Autophagy‐Sirt3 axis decelerates hematopoietic aging

ABSTRACT Autophagy suppresses mitochondrial metabolism to preserve hematopoietic stem cells (HSCs) in mice. However, the mechanism by which autophagy regulates hematopoietic aging, in particular in humans, has largely been unexplored. Here, we demonstrate that reduction of autophagy in both hematopoietic cells and their stem cells is associated with aged hematopoiesis in human population. Mechanistically, autophagy delays hematopoietic aging by activating the downstream expression of Sirt3, a key mitochondrial protein capable of rejuvenating blood. Sirt3 is the most abundant Sirtuin family member in HSC‐enriched population, though it declines as the capacity for autophagy deteriorates with aging. Activation of autophagy upregulates Sirt3 in wild‐type mice, whereas in autophagy‐defective mice, Sirt3 expression is crippled in the entire hematopoietic hierarchy, but forced expression of Sirt3 in HSC‐enriched cells reduces oxidative stress and prevents accelerated hematopoietic aging from autophagy defect. Importantly, the upregulation of Sirt3 by manipulation of autophagy is validated in human HSC‐enriched cells. Thus, our results identify an autophagy‐Sirt3 axis in regulating hematopoietic aging and suggest a possible interventional solution to human blood rejuvenation via activation of the axis.

Hematopoietic aging is often associated with a clonal shift in the HSC pool and subsequent clonal hematopoiesis caused by cell-intrinsic and cell-extrinsic factors (Chung & Park, 2017). Recent study shows that in mice, inflammation promotes HSC aging via IL27Ra (He et al., 2020). However, the cellular and molecular basis responsible for hematopoietic aging, in particular in humans, remains largely unclear.
Autophagy is a highly conserved metabolic process that is critically essential to maintain cellular homeostasis under normal and stress conditions, as its activation promotes cellular survival by limiting oxidative stress, maintaining adequate metabolic functions, and controlling bioenergetic levels and amino acid pools. In mammalian species, autophagic machinery is controlled by several autophagy-essential protein complexes, such as the microtubule-associated protein light chain 3 (LC3) conjugation system. It includes LC3 proteins and γ-aminobutyric acid receptor-associated proteins (GABARAPs), and the ATG12 conjugation system that includes the ATG12-ATG5-ATG16L1 complex. ATG7 is a critical component in maintaining the function of the above two conjugation systems, and loss of ATG7 disables the formation of double-membrane autophagosomes, thus blocking the initiation of autophagy (Mizushima, Levine, Cuervo, & Klionsky, 2008).
Since aging results from the progressive accumulation of cellular damage by chronic insults, autophagy, as a sensor of multiple stresses, has been linked to aging. Previous studies using loss-offunction mutations in autophagy genes in animal models such as yeast, C. elegans, flies, and mice have shown that alteration to autophagy reduces their life span, whereas activation of autophagy prolongs life span in yeast and mice (Fernandez et al., 2018;Hansen, Rubinsztein, & Walker, 2018;Rubinsztein, Marino, & Kroemer, 2011).
A recent study in mouse models shows that autophagy is critical for protecting HSCs from metabolic stress by clearing active, healthy mitochondria to maintain their quiescence and potency through the mitophagy pathway (Ho et al., 2017). Loss of autophagy in HSCs causes accumulation of mitochondria and a constitutively activated metabolic state, which drives accelerated myeloid differentiation and impairs the capacity of HSCs for self-renewal (Ho et al., 2017).
Sirtuins are nicotinamide adenine dinucleotide (NAD + )dependent protein deacetylases (Katsyuba, Romani, Hofer, & Auwerx, 2020). Several members of the Sirtuin family have been shown to regulate aging or life span in numerous lower organisms including yeast, nematodes, and fruit flies (Burnett et al., 2011;Haigis & Guarente, 2006), as well as higher organisms such as mice (Kanfi et al., 2012). In the process of aging, Sirtuins have been reported to be related to, but not limited to their roles in regulation of energy metabolism, response to calorie restriction, and control of cell death (Chang & Guarente, 2014;Guarente, 2013). Most of the Sirtuin family members have been documented in the regulation of mammalian hematopoietic function, in which Sirt2, Sirt3, and Sirt7 were found reduced transcription in aging HSCs (Chambers & Goodell, 2007), Sirt1 controls HSC homeostasis via the longevity transcription factor FOXO3 (Rimmele et al., 2014), Sirt6 controls HSC homeostasis through epigenetic regulation of Wnt signaling ), and Sirt7 is implicated in HSC aging via mitochondrial unfolded protein response that ensures proteostasis in the mitochondria (Mohrin et al., 2015;Mohrin, Widjaja, Liu, Luo, & Chen, 2018). In particular, Sirt3 reverses aging-associated HSC degeneration by regulating the global acetylation landscape of mitochondrial proteins and effectively reducing oxidative stress (Brown et al., 2013). Sirt3 is indispensable in aged mice, and its expression is inhibited in aging (Brown et al., 2013). However, the regulatory means by which Sirt3 is suppressed during aging remains unknown.
Although autophagy and Sirt3 have been implicated in hematopoietic aging, the mechanisms underlying their roles have just been begun to explore in model mammals. In particular, the functional interconnection between autophagy and Sirt3 in human hematopoietic aging and the solution to rejuvenating human blood remain fundamentally unexplored. In this study, we show that autophagy cooperates with Sirt3 by promoting its expression to decelerate hematopoietic aging and that positive intervention to autophagy-Sirt3 axis leads to blood rejuvenation.

| Hematopoietic autophagy activity is reduced in aged human population
Human peripheral blood analysis of a population of 4250 individuals ranging in age from 20 to over 90 years showed a progressive decrease in hemoglobin levels, as well as red blood cell (RBC) and lymphocyte counts with increase of age (Figure 1a), which are hallmarks for hematopoietic aging (Chung & Park, 2017;Latchney & Calvi, 2017;Pang et al., 2011). Recent studies in several model organisms demonstrate that autophagy declines with aging (Burnett et al., 2011;Chung & Park, 2017). While the age-related decline in autophagy has been documented in mammalian animal cells, direct association between autophagy and hematopoietic aging in humans has not been studied.
Fluorescently tagged or endogenous LC3/GABARAP family proteins are commonly used markers for autophagy in higher eukaryotes to facilitate microscopic visualization of phagophores, autophagosomes, or autolysosomes in the cell. Use of a second marker, and depending on what type of the marker, can reveal initiation of autophagy or functional autophagy when autophagic targets are degraded in the autolysosomes (Mizushima et al., 2008). To search for direct evidence linking autophagy to blood aging in human subjects who were not in medical intervention, we examined the physiological activity of autophagy in primary human blood cells from two age groups.
Expression of genes encoding autophagy machinery was highly reduced in the HSC-enriched hematopoietic cells in the aged group (over 70 years old) compared with the young group (below 40 years old) (Figure 1b). To further address the impact of functional autophagy on human blood aging, we applied image flow cytometry to observe co-localization of LC3 and the lysosomal marker Lamp1 (Phadwal et al., 2012) for measuring activity of hematopoietic autophagy during autolysosome formation in the two differently aged populations.
The results show that in comparison with the young group, there was a significant decrease in basal autophagy levels in both total bone F I G U R E 1 Reduction of autophagy is associated with aged hematopoiesis in human population. (a) Linear regression and Pearson correlation coefficients of peripheral blood counts in aging human population. A pool of peripheral blood count information from physical examination of 4250 people aged from 20 to 90 years was analyzed using SPSS statistic software. (b) Quantitative PCR measuring the expression of autophagy-essential genes in human bone marrow primary HSC-enriched hematopoietic cells (CD45CD34) from the indicated two age groups. (c, d) Quantitative ImageStream detection of basal autophagy levels in human bone marrow primary hematopoietic cells (CD45), HSC-enriched hematopoietic cells (CD45CD34) from the two age groups. Left, statistical data from individual human samples. Right, representative images of the cells, either single-stained (CD45-blue; CD34-purple; LC3-green; Lamp1-red) or stained for both markers (merge of LC3 and Lamp1). Bar, 10 μm marrow hematopoietic cells (CD45-positive) and HSC-enriched hematopoietic cells (CD45CD34-positive) in the aged human population (Figure 1c,d). Therefore, autophagy is significantly reduced in hematopoietic cells and their stem cells in the aged population of humans.

| Loss of Atg7 in the hematopoietic system causes accelerated blood aging
To study the role of autophagy in hematopoietic aging, we used an Atg7 f/f ;Vav-iCre mouse model in which the Atg7, an essential autophagy gene, is deleted (Atg7 −/− ) primarily in hematopoietic cells and their stem and progenitor cells (Figure 2a). Deletion of Atg7 blocks the formation of both ATG8 and ATG12 systems, completely disabling canonical autophagy Mizushima et al., 2008;Nishida et al., 2009). At 10 weeks old, mice carrying the Atg7 deletion displayed a significant decrease in their RBC and lymphocyte counts, as well as in their level of HGB, comparable to that of 90-week-old mice ( Figure 2b). Atg7 deletion in the mouse model also accelerated biased myeloid-lymphoid differentiation (Figure 2c), which is the most recognized hallmark for hematopoietic aging (Chung & Park, 2017;Elias et al., 2017;Latchney & Calvi, 2017;Pang et al., 2011).
Protein homeostasis, mitochondrial accumulation, and generation of ROS and DNA damage are also hallmarks of cellular aging (Martin-Pardillos et al., 2017;Pilzecker et al., 2017). Examination by flow cytometry showed that all of these parameters were increased in the Atg7-deleted mouse HSC-enriched population (Lin − c-Kit + Sca-1 + , LSK) ( Figure 2d). Telomere length is another important indicator for cell aging (Calado & Dumitriu, 2013). Telomere length is largely maintained by telomerase reverse transcriptases that catalyze the addition of bases to the end of the telomere (Lansdorp, 2017). We measured the expression of three telomerase genes, namely Tert, Terf1, and Terf2 in the HSC-enriched cells by quantitative reverse transcription-PCR, and found that expression of these telomerase genes was reduced in the hematopoietic Atg7-deleted 10-week-old mice, comparable to their expression in the wild-type mice at 90 weeks old ( Comparison of peripheral blood counts between autophagy-defective mice (Atg7 −/− ) and wild-type (young or aged) mice (Atg7 +/+ ). (c) Flow cytometric analysis of myeloidbiased hematopoietic differentiation in autophagy-defective mice and wild-type mice (young or aged). (d) Flow cytometric detection of metabolic stress levels in the HSC-enriched cells of wild-type mice (young or aged) and Atg7-deleted mice. Total cellular ROS was measured. (e) Quantitative PCR measurement of telomerase expression in the HSPCs from autophagy-defective mice and wild-type mice (young or aged). Primer information is listed in Table S2. (f, g) Flow cytometric analysis on the cell cycle of the HSC-enriched cells from wild-type and Atg7-deleted mice. Left, representative flow density plot. Right, statistic data showing distribution of the percentage of HSC-enriched population in their cell cycle aging (Garcia-Prat et al., 2016). Taken together, these results demonstrate that Atg7 deletion causes greatly accelerated hematopoietic aging, particularly in HSC-enriched population.

| Loss of Atg7 selectively suppresses Sirt3 in mouse hematopoietic system
To explore the mechanistic connection between Atg7 and hematopoietic aging, we performed RNA sequencing of HSC-enriched population from wild-type and Atg7-deleted mice. The volcano map of differentially expressed genes between Atg7 +/+ and Atg7 −/− HSC-enriched cells of the mice revealed a total of 1062 significantly upregulated genes, with 789 downregulated genes in the Atg7depleted HSC-enriched cells (Figure 3a). Sirtuins have long been recognized as important regulators of aging (Burnett et al., 2011;Haigis & Guarente, 2006;Kanfi et al., 2012). The loss of Atg7 thus causes a significant decrease in Sirt3 expression by 10 weeks of age. In accordance with the reduction of Sirt3, RNA sequencing of the HSC-enriched hematopoietic cells from 10-week-old mice discloses a decreased expression of an array of anti-aging genes due to Atg7 deletion ( Figure S1).
To further determine the expression pattern of Sirt3 in hematopoietic aging, we tracked a time course of Sirt3 transcription in wild-type mice by quantitative RT-PCR. The data show that Sirt3 increased its expression until 24 weeks and thereafter transcription progressively decreased in the HSC-enriched cells (Figure 3h), a dynamic pattern similar to deterioration of autophagy in physiological aging. Surprisingly, Sirt3 expression began to dramatically decrease at 4-6 weeks in the mouse model in which hematopoietic Atg7 is disrupted (Figure 3i), a pattern almost identical to the accelerated hematopoietic aging due to the low activity of autophagy ( Figure 1).
In 10-week-old Atg7-deleted mice, Sirt3 expression is lower than that of the wild-type mice at 84 weeks (Figure 3h,i). Together, the above data demonstrate that Sirt3 expression depends on Atg7.

| Loss of Atg5 causes loss of Sirt3 and aging of hematopoiesis, resembling Atg7 deletion
To answer whether Atg7 deletion-caused reduction in Sirt3 expression and hematopoietic aging is autophagy-dependent, we generated hematopoietic Atg5-deleted mice by crossing the Atg5 flox mice from Dr. Noboru Mizushima, Japan (Hara et al., 2006), to Vav-iCre mice (Jackson Laboratory) to examine the effect of deletion of Atg5, another autophagy-essential gene, on Sirt3 in the hematopoietic system. The results show that deletion of Atg5 disrupts the transformation of LC3-I to LC3-II by lipidation ( Figure S2a Figure S2e). These data, together with Atg7 deletion data, support that disruption of autophagy machinery, due to loss of either Atg7 or Atg5, causes loss of Sirt3 and hematopoietic aging.

| Autophagy collaborates with Sirt3 to decelerate mouse hematopoietic aging
To examine whether autophagy collaborates with Sirt3 to counteract hematopoietic aging, we began with in vitro colony-forming assays in which the Sirt3-depleted HSC-enriched cells, prepared by infec- To test whether autophagy regulates Sirt3 expression, we triggered an ex vivo autophagy response in wild-type mouse HSCenriched cells with rapamycin, an autophagy inducer that was reported to increase lysosome biogenesis via mTOR signaling (Carroll & Dunlop, 2017). While rapamycin treatment activates autophagy manifested by the upregulation of LC3a and ULK1 at the transcriptional level (Figure 5a), which was further confirmed by enhanced autophagy response measured by Amnis ImageStream imaging flow cytometer for functional co-localization between autophagosomes (LC3 marker) and lysosomes (Lamp1 marker) (Figure 5b  Ex vivo activation of autophagy by starvation selectively upregulates Sirt3 transcription in primary HSC-enriched hematopoietic cells of wild-type mice. Transcription levels for Sirtuin family members were detected by quantitative PCR. (e) In vivo activation of autophagy by progressive starvation upregulates Sirt3 transcription. Transcription levels for Sirt3 (left) and autophagy-essential genes (right) in the HSC-enriched hematopoietic cells from long-term progressively starved mice were detected by quantitative PCR. Starvation was achieved by calorie restriction where the amount of feed for the mice was increasingly reduced by 10% of that fed last week, for a total of 4 weeks.

| Activation of autophagy upregulates Sirt3 in human blood
To examine whether the expression pattern of Sirtuin family members in human HSC-enriched hematopoietic cells is similar to that in mouse HSC-enriched hematopoietic cells, we measured the transcription levels of the Sirtuin genes in CD45CD34 cells of humans under age of 40 years by quantitative RT-PCR. The result shows that Sirt3 is the most pronounced member among its family in the HSC-enriched hematopoietic cells of adult but not aged humans (Figure 6a), which is in consistency with the pattern of wild-type mice (Figure 3b). Given that hematopoietic autophagy activity is reduced in the aged human population (Figure 1), we intended to test whether Sirt3 also declines in the HSC-enriched hematopoietic cells of aged human populations.
To this end, we measured Sirt3 expression by quantitative RT-PCR in bone marrow CD45CD34 cells from two different aged human groups (below age of 40 years group and above age of 70 years group).
The results show that Sirt3 transcription level was decreased in the group of over 70 years as compared to the group of less 40 years old ( Figure 6b, left), a similar pattern to autophagy deterioration in human blood (Figure 1b,c,d). Furthermore, at the translational level, both Sirt3 and two key autophagy proteins, ULK1 and Beclin1, were extremely reduced in the hematopoietic stem cells of the aged individuals (Figure 6b, (Figure 6c). The results thus suggest a similar module in human blood in the autophagic regulation of Sirt3 that was identified in mice.

| DISCUSS ION
The study presented here provides new insights into autophagic regulation of hematopoietic aging. By examining human blood, we show that although the decrease in peripheral blood counts is mildly progressive, the deterioration of autophagy and reduction of Sirt3 expression are strikingly significant in both hematopoietic cells and their stem and progenitor cells in aged humans. Using two mouse models in which autophagy is genetically impaired in the hematopoietic system, we show that autophagy collaborates with Sirt3, to form a regulatory axis that synergistically delays hematopoietic aging. Furthermore, we demonstrate that the autophagy-Sirt3 axis is also functional in human hematopoietic system. The summary of this study is illustrated in the cartoon (Figure 6d). Aging in hematopoietic system not only deteriorates the function of blood cells, but also accelerates aging of non-hematopoietic organs (Fang et al., 2019;Yuan et al., 2020). Hematopoietic aging in cells, driven by both intrinsic and extrinsic factors, is linked to impaired HSC self-renewal and reconstitution, and increased hematopoietic malignant incidence. Emerging studies that include increased oxidative stress and compromised DNA damage response (Martin-Pardillos et al., 2017;Pilzecker et al., 2017), global epigenetic alteration (Akunuru & Geiger, 2016;Florian et al., 2018)) and cell polarity shift (Carrillo-Garcia & Janzen, 2012;Florian et al., 2018), cellular senescence (Chang et al., 2016), clonal selection of HSCs (Chung & Park, 2017;Pang, Schrier, & Weissman, 2017), frequency of myeloid-restricted repopulating progenitors (MyRPs) (Yamamoto et al., 2018), as well as chronic inflammation (He et al., 2020), have provided important understanding into the mechanisms of hematopoietic aging.
A recent study on hematopoietic autophagy demonstrated that in aged mice, the majority of HSCs have lower autophagy activity, which contributes to a lower reconstitution potential, with only onethird of aged HSCs maintaining active autophagy at similar capacity to young mice for self-renewal and multilineage differentiation (Ho et al., 2017). Here, we show that in aged mice (90 weeks old), the overall capability for HSC reconstitution is significantly low because the peripheral blood counts are markedly reduced and the biased myeloid differentiation is clearly apparent (Figure 2b,c). 10-week-old mice with a gene deletion to constitutively suppress autophagy displayed severe hematopoietic aging, mimicking the aging phenotype of 90-week-old wild-type mice (Figure 2b). However, in contrast to these observations in mice, peripheral blood counts only display a mildly progressive decline in aging humans covering a broad age range from 20 to 90 years old (Figure 1a). Nevertheless, autophagy-essential genes and their function in human autophagy machinery are strongly inhibited in aged individuals (Figure 1b,c,d; Figure 6b) to a degree comparable with that of the autophagy-defective mouse model (Cao, Cai, et al., 2015;. These results suggest that functional deterioration of the hematopoietic system may largely be attributed to insufficient autophagy activity in humans. Sirt3 increases the activity of antioxidants, such as superoxide dismutase 2 (SOD2), and promotes ROS scavenging (Qiu, Brown, Hirschey, Verdin, & Chen, 2010;Someya et al., 2010;Tao et al., 2010;Tseng, Shieh, & Wang, 2013;Yu, Dittenhafer-Reed, & Denu, 2012). Decreased expression of Sirt3 in aged HSCs is associated with a concomitant repression of mitochondrial protective programs, which can result in impaired function of the Sirt3-directed mitochondrial unfolded protein response pathway (Brown et al., 2013). Sirt3 is indispensable under stress or in aged mice, and it is downregulated with age, contributing to increased ROS levels in aged HSCs (Brown et al., 2013). In this study, we find that Sirt3 is the major Sirtuin family member in the HSC-enriched hematopoietic cells of both mice ( Figure 3) and humans (Figure 6a).
Transcription of Sirt3 relies on autophagy and is progressively reduced from early to late stages in the hematopoietic hierarchy (Figure 3e), showing an accelerated drop in autophagy-defective mice, with expression at 10 weeks of age comparable to that of the wild-type mice at 84 weeks of age (Figure 3h,i). An ex vivo study showed that Sirt3 is the only member in Sirtuin family that effectively responds to multiple autophagy inducers (Figure 5c,d). Similarly, an in vivo study with longterm calorie restriction further revealed that restriction of food intake upregulates Sirt3, along with the upregulation of autophagy (Figure 5e).
These results indicate that Sirt3 expression relies heavily on intact autophagy machinery, and Sirt3 decline is associated with low autophagy capacity, whereas activation of autophagy reverses Sirt3 decline. We thus conclude that aging-related suppression of Sirt3 is attributed to the deterioration of the capacity for autophagy. We previous found that in a non-aging blood cancer cell line, Sirt3 appears to be associated with elevated oxidative stress and is downregulated by autophagy (Fang et al., 2016). The discrepancy in the role and autophagic regulation of Sirt3 is enriched in mitochondria in the cell and initiates metabolic adaptations to enhance mitochondrial management of oxidative stress inducers; therefore, Sirt3 positively regulates mitophagy (Meng et al., 2019). Mitophagy is regulated by multiple mechanisms including post-translational modification of a long array of enzymes that catalyze phosphorylation, acetylation, and deacetylation, etc. (Wang, Qi, Tang, & Shen, 2020). Increased mitochondria are often a cause of low capacity of mitophagy that promotes aging (Bakula & Scheibye-Knudsen, 2020). Our present study with Atg7-deleted mice indicates an increase of mitochondrial mass ( Figure 2d) and a decrease in Sirt3 ( Figure 3), but activation of autophagy did not reduce Sirt3 levels and Sirt3 prevented premature hematopoietic aging in an autophagy-disrupted mouse model, and we conclude that the evidence is compelling to support the hypothesis that the autophagy-Sirt3 axis delays hematopoietic aging independently of the mitophagy pathway.
Sirt3 regulates aging via its deacetylation capacity (Brown et al., 2013;Shimazu et al., 2010;Someya et al., 2010). Although overwhelming studies indicate Sirt3 functions in the mitochondria due to its mitochondrial localization, a previous study shows that in humans, both the full-length and processed forms of Sirt3 target H4-K16 for deacetylation in vitro and can deacetylate H4-K16 in vivo when recruited to a gene, and Sirt3 is transported from the nucleus to the mitochondria upon cellular stress (Scher, Vaquero, & Reinberg, 2007). A recent work with systematic study suggests that human Sirt3 displays class-selective histone de-β-hydroxybutyrylase activities with preference for H3 K4, K9, K18, K23, K27, and H4 K16, but not for H4 K5, K8, and K12, which distinguishes it from HDACs, another group of deacetylation enzymes (Zhang et al., 2019). These reports suggest that Sirt3 is able to deacetylate proteins in the nucleus, not just in the mitochondria, albeit a transient nuclear presence for full-length Sirt3.
To answer whether Sirt3 deacetylation activity on nuclear proteins depends on autophagy machinery, we measured histone acetylation levels in the bone marrow mononuclear cells of the wild-type and Atg7-deleted mice, and the results show that histone acetylation was highly accumulated in the Atg7-deleted mice ( Figure   S4). Furthermore, transcriptional profiling of hematopoietic stem and progenitor cells indicates that Atg7 deletion does not alter the expression levels of all eleven members of HDAC family ( Figure S5).
These data appear to be in agreement with previous reports on nuclear autophagy that targets nuclear components for degradation (Luo, Zhao, Song, Cheng, & Zhou, 2016;Papandreou & Tavernarakis, 2019). Our results thus propose that Sirt3 more likely contributes to the deacetylation of histone proteins in hematopoietic cells in the context of autophagic regulation, and the acetylation activity of Sirt3 appears to be autophagy-dependent in hematopoietic aging. Similar to our observations in mice, Sirt3 is also highly expressed in the HSC-enriched hematopoietic cells of young humans, but it is highly suppressed in the hematopoietic system of aged humans ( Figure 6b). In accordance with the observation that in vivo activation of autophagy by caloric restriction upregulates Sirt3 expression in mice (Figure 5e), the induction of autophagy by various mTOR inhibitors effectively upregulates Sirt3 in the HSC-enriched hematopoietic cells of humans (Figure 6c). This suggests that the autophagy-Sirt3 axis regulating aging is similarly present in the mouse and human hematopoietic system (Figure 6d). Therefore, our finding provides promising translational avenues for decelerating human hematopoietic aging by enhancing the autophagy-Sirt3 axis with autophagy inducers, or by enhancing Sirt3 expression.
Although our study has established the functional connection between autophagy and Sirt3 transcription, there are missing parts in the regulatory autophagy-Sirt3 axis. In particular, the direct target of autophagy, which may negatively regulates Sirt3 expression in the nucleus, has yet to be identified. Our future work will continue the search for the direct targets of autophagy in the anti-aging regulatory axis.

| Clinical records and human samples
The clinical records in human blood test were provided by the
All mice were bred on a C57BL/6 genetic background. Whole blood was collected from the mouse orbit after anesthesia, subjected to an automated blood count (ADVIA 2120i). For calorie restriction, mice were starved by progressive reducing 10% food intake per week for 1 month. All animal experiments were reviewed and approved by the Institutional Committee on Animal Welfare Protection and Ethics of Soochow University.
The information of biological reagents is provided in Table S1.

| RNA-seq analysis
LSK cells were sorted from 10-week-old Atg7 +/+ and 10-weekold Atg7 −/− mice. Sequencing libraries were generated using NEBNext® UltraTM RNA Library Prep Kit for Illumina® (NEB, USA) following the manufacturer's recommendations and followed by sequencing with on an Illumina HiSeq platform and 125 bp/150 bp paired-end reads were generated. Approximately 40 million paired-end reads per sample were aligned to the mouse genome (GRCm38/mm10). HTSeq v0.6.0 was used to count the reads numbers mapped to each gene. FPKM of each gene was calculated based on the length of the gene and reads count mapped to this gene. Differential expression analysis of two conditions/ groups (two biological replicates per condition) was performed using the DESeq2 R package (1.10.1). DESeq2 provides statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting p-values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate.
Genes with an adjusted p-value <0.05 and fold change >2 found by DESeq2 were assigned as differentially expressed. LSK cells were sorted from 10-month-old Atg7 +/+ and Atg7 −/− mice. Sequencing libraries were generated using NEBNext® UltraTM RNA Library Prep Kit for Illumina® (NEB, USA) following the manufacturer's recommendations, and the quantity and quality of them were assessed by Q ubit 2.0 Fluorometer and Agilent 2100 bioanalyzer and followed by sequencing with on an Illumina HiSeq PE150 platform and 150 bp paired-end reads were generated. Approximately 40 million paired-end reads per sample were aligned to the mouse genome (GRCm38/mm10). HTSeq v0.6.0 (Anders, Pyl, & Huber, 2015) was used to count the reads numbers mapped to each gene.
FPKM of each gene was calculated based on the length of the gene and reads count mapped to this gene. The expression levels of all genes were calculated by RSEM v1.2.28 (Li & Dewey, 2011). Differential expression analysis of two conditions/groups (two biological replicates per condition) was performed using the edgeR package (v3.12.1) (Robinson, McCarthy, & Smyth, 2010).
The edgeR provides statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting p-values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate (Benjamini & Yekutieli, 2005). Genes with an p-value <0.05 and fold change >2 found by edgeR were assigned as differentially expressed and visualized as volcano map. In addition, the differentially expressed pattern of Sirtuin gene family was analyzed by using R language (v3.4.3).

| qRT-PCR analysis
Total RNA was extracted with MicroElute Total RNA Kit (R6831, Omega), and cDNA was synthesized using the RevertAid cDNA Synthesis Kit (K1622, Thermo Fisher Scientific) according to the manufacturer's instructions. RT-qPCR was performed in triplicate using Roche LightCycler 480 Instrument II with the LightCycler 480 SYBR Green I Master (04707516001, Roche). Expression of the gene of interest was normalized to the housekeeping gene GAPDH using the 2 -ΔΔCt method. All RT-qPCR primers used are listed in Table S2.

| Western blot analysis
Cells were harvested from 10-week-old mice. Lysates from cells were prepared using cell lysis (9803S, Cell Signaling Technology) for Western blotting as previously described (Cao, Cai, et al., 2015).

| Statistical analysis
The statistical significance of differences was calculated for Pearson correlation coefficient of clinical data using SPSS version 23.0 or Student's unpaired t test for experiments with animals and all primary samples including animals and humans. Graphs containing error bars show the mean ±SEM. Statistical significance is represented as *p < 0.05, **p < 0.01, ***p < 0.001 and not significant (ns).

ACK N OWLED G M ENTS
The authors thank the patients, healthy blood donors, and clini-

CO N FLI C T S O F I NTE R E S T
The authors have declared no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
YF and JW conceived the project. YF designed the study. YF, NA, LZ, YG, JQ, GJ, and RZ performed the experiments. GZ, YF, and XY analyzed the RNA sequencing data. WW and LX performed mouse genotyping. YF, NY, SP, YZ, and JW discussed and analyzed the data.
YF and JW wrote the manuscript. All authors read and approved the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors declare that the main data supporting the findings of this study are available within the article and these supplementary information files. The RNA sequencing data have been deposited in GenBank database with an accession number PRJNA506815 (http://www.ncbi.nlm.nih.gov/biopr oject/ 506815).