesting that TPA enhanced cell proliferation in Ikkα^+/− skin. We next examined phosphorylation of ERK, a downstream target of Ras. ERK activity was higher in Ikkα^+/− than in Ikkα^+/+ cultured keratinocytes under routine culture conditions (Supplementary Fig. S4). ERK activity was much higher in TPA-treated Ikkα^+/− than in TPA-treated Ikkα^+/+ skin (Fig. 5B). Furthermore, examination of several growth factors that are downstream targets of TPA (12) by real-time PCR showed excessive expression of TGFα, EGF, amphiregulin, FGF2, FGF13, VEGF-A, and TNFα in TPA-treated Ikkα^+/− skin specimens relative to that in TPA-treated Ikkα^+/+ skin specimens (Fig. 5C). TPA moderately elevated expression of HB-EGF and IL-1 in Ikkα^+/− compared with that in Ikkα^+/+ skin. The basal levels of ERK activity and expression of these growth factors were slightly higher in Ikkα^+/− than in Ikkα^+/+ skin (Fig. 5B and C). Moreover, we isolated keratinocytes from Ikkα^+/+ and Ikkα^+/− newborn mice and found that TPA treatment induced significant higher levels of ERK activities and higher expression levels of the growth factors in Ikkα^+/− keratinocytes than in Ikkα^+/+ keratinocytes (Fig. 5D and E), suggesting that the increased mitogenic activity was intrinsic to keratinocytes. Western blotting showed considerably higher ERK activity in carcinomas than in papillomas (Fig. 5F). Collectively, these results indicated that reduced IKKα markedly elevated TPA-induced ERK activity and expression of these growth factors, which provided a molecular basis for promoting keratinocyte proliferation, papilloma formation, and malignant conversion in Ikkα^+/− mice.

Because IKKα is one subunit of the IKK complex, we then evaluated IKK activity. Kinase assay showed that IKK activity was only slightly increased in TPA-treated Ikkα^+/− compared with that in TPA-treated Ikkα^+/+ skin specimens; the levels of IKKβ were similar in these skin specimens (Supplementary Fig. S5). However, the IKK activity was higher in carcinomas than in papillomas, although the IKKβ level was not increased in carcinomas (Fig. 5G). The levels of IκBα were found to be lower in some carcinomas than in papillomas (Supplementary Fig. S6A). To determine whether IKKα loss deregulated expression of IKKβ and IκBα, we examined their levels in the epidermis. Western blotting showed that Ikkα^+/+ and Ikkα^−/− epidermis expressed similar levels of IKKβ, IκBα, and p65 (Supplementary Fig. S6B). Presumably, the replacement of IKKα by IKKβ in the IKK complex might contribute to the increased IKK kinase activity in IKKα-deficient cells as suggested by three previous reports (8, 19–21).

Loss of IKKα promotes VEGF-A expression and blood vessel formation. Our skin carcinogenesis experiments showed that Ikkα^+/− mice developed 11 times more carcinomas than did Ikkα^+/+ mice (Fig. 1C). VEGF-A, an important angiogenesis factor, promotes neovascularization and the onset of tumor invasion (22). We reported previously that overexpression of IKKα repressed Ras-induced VEGF-A expression; binding of IKKα to a distal VEGF-A promoter (−2,414 to −2,065 bp) was correlated to a decrease in VEGF-A expression (1). In the present study, we observed elevated VEGF-A expression in Ikkα^+/− compared with that in Ikkα^+/+ skin specimens and keratinocytes (Fig. 5C and E). It has been reported that EGF enhances VEGF-A expression (23). We thus examined whether reduced IKKα elevated VEGF-A expression and blood vessel formation. ChIP assay showed that EGF treatment reduced binding of IKKα to the distal VEGF-A promoter, which was associated with elevated EGF-A expression in Ikkα^+/+ keratinocytes (Fig. 6A–C). Reintroduced IKKα bound to the distal VEGF-A promoter and repressed VEGF-A expression in Ikkα^−/− keratinocytes (Fig. 6A–C). IKKα did not bind to the proximal VEGF-A promoter (Fig. 6A), which was consistent with our previous finding (1). Furthermore, we found that TPA and DMBA dramatically enhanced the formation of blood vessels in the skin of Ikkα^+/− mice relative to that in the skin of Ikkα^+/+ mice (Fig. 6D). Significantly more blood microvessels in the skin stroma of Ikkα^+/− mice than in the skin stroma of Ikkα^+/+ mice were also observed after treatment with DMBA/TPA for 14 weeks (Supplementary Fig. S7A and B). Thus, IKKα loss-enhanced VEGF-A expression might foster the development of carcinomas in Ikkα^+/− mice.

In the present study, we found that Ikkα^+/− mice developed many more and larger tumors than did Ikkα^+/+ mice; the latency period for tumor appearance was shorter in Ikkα^+/− mice than in Ikkα^+/+ mice (Fig. 1A–D), indicating that a reduction in IKKα expression provided a selective growth advantage, which promoted the formation of skin tumors. Elevated ERK activity has been suggested as a signature event for Ras-initiated skin tumors (17). For example, ERK activity and H-Ras mutations were not detected in skin carcinomas that lost the remaining WT Pten allele in Pten^+/− mice (17). H-Ras mutations were detected in all the tumors in our mice and elevated ERK activity was observed in carcinomas compared with those in papillomas, which suggested that reduced IKKα provided a selective growth advantage that cooperated with DMBA-induced H-Ras mutations to promote skin carcinogenesis. Our results further showed that TPA treatment induced excessive ERK activity and excessive expression of EGF, TGFα, amphiregulin, FGF2, FGF13, and VEGF-A in Ikkα^+/− skin specimens. The elevated mitogenic activity was intrinsic to keratinocytes, which was consistent with a previous study, in which the expression levels of a group of FGFs and ERK activity were dramatically higher in Ikkα^−/− skin and keratinocytes than in Ikkα^+/+ skin and keratinocytes and that reintroduction of IKKα or kinase-inactive IKKα repressed ERK activity and expression of FGFs (9). The mitogenic activity was associated with the nuclear function of IKKα (9). Thus, a reduction in IKKα expression induced excessive mitogenic activities following TPA treatment, which provided a molecular basis for promoting keratinocyte proliferation and papilloma formation. Elevated ERK activity and excessive expression of growth factors were reported previously to contribute to the malignant conversion during skin carcinogenesis (17, 24). In addition, IKKα loss enhanced expression of VEGF-A that was likely to be important for tumor invasion (Fig. 6A–D; Supplementary S7; ref. 25). Thus, IKKα loss-mediated mitogenic and angiogenic activities might facilitate malignant conversion as well.

We showed previously that elevated IKKα expression repressed chemical carcinogen-induced mitogenic and angiogenic activities in the epidermis and the dermis of Lori.IKKα transgenic mice (1). The IKKα transgenic mice developed significantly fewer malignant carcinomas and metastases than did WT mice. Here, we report that reduced IKKα expression promoted chemical carcinogen- or growth factor–induced mitogenic and angiogenic activities and that Ikkα^+/− mice were far more prone to skin carcinogenesis than were Ikkα^+/+ mice. These findings suggest that IKKα-associated mitogenic and angiogenic activities are important mechanism in repressing or promoting skin carcinogenesis.

Moreover, we observed elevated IKK kinase activity in carcinomas compared with that in papillomas, although the levels of IKKβ were not elevated in the carcinomas. In addition, the levels of IκBα were reduced in some carcinomas (Fig. 5G; Supplementary S6A). Loss of IKKα was not found to elevate the levels of IKKβ or down-regulate the levels of IκBα in the epidermis (Supplementary Fig. S6B). Previous findings suggested that IKKβ replacement for IKKα in the IKK complex elevated the IKK kinase activity in IKKα-deficient cells because IKKβ showed a stronger kinase activity for IκBs than did IKKα (8, 19–21). Thus, IKKα loss or other indirect causes might be involved in the IKK kinase activation in the carcinomas. Increased IKK kinase activity has been implied to promote tumor progression through a variety of avenues (26). Increased IKKβ-dependent IKK activity was reported to promote human cancer development and colitis-associated cancer in mice (27, 28). Reduced IκBα levels and increased NF-κB activities were observed previously in chemical carcinogen-induced carcinomas compared with those in papillomas (29). In addition, IKKα loss was reported to promote the expression of cytokines, such as TNFα (19), and we also detected elevated expression of TNFα in IKKα-deficient keratinocytes (data not shown). Thus, elevated IKK/NF-κB activity might promote skin carcinogenesis in Ikkα^+/− mice, although the detailed mechanisms remain to be elucidated.

We previously reported somatic Ikkα mutations in human SCCs (1). In this study, we also observed Ikkα mutations in carcinomas and papillomas, suggesting that the Ikkα gene might be a susceptible target for mutagenesis during skin carcinogenesis. Recently, Greenman et al. (30) reported that Ikkα (Chuk) mutations were frequently detected in human cancers after examining mutations in 518 genes. We found more transition mutations (T→C and A→G) than transversion mutations (Supplementary Tables S4 and S5), which was consistent with our previous report (1). Some of the mutations resulted in changed amino acids, which might destabilize IKKα or alter IKKα activity. Some of the mutations created the stop codon, which presumably generated a truncated IKKα protein. We also detected deletions and insertions in the IKKα transcripts, which caused frameshift mutations for IKKα. Seemingly, there were more Ikkα mutations in carcinomas than in papillomas, suggesting that the increased numbers of Ikkα mutations might contribute to destabilization of IKKα in skin carcinomas. Furthermore, we found some repeated mutations in different tumors, which might be “hotspots.” We also noticed more Ikkα mutations in DMBA/TPA–induced mouse SCCs than those in human SCCs (1), which might be due to repeated treatments with a high dose of TPA for inducing skin tumors in mice. It is known that TPA can increase the amount of intracellular oxidative damage that influences the potential for DNA damage in cells (31). However, we previously observed markedly reduced IKKα expression in poorly differentiated human SCCs (1). Possibly, in addition to Ikkα mutations, IKKα in human SCCs can be down-regulated by alternative pathways. For example, p63 was reported to regulate IKKα expression in the formation of the epidermis, and deregulated p63 expression was involved in skin tumor development (32–34). Oncogenic stress, mutations, or unbalanced alleles can cause LOH of genes in tumors (14, 35, 36). We observed that 44% of the Ikkα^+/− papillomas and 95% of the Ikkα^+/− carcinomas lost the remaining WT Ikkα allele. Ikkα^+/− mice developed 11 times more carcinomas than did Ikkα^+/+ mice. These results suggested that the integrity of the Ikkα gene was important for suppressing malignant conversion. The development of malignancies is complex; more mechanisms remain to be revealed.

Grant support: National Cancer Institute grants CA102510 and CA117314 (Y. Hu), CA105345 (S.M. Fischer), and CA16672 (comprehensive center grant) and National Institute of Environmental Health Science grant ES07784.

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.

We thank L. Schroeder and S. Hensley for analyzing Ikkα mutations; I.B. Gimenez-Conti and N.W. Abbey for doing immunohistochemical and histological examinations; H. Thames for conducting the statistical analyses; K. Bouic (University of California, San Diego, La Jolla, CA) for generating adenoviruses; M. Aldaz and M. MacLeod for helpful discussions; and V. Edwards for editorial assistance.