Essential role for the TRF2 telomere protein in adult skin homeostasis

TRF2 is a component of shelterin, the protein complex that protects the ends of mammalian chromosomes. TRF2 is essential for telomere capping owing to its roles in suppressing an ATM-dependent DNA damage response (DDR) at chromosome ends and inhibiting end-to-end chromosome fusions. Mice deficient for TRF2 are early embryonic lethal. However, the role of TRF2 in later stages of development and in the adult organism remains largely unaddressed, with the exception of liver, where TRF2 was found to be dispensable for maintaining tissue function. Here, we study the impact of TRF2 conditional deletion in stratified epithelia by generating the TRF2∆/∆-K5-Cre mouse model, which targets TRF2 deletion to the skin from embryonic day E11.5. In marked contrast to TRF2 deletion in the liver, TRF2∆/∆-K5-Cre mice show lethality in utero reaching 100% lethality perinataly. At the molecular and cellular level, TRF2 deletion provokes induction of an acute DDR at telomeres, leading to activation of p53 signaling pathways and to programed cell death since the time of Cre expression at E11.5. Unexpectedly, neither inhibition of the NHEJ pathway by abrogation of 53BP1 nor inhibition of DDR by p53 deficiency rescued these severe phenotypes. Instead, TRF2 deletion provokes an extensive epidermal cell death accompanied by severe inflammation already at E16.5 embryos, which are independent of p53. These results are in contrast with conditional deletion of TRF1 and TPP1 in the skin, where p53 deficiency rescued the associated skin phenotypes, highlighting the comparatively more essential role of TRF2 in skin homeostasis.

Abrogation of TRF1, TRF2, POT1b, TPP1, and TIN2 in mice results in early embryonic lethality (Karlseder et al., 2003;Chiang et al., 2004;Celli & de Lange, 2005;Hockemeyer et al., 2006;Lazzerini Denchi et al., 2006;Wu et al., 2006;Kibe et al., 2010). Conditional deletion of TRF1 and TPP1 in mouse stratified epithelia leads to perinatal lethality coincidental with skin hyperpigmentation and severe skin morphogenesis defects, including absence of mature hair follicles and sebaceous glands, which are concomitant with induction of telomere-originated DNA damage, activation of the p53/p21 and p16 pathways, and cell cycle arrest in vivo (Martinez et al., 2009;Tejera et al., 2010). Importantly, p53 deficiency rescues both the stem cell defects and skin hyperpigmentation, as well as mouse survival, in both mouse models, indicating that the severe skin defects associated with TRF1 and TPP1 abrogation are mediated by p53. Interestingly, long-lived TRF1/p53 double null mice spontaneously develop invasive and genomically unstable squamous cell carcinomas (Martinez et al., 2009). These results suggest that TRF1 normally acts as a tumor suppressor in the context of the organism by preventing telomere-induced genetic instability in proliferating cells. In marked contrast to TRF1 and TPP1, targeted Rap1 deletion in stratified epithelia in Rap1 D/D K5-cre mice does not impact on mouse viability in accordance with a dispensable role for Rap1 in telomere capping .
In contrast to conditional deletion of TPP1 and TRF1 in stratified epithelia, TRF2 conditional deletion in the liver (Mx1TRF2 mice) did not impact on liver regeneration or mouse viability, indicating that TRF2 is dispensable for hepatocyte regeneration (Lazzerini Denchi et al., 2006). In particular, TRF2 deletion in hepatocytes leads to telomere damage and telomere fusions; however, these phenotypes were not accompanied by loss of liver function upon partial hepatectomy. Indeed, liver regeneration in these mice occurred by endoreduplication and cell growth, but not cell division, thereby overcoming the chromosome segregation problems associated with telomere fusions (Lazzerini Denchi et al., 2006).
To further understand the role of TRF2 in the adult organism, as well as to do a comparative analysis with conditional knock-out models for other shelterins, here we deleted TRF2 in stratified epithelia. TRF2 D∕D K5-Cre mice were born at submendelian ratios indicating partial embryonic lethality. Newborn mice die immediately after birth and show complete lack of epidermis. We show that TRF2-deleted epidermis undergoes tissular necrosis accompanied by the presence of dermal infiltrates and increased IL6 levels already at E13.5 of embryonic development that becomes apparent at E16.5 by the complete absence of the epidermal layer of the skin. Comparative studies of the phenotypic effects originated by conditional deletion of either TRF1, TPP1, RAP1, or TRF2 in stratified epithelia, a highly proliferative tissue, revealed that TRF2 deficiency causes the most severe proliferative and developmental defects of all these shelterin components.

Results
TRF2 conditional deletion in stratified epithelia results in partial embryonic lethality and impairs skin development TRF2 conditional deletion in the liver was previously found to be dispensable for liver regeneration and mouse viability (Lazzerini Denchi et al., 2006). Here, we set to address in vivo the impact of TRF2 deficiency in a highly proliferative tissue, such as stratified epithelia. To this end, we crossed conditional TRF2 flox/flox mice or TRF2 +/flox mice with K5-Cre mice that constitutively express the Cre recombinase under the control of the keratin 5 (K5) promoter from day 11.5 of embryonic development onwards (Ramirez et al., 1994), thereby generating TRF2 D∕D K5-Cre mice. TRF2 D∕D K5-Cre mice were born at submendelian ratios indicating partial embryonic lethality associated with TRF2 abrogation in stratified epithelia (Fig. 1A). In addition, newborn TRF2 D∕D K5-Cre mice presented a lower body weight than their wildtype counterparts and died immediately after birth (Fig. 1B). At the macroscopic level, newborn TRF2 Δ/Δ -K5-Cre mice showed a strongly compromised skin phenotype, with a very thin and fragile epidermis at the time of birth (Fig. 1C). In addition, histopathological analysis of newborn mice (P1) revealed severe abnormalities in the skin, palate, tongue, and esophagus. In particular, the skin of TRF2 deficient newborns showed rare small pieces of skin that were mostly necrotic and were devoid of hair follicles and sebaceous glands. Subjacent areas of dermal inflammation showed the presence of macrophages and neutrophils (red arrows in Fig 1D). TRF2 Δ/Δ K5-Cre newborns also showed severely dysplastic epithelia in the tongue and palate, showing areas of atrophy and a reduced number of cellular layers, a phenotype that was milder in the esophagus (Fig. 1D). Of note, no significant phenotype was observed in stratified epithelia of heterozygous TRF2 +/Δ -K5-Cre mice compared with the control TRF2 +/+ mice (data not shown).
Given the very severe skin defects of TRF2 deficient mice at the time of birth, we set to analyze the impact of TRF2 deletion at earlier stages of skin embryonic development. To this end, we studied embryos at embryonic stages E11.5 (time of induction of TRF2 deletion), E13.5, E16.5, and E18.5 ( Fig. 1E and Fig. S1). From E11.5, a progressive reduction in the skin cellularity was observed both in the suprabasal and in the basal skin layers (Fig. S1). At day E13.5, the TRF2 Δ/Δ -K5-Cre epidermis presented a flat epithelia containing just one keratinocyte layer with irregular cellular distribution compared with the two keratinocyte layers present in the wild-type controls (Fig. 1E). At E16.5, wild-type skin presented four keratinocyte layers, while the TRF2 Δ/Δ -K5-Cre epidermis was largely necrotic as observed by almost complete absence of the basal cell layer and showing dysplastic skin foci with hyperkeratosis ( Fig. 1E). Diffuse dermal inflammation is evident (red arrows in Fig. 1E). At this stage, the skin lacked early hair primordial and was composed of only one layer of keratinocytes showing atypically large nuclei (black arrow heads in Fig. 1E). In addition, suprabasal layer development and differentiation were completely blocked in the absence of TRF2. These skin defects progressively worsened until E18.5 at which point almost no keratinocytes could be detected except for a few skin patches with aberrantly large nuclei, suggestive of endoreduplication and massive necrosis of the tissue (black arrow heads in Fig. 1E; see Fig. 4). These rare skin patches also showed complete absence of hair primordia (Fig. 1E). From E18.5 onwards, TRF2-null skin was erosive and composed of a very thin stratum corneum completely lacking hair follicles and sebaceous glands (Fig. S1).
Immunohistochemical analysis of various epidermal differentiation markers, including cytokeratin 10, cytokeratin 14, cytokeratin 6, and loricrin, at different stages of embryonic development revealed an abnormal expression pattern in the absence of TRF2 (Fig. S1), confirming a strongly compromised skin development from day E11.5 when K5-Cre begins to be expressed (E11.5) (Ramirez et al., 1994). In particular, while loricrin and cytokeratins K10 and K14 showed a reduced expression, the cytokeratin 6 showed a higher expression and abnormal distribution compared wild-type skin (Fig. S1).
Severe epidermal stem cell defects in TRF2 Δ/Δ K5-Cre mouse epidermis Impairment of skin development in TRF2 Δ/Δ -K5-Cre mice suggested a defect in the proliferative capacity of epidermal stem cells. We examined the impact of TRF2 abrogation on epidermal stem cell functionality by studying the expression of p63, a marker of proliferating cells of the basal skin layer in stratified epithelia, and Sox9, a stem cell gene that establishes epidermal stem cell compartments in mice during embryonic and early postnatal development (Vidal et al., 2005;Nowak et al., 2008). p63 staining pattern revealed that TRF2-null skin already at E11.5 presents a strong reduction in the number of basal cells becoming completely absent at day E18.5 compared with wild-type skin (Fig. S2).
Sox9 is expressed early in skin morphogenesis and is essential to establish the stem cell niche, so that in its absence, the early hair bulge stem cell population never forms and hair bulge and sebaceous glands do not develop (Vidal et al., 2005;Nowak et al., 2008). Wild-type embryos showed an intense and widespread signal of Sox9 at E13.5, and starting on E16.5, the expression of Sox9 was restricted to the early hair primordia and to the hair bulge in the newborns. In contrast, Sox9 expression in TRF2-deleted epidermis was only detected at E13.5 although to a much lesser extend as compared to wild-type counterparts. Sox9 expression is undetectable from E16.5, in accordance with the complete absence of hair primordia (Fig. S2).
TRF2 deletion in epidermis leads to a persistent DDR activation TRF2 deletion in MEFs is known to result in induction of a DNA damage response (DDR), genome-wide telomere fusions, and onset of p53/Rbdependent cellular senescence (Celli & de Lange, 2005). Uncapped telomeres are recognized as DNA double-strand breaks (DSB) and, as such, are bound by 53BP1 and cH2AX forming the so-called telomere-induced foci (TIF) . To address the presence of DNA damage specifically at telomeres in the absence of TRF2, we made a double immunofluorescence staining with anti-TRF1 and anti-cH2AX in E11.5 embryos. We find that epidermal cells lacking TRF2 show cH2AX foci and that the majority of cH2AX foci colocalize with TRF1 foci in TRF2 Δ/Δ -K5-Cre embryos but not in wild-type embryos indicating that TRF2 deletion in stratified epithelia resulted in a rapid induction of DNA damage specifically at telomeres ( Fig. 2A, left panel: see below quantification of cH2AX-positive cells in the different genotypes). In addition, TRF2-deleted keratinocytes showed multipolar mitosis containing more than two centrosomes indicative of fused chromosomes in the absence of TRF2 ( Fig. 2A, right panel).
To address whether complete impairment of skin development in TRF2 Δ/ Δ K5-Cre mice was also associated with induction of DNA damage, we quantified the abundance of cH2AX positive cells during skin development using immunohistochemistry (Experimental Procedures). Given the almost complete absence of keratinocytes in TRF2-null skin at day E18.5, we focused our analysis in E11.5, E12.5, E13.5, and E16.5 wild-type, TRF2 +/Δ heterozygous and TRF2 Δ/Δ deficient embryos (Fig. 2). At all the analyzed time points, we observed an increased number of cH2AX-positive cells in TRF2 Δ/Δ -K5-Cre epidermis compared with wild-type epidermis. At E11.5, upon initial deletion of the gene, around 90% of cells were cH2AX-positive as compared to approximately 15% observed in heterozygous and in wildtype controls ( Fig 2B). The percentage of cells showing cH2AX foci decreased in TRF2-null skin sections in the following days, remaining stable at 75% at E13.5 and E16.5, while less than 10% were cH2AX-positive in the controls (Fig. 2B). This decrease in the number of cH2AX-positive cells could be due to a rapid induction of chromosome fusions upon TRF2 deletion at E11.5 that consequently lowered the number of exposed uncapped cH2AX labeled telomeres ( Fig. 2A, left panel).
Persistent telomere damage, such as that induced by telomere uncapping, leads to DNA damage response (DDR) characterized by activation of cell cycle inhibitors p53/p21 that leads to cessation of cellular proliferation (de Lange, 2009;. Indeed, already at E11.5 TRF2 Δ/Δ -K5-Cre embryos showed 15-fold and twofold increase in the amount of p53 and p21 positive cells, respectively, compared with heterozygous and wild-type embryos. The number of p53 and p21 positive cells gradually increased up to 60% at E16.5 in agreement with the severe proliferative defects observed in TRF2-deleted skin (Fig. 2C,D). Interestingly, TRF2 depletion was shown to initially induce cellular apoptosis. Thus, the number of active caspase-3 positive cells was progressively augmenting from E11.5 to E.13.5 in TRF2 Δ/Δ -K5-Cre embryos, from approximately 8 to 15% as compared to 3% in the heterozygous and wild-type embryos (Fig. 2E). However, at E16.5, epidermal necrosis is evident as almost no basal cells are present in TRF2 Δ/Δ -K5-Cre embryos (Fig. 1E). Indeed, not even apoptotic cells are observed at this stage (Fig. 2E). These results indicate that TRF2 deletion leads to a progressive loss of skin cellularity initially owing to induction of senescence and apoptosis and eventually resulting in complete skin loss due most likely to epidermal necrosis (see Fig. 4).
In accordance with normal skin development and mouse survival, TRF2 +/D heterozygous mice showed no significant differences in the number of cH2AX, p53, p21, and C3a expressing cells compared with wild-type embryos (Fig. 2B-E).

53BP1 deficiency does not rescue TRF2-associated developmental defects
53BP1 is a C-NHEJ component and an ATM target that accumulates at DSBs and at uncapped telomeres (Rappold et al., 2001;Fernandez-Capetillo et al., 2002;Wang et al., 2002;d'Adda di Fagagna et al., 2003;Takai et al., 2003). It has been shown that chromosome end-to-end fusions occurring upon TRF2 deletion are strongly dependent on the binding of 53BP1 to damaged chromosome ends and that in the absence of 53BP1, these fusions can be rescued (Dimitrova et al., 2008;Rai et al., 2010). To address the contributions of aberrant chromosome rearrangements to the deleterious effects of TRF2 deletion in vivo, we crossed conditional TRF2 +/Δ K5-Cre with TRF2 +/+ 53BP1 À/À mice (Ward et al., 2003), to obtain TRF2 Δ/Δ K5-Cre 53BP1 À/À mice. Similar to TRF2 D∕D K5-Cre single knock-out mice (Fig. 1A,B), doubly deficient TRF2 D∕D -K5-Cre 53BP1 À/À mice present submendelian birth ratios, neonates had significantly lower body weight as compared to control mice and died immediately after birth ( Fig. 3A-C). Neither differences in body weight nor macroscopic differences were revealed between TRF2 D∕D K5-Cre and TRF2 D∕D K5-Cre 53BP1 À/À neonates (Fig. 3B,C). These results indicate that abolishment of the C-NHEJ and consequently abolishment of chromosome fusions neither rescued nor aggravated the pathological effects of TRF2 deletion in stratified epithelia (see Fig. 6D-E), suggesting that the presence of fusions is not responsible for the mouse phenotype.
p53 deficiency does not rescue TRF2-associated developmental defects TRF2 deficiency causes severe proliferative defects in vitro (Celli & de Lange, 2005). TRF2 D∕D -K5-Cre epidermis showed a strong induction of the cell cycle inhibitors p53/p21 suggesting that p53 is an important mediator of apoptosis and cell cycle arrest produced by TRF2 deficiency (Fig. 2). To address whether p53 deficiency could rescue TRF2-mediated proliferative and developmental defects in vivo, we generated mice simultaneously deficient for TRF2 and p53, TRF2 D∕D -K5-Cre p53 À/À mice. Similar to TRF2 D∕D -K5-Cre mice, TRF2 D∕D -K5-Cre p53 À/À double mutant were born at submendelian ratios and newborn mice died immediately after birth, indicating occurrence of embryonic lethality (Fig. 3C,D). No differences in neonate body weight were observed between TRF2 D∕D -K5-Cre and TRF2 D∕D -K5-Cre p53 À/À mice (Fig. 3E), suggesting that p53 deficiency does not rescue the mouse phenotype associated with TRF2 deficiency.
Histopathological analysis of TRF2 D∕D -K5-Cre, TRF2 D∕D -K5-Cre 53BP1 À/ À and TRF2 D∕D -K5-Cre p53 À/À neonates revealed that the overall clinical picture was equally severe in the single knockout as in the compound mice under study (Fig 4). The wild-type neonatal skin clearly shows the basal layer, stratum granulosum, stratum corneum, and differentiated hair follicles. In contrast, the neonatal skin of TRF2 D∕D -K5-Cre, TRF2 D∕D -K5-Cre 53BP1 À/À , and TRF2 D∕D -K5-Cre p53 À/À mice were extensively undeveloped, consisting of rare small regions that were largely necrotic ( Fig 4A). These were highly dysplastic, with areas of hyperkeratosis and ulcerated parts. TRF2 D∕D -K5-Cre 53BP1 À/À and TRF2 D∕D -K5-Cre p53 À/À tongue, esophagus, and palate showed severe dysplasia similar to that formed in TRF2-null epithelia (data not shown). To address whether 53BP1-or p53-deficiency could be partially rescuing proliferative and stem cell defects observed during embryonic development of the TRF2 D∕D -K5-Cre single knockout mouse, we analyzed p63 expression and quantified the number of mitotic cells by phospho-histone 3 staining in E13.5 embryos (Fig. S3A,B). The p63 staining pattern in TRF2 D∕D -K5-Cre, TRF2 D∕D -K5-Cre 53BP1 À/À , and TRF2 D∕D -K5-Cre p53 À/À embryos was similar and revealed a strong reduction in the number of basal cells as compared to wild-type embryos (Fig. S3A). No differences in the number of mitotic cells between the TRF2 D∕D -K5-Cre and TRF2 D∕D -K5-Cre 53BP1 À/ À embryos were detected. However, at this stage, E13.5, p53 deficiency TRF2 is essential for skin development, P. Mart ınez et al. 659 resulted in a bypass of TRF2-mediated mitotic arrest that was clearly not sufficient to prevent TRF2-mediated degenerative epithelial pathologies or survival (Fig. S3B). In fact, almost no mitotic cells in TRF2 D∕D -K5-Cre, TRF2 D∕D -K5-Cre 53BP1 À/À , and TRF2 D∕D -K5-Cre p53 À/À neonates were detected as compared to wild-type neonates, which show 2 mitotic cells per visual field. Moreover, the scarce mitotic cells observed presented highly aberrant morphology (Black arrow heads in Fig 4A). Not a single normal mitotic figure was observed in TRF2-deficient epidermis regardless of 53BP1 and p53 status.
Histopathological examination of TRF2 D∕D -K5-Cre, TRF2 D∕D -K5-Cre 53BP1 À/À , and TRF2 D∕D -K5-Cre p53 À/À skin revealed aberrant mitotic figures and giant multinucleated cells, which are hallmarks of mitotic catastrophe. Specifically, TRF2 D∕D -K5-Cre, TRF2 D∕D -K5-Cre 53BP1 À/À , and TRF2 D∕D -K5-Cre p53 À/À neonates showed a significantly increased number of giant and multinucleated cells per visual field as compared to wild-type (Fig. 4B). Mitotic catastrophe has been delineated as a type of cell death that results from abnormal mitosis and is associated with premature chromosome condensation and the formation of large cells with multiple micronuclei (Vitale et al., 2011). Cellular composition of TRF2 D∕D -K5-Cre p53 À/À skin showed very few enormous cells, which presented giant nuclei and dense chromatin, a strong indication of the occurrence of massive endoreduplication (blue arrow heads in Fig. 4A). Of note, TRF2 D∕D -K5-Cre and TRF2 D∕D -K5-Cre p53 À/À present twofold increase in the number of giant and multinucleated cells than TRF2 D∕D -K5-Cre 53BP1 À/À indicating that 53BP1 deficiency partially rescues endoreduplication most likely due to less amount of fused chromosomes in the latter (Fig. 4B, see Fig. 6D,E). Morphologies characteristic of necrotic cells were observed at similar frequencies in TRF2 D∕D -K5-Cre, TRF2 D∕D -K5-Cre 53BP1 À/À , and TRF2 D∕D -K5-Cre p53 À/À neonates, indicating that necrotic cell death is observed independently of either 53BP1 or p53 status (Fig. 4B). Necrotic cells present pale eosinophilic cytoplasm, swollen nucleus, euchromatic chromatin that condensates into irregular patches and increased cell volume that culminate in the breakdown of the plasma membrane (Fig 4). Of note, at this stage, cells with morphology characteristic of apoptotic cells were not observed (see Fig. 2E). p53 mediates p21 induction and apoptosis upon TRF2 deletion during embryonic development

(A) (B)
To address whether either 53BP1 or p53 deficiency could have an impact on the DNA damage level, on the induction of p53/p21 as well as in apoptotic levels during fetus development, we performed a comparative study among TRF2 D∕D -K5-Cre 53BP1 +/+ p53 +/+ , TRF2 D∕D -K5-Cre 53BP1 À/À p53 +/+ , TRF2 D∕D -K5-Cre 53BP1 +/+ p53 À/À , and their correspondent wildtype and single knock-out controls in embryos at E.13.5, two days after Cre expression. Analysis of cH2AX positive cells in stratified epithelia did not reveal significant differences between TRF2 D∕D -K5-Cre p53 +/+ and TRF2 D∕D -K5-Cre p53 À/À embryos, being approximately 75% of the cells cH2AX positive (Fig. 5A,B). However, 100% of the cells in TRF2 D∕D -K5-Cre 53BP1 À/À E.13.5 embryos were positive for cH2AX. This increase could be explained by the abolishment of chromosome end-to-end fusions that temporarily leads to higher number of exposed damaged telomeres as compared to cells proficient in the C-NHEJ (see Fig. 6). The number of p53 positive cells in TRF2 D∕D -K5-Cre 53BP1 À/À embryos was also somewhat higher as compared to TRF2 D∕D -K5-Cre 53BP1 +/+ E.13.5 embryos although this difference did not reach significant values (Fig. 5).
As expected, no p53 signal was detected in p53 À/À samples. The number of p21 positive cells was similarly increased in TRF2 D∕D -K5-Cre 53BP1 +/+ p53 +/+ and TRF2 D∕D -K5-Cre 53BP1 À/À p53 +/+ embryos, suggesting that is independent of the 53BP1 expression. However, the percentage of p21positive cells was significantly decreased in TRF2 D∕D -K5-Cre 53BP1 +/+ p53 À/À epithelia as compared to TRF2 D∕D -K5-Cre 53BP1 +/+ p53 +/+ epithelia, in agreement with the fact that p21 activation is mediated by p53 (Fig. 5). Apoptotic levels were similar between TRF2 D∕D -K5-Cre 53BP1 +/+ p53 +/ + and TRF2 D∕D -K5-Cre 53BP1 À/À p53 +/+ embryos, reaching values over 10% of the total epidermal cells (Fig. 5), suggesting that the apoptotic response is not triggered by the presence of chromosome fusions. In contrast, the number of caspase-3 (Ca3) positive cells was not further increased in TRF2 D∕D -K5-Cre 53BP1 +/+ p53 À/À compared with TRF2 +/+ control embryos. These results indicate that the increase in the Ca3 positive cells observed upon TRF2 removal is p53-mediated. Similar values than with Ca3 staining were obtained by TUNEL assay indicating that the observed apoptosis was Ca3-mediated ( Figure S3C). The observation that abolition of p53/p21 response as well as apoptosis by p53 deficiency did not rescue developmental defects and survival of the TRF2 D∕D -K5-Cre p53 À/À mice indicates that these two processes, cell cycle arrest and programmed cell death induced by TRF2 deletion and mediated by p53, are not the major cellular responses responsible for the severe skin atrophy and morbidity of TRF2-deleted mice. Rather, it appears that extensive necrosis (see Fig. 4) and inflammation are the causes of the severity of the phenotype. Indeed, the presence of macrophage infiltrates became apparent by immunohistochemistry analysis of F4/80 positive cells, a marker of mature macrophages, in epidermal layer already at E13.5 embryos lacking TRF2 (Fig. 6A). Moreover, TRF2-deficient E13.5 embryos independently of 53BP1 and p53 status showed increased interleukin 6 (IL6) levels as compared to wild-type embryos, indicative of an inflammatory process triggered by tissue damage (Fig. 6B).

TRF2 deficiency does not impact on telomere length homeostasis
To address whether TRF2 deletion in stratified epithelia could lead to defects in telomere length homeostasis, as well as to confirm that 53BP1 deficiency also abolishes end-to-end fusions in vivo, we performed quantitative telomere FISH (QFISH) analysis on wild-type, TRF2 D∕D K5-Cre, TRF2 D∕D K5-Cre 53BP1 À/À , and TRF2 D∕D K5-Cre p53 À/À E13.5 embryo sections. We quantified the mean intensity fluorescence per area of the spot and the area of each individual spot in both the epidermal layer as well as in the subadjacent nonepidermal tissue of the embryos (Fig. 6C-E and Fig. S4). We did not find significant differences in the median telomere fluorescence per pixel neither when comparing the epidermis with the dermis within a given genotype nor among the different genotypes analyzed, indicating that TRF2 deletion specifically in the epidermis does not impact on telomere length (Fig. 6C and Fig. S4). Interestingly, when we calculated the median telomere area, we found significantly higher telomere area in TRF2 D∕D K5-Cre and TRF2 D∕D K5-Cre p53 À/À epidermis compared with wild-type and with TRF2 D∕D K5-Cre 53BP1 À/À epidermis (Fig. 6D,E and Fig. S4). These differences, however, were not observed in the dermis where TRF2 is not deleted, indicating that are the consequence of TRF2 deletion. These results are compatible with a higher portion of telomeric association/fusions and therefore higher telomeric area in TRF2-deficient epidermis, which are rescued in the absence of 53BP1 but not in the absence of p53 (Fig. 6D,E and Fig.  S4). These findings are in agreement with the fact that 53BP1 but not p53 is involved in the C-NHEJ pathway.

Discussion
The telomere-binding protein TRF2 has been extensively studied in vitro. In particular, In vitro studies have shown that TRF2 has a crucial role in end protection through the suppression of ATM activation and inhibiting the NHEJ pathway and thereby preventing end-to-end fusions (van Steensel et al., 1998;Karlseder et al., 2004;Celli & de Lange, 2005;Denchi & de Lange, 2007;Okamoto et al., 2013). In contrast, there are very few reports on the function of TRF2 in the organism. In particular, while whole body deletion of TRF2 causes fully penetrant embryonic lethality (Celli & de Lange, 2005), TRF2 deletion in the liver is dispensable for liver regeneration and mouse survival (Lazzerini Denchi et al., 2006). In this work, we have studied the impact of TRF2 conditional deletion in stratified epithelia during embryonic development. Previous work from our group has shown the impact of deleting other members of the shelterin complex in the skin (i.e., TRF1, TPP1, and RAP1) (Martinez et al., 2009Tejera et al., 2010). Therefore, it would be of interest to determine the relative importance of the different shelterin components in tissue homeostasis. We show that embryonic development of TRF2 D∕D -K5-Cre epidermis is fully blocked from E11.5 at the time when K5-Cre expression commences. TRF2 deletion provokes a dramatic DNA damage burden and an immediate DNA damage response as indicated by the increase in cH2AX, p53, and p21 positive cells in the epidermal layer of the skin that is coincidental with a progressive induction of apoptosis until E16.5. Indeed, epidermis of E16.5 embryos displayed only one keratinocytes layer and extensive necrosis as well as complete absence of early hair primordia. Skin of E18.5 embryos is composed of rare small areas of epidermis, which were completely necrotic and keratinocytes presented marked cytological atypia with disturbed polarity of the cells and extremely large nuclei. The clinical picture worsens with the progress of embryonic development, becoming incompatible with life in newborns. This is in agreement with the observed partial embryonic lethality and immediate death after birth of the TRF2 D∕D -K5-Cre newborns. In comparison with other available conditional knock-out mice in stratified epithelia lacking shelterin components, that is, TRF1 D∕D -K5-Cre, TPP1 D∕D -K5-Cre, and RAP1 D∕D -K5-Cre, the TRF2 D∕D -K5-Cre mouse is the only one that shows partial embryonic lethality (Martinez et al., 2009Tejera et al., 2010). Indeed, although both TRF1 D∕D -K5-Cre and TPP1 D∕D -K5-Cre mice show perinatal lethality, some neonates reached day 6 and day 9 postpartum, respectively, while no impact on survival was observed in the RAP1 D∕D -K5-Cre as compared to their wild-type counterparts (Martinez et al., 2009Tejera et al., 2010). In addition, morphological defects of the TRF2 D∕D -K5-Cre epidermis were more severe as compared to those showed by the TRF1 D∕D -K5-Cre and TPP1 D∕D -K5-Cre neonates (Martinez et al., 2009;Tejera et al., 2010). In contrast to the rescue of degenerative pathologies observed by p53 deficiency combined with TRF1 and TPP1 deletion, p53 abrogation did not bypass developmental defects or perinatal lethality associated with TRF2 loss (Martinez et al., 2009;Tejera et al., 2010). Interestingly, although the percentage of both p21 and apoptotic cells in the TRF2 D∕D K5-Cre p53 À/À E13.5 embryos remained similar to that of TRF2 +/+ wild-type embryos, this was no sufficient to recover either epidermal development or embryonic viability. Based on histological findings of TRF2-deficient epithelia revealing the presence of aberrant mitosis, giant multinucleated cells and morphologies characteristic of necrotic cells as well as on the observation that abolition of senescence and apoptosis by combined p53-deficiency did not rescue TRF2-associated phenotypes, we propose that mitotic catastrophe and necrosis are the cause of complete disappearance of the epidermal basal layer already from E16.5 onwards. Mitotic catastrophe has been extensively used as a term to indicate cell death resulting from aberrant mitosis (Kroemer et al., 2009). Some studies show that mitotic catastrophe can be followed by apoptosis and is therefore a matter of debate whether mitotic catastrophe is a specific death process or just functions as a trigger for other cell death mechanisms (Castedo et al., 2004). The actual view defines mitotic catastrophe as a mechanism that senses mitotic failure and responds to it by driving the cell to an irreversible fate, be it apoptosis, necrosis or senescence-mediated elimination of mitosis deficient, genomically unstable cells. (Vitale et al., 2011). Unfortunately, given the impossibility to cell culture TRF2-deficient keratinocytes, we cannot provide a molecular probe to ultimately demonstrate necrosis as the mechanism by which cell death is taking place upon TRF2 loss in skin (Li et al., 2012). However, the fact that TRF2-driven apoptosis and senescence were abolished in TRF2 D∕D K5-Cre p53 À/À embryos without impacting on survival supports the statement that disappearance of epidermal layer in TRF2 deficient mice is mainly necrosis-mediated.
The current view is that necrotic cell death is not the result of one welldefined signaling cascade but rather the consequence of crosstalk between several signaling pathways (Festjens et al., 2006;Kung et al., 2013). Apoptosis occurs without and sometimes even with the active sequestration of danger-associated molecular pattern. Apoptotic corpses can suppress the transcription of proinflammatory cytokine genes, promote the secretion of anti-inflammatory cytokines by phagocytes and cause antigen-presenting cells to present dead-cell-antigen leading to immunological tolerance. In contrast, necrotically dying cells initiate proinflamatory signaling cascade by actively releasing inflammatory cytokines and by spilling their contents to the environment when they lyse to attract inflammatory effectors and in particular macrophages (Zitvogel et al., 2010). Indeed, the presence of macrophage infiltrates and increased levels of IL6 became apparent in epidermal layer already at E13.5 embryos lacking TRF2 indicating commence of tissue inflammation. TRF2 deletion also releases RAP1 from telomeres. RAP1 has been shown to act as a transcriptional regulator . It is therefore tempting to speculate that release of RAP1 from telomeres upon TRF2 deletion might be upregulating pronecrosis genes. Confirmation of this hypothesis warrants further studies. Interestingly, it has been shown that extensive DNA damage induces necrotic cell death in a Poly(ADP-ribose) polymerase (PARP) dependent manner. PARP is a nuclear enzyme activated by DNA strand breaks that participates in DNA repair. Overactivation of PARP after extensive DNA damage leads to necrotic cell death by depletion of bnicotinamide adenine dinucleotide and ATP. In fact, PARP-deficient mouse fibroblasts are protected from necrotic cell death and ATP depletion but not from apoptotic death (Ha & Snyder, 1999). It would be interesting to test whether PARP-deficiency may rescue epidermal developmental defect associated with TRF2-loss. [Correction added on 30 June 2014, after first online publication: the abbreviation of Poly(ADP-ribose) polymerase was corrected from PPAR to PARP.] Abolishment of end-to-end chromosomal fusions by 53BP1 deficiency neither rescued nor aggravated TRF2-mediated developmental and molecular skin defects. This is in marked contrast to TRF1 D∕D -K5-Cre 53BP1 À/À mice that showed a clear aggravation of the skin phenotypes as compared to TRF1 D∕D -K5-Cre single mutant, which was associated with an earlier onset of DNA damage and DDR activation during embryonic development (Martinez et al., 2012). In fact, 2 days after Cre expression in E13.5 TRF2 D∕D -K5-Cre 53BP1 À/À embryos, 100% of the cells were positive for cH2AX indicating that chromosomal fusions are not the cause of embryonic death but rather the high DNA damage burden occurring upon TRF2 depletion. The differences obtained between the TRF1 D∕D -K5-Cre 53BP1 À/À and the TRF2 D∕D -K5-Cre 53BP1 À/À mice could be due to the more immediate and stronger DNA damage inflicted upon TRF2 depletion as compared to TRF1. Our results show that in vivo TRF2 abrogation in a highly proliferative tissue such as skin leads to the most severe deleterious effect as compared to TRF1, TPP1, and RAP1 loss (Martinez et al., 2009Tejera et al., 2010).
Despite the deleterious molecular consequences of TRF2 loss, conditional deletion of TRF2 in the liver did not impact on liver regeneration or mouse viability (Lazzerini Denchi et al., 2006). The very distinct outcomes obtained from deleting TRF2 in the liver or in the epidermis must be due to the different nature of the tissue. In noncycling hepatocytes, TRF2-mediated telomere dysfunction did not induce p53 or cell death and liver function was unaltered (Lazzerini Denchi et al., 2006). In contrast, in highly proliferative tissues such as skin, TRF2 deletion induces a rapid DDR that leads to inhibition of cell proliferation, cell death, and ultimately to tissular necrosis.
Skin samples and other mouse tissues were fixed in 10% buffered formalin, dehydrated, and embedded in paraffin. For histopathological analysis, 4-lm sections were deparaffinated and stained with hematoxylin and eosin according to standard procedures. A minimum of five mice or embryos per genotype were histopathologically analyzed in back skin sections. A blind subjective image analysis to estimate the keratinocyte layers was performed blindly. Multinucleated, giant cells and aberrant mitosis figures were counted on a minimum of ten randomly selected 409 high-power fields per animal. Immunohistochemistry was performed on deparaffinated skin sections processed with 10 mM sodium citrate (pH 6.5) cooked under pressure for 2 min. Slides were washed in water, then in Buffer TBS Tween20 0.5%, blocked with peroxidase, washed with TBS Tween20 0.5% again, and blocked with fetal bovine serum followed by another wash. The slides were incubated with the primary antibodies: rabbit polyclonal to p53 (CM5, Novocastra), goat polyclonal to p21 (C-19-G, Santa Cruz), mouse monoclonal to phosphohistone H2AX (ser139), rabbit polyclonal to SOX9 (Millipore), mouse monoclonal to p63 (p63-4A4, Neo Markers), rabbit polyclonal to cytokeratine 14 (AF64, Covance), rabbit polyclonal to cytokeratine 10 (Covance), rabbit polyclonal to caspase-3 active (R&D systems), rabbit polyclonal to cytokeratine 6 (Covance), rabbit polyclonal to loricrine (Covance), rabbit polyclonal to phospho-histone 3 H3 (ser10) (Millipore), and rabbit polyclonal to interleukin 6 (ABCAM). Slides were then incubated with secondary antibodies conjugated with peroxidase from DAKO. For signal development, DAB (DAKO) was used as a substrate. Sections were lightly counterstained with hematoxylin and analyzed by light microscopy. Telomere-induced foci (TIFs) were detected using a rabbit polyclonal to TRF1 (homemade) and a mouse monoclonal antibody raised against phospho-histone2 H2AX-Ser 139 (Millipore). Centrosome and microtubules were detected using a rabbit polyclonal to pericentrina (Covance) and a mouse monoclonal to a-tubulin (abcam). DeadEnd Fluorometric TUNEL System was used to assay apoptotic cells according to the manufacturer's instructions (Promega).

Quantitative immuno fish
Quantitative telomere fluorescence in situ hybridization (Q-FISH) directly on embryo sections was performed as previously described (Zijlmans et al., 1997;Martinez et al., 2013). Confocal microscopy was performed at room temperature with a laser-scanning microscope (TSC SP5) using a Plan Apo 63 A-1.40 NA oil immersion objective (HCX). Maximal projection of z-stack images generated using advanced fluorescence software (LAS) was analyzed with Definiens XD software package. The DAPI images were used to detect telomeric signals inside each nuclei.

Statistical analysis
A Student's t-test was used to calculate the statistical significance of the observed differences in telomere fluorescence, cH2AX foci, 53BP1 foci, body weight, cH2AX expression, p53 expression, p21 expression, and caspase-3 expression. A chi-square test was used to calculate statistical significance in Mendelian ratios.

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site.    TRF2 is essential for skin development, P. Mart ınez et al.