To examine the role of cellular senescence in ageing and age-related pathologies, we designed a transgenic strategy for the clearance of senescent cells in mice. We based our approach on an earlier mouse model, termed FAT-ATTAC (fat apoptosis through targeted activation of caspase), in which adipocytes were selectively killed by apoptosis upon the administration of AP20187, a synthetic drug that induces dimerization of a membrane-bound myristoylated FK506-binding-protein–caspase 8 (FKBP–Casp8) fusion protein expressed specifically in adipocytes via the minimal Fabp4 promoter6. Although a universal marker that is solely expressed in senescent cells has not been identified, most senescent cells seem to express p16Ink4a, a cyclin-dependent kinase inhibitor and tumour suppressor that enforces growth arrest by activating Rb5,7. Additionally, the expression of p16Ink4a is known to increase with ageing in several rodent and human tissues8. We replaced the Fabp4 promoter with a 2,617-bp fragment of the p16Ink4a gene promoter that is transcriptionally active in senescent, but not non-senescent cells (Fig. 1a)9. We added an internal ribosome entry site (IRES) followed by an open reading frame (ORF) coding for enhanced green fluorescence protein (EGFP) to allow for detection and collection of p16Ink4a-positive senescent cells. Injection of the resulting construct into fertilized eggs yielded nine transgenic INK-ATTAC founder lines.

Figure 1: Generation and characterization of INK-ATTAC transgenic mice. a, Schematic of the INK-ATTAC construct and the mechanism of apoptosis activation. b, GFP intensity of IAT. c, qRT–PCR analysis of the indicated tissues of 10-month-old mice. ATTAC, INK-ATTAC; H/H, BubR1H/H; SkM, skeletal muscle (gastrocnemius). d, Bone marrow cells harvested from 2-month-old mice immunostained for Flag after culture in the absence or presence of rosiglitazone for 48 h. e, SA-β-Gal stained IAT collected from 9-month-old mice of the indicated genotypes. f, Expression of senescence markers in tissues of 10-month-old mice measured by qRT–PCR. All increases are statistically significant (P < 0.05). g, FACS profile of single-cell suspensions from IAT of 10-month-old mice. Brackets indicate sorting gates. h, GFP+ and GFP– cell populations from IAT analysed for relative expression of senescence markers by qRT–PCR. All increases are statistically significant (P < 0.01). i, Bright field images of MEFs sorted into GFP+ and GFP– populations after induction of senescence and then stained for SA-β-Gal. For all experiments, n = 3 untreated females per genotype. Error bars, s.d. Scale bars in b, d and i, 20 µm. *P < 0.05, **P < 0.01, ***P < 0.001. Full size image Download PowerPoint slide

To examine whether removal of p16Ink4a-expressing cells is technically feasible and whether this affects age-associated deficits in mice, we bred each of the founder lines onto a BubR1 hypomorphic (BubR1H/H) genetic background. BubR1 encodes a key member of the mitotic checkpoint, a surveillance mechanism that ensures accurate chromosome segregation in mitosis by inhibiting the ubiquitin ligase activity of Cdc20-activated anaphase-promoting complex (APCCdc20) in the presence of unattached chromosomes10,11. BubR1H/H mice have a markedly shortened lifespan and exhibit a variety of age-related phenotypes, including infertility, lordokyphosis, sarcopenia, cataracts, fat loss, cardiac arrhythmias, arterial wall stiffening, impaired wound healing and dermal thinning12,13,14. It has been proposed that BubR1 is a determinant of natural ageing, because levels of BubR1 decline markedly with age12,13,14. BubR1H/H mice selectively accumulate p16Ink4a-positive cells in certain tissues in which age-associated pathologies develop, including adipose tissue, skeletal muscle and eye15. Inactivation of p16Ink4a in these mice is known to delay the onset of age-related phenotypes selectively in these tissues15. To screen for INK-ATTAC transgene activity in p16Ink4a-positive cells, we collected samples of inguinal adipose tissue (IAT) from each of the nine BubR1H/H;INK-ATTAC strains at 5 months of age and analysed them for GFP expression by fluorescence microscopy. We observed GFP fluorescence in two of these strains, BubR1H/H;INK-ATTAC-3 and -5 (Fig. 1b and Supplementary Fig. 1a). Quantitative reverse transcription–polymerase chain reaction (qRT–PCR) analysis of various tissues from BubR1H/H;INK-ATTAC-3 and -5 mice demonstrated that INK-ATTAC and GFP transcript levels were significantly elevated in adipose tissue, skeletal muscle and eye, but not in tissues in which endogenous p16Ink4a is not induced, including liver and heart (Fig. 1c and Supplementary Fig. 1b).

To confirm that transgenic INK-ATTAC and endogenous p16Ink4a are under the same transcriptional control mechanism outside the context of BubR1 hypomorphism, we harvested bone marrow cells from 2-month-old wild-type (WT);INK-ATTAC-3 and -5 mice and cultured them in the absence or presence of rosiglitazone, a drug that can induce cellular senescence and p16Ink4a expression through activation of PPARγ16. Immunofluorescence microscopy revealed that a high proportion of cells expressed Flag-tagged FKBP–Casp8 in the presence of rosiglitazone, but not in its absence (Fig. 1d). Furthermore, we observed selective INK-ATTAC transgene induction in tissues of WT;INK-ATTAC-3 mice showing elevated expression of endogenous p16Ink4a upon chronological ageing (Supplementary Fig. 2). Together, these data indicate that INK-ATTAC gene activity in founder lines 3 and 5 overlaps with endogenous p16Ink4a expression.

Next, we tested whether INK-ATTAC is expressed in senescent cells in BubR1 hypomorphic tissue. Fat tissue of aged BubR1H/H;INK-ATTAC mice stained strongly for senescence-associated-β-galactosidase (SA-β-Gal; Fig. 1e). qRT–PCR analysis demonstrated that INK-ATTAC expression correlates with expression of senescence markers in IAT (Fig. 1f and Supplementary Fig. 3a). Skeletal muscle and lens tissue of aged BubR1H/H;INK-ATTAC mice are SA-β-Gal negative (data not shown), but both these tissues expressed other markers of senescence (Fig. 1f and Supplementary Fig. 3a). Senescence markers were not elevated in 3-week-old BubR1H/H;INK-ATTAC mice (Supplementary Fig. 3b, c). To obtain additional evidence for selective expression of INK-ATTAC in senescent cells, we collected IAT from aged BubR1H/H;INK-ATTAC animals, prepared single-cell suspensions by collagenase treatment, separated GFP+ and GFP– cell populations by fluorescence activated cell sorting (FACS; Fig. 1g), and analysed each population for expression of INK-ATTAC and senescence markers by qRT–PCR. GFP+ cells not only expressed much higher levels of p16Ink4a than GFP– cells but also had elevated levels of other key senescence markers (Fig. 1h and Supplementary Fig. 3d). Furthermore, two conditions that induce p16Ink4a expression and senescence in primary mouse embryonic fibroblasts (MEFs), ectopic expression of oncogenic Ras and serial passaging12,17,18, produced a subpopulation of GFP+ WT;INK-ATTAC-3 MEFs that, in contrast to the remaining GFP– cells, stained positively for SA-β-Gal (Fig. 1i). Taken together, these results indicate that INK-ATTAC is selectively expressed in p16Ink4a-positive senescent cells.

To determine whether INK-ATTAC can eliminate senescent cells, we cultured bone marrow cells of WT;INK-ATTAC transgenic lines 3 and 5 in the presence of rosiglitazone to induce senescence and then monitored cell survival after activating the FKBP–Casp8 fusion protein by AP20187 treatment. We found that the vast majority of cells from both transgenic lines were either dead or in the process of dying 48 h after adding AP20187 (Fig. 2a). In contrast, parallel cultures that remained untreated consisted almost entirely of viable SA-β-Gal-positive cells. These data show that FKBP–Casp8 activation efficiently eliminates p16Ink4a-positive senescent cells in vitro.

Figure 2: BubR1H/H;INK-ATTAC mice treated with AP20187 from weaning age on show delayed onset of p16Ink4a-mediated age-related phenotypes. a, Bone marrow cells cultured in rosiglitazone for 5 days and then treated or not treated with AP20187 (AP) for 2 days before SA-β-Gal staining. Scale bar, 50 µm. b, Incidence of lordokyphosis and cataracts. c, Representative images of 9-month-old mice. d, Mean skeletal muscle fibre diameters of 10-month-old mice. ABD, abdominal muscle; Gastro, gastrocnemius muscle. e, Exercise ability of 10-month-old AP20187-treated mice relative to age-matched untreated mice. Time is running time to exhaustion; distance is distance travelled at time of exhaustion; work is the energy expended to exhaustion. f, Body and fat depot weights of 10-month-old mice. Parentheses, s.d. Mes, mesenteric; Peri, perirenal; POV, paraovarian; SSAT, subscapular adipose tissue. g, Average fat cell diameters in IAT of 10-month-old mice. h, Dermis and subdermal adipose layer thickness of 10-month-old mice. Colour codes in e, g and h are as indicated in d. Error bars, s.e.m. For all analysis n = 6 female mice per genotype (per treatment). *P < 0.05, **P < 0.01, ***P < 0.001. Full size image Download PowerPoint slide

Next, we examined whether clearance of p16Ink4a-expressing cells from BubR1H/H mice prevents or delays the onset of age-related phenotypes in this progeroid background. To this end, we established cohorts of BubR1H/H;INK-ATTAC-3 and -5 mice, which were either treated with AP20187 every third day beginning at 3 weeks of age or left untreated. Both treated and untreated mice were monitored for the development of age-associated deficits known to accompany p16Ink4a induction, including sarcopenia, cataracts and loss of adipose tissue15. Remarkably, treated mice of both BubR1H/H;INK-ATTAC lines had substantially delayed onset of lordokyphosis (a measure of sarcopenia onset in this model15) and cataracts compared to untreated mice (Fig. 2b, c). Consistent with decreased lordokyphosis, muscle fibre diameters of AP20187-treated BubR1H/H;INK-ATTAC animals were larger than those of untreated counterparts (Fig. 2d). In addition to muscle retention, treadmill exercise tests revealed that duration of exercise, distance travelled and overall amount of work performed were all significantly increased in the animals treated with AP20187 (Fig. 2e), indicating preservation of muscle function. Dual-energy X-ray absorptiometry (DEXA) scans of BubR1H/H;INK-ATTAC mice confirmed that AP20187 treatment prevented loss of adipose tissue (Fig. 2f). All major fat deposits were larger in AP20187-treated BubR1H/H;INK-ATTAC animals (Fig. 2f) and individual adipocytes were markedly increased in size (Fig. 2g). Consistent with this generally increased adiposity, lateral skin contained significantly more subdermal adipose tissue (Fig. 2h). The above age-related phenotypes were not delayed upon AP20187 treatment of BubR1H/H mice lacking INK-ATTAC (Fig. 2b and Supplementary Fig. 4).

Age-related phenotypes of BubR1H/H mice that arise in a p16Ink4a-independent fashion, such as cardiac arrhythmias and arterial wall stiffening14, were not attenuated in AP20187-treated BubR1H/H;INK-ATTAC mice (Supplementary Fig. 5a, b). This correlated with lack of INK-ATTAC induction in heart and aorta (Fig. 1c and Supplementary Fig. 5c). Cardiac failure is presumably the main cause of death in BubR1H/H mice (data not shown), which could explain why the overall survival of AP20187-treated BubR1H/H;INK-ATTAC mice was not substantially extended (Supplementary Fig. 5d). To examine whether clearance of p16Ink4a-positive cells might have any overtly negative side effects, WT;INK-ATTAC mice were continuously treated with AP20187 until 8 months of age; however, no such effects were observed (data not shown). Taken together, these results indicate that continuous removal of p16Ink4a-expressing cells from BubR1H/H;INK-ATTAC mice selectively delays age-related phenotypes that depend on p16Ink4a induction.

Next, we determined whether the delayed onset of age-related pathologies coincided with a reduction in the number of senescent cells in these tissues. The IAT of AP20187-treated BubR1H/H;INK-ATTAC mice showed a marked decrease in SA-β-Gal staining compared with the IAT of untreated counterparts (Fig. 3a). Corresponding decreases in other senescence-associated markers were also observed, as well as expected reductions in INK-ATTAC and GFP (Fig. 3b and Supplementary Fig. 6a). Skeletal muscle and eye had a similar reduction in senescence indicators (Fig. 3c, d and Supplementary Fig. 6b, c). BrdU incorporation was lower in IAT and muscle tissue of untreated than treated animals (Fig. 3e), supporting the contention that senescence-associated replicative arrest is decreased upon administration of AP20187 in BubR1H/H;INK-ATTAC transgenic animals. Together, these data indicate that senescent cells were cleared from tissues and that this delays acquisition of age-related dysfunction in BubR1 hypomorphic mice.

Figure 3: AP20187-treated BubR1H/H;INK-ATTAC mice have reduced numbers of p16Ink4a-positive senescent cells. a, Images of SA-β-Gal stained IAT of 10-month-old mice. b–d, Expression of senescence markers in IAT (b), gastrocnemius (c) and eye (d) of 10-month-old AP20187-treated and untreated BubR1H/H;INK-ATTAC-3 mice relative to age-matched untreated WT;INK-ATTAC-3 mice. Error bars indicate s.d.; n = 3 females per genotype per treatment. The expression of all genes is significantly decreased upon AP20187 treatment (P < 0.05) with the exception of GFP in the eye. e, BrdU incorporation rates in IAT and skeletal muscle. Error bars, s.e.m.; n = 6 females per genotype per treatment. *P < 0.05. Full size image Download PowerPoint slide

To investigate the effect of senescent cell clearance later in life when age-related phenotypes are apparent in BubR1H/H mice, we started AP20187 treatment of BubR1H/H;INK-ATTAC mice at 5 months instead of weaning age and measured p16Ink4a-dependent age-related phenotypes at 10 months. Cataracts had already fully maturated by the onset of AP20187 treatment and remained unchanged (data not shown). Importantly, late-life treated animals had increased mean muscle fibre diameters and showed improved performance in treadmill exercise tests (Fig. 4a, b). Furthermore, most fat depots of these animals were enlarged and adipocyte cell size and subdermal adipose layer thickness were significantly increased (Fig. 4c–e). Senescence markers were substantially reduced in both fat and skeletal muscle of AP20187-treated animals (Fig. 4f, g and Supplementary Fig. 7). Analysis of 5-month-old untreated BubR1H/H;INK-ATTAC-5 mice revealed that the observed improvements in skeletal muscle and fat of late-life treated 10-month-old BubR1H/H;INK-ATTAC-5 mice reflect attenuated progression of age-related declines rather than a reversal of ageing (Fig. 4a, c–e). Thus, late-life clearance of p16Ink4a-positive senescent cells attenuates progression of age-related decline in BubR1 hypomorphic mice.

Figure 4: Treatment of older BubR1H/H;INK-ATTAC mice with AP20187 delays progression of p16Ink4a-mediated age-related phenotypes. a, Mean skeletal muscle fibre diameters of the indicated mice. ABD, abdominal muscle; Gastro, gastrocnemius muscle. mo, months. b, Improvement of exercise ability of the indicated mice relative to age-matched untreated mice. c, Body and fat depots weights of the indicated mice. Parentheses, s.d. Mes, mesenteric; Peri, perirenal; POV, paraovarian; SSAT, subscapular adipose tissue. d, Average size of fat cells in IAT of the indicated mice. e, Subcutaneous adipose layer thickness of the indicated mice. f, SA-β-Gal-stained IAT. g, Expression of senescence markers in IAT and gastrocnemius of the indicated mice (n = 3 females per genotype per treatment). Expression of all genes, except those marked with NS, is significantly decreased (P < 0.05) upon late-life AP20187 treatment. Colour codes in d and e are as indicated in a. Error bars indicate s.e.m. except in g where they indicate s.d. For analyses in a–f: n = 5 5-month-old BubR1H/H;INK-ATTAC-5 –AP females; n = 9 10-month-old BubR1H/H;INK-ATTAC-3 +AP and –AP females; n = 7 10-month-old BubR1H/H;INK-ATTAC-5 +AP females; and n = 8 10-month-old BubR1H/H;INK-ATTAC-5 –AP females. *P < 0.05, **P < 0.01, ***P < 0.001. NS, not significant. Full size image Download PowerPoint slide

Whether and how cellular senescence is related to age-related diseases, frailty and dysfunction has been one of the major open questions in the biology of ageing and clinical geriatrics1. Here we present a novel transgenic mouse model that allows for the inducible removal of p16Ink4a-positive senescent cells. Remarkably, even though transcriptional regulation of endogenous p16Ink4a expression is highly complex, involving various transcriptional activators/repressors, epigenetic mechanisms and antisense non-coding RNA19,20,21,22, we find that expression of INK-ATTAC driven by a relatively small portion of the p16Ink4a promoter closely overlaps with that of endogenous p16Ink4a. By breeding INK-ATTAC mice into a progeroid mouse genetic background, we show that both life-long and late-life clearance of the p16Ink4a-expressing senescent cells selectively delayed age-related pathologies in tissues that accumulate these cells. Furthermore, our data indicate that acquisition of the senescence-associated secretory phenotype (SASP), which enables cells to secrete a variety of growth factors, cytokines and proteases4, contributes to age-related tissue dysfunction. There were no overt side effects of senescent cell clearance in our model, even though it has been postulated that senescent cells enhance certain types of tissue repair23,24. Our proof-of-principle experiments demonstrate that therapeutic interventions to clear senescent cells or block their effects may represent an avenue for treating or delaying age-related diseases and improving healthy human lifespan.