Mitochondrial phosphatase PGAM5 modulates cellular senescence by regulating mitochondrial dynamics

Mitochondria undergo dynamic fusion/fission, biogenesis and mitophagy in response to stimuli or stresses. Disruption of mitochondrial homeostasis could lead to cell senescence, although the underlying mechanism remains unclear. We show that deletion of mitochondrial phosphatase PGAM5 leads to accelerated retinal pigment epithelial (RPE) senescence in vitro and in vivo. Mechanistically, PGAM5 is required for mitochondrial fission through dephosphorylating DRP1. PGAM5 deletion leads to increased mitochondrial fusion and decreased mitochondrial turnover. As results, cellular ATP and reactive oxygen species (ROS) levels are elevated, mTOR and IRF/IFN-β signaling pathways are enhanced, leading to cellular senescence. Overexpression of Drp1 K38A or S637A mutant phenocopies or rescues mTOR activation and senescence in PGAM5−/− cells, respectively. Young but not aging Pgam5−/− mice are resistant to sodium iodate-induced RPE cell death. Our studies establish a link between defective mitochondrial fission, cellular senescence and age-dependent oxidative stress response, which have implications in age-related diseases.

C ellular senescence is a phenomenon originally defined as stable cell cycle arrest in response to different stresses 1 . Besides cell cycle exit, senescent cells undergo phenotypic changes including metabolic reprogramming, chromatin rearrangement and autophagy modulation 2 . Features of senescence include increased cellular volume, shortened telomeres, increased reactive oxygen species (ROS) level, persistent DNA damage response, elevated senescence-associated β-galactosidase activity, p16 Ink4A expression, etc. Senescent cells produce and secrete a combination of factors to exert non-cell-autonomous effects, referred to as senescence-associated secretory phenotype (SASP). Senescence can be a protective mechanism against stress or cancer. However, accumulation of senescence cells could drive aging and age-related diseases. Age-related macular degeneration (AMD) is a degenerative disease of the retina and the leading cause of irreversible blindness in the elderly 3 . Dry AMD is characterized by degeneration of retinal pigment epithelial (RPE) layer at late stage. The role of RPE senescence in AMD remains unclear 4 .
Mitochondria has critical roles in cellular senescence due to ROS generation from oxidative phosphorylation 5 . It undergoes dynamic fusion/fission, biogenesis and mitophagy in response to physiological stimuli or pathological stresses. The role of mitochondrial fusion/fission in aging (or senescence) has been controversial. Inhibition of mitochondrial fission by deleting a mitochondrial fission protein dynamin-related protein 1 (Drp1) or maintenance of the fused mitochondrial network is necessary for longevity in yeast or C. elegans 6,7 . However, disruption of either mitochondrial fission or fusion significantly reduces medium lifespan in C. elegans, while promoting Drp1-mediated mitochondrial fission in midlife prolongs healthy lifespan of Drosophila melanogaster 8,9 . These controversies suggest the necessity to clarify mitochondrial fission/fusion function in aging (senescence) and longevity, especially in higher vertebrate species.
Mitchondrial fission/fusion, mitophagy and biogenesis collectively control mitochondrial turnover. Mitophagy, a process of selective engulfment of dysfunctional mitochondria for degradation by lysosome, is indispensable for maintaining cellular homeostasis. Impaired mitophagy indicates less mitochondrial turnover, which leads to accumulation of dysfunctional mitochondria, senescence and age-related disorders 10 . Mitophagy is orchestrated by mitochondrial dynamics 11 . Drp1-driven mitochondrial fission facilitates mitophagy by dividing mitochondria into fragments or segregating damaged mitochondrial subdomains for autophagosome engulfment 12,13 .
Phosphoglycerate mutase 5 (PGAM5) is a mitochondrial Serine (Ser)/Threonine (Thr) phosphatase normally located in the inner mitochondrial membrane. Upon mitochondrial dysfunction, PGAM5 recruits and dephosphorylates Drp1 at Ser-637, triggers its GTPase activity and promotes mitochondrial fission 14,15 . PGAM5 regulates mitophagy by stabilizing PINK1 under stress conditions, which recruits E3 ubiquitin ligase PAR-KIN for degradation of the damaged mitochondria 15,16 . PGAM5 also regulates mitophagy independent of PARKIN by interacting and dephosphorylating FUNDC1, which interacts with LC3 (ref. 17 ). In the end, PGAM5 can be cleaved and released to the cytoplasm through PARKIN, which activates Wnt signaling and induces mitochondrial biogenesis 18 . PGAM5 also regulates antioxidative response by forming a tertiary complex with KEAP1 and NRF2 (ref. 19 ). It regulates necroptosis by acting as a RIPK3 target and recruiting the RIPK1-RIPK3-MLKL necrosis "attack" complex to mitochondria 14 . Pgam5 −/− mice show Parkinson-like movement disorder 16 . In sum, PGAM5 has multiple functions and could act as signaling hub to sense mitochondrial stress, regulate mitochondrial dynamics and antioxidative response.
Given the importance of PGAM5 in mitochondrial dynamics, we ask whether PGAM5 regulates cellular senescence and agedependent anti-oxidative response. Through in vitro and in vivo approaches, we show that PGAM5 is essential for mitochondrial homeostasis, and Pgam5 deficiency induces accelerated senescence in mice. PGAM5 deletion leads to reduced mitochondrial turnover, increased ATP and ROS levels, elevated mTOR and IRF/IFN-β signaling pathways. Collectively, these result in cellular senescence and age-related reduction in anti-oxidation capability. ATP elevation, mTOR activation and senescence in PGAM5 −/− cells can be phenocopied or rescued by overexpression of DRP1-K38A or S637A mutant, respectively, underscoring the significance of mitochondrial fission in repressing cellular senescence.

Results
PGAM5 deletion leads to accelerated cell senescence. To determine the role of mitochondrial fusion/fission in mouse aging, Pgam5 flox/flox mice were generated using verified ES cell from European Mouse Mutant Cell Repository, and Pgam5 −/− allele was made by crossing Pgam5 flox/flox mice (C57BL/6J background) with CAG-Cre mice (Fig. 1a). As a Laz cassette was infused between exon I and exon II in Pgam5 flox or knockout allele, Laz staining was used to monitor PGAM5 protein expression. In the retina, LacZ signal was enriched in the retinal pigment epithelium (RPE) layer, retinal ganglion cells and ciliary body epithelium (Fig. 1b). Successful knockout (KO) of Pgam5 in the mice was confirmed by western blot analysis using PGAM5 antibody (Fig. 1c). Pgam5 −/− mice were born at Mendel ratio, but weighed significantly less compared to the wild-type (WT) controls at 1-year and older ( Supplementary Fig. 1a, b).~50% of the Pgam5 −/− mice (aged 1-1.5 years) showed lordokyphosis (hunchback), odd gait and swollen foot, and among them~50% died from unknown causes or were too sick and killed as required by animal facility (see Kaplan-Meier survival curve in Fig. 1d). These phenomena were not observed in age-matched WT control mice. Aged Pgam5 −/− mice appeared to have fewer movements under a new environment compared with control mice (Supplementary Video 1), consistent with the Parkinson's-like movement disorders observed previously 16 . Together, these suggest an accelerated senescence (or aging) phenotype in Pgam5 −/− mice.
PGAM5 deletion induces changes in mitochondrial dynamics. To explore the underlying mechanism of PGAM5 deletioninduced cellular senescence, mitochondrial morphology and dynamics were initially evaluated 14 . Compared to the controls, mitochondria of PGAM5 −/− ARPE-19 cells exhibited more tubular and less fragmented structures, as quantified by a~30% increase in the average length of Tom20-positive mitochondrial branches ( Fig. 3a and Supplementary Fig. 2a). Consistent with PGAM5 regulation of mitochondrial fission by dephosphorylating Drp1, Drp-1 (Ser-637) phosphorylation was increased in PGAM5 −/− ARPE-19 cells, supporting that PGAM5 is required for mitochondrial fission via dephosphorylating Drp1 (Fig. 3b, and Supplementary Fig. 2b).
PGAM5 has been reported to have long and short forms, as well as full-length and cleaved forms 14 . Cleaved PGAM5 retains its phosphatase domain 25  to cytosol to interact with Axin1 (ref. 18 ). Moreover, dephosphorylation of Drp1-Ser-637 is essential for its binding to mitochondria 26 . Based on these, we tested the hypothesis that cleaved PGAM5 interacts with Axin1 and recruits Drp1 for dephosphorylation and fission processing. Both full-length and cleaved PGAM5 forms were found is ARPE-19 cells, with cleaved form dominant when induced by mitochondrial uncoupling agent carbonyl cyanide chlorophenylhydrazone (CCCP), consistent a decrease in Drp1 (Ser-637) phosphorylation (Fig. 3c). A shorter cleaved form the claimed "short form" was not detected, disapproving the existence of the PGAM5 short form. By coimmunoprecipitation assay using antibody to Axin1, both Drp1 and cleaved PGAM5 can be co-immunoprecipitated with Axin1 ( Fig. 3d). Co-localization of Axin1, Drp1 and PGAM5 was also confirmed by immunofluorescence, supporting that cleaved PGAM5 recruits and dephosphorylates Drp1 through interacting with Axin1 ( Supplementary Fig. 2c). As Drp1-mediated mitochondrial fission is required for mitochondrial homeostasis, we hypothesized that hyperfusion from PGAM5 deficiency could result in less mitochondrial turnover 11,15,[27][28][29] . The mitochondrial outer membrane protein Tom20, inner membrane-associated protein cytochrome C, and mitochondrial matrix protein, Cyclophilin F (CYPD), were all increased by 20-40% in PGAM5 −/− ARPE-19 cells, indicating more mitochondrial mass (Fig. 3e, and Supplementary Fig. 2d). Consistently, the mitochondrial DNA copy number was also increased in PGAM5 −/− ARPE-19 cells (Fig. 3f). The increase in mitochondrial mass was also confirmed in vivo by western blot using CYPD antibody in RPE/choroid tissue isolated from Pgam5 −/− mice (Fig. 3g). To test whether mitochondrial biogenesis contributes to increased mitochondrial mass in PGAM5 −/− RPE cells, the protein level of PGC1α, a critical cotranscriptional regulator for mitochondrial biogenesis, was examined. PGC1α was significantly downregulated by PGAM5 deletion, arguing against the possibility that increased mitochondrial biogenesis contributes to increased mitochondrial mass in PGAM5 −/− cells (Fig. 3e). Instead, more mitochondrial mass and less mitochondrial biogenesis support less mitochondria turnover by PGAM5 deletion. To directly monitor mitochondrial turnover, MitoTimer labeling was used 30 . MitoTimer is a mitochondriatargeting florescence protein that is green once being translated and becomes more red over time. As expected, PGAM5 deletion induced more "Red" mitochondria accumulation than control cells, supporting less mitochondrial turnover (Fig. 3h). Mitochondrial membrane potential (ΔΨ M ) is an important parameter for mitochondrial function. JC-1 dye was used to measure the ΔΨ M change, JC-1 could distinguish between low and high ΔΨ M in mitochondria. The high membrane potential allows JC-1 to form red J-aggregates, but in low membrane potential mitochondria, JC-1 forms green J-monomers. Increased number of unhealthy "green" labeled low membrane potential mitochondria were observed in PGAM5 −/− ARPE-19 cells compared to the control cells ( Fig. 3i and Supplementary Fig. 2e). Moreover, the proportion of the "red" high membrane potential mitochondria was increased in PGAM5 −/− ARPE-19 cells even after normalization to the "green" signal ( Fig. 3i and Supplementary Fig. 2f). These indicate that both "healthy" (red) and "unhealthy" (green) mitochondria are increased in PGAM5 −/− ARPE-19 cells due to the accumulation in total mitochondrial mass. Together, our data suggested PGAM5 deletion leads to mitochondrial hyperfusion and less mitochondrial turnover.
To examine the functional consequence of the mitochondrial morphological and dynamic changes, mitochondrial ATP and reactive oxygen species (ROS) production was measured. Consistent with the increased mitochondrial mass and less turnover, PGAM5 −/− ARPE-19 cells under short-term culture (1 week) produced 67% more ATP compared to the controls (Fig. 3j). However, the increased production of ATP did not sustain overtime. PGAM5 −/− RPE under long-term (8 weeks) culture produced~30% less ATP compared to the controls. Elevated ROS could contribute to cellular senescence 5 . Cellular ROS level was increased~50% in PGAM5 −/− ARPE-19 cells under short-term culture, but significantly decreased in cells under long-term culture as shown by DCFDA dye staining (Fig. 3k).
Regulation of AMPK-mTOR pathway by PGAM5. As ATP was elevated in PGAM5 −/− ARPE-19 cells, AMPK-mTOR signaling pathway was examined as underlying mechanism for PGAM5regulated senescence. Cellular AMP:ATP ratio reflects the energy requirement of a cell, and can be sensed by intracellular sensor AMP-activated protein kinase (AMPK) 31,32 . AMPK phosphorylation but not total AMPK level was reduced by more than 50% in PGAM5 −/− ARPE-19 cells cultured for a week (Fig. 4a). TSC2, a negative modulator of mTORC1, is regulated by AMPK 33,34 . As expected, TSC2 phosphorylation was significantly reduced, and mTORC1 activity indicated by the phosphorylation of S6 ribosomal protein was markedly boosted in those cells. These were confirmed by PGAM5 siRNA experiments. mTOR pathway can be activated by protein biosynthesis inhibitor cycloheximide (CHX) treatment in MEF and HEK293 cells 35 . We found CHX stimulates mTORC1 pathway in ARPE-19 cells. Higher basal level and stronger CHX-induced S6 and 4EBP1 phosphorylation were observed upon PGAM5 silencing (Fig. 4b). mTOR pathway activation by PGAM5 silencing was not restricted to ARPE-19 cells and was also observed in primary human RPE (HRPE) and human umbilical vein endothelial cells (HUVEC) ( Supplementary  Fig. 3a, b). Upregulation of S6 phosphorylation was also confirmed in vivo by western blot of RPE/choroid tissue and immunostaining of retinal tissue in 2-month-old Pgam5 −/− mice (Fig. 4c, d). To further validate the causative relationship between ATP elevation and mTORC1 activation, ATPase inhibitor oligomycin A or ROS inhibitor N-acetyl-L-cysteine (NAC) was used to treat PGAM5 −/− RPE cells. Activation of mTOR signaling was blunted by oligomycin A but not NAC (Fig. 4e). This supports that the elevated ATP level but not ROS is responsible for mTOR activation in PGAM5 −/− RPE cells. These data collectively indicate that ATP elevation by PGAM5 deletion stimulates mTORC1 activity, which could explain the accelerated senescence in PGAM5 −/− cells 36 .
Age-dependent response to oxidative stress by PGAM5 deletion. PGAM5 −/− lymphocytes and MEF cells are resistant to oxidative stress in caspase-independent manner 39 . To address whether PGAM5 deletion protects RPE cells from oxidative stress, we employed a NaIO 3 -driven RPE degeneration model 40 . Fundus pictures showed no obvious difference between Pgam5 −/− and WT mice at 2 or 18 months (Fig. 7a, f). However, at 5 days after NaIO3 injury, the number of white spots, indicative of RPE degeneration, appeared to be less in 2-month-old Pgam5 −/− fundi compared to WT mice, suggesting Pgam5 deletion renders oxidative stress resistance (Fig. 7b). Protection of RPE layer was confirmed by ZO-1 staining of RPE flatmount and histological sections (Fig. 7c-e). The damage induced by NaIO 3 -derived oxidants depends on its local concentration after tail vein injection. The ocular arteries and veins converge through the optic disc in the retina, generating a concentration gradient of NaIO 3 with higher concentration adjacent to the optic disc. By ZO-1 staining, the tight junction of RPE cells were interrupted in WT mice (Fig. 7d, left panel), as confirmed by the dislocated RPE cells from the RPE layer in histological sections (Fig. 7e, WT panels). However, in Pgam5 −/− mice, RPE damage was much less severe in the middle and periphery region compared to the corresponding regions in the WT as shown by ZO-1 (Fig. 7d, right panel) and histology staining (Fig. 7e, KO middle and periphery panels). However, when similar experiments were performed in the 18-months-old mice, RPE protection rendered by Pgam5 deletion was lost (Fig. 7f-j).
To confirm the results in vitro, neutral red assay and propidium iodide (PI) staining were used to measure RPE cell viability after NaIO 3 treatment. PGAM5 −/− ARPE-19 cells showed significantly better viability than controls at 12 h after NaIO3 treatment at several concentrations (Fig. 8a). PI staining confirmed less cell death in PGAM5 −/− cells compared to  (Fig. 8b, c). To further investigate the underlying mechanism, qRT-PCR was used to evaluate the expression of a list of anti-oxidative genes. The expression of FOXO4, SOD1 and NRF2 was significantly upregulated in PGAM5 −/− ARPE-19 cells, consistent with the observed ROS level increase (Fig. 8d). Increased NRF2 protein level was confirmed by western blot analysis (Fig. 8f, g). However, the upregulation was lost when cells were cultured for 8 weeks. Decreased expression of FoxO4, Sod1 and Nrf2 mRNA and Nrf2 protein expression was detected instead (Fig. 8e, h, i). Consistently, upregulation of anti-oxidative genes was also confirmed in the RPE/choroid of 2-months-old Pgam5 −/− mice ( Supplementary Fig. 5a). Moreover, the upregulation of Nrf2 and its downstream gene Nqo1 was lost in the RPE/ choroid tissue in 18-months-old Pgam5 −/− mice (Fig. 8j, Supplementary Fig. 5b). These suggest that PGAM5 deletion renders age-related resistance to oxidative stress.
Loss of PGAM5 promotes immune regulatory pathway. Interferon regulatory factor (IRF) pathway is critical in innate immune response activation, inflammation and senescence 41,42 . As phospho-rH2AX, a marker for DNA damage, was activated by PGAM5 deletion, we asked whether IRF pathway is activated in PGAM5 −/− cells. By western blot, IRF-3 phosphorylation is significantly upregulated in PGAM5 −/− ARPE-19 cells (Fig. 9a). Consistently, RNA level of IFNβ, a downstream gene of IRF-3, was significantly upregulated in PGAM5 −/− or PGAM5 silenced ARPE-19 cells (Fig. 9b, c). Moreover, Ιfnβ, Ccl2 and Ccl4 RNA level was increased during aging in RPE/choroid tissue, and a trend of higher expression was observed in Pgam5 −/− mice ( Fig. 9d and Supplementary Fig. 5c, d). These suggest the activation of IRF pathway by PGAM5 deletion. To examine its consequence, subretinal microglia cells were quantified after Iba-1 staining. As shown in Fig. 9e, f, subretinal microglia cells were increased during aging, and higher numbers were observed in Pgam5 −/− mice compared to the WT controls. Taken together, we found that PGAM5 deficiency promotes the IRF/IFN-β pathway activation, which results in more microglia accumulation in the RPE/subretinal space.

Discussion
Our results reveal an important function for PGAM5 in senescence. We found that: (1) PGAM5 deletion leads to accelerated senescence in vitro and in mice; (2) PGAM5 deletion leads to mitochondrial hyperfusion and less mitochondrial turnover; (3) PGAM5 deletion leads to activation of mTOR and IRF/IFN-β pathways, which are essential pathways for senescence; (4) Drp1-K38A overexpression phenocopies but Drp1-S637A overexpression rescues the mitochondrial morphology change, mTOR activation and senescence in PGAM5 −/− RPE cells; (5) PGAM5 deletion renders age-dependent RPE cells resistance to oxidative stress. These data establish a link between PGAM5, mitochondrial fission, senescence and anti-oxidative response, reinforcing the importance of mitochondrial dynamics in regulating cellular senescence (Fig. 10). Accumulation of senescent cells has been observed during aging and age-related diseases 43 . Elimination of senescent cells could restore tissue homeostasis or prevent neurodegeneration, inciting a senolytic approach for treating age-related diseases 44,45 . The molecular mechanism underlying senescence remains unclear, but ROS, mTORC1 and IRF/IFN-β each was shown to contribute to cell senescence 5,41,46 . We found that PGAM5 deletion leads to accelerated senescence, as indicated by increased cell volume, senescence β-gal staining, and a series of other markers and SASPs. PGAM5 −/− mice show reduced survival by 18-month, with~50% of them showing hunchback, swollen foot and odd gait phenotypes, and Parkinson's-like movement disorders as reported previously 16 . This phenotype was confirmed by senescence biomarker staining. In contrast to WT mice, 18months-old Pgam5 −/− mice show increased P16 Ink4a , phospho-rH2AX staining, p53 and MMP3, decreased Lamin B1, as well as increased age-related accumulation of microglial cells in the subretinal space. These support Pgam5 −/− mouse as model for accelerated senescence. As Parkinson's disease is age-related, linking senescence to Parkinson-like syndrome in vivo supports senescence could contribute to Parkinson's disease 47 . Several mouse models of accelerated cellular senescence have been reported, confirming the important of genome instability (Ercc1 −/− mouse), ROS (SOD1 −/− mouse) and nuclear membrane compromise (Lmna mouse) in senescence [48][49][50][51][52] . Our Pgam5 −/− mice represent an accelerated senescence mouse model caused by defective mitochondrial dynamics. Of note, accumulation of dysfunctional mitochondria is associated with senescence or age-related diseases (16)(17)(18). Our model represents a model more related to mitochondrial metabolism and natural aging process, underscoring the important of mitochondrial dynamics in normal life, senescence and aging, and potentially age-related diseases. We explored whether PGAM5 expression level is altered in healthy old people or AMD patients. Analysis of PGAM5 expression in GSE50195 (ref. 53 ) and GSE29801 (ref. 54 ) revealed PGAM5 mRNA level is not significantly altered with age or in AMD disease ( Supplementary Fig. 6a-c). Further studies would focus on total and cleaved PGAM5 protein level and PGAM5 mutations in age-related disease models. Alternatively, PGAM5 activity might not change during aging, but still mediate multiple pathways involved in senescence.
phosphatase to dephosphorylate Drp1 mediated by Axin1. Damaged mitochondria (such as CCCP treatment) induces PGAM5 cleavage and release to cytosol, where it binds to Axin1. Drp1 (Ser-637) is also recruited to Axin1 and subsequently dephosphorylated by PGAM5. The dephosphorylated Drp1 binds to mitochondria and promotes mitochondrial fission and turnover, therefore removing the damaged mitochondria. PGAM5 deletion leads to accumulation of phosphorylated and therefore inactive form of Drp1. This leads to impaired mitochondrial fission and less mitochondrial turnover, which generated more tubular mitochondria. The change in mitochondrial dynamics has two consequences: ROS increase and higher ATP production initially but lower ATP level over longer time. Initial higher ATP production may be good for cells energy-wise, but it stimulates NaIO3 (5d) NaIO3 (5d) NaIO3 (5d) NaIO3 (5d) NaIO3 (5d) NaIO3 (5d) NaIO3 (5d) NaIO3 affect mTOR pathway activation in our study but could contribute to senescence through other mechanisms 5 ; (2) PGAM5 may regulate mitochondrial biogenesis. We found that PGAM5 deletion leads to drastic downregulation of PGC1α, which is a critical cotranscriptional factor for mitochondrial biogenesis 55 . Consistently, cleaved PGAM5 has been shown to be released to the cytoplasm and activate mitochondrial biogenesis 18 . A decrease in mitochondrial biogenesis could contribute to less mitochondrial turnover, leading to accumulation of damaged or senescent mitochondria in the long run. Further experiments would directly confirm mitochondrial biogenesis regulated by PGAM5; (3) PGAM5 deletion leads to IRF/IFN-β pathway activation. Activation of IFNβ by either PGAM5 deletion or PGAM5 siRNA suggests PGAM5 deletion could lead to IRF/IFN-β pathway activation. Consistently, we observed increased age-related accumulation of microglial cells in the subretinal space in Pgam5 −/− mice. We propose that mitochondrial or nuclear DNA damage caused by PGAM5 deletion could lead to IRF/IFN-β pathway activation. In total, we found ROS increase, mTOR and IRF/IFN-β activation in PGAM5 −/− cells. Therefore, our study provides molecular underpinning leading to ROS elevation, and mTOR and IRF/IFN-β activation, which eventually result in senescence. Besides Drp1, other PGAM5 substrates could also contribute to PGAM5 function. For example, PGAM5 may also directly regulate mitophagy and autophagy through PINK1 and FUNDC1. Future work is needed to dissect the full spectrum of PGAM5 targets and function. Controversies exist regarding mitochondrial dynamics, especially mitochondrial fusion/fission in senescence. We found that PGAM5 deletion leads to mitochondrial hyperfusion, less mitochondrial turnover, and eventually accelerated senescence in   56 reported that knocking out hFis1 (Mitochondrial fission 1 protein, which recruits Drp1 to mitochondrial surface for fission) generates elongation of mitochondria and cellular senescence phenotypes in vitro. Rana et al. 8 showed that upregulated Drp1 expression in midlife in Drosophila lead to more mitochondrial fission and prolonged lifespan. However, contradictory results also exist. Scheckhuber et al. 57 found that reducing mitochondrial fission by deleting Dnm1p (Drp1 homologous gene) in fungal models leads to fungal with low growth rate but extended lifespan. Maintenance of the fused mitochondrial network in C. elegans is necessary for longevity, and absence of mitochondrial fusion in the Deltamgm1 mutant leads to a striking reduction of both replicative and chronological lifespan in C. elegans 7,58 . In a recent study, disruption of either mitochondrial fission or fusion significantly reduces medium lifespan in C. elegans 9 . For mouse study, Nishimura et al. 59 reported that Drp1 and filamin interaction induced by hypoxia could induce mitochondrial hyperfission and myocardial senescence. The different outcomes from these studies may reflect on the lifespan and other characteristics of different species. For example, single cell organism may have different response to mitochondrial fission/fusion from mammalians. Extreme conditions like hypoxia may induce acute mitochondrial responses, therefore short-term and long-term responses could be different.
Our study also sheds light on the relationship between senescence and anti-oxidative response. It is established that increased ROS and oxidative stress can lead to senescence and age-related disease. Aging (or senescent) cells could also have reduced antioxidative capabilities. For example, aging RPE cells have been shown to be more vulnerable to oxidative damage due to impaired to Nrf2 signaling 60 . Our results show that young Pgam5 −/− mice are more resistant to NaIO3 induced RPE damage compared to control WT mice. This phenotype correlates the upregulation of anti-oxidative genes NRF2, SOD1 and FOXO4 in short-term cultured PGAM5 −/− cells. Upregulation of NRF2 in short-term cultured RPE PGAM5 −/− cells is consistent with PGAM5's function in forming PGAM5/KEAP1/NRF2 complex for NRF2 degradation 19 . Regulation of FOXO4 by PGAM5 is also consistent with a recent finding that PGAM5 promotes lasting FoxO activation after developmental mitochondrial stress and extends lifespan in Drosophila 61 . Several groups also verified PGAM5 deficiency renders cells resistance to oxidative stress 14,39 . However, we found by surprise that the anti-oxidative resistance was lost in 18-month-old Pgam5 −/− mice, consistent with the failure to upregulate NRF2 in long-term cultured PGAM5 −/− cells and in vivo. These findings suggest that senescence may change the landscape of anti-oxidative response of the cells, and the beneficial anti-oxidative function by PGAM5 deletion is lost due to its role in senescence. Our study also exemplifies nature's balance on short-term benefit to long-term undesired effects, which draws caution to correlate short-term mechanistic studies with long-term phenotypic analyses. Further work is needed to decipher the role and mechanism of senescence in anti-oxidative response.
Our findings that PGAM5 regulates mitochondrial dynamics, senescence and age-related anti-oxidative response could have therapeutic implications. Harnessing the effect of PGAM5 deletion in anti-oxidation but preventing its undesired effect in senescence could be beneficial in age-related diseases, including Parkinson's disease and AMD.  62 . For NAIO3 injury experiments, 20 mg/kg NaIO 3 (in 80 µl) was administrated into 2-months-old or 18-months-old mice intravenously. Fundus images were captured at 5 days post injection using Micron III (Phoenix). The mice were then killed, and the eye balls were fixed in 4% PFA for histology or flatmount ZO-1 staining.

Methods
PGAM5 −/− ARPE-19 cell line generation using CRISPR/Cas9 technology. PGAM5 −/− ARPE-19 cell lines were generated according to ref. 63 . Briefly, lentiguide puro vector (addgene: #52963) 64 was used to insert guide RNA targeting human PGAM5 gene after BsmBI (Fermentas) digestion of the vector. Two different guide RNAs were designed online (https://zlab.bio/guide-design-resources) (Supplementary Table 1 Fig. 10 Proposed model for PGAM5 in senescence. In this model, PGAM5 regulates senescence through several mechanisms. First, PGAM5 is required for regulating mitochondrial dynamics. PGAM5 deletion leads to elevation of Drp1 phosphorylation at Ser-637, which tip the balance of mitochondrial dynamics toward hyperfusion and less mitochondrial turnover. As result, this results in elevated ATP and ROS production, activation of AMPK-mTOR pathway, and accelerated senescence. Second, PGAM5 deletion promotes the IRF/IFN-β pathway, which could also contribute to cell senescence. PGAM5 deletion leads to resistance to oxidative stress, possibly through mitochondrial hyperfusion and/or upregulation of anti-oxidative genes. This resistance is lost in aging possibly due to the accelerated senescence caused by PGAM5 deletion. Our model highlights a critical role for PGAM5 in regulating mitochondrial dynamics, senescence and age-related oxidative response. by puromycin (2 µg/ml) for 1 week, Ad5-CMV-Cas9 (Cat #: VVC-U of Iowa-4683), an adenovirus vector expressing Cas9 gene, was added into cells at 100:1 MOI. After 4 days, cells were examined by western blot for knockout efficiency. Two independent lines have been established with similar knockout efficiency and phenotypes.
Adenovirus and lentivirus generation. Adenoviruses were generated as described 65 . Drp1 K38A mutant was amplified from pCDNA3-Drp1 K38A mutation vectors (Gifts from Dr. Wang Wang at University of Washington) respectively by using primers: Drp1-up: 5′-AATGGTACCATGGAGGCGCTAATTCCT-3′ and Drp1-down: 5′-AACTCGAGTCACCAAAGATGAGTCTCCCGGA-3′. After digestion by KpnI and XhoI, they were subcloned into pShuttle-CVM vector (Ad-EASY system). The positive clones were cut with PmeI and transformed into E. coli with adenovirus vector for recombination. Positive clones were then cut with PacI and transfected into Ad-293 cells using Viral-Pack Transfection Kit from Stratagene. Viral titers were determined by End-Point Dilution Assay. MOI of 100 was used in the infections. For lentiviral vectors, Drp1 S637A mutant was amplified from AAV-DIO-Drp1 (S637A)-EYF mutation vector (Gift from Dr. Lobo at University of Mariland), respectively, using primers: (5′ AATGGTACCATGGAGGCGCTAATTCCT3′, and 5′AAACTCGAGTCACCAAAGATGAGTCTCCCGGA3′). Lenti-GFP or lenti-Drp1637A was transduced into control or PGAM5 −/− APRE19 cells at M.O.I. of 10. Immunostaining and western blot were performed at 48 h and 1 week after transduction respectively. ATP level was measured at one week (short-term) and eight weeks (long-term). SA-β-Gal assay was carried out at eight weeks after transduction.
For cell treatment, Cycloheximide (10 mg/ml in DMSO, −20°C), Oligomycin A (10 mM in DMSO, −20°C), Sodium Iodate (NaIO 3 , 10 mg/ml), and N-acetyl-Lcysteine (NAC) (3 mM in water, 4°C) were from Sigma-Aldrich (St. Louis, MO, USA). For in vitro NaIO 3 treatment, 5-80 mM NaIO 3 was added into ARPE-19 cells for overnight, and cell viability was evaluated by Propidium Iodide (PI) staining and neutral red assays. For in vitro CCCP treatment, 20 μM CCCP was added in to medium for indicated duration. Samples were harvested for western blot or other subsequent assays. For PI staining, 10 μg/ml of PI was added to the fresh medium and incubated for 10 min before imaging at 550 nm. Neutral red assay method was adapted from 66 . Briefly, phenol red-free DMEM (Gibco) containing 0.4 mg/ml neutral red (Sigma-Aldrich) was incubated with cells for 2 h in 37°C incubator. After incubation, cells were washed three times by PBS buffer and a solubilization solution (50% ethanol/1% acetic acid) was added. Concentration of neutral red was measured by absorbance at 540 nm. For JC-1 assay, 2 µM of JC-1 dye was added into cell culture medium. After 15 min of incubation at 37°C, medium was removed, and fresh medium was added. Then, cell images were captured under the microscopy immediately by 488 nm and 555 nm and images were analyzed by NIS-Elements software. Cellular diameter was measured using NIS-Elements software.
SA-βGal staining in cultured cells was performed similarly with previous study 67 . Briefly, after fixation with 4% PFA for 10 min, cells were washed three times by PBS buffer (pH 7.4) and then incubated with freshly prepared staining solution (5 mM Potassium Ferricyanide Crystalline (Sigma-Aldrich), 5 mM Potassium Ferricyanide Trihydrate (Sigma-Aldrich), 2 mM Magnesium Chloride (Sigma-Aldrich), 1 mg/ml X-gal (Sigma-Aldrich) in PBS buffer (pH 6.0)) for 12-24 h. After removing staining solution and washing cells with PBS buffer (pH 7.4), bright field microscopy was used to take the images.
Reactive oxygen species (ROS) was measured using DCFDA-Cellular ROS Assay Kit following the manufacture's protocol (Abcam, ab113851). Cells were counted and seeded to 96-well plates before growing to 100% confluence. After changing medium for 2 days, cells were washed by washing buffer and stained by DCFDA for 45 min at 37°C in the dark. Plates were read on a SpectraMax iD3 (Molecular Devices) plate reader at Excitation length 485 nm and emission length of 535 nm.
Mitotimer assay was used to monitor mitochondrial turnover as described 69 . pTRE-Tight-MitoTimer plasmid (addgene#50547) was transfected into ARPE-19 cells using lipofectamine 3000. After 24 h, 0.5 µg/ml doxycycline was added to induce Mitotimer expression for 4 h before changing into fresh medium. Cell was fixed after 12 h, images were captured at 488 nm and 555 nm using Nikon A1 confocal microscopy and were analyzed as described 69 . rest were incubated with 3 μg Axin1 antibody (Cell Signaling Technologist) or control rabbit IgG (Santa Cruz sc-66931) at 4°C for overnight. 30 μl of protein A and G conjugated magnetic beads (Bio-rad 1614833) were washed by lysis buffer and added into antibody-cell lysate mixture for 4 h incubation at 4°C. Magnetic rack was used to precipitate the magnetic beads, and beads were subsequently rinsed by pre-chilled lysis buffer for three times. 1× SDS sample loading buffer (Bio-rad) was used to elute the proteins. Western blot was used for the downstream assay.
Cell size measurement. For in vitro cell size quantification, cells were digested with trypsin and suspended, and images were captured by microscopy. Cell size was also quantified by flow cytometry using up to 10 5 cells (Sony SH800S Cell Sorter). Fordware Scatter (FSC) parameter along the X axis of flow cytometry result represents the relative size of cells.
For in vivo RPE cell size quantification, RPE flatmounts were stained by phoillodin or ZO-1 antibody to outline the boundaries of cells, and cell numbers were counted. Average cellular area was calculated based on the number of cells in the area.
Statistics. Graphpad Prism 6 was used to perform statistical analyses. For most experiments, two-tailed unpaired t-tests and Tukey's multiple comparisons twoway ANOVA were used to determine statistically significant differences between groups. All experiments were performed at least three times and data were presented as means ± standard deviation (s.d.) unless otherwise specified. Exact P-values were labeled in figure legends. The details of each experiment could be referred to the corresponding figure legends.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.