Adult progenitor rejuvenation with embryonic factors

Abstract During ageing, adult stem cells' regenerative properties decline, as they undergo replicative senescence and lose both their proliferative and differentiation capacities. In contrast, embryonic and foetal progenitors typically possess heightened proliferative capacities and manifest a more robust regenerative response upon injury and transplantation, despite undergoing many rounds of mitosis. How embryonic and foetal progenitors delay senescence and maintain their proliferative and differentiation capacities after numerous rounds of mitosis, remains unknown. It is also unclear if defined embryonic factors can rejuvenate adult progenitors to confer extended proliferative and differentiation capacities, without reprogramming their lineage‐specific fates or inducing oncogenic transformation. Here, we report that a minimal combination of LIN28A, TERT, and sh‐p53 (LTS), all of which are tightly regulated and play important roles during embryonic development, can delay senescence in adult muscle progenitors. LTS muscle progenitors showed an extended proliferative capacity, maintained a normal karyotype, underwent myogenesis normally, and did not manifest tumorigenesis nor aberrations in lineage differentiation, even in late passages. LTS treatment promoted self‐renewal and rescued the pro‐senescence phenotype of aged cachexia patients' muscle progenitors, and promoted their engraftment for skeletal muscle regeneration in vivo. When we examined the mechanistic basis for LIN28A's role in the LTS minimum combo, let‐7 microRNA suppression could not fully explain how LIN28A promoted muscle progenitor self‐renewal. Instead, LIN28A was promoting the translation of oxidative phosphorylation mRNAs in adult muscle progenitors to optimize mitochondrial reactive oxygen species (mtROS) and mitohormetic signalling. Optimized mtROS induced a variety of mitohormetic stress responses, including the hypoxic response for metabolic damage, the unfolded protein response for protein damage, and the p53 response for DNA damage. Perturbation of mtROS levels specifically abrogated the LIN28A‐driven hypoxic response in Hypoxia Inducible Factor‐1α (HIF1α) and glycolysis, and thus LTS progenitor self‐renewal, without affecting normal or TS progenitors. Our findings connect embryonically regulated factors to mitohormesis and progenitor rejuvenation, with implications for ageing‐related muscle degeneration.

defined embryonic factors can rejuvenate adult progenitors to confer extended proliferative and differentiation capacities, without reprogramming their lineage-specific fates or inducing oncogenic transformation. Here, we report that a minimal combination of LIN28A, TERT, and sh-p53 (LTS), all of which are tightly regulated and play important roles during embryonic development, can delay senescence in adult muscle progenitors. LTS muscle progenitors showed an extended proliferative capacity, maintained a normal karyotype, underwent myogenesis normally, and did not manifest tumorigenesis nor aberrations in lineage differentiation, even in late passages.
LTS treatment promoted self-renewal and rescued the pro-senescence phenotype of aged cachexia patients' muscle progenitors, and promoted their engraftment for skeletal muscle regeneration in vivo. When we examined the mechanistic basis for LIN28A's role in the LTS minimum combo, let-7 microRNA suppression could not fully explain how LIN28A promoted muscle progenitor self-renewal. Instead, LIN28A was promoting the translation of oxidative phosphorylation mRNAs in adult muscle progenitors to optimize mitochondrial reactive oxygen species (mtROS) and mitohormetic signalling. Optimized mtROS induced a variety of mitohormetic stress responses, including the hypoxic response for metabolic damage, the unfolded protein response for protein damage, and the p53 response for DNA damage. Perturbation of mtROS levels specifically abrogated the LIN28A-driven hypoxic response in Hypoxia Inducible Factor-1α (HIF1α) and glycolysis, and thus LTS progenitor Peng Wang and Xupeng Liu contributed equally to this study. self-renewal, without affecting normal or TS progenitors. Our findings connect embryonically regulated factors to mitohormesis and progenitor rejuvenation, with implications for ageing-related muscle degeneration.

| INTRODUCTION
The regenerative properties of muscle stem cells decline with ageing as they enter an irreversibly senescent state, thereby failing to proliferate or differentiate, with important implications for transplantation and regenerative medicine. 1 The regenerative capacity of skeletal muscles depends on muscle stem cells and muscle progenitors, which proliferate in response to tissue damage, and which either fuse and differentiate to regenerate myofibres or self-renew to restore the pool of stem cells. [2][3][4][5] In contrast to aged animals, whose prosenescent stem cells fail to proliferate and differentiate properly in response to injury, juvenile animals are able to manifest more robust regenerative responses in general. [6][7][8] The conserved relationship between juvenility and tissue regeneration was first discussed by Darwin et al., 9,10 but the precise mechanisms that underlie juvenility and rejuvenation had remained unclear.
Skeletal muscles constitute $40% of the young human body mass. Sarcopenia is the gradual decline of skeletal muscle mass and function with ageing. With ageing, muscles manifest a profound regenerative defect that contributes to elderly frailty in sarcopenia and cachexia. Both changes in the extrinsic microenvironment and stem cell-intrinsic mechanisms may contribute to this regenerative decline. 2,11,12 Recent studies have demonstrated that both the numbers and the functionality of adult muscle stem cells decline with ageing, especially after geroconversion and senescence. [13][14][15][16][17][18] In contrast, embryonic, foetal and perinatal muscle progenitors are widely known to possess extended proliferative capacities, compared with adult muscle progenitors, which in turn possess higher proliferative capacities than aged adult muscle progenitors. [19][20][21][22] This is despite the fact that embryonic and perinatal progenitors also sustained numerous rounds of mitosis during embryogenesis and foetal development. For instance, it is known that Pax3+ muscle progenitors give rise to all embryonic, foetal, and adult myoblasts and myofibres. 3,[23][24][25] While it is widely known that muscle progenitors' lifespan inexorably decline with development and ageing, the molecular principles responsible for this phenomenon have remained incompletely understood. Thus we asked, what embryonically regulated factors change with adult human muscle progenitor ageing, and what is the minimal set of embryonically regulated factors that are needed to delay the pro-senescence trend in adult human muscle progenitors?
The mammalian p53 transcription factor family comprises three members p53, p63, and p73. The transcription factors evolved from an ancestral p63/p73 gene that can be found in most invertebrates, 26,27 mediating stress responses upon DNA damage. Although the full-length isoforms p53α, TAp63α, and TAp73α all function as tumour suppressors in adulthood, 28,29 they also play important roles during embryogenesis. While p63 governs epithelial progenitor cells and epidermis development, 30,31 p73 functions in neuronal development and multiciliated cell differentiation. 32 The p53 factor is not only finely regulated in expression during embryogenesis, 33,34 but also functions in neural tube closure 35 and craniofacial skeletal, neuronal, and muscle tissue development. 36 Moreover, the dosage of p53 is also tightly regulated by the Mdm2/Mdm4 family of E3 ubiquitin ligases and the miR-125/lin-4 family of microRNAs, to facilitate normal tissue development. [37][38][39] Indeed, previous studies had shown that p53 is important in regulating many tissue progenitors' self-renewal and differentiation. [40][41][42] Two other factors that are tightly regulated during embryonic development in multiple tissue lineages, are LIN28 and TERT, both of which are RNA-binding proteins that are typically only highly expressed in embryonic progenitors in mammals. [43][44][45] Here, by examining a series of molecular markers associated with senescence in our ageing adult muscle progenitors and performing a mini-screen for embryonic factors, we found that a minimal combination of LIN28A, hTERT, and p53 shRNA (LTS) could dramatically delay the senescence of human muscle progenitors. LTS (but not TS) factors could dedifferentiate adult muscle progenitors and improve their regenerative potential in vivo. Mechanistically, we found that LIN28A collaborated with TS to optimize mitochondrial reactive oxygen species (mtROS) and stress-responsive signalling pathways, thereby inducing mitohormesis and restoring juvenility. Perturbation of mtROS levels specifically abrogated the LIN28A-driven hypoxic response in Hypoxia Inducible Factor-1α (HIF1α) and glycolysis, and thus LTS progenitor self-renewal, without affecting normal or TS progenitors. Our findings connect embryonic factors to mitohormesis during the process of muscle progenitor self-renewal and rejuvenation, with implications for ageing-related muscle degeneration in cachexia and sarcopenia.

| LTS factors delay adult progenitor senescence without transformation while permitting normal differentiation
To address this question, we used young and old primary adult human skeletal muscle (HSKM) progenitors to screen for a variety of embryonically regulated factors that are not lineage-specific, to biomimic the foetal growth phase. We defined 'young' as <5 population doublings, and 'old' as more than 20 population doublings, based on prior experience with replicative senescence. 46 First, we found that old progenitors, compared with progenitors, showed significantly higher levels of cell cycle inhibitors such as p21 WAF1 , p27 KIP1 , and p16 INK4a ( Figure 1A). Furthermore, the anti-proliferative family of let-7 microRNAs, especially F I G U R E 1 The effects of LIN28A, TERT, and sh-p53 (LTS) factors on adult human muscle progenitor senescence and differentiation. (A) Quantitative Reverse Transcription-Polymerase Chain Reaction (RT-PCR) for mRNAs of cell cycle inhibitors in aged adult human skeletal muscle (HSKM) myoblasts, relative to young adult HSKM myoblasts (N = 3 wells of cells for each group). (B) Quantitative RT-PCR for let-7 microRNAs in aged adult HSKM myoblasts(red columns), relative to young adult HSKM myoblasts (black columns; N = 3 wells of cells for each group). (C) Quantitative RT-PCR for telomere length in aged adult HSKM myoblasts, relative to young adult HSKM myoblasts (N = 3 wells of cells for each group). (D) Population doubling curves for young HSKM myoblasts (black), and young adult HSKM myoblasts transduced with viral sh-p53 (S, pink), and LIN28A (LS, green), or hTERT (TS, purple), or LIN28A and hTERT (LTS, red). While young adult HSKM and other transgenic myoblasts started to undergo senescence by 60 days, before the 30th population doubling, the LTS myoblasts (red) continued to proliferate steadily beyond 120 days and beyond the 90th population doubling. (E) Brightfield micrographs of senescence-associated β-galactosidase positive (SA-β-gal+) cells in 100-day-old adult HSKM myoblasts and 100-day-old LTS myoblasts. Quantification of senescence-associated β-galactosidase positive (SA-β-gal+) cells in 100-day-old adult HSKM myoblasts, relative to 100-dayold LTS myoblasts (N = 3 wells of cells for each group). (F) Quantitative RT-PCR for p27 KIP1 , p21 CIP1 , p16 INK4A , and let-7 microRNAs in LTS myoblasts, relative to young adult HSKM myoblasts (N = 3 wells of cells for each group). (G) Immunofluorescence staining for the myotube protein marker myosin heavy chain in young HSKM and LTS myoblasts, relative to old HSKM myoblasts and immortalized (hTERT-CyclinD1-CDK4 R24C ) myoblasts, that were cultured in myogenic differentiation media. Cells were counterstained with 4'6-Diamidino-2-Phenylindole to visualize the myonuclei. Scale bars 100 μm. *p < 0.05, **p < 0.01, ***p < 0.001; ns, not significant.
let-7b and let-7g, also accumulated to higher levels in old progenitors, compared with young progenitors ( Figure 1B), consistent with previous studies of primary human muscle samples. 47 In contrast, the telomeres of old progenitors were significantly shorter than young progenitors ( Figure 1C). Although the mRNA expression levels of p53 and p14 ARF trended towards downregulation, we found that p53 protein was accumulating to higher levels in old progenitors ( Figures S1 and S2). In contrast, we had found that the p53 inhibitor Mdm4 and the telomerase complex component Tep1 were amongst the most highly upregulated genes in mouse Lin28a + MuSCs (Preprint). 48 These preliminary results spurred us to test some embryonically regulated factors that are not lineage specific, to attempt to prevent or delay the pro-senescence trend.
We tested LIN28A to regulate the accumulating let-7 miRNAs and other developmental genes, hTERT to lengthen and restore the telomeres, and short hairpin RNAs (shRNAs) against p53 (which also transactivates p21 WAF1 ), p16 INK4a , p14 ARF , or Rb, all of which are tightly regulated in embryogenesis and accumulate with senescence. 40,42,[49][50][51][52] In contrast, we omitted screening any mutant oncogenes which overcome existing senescence, instead of delay senescence. Oncogenes which overcome (instead of delay) senescence might cause irreversible transformation, aberrant differentiation of progenitors, 46 and thus tumour growth. We also omitted screening pluripotency factors, such as OCT4, SOX2, NANOG, and so forth, because they are lineage-specific to pluripotent stem cells, and we found that none of the pluripotency markers were increased in mouse Lin28a + muscle stem cells (Preprint). 48 Our results showed that young progenitors quickly undergo replicative senescence, plateauing sigmoidally at $30 population doublings, and no single factor alone could prevent the senescence ( Figures 1D and S3). However, when LIN28A was combined with hTERT and a shRNA against TP53 (LTS; Figure S4), human muscle progenitors could self-renew and proliferate beyond 90 population doublings, with a linear population doubling curve up till $206 population doublings. Furthermore, each factor was necessary in the LTS combo, as lacking any one of the three factors resulted in either senescence ( Figure 1D) or apoptosis (data not shown).
We then tested if the LTS progenitors could still differentiate properly. We expect oncogene-transformed progenitors to be unable to differentiate properly, since progenitors need to completely withdraw from the cell cycle and properly activate epigenetic remodelling before they can undergo terminal differentiation. Surprisingly, we found that the rejuvenated 100-day-old LTS progenitors could still robustly differentiate into myosin heavy chain (MHC) + , α-actinin + , myogenin + multinucleated myotubes, just like young HSKM progenitors ( Figures 1G and S7).
In contrast, the immortalized progenitors harbouring oncogenic CDK4 R24C , cyclin D1, and hTERT failed to differentiate properly, and only expressed very weak levels of MHC, with few multinucleate myotubes ( Figure 1G). Similarly, the 100-day-old senescent progenitors could not differentiate into multinucleated myotubes ( Figure 1G).
We also tested if LTS progenitors were reprogrammed into primitive mesenchymal or mesodermal progenitors by subjecting them to adipogenic, osteogenic, and chondrogenic differentiation conditions. We did not detect significant adipogenesis, osteogenesis, or chondrogenesis (data not shown). We also profiled LTS cells for their mRNA Secondly, we tested if the LTS factors would lead to genomic instability, given the partial inhibition of TP53 by RNAi. When we subjected 100-day-old LTS progenitors to chromosomal analysis, the highly passaged cells still displayed a normal diploid karyotype, similar to highly passaged ES cells ( Figure S8). In contrast, 100-day-old TS progenitors displayed a variety of aneuploid karyotypes, including the loss of numerous chromosomes, and the appearance of dicentric, ring, and marker chromosomes ( Figure S8). We also performed immunofluorescence staining for the expression of a series of DNA damage markers in LTS progenitors, and found that the expression of PARP-1, XRCC1, γ-H2A.X, and 53BP1 in LTS progenitors was relatively lower than in TS progenitors ( Figure S9A). These results indicate that LTS progenitors can properly activate a normal DNA repair and selfrenewal programme.
Luciferase-labelled LTS progenitors gradually declined and disappeared within 2 weeks of subcutaneous transplantation ( Figure S9B).
In comparison, all other transformed cancer cell lines form tumours upon subcutaneous transplantation in NSG mice. These results suggested that LIN28A could ensure proper self-renewal with normal differentiation potential and genomic stability, and thus prevent tumorigenicity.

| LTS factors can rejuvenate aged muscle progenitors for transplantation
To test the utility of the LTS factors in aged muscle progenitors of elderly patients, we transduced aged muscle progenitors from elderly cachexia patients. We previously found that cachexia can lead to suppression of myocyte growth. 53 When we cultured aged patients' HSKM progenitors, we also found that they rapidly underwent senescence within 6 population doublings (Figures 2A,B), which is significantly less than healthy young adult muscle progenitors ($30 population doublings, Figure 1B). However, upon treatment with the LTS factors (Figure 2A  To confirm this observation, we performed qRT-PCR for a variety of myogenic markers to assess the differentiation status of young adult HSKM and LTS progenitors. Our results showed that LTS progenitors were significantly higher in expression of PAX3, an embryonic myogenic transcription factor, than in HSKM progenitors ( Figure 3B). While the myoblast commitment factor MYOD was higher ( Figure 3C), the myogenic determination factor MYF5 was lower ( Figure 3C), and the myogenic differentiation factor myogenin MYOG remained similarly low ( Figure 3C). MYOD and MYF5 are homologous in sequence and cobelong to the MRF family, thus they have functionally overlapping roles in promoting myoblast specification. 56 Because in LTS muscle progenitors, both MYOD and the muscle stem cell factor PAX3 are already highly expressed in LTS, it is possible that MYF5 was slightly decreased in expression as a compensatory response. Most other myogenic differentiation markers were even lower in LTS progenitors than in HSKM progenitors, including skeletal muscle actin A1 (ACTA1; Figure 3C), and several MHC isoforms ( Figure 3D). These results indicated again that LTS myoblasts were dedifferentiated. In contrast, in the LTS myotubes, the expression of MYF5 and MYOD, as well as MYOG and MYH3 are all slightly enhanced, whereas some myogenic markers are slightly lower.
Overall, we believe these differences are too slight to be biologically significant, and the myogenic differentiation capacity of LTS cells can be considered similar to HSKM cells ( Figure S7).
We also performed western blots for PAX3 and TWIST2 protein expression in LTS, TS and HSKM cells. LTS progenitors showed the  Figure S2. *p < 0.05, ***p < 0.001; ns, not significant. highest PAX3 protein levels, compared with HSKM and TS progenitors ( Figures 3E and S2). All the progenitors showed significant declines in PAX3 and TWIST2 protein levels, as they differentiated into myotubes ( Figures 3E and S2), and as expected of myogenesis. 57 The increased PAX3 protein, an embryonic myogenic transcription factor, supports the transcriptomic observations on LTS progenitors.

| LIN28A aids rejuvenation via let-7independent effects on mRNA expression
To dissect the mechanism for LIN28A's rejuvenating effect, we returned to primary human progenitors cultured ex vivo. Our RNA profiling studies earlier had shown that let-7 family members accumulated to higher levels in ageing progenitors ( Figure 1B). Both TS and LTS transduction lowered the let-7 levels, with LIN28A further suppressing let-7b, let-7e, and let-7g levels ( Figure 4A). To test if LIN28A's rejuvenation effect in muscle progenitors was mediated by let-7, we transfected mature let-7 mimics in an attempt to abrogate transgenic LIN28A's effect on proliferation ( Figure 4A). However, we found that neither let-7a, nor let-7b, nor both, had any effect on LTS progenitors' proliferation rate ( Figure 4B).
Moreover, let-7 overexpression also failed to increase the numbers of senescent SA-β-gal + cells ( Figure 4C), even though important let-7 targets such as IMP1/2/3 and HMGA2 45 had been almost completely depleted by the let-7 overexpression (Figures 4D and S2). These results suggest that let-7 repression alone cannot fully explain LIN28A's effect on muscle progenitor self-renewal.
Besides regulating the let-7 microRNAs, Lin28a also binds and regulates mRNAs. Thus, we analysed the transcriptomic profiles of LTS progenitors, and compared them against TS progenitors by gene set enrichment analysis (GSEA). The top downregulated signatures in LTS progenitors were almost completely dominated by response targets ( Figure 4E). This is consistent with findings that an interferon-driven innate immune or inflammageing response is activated during senescence, 58 and further corroborated the senescence of TS progenitors, which the addition of LIN28A prevented. Our analysis further revealed that LIN28A's upregulated signatures were primarily metabolic in nature, consisting of mitochondrial oxidative phosphorylation (OxPhos) genes ( Figure 4F,G), stress-responsive signatures such as the unfolded protein response (UPR) mediated by XBP1s activity ( Figure 4H,I), and the hypoxic stress response and glycolysis targets mediated by HIF1α ( Figure 4J,K). Interestingly, the HIF1α-mediated hypoxic stress response was also observed in mouse Lin28a + MuSCs (Preprint), 48 indicating that Lin28a's induction of HIF1α is a conserved mechanism in both adult human muscle progenitors and mouse MuSCs.

| LIN28A optimizes OxPhos and mtROS to induce HIF1α-glycolysis
To validate that OxPhos and glycolysis were indeed upregulated, we used the Seahorse extracellular flux analyser. Our analysis of oxygen consumption rates in the LTS progenitors revealed that LTS progenitors do indeed show significantly higher basal and maximal OxPhos rates ( Figure 5A,B). Moreover, LTS progenitors also showed higher glycolysis rates ( Figure 5C). We also found that the mitochondrial membrane potential Δψ m in LTS progenitors was significantly increased ( Figure 5D). In contrast, mitochondrial DNA and protein biogenesis did not increase, indicating that while the mitochondrial OxPhos activity levels were increased, total mitochondrial biogenesis did not increase (Figures S2D and S10). Since LIN28A has already been shown to bind OxPhos mRNAs to regulate OxPhos protein expression, [59][60][61] it is plausible that LIN28A-induced OxPhos would increase mtROS via an increased ETC flux and increased Δψ m . Previous studies had shown that a high mitochondrial membrane potential Δψ m can shift the mtROS-producing sites to a more reduced state, thereby increasing the propensity for mtROS production. 62,63 To validate this hypothesis, we used the mitochondrial Grx1-roGFP2 reporter 64 Figure 5I). Using a HIF1α/hypoxiaresponse-element luciferase reporter, 67 we further validated that LIN28A does induce a HIF1α-mediated hypoxic response in LTS progenitors relative to TS and HSKM progenitors, and that this stress response was dependent on ROS levels ( Figure 5J).
When we subjected the progenitors to pro-oxidant and antioxidant treatment, we found that both higher ROS and lower ROS could suppress the HIF1α targets ( Figure 5K) and glycolysis genes ( Figure 5L) specifically in LTS progenitors, but not in TS progenitors.
This indicates that LIN28A was optimizing the mtROS level to upregulate HIF1α targets and glycolysis genes in LTS progenitors. In contrast, both H 2 O 2 and DTT treatments failed to perturb the mildly higher OxPhos mRNA levels in LTS progenitors ( Figure 5M), indicating that OxPhos mRNA regulation laid upstream of mtROS regulation. These results support a model whereby LIN28A à optimal OxPhos and mtROS à HIF1α-driven hypoxic and glycolysis stress responses. We also used another antioxidant N-acetyl-L-cysteine (NAC), to examine the above model. NAC treatment led to significant down-regulation of some HIF1α targets and glycolysis genes specifically in LTS progenitors, more so than in TS progenitors ( Figure S11A,B). NAC treatment also decreased the HIF1α/hypoxia-response-element luciferase reporter activity more drastically in LTS progenitors, than in TS or HSKM progenitors ( Figure S11C), consistent with the model.

| LIN28A-mtROS-HIF1α promotes selfrenewal
However, it remained unclear if these molecular changes were functionally responsible for LIN28A's effects on muscle progenitor selfrenewal. We tested and found that both pro-oxidant and antioxidant treatments led to disproportionately more suppression of LTS progenitor proliferation, than of TS or HSKM progenitor proliferation ( Figure 6A), supporting the model that mtROS optimization to promote proliferation is a mechanism specific to LIN28A + progenitors. Consistent with the model above, we also found that treatment with the HIF1α inhibitor (KC7F2) led to disproportionately more suppression of LTS progenitor proliferation, than of TS or HSKM progenitor proliferation ( Figure 6B).
Live-dead staining of the cells revealed that in control vehicletreated cells, TS progenitors had more cell death than normal HSKM progenitors at baseline ( Figure 6C), indicating that TS increased both proliferation and cell death, but LIN28A in LTS progenitors suppressed the increased cell death. Addition of the HIF1α inhibitor increased the proportion of dead cells similarly across HSKM, TS and LTS progenitors ( Figure 6C), suggesting that LIN28A-mediated suppression of cell death was over-ridden by the HIF1α inhibitor, and that Lin28a promoted self-renewal via HIF1α. The glycolysis inhibitor 2-deoxyglucose (2DG) led to mild but significant decreases in all progenitors' proliferation, but the decrease was most pronounced in LTS progenitors ( Figure 6D), again supporting the model that LIN28A is driving HIF1α and glycolysis to promote self-renewal.
To further confirm the role of mtROS in this mechanistic model, we used genetic perturbation of mtROS-specific enzymes to examine the necessity of mtROS to LIN28A's mechanism of action in cell proliferation. Targeted overexpression of mitochondrial superoxide dismutase 2 and mitochondrial catalase led to a significant decrease in mtROS ( Figure 6E), more so in LTS progenitors. This decrease in mtROS translated to disproportionately lower cell viability in LTS progenitors than in TS or HSKM progenitors ( Figure 6F). Treatment with NAC also further confirmed that the proliferation and survival of LTS progenitors were specifically more sensitive to antioxidants than TS or HSKM progenitors ( Figure 6G,H).
Taken together, these results demonstrate that LIN28A specifically optimized mtROS-HIF1α signalling and various stress response/repair pathways ( Figure 6I) to promote the self-renewal of human muscle progenitors. These results are consistent with the stress-responsive signatures that we uncovered in the Lin28a + MuSCs in vivo (Preprint), 48 and suggest that LIN28A activates molecular repair pathways in an evolutionarily conserved manner to maintain progenitors in a juvenile state. Strikingly, the mtROS induced a hypoxic stress response, glycolytic response, p53 response, and UPR to rejuvenate muscle progenitors ( Figure 6I). In other contexts, such rejuvenative mtROS-induced stress responses are collectively termed as mitohormesis. 70

| DISCUSSION
We have shown that pre-senescent but old human muscle progenitors accumulate let-7 microRNAs, p53, p21 WAF1 , p16 INK4a , and show telomere loss during ageing. A mini-screen with embryonic factors revealed that a minimal set of LIN28A, telomerase, and shRNAmediated partial knockdown of p53 (LTS) was sufficient to significantly extend the lifespan of adult human muscle progenitors. We demonstrate that these rejuvenated muscle progenitors are endowed with extended self-renewal and myogenic differentiation capacities, without oncogenic transformation into cancer cells. In fact, when aged and dysfunctional muscle progenitors from cancer cachexia patients were treated with LTS factors, their intrinsic self-renewal and myogenic differentiation capacities were restored, and they could now contribute to the regeneration of muscles in vivo. Mechanistically, we found that let-7 microRNAs alone and their effects on IMPs, HMGA2, and associated IGF-PI3K-mTOR signalling 71 Few studies have investigated the role of LIN28A in human progenitors. Our and others' previous studies have shown that transgenic LIN28A can enhance foetal-like morphogenesis and tissue regeneration in multiple mouse tissues, including the limb digits and skin, 60 kidneys, 73,74 brain, 75,76 and retina. 77 But, few studies have investigated its role in human cells, particularly in the skeletal muscles. Moreover, while previous studies have suggested Lin28a's role in regulating both glycolysis and OxPhos, 61,72,78,79 none have uncovered its importance in optimizing mtROS and mitohormesis to promote self-renewal hitherto.
Our mechanistic studies suggest that LIN28A fine-tunes mtROS to optimize mitohormesis, and that both excessive upregulation or downregulation of ROS can disrupt LIN28A's beneficial effects.
Mitohormesis is defined as the coordinated compensatory response to mild mitochondrial stress that rapidly activates nucleocytoplasmic signalling pathways, and which ultimately alters gene expression to protect the cell against stressful perturbations and molecular damage. 70 While the TS combo had little effect on mtROS and ultimately failed to enhance self-renewal of muscle progenitors, the LTS combo upregulated mtROS to an optimal level and successfully delayed muscle progenitor senescence. These results support the emerging notion that, while excessive mtROS can cause a variety of pro-ageing phenotypes in many contexts, mild mtROS can be beneficial at the right amounts in the right cell types. 63 Interestingly recent studies have also shown that mouse embryonic myogenesis requires low but not excessive amounts of mtROS, tightly regulated by Pitx2/3. 80,81 Thus, our study validated the notion that mtROS and progenitor lifespan share a non-linear relationship, and provided a genetic combo that optimizes mtROS to extend primary human progenitors' self-renewal.
Our gene expression analyses showed that, in the context of LTS treatment, optimal levels of mtROS can robustly stimulate the HIF1αmediated hypoxic stress response against metabolic damage, XBP1Smediated UPR against protein damage, and p53-mediated response against DNA damage (DDR, DNA Damage Response), all of which are already known to promote stem cell survival. It is interesting, but not surprising, that in the context of partial p53 knockdown, LIN28A-mediated mtROS would restore some p53-DDR activity. 63 Previous studies had shown that the p53 gene network is tightly regulated and fine-tuned through paradoxical signalling in progenitor cells. 37 LIN28Amediated restoration of some p53 activity might be critical for ensuring proper muscle progenitor self-renewal and differentiation, by preventing chromosomal instability and oncogenic transformation at the same time. Indeed, previous studies had shown that p53 is important in regulating myogenesis and many other lineage progenitors' proper differentiation. [40][41][42] Besides the p53-mediated DDR, the XBP1S-mediated UPR proteotoxic stress response was also activated by LIN28A-mediated mtROS. ROS-mediated protein oxidation has long been known to cause proteotoxicity, and thus activate the UPR proteotoxic stress response. 82 In fact, recent studies have shown that the UPR is needed for proper functioning of DNA repair enzymes, including the base excision repair pathway, to remove 8-oxoguanine 83,84 . Thus the UPR also cross-talks with the DDR to pre-emptively protect progenitors against genotoxic stress. Another pathway stimulated by UPR is autophagy, 85 which was previously shown to be important in rescuing geriatric mouse muscle stem cells from senescence. 86 Another mitohormetic pathway that cross-talks with the p53-mediated DNA damage response is the HIF1αmediated hypoxic stress response. 68 Multiple previous studies had already shown that ROS activates HIF1α to drive the hypoxic stress response. 63,68 Moreover, it is well-known that HIF1α can transactivate glycolysis genes. 68 Studies have shown that the hypoxic stress response is important for mammalian regeneration and the biology of naked mole rats, 87,88 which are remarkably resistant to cancer and ageing for their size, suggesting that the hypoxic stress response could be an important arm of mitohormesis in mammalian rejuvenation.
Interestingly, ROS can also directly stimulate glycolytic flux, especially into the pentose phosphate pathway (PPP 89 ). Glycolysis intermediates such as fructose-6-phosphate can be diverted into the PPP to fuel NADPH synthesis for reducing oxidative stress, and nucleotide synthesis for stem cell proliferation. Other glycolytic intermediates like 3-phosphoglycerate and pyruvate can be diverted into amino acid synthesis, whereas dihydroxyacetone-phosphate and acetyl-CoA can be diverted into lipid synthesis for anabolic growth. 90   glutathione in the context of an oxidative stress response, which is also activated by the UPR (Ron and Walter, 2007). This is attested by the rise in the total pool of oxidized and reduced glutathione in LTS progenitors, compared with HSKM progenitors. Both NAD+ and NADH were also upregulated by LTS, but there were no significant changes in the NAD+/NADH ratio, suggesting that while changes in NAD+ and Sirt1 activity are both crucial for regulating muscle stem cell self-renewal and muscle regeneration, 61,91 NAD+ and Sirt1 are less likely to be responsible for LTS' effects in this context. Nevertheless, the mutually reinforcing nature of these protective metabolic pathways suggests that mitohormesis is a conserved and well-coordinated programme of stress-responsive pathways to control progenitor lifespan.
These mitohormetic stress response mechanisms might also shed more light on why LIN28A alone is insufficient to induce oncogenic transformation or induced pluripotent stem cell reprogramming on its own, but rather facilitates progenitor self-renewal instead. It should also be noted that both Lin28 KO and Tert KO cells can still undergo tumour initiation through other pathways, that is, they are not as critical for cancer as they are critically needed for normal development. 45,92 Prior research has demonstrated that telomere shortening or telomerase inhibition is more toxic to normal stem cells, than cancer cells, 92,93 which has led to the failure of such approaches clinically. In fact, TERT deficiency promotes (not prevent) malignant transformation by accelerating senescence and SASP and chromosomal instability, whereas a transient upregulation of TERT delays senescence and prevents transformation. 94,95 It is also interesting that LIN28A induced some level of dedifferentiation to produce PAX3 high skeletal muscle progenitors.

| Cachexia patient cell lines
Aged primary HSKM progenitors were derived from the rectus abdominus of two aged patient donors with cachexia (age/weight:

| Quantitative PCR
RNA was extracted by TRIzol (Thermo Fisher) and reverse transcribed with Superscript III (Thermo Fisher) according to manufacturer's instructions. The resulting cDNA was diluted 5Â before performing qPCR with KAPA SYBR FAST on ABI Prism 7900HT (Applied Biosystems) according to manufacturers' instructions.

| Cytogenetics
Cells were seeded in gelatin-coated 6-well plates and cultured up to a maximum of 50% confluency to avoid myoblast fusion. Cells were first treated with colcemid and 5-Bromo-2-DeoxyUridine overnight before harvesting with EDTA. A fixative solution (1:3 = glacial acetic acid: methanol) was used to fix pelleted cells prior to slide preparation, Giemsa banding, and mounting. Twenty metaphase spreads were prepared for each cell line for detailed analyses and karyotyping.

| Statistics and reproducibility
All statistical analyses were performed using GraphPad Prism 6 (GraphPad Software). Data are presented as mean ± SEM. Differences between groups were tested for statistical significance by using the two-sample t-test. A p < 0.05 was considered significant. The number of biological (non-technical) replicates for each experiment is indicated in the legends.