LUBAC prevents lethal dermatitis by inhibiting cell death induced by TNF, TRAIL and CD95L

HOIP and HOIL-1 are essential to maintain skin homeostasis

To understand the role of LUBAC in the skin, we generated mice that lack HOIP or HOIL-1 selectively in epidermal keratinocytes by crossing Hoip- and Hoil-1-floxed mice with mice expressing the Cre recombinase under the control of the human keratin 14 (K14) promoter. The genotype of the mice was confirmed by PCR (Supplementary Fig. 1a). At the protein level, deletion of HOIP or HOIL-1 in keratinocytes was verified by western blot and immunohistochemistry (Supplementary Fig. 1b, c). As expected, HOIP deficiency abrogated linear ubiquitination at the TNFR1-SC (Supplementary Fig. 1d) and reduced TNFR1-mediated NF-κB activation in primary murine keratinocytes (PMKs) without preventing it (Supplementary Fig. 1e). Mice homozygous for keratinocyte-specific deletion of HOIP or HOIL-1 (Hoip^E-KO and Hoil-1^E-KO mice, respectively) were born at the expected Mendelian frequencies and were macroscopically indistinguishable from littermates up to postnatal day (P) 2 (data not shown). From this day onwards, however, both Hoip^E-KO and Hoil-1^E-KO mice developed severely damaged and scaly skin, which, invariably, resulted in the death of these mice between P4 and P6 (Fig. 1a). No Hoip or Hoil-1 gene dosage effect was observed as Hoip^fl/wtK14Cre⁺ and Hoil-1^fl/wtK14Cre⁺ mice developed normally into adulthood without showing any signs of skin disease (data not shown).

Histological analysis of Hoip^E-KO and Hoil-1^E-KO mice at P4 revealed increased epidermal thickness, parakeratosis, hyperkeratosis and keratinocyte differentiation defects (Fig. 1b, c). These pathologies were accompanied by abnormal myeloid cell infiltration and high levels of cell death as demonstrated by increased cleaved caspase-3 and terminal deoxinucleotidyl transferase-mediated dUTP-fluorescein nick end labelling (TUNEL) staining (Fig. 1b, d, e and Supplementary Fig. 1f, g). Together, these observations reveal that HOIP and HOIL-1 are essential to prevent fatal dermatitis characterised by disruption of the normal epidermal structure, inflammation and aberrant keratinocyte death.

Lethal dermatitis is only partially mediated by TNFR1

The inflammatory syndrome of Sharpin^cpdm/cpdm mice can be abrogated by the absence of TNF and TNFR1 or by the loss of the kinase activity of RIPK1^22,27,32,34. We therefore first tested whether genetic ablation of TNFR1 could also prevent the morbidity and mortality in Hoip^E-KO and Hoil-1^E-KO mice. Unexpectedly, however, inflammation was only delayed in Tnfr1^KO;Hoip^E-KO and Tnfr1^KO;Hoil-1^E-KO mice as they progressively developed severe skin lesions resulting in a median survival of 70 days (Fig. 2a and Supplementary Fig. 2a). Sick Tnfr1^KO;Hoip^E-KO and Tnfr1^KO;Hoil-1^E-KO mice presented with epidermal disruption and thickening, parakeratosis, hyperkeratosis and inflammation (Fig. 2b, c). Crucially, infiltration by myeloid and lymphoid cells and cell death were significantly augmented in the epidermis of adult Tnfr1^KO;Hoip^E-KO mice compared to control animals (Fig. 2b, d and Supplementary Fig. 2b–d).

Next, we addressed whether ablation of the kinase activity of RIPK1 was sufficient to prevent inflammation in Hoip^E-KO mice. Surprisingly, genetic ablation of the kinase activity of RIPK1 was substantially less effective than loss of TNFR1 in preventing dermatitis as Ripk1^D138N;Hoip^E-KO mice died at around P8 from severe skin disease (Fig. 2f, g). Thus lethal dermatitis caused by keratinocyte-specific deficiency in either HOIP or HOIL-1, the two essential components of LUBAC²⁸, is only partially dependent on TNFR1 and, in the presence of TNFR1, almost completely independent of the kinase activity of RIPK1.

Increased cell death precedes inflammation

We next investigated the temporal relationship between aberrant cell death and inflammation in Hoip^E-KO and Hoil-1^E-KO mice. Abnormally increased cell death in the epidermis of Hoip^E-KO and Hoil-1^E-KO mice was already apparent in utero at E18.5 and at birth (P0) (Fig. 3a, b and Supplementary Fig. 3a, b). This implies that lack of linear ubiquitination in keratinocytes results in aberrant cell death in sterile conditions. Hoip^E-KO and Hoil-1^E-KO mice displayed abnormally increased immune cell infiltration at P2 but not at birth (Fig. 3a, c and Supplementary Fig. 3c–e). Accordingly, keratinocyte differentiation and epidermal thickness appeared abnormal at P2 but not at E18.5 or P0 (Fig. 3a and Supplementary Fig. 3f). Thus excessive cell death precedes the inflammatory response suggesting that keratinocyte death upon loss of HOIP or HOIL-1 may trigger lethal dermatitis.

To understand the mechanism of cell death induction in the skin of Hoip^E-KO and Hoil-1^E-KO mice, we analysed the formation of the signalling platform known to trigger cell death downstream of various death receptors³⁵ by immunoprecipitating the adaptor protein FADD in PMKs in the presence of the caspase inhibitor, Z-VAD-fmk. This revealed that, even without an exogenous stimulus, a FADD/caspase-8/RIPK1-containing complex was readily detectable in HOIP-deficient but not in control PMKs (Fig. 4a). Consistent with apoptotic signalling by such a complex, the HOIP-deficient cells were less viable even in the absence of exogenous stimuli (Fig. 4b). This loss in cell viability was significantly reduced by inhibition of caspases and RIPK1 kinase activity but not by blocking the kinase activity of RIPK3 (Fig. 4b). Inhibition of TNF or genetic ablation of TNFR1 also increased the viability of Hoip^E-KO PMKs (Fig. 4c, d). These results indicate that in PMKs HOIP prevents aberrant RIPK1 kinase-dependent apoptosis triggered by spontaneously produced autocrine TNF.

LUBAC loss in adulthood induces cell death-driven dermatitis

To assess the impact of acute loss of HOIP in keratinocytes in adult mice, we treated Hoip^fl/flK14CreER^Tam mice with 4-hydroxytamoxifen (4-OHT) in a localised area of the skin. This treatment resulted in rapid cell death induction, followed by increased immune cell infiltration (Fig. 5a–c and Supplementary Fig. 4a). This was accompanied by epidermal thickening, hyperplasia, hyperkeratosis as well as parakeratosis and defects in keratinocyte differentiation (Fig. 5d and Supplementary Fig. 4b–e). These findings demonstrate that HOIP is required to maintain normal skin architecture and function in adult mice and that also in the adult skin cell death precedes inflammation in the absence of HOIP.

Aberrant apoptosis is responsible for lethal dermatitis

Since aberrantly increased cell death was the first abnormal event we could detect in the epidermis of Hoip^E-KO and Hoil-1^E-KO mice and because it is also observed in the absence of TNFR1 in adult mice, we next evaluated genetically whether and, if so, which form(s) of aberrant cell death cause the early and the late dermatitis.

Consistent with the apoptotic cell death observed in vitro, genetic ablation of Ripk3 in Hoil-1^E-KO or the loss of Mlkl in Hoip^E-KO mice failed to prevent aberrant cell death and skin inflammation and did not delay the postnatal lethality (Fig. 6b, c, e and Supplementary Fig. 5).

Since caspase-8 deficiency is embryonically lethal due to sensitisation to RIPK3- and MLKL-induced necroptosis^9,10,36–39, it is not possible to generate viable Casp8^KO;Hoil-1^E-KO mice. We therefore first evaluated the effect of Casp8 heterozygosity in Hoil-1^E-KO and Ripk3^KO;Hoil-1^E-KO mice. Heterozygosity of Casp8 was able to extend the survival of Hoil-1^E-KO and Ripk3^KO;Hoil-1^E-KO mice to around P8 and day 20, respectively (Fig. 6e and Supplementary Fig. 6a–d).

We next determined the effect of complete absence of caspase-8. Remarkably, both Mlkl^KO;Casp8^KO;Hoip^E-KO and Ripk3^KO;Casp8^KO;Hoil-1^E-KO mice reached adulthood without any signs of skin disease (Fig. 6a and Supplementary Fig. 6e, f). Epidermal structure and keratinocyte differentiation were completely normal in Ripk3^KO;Casp8^KO;Hoil-1^E-KO mice, and these animals neither exhibited aberrant cell death nor immune cell infiltration in their skin (Fig. 6b–d and Supplementary Fig. 6g). Ripk3^KO;Casp8^KO;Hoil-1^E-KO mice survived without developing any signs of skin inflammation well beyond the 70-day time point at which the Tnfr1^KO; Hoil-1^E-KO mice succumb to severe dermatitis (Fig. 6e), although they had to be sacrificed at later times because of lymphadenopathy and splenomegaly (Supplementary Fig. 6h) caused by the combined deficiency in caspase-8 and RIPK3 or MLKL, as previously reported^10,38,40. Collectively, these results demonstrate that caspase-8-mediated apoptosis is responsible for the lethal dermatitis in mice lacking HOIP or HOIL-1 in keratinocytes and that RIPK3/MLKL-mediated necroptosis does not contribute to the disease.

RIPK1 kinase is required for TNFR1-independent dermatitis

Since aberrant cell death is the cause of dermatitis in Tnfr1^KO;Hoip^E-KO and Tnfr1^KO;Hoil-1^E-KO mice, we aimed to further characterise this type of cell death. In order to evaluate the contribution of necroptosis to the pathology of Tnfr1^KO;Hoip^E-KO mice, we generated Mlkl^KO;Tnfr1^KO;Hoip^E-KO mice. These mice developed less severe skin lesions at day 70 and their survival was significantly prolonged (Fig. 7a–c). Thus, in the absence of TNFR1, necroptosis contributes to skin inflammation.

To investigate the involvement of the kinase activity of RIPK1 in the TNFR1-independent disease caused by LUBAC deficiency in the skin, we next fed Tnfr1^KO;Hoil-1^E-KO and control mice from E14.5 until day 100 with the RIPK1 inhibitor GSK’547A⁴¹. As a control, Sharpin^cpdm/cpdm and Hoip^E-KO mice were also fed with GSK’547A. In line with the genetic analysis of Hoip^E-KO mice (Fig. 2f, g) and previous reports with Sharpin^cpdm/cpdm mice³⁴, this treatment delayed disease development and death of Hoip^E-KO mice up to P8 and prevented dermatitis in Sharpin^cpdm/cpdm mice (Supplementary Fig. 7a, b). Strikingly, pharmacologic inhibition of the kinase activity of RIPK1 rescued the majority of Tnfr1^KO;Hoil-1^E-KO mice from dermatitis for the duration of the treatment with only three of the ten treated mice developing small punctate scales (Fig. 7c–e and Supplementary Fig. 7c). Thus the lethal dermatitis caused by keratinocyte-specific HOIL-1 that occurs in the absence of TNFR1 is mediated by RIPK1 kinase-dependent apoptosis and necroptosis.

TNF, TRAIL and CD95L drive cell death and dermatitis

Finally, we sought to identify the instigator(s) of the TNFR1-independent cell death that is responsible for the fatal dermatitis in Hoip^E-KO and Hoil-1^E-KO mice. Consistent with our previous findings in other cell types^17,42, PMKs derived from Tnfr1^KO;Hoil-1^E-KO mice were more sensitive than control cells to induction of cell death by TRAIL, CD95 (Fas/APO-1) ligand (CD95L) or Polyinosinic:polycytidylic acid (Poly(I:C)) (Fig. 8a). We, therefore genetically ablated the death domain (DD) of CD95 specifically in keratinocytes or deleted TRAIL-R or TLR3 systemically in Tnfr1^KO;Hoil-1^E-KO mice. However, Cd95^E-DD;Tnfr1^KO;Hoil-1^E-KO, Trail-r^KO;Tnfr1^KO;Hoil-1^E-KO and Tlr3^KO;Tnfr1^KO;Hoil-1^E-KO mice all suffered from skin lesions that were indistinguishable in severity from those seen in the Tnfr1^KO;Hoil-1^E-KO mice (Supplementary Fig. 8).

Reasoning that these death receptors might be able to drive disease in a redundant manner in Tnfr1^KO;Hoil-1^E-KO mice, we next assessed the impact of their combined loss on the onset of dermatitis. Co-deletion of TRAIL-R and TLR3 in Tnfr1^KO;Hoil-1^E-KO mice resulted in slightly milder skin lesions at day 70 (Fig. 8b, c), yet these mice still succumbed to inflammatory skin disease with a median survival of 77 days (Fig. 8d). Strikingly, however, combined loss of TRAIL-R with keratinocyte-specific deletion of the DD of CD95 resulted in prevention of dermatitis in Tnfr1^KO;Hoil-1^E-KO mice reflected by the significantly reduced severity score of skin lesions and prolonged survival (Fig. 8b–d). We therefore conclude that the cell death-driven inflammatory disease caused by HOIL-1 deficiency in keratinocytes is induced by multiple death receptor-ligand systems, namely the TNF/TNFR1, TRAIL/TRAIL-R and CD95L/CD95 systems. Thus, cell death induction via either of these receptors is sufficient to cause disease (Supplementary Fig. 9) in the absence of linear ubiquitination caused by HOIL-1 deficiency.

Mice

To generate Hoip^E-KO and Hoil-1^E-KO mice, Hoip^fl/fl26 and Hoil-1^fl/fl28 mice were crossed with mice expressing the Cre recombinase under the control of the human K14 promoter (obtained from Geert van Loo)⁵⁶, strain AZO-Nn4Cre (K14). The K14CreER^Tam57, Mlkl^KO28, Ripk3^KO58, Casp8^KO59, Trail-r^KO60 and Ripk1^D138N61 mice have been previously described. Tnfr1^KO, Tlr3^KO, Cd95-DD^fl/fl mice (C57BL/6-Fastm1Cgn/J) and Sharpin^cpdm/cpdm (C57BL/Ka) mice were purchased from The Jackson Laboratories. To induce deletion of HOIP in epidermal keratinocytes of adult mice, Hoip^fl/flK14CreER^Tam mice were treated as previously described⁶². Briefly, a small shaved area of the dorsal neck was treated with 50 μL of 4-OHT 20 mg mL⁻¹ dissolved in ethanol every other day for a total of 1, 2, 3 or 4 treatments, as indicated. As a vehicle treatment, a small dorsal area close to the tail was shaved and treated with ethanol. Hoip^fl/wtK14CreER^Tam mice were used as tamoxifen controls. Mice were analyzed 2 days after the last treatment or as indicated in the figure legends. Timed matings were performed as previously described²⁶. All mice were genotyped by PCR analysis. Colonies were fed ad libitum. All animal experiments were conducted under an appropriate UK project license in accordance with the regulations of UK home office for animal welfare according to ASPA (animal (scientific procedure) Act 1986).

Pharmacological inhibition of RIPK1 kinase activity

Pregnant females were fed with rodent chow containing 100 mg kg⁻¹ day⁻¹ GSK3540547A (GSK’547A) (GlaxoSmithKline LLC) from 14 days post coitum and continued the special diet throughout the nursing period. At weaning age, Sharpin^cpdm/cpdm and Tnfr1^KO;Hoil-1^E-KO mice and littermate controls were continuously treated with GSK’547A for another 100 days.

Immunostaining and quantification

Four-μm-thick formalin-fixed paraffin-embedded skin sections were stained following standard protocols. Briefly, sections were boiled in 10 mM sodium citrate buffer (pH 6.0) in a microwave. Slides were blocked in buffer containing Tween 20 0.5% and bovine serum albumin 0.2%. For CD45 staining, slides were boiled in Retrievagen A (BD) and blocked with buffer without Tween. Next, slides were incubated with the primary antibody overnight at 4 °C. The following antibodies were used: anti-K14 (1/1000, PRB-155P), anti-K10 (1/100, MMS-159S), anti-loricrin (1/500, PRB-145P) and anti-K6 (1/500, PRB-169P) (Covance); anti-Ki-67 (1/100, Abcam); anti-CD45 (1/100, BD Biosciences); anti-cleaved caspase-3 (1/250, 9661, Cell Signaling); anti-HOIP (custom-made, Thermo Fisher Scientific); and anti-HOIL-1²¹. Slides were incubated with the following secondary antibodies: Alexa Fluor 488 Goat anti-Rabbit IgG, 594 Goat anti-Rabbit IgG (Invitrogen), or goat anti-rat horseradish peroxidase (HRP; Cambridge Bioscience) at room temperature for 1 h. Where an HRP-conjugated antibody was used, the TSA™ Plus Cyanine 3 System (Perkin Elmer) was applied according to the manufacturer’s instructions. Sections were counterstained with 4,6-diamidino-2-phenylindole (DAPI; Roche). For HOIP and HOIL-1 staining, conventional immunohistochemistry (antibody dilution 1/100) was performed on BOND-III (Leica Microsystems) and BenchMark Ultra (Ventana-Roche Medical System) according to a protocol previously described⁶³. For TUNEL staining, which was performed in combination with cleaved caspase-3 staining, the ApopTag Red In Situ Apoptosis Detection Kit (Merck Millipore) and streptavidin conjugate Cy2 (Jackson Immuno Research), respectively, were used according to the manufacturer’s instructions. Sections were analyzed by fluorescent microscopy.

At least ten different images (×40) per slide were acquired. Quantification was performed by an experimenter who was blinded to the genotype of the samples by using the ImageJ Software on monochrome images as the percentage of cells positive for the specific staining in relation to the total number of cells (DAPI-positive) within the epidermis.

Epidermal thickness quantification

The epidermal thickness was measured in five different positions per microscopic field for at least ten different fields per mouse. Quantification was performed by an experimenter who was blinded to the genotype of the samples by using the ImageJ Software.

Dermatitis scoring criteria

Mice were assessed macroscopically based on two main clinical criteria. Each region of the body, comprising head, neck, back and flank, affected by lesions, was given a score of 1 and the sum of these provided information of how expanded the lesions were. The other criteria were the characteristics of the lesion: punctuated small crusts, coalescent crusts, and ulceration, with a score of 1–3, respectively. The sum of both criteria represented the total severity score of the lesions. Scoring was performed by two independent researchers.

Isolation, culture and viability of PMKs

PMKs were obtained from Hoip^E-KO newborn pups, Tnfr1^KO;Hoip^E-KO and Tnfr1^KO;Hoil-1^E-KO adult tails according to established protocols⁶⁴. Briefly, skin was incubated with 0.25% Trypsin in Hank’s Balanced Salt Solution without calcium and magnesium (Stratech Scientific Ltd) overnight at 4 °C. On the following day, the dermis and epidermis were separated. Cell suspensions were cultured in Eagle’s minimal essential medium (Lonza) without calcium with 8% chelate foetal calf serum and penicillin–streptomycin (Sigma). PMKs were seeded in plates pre-coated with collagen I (Life technologies) for subsequent experiments. PMKs were cultured in medium supplemented with 20 µM Z-VAD-fmk (Abcam), 10 µM Necrostatin-1s (Cambridge Bioscience), 1 µM RIPK3 inhibitor (GSK2399872B) or 50 µg mL⁻¹ Etanercept (Enbrel®) (Pfizer and Pentaglobin from Biotest) for 4 days, with supplemented medium replaced every day. On the last day, cell viability was measured using the CellTiter-Glo Luminescent Cell Viability Assay Kit (Promega) following the manufacturer’s instructions. Alternatively, PMKs were treated for 24 h with the following ligands as indicated: 50 ng mL⁻¹ mouse iz-TRAIL, 50 ng mL⁻¹ CD95L-Fc, or 100 µg mL⁻¹ Poly(I:C) (Invitrogen).

Western blotting and immunoprecipitation (IP)

Western blotting was performed as previously described²¹. Briefly, PMKs were lysed in IP-lysis buffer (30 mM Tris-HCl [pH 7.4], 120 mM NaCl, 2 mM EDTA, 2 mM KCl, 1% Triton X-100, EDTA-free proteinase inhibitor cocktail (Roche) and 1× phosphatase-inhibitor cocktail 2 (Sigma) at 4 °C for 20 min. Lysates were denatured with reducing sample buffer and dithiothreitol at 95 °C for 10 min. Proteins were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (NuPAGE) and analyzed by western blotting with antibodies (all primary antibodies were used at a 1/1000 dilution) against HOIP (custom-made, Thermo Fisher Scientific), HOIL-1²¹, Sharpin (14626–1-AP, ProteinTech), actin (A1978, Sigma), tubulin (T9026, Sigma), FADD (Assay Design, AAM-121, RIPK1 (610459, BD), cleaved caspase-8 (9429, Cell signaling), MLKL (MABC604, Millipore), TNFR1 (ab19139, Abcam), phosphorylated IκBα (9246, Cell Signaling), IκBα (9242, Cell Signaling) and linear ubiquitin (MABS199, Millipore). Isolation of native TNFR1-SC and FADD IP were performed as previously described²⁶. Briefly, PMKs were cultured in the presence of 20 µM Z-VAD-fmk (Abcam; broad spectrum caspase inhibitor) and, in the case of TNFR1-SC analysis, stimulated with 0.5 μg mL⁻¹ 3xFlag-2xStrep-TNF for the indicated times. Control cells were left untreated. Cellular lysates were subjected to anti-Flag IP using M2 beads (SIGMA; Schnelldorf, Germany) for 16 h. For FADD IP, lysates were incubated with anti-FADD antibody (sc-5559, Santa Cruz) and protein G Sepharose Beads (GE Healthcare) at 4 °C for 4 h.

Flow cytometry

Cell suspensions obtained from skin samples were fluorescently labelled with Fixable Viability Dye eFluor® 780 (eBioscience). Samples were then stained with antibodies against the following cell surface markers: CD45-APC, CD45-AF700, CD3-PerCP/Cy5.5, CD4-FITC, CD8-PE/Cy7, GR1-FITC, GR1-PE/Cy7, F4/80-PE, F4/80-BV786, CD11b-Percp/Cy5.5 (Biolegend), and CD19-BV650 (Biolegend). Data were acquired with a LSRFORTESSA X-20 (BD) or Accuri (BD) with subsequent analysis using the FlowJo software.

Statistics

Data were analyzed with the GraphPad Prism 6 software (GraphPad Software) or Microsoft Excel. Data shown in graphs represent the mean values ± s.e.m., as indicated in the figure legends. Preliminary data sets were used to determine the group size necessary for adequate statistical power. Statistical analyses were performed by unpaired two-tailed Student’s t test. For multiple grouped comparisons, two-way analysis of variance was applied. Statistical significance in survival curves was determined using a log-rank test. A P value of >0.05 was considered not significant (NS), whereas P ≤ 0.05 was indicated with one asterisk (*), P ≤ 0.01 with double asterisks (**), P ≤ 0.001 with triple asterisks (***) and P ≤ 0.0001 with four asterisks (****). In all cases, comparisons were made between the indicated knockout (KO) mice and the respective littermate controls.

Competing interests

J.B. and P.J.G. are employees of GSK. The remaining authors declare no competing interests.