Hyperglutaminolysis drives senescence and aging through arginine-mTORC1 axis activation

Abstract

The catabolism of glutamine is essential for living organisms, so that its first step, driven by glutaminase 1 (GLS1), generally referred to as glutaminolysis, plays important roles in physiological metabolism. However, the status and impact of glutaminolysis in pathological contexts such as aging and age-related diseases remain elusive. In this study, through metabolomics analysis and different aging models, we verified the hyperactivation status of glutaminolysis in senescent cells and aged Drosophila and mice, which we term “hyperglutaminolysis”. We further confirmed the aging-promoting role of this hyperglutaminolysis by addition and removal intervention experiments. Intriguingly, a novel signaling axis connecting to senescence-associated persistent mTORC1 activation was found. This pathway begins with glutaminase-catalyzed production of ammonium and glutamate, which drives arginine biosynthesis and is subsequently sensed by CASTOR1, leading to persistent mTORC1 activation. The regulatory roles of two key enzymes within this cascade, GLS1 and argininosuccinate lyase (ASL), were specifically investigated and verified by cellular and in vivo experiments, including those using stress-promoted and naturally aged animals, combined with GLS1 and ASL knockdown, and multiple rounds of metabolite analysis. In conclusion, our work positions dysregulated glutaminolysis as a key driver of aging and delineates a previously unrecognized molecular cascade that directly links glutaminolysis, arginine biosynthesis, and mTORC1 activation. These findings significantly expand our understanding of the relationship between glutamine catabolism and aging and are valuable for identifying novel intervention targets aimed at mitigating aging-related processes.

Introduction

Aging is characterized by a progressive loss of physiological integrity and function of organs and organisms. It is also the primary risk factor for aging-related diseases.^1–3 Recent studies further confirm that cellular senescence is a defined hallmark of organismal aging, and its accumulation plays a crucial role in aging progression.^1,2,4,5 Not only, the elimination of senescent cells has also been shown to improve multiple age-related diseases in animal models.^6–8 With the rapid development of geriatric medicine and aging biology, increasing evidence reveals the close link between the aging progression/senescence development and the dysregulated cellular metabolism, especially the dysregulated metabolism of glucose and lipids.^9–11 However, many gaps remain on our way to make clear the mechanism underlying the aging-related metabolic dysregulation, which largely hinder the geriatric research and the development of human health. More specifically, we know less about amino acid metabolism in senescent cells, compare to that of glucose and lipids metabolism, and much less about the status and impact of glutamine (Gln) metabolism on aging and senescence.

Glutamine, the most abundant free amino acid in human bodies, acts as the major carbon source and nitrogen source in many organisms.^12–15 It plays roles in various cellular processes through its catabolism.^14,15 The first and rate-limiting step of glutamine catabolism is known as glutaminolysis.¹⁵ This process, catalyzed by glutaminase (GLS), converts glutamine into glutamate and ammonium.^15,16 These products then enter branched cascades of metabolic reactions.^17,18 Glutamate is further metabolized into α-ketoglutarate, which fuels the tricarboxylic acid (TCA) cycle and the glutathione (GSH) generation, as well as some nonessential amino acids synthesis, such as the synthesis and aspartate and alanine.^14,15,17 On the other hand, ammonium is assimilated into different pathways, including those linking to the urea cycle and nucleotide synthesis, notably contributing to the synthesis of citrulline.^15,18–20 Although the interplay between these metabolites and pathways has not been precisely explored, its impacts on cell metabolism can be imagined. In this unclear map, we noticed one issue, that is, two amino acids included in the downstream metabolites of glutaminolysis, aspartate and citrulline, are synergistically involved in arginine biosynthesis.^14,18,19,21 Despite being somewhat overlooked before, this implication raises the question of whether glutaminolysis might influence arginine biosynthesis. Considering arginine has been recognized for its capacity to activate mTORC1 and persistent mTORC1 activation is a well-documented inducer event of cellular senescence,^22–24 we are interested to elucidate the relationship between glutaminolysis and mTORC1 activation, with particular attention to investigate if the senescence-associated reprogramming of glutaminolysis is connected with the senescence-associated persistent mTORC1 activation, i.e. aberrant mTORC1 activation, through a way involving arginine biosynthesis.

Glutaminolysis is often endowed with antioxidant properties and anti-aging potential, basically standing on the beneficial effects of glutamine supplementation in the elevation of GSH levels and the relieve of oxidative stress in vivo.^25–27 However, the evidence supporting the anti-aging potential of glutaminolysis is not solid enough, at least in our view. For example, repeated studies demonstrate that glutamine level is not decreased in senescent cells and tissues of aged individuals or those with aging-related diseases.^28–34 Moreover, enhanced glutaminolysis and high levels of its downstream metabolites, such as glutamate, ammonium and α-ketoglutarate, have been observed in human and murine tissues of aged contexts.^29,35–39 Additionally, a recent study reports that glutamine metabolic flux is significantly elevated in aged mice.³⁹ These findings strongly suggest the necessity to comprehensively explore the impact of glutaminolysis in aging process from perspectives beyond GSH synthesis and TCA cycling. In fact, despite GSH generation, other downstream metabolic pathways of glutaminolysis have not been extensively studied, specifically in the area of aging research.

This study delves into the intricate relationship between glutaminolysis and aging by sequentially examining the aging-specific status of glutamine catabolism, the roles of glutamine and glutaminolysis in aging progression, and their impact on the arginine biosynthesis-mTORC1 pathway. Our findings based on cellular and animal experiments demonstrate that hyperglutaminolysis status plays pro-aging roles in insect and mammalian cells, via a way connecting glutaminolysis, arginine synthesis and persistent mTORC1 activation. This work broadens our perspective on the impacts of glutamine and glutaminolysis on aging progression with mechanistic investigation. Furthermore, it also provides new insight into how altered glutaminolysis contributes to cellular dyshomeostasis, particularly through mTORC1-driven mechanisms.

Results

Alteration of glutamine metabolism in senescent cells

To elucidate the metabolomic signature of senescent cells, we employed two stress-induced premature senescence (SIPS) models and one replicative senescence model (Supplementary Fig. S1a–c). Metabolomics assay allowed us to identify certain water-soluble compounds that exhibit significant abundance changes in senescent cells compared to their proliferating counterparts. Among 147 metabolites detected, 88 exhibited an increase while 7 displayed a decrease in H₂O₂-induced senescent cells. Notably, the majority of α-amino acids were found to be upregulated (Fig. 1a; Supplementary Table 1). KEGG pathway enrichment analysis further underscored the importance of α-amino acid metabolism, with 19 related pathways ranking among the top 30 enriched pathways (Fig. 1b). These findings highlight the profound influence of α-amino acid metabolic reprogramming on the overall metabolic alterations associated with senescence.

Fig. 1.

Glutaminolysis alters in senescent cells and aged animals. a Volcano plot showing significantly changed (p ≤ 0.05) metabolites in H₂O₂-induced senescent cells. Upregulated amino acids are listed on the right (n = 6 per group). b Top 30 enriched KEGG pathways associated with the significantly changed metabolites in senescent cells. Red bars indicate amino acid-related pathways, white bars indicate amino acid-irrelated pathways. Numbers inside bars represent log₁₀(FDR). c The network of metabolites included in the 19 α-amino acid-related KEGG pathways; amino acids with adjusted centrality values ranked in the top 10 metabolites are labeled. d–g Glutamine (Gln) consumption rate and glutaminase expression/activity in senescent and control cells (n = 3 per group). Gln consumption rates (d), images and relative folds of GLS1 protein level (e, f), and GLS activity levels (g). Direct products of glutaminolysis in cells. NH₄⁺ levels (h) and glutamate levels (i). Senescent NIH3T3 cells were induced by H₂O₂ and adriamycin (ADR) treatments, senescent hFB cells were produced by 19 serial passages, and their proliferative counterpart controls were labeled separately as PBS, DMSO and P10 (n = 3 per group). j Relative GLS activities in naturally aged (Day 60) and aging-like (H₂O₂) flies and their younger counterparts; pooled homogenates of 20 flies for each group are used with n = 3 technical replicates. k Related GLS activity in kidney, spleen and muscle tissues from young (3 months) and old (27 months) mice; pooled tissue homogenates (4–5 mice per group) were analyzed with n = 5 technical replicates. All histogram data are shown as mean ± SD. Student’s t-test was performed: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001

A network encompassing 74 metabolites from these 19 α-amino acid-related pathways was subsequently constructed. In this network, eight amino acids, highlighted in red, exhibited higher adjusted centrality values, securing their positions within the top 10 (Fig.1c). Among them, glutamine emerged as the most central metabolite, displaying the highest adjusted centrality value (Fig.1c). This observation reveals significant alterations in amino acid metabolism, particularly glutamine metabolism, in senescent cells.

Hyperglutaminolysis occurs in senescent cells and aged animals

To validate the reprogramming of glutamine metabolism in senescent cells, we quantified glutamine consumption rates in three distinct senescent cell models. As shown in Fig. 1d, higher consumption rates of glutamine were exhibited in all these senescent cells, compared to their proliferative counterparts.

Given the reliance of glutamine consumption on its catabolic process, we investigated the state of glutaminolysis, the initial and crucial step of glutamine catabolism, by examining the expression and activity of glutaminase (GLS), especially focusing on the GLS1 isoform as it is widely expressed in mammalian cells.^40,41 Our data revealed an upregulation of GLS1 protein levels and an increase in overall GLS activity in senescent cells (Fig. 1e–g). Additionally, we monitored cellular levels of glutamate (Glu) and ammonium (NH₄⁺), the two direct products of glutaminolysis. As anticipated, both metabolites exhibited a marked elevation in senescent cells (Fig. 1h, i). Together, these findings demonstrate the upregulation of glutaminolysis in senescent cells, a metabolic state we define as hyperglutaminolysis.

To investigate the in vivo relevance of hyperglutaminolysis, we first assessed glutaminolysis states in naturally aged flies and aging-mimic flies modeling by chronic H₂O₂ exposure (H₂O₂-stressed, Supplementary Fig. S1d–f). Consistent with our results based on senescent cells, GLS activities were increased in the bodies of aged flies and H₂O₂-pre-exposed flies, compared to their counterparts without aging (Fig. 1j). Furthermore, hyperglutaminolysis was observed in the bodies of naturally aged mice. Specifically, GLS activities were markedly higher in kidney, spleen, and muscle tissues of 27-month-old mice compared to those of 3-month-old mice (Fig. 1k; Supplementary Fig. S1g-j).

The above findings collectively demonstrate that hyperglutaminolysis is a conserved metabolic state in aging models, from murine and human senescent fibroblasts, to aged flies and mice.

Glutaminolysis restriction alleviates cellular senescence and fly aging

Then, we delved into the functional impact of hyperglutaminolysis on senescence and aging progression. We employed three distinct approaches to inhibit glutaminolysis for in vitro evaluation.

First, glutamine-deprived DMEM supplemented with 10% FBS (low-glutamine medium cultivation) was applied to NIH3T3 cells together with H₂O₂ or ADR treatments, or hFB cells during their 18th to 19th passages, i.e., in the period of senescence induction. The low-glutamine medium cultivation treatment markedly reduced the proportions of SA-β-gal-positive cells in both stress-induced and replicative senescent populations (Fig. 2a, b). Second, the GLS inhibitor loading approach was adopted. Similar to low-glutamine medium cultivation, two GLS inhibitors (DON and CB-839) effectively reduced the proportions of SA-β-gal-positive cells in senescent populations (Fig. 2c, d; Supplementary Fig. S2a, b). In addition, p16 expression was also markedly reduced while low-glutamine medium cultivation GLS inhibitors was applied (Fig. 2e, f; Supplementary Fig. S2c, d, S3a, b)

Fig. 2.

Glutaminolysis restriction alleviates cellular senescence and fly aging. a–f Cell senescence was induced by H₂O₂ or ADR treatment (NIH3T3), or serial passaging (hFB); some senescent cells were adapted to cultivation in low glutamine medium (LQ, a, b, e) or 10 μM Diazo-5-oxo-L-nor-leucine (DON, c, d, f); proliferative and senescent cells with normal cultivation served as controls. SA-β-gal staining (a, c), quantitation of SA-β-gal-positive cells (b, d) and representative immunoblotting images of p16 protein (e, f). siRNA for GLS1 gene (siGls1) and scrambled siRNA (siNC) were transfected into NIH3T3 cells and some cells were brought into senescence (H₂O₂) induction. RT-qPCR-based Gls1 knockdown verification (g), SA-β-gal staining (h), quantitation of SA-β-gal-positive cells (i) and representative immunoblotting images for p16 protein (j). Gls gene knockdown (Gls-KD) and wild-type (Control) flies were included. RT-qPCR assay for Gls mRNA monitoring (k), GLS activities in whole body homogenates (l), cumulative survival probability (CSP), median and maximum lifespans (m), climbing scores, Smurf-positive rates and egr (homologous gene of TNF in fly) mRNA levels of 90-day-old flies (n–p). Wild-type flies were pretreated with 1% H₂O₂ for three days to induce accelerated aging, followed by combined treatment with H₂O₂ (1%) and DON (0.1 μM). Test data include cumulative survival probability (CSP) and lifespans (q), climbing scores (r) and Smurf-positive rates (s). For lifespan assays, each group contained 100 flies; for climbing, Smurf staining and RT-qPCR assays, each group contained 15–20 flies. All assays were repeated three times. Scale bar = 50 µm. All histogram data are shown as mean ± SD. Student’s t-test: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Kaplan-Meier log-rank test p-values are shown

Third, the way of silencing Gls1 expression, by transfecting cells with siRNA targeting GLS1 (siGls1), also reduced the proportion of SA-β-gal-positive cells and p16 expression in senescent populations. (Fig. 2g–j; Supplementary Fig. S3c).

Meanwhile, we investigated the influence of glutaminolysis restriction on SASP expression, another set of senescence markers. Unsurprisingly, the expressions of several SASP factors (Cxcl10, Tnf, Il6 and Mmp9) were significantly suppressed by low-glutamine medium cultivation and GLS inhibitors in both H₂O₂- and ADR-induced senescent cells (Supplementary Fig. S2e, f, S3d, e).

To further validate, we gathered in vivo evidence by two experimental paradigms. The first one is Drosophila melanogaster with RNAi-mediated knockdown of GLS (GLS-KD) (Fig. 2k, l). As shown in Fig. 2m, GLS-KD flies not only displayed extended median (102 vs. 59 days) and maximum (147 vs. 116 days) lifespans, but also had a significantly improved cumulative survival probability (CSP, log-ranked p value: 1.0168 × 10⁻²⁷). Moreover, these flies exhibited enhanced physical activity, as evidenced by increased climbing scores, and a reduced Smurf phenotype, a marker of intestinal permeability and a hallmark of aging in Drosophila (Fig. 2n, o). In addition, the expression of egr, a homolog gene of Tnf (one of SASP factor gene), was also decreased in Gls-KD flies (Fig. 2p).

The second set of in vivo data was derived from studies on flies subjected to chronic H₂O₂-stress and treated with DON or CB-839. Both treatments significantly extended the median (DON: 9 vs 12 days; CB-839: 9 vs 11 days) and maximum (DON: 16 days vs 18 days; CB-839: 16 days vs 18 days) lifespans, compared to the unstressed control group (Fig. 2q; Supplementary Fig. S2g). The survival benefits of both treatments were further supported by an increase in the cumulative survival probability, as evidenced by the log-rank p-values (DON: 0.004063; CB-839: 0.034132) (Fig. 2q; Supplementary Fig. S2g). GLS inhibitor-treated flies also exhibited improved climbing capacity and a reduced Smurf phenotype (Fig. 2r, s; Supplementary Fig. S2h, i).

These results from in vitro and in vivo experiments collectively reveal the suppressive effect of glutaminolysis restriction on senescence and aging.

Glutaminolysis restriction-caused senescence alleviation does not accompany with senolytic cell death

Building on the senolytic effect of glutaminolysis inhibitors reported by a previous study,⁴² we investigated whether the repressive role of glutaminolysis restriction observed in our system is associated with the induction of senolytic apoptosis.

As indicated by the enhanced PI +/Annexin V+ and PI +/Annexin V- cell populations, glutaminolysis restriction triggered apoptosis when low-glutamine medium cultivation or GLS inhibitor DON was applied to senescent populations, i.e., after several days of senescence induction. This effect is similar as that of ABT-263, a well-established senolytic agent (Supplementary Fig. S4a, b), and consistent with the demonstration of Johmura et al.⁴² However, glutaminolysis restriction did not cause apoptotic cell death when low-glutamine cultivation was applied from the onset of senescence induction (Supplementary Fig. S4c, d).

Consistent results were obtained by TUNEL staining assays. It shows that, like ABT-263, glutaminolysis restriction only increased TUNEL-positive cells when low-glutamine medium cultivation and DON were administered to cell populations after senescence induction (Supplementary Fig. S4e, f), but not to cells under senescence induction (Supplementary Fig. S4g, h).

These results suggest that glutaminolysis restriction caused amelioration of senescence detected in our study mainly relied on its blocking effect on senescence induction but not on its senolytic effect.

Hyperglutaminolysis contributes to aberrant mTORC1 activation in senescent cells

mTORC1 is known to be aberrantly activated in senescent cells.^43,44 However, the relationship between hyperglutaminolysis and mTORC1 in the context of senescence remains unclear. Given that both processes represent significant metabolic changes associated with aging, we hypothesized a potential link between them. Our initial investigations confirmed the aberrant mTORC1 activation in our senescence model, as evidenced by elevated phosphorylation of p70/S6K and 4EBP1 under serum starvation conditions (Fig. 3a, b).^43,44

Fig. 3.

Glutaminolysis activates mTORC1 and suppresses autophagy via mTORC1 activation. Representative immunoblotting images (a) and quantitation (b) for phosphorylated p70/S6K (P-p70/S6K) and 4EBP1(P-4EBP) in NIH3T3 cells with 24 hours of serum starvation. Pro: proliferating cells, Sen: senescent cells induced by H₂O₂. Senescence induction was combined with 10 μM DON or siGls1 treatment for three days in NIH3T3 cells. mTORC1 activity in senescent cells was examined by testing p70/S6K and 4EBP1 phosphorylation (c, d) and mTOR-LAMP2 colocalization (e, f). Autophagy flux was examined by detecting RFP+/GFP- (red) puncta in senescent populations stably expressing LC3-RFP-GFP protein (g, h), and p62 protein abundance in senescent cells, without (i, j) or with (k, l) 25 nM BafA1 treatment combined during the last 24 hours. m–q Autophagy flux in senescent cells under conditions with 20 mM glutamine loading (Gln, 24 hours), or mTORC1 inactivation caused by siRNA for Raptor (siRap) and rapamycin (Rapa, 0.1 μM) treatments. m RT-qPCR based Raptor knockdown verification. RFP + /GFP- cell ratio (n, o), and p62 protein abundance (p, q). p70/S6K phosphorylation levels in serum-starved proliferating (Pro) and H₂O₂-induced senescent cells (Sen), with 2 μM DON treatment for 2 hours (r, s), and 20 mM Gln treatment for 6 hours (t, u). All assays were conducted three times. Scale bar = 50 μm. All histogram data are shown as mean ± SD. Student’s t-test: * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, n.s. non-significant

We then delved into the ability of glutaminolysis inhibitors, DON and siGls1, to modulate mTORC1 activity. A significant reduction in the phosphorylation levels of p70/S6K and 4EBP1 was observed following treatments with these inhibitors (Fig. 3c, d). Furthermore, the lysosomal localization of mTOR was diminished, as exhibited by the reduced colocalization of mTOR with lysosome-associated membrane protein 2 (LAMP2) (Fig. 3e, f). These results suggest a decrease in mTORC1 activity.

Considering the established link between mTORC1 dysregulation and autophagy impairment in senescent cells,²⁴ we further evaluated the effect of glutaminolysis restriction on autophagic flux. Using NIH3T3 cells stably expressing RFP-GFP-LC3 tandem protein, we observed that DON and siGls1 both increased red LC3 puncta in normally cultivated cells, indicative of enhanced autophagic flux (Fig. 3g, h). Concurrently, the accumulation of p62, a marker of impaired autophagic flux, was markedly reduced in senescent cells that received these treatments (Fig. 3i, j). To confirm the effect of DON and siGls1 on autophagic flux enhancement, we included Bafilomycin A1 (BafA1)-treated groups, since it works as an autolysosomal inhibitor. As expected, BafA1 impeded the reduction in p62 accumulation caused by DON and siGls1 (Fig. 3k, l).

Another line of obstructive experiments was conducted to validate whether mTORC1 suppression modulates the role of hyperglutaminolysis in autophagy flux. Results showed that siRNA of Raptor, a unique protein of mTORC1 complex, significantly increased the rate of red LC3 puncta and reduced p62 accumulation levels in senescent cells treated with 20 mM glutamine (Fig. 3m-q). Similar effects were observed in cells treated with Rapamycin (Fig. 3n-q).

Results of these experiments demonstrate that hyperglutaminolysis significantly contributes to aberrant mTORC1 activation in senescent cells, and that inhibiting glutaminolysis can decrease mTORC1 activation and enhance downstream autophagy flux.

Further, we compared the influence of glutaminolysis on mTORC1 activation in senescent cells versus proliferating cells. Results revealed that DON treatment reduced the phosphorylation of p70/S6K in senescent cells but not in proliferating cells (Fig. 3r, s). Conversely, high concentration of glutamine supplementation induced a marked increase of p70/S6K phosphorylation exclusively in senescent cells (Fig. 3t, u). These findings demonstrate that the impact of glutaminolysis on mTORC1 activation is more pronounced in senescent cells than in proliferating cells, suggesting that hyperglutaminolysis occurring in senescent cells is a relative leading metabolic disorder involved in cell dyshomeostasis and senescence development.

Hyperglutaminolysis results in enhanced arginine biosynthesis

To precisely understand the molecular cascade linking hyperglutaminolysis to mTORC1 activation, we performed a new round of metabolomics analysis to compare those metabolites associated with the glutaminolysis-arginine biosynthesis axis in senescent and proliferative cells, because arginine is a mTORC1 promoting amino acid.⁴⁵ The changes of these metabolites caused by GLS inhibitor CB-839 in senescent cells were measured together. This analysis detected 71 metabolites increased in senescent cells (versus proliferative cells), and 33 metabolites decreased in CB-839 treated senescent cells (versus untreated counterparts; Supplementary Table 2–4). Furthermore, 24 metabolites overlapped between these two datasets, with 12 of them implicated in KEGG pathways involving glutaminase (GLS; Supplementary Tables 2–4). Notably, arginine and its two-predecessor substances, aspartate and citrulline,^46–48 were included in these 12 metabolites (Fig. 4a, b).

Fig. 4.

Hyperglutaminolysis enhances arginine biosynthesis in senescent cells and aged flies. a Venn diagram of metabolites detected by LC-MS/MS-based metabolomics assay (n = 5 per group). These metabolites are classified into three subsets: the upregulated ones in senescent NIH3T3 cells (green, H vs C UP, p ≤ 0.05, fold change ≥ 1), the downregulated ones in CB-839 (1 μM, 36 hours)-treated senescent cells (red, HCB vs H DOWN, p ≤ 0.05, fold change ≤ 1), and those engaged in the “GLS-related pathway” (blue, GLS-related). 12 metabolites situated in the intersection of these 3 subsets are listed on the right side. b A sketch map of the arginine synthesis pathway linked to glutaminolysis. c–n Levels of metabolites involved in arginine synthesis detected by LC-MS/MS assays and CB-839-caursed alterations (delta levels) of these metabolites, in proliferating (Pro) and senescent (Sen) cells. These metabolites are glutamine (Gln, c, d), arginine (Arg, e, f), glutamate (Glu, g, h), ammonia (NH₄⁺, i, j), aspartate (Asp, k, l) and citrulline (Cit, m, n). Levels of arginine, aspartate and citrulline in naturally aged (o) and H₂O₂-stressed (p) flies, with young and unstressed controls. Arginine (q) and NH₄⁺ (r) levels in the control and H₂O₂-stressed flies treated without or with DON (0.1 μM, 3 days). s Levels of arginine, aspartate and citrulline in wild-type and Gls-KD flies. All metabolite assays used pooled homogenates of at least 20 flies per group and were executed with three technical replicates. Histogram data are shown as mean ± SD. Student’s t-test: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, n.s. non-significant

Following glutaminolysis, arginine can be synthesized with intermediate steps including aspartate synthesis and citrulline synthesis (Fig. 4b).^46–48 To explore the connection between the upregulated arginine biosynthesis and its dependence on hyperglutaminolysis in senescent cells, we quantified the levels of relevant metabolites in cells. As expected, glutamine level was increased in senescent cells and elevated by CB-839 treatment (Fig. 4c, d). Notably, the CB-839-elevated glutamine level was more pronounced in senescent cells than in proliferating cells, consistent with the status of hyperglutaminolysis in senescent cells (Fig. 1d–g). Likely, arginine level was higher in senescent cells. However, they were decreased in CB-839-treated senescent cells (Fig. 4e), that is match to its position under downstream of GLS inhibition. and had a more pronounced reduction upon CB-839 treatment compared to proliferative cells (Fig. 4e, f). The chances of glutamate, NH₄⁺, aspartate and citrulline, all of which are required for arginine biosynthesis, tended similarly like arginine in the condition with CB-839 treatment (Fig. 4g–n). However, glutamate levels did not rise in senescent cells at the assessed time point.

We extended this investigation to animal models. Compared to control groups, naturally aged and oxidatively stressed Drosophila exhibited elevated levels of arginine and its biosynthetic precursors (Fig. 4o, p). Notably, DON treatment significantly reduced the elevated levels of arginine and NH₄⁺ in oxidatively stressed flies (Fig. 4q, r). Supporting evidence was obtained by using Gls-KD flies, as the knockdown of Gls led to decreased levels of arginine, aspartate, and citrulline (Fig. 4s).

These results highlight the intricate connection between hyperglutaminolysis and the upregulation of arginine biosynthesis, emphasizing its relevance to senescent cells as well as aged flies.

The glutaminolysis-arginine axis contributes to mTORC1 activation

Next, we tried to evidence the role of glutaminolysis-driven arginine biosynthesis in mTORC1 activation and aging progression, and, thus, proliferative cells were used in this section. Firstly, experiments appending arginine and contributors to glutaminolysis-driven arginine biosynthesis were conducted. A condition of glutaminolysis restriction was set by DON treatment in order to highlight the role of these additives. Results showed that, compare to DON treatment alone, the supplementation of glutamate plus NH₄Cl, aspartate plus citrulline, and arginine all enhanced mTORC1 activation, indicating by elevated phosphorylation of p70/S6K and 4EBP1 (Fig. 5a, b). Secondly, we carried out gene silencing experiments, by targeting enzymes working for arginine biosynthesis, which are GOT2 (for aspartate synthesis), OTC (for citrulline synthesis), ASS1 and ASL (for arginine synthesis, Fig. 5c). In order to highlight the role of gene silencing, glutamine supplementation was applied to cells after a 24-hour low glutamine medium cultivation. As shown, lowered expressions of these enzymes suppressed mTORC1 activity, as evidenced by decreased p70/S6K phosphorylation and reduced colocalization of mTOR with LAMP2 (Fig. 5d–h). We also tested the effects of siRNAs for Ass1 (siAss1) or for Asl (siAsl) on arginine levels in cells, in a condition as descript in the section of Materials and Methods. The results from LC-MS/MS assay showed that arginine levels decreased in cells transfected with siAss1 or siAsl (Fig. 5i).

Fig. 5.

Glutaminolysis-driven arginine biosynthesis promotes mTORC1 activation in cells and mice. a–i Proliferating NIH3T3 cells were used for assays. Representative immunoblotting images (a) and quantitation (b) of p70/S6K and 4EBP1 phosphorylation in cells treated with 4 mM of the indicated metabolites alone or in combination for five hours after a one-hour DON (2 μM) pretreatment. Glu: glutamate, NH₄Cl: ammonium, Asp: aspartate, Cit: citrulline, Arg: arginine. Three biological replicates were tested. c A sketch map showing the enzymes involved in the glutaminolysis-arginine synthesis pathway. GLS1: glutaminase 1, OTC: ornithine transcarbamylase, GOT2: glutamic-oxaloacetic transaminase 2, ASS1: argininosuccinate synthase 1, ASL: argininosuccinate lyase. d RT-qPCR based detections for mRNAs of Got2, Otc, Ass1, and Asl genes in cells transfected with indicated siRNAs. p70/S6K phosphorylation (e, f) and mTOR-LAMP2 colocalization (g, h) in siRNAs-transfected cells cultured in fresh medium with 4 mM glutamine for 6 hours after a 24-hour period of glutamine starvation. i Arginine levels in cells transfected with indicated siRNAs. All cellular experiments were repeated three times. j–l Pooled homogenates of indicated tissues from 4-5 mice per group were assayed with at least three technical repetitions. Arginine levels in kidney, spleen and muscle of young (3 months) and old mice (27 months), measured by using pooled homogenates of tissues from 4-5 mice (j). p70/S6K phosphorylation levels in gastrocnemius muscle of mice (k, l). m–q Mice with AAV9-shAsl or AAV9-GFP injection (1 × 10¹¹ viral genomes) were applied for assays after one to three months of viral injection, with three mice per group. Fluorescence intensity in gastrocnemius muscle with AAV9-GFP injection (m). mRNA levels of the Asl gene in gastrocnemius muscle of mice infected with AAV9-shAsl (shAsl) and control mice (shNC) (n). Arginine levels (o) and p70/S6K phosphorylation levels (p, q) in gastrocnemius muscle. Scale bar = 50 μm. Histogram data are shown as mean ± SD. Student’s t-test: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001

In vivo evidence was also collected based on mouse experiments, by AAV-mediated Asl gene knockdown. First, arginine levels were high in kidney, spleen, and gastrocnemius muscle tissues from aged mice (27 months vs 3 months, Fig. 5j). Consistently, a higher p70/S6K phosphorylation level was also been observed in muscle tissues from the aged mice (27 months) compared to the young ones (3 months, Fig. 5k, l). It revealed that AAV-mediated Asl knockdown (shAsl) effectively decreased the levels of arginine and p70/S6K phosphorylation in gastrocnemius muscle tissue of mice, compared to wild-type mice (Fig. 5m–q).

These in vitro and in vivo results underscore the critical role of glutaminolysis-fueled arginine biosynthesis in mTORC1 activation.

Upregulated glutaminolysis-arginine axis promotes aging progression

To clarify the connection of hyperglutaminolysis-augmented arginine biosynthesis with aging progression, further experiments were conducted in vitro, that proliferative NIH3T3 cells were cultivated in media with extra amounts of metabolites situated in the glutaminolysis-arginine axis for 20 days. Results showed that the addition of high levels of these metabolites, grouped as glutamine alone, glutamate plus NH₄Cl, aspartate plus citrulline and arginine alone, increased SA-β-gal staining and SASPs expression in cell populations, revealing prompted senescence (Fig. 6a–c). Conversely, siRNA targeting Asl (siAsl) reduced SA-β-gal staining in H₂O₂-stressed cell population (Fig. 6d, e).

Fig. 6.

Glutaminolysis-driven arginine biosynthesis contributes to senescence development and lifespan shortening of flies. Representative SA-β-gal staining (a), quantitation (b) and mRNA levels of SASP genes (c) of NIH3T3 cells adapted to a 20-day treatment with 20 mM of indicated metabolites. Gln: glutamine, NH₄Cl: ammonia, Glu: glutamate, Asp: aspartate, Cit: citrulline, Arg: arginine. Representative SA-β-gal staining (d) and quantitation (e) of NIH3T3 cells transfected with siNC or siAsl and subjected to H₂O₂-stimulated senescence development for three days; proliferative cells served as a negative control. Cumulative survival probability (CSP) and lifespans of flies fed by food supplemented with 20 mM and 100 mM glutamine (f) or arginine (g). Panels f and g share the same control group. n = 100 flies per group. Three repetitive assays were performed. Scale bar = 50 μm. Histogram data are shown as mean ± SD. Student’s t-test: ** p ≤ 0.01, ***p ≤ 0.001. Kaplan-Meier log-rank p-values are shown, n.s. non-significant

The in vivo experiments showed that supplementation with glutamine or arginine shortened the lifespans of Drosophila (Fig. 6f, g), as indicated by decreases in median and maximum lifespans, and also lowered their cumulative survival probabilities (CSP).

Together, these in vitro and in vivo findings demonstrate that hyperglutaminolysis-elevated arginine biosynthesis acts essentially for aging progression.

Glutaminolysis-induced mTORC1 activation mediated by arginine-sensing CASTOR1

Given that CASTOR1 is a unique arginine sensor protein and works for mTORC1 upregulation, we assessed the action of CASTOR1 on hyperglutaminolysis-fueled mTORC1 activation. We found that silencing of CASTOR1 by siCas1 attenuated the inhibitory effects of DON and siGls1 on mTORC1 activity and blocked their ability to improve autophagy flux (Fig. 7a-i). In addition, siCas1 diminished the ability of DON and siGls1 to ameliorate senescence (Fig. 7j–m). Conversely, CASTOR1 overexpression (Cas1 OE) suppressed p70/S6K phosphorylation induced by glutamine supplementation following a 24-hour glutamine starvation (Fig. 7n–p). These results suggest that glutaminolysis-induced mTORC1 activation is mediated by the arginine-CASTOR1 cascade.

Fig. 7.

CASTOR1 mediates the effects of hyperglutaminolysis on mTORC1 activation, autophagy suppression and senescence development. a–m siRNA for CASTOR1 (siCas1 + ) and scrambled siRNA (siCas1-) transfected NIH3T3 cells were adopted for senescence induction and used for assays, and a portion of siCas1+ cells treated with 10 μM DON or transfected with siGls1. RT-qPCR-based Castor1 knockdown verification (a). p70/S6K phosphorylation exhibited by immunoblotting assay (b, c), mTOR-LAMP2 colocalization indicated by immunofluorescence assay (d, e), RFP+/GFP- cells f, g and p62 protein abundance (h, i), SA-β-gal staining (j, k) and p16 protein abundance (l, m), n–p Castor1 was overexpressed in HEK393T cells via lentivirus-infection approach (Cas1 OE), with empty virus infection as the control (Vector). Data include mRNA levels of Castor1 in infected cells (n). p70/S6K phosphorylation levels in Cas1 OE cells and Vector cells, and portions of these cells were subjected to a 24-hour glutamine starvation followed by a 2-hour 4 mM glutamine supplementation (o, p). All experiments were conducted at least three times and with three biological replicates per experiment. Histogram data are shown as mean ± SD. Scale bar = 50 μm. Student’s t-test: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001

Discussion

This study identified a novel metabolic feature of senescent cells- hyperglutaminolysis, and validated its contribution to aging development. It further revealed that this upregulated glutamine catabolism accompanies enhanced arginine biosynthesis, aberrant mTORC1 activation, and autophagy suppression. Through mechanistic analysis, we demonstrate that these events are connected by a special molecular cascade, which starts from elevated GLS activity, leading to increased generation of ammonium and glutamate, then to enhanced arginine biosynthesis, which results in mTORC1 activation and autophagy impairment, and ultimately senescence/aging development (Fig. 8). These findings highlight a distinct role and mechanism of over-activated glutaminolysis in aging acceleration, and establish a new signaling pathway associated with the glutaminolysis-arginine biosynthesis-mTORC1 axis.

Fig. 8.

Aging-promoting role of hyperglutaminolysis in cells and the underlying mechanism. Overactivated glutaminolysis occurs in senescent cells, aged flies and mice, and plays an aging-promoting role. Mechanistically, this aging-promoting role is mediated by hyperglutaminolysis-induced enhancement of arginine biosynthesis and aberrant mTORC1 activation. This novel signaling pathway enriches our understanding of aging progression, and establishes a new connection between the dysregulation of glutamine catabolism and disordered mTORC1 activation, which is powerful to disturb whole cell homeostasis and acts pathologically for overall metabolism. This image was created using Adobe Illustrator and has been optimized and polished by Gemini 3.0

Glutamine metabolism, particularly glutaminolysis, plays an important role in physiological processes, such as TCA cycle processing, GSH synthesis, and amino acid metabolism.^14–16 Among these functions, GSH synthesis has been extensively studied in the field of aging research due to GSH is one of the most critical endogenous antioxidants in mammalian cells.^49,50 The role of glutaminolysis in TCA cycling also supports its anti-aging function owing to its contribution to mitochondrial homeostasis.¹ However, a pivotal shift occurred in 2021 when Johmura and colleagues demonstrated the senolytic effect of glutaminolysis inhibitors.⁴² This study displayed the mitigative role of glutaminolysis inhibition in aging, mechanistically through reducing cytoplasmic ammonium levels and elevating cytoplasmic pH, which promotes senescent cell death. Nevertheless, the downstream molecular pathways linking the glutaminolysis-ammonium axis to senescence development and the central hub of cellular homeostasis, such as mTORC1, remain unexplored. In 2022, Choudhury and colleagues proposed another mechanism underlying the aging protection effect of glutaminolysis inhibition, regarding the alleviation of NH₄⁺/urea accumulation-induced mitochondrial dysfunction.⁵¹ They identified three contributors to NH₄⁺/urea accumulation in senescent cells, which are upregulated glutaminase 1 (GLS1, for ammonia production), arginase 2 (ARG2, for urea synthesis), and downregulated urea transporter (SLC14A1, for urea efflux). However, the molecular connection from NH₄⁺/urea accumulation to mitochondria dysfunction has not been addressed clearly in their study. We hold the opinion that two questions need to be answered in near future: (1) whether urea accumulation occurs in our system, and (2) whether the glutaminolysis-ammonium-mTORC1 axis interacts with the glutaminolysis-ammonium-urea axis to drive aging.

Our interest in the interplay between glutaminolysis and aging was initially driven by the contrasting changes of two key enzymes in senescent cells and aged tissues. They are glutaminase (GLS), the rate-limiting enzyme for glutaminolysis, and glutamate-cysteine ligase (GCL), the rate-limiting enzyme for glutathione (GSH) synthesis. While GLS expression/activity increases with aging,^35–37 GCL activity declines.^52,53 This opposing trend reminds us that hyperglutaminolysis may not effectively support GSH production but serve alternative roles in aged systems. Thus, aiming to uncover these alternative roles, we investigated the impact of hyperglutaminolysis on amino acid metabolism, particularly arginine biosynthesis, in aged conditions, since the direct products of glutaminolysis, glutamate and ammonium, converge at arginine biosynthesis through distinct routes: glutamate metabolism yields aspartate and ammonium metabolism involves in citrulline formation, whereas both aspartate and citrulline are precursors for arginine synthesis (Fig. 4b).^46,47 Given that arginine is an assured activator of mTORC1 and mTORC1 dysregulation is a hallmark of senescence,^22,54 our investigations established a fresh connection that bridges hyperglutaminolysis to aberrant mTORC1 activation in aging, that is, the glutaminolysis-arginine-mTORC1 axis. Obviously, our findings propose a novel mechanism underlying the relationship of glutaminolysis with aging, by emphasizing glutaminolysis-fueled mTORC1 activation. Notably, while the current study demonstrates the aging-promoting effects of hyperglutaminolysis, our prior work revealed that chronic glutamine deprivation actually induces cellular senescence (Fig. S5),⁵⁵ and p16 level also increased when siGls1 applied to proliferative cells (Fig. 2j; Supplementary Fig. S3c). These seemingly paradoxical observations suggest that dysregulated glutamine catabolism—whether overactivated or suppressed—is detrimental to healthy aging. Based on existing knowledge and our hypothesis, these phenomena likely operate through distinct mechanisms. On the one hand, glutamine deprivation forces cells into a stressful state with energy depletion and glutathione (GSH) deficiency, which ultimately cause redox imbalance and senescence. Conversely, hyperglutaminolysis generates excessive glutamate/ammonium and enhances arginine biosynthesis, driving mTORC1 hyperactivation and subsequent pro-senescence pathways. Undoubtedly, it is interesting to continue studying the dynamics and balance of such reliance on glutamine during senescence.

Arginine-sensing mTORC1 activation was first elucidated in 2004 by Ban et al.⁴⁵, who identified CASTOR1 as an arginine sensor. They found that arginine binding to CASTOR1 releases the latter’s inhibitory effect on mTORC1 activation.²² Our study extends this paradigm by revealing the fact that arginine sensing of CASTOR1 is a critical procedure for hyperglutaminolysis-fueled aberrant mTORC1 activation in senescent cells. To our knowledge, this work provides the first experimental evidence delineating a comprehensive molecular cascade that connects glutaminolysis, arginine biosynthesis, arginine sensing of CASTOR1, and aberrant mTORC1 activation together. This cascade has not been previously recognized in the context of aging and provides a new framework for understanding the role of metabolic dysfunction in aging processing. To date, only a few studies have explored potential mechanisms of aberrant mTORC1 activation, particularly in senescent cells. This activation is aberrant because it is sustained and cannot be suppressed by serum starvation and growth factor deficiency, which is different from the mTORC1 activation observed in proliferating cells.^43,44 Carroll et al. proposed that mTORC1 activity in senescent cells is resistant to serum and amino acid starvation due to plasma membrane depolarization and impaired autolysosomal function.⁴⁴ Liang et al. demonstrated that branched-chain amino acid accumulation could fuel aberrant mTORC1 activation in senescent cells.⁵⁶ Moreover, Xiong et al. evidenced that mTORC1 activation in senescent cells was associated with the upregulated ARG2 expression, and this effect was dependent on the non-enzymatic function of ARG2.⁵⁷ The hyperglutaminolysis-arginine-mTORC1 axis we identified in this study offers a new perspective pointing to the metabolic origins of aberrant mTORC1 activation. We propose that sustained hyperglutaminolysis in senescent cells acts as a metabolic driver, initiating arginine biosynthesis, activating the arginine/CASTOR1-mediated mTORC1 axis, and ultimately leading to growth factor-independent, persistent mTORC1 activation. Based on these findings, we propose that dysregulated amino acid metabolism, particularly hyperglutaminolysis, serves as a continuous stimulus for mTORC1 in cellular senescence. This occurs through a molecular cascade involving elevated arginine biosynthesis and sustained inhibition of CASTOR1, offering a mechanistic explanation for the metabolic roots of mTORC1 dysregulation in senescence.

Consistent with our study, other investigations also demonstrated the aging-associated elevation of arginine metabolism.^58,59 Notably, however, the detected levels of arginine under aging conditions are inconsistent,^{32,58,60–62}, and there is substantial evidence showing the upregulation of arginase (ARG), either expression or activity, in senescent cells and aged tissues, and some of them suggest further the aging-promoting role of this ARG upregulation.^57,63 Two questions arose then. The first, can ARG activation can significantly counteract glutaminolysis-driven arginine accumulation and therefore suppress mTORC1 activity in the senescence context? Second, whether and how does the arginine metabolism reprogramming in the senescent context? Although ARG1/2 expression was elevated in our senescent cells (data not shown), arginine accumulation was consistently detected concurrently. It suggests that, compared to ARG-catalyzed arginine decomposition, glutaminolysis-fueled arginine production probably plays predominantly for the steady level of arginine in our system. Without doubt, more precise investigations on the state of arginine flux under aging conditions are required.

We would like to briefly discuss the inconsistent changes in NH₄⁺ and glutamate levels in senescent cells, exhibiting as NH₄⁺ level elevated persistently, whilst glutamate level only increased in a comparatively short period (Fig. 1h, I; Fig. 4g, i). Given both NH₄⁺ and glutamate are produced directly by GLS-driven reaction, two potential explanations may account for this discrepancy. First, the turnover of glutamate may be more rapid than that of ammonia in cells, as it links to multiple metabolic processes with high flux, such as the TCA cycle and GSH synthesis.^14,15 In fact, our metabolomics analysis displayed a significant increase in two key downstream metabolites of glutamate, α-ketoglutarate and succinic acid, in senescent cells (data not shown). Considering these high-fluxed processes work for cells fundamentally, the steady level of glutamate should hardly remain high in cells. On the other hand, the cellular processes downstream of ammonia are relatively fewer and not linked to constant energy expenditure,⁶⁴ which makes the elevated level of ammonia easier to detect. Second, although GLS catalyzes the simultaneous production of one glutamate and one ammonia molecule per reaction, an additional molecule of ammonia is actually produced during glutamate turnover.⁶⁴ This results in a higher yield of ammonia molecules from glutaminolysis.

In conclusion, our study sheds new light on the role of dysregulated glutaminolysis in aging progression, by demonstrating that hyperglutaminolysis fuels the upregulation of the arginine-mTORC1 axis through a CASTOR1-mediated mechanism. While further research is needed to elucidate the precise triggers of hyperglutaminolysis in senescent cells, our findings establish a causal relationship between dysregulated glutamine catabolism and senescence-associated mTORC1 activation. These results provide a novel perspective on the contribution of glutamine catabolism in aging and offer valuable insights for developing targeted strategies to delay or mitigate aging-related processes. Although the biological necessity of glutamine is known, how to precisely control and utilize glutaminolysis for health improvement requires further investigation.

Materials and Methods

Reagents and antibodies

Adriamycin HCl (ADR, S1208), Bafilomycin A1 (BafA1, S1413) were purchased from Selleck (Houston, USA); ammonium chloride (NH₄Cl) and hydrogen peroxide (H₂O₂) were purchased from CHRON Chemical (Chengdu, China); 6-Diazo-5-oxo-L-nor-leucine (DON, HY-108357) was purchased from MedChemExpress (NJ, USA); CB-839 (Telaglenastat, T6797) was purchased from TargetMol (Massachusetts, USA); L-glutamine (Gln, G8540) was purchased from Merck (Darmstadt, Germany); sodium glutamate hydrate (glutamate, Glu, A602012), L-aspartic acid sodium salt monohydrate (aspartate, Asp, A601178), L-citrulline (Cit, A604057) and L-arginine (Arg, A600205) were purchased from Sangon Biotech (Shanghai, China). All HPLC-grade reagents used in the LC-MS/MS assays were purchased from Merck (Darmstadt, Germany). Antibodies against p16 (P21212) were purchased from ProMab Biotechnologies, Inc (Hunan, China); antibodies against p62 (ab109012) were purchased from Abcam (Cambridge, UK); antibodies against GLS1 (A5125) and 4EBP1 (A5090) were purchased from Selleck (Houston, USA) were purchased from Absin (Shanghai, China); antibodies against phosphorylated (P)-p70/S6K (9205S) were purchased from CST (Massachusetts, USA); antibodies against p70/S6K (A4898) and P-4EBP1 (AP0030) were purchased from ABclonal Technology (Hubei, China); antibody against mTOR (R1510-21) was purchased from Huabio (Zhejiang, China); antibody against LAMP2 (MA1-165) was purchased from Thermo Fisher Scientific (Massachusetts, USA); antibodies against β-actin (bs-10966R) were purchased from Bioss (Beijing, China); goat anti-rabbit IgG (H&L) (HRP conjugate) (701051) was purchased from Zenbio (Chengdu, China); and goat anti-rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^TM Plus 488 (A32731) and goat anti-rat IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^TM Plus 594 (A48264) were purchased from Thermo Fisher Scientific (Massachusetts, USA). High-glucose Dulbecco’s Modified Eagle Medium (DMEM, 12100046, Gibco) was purchased from Thermo Fisher Scientific (Massachusetts, USA) and glutamine-free high-glucose DMEM (SH30081.02) was purchased from Cytiva (Delaware, USA). siRNAs were purchased from GenePharma (Shanghai, China).

Cell culture and cellular senescence model

NIH3T3 cells (murine fibroblast line), and HEK293T (human embryonic kidney) cells were purchased from the Shanghai Institutes for Biological Sciences of the Chinese Academy of Sciences (Shanghai, China). Primary human fibroblasts (hFBs) were a gift from Lv, Liang Ph. D. All cell lines were maintained in high-glucose Dulbecco’s Modified Eagle Medium (DMEM; 12100046; Gibco) supplemented with 10% fetal bovine serum (FBS; ST30-3302; PAN Biotech, Bavaria, Germany), at 37 °C in a humidified 5% CO₂ incubator. Cells were routinely passaged at ~80% confluence using 0.25% trypsin (Merck) with 0.02% EDTA.

For H₂O₂ stress-induced premature senescence (SIPS) models, 60% confluent NIH3T3 cells cultured in plates or dishes were exposed to 400 µM (for NIH3T3) or 200 µM (for ptfLC3 plasmid-transfected NIH3T3) H₂O₂H₂O₂ in PBS at 37 °C for 60 min. Treatment was terminated by replacing H₂O₂ containing PBS with complete DMEM (10% FBS, 4 mM glutamine). Glutamine restriction treatment was applied after H₂O₂ treatment terminated (see “Glutaminolysis restriction” section). Cells were then maintained for 3 days before downstream assays. For ADR-induced SIPS, 30% confluent NIH3T3 cells were exposed to 1 µM ADR in complete DMEM for 3 days. Following ADR removal, cells were maintained in ADR-free medium for an additional 4 days before downstream assays. Glutamine restriction treatment was applied simultaneously with ADR treatment where indicated. For the replicative senescence model, primary hFBs were seeded at 40% confluence in 10 cm dishes and serially passaged upon reaching 80% confluence, and the medium was refreshed every other day. Cells at passage 10 were designated “proliferative control”, whereas cells at passage 19 were considered “replicative senescent”. Glutamine restriction treatment was applied during the 18th to 19th passages where indicated.

Glutaminolysis restriction

To interrogate the role of glutaminolysis in senescence and related signaling pathways, three complementary approaches were employed to restrict glutaminolysis in cells. 1) Low-glutamine medium cultivation. After H₂O₂ treatment or simultaneously with ADR addition, cells were maintained in glutamine-free DMEM supplemented with 10% FBS. 2) Pharmacological Inhibition: 6-Diazo-5-oxo-l-norleucine (DON) or CB-839 (telaglenastat) was added at 10 µM after H₂O₂ treatment or simultaneously with ADR addition. 3) Genetic Knockdown: Gls1-targeting siRNA was transfected 24 h prior to senescence induction. The duration of glutaminolysis restriction was 3 days for H₂O₂-SIPS and 7 days for ADR-SIPS. In replicative senescence, low-glutamine culture or 10 µM DON treatment was applied during the 18th to 19th passages (3 days).

To investigate the contribution of glutaminolysis to the downstream metabolic pathways, H₂O₂-induced senescent and proliferative NIH3T3 cells in 10 cm dishes were pre-equilibrated in fresh medium 12 hours prior to CB-839 (1 µM) or vehicle addition. Medium (with CB-839 or vehicle) was refreshed at 12 hours after treatment initiation, and cells were harvested at 36 hours after treatment initiation for metabolomic profiling (see “Cellular metabolomics analysis”).

Cellular metabolomics analysis

Cellular metabolomic profiling was conducted via an LC-MS/MS platform. Adherent cells grown in 10 cm dishes were rapidly chilled on ice to quench metabolism. After aspirating culture medium, each dish was washed with 5 mL ice-cold phosphate-buffered saline (PBS, pH 7.4) for 5 min on ice to remove extracellular metabolites. PBS was discarded, and cells were scraped into 200 µL ice-cold PBS; this wash was repeated once. Cell suspensions were centrifuged at 800 rpm for 5 min at 4 °C, and supernatants were discarded. Pelleted cells were extracted by adding 1 mL pre-chilled (-80 °C) 80% MeOH spiked with two internal standards (¹³C2-succinic acid and ¹³C5-¹⁵N-L-glutamic acid), then incubating at -80 °C for 30 min to achieve metabolic quenching. Samples were homogenized by ultrasonication at 4 °C, and then centrifuged at 13300 rpm for 15 min at 4 °C. A total of 800 µL supernatant was collected and dried under vacuum at 30 °C for approximately 3 h. The pellets were reconstituted in 1 mL of 10 mM ammonium acetate in 40% water/ 60% ACN + 0.2% acetic acid containing three internal standards (D4-glutaric acid, ¹³C5-¹⁵N-L-tyrosine and ¹³C1-L-lactate), sonicated at 4 °C for 15 min, and centrifuged at 13300 rpm for 15 min at 4 °C. Aliquots of 100 µL supernatant were transferred to LC-MS vials; 20 µL from each sample was pooled to generate a quality-control (QC) mixture. Metabolite separation was performed on a SHIMADZU LC-30AD coupled to a triple-quadrupole MS/MS. Chromatography was performed using a Waters BEH Amide column (2.1 × 100 mm, 1.7 µm) at 40 °C. Mobile phase A was 1 mM ammonium acetate with 0.02% acetic acid in water, and mobile phase B was 9 mM ammonium acetate with 0.18% acetic acid in water. The flow rate was 0.3 mL/min, and the gradient was from 10% A during the first 1.5 min and 55% A during the 5 to 10 min, followed by 10% A during the 12 to 25 min. Electrospray ionization (ESI) mode was used to detect positive and negative ions. The capillary voltage was ±4 kV, the ion source temperature was 650 °C, and the multistage reaction detection (MRM) mode was used to determine the metabolites and internal standards. Raw data were processed in MultiQuant to integrate peak areas. Statistical analysis and pathway enrichment were carried out in MetaboAnalyst 5.0, applying two-tailed Student’s t-test (p < 0.05) for differentially abundant metabolite screening and KEGG pathway mapping. Metabolite interaction networks were constructed and visualized in Cytoscape 3.9.1. Adjusted betweenness centrality (Cen_adj) was calculated as:

$${\text{Cen}}_{\text{adj}\left(\text{x}\right)}={\text{Cen}}_{\text{Btw}\left(\text{x}\right)}\times {\log }_2\mathrm{Fold\; change}\left(\mathrm{x}\right)$$

where Cen_Btw(x) is the betweenness centrality of metabolite x, and FoldChange(x) is its abundance ratio between experimental and control groups.

For investigation of the contribution of glutaminolysis to the downstream metabolic pathways. Metabolite change (“ΔAA”) was calculated as:

$${\mathrm{\Delta }\text{AA}}_{\left(\text{x}\right)}={\text{AA}}_{\text{CB}-839\left(\text{x}\right)}-{\overline{\text{AA}}}_{\text{Vehicle}\left(\text{x}\right)}$$

where x denotes either proliferating (Pro) or senescent (Sen) cells.

Amino acid quantification

To quantify extracellular amino acid abundance, cells were seeded in 6-well plates, and after the establishment of the cellular senescence model, the medium was refreshed with 2 mL of DMEM supplemented with 4 mM glutamine and 10% FBS per well. Fresh medium added to a 6-well plate with no seeded cells (n = 3) served as blank. After 24 h incubation at 37 °C in a humidified 5% CO₂ incubator, culture medium was collected on ice, pooled per condition, and vortex-mixed. For sample preparation, aliquots of 300 µL medium were deproteinized by adding 900 µL ice-cold methanol, vortexing for 5 sec, and centrifuging at 13300 rpm for 20 min at 4 °C. Supernatants were transferred to LC-MS vials for single amino acid quantification as below, in order to calculate the glutamine consumption rate. The cells were collected for protein quantification using the bicinchoninic acid (BCA) method.

For single intracellular amino acid quantification, cells were collected as described in the “Cellular metabolomics analysis” section, and cell pellet was re-suspended in 500 µL of 80% methanol, sonicated on ice, then centrifuged at 13300 rpm for 20 min at 4 °C. The supernatant was collected and stored at −80 °C for further LC-MS/MS analysis, and the sediment was used for protein quantification. via the BCA assay. Amino acids were separated on an AB Sciex QTRAP 6500 + LC-MS/MS using an ACQUITY UPLC® BEH-C18 column (2.1 × 100 mm, 1.7 µm) at 35 °C. Mobile phase A was 0.1% formic acid in water; B was acetonitrile. The gradient proceeded from 90% A to 10% A over 2 min, held for 3 min, then re-equilibrated; flow rate was 0.4 ml/min. ESI positive mode used a 5.5 kV capillary voltage, 500 °C source temperature, and 60 V declustering potential. MRM was used to determine the amino acids and internal standard ¹⁵N-glutamate. The corresponding monitoring ion pair and collision energy are shown as follows.: Glutamine: parent ion, m/z 147.1; daughter ion, m/z 84.0; dwell time, 0.1 sec; collision energy, 12 eV; arginine: parent ion, m/z 175.1; daughter ion, m/z 116.0; dwell time, 0.1 sec; collision energy, 15 eV; glutamate: parent ion, m/z 148.1; daughter ion, m/z 84.1; dwell time, 0.1 sec; collision energy, 13 eV; internal standard: parent ion, m/z 149.1; daughter ion, m/z 85.0; dwell time, 0.1 sec; collision energy, 13 eV.

To evaluate the impact of Ass1 and Asl knockdown on intracellular arginine level, cells transfected with siAss1, siAsl or siNC were subjected to a 24-hour treatment by Hank’s buffer supplied with 4 mM glutamine and 0.4 mM arginine before LC-MS/MS assay, to avoid the interference from other amino acids. After then cells were collected and processed as described above.

The glutamine consumption rate was calculated using the following formulations.

$${\text{C}}_{\text{norm}\left(\text{x}\right)}=\frac{\left({\text{R}}_{\text{blank}}-{\text{R}}_{\text{cell}\left(\text{x}\right)}\right)}{{\text{CON}}_{\text{prt}\left(\text{x}\right)}}$$

$${\text{C}}_{\text{rel}\left(\text{x}\right)}=\frac{{\text{C}}_{\text{norm}\left(\text{x}\right)}}{{\stackrel{\circledR }{\text{C}}}_{\text{norm}\left(\text{control}\right)}}$$

R_cell(x): The absolute residual glutamine abundance of medium x. R_blank: The absolute residual glutamine abundance of the blank medium sample. CON_prt(x): The cellular protein concentration of cell x. C_norm(x): Normalized glutamine consumption rate of x. C_rel(x): The relative glutamine consumption rate of x. $\stackrel{\circledR }{\text{C}}$_{norm(control)}: The average value of the normalized consumption rate of the control.

The intracellular amino acid level was calculated as follows.

$${\text{AA}}_{\text{norm}\left(\text{x}\right)}=\frac{{\text{AA}}_{\left(\text{x}\right)}}{{\text{CON}}_{\text{prt}\left(\text{x}\right)}}$$

$${\text{AA}}_{\text{rel}\left(\text{x}\right)}=\frac{{\text{AA}}_{\text{norm}\left(\text{x}\right)}}{{\stackrel{\circledR }{\text{AA}}}_{\text{norm}\left(\text{control}\right)}}$$

AA_norm(x): The normalized intracellular level of certain amino acids in x. AA_(x): The raw result from LC-MS/MS testing of certain amino acids in x. CON_prt(x): Protein concentration of x. AA_rel(x): The relative level of certain amino acids in x. ${\stackrel{\circledR }{\text{AA}}}_{\text{norm}\left(\text{control}\right)}$: The average value of the normalized level of certain amino acids in controls.

Glutaminase enzymatic activity assay

Glutaminase activity was measured using a glutaminase (GLS) test kit (A124-1-1; Nanjing Jiancheng Bioengineering Institute, Jiangsu, China). In brief, for cells, samples were collected in the same way as described in the “Cellular metabolomics analysis” section, and 500 μl of “Solution I” was added to each sample, which was subsequently homogenized by ultrasonication; for flies, about 15 flies were collected from each group in a 1.5 mL EP tube, and 500 μl of “Solution I” was added to each sample, which was subsequently homogenized by a tissue homogenizer. After a 10-minute centrifugation at 8 × 10³g at 4 °C, the supernatant was collected for further measurement. Then, all the samples were tested, and the results were calculated following the manufacturer’s instructions. The results were normalized to the protein concentration.

Ammonium measurement

The intracellular ammonium concentration was measured using an enzymatic assay kit (AA0100-1KT; Merck, Darmstadt, Germany). In brief, cells were collected in the same way as described in the “Metabolomics” section, and 300 μl of ddH₂O was added to each sample, which was subsequently homogenized by ultrasonication. The samples were centrifuged, and the supernatant was collected and deproteinized using trichloroacetic acid at a final concentration of 20%. Then, these samples were further centrifuged, the supernatant was collected, and the pH was adjusted to approximately 7 ~ 9. After that, 100 μl of each sample was used for the ammonium assay, and 1 ml of Ammonium Assay Reagent for each sample was added, followed by a 5 min incubation at room temperature, after which the absorbance at 340 nm was measured. Then, 10 μl of L-glutamate dehydrogenase was added to each sample, followed by another 5 min of incubation at room temperature, after which the absorbance at 340 nm was measured. The results were calculated following the manufacturer’s instructions and normalized to the protein concentration.

High metabolite cultivation

To assess the effects of elevated extracellular metabolites, complete DMEM (5% FBS) was supplemented with 20 mM of each indicated metabolite as following: glutamine, glutamate plus NH₄Cl (equimolar), L-aspartate plus L-citrulline (equimolar), L-arginine. The medium was replaced with a normal medium every 2 days, to avoid serious cell death, and the metabolites were re-added another 2 days later. This medium switch cycle was repeated until the end of the experiment, and cells were passaged when necessary. All treatments lasted 20 days.

Cellular Immunofluorescence assay

Cells were fixed in 4% paraformaldehyde for 15 min at room temperature, washed thoroughly in PBS, and permeabilized with 0.1% Triton X-100 for 12 min. After three PBS washes, samples were blocked in 3% BSA for 1 h and incubated overnight at 4 °C with primary antibodies diluted in 3% BSA. The following day, coverslips were washed and probed with fluorescently labeled secondary antibodies in 3% BSA for 1 h at room temperature, and then counterstained with 1 µg/mL DAPI for 15 min. Slides were mounted and imaged on a Zeiss LSM 870 confocal microscope using a 63 × oil-immersion objective (image field 214.36 × 214.36 µm at 9.5542 pixels/µm). Colocalization of mTOR with LAMP2 was quantified in Fiji (ImageJ2) by calculating the percentage of mTOR (magenta) puncta overlapping LAMP2 (green) signal, and overlapping puncta are shown in white; at least five fields and about 50 cells per condition were analyzed.

LC3-RFP-GFP assay

NIH3T3 cells stably expressing the LC3-RFP-GFP tandem protein were constructed and preserved in our laboratory.⁶⁵ After being subjected to the indicated treatments, images were obtained using a Zeiss LSM 870 fluorescence confocal microscope. For each sight, no less than 50 cells were counted and at least 150 cells for each group were counted in total, and the results of 3 sights were used to calculate the percentage of red (RFP + /GFP-) cells for each group.

siRNA and plasmid transfection

For siRNA transfection, cells were transiently transfected with target-specific siRNA oligonucleotides (GenePharma) using the JetPRIME® transfection reagent (Polyplus, Bas-Rhin, France) in accordance with the manufacturer’s protocol, and downstream assays were performed 48 h post-transfection.

Lentiviral particles were generated by co-transfecting HEK293T cells with our gene-of-interest plasmid (pTRIPZ–Mm-Castor1 for Castor1 overexpression or empty pTRIPZ vector) together with the packaging plasmids psPAX2 and pMD2.G using the PEIpro reagent (101000029; Polyplus). Sixteen hours post-transfection, the medium was replaced with fresh complete DMEM, and viral supernatants were harvested 48 hours later, clarified through a 0.45 μm filter, and stored on ice. Target HEK293T cells were then transduced with the filtered viral supernatant and, after 48 hours, subjected to selection in 1 μg/mL puromycin (A424862, Sangon Biotech) until resistant colonies were established. For induction of Castor1 overexpression, stable HEK293T cells harboring either the Mm-Castor1 or control pTRIPZ construct were treated with 0.2 μg/mL doxycycline (S5159, Selleck) for 48 hours. Plasmids psPAX2, pMD2.G, and PEIpro transfection reagent were kindly provided by Dr. Biao Dong’s laboratory, and the pTRIPZ–Mm-Castor1 vector was obtained from GENEWIZ (NJ, USA). All procedures were performed under sterile conditions in a biosafety cabinet, and viral work complied with institutional biosafety regulations.

SA-β-gal staining

Cellular senescence was evaluated by histochemical detection of senescence-associated β-galactosidase (SA-β-gal) activity using the SA-β-gal Staining Kit (G1580; Solarbio, Beijing, China) in accordance with the manufacturer’s instructions. Briefly, cells were fixed and incubated with the X-gal staining solution at 37 °C (no CO₂) until bluish-green granules became evident in the cytoplasm, indicating SA-β-gal activity. Five microscopic views were randomly selected under bright-field microscopy, and over 300 cells per group were counted; the proportion of SA-β-gal–positive cells was calculated to quantify senescence induction.

Drosophila experiments

For Drosophila stocks and husbandry, wild-type w1118 fruit flies (Drosophila melanogaster) were obtained from Fungene (Beijing, China). Gls-RNAi lines were sourced from the Tsinghua Fly Center (Beijing, China). Tubulin-Gal4 and UAS-lacZ lines were kindly provided by Prof. Haiyang Chen (PhD). All stocks were maintained at 25 °C under a 12 h:12 h light–dark cycle on standard cornmeal-molasses-yeast medium.

To achieve tissue-wide Gls depletion, Tubulin-Gal4 virgins were crossed to UAS-Gls-RNAi males. Parallel crosses to UAS-lacZ served as genetic controls. Progeny were reared at 25 °C on standard medium until assays.

To detect the relevant indicators of fly aging. Naturally aging and H₂O₂-stressed flies were used. For Naturally aging, flies were maintained under the standard condition and the 10-day-old female flies were regarded as young control and 60-day-old female flies were regarded as aging flies. To establish the H₂O₂-stressed fly model, 10-day-old female flies were pretreated for 3 days in vials containing 1% H₂O₂ in standard medium. Approximately 100 pretreated flies per group were then transferred to fresh vials supplemented with either vehicle (DMSO), 0.1 µM DON, or 0.1 µM CB-839. Cohorts were maintained under standard conditions, and daily survivors were counted. Median and maximal lifespan were derived using Kaplan-Meier survival analysis in SPSS. To evaluate the mobility of flies, a climbing score assay was performed. Age-matched cohorts (about 15 flies/vial) were gently tapped to the bottom of a clean vial marked at 8 cm. The percentage of flies that climbed past the mark within 5 s was recorded. Each group was tested in triplicate; mean percentages are reported. To evaluate the gut permeability of flies, a Smurf assay was performed. Flies were fed medium containing 2.5% brilliant blue for 16 h. Flies exhibiting dye leakage into the hemocoel (“Smurf” phenotype) were scored as positive; the percentage of Smurf-positive flies per cohort was calculated.

For sample collection and metabolite extraction, adult flies were anesthetized with CO₂, decapitated to isolate cephalothoraxes, which were immediately flash-frozen on dry ice (abdomen discarded). Batches of 15–20 cephalothoraxes were pooled per sample and stored in liquid nitrogen. Prior to metabolomic analysis, samples were homogenized in 1 mL of pre-chilled (−80 °C) 80% methanol containing isotopic internal standards using a high-throughput bead-mill homogenizer. Extracts were processed exactly as described in “Cellular metabolomics analysis,” including homogenization, protein precipitation, sonication, vacuum concentration, and HILIC reconstitution.

For dietary supplementation regimens, to model elevated amino acid intake, 10-day-old female w1118 flies (∼100 per group) were maintained on standard medium supplemented with specified concentrations of either l-glutamine or l-arginine for the duration of the experiment. Survival assays were performed as described above.

Mouse experiments

Male C57BL/6J mice (10 weeks old) were obtained from Beijing HFK Bioscience Co., Ltd. (Beijing, China). Animals were housed at the Experimental Center of West China Hospital under specific pathogen-free (SPF) conditions, with ad libitum access to standard chow and water, a controlled temperature (22 ± 2 °C), and a 12 h:12 h light-dark cycle. All procedures were approved by the Laboratory Animal Ethics Committee of West China Hospital, Sichuan University (20240227038).

To compare young versus naturally aged phenotypes, an independent cohort of C57BL/6 J mice was maintained from 10 weeks to 27 months of age under identical husbandry conditions. A separate group of age-matched young controls (3 months old) was generated by purchasing 10-week-old mice and acclimating them for 2 weeks prior to sacrifice. Mice were sacrificed under anesthesia, and tissues were harvested immediately: epididymal white adipose tissue (eWAT) for SA-β-gal staining; spleen, kidney and gastrocnemius muscle for glutaminase (GLS) enzyme activity assays and targeted LC-MS/MS metabolomic analysis.

To establish AAV9-mediated argininosuccinate lyase (Asl) knockdown, 3-month-old male C57BL/6 J mice received a single tail-vein injection of AAV9-shAsl (1 × 10¹¹ viral genomes [vg]) to achieve systemic Asl knockdown; AAV9-Vector (1 × 10¹¹ vg) served as a negative control. Two weeks post-injection, mice were randomized to receive either standard drinking water or water supplemented with 200 mM glutamine for 3 consecutive days. At the end of the study, mice were anesthetized and subjected to in vivo fluorescence imaging (IVIS Spectrum, Guangzhou Biolight Biotechnology Co., Ltd.) to confirm transduction efficiency via GFP signal. Gastrocnemius muscles were then dissected for downstream analyses, including RT-qPCR, LC-MS/MS quantification of metabolites, and western blotting. AAV9 vectors were produced by PackGene Biotech (Massachusetts, USA), and the shAsl sequence was as previously described.⁶⁶

For SA-β-gal staining, fresh eWAT was fixed with 10% of paraformaldehyde, and stained using the SA-β-gal Staining Kit (G1580; Solarbio) according to the manufacturer’s protocol. To detect GLS activity in mice, tissue homogenates were prepared in ice-cold assay buffer; GLS activity was measured using a glutaminase (GLS) test kit (A124-1-1; Nanjing Jiancheng Bioengineering Institute) following the vendor’s instructions. To measure the metabolite levels in tissues, the samples were first flash-frozen in liquid nitrogen, and tissues from 3-5 mice were pooled together and homogenized with pre-chilled 80% methanol containing isotopic internal standards. Extracts were processed and analyzed exactly as described in “Cellular metabolomics analysis.” Total RNA was isolated from gastrocnemius muscle using RNAiso Plus (9109; Shiga, Japan), and 1 μg of total RNA was reverse transcribed with HiScript III RT SuperMix for qPCR (+gDNA wiper; R323-01; Vazyme, Jiangsu, China). RT-qPCR was subsequently performed using Taq Pro Universal SYBR qPCR Master Mix (Q712; Vazyme) with the gene-specific primers listed in Supplementary Table 5. Protein lysates were prepared in RIPA buffer (R0020; Solarbio) supplemented with proteinase inhibitor cocktail (B14001; Selleck), and protein concentration was quantified by BCA assay (C503021; Sangon Biotech), and subjected to SDS-PAGE and immunoblotting for the indicated proteins.

Western Blotting

Whole-cell proteins were solubilized in RIPA buffer (R0020; Solarbio) supplemented with protease inhibitors (B14001; Selleck), lysates were rotated at 4 °C for 30 min and clarified by centrifugation, and total protein was quantified by BCA assay (C503021; Sangon Biotech). Equal amounts (30 µg) of protein were resolved by SDS-PAGE and electrophoretically transferred onto PVDF membranes (IPVH00010; Millipore). Membranes were blocked in 5% skim milk in TBS-T, incubated with primary antibodies against target proteins under optimized dilutions overnight at 4 °C, and then incubated with HRP-conjugated secondary antibodies. Chemiluminescent signals were developed using Ultrasensitive ECL Kit (PD202; Oriscience, Sichuan, China) and captured on a Fusion Solo imaging system. Band intensities were normalized to loading controls and quantified in Fiji (ImageJ2).

RT-qPCR

For gene expression analysis, total RNA was extracted from cultured cells using RNAiso Plus (9109; Shiga, Japan). One microgram of RNA was reverse-transcribed into cDNA using HiScript III RT SuperMix for qPCR (+gDNA wiper; R323-01; Vazyme). RT-qPCR was subsequently performed using Taq Pro Universal SYBR qPCR Master Mix (Q712; Vazyme) with the gene-specific primers listed in Supplementary Table 5. Reactions were run in triplicate, and relative transcript abundance was determined by the 2^–ΔΔCt method, normalizing target gene Ct values to the 18S rRNA reference. All experiments were independently repeated three times.

Statistical analysis

The data from at least three independent experiments are presented as the mean ± SD. Differences between the two groups were assessed by a two-tailed unpaired Student’s t-test, while survival curves of fly cohorts were compared by the Kaplan–Meier method with a log-rank test. p ≤ 0.05 was considered significant (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001).

Author contributions

All authors have read and approved the article. Conceptualization: H.X. and H.C. Data curation and formal analysis: H.C., N.H., and W.X. Funding acquisition: H.X., N.H., and H.G. Investigation: H.C., N.H., W.X., Y.Y., F.W., H.G., J.Z,. and T.Z. Methodology: H.C., F.W., G.L., M.T., and J.L. Project administration and resources: H.X. and H.C. Software: H.C., N.H., and W.X. Supervision: H.X. Validation: F.W., J.Z., H.T., M.Y., and Y.L. Visualization: H.C. Writing - Original Draft Preparation: H.C., N.H,. and W.X. Writing - Review and Editing: H.X., H.T., X.H, and J.L.

Data availability

All data related to this article are included in the main text and supplementary materials. The original Western blot images used to generate the figures throughout the manuscript can be found within the supplementary information.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41392-026-02576-w.