Sirt6 prevents the age-related decline of H₂S through the control of one-carbon metabolism

Significance

Aging is a major risk factor for multiple diseases, facing humanity with the challenge of how to prolong healthspan. Here, we explore a molecular mechanism underlying the prolongevity activity of the Sirt6 enzyme in supporting healthy aging. We show that Sirt6 maintains youthful hepatic levels of hydrogen sulfide (H₂S), a gasotransmitter linked to the benefits of caloric restriction, by regulating cystine uptake and methionine metabolism. Sirt6 also prevents age-related increase in S-adenosylmethionine (SAM), the main methyl donor for epigenetic and protein methylation, through posttranslational acetylation. In addition, we define a link between one-carbon metabolism and the transsulfuration pathway. These findings reveal a mechanism of Sirt6 action and suggest potential therapeutic targets to support healthy aging.

Abstract

Mice overexpressing Sirt6 or fed a caloric restriction (CR) diet live longer with improved health. CR increases Sirt6 levels, and its beneficial effects are mediated by the gasotransmitter H₂S, a one-carbon pathway product. Yet, the role of this pathway in Sirt6-regulated longevity remains elusive. Here, we show that Sirt6 controls hepatic one-carbon metabolism, preventing the aging-dependent H₂S reduction, and the elevation of the methyl donor, S-adenosylmethionine (SAM). Sirt6 downregulates Slc7a11 expression in an Sp1-dependent manner, decreasing cystine uptake and increasing Cgl H₂S production activity. Additionally, comparative acetylome in old livers revealed Sirt6-related differential acetylation of most of the one-carbon enzymes. Specifically, Sirt6-dependent Matα1 K235 deacetylation reduces its SAM production activity and Cbs binding, thereby reducing its activation of Cbs-dependent H₂S production. The net outcome is H₂S and SAM levels as observed in young animals. Thus, we unveil a fundamental mechanism for the promotion of healthy longevity by Sirt6.

The average human lifespan has almost doubled over the last century, leading to the challenge of a growing elderly population suffering from multiple aging-related pathologies, including diabetes, cancer, inflammation, and neurodegenerative diseases (1). Hence, exploring the mechanisms underlying healthy aging is crucial. One of the key regulators of aging is the sirtuin family of proteins. The seven mammalian sirtuins, homologues of yeast sir2, are NAD⁺-dependent enzymes with catalytic activities mediating lysine deacetylation, deacylation, or monoADP ribosyltransferase activity (MADPR) (2, 3). All sirtuins have a major effect on key cellular processes, yet Sirt6 is best known for its lifespan- and healthspan-prolonging properties.

Sirt6 is a key regulator of aging, metabolism, and DNA repair (4–7). Along with its established localization to the nucleus and the endoplasmic reticulum (ER), it was shown to include a cytosolic fraction (8, 9). Sirt6 possesses all three enzymatic activities attributed to sirtuins and is active on a variety of protein substrates. Particularly, deacetylating lysines 9, 18, and 56 of histone H3 (H3K9, H3K18, and H3K56, respectively) (10–13). For efficient histone deacetylation, Sirt6 requires binding to the nucleosome rather than to free histones (14). Sirt6 binds to various transcription factors and deacetylates the promoter regions of their target genes. For example, Sirt6 preserves glucose homeostasis via Hif1α (15) and negatively regulates inflammation via NFκB (13). In addition, Sirt6 controls inflammation through demyristoylation of tumor necrosis factor alpha (TNFα) (12), cholesterol metabolism by sterol regulatory element-binding protein (SREBP) downregulation and AMP-activated protein kinase (AMPK) activation, and fatty acid beta oxidation through peroxisome proliferator-activated receptor alpha (PPARα) (16, 17).

Mice deficient for Sirt6 die prematurely several weeks after birth, due to severe hypoglycemia and aging-like phenotypes (18, 19). In rhesus monkeys, knockout (KO) of the SIRT6 gene causes severe prenatal developmental retardation, which leads to mortality a few hours after birth (20). In humans, a homozygote inactivating mutation in SIRT6 led to embryonic lethality (21). Conversely, overexpression of Sirt6 in Drosophila melanogaster significantly extends lifespan and increases resistance to oxidative stress (22). Transgenic mice overexpressing Sirt6 (Sirt6 TG, known as MOSES mice) also live significantly longer with improved healthspan (4). These mice maintain better glucose homeostasis and fatty acid beta-oxidation, experience reduced occurrence of malignancies and inflammation, are protected against the physiological sequelae of high-fat-diet-induced obesity, maintain young-like physical activity, and exhibit overall improved physiology compared to wild-type (WT) mice, specifically in old age (4, 5, 23). Interestingly, a specific SIRT6 variant was found in human centenarians, exhibiting enhanced MADPR activity, contributing to human longevity by improving genome maintenance (24).

Caloric restriction (CR) is defined as a 20 to 40% reduction in total calorie intake without malnutrition and is the most extensively studied and well-established manipulation to extend lifespan and healthspan in a variety of organisms (25, 26). In rodents fed CR diet, Sirt6 expression is upregulated in several tissues, including the liver, indicating a pivotal role of Sirt6 in CR-mediated beneficial effects. Indeed, Sir6 TG mice show a transcriptional profile and regulated pathways highly similar to those affected by CR (27), including the above-mentioned pathways.

A key factor in the CR response is the gasotransmitter hydrogen sulfide (H₂S) (28, 29). In the cell, H₂S is produced through the transsulfuration pathway (TSP) by two pyridoxal 5’-phosphate (PLP)-dependent cytosolic enzymes, cystathionine beta synthase (Cbs) and cystathionine gamma lyase (Cgl). Additionally, it is produced by the mitochondrial 3-mercaptopyruvate sulfurtransferase (3-MST). H₂S can be catabolized via its oxidation to sulfane sulfur by sulfide quinone oxidoreductase (Sqor) (30). Importantly, H₂S levels must be maintained within a defined range, between the nanomolar to the low micromolar, as high H₂S levels are toxic and even lethal (31). Interestingly, TSP-related H₂S production was shown to increase under CR and is essential for the benefits of dietary restriction (28, 32). Among its numerous effects, H₂S is known to enhance tissue repair and wound healing, moderate ischemia–reperfusion injury (IRI), improve the cardiovascular condition and blood flow in the vessels, and positively affect cognition (33–35). Exogenous supplementation of H₂S donors may also reduce inflammation, enhance autophagy, and prevent oxidative stress (36, 37). However, although H₂S is strongly linked to aging and its hallmarks and pathways, the effect of old age on H₂S metabolism is yet to be unraveled.

The TSP is a part of the methionine and cysteine pathway, also known as one-carbon metabolism, which also includes the folate and methionine cycles. These cycles are connected by methionine synthase (Mtr), which uses homocysteine and 5-methyltetrahydrofolate (5-MTHF) to regenerate methionine and tetrahydrofolate, sustaining both cycles. TSP starts with the consumption of homocysteine by Cbs instead of its remethylation, making Cbs the rate-limiting enzyme for TSP activation. The pathway is self-regulated at several levels. For example, 5-MTHF inhibits glycine N-methyltransferase (Gnmt), preventing demethylation of S-adenosylmethionine (SAM) to S-adenosylhomocysteine (SAH) (38). Moreover, the production of SAM, the source for all cellular methylation reactions, by the hepatic methionine adenosyltransferase alpha 1 (Matα1) drives homocysteine to feed the TSP rather than the methionine cycle both by allosterically activating Cbs and by inhibiting Methylenetetrahydrofolate reductase, the producer of 5-MTHF (38, 39). Reduced 5-MTHF activity decreases Mtr catalysis. The activity of the TSP is further controlled by the availability of cysteine. Usually, cysteine is mainly provided through the diet and is transferred from the blood into the cell through the cystine-glutamate antiporter Slc7a11, also known as the xCT system. However, under cysteine restriction, cysteine levels are maintained by the upregulation of Cgl expression to enable enhanced de novo cysteine synthesis (40). Accordingly, Cgl-deficient mice were shown to suffer acute lethal myopathy and oxidative injury when deprived of dietary cysteine (41).

Besides enhancing H₂S production, CR leads to upregulation of Cbs, Cgl, and Matα1 protein expression (28, 42). In addition, CR induces hepatic Gnmt-dependent reduction in SAM, accompanied by an increase in SAH (42). This results in an overall induction of the one-carbon outcomes, including the TCA cycle, methylation processes, nucleotide biosynthesis, and oxidative homeostasis.

Despite the phenotypic overlap between the role of the TSP and Sirt6 in CR-mediated healthy longevity, their connection is still elusive. Here, we describe the effect of Sirt6 on one-carbon metabolism in old age. We found that Sirt6 restores hepatic H₂S production in old age back to young animals’ levels. To increase H₂S production, Sirt6 downregulates Slc7a11 by inhibiting the transcription factor Sp1. This reduces cystine uptake, which encourages enhanced Cgl activity. To maintain H₂S levels in the beneficial range, Sirt6 decreases Matα1 activity by controlling its acetylation. Through this mechanism, Sirt6 reduces SAM production and prevents both the aging-related increase in SAM and the overactivation of the TSP. Overall, Sirt6 maintains a dual mechanism by which it controls aging-related changes in the one-carbon metabolism, thereby providing a pathway for its longevity-enhancing effect.

Results

Sirt6 Prevents Aging-Related Reduction of H₂S Production in the Liver.

One-carbon metabolism, particularly the TSP, has been shown to regulate CR-mediated healthy aging (28, 32, 43). To highlight the involvement of one-carbon enzymes in aging, a longitudinal analysis was performed in mice aged 6, 12, 18, 24, 30, and 33 mo. This revealed an age-dependent transcriptional downregulation of one-carbon metabolism enzymes. In line with that, CR-fed mice showed a global increase in the expression of many of these genes compared to chow-fed controls (Dataset S1). These findings emphasize the pathway’s importance in healthy aging.

Sirt6 plays a major role in healthy longevity. Therefore, we examined whether Sirt6 overexpression regulates H₂S production, a key one-carbon cycle component mediating CR’s benefits. H₂S production capacity was measured in liver lysates of young (3 mo) and old (23 to 25 mo) WT and Sirt6 TG mice. Old WT mice showed significantly reduced H₂S production and hepatic H₂S levels compared to young WT mice (Fig. 1A and SI Appendix, Fig. S1A), whereas old Sirt6 TG mice maintained youthful H₂S levels (Fig. 1B and SI Appendix, Fig. S1A). Both young and old Sirt6 TG mice had higher H₂S production than WT controls (Fig. 1 C and D), with a pronounced increase in old mice. This effect was liver-specific, as it was not seen in kidney and brain tissues, in which old Sirt6 TG mice had similar or lower H₂S production compared to WT littermates, respectively (SI Appendix, Fig. S1 B and C). In comparison, overexpression of Sirt1, another known aging-related sirtuin (3), had no effect on H₂S production capacity in either young or old age (SI Appendix, Fig. S1D). Importantly, the twofold increase in H₂S in old Sirt6 TG mice compared to old WT mice was not due to changes in the expression of the H₂S-producing enzymes Cbs and Cgl, which were similarly expressed in both groups (SI Appendix, Fig. S1 E and F).

Fig. 1.

Sirt6 prevents aging-related reduction of liver H₂S production. H₂S production capacity assay of: (A) Young (3 mo) and old (23 to 25 mo) WT mouse liver lysate. (B) Young WT and old Sirt6 TG mouse liver lysate. (C) Young WT and young Sirt6 TG mouse hepatocyte lysate. (D) Old WT and old Sirt6 TG mouse liver lysate. ImageJ quantification of H₂S production normalized to the average of the control group, and represented as mean ± SEM, is shown below each assay. *P < 0.05, ***P < 0.001, ****P < 0.0001, Student’s t test, n = 6-7.

Sirt6 Activates Cgl through Sp1-Related Slc7a11 Downregulation.

Dietary restriction lowers cysteine availability, reducing its uptake through the cystine/glutamate antiporter Slc7a11 (xCT). This triggers Cgl upregulation to increase cysteine de novo synthesis to meet cellular demands (40). Slc7a11 overexpression in HEK293T cells significantly decreased H₂S production capacity (SI Appendix, Fig. S2A). Therefore, the expression levels of Slc7a11 were examined in young and old WT and Sirt6 TG isolated hepatocytes. A significant increase in Slc7a11 messenger RNA (mRNA) expression was observed in old WT compared to young WT mice. However, in old TG mice, Slc7a11 expression levels were significantly reduced compared to old WT littermates, resembling its expression in young mice (Fig. 2A). This suggests that the age-dependent reduction in H₂S production is due to an increase in Slc7a11. Likewise, compared to SIRT6-deficient HEK293T cells, reexpression of human SIRT6 resulted in a significant downregulation in the expression levels of SLC7A11 (Fig. 2B). To explore whether SIRT6 enzymatic activity is required for its regulation of SLC7A11, a catalytically inactive mutant of SIRT6, H133Y (44), was also expressed in these cells. Surprisingly, the repression of SLC7A11 by SIRT6 was independent of its enzymatic activity (Fig. 2B). This suggests that SIRT6 may downregulate SLC7A11 levels via a transcription factor that controls its expression. Using the ENCODE ChIP database (45, 46), we identified that both SIRT6 and SP1 are localized to the SLC7A11 promoter (SI Appendix, Fig. S2B). Thus, the downregulation of SLC7A11 by SIRT6 is potentially SP1 dependent. To address this possibility, SIRT6 or a GFP control was expressed in HEK293T cells lacking either SIRT6, SP1, or both. In comparison to GFP control, SIRT6 overexpression or SP1 KO significantly reduced SLC7A11 expression. However, this effect was not additive, as overexpression of SIRT6 in SP1 KO cells had no further effect on SLC7A11 expression compared to SP1 KO alone (Fig. 2C and SI Appendix, Fig. S2C). Notably, SP1 KO cells also had higher H₂S production capacity compared to WT cells (Fig. 2D). Furthermore, in comparison to old WT mice, Sirt6 TG littermates had significantly lower liver Sp1 protein levels (Fig. 2E). These findings demonstrate that Sirt6 regulates H₂S production via Slc7a11 in a Sp1-dependent manner. Interestingly, no change was found in the mRNA expression levels of Sp1 between WT and Sirt6 TG mice (SI Appendix, Fig. S2D). This indicates that Sp1 regulation by Sirt6 is at the protein level alone.

Fig. 2.

Sirt6 activation of TSP through Sp1-dependent xCT downregulation. Slc7a11 mRNA expression levels in: (A) Isolated hepatocytes of young WT, old WT, and old Sirt6 TG mice. n = 3-7. (B) HEK293T SIRT6 knockout cells expressing GFP control, WT, or H133Y human SIRT6. n = 3-5. (C) HEK293T SIRT6^−/− SP1^+/+ or SIRT6^−/− SP1^−/− cells transiently expressing GFP control or human SIRT6 WT. n = 3. (A–C) analyzed with one-way ANOVA test, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. (D) H₂S production capacity in control WT and SP1 KO HEK293T cells. n = 3. (E) Sp1 protein expression in old WT and old Sirt6 TG mouse livers. n = 11-14. For (D and E) ImageJ quantification of H₂S or protein levels, respectively, normalized to Ponceau staining is shown on the Right. Student’s t test, **P < 0.01, ***P < 0.001. (F) Serum IGF-1 levels in adult WT and Sirt6 TG mice, injected with Cgl inhibitor PAG or control PBS. Two-way ANOVA test, *P < 0.05. All data are represented as mean ± SEM.

To test a potential further effect of Sirt6 over H₂S catabolism, we measured the expression of sulfide quinone oxidoreductase (Sqor), the main enzyme that lowers H₂S levels by oxidation to sulfane sulfur (30). No change was detected between old WT and Sirt6 TG livers in Sqor mRNA nor protein expression levels (SI Appendix, Fig. S3 A and B). Reduced IGF-1 levels and signaling, key mediators of CR, were previously suggested as an underlying mechanism for Sirt6-dependent lifespan extension (5). Thus, we checked whether Sirt6’s effect on IGF-1 levels is mediated by Cgl-related H₂S production. As seen in Fig. 2F, in comparison to WT mice, Sirt6 TG mice have significantly lower sera IGF-1 levels. Yet, strikingly, an injection of DL-propargylglycine (PAG), a potent Cgl inhibitor (SI Appendix, Fig. S3C), specifically abolished this effect in vivo. This indicates that Sirt6-dependent increased H₂S production mediates its longevity-associated phenotypes.

Global Liver Acetylome in Old Mice.

In search for additional regulation of Sirt6 on one-carbon metabolism, LC-MS/MS based proteomics was performed in the livers of old (23 to 25 mo) WT and Sirt6 TG mice. Out of 3,358 MS-identified proteins, only 203 exhibited a significant difference between the two genotypes. Of these, the expression levels of 70 and 133 proteins were higher or lower in old Sirt6 TG compared to WT mice, respectively. These proteins comprise the Sirt6-dependent old liver proteome (SI Appendix, Fig. S4A). Nevertheless, no significant alterations in one-carbon nor TSP enzymes were observed (Dataset S2).

Next, we explored whether Sirt6 regulates one-carbon cycle at the protein acetylation level. To date, the vast majority of Sirt6-related phenotypes were connected to its deacetylation activity on histone/chromatin-associated proteins, and studied mainly in young mice. To explore the broader effects of Sirt6 overexpression, a label-free LC-MS/MS-based analysis of the whole liver acetylome, spiked with reference peptide (47), was performed in old WT and Sirt6 TG mice. Nine mice from each genotype were analyzed. First, all 6,349 lysine acetylation sites on 1,722 proteins identified in both groups were collected to characterize the global old liver acetylome (Dataset S3). Cellular compartment (CC) analysis of these sites revealed that cytoplasmic and mitochondrial proteins were most enriched (SI Appendix, Fig. S4B). Pathway analysis of the global old liver acetylome identified multiple metabolic processes, including carbon metabolism, amino acid metabolism, sulfur compound metabolic process and fatty acid degradation (SI Appendix, Fig. S4 C and D). Further, Proteomaps (48) analysis showed that acetylated proteins in the old liver mostly relate to biosynthesis, central carbon metabolism, protein translation and folding, sorting, and degradation (SI Appendix, Fig. S4D). As previously found in young animals (49), a pLogo (50) analysis of all recognized 6,349 acetylation sites in the old liver found no specific motif of acetylated regions (SI Appendix, Fig. S4E). This suggests that like in young mice, multiple enzymes deacetylate/acetylate the vast majority of the old liver acetylome.

Sirt6-Dependent Acetylome in the Liver of Old Mice.

A comparison between the liver acetylomes of old Sirt6 TG and WT mice found a significant change in the acetylation levels of 374 lysine residues on 291 proteins (P < 0.05) (Dataset S3). Within these, the acetylation levels of 206 residues on 170 proteins and 168 residues on 138 proteins were significantly decreased or increased in Sirt6 TG mice, respectively (Fig. 3A). CC analysis found that, surprisingly, despite the primarily nuclear localization of Sirt6, nuclear proteins were underrepresented in the Sirt6-dependent acetylome. Sirt6-dependent differentially acetylated proteins were mostly localized to the cytoplasm, mitochondria, and the endoplasmic reticulum (Fig. 3B). 55.49% of the mitochondrial Sirt6-dependent acetylation sites were increased in Sirt6 TG vs. WT mice. As mitochondrial proteins can be nonenzymatically acetylated upon increased acetyl co-A levels (51), the expression of ATP citrate lyase (Acly), which produces acetyl co-A, was measured. Compared to WT littermates, in the liver of old Sirt6 TG mice the expression of Acly was significantly increased (SI Appendix, Fig. S5A), suggesting an enhanced acetyl co-A production and nonenzymatic mitochondrial protein acetylation. While in most CCs, differentially acetylated sites were increased or decreased to a similar degree, in the plasma membrane, there was a substantial enrichment of proteins that were more acetylated in Sirt6 TG compared to WT mice (Fig. 3C). Among the Sirt6-dependent altered acetylations found on nuclear proteins, 35 acetylation sites were reduced in Sirt6 TG mice, indicating a possible set of Sirt6 direct substrates (Dataset S4). To unravel potential Sirt6-dependent specific motifs, a pLogo analyses were performed on general, decreased, and increased Sirt6-dependent acetylated sites. In contrast to the global liver acetylome, a partial motif was found when examining all 374 Sirt6-dependent acetylation sites. Particularly, differentially acetylated proteins were characterized by an overrepresentation of F/R at positions +1 from the acetylated lysine, R/H at position +9, and D/E at position −1. In addition, I was globally overrepresented after the acetylation site (Fig. 3D). Interestingly, Sirt6-dependent increased acetylation sites showed only D and R at positions −1 and −9, respectively, whereas decreased acetylation sites had E at positions −1 and −9, and F at position +1 (Fig. 3 E and F). In regard to underrepresented amino acids, in the general Sirt6-dependent acetylome T was underrepresented at position +1, and P and R at positions −1 and −2, respectively. Increased acetylation sites showed KK/P at positions −1 and −2, whereas decreased acetylation sites had E at positions −1 and −9, and F at position +1. S was underrepresented in both sides of the acetylated lysine in all three analyses (Fig. 3 D–F).

Fig. 3.

Sirt6-dependent liver acetylome of old mice. (A) Heatmap representation of differentially acetylated sites between old WT and Sirt6 TG livers. (B) CC analysis represented as a heatmap in each CC. Heatmap scale is shown on the Right. (C) Percentages of lysine residues with increased (red) or decreased (blue) acetylation in Sirt6 TG out of total differentially acetylated sites in each CC. (D–F) Amino acid sequence motif (pLogo) analysis of: Sirt6-dependent total differentially acetylated (D), decreased acetylation (E), and increased acetylation (F) sites. Student’s t test P < 0.05, n = 9.

Enrichment analysis of differentially acetylated proteins in Sirt6 TG mice revealed a PTM-based regulation of Sirt6 on various metabolic pathways. These include small molecule catabolic processes, amino acid metabolism, purine-containing compound metabolic processes, valine, leucine, and isoleucine (branched chain amino acids) metabolism, nicotinamide nucleotide metabolic processes, and others (Fig. 4A and SI Appendix, Fig. S5 B and C). Moreover, Proteomaps analysis showed that the old Sirt6-dependent acetylome is mostly composed of proteins involved in amino acid metabolism, lipid and steroid metabolism, and glycolysis, strengthening the pathways analysis, in addition to ribosome and exosome-related proteins (Fig. 4B). Volcanoplot analysis of the Sirt6-dependent acetylome is shown in Fig. 4C. The abovementioned proteomic analysis of the same livers showed that alterations in the acetylome are not due to changes in protein expression levels. Only 28 common proteins were found between Sirt6-dependent acetylome and proteome (SI Appendix, Fig. S6A and Dataset S2). Moreover, Volcanoplot analysis showed that, except for Gstm3, the highest differentially expressed Sirt6-dependent proteins were not part of the Sirt6-dependent acetylome (SI Appendix, Fig. S6B). Importantly, Proteomaps analysis of Sirt6-dependent old liver proteome showed an enrichment in metabolic pathways including some that were also enriched in the Sirt6-dependent old acetylome, i.e., amino acid metabolism, lipid and steroid metabolism, and glycolysis (SI Appendix, Fig. S6C).

Fig. 4.

Sirt6-dependent old liver acetylome pathways. (A) Metascape pathways and (B) Proteomaps analyses of Sirt6-dependent acetylation sites. (C) Volcanoplot representation of significantly increased (red)/decreased (green) Sirt6-dependent acetylation sites, Student’s t test P < 0.05, and fold change > 2, n = 9.

Matα1 K235 Acetylation Controls Its Enzymatic Activity.

Strikingly, we identified most of the enzymes in the one-carbon metabolism pathway within the Sirt6-dependent acetylome (Fig. 5A, differently acetylated enzymes are encircled in red). Within those, the two most affected were Cbs and Matα1 (Fig. 4C and Dataset S3). In comparison to WT mice, Sirt6 TG mice had higher acetylation on lysine 319 (K319) of Cbs, the rate-limiting enzyme of the TSP (28.3 fold change, P = 0.001). Thus, to explore whether K319 acetylation controls Cbs activity, two mutations of the K319 residue were generated: K319R, which mimics constitutively deacetylated K319, and K319Q, which mimics constitutively acetylated lysine (52). Nevertheless, compared to WT Cbs, recombinant proteins of these mutants showed no change in H₂S production activity (SI Appendix, Fig. S7A). Next, we followed the acetylation on Matα1 lysine 235 (K235). Compared to WT mice, Matα1 K235 acetylation was significantly reduced in old Sirt6 TG mice (0.038 fold change, P = 0.02). As seen in Dataset S2 and SI Appendix, Fig. S7B, both proteomics and western blot analyses showed no significant change in Matα1 protein expression levels between the two groups. In addition, coimmunoprecipitation (co-IP) of flag-tagged Matα1 expressed in HEK293T, but not of GFP control, showed a direct interaction with endogenous SIRT6. Vice versa, flag-tagged Sirt6 precipitated Matα1 (Fig. 5 B and C), suggesting a direct deacetylation of Matα1 K235 by Sirt6. Thus, the role of K235 acetylation was investigated. First, an in vitro activity assay of recombinant WT, K235Q, and K235R Matα1 was performed. Compared to the acetylation-mimicking mutant K235Q, the deacetylation-mimicking mutant K235R had a significantly reduced intrinsic activity (Fig. 5D). This suggests that old Sirt6 TG mice will have lower levels of Matα1 product, SAM.

Fig. 5.

Matα1 K235 acetylation controls its enzymatic activity. (A) Illustration of the one-carbon metabolic pathway. Sirt6-dependent differentially acetylated proteins are highlighted by red circles. (B and C) Coimmunoprecipitation of GFP control vs. Matα1-flag or Sirt6-flag expressed in HEK293T cells, followed by IB for SIRT6 or Matα1, respectively. n = 1-3. (D) Inorganic phosphate production activity of recombinant Matα1 WT vs. K235Q and K235R mutants. For (A–D) data are represented as mean ± SEM. n = 7-9, one-way ANOVA test, *P < 0.05, **P < 0.01, ***P < 0.001. (E–G) MS-based SAM (E) and SAH (F) measurement and SAM/SAH ratio (methylation index, G) in the livers of young (3 mo) and old (23 to 25 mo) WT and Sirt6 TG mice.

The levels of SAM and its downstream metabolite, SAH, were measured by MS-based targeted metabolomics in the liver of young and old, WT, and Sirt6 TG mice. Compared to young WT mice, SAM levels significantly increased in old WT mice. Importantly, old Sirt6 TG mice maintained young-like SAM levels (Fig. 5E). No difference was found between WT and Sirt6 TG mice at a young age. SAH levels did not significantly differ between the four groups (Fig. 5F). This suggests that the higher concentrations of SAM seen in old WT mice were derived from upregulation in its production by Matα1, rather than its demethylation to SAH. The ratio between SAM and SAH, also known as the methylation index, was significantly higher in old WT mice relative to young mice, while old Sirt6 TG mice had a young-like methylation index (Fig. 5G). Thus, old Sirt6 TG mice maintain a “younger” methylation index, by repressing the age-dependent increase in SAM levels.

Sirt6 Overexpression Reduces Cbs Activation through Matα1 K235 Deacetylation.

An appealing scenario is that Matα1 or other Sirt6-dependent differentially acetylated one-carbon enzymes, or their acetylation, might also be involved in the regulation of H₂S production. To test this hypothesis, the effect of one-carbon enzymes differentially acetylated in Sirt6 TG mice on H₂S production was analyzed. Matα1, Gnmt, Shmt2, Csad, Got2, Dmgdh, Adk, Gclc, and Aldh1l1 were overexpressed in human embryonic kidney HEK293T cells, and screened for H₂S production. Importantly, proteomics validation showed that all of these proteins are expressed in both kidney and liver tissues (SI Appendix, Fig. S8A). Only the overexpression of the first two increased or decreased H₂S production capacity, respectively (Fig. 6A). To elucidate the mechanism underlying the increase in H₂S production under Matα1 overexpression, Cbs and Cgl protein levels were examined. No change in their protein expression was detected in Matα1 overexpressing cells (SI Appendix, Fig. S8B), suggesting another mechanism for its enhancing effect on H₂S production. To define which of the TSP enzymes is affected by Matα1, H₂S production capacity of recombinant Cbs or Cgl, with or without recombinant Matα1 supplementation, was tested. No effect of Matα1 supplementation on Cgl-derived H₂S production was found. However, compared to Cbs alone, H₂S production by Cbs supplemented with Matα1 was significantly increased (Fig. 6 B and C and SI Appendix, Fig. S8C). Matα1 is the main producer of SAM in the liver, an established allosteric activator of Cbs (53). To test whether the effect of Matα1 over Cbs activity is solely through SAM, we compared the Matα1-dependent activation of Cbs WT and Cbs SAM-insensitive mutant, D440N (39). While Matα1 addition resulted in a dose-dependent activation of WT Cbs, the activity of Cbs D440N was not affected by Matα1 (Fig. 6D). This demonstrates that Matα1 enhances H₂S biosynthesis by Cbs, specifically via SAM production. We then examined whether Matα1 and Cbs physically interact. As seen in Fig. 6E, flag-tagged Matα1 but not the GFP negative control was co-IPed with endogenous Cbs in HEK293T cells. Reciprocally, flag-tagged Cbs was specifically co-IPed with endogenous Matα1 (Fig. 6F). These results show that Matα1 provides Cbs with direct supplementation of SAM, which allosterically activates its H₂S production activity.

Fig. 6.

Sirt6-dependent Matα1 deacetylation decreases Cbs binding and H₂S production activity. (A) Representative lead acetate–based H₂S production capacity assay of HEK293T cells transfected with either GFP (control) or flag-tagged Mata1, Gnmt, Shmt2, Csad, Got2, Dmgdh, Adk, Gclc, or Aldh1l1. The expression of each enzyme is shown underneath. Tubulin was used as loading control. ImageJ quantification of H₂S production normalized to Tubulin is shown on the Right. Data are represented as mean ± SEM. n ≥ 3, one-way ANOVA test, *P < 0.05, ****P < 0.0001. (B and C) Lead acetate–based H₂S production capacity assay of recombinant mouse Cgl (B) or Cbs (C) with or without recombinant mouse Matα1. ImageJ quantification of H₂S production normalized to Ponceau staining is shown on the Right. n ≥ 4, Student’s t test, ****P < 0.0001. (D) H₂S production of WT Cbs or SAM-insensitive Cbs mutant D440N, supplemented with increasing doses of Matα1. Linear regression of H₂S production normalized to Ponceau staining is shown on the Right. n = 3, simple linear regression represented as mean ± SEM, Extra Sum-of-Squares F-test was used to compare between slopes, ****P < 0.0001. (E and F) Coimmunoprecipitation of GFP control, Matα1-flag (E) or CBS-flag (F) expressed in HEK293T cells, followed by IB for CBS or Matα1, respectively. n = 3. (G) H₂S production activity of recombinant WT Cbs, with increasing doses of recombinant WT, K235Q, or K235R Matα1. n = 5, simple linear regression represented as mean ± SEM, Extra Sum-of-Squares F-test was used to compare between slopes, ****P < 0.0001. (H) Co-IP of GFP control or truncated (amino acids 195 to 275) WT, K235Q, or K235R Matα1-flag expressed in HEK293T cells, with endogenous CBS. n = 3.

Building on Matα1-driven H₂S production via Cbs, we further examined the Sirt6-dependent mechanism underlying this effect. An in vitro H₂S production capacity assay of WT Cbs supplemented with increasing doses of WT, K235Q, or K235R Matα1 resulted in an enhanced Cbs activation by Matα1 K235Q (Fig. 6G). This indicates that the deacetylation of Matα1 K235 reduces its activating effect over Cbs, potentially via lower SAM production. Structural analysis of human K234 (equivalent to mouse K235) on the Matα1 dimer revealed that this residue is located on an unstructured region and is exposed to the solvent (SI Appendix, Fig. S9A). Therefore, it is accessible for acetylation/deacetylation, and its modification may regulate Matα1 association with Cbs. Interestingly, the nearby tyrosine (human Y235, mouse Y236) is oriented to create an electrostatic bond between the ε-N of K234 and its hydroxyl group. However, acetylation causes only a slight change in the distance between these two residues. To examine whether K235 acetylation also regulates the Cbs–Matα1 interaction, a flag-tagged truncated Matα1_195-275 was generated and expressed in HEK293T cells. This fragment includes both K235 and the SAM-binding region of Matα1. As seen in Fig. 6H, all truncated variants co-IPed endogenous Cbs. Of the three truncated forms, the Matα1_195-275 K235Q mutant precipitated significantly more Cbs than the other truncated forms. Hence, Sirt6 likely reduces both Matα1 activity and its binding to Cbs through deacetylation of K235. Thus, Sirt6 prevents excessive activation of the TSP through negative regulation of SAM production. Interestingly, mimicking the Sirt6-dependent acetylation status of Dmgdh also reduced H₂S production in HEK293T cells (SI Appendix, Fig. S9B), indicating that further research is required in order to reveal the full effect of Sirt6 OE on the TSP.

Overall, we present a dual regulatory role of Sirt6 in both one-carbon and TSP activity. On the one hand, Sirt6 activates H₂S production by downregulating the xCT system. This is achieved by nonenzymatic inhibition of Sp1, as well as reducing Sp1 protein expression. On the other hand, it prevents H₂S overproduction by maintaining young-like SAM levels via Matα1 deacetylation, which lowers Cbs activation (Fig. 7).

Fig. 7.

The dual regulation of TSP activity by Sirt6. Sirt6 inhibits Sp1 expression and transcriptional activity, which leads to reduced Slc7a11 expression, and upregulates Cgl-derived H₂S production in the old liver. Concurrently, Sirt6 maintains young-like SAM levels and prevents TSP overactivation by reducing Matα1 K235 acetylation, which reduces its SAM production activity and its Cbs binding.

Discussion

Here, we present a dual mechanism through which in old age, Sirt6 controls hepatic TSP activity and maintains its H₂S production within the beneficial range, simultaneously maintaining young-like SAM levels (Fig. 7). This provides an underlying mechanism for the positive effect of Sirt6 on healthy longevity.

CR mediates healthy longevity in multiple organisms, from yeast to primates (25, 26). We previously showed that Sirt6 and CR share phenotypes such as delay of age-related diseases, lifespan extension, overlapping metabolic pathways, and similar liver transcriptional profile (4, 29). Here, we found several Sirt6-dependent acetylome pathways including the TCA cycle, branched chain amino acid metabolism, and fatty acids beta oxidation, which are also regulated by CR (29) (Figs. 3 and 4). Likewise, both Sirt6 and CR enhance H₂S production. However, while in CR these changes correlate with increased expression of TSP enzymes, here no such increase was observed.

Similar to CR, cysteine restriction increases Cgl expression (28). Yet despite reduced expression of Slc7a11, which reduces cysteine uptake, Cgl levels are not changed in Sirt6 TG livers. This might be explained by reduced Sp1 protein levels and activity in Sirt6 TG mice (Fig. 2E). Thus, the upregulation of Cgl by reduced cysteine uptake may be counteracted by Sirt6’s inhibition of Sp1, a known Cgl transcription factor (44, 54) (Fig. 2 B and C), resulting in unchanged expression. Additionally, the reduction in xCT levels in Sirt6 TG mice may activate Cgl, but not lower cysteine enough to trigger additional Cgl expression as seen in cysteine restriction. Importantly, further research may elucidate other mechanisms, in addition to this Sirt6-regulated indirect link between xCT and Cgl.

Increased H₂S production, as seen in CR, has a substantial positive effect on healthspan (35). At low concentrations, H₂S serves as an electron donor to the electron transport chain, supporting mitochondrial function. Historically, however, H₂S was known mainly for its toxicity, including inhibition of mitochondrial complex IV at high concentrations, leading to cell death (55). In humans, elevated neuronal H₂S due to trisomy 21 (containing the CBS gene) contributes to Down syndrome pathologies (56). Prolonged exposure is also linked to CNS and respiratory depression and even death. Therefore, endogenous H₂S synthesis and degradation must be tightly regulated. Since the discovery of endogenous H₂S production (57), the mechanisms preventing its overproduction have remained unclear. Here, we describe a downregulatory mechanism of H₂S levels that does not involve its decomposition, but rather on its production, maintaining H₂S within its beneficial range. Specifically, regulation by acetylation, a reversible posttranslational modification, allows rapid adaptation of H₂S levels, providing plasticity to support homeostasis and healthy lifespan. Further studies are needed to understand how PTMs adapt to environmental changes to mediate longevity as suggested (58). Interestingly, recently we found that acetylated Cbs K386 is abolished in long-living mammals by the replacement of the acetylated lysine to arginine (58). In comparison to old WT mice liver, a nonsignificant trend of increased acetylation (1.9 fold change, P = 0.2, Dataset S3) was found. This suggests that Sirt6 OE might have a minor effect on the activity of Cbs in old mice through K386 acetylation. Nevertheless, further research is needed to determine the exact role of this acetylation in Sirt6-mediated longevity.

Interestingly, Sirt6 phenotypes are more pronounced at old age, as seen here for liver H₂S production and in glucose levels during fasting or on promoting physical activity at old age (6). This suggests that its activity is required to respond to stress conditions, as found in aging. During aging both Sirt6 activity and levels decrease (6), as refelected by the age-dependent increase in Slc7a11 and SAM levels. Yet, in Sirt6 TG mice the continuous activity of Sirt6 maintains young-like levels of SAM and Slc7a11 (Figs. 2A and 5C, respectively). Still, it is an open question whether inducing higher Sirt6 levels at very old age would be enough to reverse these phenotypes.

Although Sirt6 is primarily a nuclear histone deacetylase with minor ER localization, most differentially acetylated proteins we detected were cytoplasmic or mitochondrial. Sirt6 has been reported to translocate to the cytosol in response to palmitate, viral infection, and other stimuli, where it may deacetylate new, nonhistone substrates. Under such stress, it may regulate cytosolic proteins. However, this cannot explain the 45% of Sirt6-dependent proteins with increased acetylation, or its impact on mitochondrial protein acetylation. This regulation is likely mediated by other specific lysine deacetylases or acetyltransferases. Sirt6 also controls hepatic acetyl-CoA levels through regulation of ATP-citrate lyase expression. In parallel, it promotes fatty acid beta oxidation via PPARα, both contributing to acetyl-CoA levels. Elevated acetyl-CoA may lead to nonenzymatic acetylation of mitochondrial proteins. Moreover, since increased acetyl-CoA was shown to promote nonenzymatic histone acetylation, this may also explain the elevated nuclear protein acetylation observed in Sirt6 TG mice.

As stated above, we speculate that Sirt6 activation or inhibition of unknown KAT/s or KDAC/s, respectively, mediates increased Sirt6-dependent acetylation. Yet, to date, no known consensus exists for any KAT or KDAC. Indeed, no specific motif was found for the global old liver acetylome. This is probably due to the fact that this acetylome involves the redundant activities of KDACs or KATs. However, the Sirt6-dependent acetylome does show a partial motif (Fig. 3 D–F). Specifically, in the motif of Sirt6-dependent increased acetylation, the underrepresentation of serine, and also threonine (with lower significance), upstream of acetylated lysine is particularly apparent. This suggests that a search for such hypothetical Sirt6-regulated KAT/KDAC should focus on those that exhibit selectivity toward nonphosphorylated sites adjacent to the acetylation site. Additional filters that may be applied in such a search would be their activation upon physiological conditions known to upregulate Sirt6, e.g., starvation, oxidative stress, and DNA damage.

During fasting, hepatic SAM levels decrease while SAH levels increase, indicating enhanced SAM consumption. In old Sirt6 TG mice compared to WT, SAM levels were reduced, but without a CR-like increase in SAH. This pattern suggests reduced Matα1 activity rather than elevated SAM usage. We found that Sirt6 reduces Matα1 K235 acetylation, thereby lowering SAM production and preserving young-like SAM levels. Whether CR or fasting also influence Matα1 K235 acetylation remains to be determined.

TSP activation by SAM is a crucial health necessity, as Cbs insensitivity to SAM causes homocystinuria (39, 59). However, no direct connection between Matα1, Sirt6, and Cbs was shown until now. Sirt6-dependent control over Matα1 activity, and its binding to Cbs, through K235 deacetylation adds a direct interaction between the methionine cycle and the TSP. One may hypothesize that H₂S production should increase even further under the overexpression of an inactive Sirt6 H133Y mutant, which decreases Slc7a11, but do not decrease Cbs activation via Matα1 deacetylation. However, such an approach is not feasible, as the expression of an inactive SIRT6 mutant was shown to cause embryonic lethality in humans (21). Interestingly, such regulation of Cbs may also be performed by Matα1’s orthologue, Matα2, in other tissues. The high connectivity between the one-carbon enzymes raises the question of which other Sirt6-dependent acetylations in this pathway are involved in the regulation of H₂S production.

Structural analysis of the Matα1 K235 site identified its location in an unstructured region (SI Appendix, Fig. S9). Similarly, the acetylations on p53 K382 and K539, or K542 on Ku70 are also within flexible unstructured linkers, and affect their activity and interaction with other proteins. Thus, K235 acetylation may lead to a conformational change in this region, inducing a Matα1-Cbs association. In addition, this acetylation might also affect Matα1 activity by either altering tetramer formation or its binding to ATP.

Similar to H₂S, SAM levels should be maintained in defined ranges. For example, though increased SAM is shown here in a context of aging, in muscle stem cells depletion of SAM leads to loss of heterochromatin and drives aging (60). Thus, the acetylation-based regulation of Sirt6 on SAM production suggests an avenue for Sirt6-mediated healthy longevity. Further study is needed to determine additional regulatory layers of Sirt6 on SAM production, and to link SAM levels to specific Sirt6-related positive effects.

Altogether, we show that Sirt6 preserves H₂S and SAM homeostasis in old age. Thus, its activation can achieve the beneficial effects of CR without dramatic dietary changes. However, the mechanisms underlying the effects of CR and Sirt6 do not overlap fully. Thus, future therapeutic approaches should consider the combination of CR with Sirt6 activators to attain a synergetic effect enhancing a healthy lifespan.

Materials and Methods

All mice procedures were approved by the Institutional Animal Care and Use Committee of Bar Ilan University. Detailed descriptions of animal experimentation, PAG injection and IGF-1 measurement, mouse primary hepatocytes isolation, cell culture and treatment, transfection, H₂S production capacity and concentration measurements, label free LC-MS/MS based acetylome and proteomics including computational analysis, cloning and mutagenesis, recombinant protein production, co-IP, SAM and SAH targeted metabolomics, Matα1 activity assay, RNA extraction and qRT-PCR, SP1 KO cells generation, WB analysis, structural analysis of Matα1 and lists of primers, antibodies, and materials used in this research are provided in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Data, Materials, and Software Availability

Proteomic and acetylome data are available in https://www.ebi.ac.uk/pride/archive/projects/PXD061654 (61).