https://orcid.org/0000-0002-0151-6477
Abstract
Epidermal development and maintenance are finely regulated events requiring a strict balance between proliferation and differentiation. Alterations in these processes give rise to human disorders such as cancer or syndromes with skin and annexes defects, known as ectodermal dysplasias (EDs). Here, we studied the functional effects of two novel receptor-interacting protein kinase 4 (RIPK4) missense mutations identified in siblings with an autosomal recessive ED with cutaneous syndactyly, palmoplantar hyperkeratosis and orofacial synechiae. Clinical overlap with distinct EDs caused by mutations in transcription factors (i.e. p63 and interferon regulatory factor 6, IRF6) or nectin adhesion molecules was noticed. Impaired activity of the RIPK4 kinase resulted both in altered epithelial differentiation and defective cell adhesion. We showed that mutant RIPK4 resulted in loss of PVRL4/nectin-4 expression in patient epidermis and primary keratinocytes, and demonstrated that PVRL4 is transcriptionally regulated by IRF6, a RIPK4 phosphorylation target. In addition, defective RIPK4 altered desmosome morphology through modulation of plakophilin-1 and desmoplakin. In conclusion, this work implicates RIPK4 kinase function in the p63-IRF6 regulatory loop that controls the proliferation/differentiation switch and cell adhesion, with implications in ectodermal development and cancer.
Introduction
The receptor-interacting protein kinase 4 (RIPK4) is a member of the RIP serine/threonine kinase family, key regulators of cellular stress response, host defense and barrier integrity (1). RIPK4 presents a highly conserved N-terminal region composed by a catalytic N-lobe and a regulatory C-lobe with conformational flexibility crucial to control kinase activity and switching between different cellular responses. Instead, the C-terminal region consists of a protein–protein interaction domain with 12 ankyrin repeats, which modulate RIPK4 function and specificity (2). RIPK4 is widely expressed in embryonic and mature tissues and plays its major role in epithelial development, homeostasis and inflammatory response (3,4).
In humans, biallelic variants in RIPK4 cause different genetic syndromes featuring ectodermal derivative defects (i.e. EDs) variably associated to craniofacial (ankyloblepharon, cleft lip and/or palate, oral synechiae), limbs (pterygia) and/or genital anomalies (5). Clinical variability ranges from an early lethal phenotype observed in Bartsocas-Papas syndrome (BPS or autosomal recessive Popliteal Pterygium Syndrome—PPS; OMIM#263650) (6,7), to the milder form termed Curly Hair, Ankyloblepharon and Nail Dysplasia (CHAND; OMIM#214350) (8,9). Accordingly, Ripk4-deficient mice display severe anomalies of ectodermal derivatives, in addition to short limbs, a tail partially sticked to the body, fusion of external orifices, including the nose, mouth and anus, and impaired keratinocyte differentiation and skin barrier formation. These mice die within few hours after birth because of suffocation (10–12).
The role of RIPK4 in keratinocyte differentiation is mediated by the phosphorylation of its target interferon regulatory factor 6 (IRF6), whose mutations cause a distinct PPS with pits and/or sinuses of the lower lip, and cleft lip and/or cleft palate, inherited in autosomal dominant fashion (also known as Van der Woude syndrome; OMIM#119300) (13). Indeed, RIPK4 and IRF6 were shown to operate in the same biological processes in the epidermis by functioning as a signaling axis to ensure epidermal development and differentiation and skin integrity, with RIPK4 required for IRF6 activity (4,14). In addition, as observed in Ripk4 mouse models, Irf6-deficient mice display abnormal skin and epidermal adhesions at several sites (15,16). Pathological adhesion between intimately apposed epithelia during embryogenesis is accounted to defect of the ‘fence function’ of the periderm, a transient epithelial layer composed by highly polarized cells that express a specific pattern of cell adhesion proteins. Both RIPK4 and IRF6 have been indicated as key players in periderm formation, and their expression, together with that of several components of junctional complexes, is controlled by p63, the master regulator of periderm formation and maintenance (17–19). Indeed, further phenotypic overlaps exist between the RIPK4-disease spectrum and TP63-pathies (Hay-Wells, OMIM#106260, and ectrodactyly, ectodermal dysplasia, and cleft lip/palate syndrome, OMIM#604292), mostly revealed by anomalies of periderm function (e.g. pterygia and synechiae formation) (20).
Here, we report novel biallelic mutations in RIPK4 leading to a distinct EDs dominated by hair and nail anomalies, cutaneous syndactyly and hyperkeratosis. Such phenotype coincides with Ectodermal Dysplasia Syndactyly Syndrome (EDSS1, OMIM#613573), an independent clinical entity which we previously linked to genetic alterations in nectin-4, a cell adhesion molecule relevant for adherens junctions function and formation (21). By studying the functional effects of identified RIPK4 variants, we unveiled a novel role of RIPK4 in fine-tuned regulation of cell–cell adhesion, thus strengthening the importance of adhesion molecules and their accurate regulation in the pathogenesis of craniofacial and skin defects.
Results
Patient phenotypes
Two siblings were seen in our rare disease outpatient clinics seeking for a diagnosis of their disorder. Both were affected by a previously undiagnosed form of EDs with cutaneous syndactyly of hands and feet associated to mild hyperkeratosis of palms and soles. The proband (Supplementary Material, Fig. S1A, II:1) is a male individual seen at 32 years of age displaying early scalp alopecia, sparse, curly and brittle hair with tendency to break. He was born from non-consanguineous healthy parents after an uneventful pregnancy; at birth cutaneous syndactyly affecting digits 3–4 was evident, while he was told by his parents to be born with ‘closed eyes’ and that the thin line of skin that joined the eyelids had been surgically removed at birth. He had normal psychomotor development and as a teenager he started suffering of hyperkeratosis mainly localized at the skin folds of the hands and the soles. Nails were dysplastic (Fig. 1A). His younger sister (Supplementary Material, Fig. S1A, II:3), seen at age 27 years, suffered from an overlapping phenotype dominated by abnormal, brittle, uncombable and sparse hair, nail dysplasia and cutaneous syndactyly of hands and feet. Her nose was large with broad base and distinctive pseudo-clefts starting from the nostrils were evident. The skin was thin and dry with hyperkeratosis observable at the skin folds of the hands and at the soles (Fig. 1A).
Compound heterozygous RIPK4 mutations cause Ectodermal Dysplasia Syndactyly Syndrome in two Italian siblings. (A) Clinical and histological presentation of the affected siblings. Upper panels, images from the proband (II:1), lower panels from the younger sister (II:3), showing early alopecia with sparse, curly and brittle hair, cutaneous syndactyly affecting toes and/or digits, dysplastic nails, hyperkeratosis observable at the soles and at the skin folds of the hands, broad nasal base with pseudo-clefts starting from the nostrils and oral synechiae. Histopathology of the lesional skin from the plantar region of the patient II:1 shows epidermal hyperplasia with marked hyperkeratosis and focal parakeratosis, initial spongiosis and lymphocytic perivascular dermal infiltrate. Hematoxylin–eosin staining. Scale bar 200 μm. (B) Summary of known and novel variants associated to RIPK4-pathies. In red, blue and green are shown mutations associated with the early lethal PPS (Bartsocas Papas syndrome), adult PPS and CHAND syndrome, respectively; in bold, the novel variations here described. Numbering is as for human RIPK4 NP_065690. *Homozygous mutations; **Compound heterozygous mutations. Sequence alignments of the conserved regions in the kinase (left side) and ankyrin (right side) domains of human RIPK4 orthologous proteins are shown below the scheme. The multiple sequence alignment was performed via Clustal Omega software. The conserved residues at positions 241 and 428 mutated in the patients, the tryptophan residue at position 266, and two degrons (residue 378–383 and 436–440) are highlighted with boxes. ID, Intermediate Domain; ANK, Ankyrin repeat sequence. (C) Cartoon representations of the 3D models of the separated kinase and ankyrin domains of human RIPK4. On the left panel, the kinase domain (residues 1–288), composed of N-lobe, containing the ATP-binding site, and C-lobe containing the GHI helical subdomain (αG, αH and αI). The Arg241/Trp266 interacting couple of the GHI helical subdomain are shown in sticks and the magnification highlights their cation-pi interaction. The ankyrin domain (residues 375–784) is shown in the right panel. Pro428 localizes between the 378–383 and the 436–440 phosphodegrons. The latter phosphodegron and P428 belong to the second ankyrin repeat.
Identification of two novel missense variants in RIPK4
Based on the patient clinical manifestations compatible with EDSS1, the Poliovirus receptor-related protein 4-like gene (PVRL4) was first analyzed by Sanger sequencing as previously described (21), and disease-causative mutations were excluded. Clinical exome sequencing datasets were then generated from constitutive DNAs of both affected siblings to identify variants in known disease-causative genes. Variant filtering process led to the identification of 37 single nucleotide variations shared between the affected siblings in 9 distinct genes (Supplementary Material, Tables S1 and S2), including two heterozygous missense variants in the RIPK4 gene (NM_020639.2): c.722G > A (p.Arg241His) in exon 5 and c.1283C > T (p.Pro428Leu) in exon 8. Both were classified as damaging/deleterious by a variety of prediction algorithms, including PolyPhen, SIFT and Provean. In addition, these variants were extremely rare in population databases: the c.1283C > T was found only at the heterozygote state in 1/249312 and in 1/152226 subjects in GnomAD v2.1.1 and GnomAD v3.1.2, respectively (Supplementary Material, Table S3). Both variants were annotated in the COSMIC catalogue as somatic mutations detected in two gastric cancers (c.722G > A) and one thyroid carcinoma (c.1283C > T). Sanger sequencing confirmed the c.722G > A and c.1283C > T variants at the heterozygote state in the healthy mother and father, respectively, in line with autosomal recessive inheritance (Supplementary Material, Fig. S1A). Neither change affected exon splicing as witnessed by normal exon 5 and exon 8 amplification by reverse transcriptase (RT)-PCR analysis of the RNA purified from patient cultured keratinocytes.
Effect of the newly identified mutations on RIPK4 activity
Modeling studies of the newly identified RIPK4 mutations showed that arginine 241 locates in the C-lobe of the kinase domain in the loop of the GHI helical subdomain that contains the dimerization site and regulates the kinase activity through the precise orientation of helix F (residues 201–219) (22,23). In silico analysis predicted an energetically significant interaction between the positively charged residue Arg241 and the aromatic residue Trp266. This type of non-conventional interactions, called cation-pi interactions, is crucial for protein stability and function as reflected by their conservation in several protein families (24,25). Interestingly, also the Arg241/Trp266 interacting pair is evolutionarily conserved in RIPK4 proteins indicating a role in stabilizing the GHI helical subdomain architecture (Fig. 1B and C). We hypothesized that the loss of this interaction, due to the p.Arg241His mutation, disrupts both the tight interconnection network between the C- and N-lobe of the RIPK4 kinase domain and the propensity to form active homo-oligomers, lowering the (auto-) phosphorylating efficiency of the kinase domain. Indeed, the tendency of histidine residue to form cation-pi interactions is lower than arginine due to its scarce propensity to be protonated at neutral pH (24). Proline 428, instead, falls in the second ankyrin repeat and lies in between two experimentally validated phosphodegron consensus sequences, structural determinants recognized by ubiquitin E3 ligases only when phosphorylated (378-DSAFSS-383 and 436-DSGAS-440) (26) (Fig. 1B and C). Substitution of a constraining proline residue may alter the protein local conformation and impair the interface regions that mediate interaction with regulatory kinases or the proteasomal complex.
To ascertain the pathogenicity of the two novel mutations, wild-type and mutated recombinant RIPK4 were ectopically expressed in HEK293 cells and kinase activity was evaluated by autophosphorylation analysis. As expected, wild-type RIPK4 occurred in two different forms and phosphatase treatment resulted in loss of the higher molecular weight hyper-phosphorylated protein. In contrast, the expression of RIPK4 protein carrying the p.Arg241His variant resulted in the complete loss of the hyper-phosphorylated pool, suggesting that this mutation inhibits the RIPK4 kinase activity. Differently, in vitro autophosphorylation was only reduced for the p.Pro428Leu mutant, denoting that homodimerization and kinase activity were not totally impaired by this aminoacidic substitution (Fig. 2A). The effect of these RIPK4 mutations was further investigated by evaluating activation and subsequent nuclear translocation of the RIPK4 phosphorylation target, IRF6, following their co-expression in HEK293 cells. By western blotting and immunofluorescence analysis, IRF6 was mainly detected in the cytoplasm of the mock co-transfected cells, and as expected an increased nuclear localization was observed in cells co-expressing wild-type RIPK4. However, IRF6 nuclear translocation was significantly reduced when it was co-expressed with the mutant RIPK4 proteins, either p.Arg241His or p.Pro428Leu (Fig. 2B and Supplementary Material, Fig. S1B). The decreased IRF6 activation was not due to reduced expression of mutant RIPK4 since wild-type and mutant proteins were expressed at equivalent levels (Supplementary Material, Fig. S1C).
Novel identified variants interfere with RIPK4 activity and its proteasomal regulation. (A) In vitro phosphorylation analysis showing impaired autophosphorylation of RIPK4 mutant proteins. Myc-tagged wild-type (WT) and mutated (241H and 428L) RIPK4 recombinant proteins were transiently expressed in HEK293 and the presence of the hyperphosphorylated form was assessed by immunoblotting on total cell lysate using anti-Myc antibody. To confirm the hyperphosphorylated status of RIPK4-WT protein, the cell extract was treated with CIP before immunoblotting (left panel). The data are representative of two independent experiments. (B) In vitro IRF6 activation assay showing impaired activity of mutant RIPK4 proteins. HA-tagged IRF6 protein was transiently co-expressed with Myc-tagged WT or mutated (241H and 428L) RIPK4 proteins in HEK293. Thirty hours after transfection, protein fractions enriched for nuclear or soluble cytosolic components (nuclei and cytosol, respectively) were separated and analyzed by western blotting using anti-HA antibody to evaluate the IRF6 subcellular distribution. Anti-Myc antibody was used to confirm recombinant RIPK4 expression, anti-GAPDH and anti-Lamin B1 antibodies were used as cytosolic and nuclear fractions loading controls, respectively. On the right, graphical representation of the relative ratio between nuclear (active) and cytosolic (inactive) IRF6, calculated on four biological replicates, is shown. P-values were calculated using two-tailed paired t-test. *P < 0.05. (C) Mutant RIPK4 proteins are not targeted for proteasomal degradation in keratinocytes. Total protein extracts from patient (Pt) and control (CTR) primary keratinocytes treated with the proteasome inhibitor MG132 (10 μM) for 4 h were analyzed by immunoblotting with anti-RIPK4 monoclonal antibody. GAPDH levels were used as loading control. Pt cells present higher levels of RIPK4 protein that do not further increase after MG132 treatment. Slight phosphorylated protein accumulation, indicated with an asterisk (*), can be observed only in CTR cells treated with MG132. A representative experiment out of four biological replicates is shown. Results of densitometric analysis of RIPK4 levels, normalized to GAPDH, are shown below ±S.D.
Altogether these results indicate that the novel identified RIPK4 mutations lead to functionally defective proteins. In particular, the p.Arg241His abrogates the kinase function, whereas the p.Pro428Leu, though still able to preserve RIPK4 kinase activity, affects its ability to activate IRF6 and partially prevents IRF6 nuclear translocation.
Effect of the RIPK4 mutations on protein expression and proteasomal degradation in patient keratinocytes
To deeply investigate the pleiotropic RIPK4 functions in human epidermis, a primary keratinocyte cell culture from patient II:3 skin biopsy was established. First, we compared the endogenous RIPK4 protein expression and subcellular localization between patient and age- and sex-matched control cells. No significant differences in protein distribution were detected by immunofluorescence, but surprisingly the signal intensity was increased in patient cells (Supplementary Material, Fig. S2A). Immunoblot analysis confirmed a significant increase of RIPK4 protein levels in in vitro cultured patient keratinocytes (Fig. 2C) not paralleled by an increment at mRNA level (Supplementary Material, Fig. S2B). To establish if the augmented RIPK4 stability observed in the patient cells was due to impaired proteasomal degradation, primary keratinocytes were grown in presence of the proteasome inhibitor MG132. While, as expected, in control cells the treatment resulted in accumulation of RIPK4 and of its hyper-phosphorylated pool, RIPK4 protein did not show any further increment in patient keratinocytes, but it was even slightly diminished, probably due to drug toxicity (Fig. 2C). Lack of protein degradation is actually expected for the p.Arg241His protein, since this mutation affects its catalytic activity and impedes hyper-autophosphorylation and physiological turnover (26). On the other hand, we cannot exclude that the p.Pro428Leu mutant accumulates as well, as loss of the constraining proline residue may affect the interaction of RIPK4 with regulatory kinases or the proteasomal complex, thus protracting mutant RIPK4 half-life.
Altogether these results suggest that the novel identified mutations impair the RIPK4 proteasomal regulation and cause accumulation of long-lasting but malfunctioning kinases in patient keratinocytes.
Altered differentiation in RIPK4 mutated patient keratinocytes
Microscopical analysis of patient keratinocyte colonies revealed relevant morphologic differences: 72 h after plating on 3T3-J2 feeder layer, patient keratinocytes generated larger colonies compared to control cells. No significant variation of cell number per colony was observed suggesting that the increased colony dimension was not due to increased proliferation (Fig. 3A). Colonies were less stratified indicative of a defect in the differentiation program. Keratinocytes were less compacted and wider intercellular spaces could be detected (Fig. 3A, lower panels). The expression of several early and late keratinocyte differentiation markers (i.e. KRT10, IVL, LOR, FLG and SPINK5) and of transcriptional regulators of the differentiation switch (i.e. IRF6 and OVOL1) was analyzed in primary keratinocytes grown in conditions that promote either proliferation or differentiation (low and high calcium supplement, respectively). Consistently with a differentiation defect, transcription of these genes was not correctly induced in patient cells after calcium supplement. In addition, while no significant differences were observed in TP63 expression at transcriptional level, accumulation of p63 protein was revealed by western blotting in in vitro patient differentiating keratinocytes, likely imputed to the reduced IRF6 activity (Fig. 3B and Supplementary Material, Fig. S3). Defective differentiation was paralleled by augmented clonogenic potential, as demonstrated by the Colony Forming Efficiency assay (Supplementary Material, Fig. S3B). A strong reduction of abortive colonies was also observed, compatible with a reduced number of terminally differentiated cells.
Altered differentiation and cell adhesion in RIPK4 mutated patient keratinocytes. (A) Morphological analysis of keratinocyte colonies showing larger colony size and reduced cell density in patient’s cell cultures. Patient (Pt) and control (CTR) primary keratinocytes were seeded on 3T3-J2 feeder layer and grown in ‘keratinocyte medium’ supplemented with growth factors and FCS, in order to generate stratified colonies composed of both proliferating and differentiated keratinocytes. Cells were cultured for 72 h before fixation and immunofluorescence analysis with anti-E-cadherin antibody. Lower panels, higher magnification showing wider intercellular spaces between patient cells (highlighted in the dashed box). Note that patient keratinocytes do not stratify, indicative of impaired differentiation. Scale bars 30 μm. Graphical representation of the morphological analysis is shown below. The area of more than 50 colonies from different E-cadherin-stained samples was evaluated for both patient and control keratinocyte cultures using ImageJ software. Average values of colony area, number of cells per colony and cell area are shown. a.u., arbitrary units; ****, significant differences (P < 10−4, two-tailed unpaired t test with Welch’s correction). (B) Expression analysis of keratinocyte differentiation markers in RIPK4 defective patient showing aberrant differentiation program. Relative INV, LOR, KRT10, FLG, SPINK5, OVOL1, IRF6 and TP63 mRNA expression, as determined by RT-qPCR using cDNA obtained from primary keratinocytes of patient II:3 (Pt) and unaffected individual (CTR) cultured in proliferative (P, low calcium medium) or differentiating (D, 1.2 mm CaCl2 for 72 h) conditions. Levels are normalized to GAPDH or HPRT1 values. The relative expression values were determined via the ΔΔCt method. Average values ±SD of two independent experiments is shown.
In conclusion, RIPK4 defective function in patient keratinocytes affects proliferation/differentiation pathways and results in morphological changes paralleled by accumulation of p63 and defective induction of the expression of both early and late keratinocyte differentiation markers.
RIPK4 regulates PVRL4 gene transcription through IRF6 activation
Given the overlap of patient clinical features with other EDs, the expression of the EDSS1 defective protein nectin-4, which is normally induced during keratinocyte differentiation, was analyzed. Almost total absence of nectin-4 was observed in patient II:1 plantar skin, while membrane staining was detected in all suprabasal keratinocytes of the control skin biopsy. However, E-cadherin (E-Cad) appropriately decorated the cell borders indicating that adherens junctions were assembled (Fig. 4A). Altered PVRL4/nectin-4 expression was also confirmed in in vitro differentiated keratinocytes from patient II:3, both at protein and mRNA level (Supplementary Material, Fig. S4A–C).
RIPK4 regulates transcriptional expression of PVRL4 through IRF6 activation. (A) Analysis of expression and localization of adherens junction markers in patient skin showing loss of nectin-4 staining at cell–cell contact sites. Frozen sections of plantar skin obtained from an unaffected individual (CTR) and patient II:1 (Pt) were immunostained with anti-nectin-4 (N4) and anti-E-cadherin (E-Cad) antibodies. In the control sample, both N4 and E-Cad preferentially stains along cell–cell adhesion sites in all keratinocytes. In the patient skin, although a similar staining is observed with E-Cad, nectin-4 expression is strongly reduced, and its distribution altered. Scale bars 20 μm. (B) Analysis of the effects of RIPK4 activation in human primary keratinocytes showing a functional relationship between RIPK4 and PVRL4, IRF6 and TP63 gene expression. Relative PVRL4, IRF6 and TP63 mRNA levels were evaluated by quantitative RT-PCR (RT-qPCR) in human primary keratinocytes obtained from an unaffected individual (CTR) and patient II:3 (Pt) after treatment with the PKC activator, phorbol 12-myristate 13-acetate (PMA). Levels were normalized to GAPDH. The relative expression values were determined via the ΔΔCt method. Average values ±SD of 3 independent experiments is shown. Asterisks, significant differences (**, P < 10−2; ***, P < 10−3; ****, P < 10−4, One-way ANOVA multiple comparisons test with Bonferroni’s correction). (C) Analysis of the effects of IRF6 gene silencing in human keratinocytes showing loss of PVRL4 expression. Relative IRF6, PVRL4, PVRL1 and RIPK4 mRNA levels were evaluated by RT-qPCR in human primary keratinocytes transfected with IRF6-specific siRNA (siIRF6), and cultured in vitro in proliferating (low calcium, T0 = 24 h after transfection) or differentiating conditions (T2 and T3 = 48 and 72 h after 2 mm CaCl2 supplementation, respectively). Transfection with a scramble sequence (siCTR) was used as control. Levels were normalized to GAPDH. The relative expression values were determined via the ΔΔCt method. Average values ±SD of 3 independent experiments is shown. Asterisks, significant differences (*, P < 10−1; **, P < 10−2; ****, P < 10−4, One-way ANOVA multiple comparisons test with Bonferroni’s correction). (D) Chromatin immunoprecipitation assay showing IRF6 binding to the PVRL4 regulatory region (binding site 1, BS1). Chromatin immunoprecipitation using anti-IRF6 antibody was performed on primary human keratinocytes grown in proliferative (P, low calcium medium) and differentiating (D, 1.2 mm CaCl2 for 72 h) conditions, and followed by quantitative PCR amplification of BS1 region in PVRL4 intron 4 (arrows in the lower panel represent the primer pair used in PCR). Data are shown as percentage of input.
To establish whether a direct functional relationship exists between RIPK4 and nectin-4, PVRL4 expression was analyzed in control and patient keratinocytes treated with phorbol 12-myristate 13-acetate (PMA), a potent PKC activator that induces keratinocyte differentiation through RIPK4. A significant increase in PVRL4 levels was observed in normal human keratinocytes treated with PMA, concurrent with IRF6 induction and TP63 inhibition. Interestingly, patient cells showed neither the TP63 reduction nor the PVRL4 and IRF6 induction (Fig. 4B), supporting the hypothesis that PVRL4 expression is under RIPK4 control.
To obtain further insights into the regulation of PVRL4 by RIPK4/IRF6 axis, PVRL4/nectin-4 expression was analyzed in IRF6-depleted primary keratinocytes. While PVRL4/nectin-4 expression increased following calcium induced differentiation in control keratinocytes, no induction was observed in cells silenced for IRF6. Expression of the PVRL4 homologous gene, PVRL1 and RIPK4 remained unaffected (Fig. 4C and Supplementary Material, Fig. S4D). In addition, a repeated analysis of a genome-wide IRF6 ChIP-seq dataset previously generated in differentiating human primary keratinocytes (27) revealed an IRF6-binding region in the intron 4 of the PVRL4 gene (binding site 1, BS1). This region, highly conserved in mammalian, contained DNAse hypersensitive sites, evocative of an open chromatin state, and was enriched for histone H3K4Me1 marking transcriptional enhancers (Supplementary Material, Fig. S4E). To confirm this observation, ChIP-qPCR analysis of the IRF6-binding region identified in PVRL4 genomic locus (BS1) was performed in human primary keratinocytes in proliferative and differentiating condition. Consistent with a direct regulation of PVRL4 expression by IRF6 binding, a 30% enrichment of IRF6 protein bound to BS1 region in PVRL4 intron 4 was observed in differentiating versus proliferating cells (Fig. 4D).
Taken together, these data indicate that PVRL4 is a direct transcriptional target of IRF6.
Functional effects of the RIPK4 missense mutations on cell–cell adhesion in patient keratinocytes
Histological observation of lesional patient plantar skin by hematoxylin–eosin staining showed epidermal hyperplasia, spongiosis and superficial dermal lymphocytic infiltrate (Fig. 5A). Dilated intercellular spaces, cell–cell adhesion interruptions and loosening of cell–cell cohesion were features observed in in vitro cultured patient keratinocyte colonies (Fig. 3A, higher magnification). None of these features were previously reported in EDSS1 specimens suggesting that additional adhesion complexes could be affected by mutant RIPK4.
Desmosomal defects in RIPK4 defective patients. (A) Histological analysis of patient lesional skin, showing focal loss of cell cohesion and widening of intercellular spaces in the spinous layer. Higher magnification of the hematoxylin–eosin staining of patient II:1 plantar skin shown in Figure 1B and here reproduced in miniature, above the frame. Spinous layer keratinocytes have lost their typical polygonal shape and several micro-detachments are evident between the rounded cells (arrows). Insert: hematoxylin–eosin staining of control plantar skin. Scale bar 20 μm (scale bar of the miniature 100 μm). (B) Analysis of expression and localization of desmosomal markers showing altered expression and distribution in patient epidermis. Frozen sections of plantar skin obtained from an unaffected individual (CTR) and patient II:1 (Pt) were immunostained with anti-desmoglein-1 (DSG1), anti-plakophilin-1 (PKP1) and anti-desmoplakin (DSP) antibodies (inserts, higher magnification pictures). In the control sample, both DSG1, PKP1 and DSP preferentially stains along cell–cell adhesion sites in all keratinocytes. In the patient skin, DSG1 staining at cell membrane is outstretched suggestive of widening of intercellular spaces between keratinocytes; PKP1 staining intensity is strongly reduced and almost absent at cell–cell contact sites (insert, longer exposure of the higher magnification); DSP is aberrantly distributed and shows a diffuse staining in the cytoplasm. Scale bars 20 μm. (C) TEM micrographs showing numerous wide intercellular spaces and low electron-density patches, resembling vacuoles, in the proximity of cell borders in patient II:1 skin (Pt). A representative micrograph of control patient skin (CTR) showing well-preserved ultrastructure is shown in the panel on the right. Scale bars: 0.5 μm.
Regarding desmosomes, desmoglein-1 (DSG1) immunofluorescence analysis showed intercellular junctional staining throughout the different epidermal layers in both patient and control plantar skin. However, a more diffusive DSG1 staining was detected in the patient lesional skin, compatible with the widening of intercellular spaces between keratinocytes (Fig. 5B). In addition, plakophilin-1 (PKP1) expression was remarkably reduced (Fig. 5B), as also confirmed in in vitro cultured keratinocytes from patient II:3, at both protein and mRNA level (Supplementary Material, Fig. S5A and B). Finally, reduced desmoplakin (DSP) recruitment at cell–cell junctions was observed with a more diffuse cytosolic staining (Fig. 5B). Electron microscopic examination of the patient skin showed widening of intercellular spaces and low electron-density patches, resembling vacuoles, in proximity of cell borders, suggestive of impaired cell–cell adhesion (Fig. 5C). The most evident alteration was the reduction of keratin filaments crossing the desmosomes, as evidenced by electron-negative areas with a vacuolated aspect adjacent the inner and the outer dense plaques. Consequently, plasma membranes were probably stretched, with points of intracellular shredding at the level of desmosomal plaque. The electron-dense extracellular mid-line was less marked in patient skin; however, cross bridges between the plaques with intercalated particles were often seen.
In conclusion, mutations in RIPK4 result in structural perturbation of desmosomes, likely consequent at least in part to altered PKP1 expression. However, clinical symptoms are detectable only in particularly stressed body sites exposed to friction/pressure (i.e. plantar skin), and this defect is apparently compensated by the ability of keratinocytes to still adhere in homeostatic conditions.
Discussion
Biallelic variants in RIPK4 give rise to clinically distinct ED syndromes presenting a phenotypic continuum ranging from early lethal BPS to milder forms such as PPS and CHAND (5,9). In the current study, we identified novel RIPK4 biallelic variants in two siblings with cutaneous syndactyly associated to hair defects, alopecia, nail dysplasia and hyperkeratosis. This phenotype further expands the clinical spectrum seen in RIPK4-pathies. In fact, both our siblings displayed a cutaneous syndactyly that is evocative of the skin webbings across major joints seen in PPS (Fig. 1A) and not reported in patients with CHAND syndrome (8,9,28). Seeking for genotype–phenotype correlations, we reviewed the clinical pictures resulting from RIPK4 allelic variants (Fig. 1B). Interestingly, in nearly all cases, homozygous missense changes located in the kinase domain or truncating variants resulted in early lethality (BPS phenotype), while missense variants in the ankyrin repeat domain were found in survivors with PPS (featuring cutaneous webbing across one or more major joints, cleft lip and/or palate). The only exceptions are the two homozygous missense changes p.Gly163Asp and p.Glu284Lys affecting the kinase domain found in CHAND syndrome, representing the mildest end of the RIPK4-phenotypic spectrum (8,9,28). It is likely that variable effect of distinct mutations on the kinase activity, combined with the RIPK4 pleiotropic role in multiple signaling pathways, reflects variability in clinical presentation, although further studies would be needed to confirm this hypothesis. In line with such observations, the herein reported p.Arg241His and p.Pro428Leu variants, located in the kinase domain and in the second ankyrin repeats of RIPK4, respectively, the first to be reported in compound heterozygosity, give rise to an intermediate phenotype (Fig. 1A and B).
Based on the revision of RIPK4-associated phenotypic spectrum, we suggest handles for the diagnosis of RIPK4-pathies to include as mandatory feature any trait resulting from epithelial fusion (mainly oral/ocular synechiae) associated to skin webs at flexural surfaces (including syndactyly of hands/feet) and typical EDs features (hair/nail dysplasia). The epithelial fusion manifestations are thus characteristics of RIPK4 deficiency in humans as well as in mice models and has been mainly ascribed to abnormalities of periderm, the uppermost layer of embryonal epidermis, consisting of a highly polarized distribution of adhesion complex proteins. Indeed, analysis of early features of Ripk4−/− mouse embryos evoked a role of periderm development failure in the pathogenesis of the abnormal inter-epithelial adhesions that characterize RIPK4-pathies (29). In particular, Ripk4−/− periderm presents an aberrant apical localization of E-Cad in several surface epithelia (11) and keratinocyte-specific depletion of E-Cad rescued the tail-to-body and oral fusions in the double KO embryos (12). Similar epithelial fusions are observed in mice carrying loss-of-function mutations in genes involved in pathways regulated by RIPK4 activity, such as Irf6 (15) and Ikka (also known as Chuk) (30), whose mutations in humans cause Van der Woude and cocoon (OMIM#613630) syndromes, respectively, manifesting pterygia/epithelial fusions. Interestingly, most of the genes involved in periderm formation and/or functions, including RIPK4, IRF6 and CHUK, are direct transcriptional target of p63. Of note, TP63-pathies such as AEC syndrome show, in addition to ED, cleft lip/palate and limb defects, characteristic eyelid fusions (also called ankyloblepharon) similarly to what observed in our family (20). Surprisingly, even if p63−/− mice lack the periderm, they do not show the epithelial fusion phenotype. However, in addition to its effect on genes involved in periderm development, p63 also controls the expression of several adhesion molecules (including E-Cad), and their downregulation prevents epithelial fusions from occurring despite the absence of periderm (19).
Interestingly, the here described clinical picture presents phenotypic overlap not only with IRF6- and TP63-pathies but also with EDSS1, caused by biallelic mutations in PVRL4, that encodes for the cell adhesion molecule nectin-4, a major constituent, together with E-Cad, of adherens junctions (31). We previously demonstrated that PVRL4 expression is regulated during keratinocyte differentiation and postulated the existence of a regulatory link between IRF6 and PVRL4 (21,32). Consistently, PVRL4/nectin-4 expression was not induced in RIPK4 defective patient keratinocytes following in vitro differentiation either by high calcium or through direct PKC activation by PMA. Recent studies showed that RIPK4 mediates PKC-induced keratinocyte differentiation through activation of the IRF6 transcription factor, and that this activity is essential to promote skin development and keratinocyte differentiation (14). We demonstrated now that the newly identified RIPK4 mutations affect either the enzymatic activity and/or the regulative proteolytic degradation of the kinase and resulted in reduced IRF6 trans-activation. In addition, we outlined that IRF6 binds to PVRL4 intron 4 region, and that PVRL4 expression is strongly reduced in IRF6 depleted human keratinocytes. Accordingly, IRF6 represents the molecular link between RIPK4 and the cell adhesion function in keratinocytes by directly controlling PVRL4 expression.
However, nectin-4 deregulation could not justify all the cell adhesion-related morphological aspects observed in RIPK4 patient keratinocytes. Besides E-Cad staining was correctly present at cell–cell contact sites as previously observed in EDSS1 keratinocytes (33), and the signal was less sharp as compared to control skin, indicative of a loosening of the junctional complexes. In addition, several intercellular gaps were detected in patient keratinocyte colonies. None of these features was previously reported in EDSS1 specimens (21,33) suggesting the possible involvement of additional adhesion molecules. PKP1, a member of the armadillo repeat family of proteins crucial in maintaining desmosomal integrity, has recently been identified as a RIPK4 target molecule required for proper epidermal differentiation (34). Interestingly, biallelic mutations in PKP1 result in ED/skin fragility syndrome (EDSF, OMIM:604536) (35,36), an autosomal recessive disorder with several clinical, cellular and molecular overlaps with the herein described phenotype, such as the thickening of the plantar skin, dystrophic nails, altered hair follicle development/alopecia. Hence, we investigated by electron microscopy analysis the desmosome structure in our patient skin and demonstrated an increase of the medium interspace, beside desmosome length and density were not significantly altered. Consistent with these observations, PKP1 staining in patient lesional skin showed aberrant distribution and an overall reduction at both protein and transcript level. Concurrently, the PKP1 interactor DSP had a more diffuse cytoplasmic localization similarly to what observed in the skin of EDSF patients (37), while the transmembrane desmosomal cadherin DSG1 was not affected.
Altogether these data disclose a critical function of RIPK4 in the regulation of cell adhesion, playing therefore a crucial role both during development and in epithelial homeostasis. We hypothesize that the unbalanced expression of adhesion molecules observed in RIPK4-defective patients results, during embryogenesis, in loss of the ‘fence function’ that physiologically prevents spread of adhesion complexes onto the apical surface of peridermal cells, thus favoring intraepithelial fusions resulting in synechiae, pterygia, syndactyly and orofacial clefts. Moreover, alteration of the expression and/or function of either cell adhesion molecules or epithelial transcription factors affects the fine regulation of keratinocyte proliferation/differentiation program and results in altered homeostasis. These alterations are likely to be crucial in tissues with rapid renewal or in annexes, such as plantar skin, hair follicles and nails, where cell–cell junction dynamics plays a key role.
Our results broaden the description of the molecular framework involved in ectodermal morphogenesis and homeostasis, which includes intricate feedback regulatory loops between transcription factors and protein kinases, and downstream functional regulation of cell adhesion structures. Putting together available evidence, we propose the p63-dependent expression of both IRF6 and RIPK4 as the first mandatory step of the ectodermal-specific differentiation program (Fig. 6). IRF6 activation by RIPK4-dependent phosphorylation is then necessary to switch-off the initiation signal (i.e. p63 degradation), inhibit cell proliferation through OVOL1 expression and activate transcription of epithelial differentiation markers, including PVRL4/nectin4. In parallel, RIPK4 regulates cell adhesion function by phosphorylation of cell adhesion molecules (i.e. the desmosomal protein PKP1). p63, on the other side, controls also the expression of several other molecules that take part to the cell-adhesion program (Fig. 6). We envisage that such regulatory loop will be further implemented including signaling pathways downstream to cell adhesion.
Schematic representation of the link between RIPK4, p63, IRF6 and adhesion molecules. RIPK4 is part of the p63-IRF6 regulatory loop that control epidermal development and proliferation/differentiation balance in skin homeostasis through a multistep regulation of the cell–cell adhesion function. RIPK4, under the balanced control of p63 and IRF6 transcription factors, phosphorylates and activates IRF6 protein, which in turn induces p63 degradation, reducing proliferative potential of epithelial cells and inducing keratinocytes differentiation. Both p63, IRF6 and RIPK4 regulate expression or function (by phosphorylation) of cell-adhesion molecules mainly involved in adherens junction and desmosome assembly.
Altogether, these observations suggest that different pathways involved in distinct EDs (TP63, IRF6, CHUK and RIPK4-pathies) converge in altered cell adhesion programs that include both adherens junctions (e.g. through nectin-4) and desmosomes (e.g. through PKP1), contributing to delineate the embryological processes and molecular players implicated in ectodermal organogenesis and skin homeostasis. Alteration of this axis is thus essential for the patterning and polarized migration of cells during morphogenesis but also for homeostasis in mature tissues by perturbing the proliferation and differentiation balance and the adhesion properties of the cells. This aspect might be relevant also for cancer pathogenesis, as clearly outlined for single components of the pathway, such as p63 (38,39) and IRF6 (27), nectin-4 (40), PKP1 (37) and notably RIPK4 (34). Of note, the germline RIPK4 mutations here described were somatically detected in gastric and thyroid carcinomas (COSMIC catalogue). Accordingly, understanding of the mechanisms by which these activities are coordinated will be likely key to develop novel therapeutic approaches in cancer.
Materials and Methods
Transient transfections, protein phosphatase treatment and in vitro IRF6 activation assay
HEK293 cells grown to 70% confluence were transiently transfected with pCMV-Myc-RIPK4-wt, pCMV-Myc-RIPK4-241H, pCMV-Myc-RIPK4-428L or an empty vector, using Lipofectamine 2000 reagent (Invitrogen) and following manufacturer’s instruction. Thirty hours after transfection cells were harvested in Laemmli buffer and analyzed by western blotting. Cells transfected with pCMV-Myc-RIPK4-wt were also lysed in NP40 buffer (50 mm Tris pH 8.0, 150 mm NaCl, 1% NP40) supplemented with Complete protease inhibitor cocktail (Sigma-Aldrich). Lysates were cleared by centrifugation and treated with calf intestinal phosphatase (CIP; New England Biolabs), 100 units of enzyme in 50 μL. Treated and untreated samples were incubated at 30°C for 30 min and analyzed by western blotting. For the IRF6 activation assay, pCDNA-HA-IRF6 was co-transfected with pCMV-Myc-RIPK4-wt, pCMV-Myc-RIPK4-241H, pCMV-Myc-RIPK4-428L or an empty vector. Thirty hours after transfection cells were either fixed in 4% paraformaldehyde and processed for immunofluorescence or recovered by scraping in ice cold fractionation buffer (250 mm Sucrose, 20 mm Hepes pH 7.4, 10 mm KCl, 1.5 mm MgCl2, 1 mm EDTA, 1 mm EGTA) supplemented with Complete protease inhibitor cocktail. Cell suspension was then passed through a 25-gauge needle syringe 10 times (or until all cells were lysed) and left on ice for 20 min. The fraction N, enriched for the nuclear components, was recovered by centrifugation at 720 g (3000 rpm) for 5 min at 4°C, while supernatants, containing cytoplasm, cytoskeleton, membrane and mitochondria were collected as cytosolic fraction (Cy). Pellets were resuspended in 0.5 volumes of 2× Laemmli sample buffer. Equal volume of N and Cy fractions in Laemmli buffer were boiled 10 min at 95°C, separated on 10% SDS-PAGE, and analyzed by western blotting. To evaluate the RIPK4 activity, nuclear localization of IRF6 protein was assessed either by immunofluorescence (number of cells with nuclear IRF6 staining/total IRF6 positive cells ratio) or densitometric analysis of western blotting bands (nuclear IRF6/cytosolic IRF6 ratio).
IRF6 silencing
Primary keratinocytes from healthy individuals were transfected with stealth siRNA duplexes targeting human IRF6 (HSS105511–HSS105512–HSS105513) or siRNA negative control duplexes (Invitrogen) using Lipofectamine RNAiMax reagent (Invitrogen) according to the manufacturer’s instructions in KGM medium. Twenty-four hours after transfection (T = 0) the cells were induced to differentiate by supplementation with 1.2 mm Ca++ for 48 (T = 2) or 72 (T = 3) h. At each time point, cells were recovered and analyzed by western blotting and RT-qPCR. IRF6 silencing efficiency was verified at mRNA and protein levels.
Light microscopy and transmission electron microscopy
Tissue biopsies collected from the plantar skin of patient II:1 and a control individual were washed in phosphate-buffered saline (PBS) and immediately fixed in 2.5% glutaraldehyde (Agar Scientific, Cambridge Road Stansted Essex, UK) in PBS. To allow a correct penetration of the fixative, all samples were cut with a razor blade to a size of ~2 × 2 mm and stored with fresh glutaraldehyde (2.5%/PBS) at 4°C for at least 48 days.
Samples were, then, rinsed in fresh PBS (3 changes, 30 min each in stirring), post-fixed with 1% osmium tetroxide (OsO4) (Agar Scientific, Stansted, UK) in PBS for 1 h and rinsed again in PBS (1 × 30’). After that, biopsies were dehydrated in the ascending series of ethanol, with the following subsequent passages: EtOH 30%, EtOH 50%, EtOH 70% (1 × 10’), EtOH 95% (2 × 10’) and EtOH 100% (4 × 15’). Following the immersion in propylene oxide (PO) (2 × 20’) for solvent substitution, samples were infiltrated in PO/Epoxy resin (1:1) overnight, embedded in the epoxy resin EMbed-812 (Electron Microscopy Sciences, 1560 Industry Road, Hatfield, PA, USA) for 48 h at 60°C and sectioned using a Reichert-Jung Ultracut E ultramicrotome. Semi-thin sections (1 mm thick) were stained with Toluidine Blue, examined using light microscopy (LM; Zeiss Axioskop) and photographed using a digital camera (Leica DFC230).
Ultrathin sections (60–80 nm) were cut with a diamond knife, mounted on copper grids and contrasted with Uranyless and lead citrate (SIC, Rome, Italy). They were examined and photographed using Zeiss EM10 and Philips TEM CM100 Electron Microscopes operating at 80 kV.
Assessment of desmosome diameter and numerical density
ImageJ software was used to measure the diameter of desmosomes and their numerical density on low-magnification transmission electron microscopy (TEM) micrographs of control and patient. The numerical density was determined by counting the number of desmosomes present on an area of 25 μm2 and on different serial sections (distance between the sections: 3–4 μm). Values were expressed as desmosome numerical density/25 μm2 of the sample area. All data are expressed as means ± standard deviation (SD). Statistical comparisons were performed using unpaired t test with Welch’s correction for post-hoc analysis (GraphPad InStat. GraphPad Software, La Jolla, USA). Differences in values were considered significant if P < 0.05.
Acknowledgements
We are thankful to the family for participating to this study. We would like to thank Dr Elena Dellambra for critical reading of the manuscript and helpful discussion, and Dr Arianna Di Daniele for the editing of the figures.
Conflict of Interest statement. The authors declare that they have no conflict of interest.
Funding
This work was supported by a grant from the Italian Ministry of Health (GR2013-02356227 to F.B.) and (RC2020-2756828 toD.C.).
Members of the Italian Undiagnosed Diseases Network: Domenica Taruscio (Coordinator, National Center Rare Diseases, Istituto Superiore di Sanità, Rome, Italy), Federica Censi, Agata Polizzi, Cinzia Mallozzi, Giuseppe Novelli, Federica Sangiuolo, Erica Daina, Giuseppe Remuzzi, Alessandra Ferlini, Marcella Neri, Dario Roccatello, Simone Baldovino, Elisa Menegatti, Savino Sciascia, Maria Rosaria Dariol, Marco Castori, Manuela Priolo.
