Combining stem cell rejuvenation and senescence targeting to synergistically extend lifespan

Introduction

Life is a constant struggle. This is true at cellular and molecular levels where tissue homeostasis requires constant surveillance, repair and replacement of cells damaged or lost due to intrinsic and extrinsic insults [1]. Stem cells play a pivotal role in this tissue homeostasis by providing a reservoir of pluripotent precursor cells, needed to replace fully differentiated cells that are lost or damaged [2]. At the opposite end of the cell-fate spectrum are senescent cells, or cells that have permanently withdrawn from the cell cycle [3]. Cellular senescence can be replicative, where it is triggered by telomere shortening or mediated by stochastic damage, such as oxidative damage to DNA. Senescent cells can also arise as a response to oncogene activation to oppose transformation and cancerous growth [4]. By entering permanent replicative arrest, senescent cells prevent mutations from expanding, thereby providing a sink for genotoxic damage. This senescent state does not simply result in passive replicative arrest but instead leads to transcriptional changes causing resistance to apoptosis and increased secretion of pro-inflammatory signaling molecules, a process known as Senescence Associated Secretory Phenotype (SASP). Senescent cell induced SASP in turn promotes inflammation and contributes to age-dependent dysfunction and to the development of age-related diseases [5].

While the number of stem cells decreases in aging animals, senescent cells accumulate with age [6]. Manipulating cell fates by cellular reprogramming (to rejuvenate somatic cells) and by senolytic interventions (to remove senescent cells) are two promising approaches to restore homeostasis in aged individuals and to prevent age-dependent diseases. Cellular reprogramming allows differentiated cells to regain plasticity and to take on more stem cell-like qualities. A major step towards this goal was the demonstration of cellular reprogramming of terminally differentiated cells into pluripotent embryonic-like stem cell states [7]. Such reprogramming reverses epigenetic aging marks, demonstrating that even mature, terminally differentiated cells can be returned to a younger state [8]. While continuous expression of the Yamanaka factors (Oct4, Klf4, Sox2, c-Myc; OKSM) in mice led to the formation of teratomas and decreased lifespan [9, 10], repeated short term expression in adult mice succeeded in ameliorating cellular and physiological signs of aging [11–13]. Subsequently, several studies have suggested that this approach can be applied to human aging and age-related disease [14–18], and cycling expression can rejuvenate stem cells in vitro [19].

Ablation of senescent cells has been shown to reverse tissue dysfunction and extend healthspan in mice [20, 21]. A recent study using a senolytic construct (FOXO4-DRI peptide) that induced apoptosis in senescent cells, by interfering with the binding of p53 to FOXO4 thereby freeing p53 to activate apoptosis, showed that the clearing of senescent cells both counteracted senescent cell induced chemotoxicity and restored age-dependent declines in physical performance, fur density, and renal function in aging mice [22]. Several studies have further explored applications of different senolytic strategies to ameliorate age-related decline and disease [6, 23–26].

Accumulation of senescent cells and loss of stem cells are not independent processes. Through SASP, senescent cells release pro-inflammatory cytokines which contribute to chronic inflammation and mTOR activation, ultimately leading to stem cell exhaustion [27]. This interaction suggests that senolytic therapies might interact with cellular reprogramming strategies in delaying age-dependent decline and disease. We have previously explored drug-drug interactions as synergistic aging interventions [28], and here we ask whether a combinatorial treatment of OKSM and senolytic (Sen) expression could mitigate or reverse the effects of aging more efficiently than either intervention alone. To test this hypothesis, we induced expression of OKSM, Sen and an OKSM-Sen combination in adult flies and compared their effects on health and lifespan. We find that each treatment alone had limited benefits, with OKSM alone benefiting maximum lifespan while Sen expression alone increased mean lifespan but had no effect on maximum lifespan. In contrast, animals subjected to the combined intervention experienced substantially longer mean and maximum lifespan. Our data is consistent with a synergistic interaction between the two interventions, simultaneously rejuvenating stem cells and removing senescent cells.

Results

To test the interaction between senolytic removal of senescent cells and cellular reprograming, we designed a model combining these two interventions in an inducible overexpression system in Drosophila. First, we used the four Yamanaka factor based OKSM approach as this had been previously shown to induce pluripotent stem cells in mice [7], humans [29–31] and non-mammalian vertebrate and invertebrate species [32]. To make a senolytic factor for Drosophila, we took advantage of the mouse sequence (FOXO4-DRI [22]) to design an orthologous peptide based on the Drosophila foxo (forkhead box, sub-group O) gene [33]. We then characterized effects of these two interventions independently as well as in combination.

We began by looking at the effect of OKSM and Sen on stem cells in an intestinal stem cell (ISC) model [34, 35]. We chose to investigate phenotypic effects specifically in the digestive system of Drosophila (Supplementary Figure 1). As in mammals, the Drosophila gastric lining has a high turnover of cells which is enabled by stem cell pools that replenish the epithelia [34]. Age-dependent loss of stem cells and degradation of barrier function has been shown to contribute to age-dependent functional decline and mortality in Drosophila [36]. The Drosophila gut is composed of four cell types: enterocytes (ECs or absorptive cells), enteroendocrine (EEs or secretory cells), enteroblasts (EBs or transit amplifying cells) and intestinal stem cells (ISCs). ISCs rest on the external surface of the gut epithelium away from the gut lumen, and divide symmetrically to generate more ISCs, or asymmetrically to form EBs [37]. The small, bright green cells or ISCs, can be observed either by expression of the stem cell determinant escargot (esgGal4>UAS-GFP), or by using a marker of Wnt activation, β-catenin (armadillo or arm in Drosophila), observable by GFP construct inserted into the endogenous locus [38].

We looked at the effect of constant expression of the two factors separately or together over a time course of 28 days. We observed a marked increase in ISC numbers starting at day 7 and continuing into day 28 in all three experimental conditions (Figure 1). We observed an increase in ISCs and transit amplifying cells in OKSM expressing epithelia (Figure 1B, 1F, 1J, 1N), an effect likely explained by the presence of Myc and suggesting that stem cell exhaustion may occur [39]. We saw a similar increase in ISCs and transit amplifying cells in Sen expressing epithelia possibly due to the effect of p53 on stem cells [40]. The increase in ISC numbers in animals expressing OKSM was expected, but surprisingly we observed a large increase in ISCs when Sen was expressed (Figure 1C). Overall, the treatments showed higher numbers of stem cells over time as compared to wildtype flies.

We next looked at lifespan effects. Continuous expression of OKSM is detrimental in mice while repeated short-term expression was beneficial [13]. We expressed OKSM in ISCs only (Supplementary Figure 2A) or ubiquitously (Figure 4C), both of which led to a significant detriment in lifespan.

Recent studies have identified factors involved in senescence and SASP in Drosophila [41–43]. Based on these studies, we used an assay for senescence-associated β-galactosidase (SA-β-gal) to examine whether Sen expression affected the number of senescent cells. We observed a marked absence of SA-β-gal in fly midguts constantly expressing Sen, OKSM and the Sen-OKSM combination at day 40 as compared to both younger and older control flies (Figure 2A–2E and quantified in 2F). The surprising finding was that all three treatments led to the lowering of senescent cells.

To look more closely at the effect of Sen and OKSM, we determined the transcriptional profiles of the three conditions as compared to wildtype. We dissected the midguts of flies constantly expressing OKSM, Sen, OKSM and Sen (OKSM-Sen), and control flies (WT) expressing only a fluorescent protein, either in the ubiquitous expression (i.e., under the control an armadillo driver) or the ISC-restricted expression model (i.e., under control of an escargot driver) and performed RNA-seq (Figure 3, Supplementary Figure 3).

In the ubiquitous expression model, 1282 genes (FDR < 10%) were identified as significantly differentially expressed (Supplementary Table 1). Clustering of these genes identified seven distinct clusters, each representing groups of genes with similar expression profiles across the four conditions (Figure 3A). Each of the seven clusters was enriched for distinct pathways and processes (Figure 3B, Supplementary Table 1). Cluster I (N = 300) contained genes that were downregulated by Sen but were unaffected in the other conditions. Genes in this cluster were associated with cytokinesis, cell cycle and DNA replication. These would be expected results as p53 release from FoxO should lead to apoptosis rather than cell proliferation. Genes in Cluster II (N = 201) were upregulated in all conditions compared to WT. This cluster was enriched for processes associated with Glutathione and sugar metabolism and lysosomal activity, which are related to tissue building and repair. Cluster III (N = 157) consisted of genes that were upregulated only in the OKSM condition and was enriched for lipid and amino acid metabolism. Cluster IV (N = 173) consisted of genes that were downregulated by all conditions relative to WT and was enriched for genes associated with secretion and phagocytosis. Cluster V (N = 147) represented genes that were downregulated in both OKSM and in the combined OKSM-Sen conditions and contained genes related to translation. Cluster VI (N = 204) contained genes that were upregulated specifically by Sen alone and consisted mainly of genes involved in homeostasis. Cluster VII (N = 100) contained genes upregulated by the OKSM-Sen combination which were involved in Arginine and Proline amino acid metabolism. Overall, our results suggested that genes affected by expression of Sen correlated with lower cell division and FoxO signaling, and a higher level of amino acid metabolism. Expression of OKSM upregulated amino acid and lipid metabolism and proteolysis while downregulating translation and sugar metabolism.

In the ISC-restricted expression model, we identified 3791 genes (FDR < 10%) as significantly differentially expressed across conditions (Supplementary Table 2), which were subsequently clustered into seven distinct groups (Figure 3C, 3D, Supplementary Figure 2). Cluster I (N = 991) contained genes that were upregulated in all three conditions relative to WT. Genes in this cluster were associated with cytokinesis, cell cycle, cell migration and DNA replication, further supporting that in ISCs expression of these factors, in any combination, results in increased proliferation and migration of ISCs. Genes in Cluster II (N = 271) were downregulated by the expression of OKSM or Sen. This cluster contained transcription regulators and genes involved in ISC homeostasis. Cluster III (N = 233) consisted of genes that were specifically upregulated by OKSM, and downregulated in the other conditions, and was enriched for fatty acid degradation and peroxisome function. Cluster IV (N = 410) consisted of genes that were upregulated in all conditions, although to a lower extent in Sen, and was enriched for genes involved in vesicle transport and Toll signaling. Cluster V (N = 825) contained genes that were downregulated in all conditions, and contained genes involved in protein translation and signaling pathways. Genes in Cluster VI (N = 642) were downregulated by all of the constructs investigated, albeit not as much by expression of OKSM, and contained genes involved in peroxisome function. Cluster VII (N = 419) contained genes that were upregulated in either Sen or Sen-OKSM and was enriched for genes involved in key signaling pathways and apoptosis. Overall, our transcriptional analysis suggested that gene expression was affected by the expression of Sen or OKSM individually, but that there was not a distinct group that was differentially expressed specifically in response to the combination (Sen-OKSM). Overall, OKSM expression upregulated misfolded protein response and Toll signaling, while downregulating Insulin secretion and protein translation, whereas Sen expression activated Wnt and Hedgehog signaling and downregulated Toll and mTOR pathways.

Next, we compared the sets of differentially expressed genes identified in both models to determine if there was also a shared core transcriptional program was altered in both systems. Overall, 812 genes were differentially expressed in both models (Figure 3E, p < 1 × 10⁻⁶), with these genes being enriched for DNA replication, regulation of epithelial cell migration, mitosis, inflammation, various metabolic processes and specific developmental signaling pathways (Figure 3F). The majority of these genes have similar responses to transgene expression in either model, while some exhibit differences in their response between models.

To evaluate lifespan effects under optimized conditions, we designed two approaches for cycling expression to overcome the continuous expression detriment. We first used a drug induced expression model where the polycistronic OKSM transgene, under the transcriptional control of UAS regulatory sequences, was driven by the Actin-Switch-GAL4 driver activated by RU486 [44]. Flies were placed on fresh food supplemented with the drug weekly leading to periodic, ubiquitous expression. We found that OKSM expression alone resulted in mean lifespan extension in both male and female flies, with female flies showing an increase maximum lifespan as well (Supplementary Figure 2B). The advantage of this system was that flies could be cultured at higher temperatures reducing the overall length of lifespan studies and allowing rapid confirmation of lifespan benefits, however, this system does not allow for more precise control of expression due to drug half-life, consumption, and distribution to all tissues. For this we turned to a temperature sensitive expression system where a ubiquitous GAL4 driver was combined with a ubiquitous, temperature sensitive GAL80 inhibitor. This system allowed us to generate adults with no embryonic expression, and by modulating the temperature of culture, we were able to precisely induce expression in all tissues for defined periods ranging from constant to once per week. For each of these experiments the control cohort expressing a fluorescent protein alone was subjected to the same temperature cycling profile as the experimental strains. Using this approach, we found that continuous expression of OKSM was detrimental (Figure 4C), expression for 24 hours twice per week was mildly beneficial (Figure 4B), and for 12 hours twice per week showed lifespan extension (Figure 4A).

We carried out similar optimization experiments for the senolytic peptide (Sen) alone and found that, in terms of median lifespan, continuous expression was also detrimental while expression for either 24 or 12 hours twice per week resulted in significant lifespan extension (Figure 4). Having established conditions under which each individual intervention was beneficial, we then tested whether simultaneous removal of senescent cells (Sen) and cellular reprogramming (OKSM) would result in additive or synergistic benefits in aging flies. The combined intervention again was detrimental when expression of Sen and OKSM was induced continuously, but extended both maximum lifespan and median lifespan when expressed for 24 hours twice per week (Figure 4). Most striking was the significant mean and maximum lifespan extension noted in flies with OKSM and Sen expressed together for 12 hours twice per week (Figure 4A).

We reasoned that longer-lived flies should maintain stem cell pools for longer due to stem cell rejuvenation through OKSM expression. To test this hypothesis, we examined the midguts from flies expressing the various transgenes over a time course of 28 days using cycling expression of 12 hours twice per week. As these flies were expressing the factors ubiquitously, we could not use the esg>GFP marker for ISCs and instead used an endogenously GFP-tagged allele of the Wnt responsive β-catenin gene [38]. ISCs show higher levels of cytoplasmic β-catenin protein making them readily identifiable, but in addition other cell types in the epithelium show junctional β-catenin [45]. We visualized ISCs in gastric epithelia of flies with cycling expression of OKSM, Sen or both at four weeks (Figure 5A–5D), eight weeks (Figure 5G–5J), and twelve weeks (Figure 5M–5P). We observed and quantified a higher number of ISCs in flies when OKSM was expressed (Figure 5E, 5F, 5K, 5L, 5Q, 5R). These findings were not consistent with a loss of stem cells in aging organisms, but rather may reflect the loss of stem cell functionality that comes with accumulated damage [46].

Our observations for periodic expression of Sen were consistent with data from mice subjected to senolytic interventions. Sen flies experience a substantial increase in mean but not in maximum lifespan, indicating compression of mortality with excess late deaths compensating for protective effects earlier in life. The same is not true for OKSM flies which experience a statistically significant maximum lifespan extension with both 24 h and 12 h induction. Strikingly, simultaneous application of Sen and OKSM, especially for 12 h induction, result in a mortality trajectory that combines beneficial features of both individual interventions and result in mean and maximum lifespan extension benefit that exceed either. To further investigate this interaction, we performed a quantitative analysis of age-dependent mortality.

Biological aging is defined by an exponential increase in mortality rate over time. Mathematically, this is expressed by the Gompertz–Makeham law of mortality [47]:

Where A is the initial mortality rate in young animals, B is the age-independent mortality and MRDT is a characteristic time interval over which age-dependent mortality doubles. To quantitatively compare the impact of each single intervention and of the combined intervention on age-dependent mortality, we followed an approach recently described by Axel Kowald and Tom Kirkwood, fitting Gompertz–Makeham survival functions to our experimental survival data [48]. This fit yielded estimates for the initial mortality A and the MRDT parameters of flies (Supplementary Table 3). All fits resulted in a good agreement between experimental data and the Gompertz curve, with a mean residual standard error of 0.03 (3% survival) across all conditions (Figure 6B, 6D, 6F, Supplementary Table 3). Mortality trajectories were then visualized by plotting the logarithm of mortality against age (Figure 6A, 6C, 6E). In this graph the initial mortality A is the intercept of the mortality trajectory at time zero while the slope of the line is proportional to the inverse of the MRDT. Sen-driven lifespan extension showed a substantial decrease in early mortality A, relative to control. However, this decrease was associated with a significant penalty in the form of age-acceleration (decreased MRDT). For the 12 h induction, early mortality decreases almost 100-fold while MRDT decreases from 22.7 days to 7.9 days (p < 0.05). In other words, while initial mortality is substantially lower following Sen treatment, the treated flies age approximately 3 times faster than WT. This pattern was consistent with previously described mouse data and explains why Sen treatment results in mortality compression with increased mean but not maximum lifespan. By contrast, the impact of OKSM induction on initial mortality and MRDT was much smaller. OKSM induction for 24 h and 12 h significantly reduced initial mortality A by 46% and 30%, (p < 0.05), respectively (Supplementary Table 3). While OKSM was also associated with a slight age acceleration penalty in terms of MRDT, this effect was much smaller than for Sen. For 12 h OKSM induction, MRDT only decreased from 22.7 days to 19.6 days (p < 0.05); a 15.8% increase in aging rate. The result of this change can be seen when plotting log mortality as a function of age (Figure 6C, 6E). OKSM mortality was shifted downwards relative to control but ran nearly parallel to control mortality, meaning that age-dependent mortality for 12 h OKSM animals remained lower than for WT at later ages. This pattern explains why OKSM impacted maximum lifespan and was not associated with mortality compression.

OKSM-Sen flies with 24 h of induction experienced a significant reduction in the age-acceleration penalty (p < 0.05) relative to 24 h Sen-only flies (Figure 6C, Supplementary Table 3). OKSM-Sen flies with 12 h induction also showed this trend, but the difference was not statistically significant (Figure 6E, Supplementary Table 3). In contrast, the initial mortality A was decreased by over 330-fold in 12 h OKSM-Sen compared to WT. This means that initial mortality rate was significantly lower in OKSM-Sen than even in Sen-only animals (p < 0.05), suggesting that adding OKSM to Sen partially rescued the age-acceleration penalty while further augmenting Sen benefits in terms of early mortality. The resulting survival trajectories consequently show both mean and maximum lifespan extension, with mortality compression occurring only late in life, after approximately day 150 when the mortality of OKSM-Sen crosses that of controls. Mortality compression therefore only becomes apparent after most WT animals have already died.

When investigating the interaction between Sen and OKSM treatments, it is useful to compare observed effects to a hypothetical survival and mortality trajectory constructed by assuming that the two interventions act independently (see materials and methods for details). Comparing the OKSM-Sen group to this hypothetical cohort (purple dashed lines in Figure 6) revealed that the actual, observed effects cannot be explained without a direct interaction between Sen and OKSM in terms of aging rate. On their own, both interventions accelerate aging rate (decrease MRDT) but lower early mortality. However, when combined, these interventions result in a reduction, rather than further increase, in the age-acceleration penalty while further augmenting early mortality benefits. This synergistic interaction between OKSM and Sen is the reason why the combined treatment improves both maximum and median lifespan more significantly than either of the two interventions alone. Mechanistically, these data imply a direct interaction between partial reprogramming via OKSM and the Sen-driven senescent cell apoptosis. Indeed, this is what we observed on the cellular level, as expression of Sen impacted the number of stem cells directly (Figure 1) even without OKSM expression.

Discussion

Here we show that it is possible to extend both the mean and maximum lifespans by combining strategies targeting two different ageing mechanisms related to cell fate. Pulsed expression of the four Yamanaka transcription factors to rejuvenate cells combined with a Senolytic factor kept flies healthier and extended their lives. To the best of our knowledge, this is the first study to show lifespan extension in an otherwise normal animal through the expression of Yamanaka factors. Although not tested in our study, reprogramming leads to a change in DNA methylation and other epigenetic markers leading to a more youthful gene expression signature [16]. The periodic removal of senescent cells leads to fewer chemotoxic molecules being produced and in rejuvenation of organs [22]. Both interventions are rejuvenating in the sense that they reverse cellular tissue composition towards a more youthful state (fewer senescent cells, preserved stem cell pools). The substantial reduction in initial mortality following both interventions is consistent with this mechanism. Here we report that these two interventions are more closely related than previously appreciated.

It is important to note that in this study we used the mammalian versions of OKSM and based the Sen fragment on the published mammalian interaction domain. We did not use the Drosophila homologs of OKSM, but rather showed that at least some functionality is conserved from flies to mammals. The same is true for the Sen peptide, which may function through a different mechanism in Drosophila than that proposed for mammals due to the differences in p53 function and FoxO/p53 interactions between the species [49]. Here we focused on the conserved aspects affecting lifespan, rather than the possible differences in mechanism.

Senescent cells show persistent activation of the mTOR pathway, a state that promotes secretion of a wide range of signaling molecules, including proinflammatory cytokines [50–52]. As these molecules are secreted, they have the potential to impact neighboring or distant cells increasing the number of senescent cells, impairing tissue homeostasis [53, 54]. In the ubiquitous expression model, we found that expression of the Senolytic peptide led to a decrease in Tgfβ (−0.51 log₂ Fold decrease, FDR = 7.51 × 10⁻²) and the cytokine upd3 (−1.0 log₂ Fold decrease, FDR = 2.91 × 10⁻⁴) compared to Control (Supplementary Figure 4). Upd3 activates Jak/Stat signaling often related to stem cell activation upon injury leading to asymmetric divisions and lower numbers of stem cells [55, 56]. Tgfβ is upstream of upd3 and involved in promoting senescence [57–59]. SASP-mediated activation of cytokines and mTOR therefore directly link age-dependent accumulation of senescent cells to accelerated loss of stem cells and declining capacity for repair and tissue regeneration. This mechanism suggests a model by which removal of senescent cells would promote increased resilience and improved maintenance of stem cell pools, a phenotype we observe in the fly.

The pathways and processes perturbed by expression of OKSM, Sen or Sen-OKSM are associated with those previously identified as been involved in the hallmarks of aging [46]. The genes affected by expression of Sen reflect those that would be expected to be altered by disrupting FOXO: p53 interactions, i.e. apoptosis, whereas those altered by expression of OKSM include genes involved in ISC function and homeostasis. RNA-seq analysis indicates that the extension of lifespan is the result of two largely distinct transcriptional programs, and is not the result of Sen-OKSM specifically activating or repressing a shared transcriptional pathway. A major class of genes affected were in various metabolic pathways. Although we have not investigated this here, the metabolic changes resulting from rejuvenation and senolytic treatments will be interesting to consider in future work.

Although OKSM must function through partial reprogramming of cells, the exact mechanism of how this works in adult tissues is not entirely clear [13, 16–18]. We observed changes in Hedgehog signaling recently proposed as a neuroprotective and life extending pathway [60] along with genes associated with cytokinesis and DNA replication. Importantly, we observe that OKSM has limited effect on maximum lifespan unless senescent cells are removed suggesting that SASP counteracts the benefits of rejuvenation. Previous studies have shown either OKSM or Sen to be anti-aging, but in both cases the effects did not affect maximum lifespan. In our combinatorial approach, we can now extend both mean and maximum. We have further established that both approaches can be studied in the easily, genetically manipulatable Drosophila model. We suggest that reprogramming accomplished through gene therapy, or another method combined with senolytic peptides or drugs could promote both tissue repair and reverse age-related decline.

Materials and Methods

Supplementary Materials

Supplementary Figures

Supplementary Table 1

Supplementary Table 2

Supplementary Table 3