Tumor necrosis factor receptor 1-mediated signaling is required for skin cancer development induced by NF-κB inhibition

Abstract

NF-κB signaling plays an important role in skin development and epidermal growth control. Moreover, inhibition of NF-κB signaling in murine epidermal keratinocytes in vivo, by expression of a keratin 5 (K5)-directed superrepressor form of inhibitor of NF-κB (IκBα), results in an inflammatory response characterized by a massive dermal infiltration of neutrophils, epidermal hyperplasia, and a rapid development of aneuploid squamous cell carcinomas (SCC). We now show that by crossing K5-IκBα mice onto a tumor necrosis factor receptor 1(Tnfr1)-null background, both the inflammatory and the tumorigenic responses are blocked. The specificity of the block is illustrated by the fact that K5-IκBα mice lacking the IL-1 receptor type 1 (Il1r1) develop inflammation and squamous cell carcinomas. Reconstitution of lethally irradiated K5-IκBα/Tnfr1^-/- mice with Tnfr1^+/- bone-marrow does not induce the inflammatory or the tumorigenic phenotype, indicating a critical dependence on Tnfr1-mediated signaling in skin cells or nonimmune cells. Our results suggest a critical role of local Tnfr1-mediated signaling and associated inflammatory response cooperating with repressed keratinocyte NF-κB signaling in driving skin cancer development.

The Rel/NF-κB transcription factors have a central role in several cellular processes, including proliferation, cell adhesion, apoptosis, and regulation of the immune response (1). Under nonstimulating conditions, NF-κB is retained in the cytoplasm in an inactive form because of its interaction with the inhibitory proteins IκBs. In response to an activating stimulus, IκB is phosphorylated by an IκB kinase (IKK) complex, which targets it for degradation by the proteasome, releasing NF-κB, which translocates to the nucleus, where it regulates the transcription of target genes.

Evidence gathered from studies during recent years have shown that NF-κB has a growth inhibitory function in the skin. Initial studies showed that overexpression of the NF-κB subunits p50 or p65 in the basal layer of the murine epidermis by using a keratin 14 (K14) promoter leads to hypoplasia of the epidermis, whereas expression of a superrepressor form of the inhibitor of NF-κB type α (IκBα) results in hyperplasia (2). Although K14-IκBα mice die shortly after birth (day 7), we have shown that mice with keratin 5 (K5)-directed expression of IκBα (K5-IκBα) to the basal layer of the epidermis survive to adulthood (3). Not only do these mice develop hyperplasia of the epidermis, but the skin phenotype of K5-IκBα mice is also characterized by an intense neutrophil-dominated inflammation of the skin and an early development of squamous cell carcinomas (SCC).

Recent data suggest that human sporadic SCC may show a block in NF-κB signaling based on nuclear exclusion of the RelA NF-κB subunit (ref. 4 and M.v.H., unpublished results). Furthermore, it was recently shown that coexpression of a superrepressor form of IκBα and Ha-Ras results in neoplasia resembling invasive SCC in human keratinocytes transplanted to severe combined immunodeficient (scid/scid) mice, supporting the relevance of a NF-κB block in the induction of human SCC (4).

Inflammation of the skin with neutrophil infiltration and hyperproliferation is also seen in female heterozygous Ikkγ-deficient mice, which develop a condition similar to the human X-linked disorder Incontinentia Pigmenti (IP) (5, 6). IKKγ is the regulatory subunit of the IKK complex, which, in addition, contains the two catalytic subunits IKKα (IKK1) and IKKβ (IKK2) (1). It is now well established that human IP results from loss-of-function mutations in the IKKγ gene (7). Interestingly, subungual keratoacanthoma-like tumors and cases of SCC have been reported as late manifestations of IP (8-11).

The catalytic IKKβ subunit is necessary for activation of NF-κB in response to proinflammatory stimuli (1). Skin-specific deletion of the IKKβ gene, K14-Cre/Ikk2^FL/FL, was recently shown to result in an inflammatory phenotype with concomitant hyperproliferation of the epidermis (12), very similar to what is seen in the K5-IκBα mice. Because the K14-Cre/Ikk2^FL/FL mice die between day 7 and day 9 after birth, it is currently not known whether they, similar to K5-IκBα mice, are prone to develop SCC.

Besides inf lammation and hyperproliferation, an up-regulation of the proinflammatory cytokine tumor necrosis factor type α (Tnfα) in the skin is commonly observed in K5-IκBα, Ikkγ heterozygous females and K14-Cre/Ikk2^FL/FL mice (3, 5, 12). Here, we show that the Tnfα response is a major driving force of the disease in the K5-IκBα mice. Removal of Tnfr1 prevents both the development of inflammation and the hyperproliferation of the skin, in line with what was recently reported for K14-Cre/Ikk2^FL/FL mice (12). Importantly, we also show that the development of SCC in K5-IκBα/Tnfr1^-/- mice is abolished. Reconstitution of lethally irradiated K5-IκBα/Tnfr1^-/- mice with Tnfr1^+/- bone-marrow-derived cells does not cause reinduction of the inflammatory phenotype, hyperproliferation, or cancer development. Up-regulation of Tnfα can be detected in the skin of K5-IκBα/Tnfr1^-/- mice when compared with Tnfr^-/- mice, indicating that both the primary cells responsible for the up-regulation of Tnfα and the critical Tnfr1-responding cells reside in the skin. Our data show that local Tnfr1-mediated signaling and an associated inflammatory response cooperate with repressed keratinocyte NF-κB signaling in driving SCC development.

Materials and Methods

Transgenic Mice. FVB/N K5-IκBα mice were generated as described in ref. 3. These mice were crossed with C57BL/6 Tnfr1^-/- mice (Tnfrsfla^tmlMak kindly provided by Tak Mak, Advanced Medical Discovery Institute, Toronto, Canada). Intercross of F1 K5-IκBα/Tnfr1^+/- generated F2 K5-IκBα/Tnfr1^-/- and K5-IκBα/Tnfr1 wild-type (Tnfr1^+/+ and Tnfr1^+/-) littermates. Mice were killed by cervical dislocation at 3.5 weeks of age, and skin samples were taken. A total of 65 mice were analyzed: 15 K5-IκBα/Tnfr1^-/-, 15 K5-IκBα/Tnfr1^+/-, 15 K5-IκBα/Tnfr1^+/+, 10 Tnfr1^+/-, and 10 Tnfr1^-/-. No phenotypic differences were observed between K5-IκBα/Tnfr1^+/- and K5-IκBα/Tnfr1^+/+ mice; hence, they were treated as one group: K5-IκBα. The cross between FVB/N K5-IκBα mice and IL-1 receptor type 1 (Il1r1)^-/- mice (C57BL/6 Il1r1^tmllmx) was performed as described for the Tnfr1^-/- cross above. F2 littermates from two different litters were analyzed (n = 2). Mice were kept according to Swedish national requirements, and ethical permission was obtained for all animal manipulations.

Histology and Immunohistochemistry. Skin samples were fixed in 10% neutral buffered formalin overnight and embedded in paraffin. Sections were stained with hematoxylin/eosin for histological analysis. For immunohistochemistry, paraffin sections were deparaffinized in xylene and passed through a graded alcohol series. In most cases, the sections were microwaved in 10 mM sodium citrate buffer (SCB; pH 6.0) before incubation with antibody (indicated below). Antibodies and dilutions used were as follows: monoclonal rat anti-CD3 (NovoCastra, Newcastle upon Tyne, U.K.), 1:200 (SCB); polyclonal rabbit anti-Cdk4 (Santa Cruz Biotechnology), 1:200 (SCB); monoclonal rat anti-CD-45R/B220 (Pharmingen), 1:100; polyclonal goat anti-IL-1β (R & D Systems), 0.5 μg/ml (SCB); rabbit polyclonal anti-Ki67 (NovoCastra), 1:1000 (SCB); rabbit polyclonal anti-myeloperoxidase (DAKO), 1:4000 (SCB); polyclonal rabbit anti-phosphop44/42 mitogen-activated protein kinase (MAPK; Cell Signaling Technologies, Beverly, MA), 1:100 (SCB); polyclonal rabbit anti-active c-Jun N-terminal kinase (JNK) (pTPpY, Promega), 1:100 (SCB); and polyclonal goat anti-Pax-5 (Santa Cruz Biotechnologies), 1:2000 (SCB). Bound antibodies were visualized by diaminobenzidine, and sections were counterstained with hematoxylin.

Cytokine mRNA Arrays, RT-PCR, and Real-Time PCR Analysis. Total RNA was prepared from skin biopsies by using RNABee solution (Tel-Test, Friendswood, Texas) and was treated with RQ1 RNase-Free DNase (Promega) to remove contaminating genomic DNA. Five micrograms of RNA was used as a template for ³²P-labeled cDNA probe synthesis that was subsequently used in hybridization to a GEArray Q Series Mouse Common Cytokine Gene Array (SuperArray, Bethesda) according to the manufacturer's instructions. Radioactive signal intensities were analyzed on a PhosphorImager. The signal from expression of each cytokine gene was normalized to the signal derived from β-actin on the same array. One microgram of RNA was used for cDNA preparation by using SuperScript RNase H^- Reverse Transcriptase (Promega), which was subsequently used as a template for PCR or Real-Time PCR quantification. For PCR amplification, Taq Gold (Promega) was used. The primers and procedures are as follows: Il-1α: 5′-TGCCAT TGACCATCTCTCTCTG-3′ and 5′-TGGCAACTCCTTCAGCAACACG-3′; 94°C for 30 seconds, 54°C for 1 min, 60°C for 2 min, 25 cycles; Il-1β: 5′-GCAACTGTTCTGAACTCA-3′ and CTCGGAGCCTGTAGTGCAG-3′; 94°C for 30 seconds, 49°C for 1 min, and 60°C for 2 min, 37 cycles. PCR for mouse β-actin was run as a control as described in ref. 3. Real-Time PCR analysis for Tnfα expression was done by using an Assays-on-Demand gene expression kit (Assay ID Mm00443258-m1, Applied Biosystems). GAPDH was used as endogenous control (TaqMan Rodent GAPDH reagents, Applied Biosystems).

Bone Marrow Transplantation. FVB/N K5-IκBα mice were crossed with C57BL/6 Tnfr1^-/- mice. The F1 K5-IκBα/Tnfr1^+/- were backcrossed to C57BL/6 Tnfr1^-/- giving K5-IκBα/Tnfr1^-/-, K5-IκBα/Tnfr1^+/-, Tnfr1^+/-, and Tnfr1^-/- mice. K5-IκBα/nfr1^+/- mice developed inflammation of the skin within 1 month of age and could not be included in the experiment (data not shown). MHC class I typing for H-2D^b and H-2D^q/H-2L^q alleles was performed by fluorescent-activated cell sorter analysis on peripheral blood cells (KH95 and KH117, respectively; Pharmingen). Bone marrow-derived cells from Tnfr1^+/- or Tnfr1^-/- littermates were obtained by flushing femurs and tibias with PBS and transplanting 1 × 10⁶ cells i.v. into 1-month-old MHC-matched, natural killer (NK) cell-depleted and lethally irradiated mice (6.5 + 5.5 Gy, 5-h interval; n = 7 per group). Successful engraftment was verified by genotyping peripheral blood DNA. Skin biopsies for histopathology were taken 3 months after engraftment. NK cell depletion was performed by i.p. administration of 200 μg of purified anti-NK1.1 antibody (PK136; Pharmingen) in 200 μl of PBS 2 days before initiation of the experiment.

Results

Tnfr1 Signaling Is Required for Development of Inflammation and Epidermal Hyperplasia in K5-IκBα Mice. The development of inflammation and epidermal hyperplasia in K5-IκBα transgenic mice is associated with a strong up-regulation of Tnfα (3). To evaluate the role of Tnfα in the development of this disease, we crossed the K5-IκBα mice onto a Tnfr1-null background. The K5-IκBα mice developed macroscopic skin changes within 3 weeks of age, seen as verrucous keratinized lesions on the back (Fig. 1a). Histological analysis of dorsal skin from 3.5-week-old K5-IκBα mice showed severe focal changes with hyperplasia/dysplasia and hyperkeratosis of the epidermis also involving the hair follicles with a massive inflammatory dermal infiltrate consisting mainly of polymorphonuclear granulocytes (Fig. 1c). In contrast, age-matched K5-IκBα/Tnfr1^-/- mice showed no inflammatory skin changes (Fig. 1 a and c).

Fig. 1.

Tnfr1 signaling is required for development of inflammation and epidermal hyperplasia in K5-IκBα mice. (a) K5-IκBα/Tnfr1^-/- (Upper) versus K5-IκBα (Lower) mice. Mice were 3.5 weeks old. (b) Hematoxylin/eosin-stained SCC in a 3.5-week-old K5-IκBα mouse. Overview shows the histology of a clinically verrucous skin lesion. Shown at the left margin is a slightly hyperplastic epidermis, with gradual transition into irregular hyperplasia with dysplasia of both surface and follicular epithelium with concomitant hyperkeratosis. In the deep portion of the follicular structures, progression into infiltrative SCC occurs (boxed areas). The squamous cell populations in this area are characterized by increased pleomorphism and dyskeratosis. Suprabasal mitotic figures are noted. (c) Hematoxylin/eosin-stained dorsal skin sections from wild-type, K5-IκBα/Tnfr1^-/-, and K5-IκBα mice. (Scale bar, 100 μm.) (d) Ki-67 staining, labeling proliferating cells in wild-type, K5-IκBα/Tnfr1^-/-, and K5-IκBα dorsal skin. (Scale bar, 50 μm.)

In K5-IκBα mice, an increased proliferation of epidermal cells, including suprabasally located cells, was observed by immunohistochemical staining for the proliferation marker Ki67 (Fig. 1d). Wild-type and K5-IκBα/Tnfr1^-/- skin only showed sparse staining of basal keratinocytes (Fig. 1d). Thus, the inflammation and the epidermal hyperplasia observed in the K5-IκBα mice appears secondary to the Tnfr1-mediated signaling, similar to what was recently reported for K14-Cre/Ikk2^FL/FL mice (12).

Removal of Tnfr1 Abolishes the Development of SCC in K5-IκBα Mice. In >90% of the 3.5-week-old K5-IκBα mice analyzed, dysplastic changes were seen, which, in several animals, developed into infiltrative SCC (Fig. 1b). However, 1-year-old K5-IκBα/Tnfr1^-/- mice (n = 12) still showed no signs of tumor development. Our data show that not only does transfer of the K5-IκBα transgene onto a Tnfr1^-/- background prevent the inflammation and hyperproliferation of the skin, but it also abolishes the development of SCC. Data obtained here were from a mixed FVB/N-C57BL/6 background. We have also repeated the experiment on a FVB/N background (six backcrosses) with the same results (data not shown). It should be noted that the penetrance for SCC in K5-IκBα mice of either genetic background is 100% (3).

Massive Neutrophil Invasion and Presence of Plasmacytoid Dendritic Cells in Lesional Skin of K5-IκBα Mice. The inflammatory infiltrate observed in K5-IκBα mice is dominated by polymorphonuclear leukocytes, mainly neutrophils, which increase in number during disease progression (Fig. 2 and M.v.H., unpublished data). In contrast, T lymphocytes are only present in low numbers and do not change in number over time (Fig. 2 and M.v.H., unpublished data). In K5-IκBα/Tnfr1^-/- mice, no immune cell infiltration can be detected, which confirms the results from the histological analysis (Fig. 2). Positive staining for T cells (CD3) in the epidermis of K5-IκBα/Tnfr1^-/- mice labels the resident dendritic T cell population, which shows an identical staining pattern as in wild-type mice (data not shown).

Fig. 2.

The immune-cell infiltration in the skin of K5-IκBα mice is abolished in K5-IκBα/Tnfr1^-/- mice. Shown is staining for neutrophils (myeloperoxidase), T lymphocytes (CD3), B lymphocytes, and plasmacytoid dendritic cells (B220/CD-45R). Arrows indicate positive staining in K5-IκBα mice. Note the two populations seen by B220/CD-45R staining. (Scale bar, 100 μm.)

When staining K5-IκBα mice for the B lymphocyte marker B220, two different cell populations are seen (Fig. 2): one with typical lymphocyte morphology that can be found deep in the dermis and hypodermal fat layer of inflamed skin and the other with a dendritic morphology that is seen throughout the dermis, invading also the epidermal cell layers. Staining with another B lymphocyte marker, Pax5, only stains the lymphocyte population (data not shown). Although the B220(+) lymphocyte population remains stable, the B220(+) Pax5(-) dendritic cells increase in number with progression of the disease (data not shown). Putatively, these cells are identified as plasmacytoid dendritic cells, which have recently been suggested to be involved both in inflammation and in the development of cancer (13, 14).

Il1r1 Signaling Is Not Required for Development of Inflammation, Epidermal Hyperplasia, or SCC in K5-IκBα Mice. By using array analysis for cytokine expression, we detected strong up-regulation of Il-1α and Il-1β in lesional skin from K5-IκBα mice (data not shown) and confirmed this detection by RT-PCR analysis (Fig. 3a). Immunohistochemical staining for Il-1β showed strong up-regulation in inflamed skin in a mixed population of dermal cells; the majority of which were identified as polymorphonuclear granulocytes based on their morphology (Fig. 3b). To assess a potential role for Il-1 in the inflammatory process and the development of cancer, we transferred the K5-IκBα transgene onto an Il1r1-null background. Il1r1 is the main Il-1 receptor responsible for signal transduction induced by both Il-1α and Il-1β (15). No difference was seen in development of inflammation or cancer between K5-IκBα/Il1r1^-/- mice and K5-IκBα mice (Fig. 3c). Thus, Il1r1 signaling is not required either for development of inflammation or SCC in this model. Interestingly, the up-regulation of Il-1α and Il-1β is downstream of Tnfr1 signaling, which is also indicated by the absence of Il-1 up-regulation in the K5-IκBα/Tnfr1^-/- mice (Fig. 3a).

Fig. 3.

Il1r1 signaling is not required for development of inflammation, epidermal hyperplasia, or cancer in K5-IκBα mice. (a) RT-PCR showing up-regulation of Il-1α and Il-1β in K5-IκBα mice. (b) Immunohistochemical staining for IL-1β in wild-type and K5-IκBα mice. (c) Hematoxylin/eosin staining of dorsal skin from K5-IκBα/Il1r1^-/- versus K5-IκBα mice. (Scale bars, 100 μm.)

Tumor Development Driven by Tnfr1-Mediated Signaling and Inflammation. It was recently reported that coexpression of a superrepressor form of IκBα and Ha-Ras results in neoplasia resembling invasive SCC in human keratinocytes transplanted to scid/scid mice (4). In this setting, expression of IκBα was shown to cause an up-regulation of cyclin-dependent kinase 4 (Cdk4). We have previously shown that the tumors in the K5-IκBα mice do not display mutations in Ha-Ras (16). Nor could we detect any increase in active MAPK in skin or in tumors from K5-IκBα mice compared to wild-type skin (Fig. 4 a and b). Fig. 4a shows the staining pattern for active MAPK in the tail skin of 3.5-week-old wild-type and K5-IκBα mice, with a strong labeling of cells in the suprabasal cell layers. In most tumors, no positive staining can be seen, except for a fraction of well differentiated tumors that show occasional staining in suprabasal cell layers (Fig. 4b). Furthermore, we could not find any up-regulation of Cdk4 either in primary keratinocytes from K5-IκBα mice in vitro (Fig. 4c) or in vivo where Cdk4 is abundantly expressed in the basal cell layers with no evident difference between wild-type and transgenic skin (Fig. 4d). In the inflamed, severely hyperplastic K5-IκBα skin of older mice, the expression of Cdk4 is increased compared with wild-type mice, probably reflecting the increased proliferation (Fig. 4e).

Fig. 4.

Activation of the Ras-pathway or overexpression of Cdk4 does not explain the development of SCC in K5-IκBα mice. (a) Tail skin from 3-week-old mice stained for active MAPK. (Scale bar, 50 μm.) (b) Dorsal skin from wild-type mice and K5-IκBα mice with SCC stained for active MAPK. (Scale bar, 100 μm.) (c) Whole-cell extracts from keratinocytes isolated from wild-type or K5-IκBα mice, blotted and stained for Cdk4 and β-actin. (d) Tail skin section from 3-week-old mice stained for Cdk4. (Scale bar, 50 μm.) (e) Tail skin sections from adult mice stained for Cdk4. (Scale bar, 100 μm.)

Tnfα Is Up-Regulated in the Skin of K5-IκBα Independent of the Inflammation. Real-Time PCR analysis showed that Tnfα expression was increased 5-7 times in inflamed K5-IκBα skin compared with wild-type skin. Interestingly, a moderate up-regulation of 2-3 times could also be detected in the skin of K5-IκBα/Tnfr1^-/- compared with Tnfr1^-/- mice. Because the K5-IκBα/Tnfr1^-/- animals do not show any inflammation in the skin, our interpretation is that the moderate up-regulation of Tnfα seen in the K5-IκBα/Tnfr1^-/- animals depends on the inhibition of NF-κB in the keratinocytes. This up-regulation of Tnfα may be the first priming event of the disease, preceding the development of inflammation and SCC.

Development of Inflammation, Epidermal Hyperplasia, and SCC Cannot Be Recovered in K5-IκBα/Tnfr1^-/- Mice by Tnfr1^+/- Bone Marrow-Derived Cells. If up-regulation of Tnfα is the initiating event driving the development of disease in the K5-IκBα mice, the next crucial question that follows is which cells are responding to Tnfα: the immune cells or the nonbone-marrow-derived cells in the skin. To address this question, Tnfr1^+/- or Tnfr1^-/- bone-marrow cells were transplanted to lethally irradiated K5-IκBα/Tnfr1^-/- mice. Surprisingly, reconstitution with Tnfr1^+/- bone marrow did not recover the inflammatory phenotype, epidermal hyperplasia, or tumor development (animals were monitored for 6 months after transplantation) (Fig. 5). This suggests that critical Tnfr1 signaling occurs within the resident skin cells or nonimmune cells. Recent data supports this as it was shown that Tnfr1 signaling and downstream JNK activity drives hyperproliferation of RelA^-/- epidermis transplanted onto scid/scid mice (17). Similar to the reported increase in the levels of active JNK in RelA^-/- epidermis, we also observe a strong increase in JNK activity in K5-IκBα hyperplastic epidermis, which is abolished in K5-IκBα/Tnfr1^-/- mice (Fig. 6, which is published as supporting information on the PNAS web site).

Fig. 5.

Development of inflammation, epidermal hyperplasia, and SCC cannot be recovered in K5-IκBα/Tnfr1^-/- mice by Tnfr1^+/- bone-marrow-derived cells. Genotype of recipients (R) and transplants (T) are denoted in the figure. (Scale bar, 50 μm.)

Discussion

We have earlier shown that selective inhibition of Rel/NF-κB signaling in the murine skin by targeted expression of a superrepressor form of IκBα results in inflammation, hyperproliferation, increased apoptosis, and spontaneous early development of SCC, with a penetrance of 100% (3). In this study, we show that the inflammation and hyperproliferation seen in the K5-IκBα mice depend on Tnfr1 signaling, similar to what was recently reported for K14-Cre/Ikk2^FL/FL mice, also having a defect in NF-κB signaling (12). More importantly, no spontaneous cancer development was observed in K5-IκBα/Tnfr1^-/- mice. The unique role of Tnfα for the development of the disease in these mice is further demonstrated by the fact that removal of the receptor for the proinflammatory cytokines Il-1α and Il-1β, both of which are shown to be highly overexpressed in affected skin, does not influence the development of disease.

Our data show that inhibiting NF-κB in keratinocytes in vivo leads to an up-regulation of Tnfα in the skin independent of inflammation, given that it can also be seen in K5-IκBα/Tnfr1^-/- mice. This supports a model in which up-regulation of Tnfα in the skin is the first event in the disease process that precedes the development of inflammation and SCC. The cellular source of the increase in Tnfα expression is currently not known. Keratinocytes from K5-IκBα mice cultured in vitro do not show up-regulation of Tnfα compared with wild-type (data not shown). However, that keratinocytes do not show up-regulation in vitro does not exclude them from being responsible for the up-regulation in vivo.

Even more intriguing, our data strongly indicate that the Tnfα-responding cells also reside in the skin, given that reconstitution of lethally irradiated K5-IκBα/Tnfr1^-/- mice with Tnfr1^+/- bone marrow-derived cells could not recover the inflammatory phenotype, epidermal hyperplasia, or tumor development. These results support a two-step model in which the Tnfr1 response in resident skin cells leads to induction of other cytokines/chemokines that in turn attract the inflammatory cells to the skin. However, additional experiments are needed to verify this notion. Bone-marrow reconstitution is limited to experiments on adult mice, and at the present time, we cannot definitely exclude the possibility of a different result in a case in which Tnfr1 had been present in the bone-marrow cells from birth. It is also possible that Tnfr1 is needed both on the keratinocytes and on the immune cells to induce the inflammation. Additional experiments will allow further investigation of these questions. Support for an important role of Tnfr1 signaling in the keratinocytes is provided by the recent report that Tnfr1 and downstream JNK activity drives hyperproliferation in RelA^-/- epidermis (17). Similar to RelA^-/- epidermis, hyperplastic skin in K5-IκBα mice shows a strong up-regulation of active JNK (Fig. 6).

There is increasing evidence for a role of Tnfα in tumor promotion in skin carcinogenesis. Removal of the Tnfα response in mice by deletion of Tnfα or Tnfr1 has been shown to attenuate 12-O-tetradecanoylphorbol-13-acetate-induced inflammation and confers resistance in two-stage 7,12-dimethylbenz(a)anthracene/12-O-tetradecanoylphorbol-13-acetate carcinogenesis experiments (18, 19).

Although critical, increased expression of Tnfα in the skin is not likely to be the only explanation for the phenotype observed in the K5-IκBα mice. Transgenic overexpression of TNFα in the murine epidermis leads to inflammation of the skin but without hyperplasia and development of SCC; rather, such mice develop hypoplasia of the epidermis (20). In contrast, the extremely fast development of SCC in K5-IκBα mice, in which the earliest tumors were seen within 3 weeks of birth, suggests an intricate interplay between the inflammatory response, Tnfr1 signaling, and blockade of NF-κB signaling in the target keratinocytes.

We propose that inhibition of NF-κB in epidermal/follicular keratinocytes disturbs the stress-induced growth arrest response. The cellular stress in the K5-IκBα mouse is induced by inflammation of the skin, which depends on the Tnfr1-mediated signaling, and we predict that other origins of stress may have similar effects. In this context, it is interesting to note that coexpression of a superrepressor form of IκBα and Ha-Ras results in neoplasia resembling invasive SCC in human keratinocytes transplanted to scid/scid mice (4). The expression of Ha-Ras in human keratinocytes may be seen as a stress factor, similar to the inflammation in the K5-IκBα model, and the requirement of both Ras and IκBα to induce neoplasia in the skin grafts likely reflects the absence of an intact immune system in scid/scid mice. We have previously shown that tumors in K5-IκBα mice do not display mutations in Ha-Ras (16), and we now find that there is no increase in activated MAPK in the skin or tumors of K5-IκBα mice, essentially ruling out Ras as a factor critical to cancer development in this model. Interestingly, both Ras and Tnfα induce growth arrest in keratinocytes in vitro (4, 21), and an attractive possibility is that keratinocytes with blocked NF-κB signaling have lost the growth arrest response to both Ras and cytokines.

We have previously reported that keratinocytes in K5-IκBα mice show an abnormal cell cycle arrest in response to gamma irradiation (16). The effect is only seen in hyperplastic, inflamed skin whereas primary keratinocytes isolated from K5-IκBα mice grown in vitro respond normally to γ irradiation (data not shown), suggesting that inflammation and/or increased expression of cytokines contribute to the altered DNA-damage response in the K5-IκBα mice.

Expression of IκBα in human keratinocytes has also been shown to cause an up-regulation of Cdk4 (4). However, we do not find such an up-regulation in primary keratinocytes from K5-IκBα mice in vitro or in K5-IκBα skin in vivo before the development of inflammation. In inflamed hyperplastic K5-IκBα skin, the expression of Cdk4 is increased. However, this increase likely only reflects increased proliferation (Fig. 4e). Cdk4 is expressed in proliferating cells of the basal cell layer in normal skin (Fig. 4d). Although not excluding a role for Cdk4 up-regulation, our data indicate that other IκBα-dependent alterations contribute to the phenotype.

The massive infiltration of granulocytes, with a domination of neutrophils that is seen in the skin of K5-IκBα mice, is striking (Fig. 2 and M.v.H., unpublished observations). A critical role of neutrophils in driving the hyperproliferation in the flaky skin mouse mutant has been demonstrated, and it is possible that neutrophils will prove to play a similar role in K5-IκBα mice (22). It is known that neutrophils produce reactive oxygen species, lipid mediators, and proteolytic enzymes that potentially can affect keratinocyte proliferation (23-25). Of interest is also the infiltration of B220(+) Pax5(-) cells with dendritic morphology, which we identify as plasmacytoid dendritic cells. Plasmacytoid dendritic cells have recently been shown to play a role both in inflammation of the skin and in cancer development, where they have been proposed to be attracted by the tumor cells and inhibit tumor-specific immunity (13, 14).

Although only a minor fraction of human SCC harbor Ras mutations (26, 27), it is interesting to note that recent data suggest that human SCC may show a block in NF-κB signaling based on nuclear exclusion of the RelA NF-κB subunit (ref. 4 and M.v.H., unpublished results). Furthermore, inflammation and chronic wounds are associated with development of human SCC, as in the development of SCC in burn scars, chronic venous stasis ulcers, and discoid lupus erythematosus (28-30). Progression of actinic keratosis to SCC has recently also been shown to be associated with an inflammatory stage (31).

Taken together, these findings imply that the combination of inflammatory stress and inhibition of NF-κB signaling is of relevance for SCC development, creating a cellular environment enhancing generation of DNA alterations [e.g., induced by oxidative stress (32)], and/or selecting cells with preexisting mutations (33). The K5-IκBα mouse model we describe will be valuable for unraveling the detailed molecular mechanisms and signals involved in the interplay between perturbed NF-κB signaling in keratinocytes and inflammation in SCC development. Furthermore, the critical role of local NF-κB and Tnfr1-mediated signaling provides the rationale for novel approaches to treat SCC and prevent progression of premalignant lesions.