Main

Mice lacking IKK2 die in utero as a result of tumour necrosis factor (TNF)-induced hepatocyte apoptosis3,4,5. To study the function of IKK2 in different tissues we generated mice carrying a loxP-site-flanked (floxed (FL)) Ikk2 allele (Fig. 1a). Ikk2FL/FL mice develop normally and express normal levels of IKK2 (Fig. 1b). Excision of the floxed Ikk2 sequences by crossing to a Cre-deleter transgenic strain20 produced mice carrying deleted (D) Ikk2 alleles. As expected, Ikk2D/D mice die in utero and exhibit liver degeneration (data not shown). Analysis of mouse embryonic fibroblasts (MEFs) revealed that Ikk2D/D cells lack expression of IKK2 (Fig. 1b) and show impaired activation of NF-κB on stimulation with TNF, interleukin (IL)-1 or bacterial lipopolysaccharide (LPS) (ref. 7 and data not shown). To ablate IKK2 specifically in the epidermis we used a transgenic mouse line expressing Cre under the control of the human keratin 14 (K14) promoter. LacZ staining of tissue sections from a K14-Cre mouse carrying a ROSA26FL Cre-reporter allele21 (Fig. 1c and data not shown) and Southern blot analysis of DNA isolated from various tissues of a K14-Cre/Ikk2FL/FL mouse (Fig. 1d) showed Cre recombination specifically in the epidermis, hair follicles and the epithelium of the tongue. Western analysis of epidermal extracts prepared from newborn pups failed to detect expression of IKK2 in K14-Cre/Ikk2FL/FL mice, demonstrating that IKK2 protein is absent from the epidermis at birth (Fig. 1e).

Figure 1: Epidermis-specific knockout of IKK2.
figure 1

a, Diagram showing the wild-type Ikk2 genomic locus (WT), the neomycin-resistance-containing (neo), the loxP-flanked (FL) and the deleted (D) IKK2 alleles. Filled boxes indicate exons (E6–E9). Restriction enzyme sites and the location of the probe used for Southern blot analysis are depicted. StuI fragments are in kilobases. B, BamHI; E, EcoRI; Bg, BglII; S, StuI. LoxP sites are indicated by arrowheads. b, Western blot analysis of IKK2 and IKK1 expression in wild-type and Ikk2D/D MEFs and in splenocytes from wild-type and Ikk2FL/FL mice. c, Analysis of β-galactosidase expression in the skin of a K14-Cre, ROSA26FL/+ mouse shows Cre recombination specifically in the epidermis and in hair follicles. Scale bar, 200 µm. d, Tissue specificity of IKK2 inactivation. DNA isolated from various tissues of a K14-Cre/Ikk2FL/FL mouse was subjected to Southern blot analysis after digestion with StuI. Deletion is detected in DNA from epidermis, total skin and tongue. e, Western blot analysis of IKK2 and NEMO expression in the epidermis of newborn mice with the indicated genotypes. f, K14-Cre/Ikk2FL/FL and control mice at P8.

Full size image

K14-Cre/Ikk2FL/FL pups are born at the expected mendelian ratios and are macroscopically indistinguishable from their littermates until postnatal day 4–5 (P4–P5), when their skin starts becoming hard and inflexible. This phenotype progresses rapidly and by P7–P8 the mice exhibit a highly rigid, shell-like skin with widespread scaling (Fig. 1f). At this stage the mice become ‘runted’ and they subsequently die between P7 and P9. The cause of death is not known at present; however, it does not seem to be the result of impaired skin barrier function as no differences were detected in toluidine blue penetration into the skin between mutant and control mice at P8 (data not shown). Ikk2FL/FL mice lacking the K14-Cre transgene and Ikk2FL/+ and Ikk2+/+ mice with or without K14-Cre did not show any pathological skin phenotype and are collectively referred to as controls. Histological analysis of skin sections from K14-Cre/Ikk2FL/FL mice at P7 revealed a markedly thickened epidermis with loss of the granular layer, pronounced hyperkeratosis, focal parakeratosis, subcorneal pustule formation and increased cellularity and dilated blood vessels in the dermis (Fig. 2a).

Figure 2: Histology of the skin in mice with epidermis-specific IKK2 deficiency.
figure 2

a, b, Skin sections from K14-Cre/Ikk2FL/FL (mutant, mt) and control (ct) mice were stained with haematoxylin/eosin (a) or immunostained for the indicated epidermal differentiation markers (green staining in b). Red counterstaining in b shows tissue structure; blue staining shows nuclei. c, Analysis of cell proliferation by staining for Ki67 (green); red staining shows nuclei. d, TUNEL staining for the detection of apoptotic cells (green); red staining shows nuclei. A white arrow points to an intraepidermal pustule with surrounding apoptotic keratinocytes. Dotted lines indicate the border between epidermis and dermis. Scale bars: a, 150 µm; b, 100 µm; c, d, 200 µm.

Full size image

We undertook detailed immunohistological analysis of skin sections from K14-Cre/Ikk2FL/FL and control mice at different time points after birth, using antibodies against various epidermal differentiation markers. At P0 and P1 the skin of mutant mice did not show any differences compared to controls (Fig. 2b, and data not shown), suggesting that, in contrast to IKK1, IKK2 is not essential for epidermal differentiation. At P3 the epidermis of K14-Cre/Ikk2FL/FL mice showed upregulation of keratin 6 (K6) expression, a marker for inflamed and hyperproliferative epidermis, and suprabasal expression of K14, which is normally confined to the basal epidermal layer (Fig. 2b). K6 and K14 were expressed in all viable layers of the mutant epidermis at P7 (Fig. 2b). In addition, expression of the suprabasal keratin 10 (K10) and of the keratinocyte terminal differentiation markers loricrin (Fig. 2b) and filaggrin (not shown) was markedly reduced in K14-Cre/Ikk2FL/FL animals at this stage. Staining for Ki67 revealed increased proliferation of keratinocytes in mutant epidermis at P4 and P7, but not P0 and P1 (Fig. 2c, and data not shown). TdT-mediated dUTP nick end labelling (TUNEL) staining showed no differences between K14-Cre/Ikk2FL/FL and control skin at P2 (Fig. 2d); however, increased numbers of apoptotic keratinocytes in proximity to aggregates of inflammatory cells were observed in mutant epidermis at P8 (Fig. 2d).

To determine whether the cutaneous phenotype observed in K14-Cre/Ikk2FL/FL mice is caused by a cell-autonomous defect in proliferation and/or differentiation of IKK2-deficient keratinocytes, we analysed ex vivo keratinocyte cultures from mutant and control mice. Keratinocytes isolated from newborn K14-Cre/Ikk2FL/FL mice lack expression of IKK2 (Fig. 3a) and, similarly to Ikk2-knockout MEFs3,4,5,7, show impaired NF-κB activation in response to stimulation by TNF or IL-1β (Fig. 3b, c). IKK2-deficient keratinocytes were not impaired in their capability to terminally differentiate in suspension culture, as demonstrated by similar expression of K10, involucrin, loricrin and filaggrin compared to control keratinocytes (Fig. 3e–f, see also Supplementary Information Table 1). Growth curves revealed that primary keratinocytes lacking IKK2 exhibit decreased proliferation compared with control cells (Fig. 3d). These findings show that, in contrast to IKK1, IKK2 deficiency does not lead to impaired differentiation and hyperproliferation of keratinocytes. Previous in vivo and in vitro experiments showed hyperproliferation and epidermal hyperplasia after NF-κB inhibition and growth arrest after NF-κB activation in epidermal keratinocytes, and suggested a role for NF-κB in the regulation of keratinocyte growth12,13,14. In these studies NF-κB activity in keratinocytes was modulated downstream of the IKK complex. Our experiments demonstrate that the proposed growth regulatory function of NF-κB in epidermal keratinocytes can be mediated through an IKK2-independent pathway and suggest that the skin phenotype of K14-Cre/Ikk2FL/FL mice is not caused by cell-autonomous hyperproliferation of IKK2-deficient keratinocytes. TUNEL staining revealed that keratinocytes lacking IKK2 do not undergo increased spontaneous apoptosis in culture and show only mildly increased apoptosis on TNF treatment (1.9–2.8% compared with 0.1–0.2% in the controls). In agreement with the presence of only a few apoptotic keratinocytes in the epidermis of K14-Cre/Ikk2FL/FL mice, these results suggest that TNF-induced killing of IKK2 knockout keratinocytes does not have an important role in the pathogenesis of the skin phenotype in our mice.

Figure 3: Analysis of NF-κB activation, proliferation and differentiation in cultured keratinocytes.
figure 3

a, Western blot analysis showing expression of IKK2, IKK1, NEMO and actin in keratinocytes isolated from mice with the indicated genotypes. b, IκB kinase activity (KA), IκBα degradation and NF-κB activation in primary keratinocytes isolated from control (ct) and K14-Cre/Ikk2FL/FL (mt) mice. Cells were treated with TNF (20 ng ml-1) for the indicated times. The kinase activity of IKK complexes immunoprecipitated with anti-NEMO antibody was measured. NEMO blot is shown as loading control. IκBα degradation was assayed by western analysis of cytoplasmic extracts and NF-κB DNA-binding activity was measured by gel mobility shift analysis of nuclear extracts. c, Immunostaining for p65/RelA in unstimulated or IL-1β-treated (20 ng ml-1 for 20 min) mutant and control keratinocytes. d, Growth curves of primary keratinocytes in culture. A total of 30,000 cells isolated from mutant (red circles) and control mice (blue squares) were seeded per well, cultured for 17 days and collected at different time points. The graph shows mean values ± s.d. from triplicate wells. e, f, Expression of differentiation markers in IKK2-deficient primary keratinocytes before (start) or after 24 h in suspension culture (susp). Fluorescence images of keratinocytes expressing keratin 10 (red in e) and involucrin (red in f) are shown. Blue signal shows nuclei stained with 4,6-diamidino-2-phenylindole (DAPI). Scale bars: c, e, f, 200 µm.

Full size image

The progressive disturbance of epidermal differentiation and proliferation in the skin of K14-Cre/Ikk2FL/FL mice seems to be associated with an inflammatory response, suggested by the presence of subcorneal pustules (Fig. 4a) and increased cellularity in the dermis at P7. Immunohistological analysis revealed the presence of increased numbers of macrophages, granulocytes and CD4-positive T cells in the dermis of mutant mice at P7 (Fig. 4a). Examination of inflammatory cytokine gene expression showed increased levels of IL-1β messenger RNA in the skin of K14-Cre/Ikk2FL/FL mice as early as P2, but not at P1 (Fig. 4b and data not shown). Curiously, IKK2-deficient epidermal keratinocytes are identified as the major source of these elevated IL-1β levels, as demonstrated by immunostaining with an antibody against IL-1β (Fig. 4d) and by analysis of IL-1β mRNA in epidermal sheets (data not shown). However, the increased expression of IL-1β does not seem to be induced by a cell-autonomous mechanism triggered by the lack of IKK2, as cultured IKK2-deficient keratinocytes do not show increased expression of IL-1β (Fig. 4c and data not shown). Immunohistology with anti-TNF antibodies revealed increased expression of TNF by cells in the dermis—most probably macrophages and granulocytes—of K14-Cre/Ikk2FL/FL mice at P4 and P7, but not at P1 and P2 (Fig. 4d and data not shown). To identify the signals that attract the inflammatory cells into the skin we performed in situ hybridization analysis for various chemokine mRNAs. These experiments showed expression of MCP-1 (Fig. 4e), LIX, C10, IP-10 and Mig (data not shown) in the dermis of K14-Cre/Ikk2FL/FL mice at P4 and P7, but not in the controls. MIP-2, one of the murine homologues of human IL-8, was expressed mainly in pustules of inflammatory cells in the epidermis of K14-Cre/Ikk2FL/FL mice at P7 (Fig. 4e). These results suggest that an immune response orchestrated by the expression of inflammatory cytokines and multiple chemokines in the dermis, and involving T lymphocytes, macrophages and granulocytes, is implicated in the pathogenesis of the skin disease that develops in these mice.

Figure 4: Inflammation in the skin of K14-Cre/Ikk2FL/FL mice.
figure 4

a, Sections were stained with Giemsa or immunostained with antibodies recognizing T cells (CD3, CD4), granulocytes (Gr-1) and macrophages (F4/80) (green signal). Counterstaining and nuclear staining is as in Fig. 2b. Arrows point to intraepidermal pustules. b, RNase protection assay showing IL-1β expression in the skin of mice 2 or 4 days after birth. Left lane shows positive controls; tRNA is used as negative control. c, RT-PCR analysis of mouse IL-1β expression in skin samples from mutant and control mice at P4 and in cultured IKK2-deficient (mt) and control (ct) keratinocytes (cells). d, Immunostaining using antibodies specific for either IL-1β or TNF (green) in skin sections from mice at P3, and P7–P8; red staining shows nuclei. e, In situ hybridization for MCP1 and MIP-2 mRNA (black) in skin sections from mice at P4 (MCP-1) or P7 (MIP-2). mt, mutant; ct, control mice. Magnification × 250. Dotted lines indicate the border between epidermis and dermis. Scale bars, a, d, 200 µm.

Full size image

To investigate whether αβ T lymphocytes are essential for the development of the skin disease we crossed the K14-Cre/Ikk2FL/FL mice with T-cell antigen receptor-α (TCRα) knockout mice, which lack mature αβ T cells22. K14-Cre/Ikk2FL/FL/TCRα-/- mice develop skin disease with similar kinetics and histological characteristics as the K14-Cre/Ikk2FL/FL mice (Fig. 5b), demonstrating that αβ T cells are not required for the development of the inflammatory skin lesions. This finding, together with the early age of disease onset, suggests that the inflammation of the skin in K14-Cre/Ikk2FL/FL mice is not triggered by an antigen-specific immune response but rather by a reaction driven from the innate immune system.

Figure 5: Requirement of TNFRI for the development of the skin phenotype in K14-Cre/Ikk2FL/FL mice.
figure 5

a, K14-Cre/Ikk2FL/FL/TnfrI-/- mice do not develop inflammatory skin disease. b, Analysis of skin samples from K14-Cre/Ikk2FL/FL/TnfrI-/- and K14-Cre/Ikk2FL/FL/TCRα-/- mice. Sections were stained with haematoxylin/eosin (H/E) or Giemsa, or with antibodies against the indicated differentiation, proliferation or immune cell markers (green signal; see Fig. 4 for abbreviations). Nuclei are in red. Top scale bar, 100 µm; bottom scale bar, 200 µm. Dotted lines indicate the border between epidermis and dermis.

Full size image

To investigate the role of TNF signalling in the pathogenesis of the skin disease we crossed the K14-Cre/Ikk2FL/FL mice with TNF-receptor I (TNFRI)-deficient mice23. Surprisingly, K14-Cre/Ikk2FL/FL mice lacking TNFRI do not develop inflammatory skin disease (Fig. 5a). Immunohistological analysis of skin sections from K14-Cre/Ikk2FL/FL/TnfrI-/- mice revealed a normal pattern of keratinocyte differentiation and proliferation in the epidermis and the absence of inflammatory cells from the dermis (Fig. 5b). These results demonstrate that TNFRI signalling is critical for the inflammatory response that causes the skin disease in K14-Cre/Ikk2FL/FL mice. As TNFRI mediates its effects largely through activation of NF-κB-dependent gene transcription, the presence of an intact NF-κB signalling pathway in all other cell types except epidermal keratinocytes in K14-Cre/Ikk2FL/FL mice may be critical for disease pathogenesis. Finally, the normal development of the epidermis in K14-Cre/Ikk2FL/FL/TnfrI-/- mice confirms that IKK2 deficiency does not interfere directly with keratinocyte differentiation and proliferation, and suggests that the observed hyperplasia and the disturbed expression of keratinocyte differentiation markers in the mutant epidermis after P3 are secondary to the inflammation of the skin.

How can deletion of IKK2 in epidermal keratinocytes initiate an inflammatory skin disease? A tightly regulated microenvironment of cytokines and growth factors produced by keratinocytes, dermal fibroblasts and immune cells is critical for the maintenance of physiological skin homeostasis24. The NF-κB signalling pathway may be essential for epidermal keratinocytes to respond to such regulatory mediators and/or to produce factors that are important for the maintenance of a well balanced interplay between the epidermis, the dermis and the immune system. Although the exact mechanism of disease initiation remains unclear, ablation of IKK2 from epidermal keratinocytes seems to interfere with this balance and triggers an inflammatory response that is marked by the induction of inflammatory cytokines such as IL-1β and TNF, resulting in the development of skin disease in K14-Cre/Ikk2FL/FL mice. The neutralization of the skin disease in K14-Cre/Ikk2FL/FL mice lacking TNFRI demonstrates the essential pathogenic role of TNF for the development of this phenotype. This result does not exclude a function for IL-1 in our model, but it confirms the role of TNF as a master cytokine in inflammation and highlights its significance in the pathogenesis of inflammatory skin disease. The initiation of the skin disease during the first few days after birth may indicate the involvement of environmental factors. Alternatively, mechanisms related to developmental changes of the skin at this early age might be involved in disease initiation. Our finding that interference with NF-κB signalling in epidermal keratinocytes triggers skin inflammation provides new insight into the function of NF-κB in the epidermis and suggests that keratinocytes can act as the initiating cell type in inflammatory skin disease.

Methods

Gene targeting and transgenic mice

A clone containing the mouse Ikk2 genomic locus was isolated by screening a P1 embryonic stem cell mouse library (Genome Systems). A 12-kilobase (kb) BamHI fragment containing exons 6–9 of the Ikk2 gene was mapped and sequenced. The IKK2 targeting vector was constructed using the pEasyFlox plasmid (provided by M. Alimzhanov) by placing a 2-kb EcoRI–BglII genomic fragment, containing exons 6 and 7, between the loxP-flanked PGKneo cassette and the third loxP site. An upstream 2.6-kb ScaI–EcoRI fragment and a 4.2-kb downstream BglII fragment were used as arms for homology. Bruce-4 embryonic stem cells derived from C57Bl/6 mice25 were cultured, transfected and selected as described previously7. Homologous recombinant clones were isolated and the loxP-flanked PGKneo cassette was excised by transient expression of Cre recombinase. Germ-line transmitting chimaeras were generated using embryonic stem cells carrying the floxed Ikk2 gene. Deletion of exons 6 and 7, which correspond to nucleotides 389–567 of the coding IKK2 complementary DNA sequence, introduces premature termination codons producing an Ikk2 null allele. The K14-Cre transgenic mouse will be described elsewhere (M.H., manuscript in preparation). TNFRI-deficient mice were provided by K. Pfeffer.

Western blotting, IKK and NF-κB assays

Immunoblot analysis, IKK kinase assays and NF-κB gel mobility shift assays of cytoplasmic and nuclear extracts from cultured keratinocytes were done as described7,26.

Histology and immunostaining

Immunostainings were carried out on paraffin or cryostat sections using polyclonal antibodies against mouse loricrin, filaggrin, K14 and K10 (BabCo) and involucrin, or monoclonal antibodies against mouse CD3, CD8 (Chemicon), CD4 (clone GK1.5/4), Gr-1 (Ly-6G, clone RB6-8C5), B220 (clone RA3-6B2), F4/80 (Serotec), IL-1β (R&D Systems) and TNF (PharMingen). We used secondary antibodies coupled to Alexa 488 (Molecular Probes). Sections were counterstained with phalloidin labelled with tetramethyl rhodamine B isothiocyanate (Sigma) and with ToPro III (Molecular Probes) to visualize tissue structure and nuclei. We performed TUNEL staining using the Apoptosis Detection System (Promega). Fluorescent staining was analysed using a Leica TCS upright confocal laser-scanning microscope at excitation wavelengths of 488, 543 and 633 nm.

Keratinocyte proliferation, differentiation and apoptosis

Primary epidermal keratinocytes were isolated from the skin of newborn mice as described27 and cultured in FAD medium containing 50 µM calcium ions in the absence or presence of mitomycin C-treated 3T3 fibroblasts as feeders. For induction of differentiation, keratinocytes were transferred to FAD medium containing 1.8 mM calcium ions and methylcellulose28 and cultured in suspension for 24 h. Cells were recovered, washed, attached to chamber slides (Invitrogen), fixed with methanol and stained for differentiation markers. For the detection of TNF-induced apoptosis, cells were treated with 30 ng ml-1 TNF for 18 h and were analysed by TUNEL staining. Images were digitized and positive cells were counted.

In situ hybridization and RNA expression analysis

In situ hybridization experiments were performed as described previously29. Mouse cDNA probes were provided by J. M. Farber (Mig), T. A. Hamilton (IP-10), J. B. Smith and H. R. Herschman (MIP-2 and LIX), T. Yoshimura (MCP-1) and A. Orlofsky (C10). Briefly, paraffin-embedded tissue was cut, deparaffinized, rehydrated and acetylated. Sections were then overlaid with the hybridization solution containing 35S-labelled antisense or, for control, sense probes. Slides were then dipped into Kodak NTB-2 solution, exposed for autoradiography, counterstained and microscopically evaluated. We performed RNase protection assays as described30. For polymerase chain reaction with reverse transcription (RT-PCR) analysis the following primers were used: mouse IL-1β, 5′-CTGAAGCAGCTAT GGCAACT-3′ and 5′-GGATGCTCTCATCTGGACAG-3′; mouse actin, 5′-TAAAACGC AGCTCAGTAACAGTCCG-3′ and 5′-TGGAATCCTGTGGCATCCATGAAAC-3′.