Branched-chain amino acids impact health and lifespan indirectly via amino acid balance and appetite control


Elevated branched-chain amino acids (BCAAs) are associated with obesity and insulin resistance. How long-term dietary BCAAs impact late-life health and lifespan is unknown. Here, we show that when dietary BCAAs are varied against a fixed, isocaloric macronutrient background, long-term exposure to high BCAA diets leads to hyperphagia, obesity and reduced lifespan. These effects are not due to elevated BCAA per se or hepatic mammalian target of rapamycin activation, but instead are due to a shift in the relative quantity of dietary BCAAs and other amino acids, notably tryptophan and threonine. Increasing the ratio of BCAAs to these amino acids results in hyperphagia and is associated with central serotonin depletion. Preventing hyperphagia by calorie restriction or pair-feeding averts the health costs of a high-BCAA diet. Our data highlight a role for amino acid quality in energy balance and show that health costs of chronic high BCAA intakes need not be due to intrinsic toxicity but instead are a consequence of hyperphagia driven by amino acid imbalance.


The role of macronutrients (proteins, fats and carbohydrates) in linking diets to health has been the focus of much research. Recent work has underscored the necessity of examining these links within a mixture framework, which is sensitive not only to the individual effects of macronutrients, but also to their interactive effects1. Protein, in particular, has been shown to interact powerfully with dietary fats and carbohydrates to influence health via effects on appetite and post-ingestive physiology. One such interaction, observed in many animals including humans, is ‘protein leverage’, where the strong appetite for protein causes the overconsumption of fats and carbohydrates when feeding on protein-dilute diets2,3. Another is the demonstration that the dietary ratio of protein to carbohydrate impacts reproduction, ageing, immune function, microbiome, late-life cardiometabolic health, brain health and lifespan4,5,6,7.

Proteins, however, are themselves complex mixtures of amino acids that when modified can have profound effects on growth, early life health and longevity8,9,10. Restriction of specific amino acids, such as methionine, mimic the health and lifespan effects of chronic dietary restriction, despite an increase in energy intake11. Indeed, most amino acids have important functions in metabolic health outside their role in protein synthesis. The branched-chain amino acids (BCAAs) Ile, leucine (Leu) and valine (Val) have assumed particular prominence, both because of their role in influencing insulin, insulin-like growth factor 1 (IGF1) and mammalian target of rapamycin (mTOR)—key pathways linking nutrition with health and ageing12—and also because their circulating levels are positively associated with obesity, insulin resistance and metabolic dysfunction in rodents5,13,14,15,16 and with obesity, insulin resistance and type 2 diabetes in humans17,18,19,20. We previously reported that mice fed a high-protein, low-carbohydrate diet throughout life were hypophagic, metabolically impaired in late-life, had elevated circulating BCAAs, increased hepatic mTOR activation and reduced median lifespans compared to mice fed a low-protein, high-carbohydrate diet5. These results suggest that health and ageing in mice can be altered by titrating the balance of macronutrients to influence circulating BCAA and mTOR activation.

In this study, we sought to determine whether and how dietary BCAA manipulation influences healthspan and lifespan in mice. We demonstrate that the metabolic and lifespan costs of high BCAA:non-BCAA intakes, when paired with a high-carbohydrate, low-fat nutritional background, are not associated with increased hepatic mTOR activation; rather, they can be explained by their interactions with other key metabolically essential amino acids leading to extreme hyperphagia. Rebalancing the amino acid profile in the diet with the addition of tryptophan (Trp) or threonine (Thr) suppressed food intake, whereas preventing hyperphagia in BCAA-supplemented diets through either 20% calorie restriction or pair-feeding reversed metabolic dysfunction and lifespan costs. These findings illustrate the complex nutritional interactions that influence appetite signalling, metabolic health and lifespan in mice.


Dietary amino acid imbalance drives hyperphagia and shortens lifespan

Three hundred and twelve male and female C57BL/6J mice were fed one of four isocaloric diets (all 18% total protein, 64% carbohydrate, 18% fat; total energy density 14.3 kJ g−1) varying in protein quality and amino acid balance. Mice were either killed at 15 months of age for tissue analysis or maintained for lifespan determination. The data shown are the averages of males and females combined, unless otherwise stated for variables showing a statistically significant sex difference. Manipulations of protein quality were performed so that the BCAA200 diet contained twice the BCAAs of the control diet BCAA100 (AIN-93 G), which contains the standard amount of BCAAs. The BCAA50 and BCAA20 diets contained one-half and one-fifth of the standard content of BCAAs, respectively. These adjustments to BCAAs resulted in a change in the ratio of BCAAs to other amino acids (henceforth, ‘non-BCAAs’) within the fixed total protein complement of 18% (Supplementary Table 1). Mice confined to the BCAA200 diet (that is, the diet with the highest ratio of BCAAs to non-BCAAs) were hyperphagic, consuming approximately 20% more energy than the mice on the other diets (Fig. 1a). When non-BCAA intake was plotted, mice on all treatment groups, except BCAA200, maintained their average intake of non-BCAAs to a similar level, suggesting regulation of intake of one or more non-BCAAs to a target intake, with BCAA and total energy intakes following passively (Fig. 1b).

Fig. 1: Dietary BCAA imbalance drives hyperphagia and obesity, and shortens lifespan.

a,b, Energy intake (a) and non-BCAA intake (b) averaged over 12–15 months of age (200, 50 and 20%, n = 18; 100%, n = 24 independent cages). c, Body weight trajectories over time (200, 50 and 20%, n = 72; 100%, n = 96 biologically independent mice). The green dashed line indicates the 15-month tissue collection time point from which plasma and tissue were analysed. df, body weight (d) (200%, n = 15; 100%, n = 19; 50%, n = 15; 20%, n = 9 independent cages), fat mass (e) (200%, n = 15; 100%, n = 20; 50%, n = 15; 20%, n = 8 independent cages) and lean mass (f) (200%, n = 15; 100%, n = 19; 50%, n = 15; 20%, n = 9 independent cages) measured longitudinally using EchoMRI. g, Representative DXA scans of mice measured once at 15 months of age (200%, n = 9; 100%, n = 8; 50%, n = 11; 20%, n = 10 biologically independent mice). h, Plasma BCAAs versus BCAA intake from animals collected at 15 months of age (n = 47 biologically independent mice). i, Percentage body fat (r = 0.329, Pearson correlation, P = 0.0003) measured by EchoMRI versus plasma BCAAs at 15 months of age (n = 47 biologically independent mice). The red lines show the 95% confidence interval. j, Plasma leptin levels at 15 months of age (200%, n = 17; 100 and 20%, n = 16; 50%, n = 18 biologically independent mice). km, Hepatic mTOR (n = 12 biologically independent mice for all groups) (k), S6K (n = 12 biologically independent mice for all groups) (l) and PKB/Akt (200, 100 and 20%, n = 16; 50%, n = 18 biologically independent mice) (m) activation analysed by western blot. AU, arbitrary unit. n, Survival curves analysed by CPHM. The dotted line indicates the median lifespan. The data shown are for the combined sexes analysed at 15 months of age. For all bar graphs, ANOVA was used for normal and log-normal data, and Kruskal–Wallis tests were used for non-normal data. Pairwise comparisons among diets for normal and log-normal data were made using two-sided t-tests. For non-normal data, pairwise comparisons among diets were made using a Kruskal–Wallis test. All bars indicate the mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 based on post-hoc analysis following correction for multiple testing.

We hypothesized that one or more non-BCAAs were regulated to a target intake; if this target could not be met due to dietary imbalance, mice overconsumed food (and consequently both total energy and BCAAs) to reach this intake target. Because of such compensatory feeding, hyperphagic mice had substantially increased body weight and fat mass with no change in lean mass (Fig. 1c–g). Circulating levels of BCAAs were strongly associated with BCAA intake, plateauing at approximately 40 µg ml−1 (Fig. 1h). The plasma BCAAs of mice on the BCAA200 and BCAA100 diets did not differ (Supplementary Fig. 4h), despite hyperphagia and the doubling of BCAA content in the BCAA200 diet, suggesting that the blood levels of BCAAs are regulated post-ingestively. Adiposity was strongly associated with circulating BCAA levels (Fig. 1i) and supported by elevated leptin levels in mice fed a BCAA200 diet (Fig. 1j). Elevated circulating leptin was not accompanied by changes in hypothalamic gene expression of Lepr or Socs3, key features of leptin resistance seen in many obese animals (Supplementary Fig. 2c,d).

Given that BCAAs activate the mTOR signalling pathway, we quantified activation of mTOR, ribosomal protein S6 kinase I (S6K) and protein kinase B PKB/Akt in the liver and found few differences between the groups (Fig. 1k–m), which may be attributed to the dietary macronutrient background on which BCAAs were manipulated (that is, high carbohydrate, low fat). Despite no increase in hepatic mTOR activation, the median lifespan of BCAA200 mice was reduced by approximately 10% when compared to the other dietary groups (Fig. 1n; BCAA200: 92.8 weeks; BCAA100: 102.3 weeks; BCAA50: 102.6 weeks; BCAA20: 104.6 weeks; all comparisons P < 0.05); this is probably due to the effects of hyperphagia and obesity.

Rebalancing the BCAA200 diet with Trp and Thr prevents hyperphagia

To determine whether there were specific non-BCAAs that mediated this effect, the intake of each essential amino acid (EAA) was calculated21,22. The ratio among amino acids in the non-BCAA complement was not identical between treatment diets (Supplementary Table 1), offering the opportunity to disentangle the non-BCAA effects. The intakes of three EAAs, Trp, Thr and methionine (Met), were maintained consistently across diet treatments (Fig. 2a), suggesting that these amino acids are prioritized and regulated, influencing food intake and feeding behaviour.

Fig. 2: Trp and Thr supplementation prevents hyperphagia.

a, Average intake of essential amino acids over 12–15 months (200, 50 and 20%, n = 18; 100%, n = 24 biologically independent mice). Amino acids were categorized as those that remained stable in intake across diets, BCAAs and those that were unstable across diets. The three amino acids that remained stable (Trp, Thr and Met) were used in a six-week feeding study. His, histidine. b, Over 6 weeks of feeding, male mice on a BCAA200 diet were hyperphagic as seen in the long-term study. Adding back 150% of Trp or Thr significantly suppressed hyperphagia (100%, n = 82; 200%, n = 79; 200% + Thr, n = 99; 200% + Trp, n = 91; 200% + Met, n = 105 independent daily measurements of food intake). For all bar graphs, ANOVA for normal and log-normal data, and Kruskal–Wallis tests for non-normal data, were used to determine significant differences between groups. Pairwise comparisons among diets for normal and log-normal data were made using two-sided t-tests. For non-normal data, pairwise comparisons among diets were made using a Kruskal–Wallis test. One-way ANOVA was performed with Tukey’s multiple comparisons test (100 versus 200%, P < 1 × 10−15; 100 versus 200% + Thr, P = 8.3 × 10−5; 100 versus 200% + Trp, P = 3.6 × 10−4; 100 versus 200% + Met, P < 1 × 10−15; 200 versus 200% + Thr, P = 0.010; 200 versus 200% + Trp, P = 0.005; 200% + Thr versus 200% + Met, P = 2.1 × 10−6; 200% Trp versus 200% Met, P = 9.5 × 10−7). All bars indicate the mean ± s.e.m. Groups that do not share common letters indicate significant differences (P < 0.05) based on post-hoc analysis.

We next tested the hypothesis that there is an intake target for one or more of these three EAAs, which mice will attempt to achieve despite the overconsumption of energy and BCAAs, as observed in BCAA200 mice. A six-week intervention was performed where mice were fed BCAA200 diets supplemented with either Met, Thr or Trp (by 150% of standard chow concentrations, thereby normalizing the ratio between these individual amino acids and dietary BCAAs; Supplementary Table 2). Supplementation of Trp or Thr, two metabolically essential amino acids22, but not Met, suppressed food intake on BCAA200 diets towards levels seen in control mice on the BCAA100 diet (Fig. 2b). When matched to the exome23, Trp and Thr, but not Met, were the limiting amino acids on BCAA200 diets (Supplementary Fig. 1). These results suggest that food intake can be regulated by the interaction between BCAAs, Trp and Thr, whereby adding these EAAs largely reversed the hyperphagia induced by imbalanced high-BCAA diets by normalizing their ratio.

Hyperphagia is linked to Trp-mediated serotonin depletion

Plasma metabolomics was performed using targeted liquid chromatography with triple-quadrupole mass spectrometry (LC–QQQ–MS) and showed that the glutathione metabolism and aminoacyl-transfer RNA synthesis pathways were positively associated with high BCAA:non-BCAA intakes, while the tricarboxylic acid cycle and Trp metabolism pathways were negatively associated (Fig. 3a,b and Supplementary Table 3). The downregulation of Trp metabolism is of particular importance in feeding behaviour since Trp is the sole precursor for the production of 5-hydroxytryptamine (5-HT, serotonin), a monoamine neurotransmitter that controls appetite. The reduction in Trp, 5-HT and its main metabolite 5-hydroxyindoleacetic acid in the plasma of mice with high BCAA:non-BCAA intakes reflects peripheral Trp limitation (Fig. 3c). Indeed, the inverse relationship between BCAAs and Trp, which compete for peripheral and central transport24 by the l-type amino acid transporter 1 (LAT1)25 is also reflected in the ratio of Trp:BCAA in the plasma and cortex, where BCAA200 mice exhibit the lowest Trp:BCAA ratio (Fig. 3d).

Fig. 3: Hyperphagia in BCAA200 mice is linked to Trp-mediated serotonin (5-HT) depletion.

a, Metabolic pathways (KEGG) positively (red) and negatively (blue) correlated with high BCAA: non-BCAA intake measured in the plasma of 15-month-old mice using LC–MS. b, Metabolites in each pathway (200 and 50% n = 17; 100%, n = 14; 20% n = 16 biologically independent mice). NADP+, nicotinamide adenine dinucleotide phosphate. c, Diagram depicting the relationship between dietary BCAA and non-BCAA intake, Trp metabolism and the effects on food intake. 5-HIAA, 5-hydroxyindoleacetic acid. d, The ratio of Trp:BCAA in the plasma and cortex of mice collected at 15 months of age. Plasma: 200, 100 and 20%, n = 12; 50%, n = 11 biologically independent mice. Cortex: 200%, n = 16; 100%, n = 24; 50%, n = 18; 20%, n = 17 biologically independent mice. e, Example traces from electrophysiological patch-clamp recordings showing 5-HT synaptic responses from mice fed either the BCAA100 (control) or BCAA200 (predicted 5-HT-depleted) diet for 6 weeks (n = 5 biologically independent mice). The evoked IPSC is shown in black; blocking by the 5HT1A receptor antagonist NAN-190 is shown in blue. f, Average 5-HT evoked IPSC amplitude without and with the addition of Trp to the media. The numbers of neurons measured are shown above the bars. g, Food intake of mice on BCAA200 diets following 4 d of oral administration of either fluoxetine or saline (saline, n = 7; fluoxetine, n = 8 independent cages). a,b, Pearson correlation. d,e, The data are from males and females combined. d, Pairwise comparisons for normal and log-normal data were made using two-sided t-tests. For non-normal data, pairwise comparisons among diets were made using a Kruskal–Wallis test. f, Pairwise comparisons were made between the 200 and 100% groups, without and with Trp, using two-sided t-tests. g, The caret symbol denotes a significant difference between treatments (day 1, P = 0.030; day 3, P = 0.028); the hash symbol indicates near significance (P = 0.078) based on a one-sided t-test. All bars indicate the mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.001, unless otherwise shown, based on post-hoc analysis following correction for multiple testing.

To determine if a reduced Trp:BCAA ratio in the plasma and cortex reflects central Trp-mediated 5-HT depletion, we fed mice for 6 weeks on either the BCAA200 or BCAA100 diet as a control. We made patch-clamp recordings from the dorsal raphe nuclei and used local electrical stimulation to evoke the synaptic release of 5-HT (Fig. 3e). Stimulation of the dorsal raphe nuclei evoked an inhibitory postsynaptic current (IPSC) in control BCAA100 mice (shown in black) that was blocked by the 5-hydroxytryptamine receptor 1A (5HT1A) receptor antagonist NAN-190 (shown in blue). BCAA200 mice showed a significantly blunted evoked IPSC on stimulation, consistent with the hypothesized diet-related 5-HT depletion. The average evoked IPSC amplitude was significantly lower in BCAA200 mice compared to BCAA100 mice (P < 0.001), whose response was rescued by adding Trp to the media (Fig. 3f). This suggests that increased dietary BCAAs, relative to Trp, are sufficient to lower Trp availability for central Trp uptake and 5-HT synthesis. We then tested if increasing 5-HT availability in the brain would reduce food intake in mice on a BCAA200 diet. An additional group of mice was fed for 6 weeks on the BCAA200 diet and treated with either saline or fluoxetine, a widely used antidepressant and selective serotonin reuptake inhibitor (SSRI), via oral gavage for 4 d. Mice treated with fluoxetine showed a rapid and significant decrease in food intake compared to saline controls from day 1 of treatment (Fig. 3g), a pattern that persisted for the next 3 d.

High BCAA:non-BCAA intake alters hypothalamic gene expression

To investigate how dietary BCAA:non-BCAA influences central appetite signalling, we performed RNA sequencing (RNA-seq) in the hypothalamus of 15-month-old mice. Differentially expressed genes were correlated with BCAA intake; genes with significant correlations were plotted on a heat map. Mice with the highest intake of BCAAs (and hence low non-BCAA intake) had a contrasting pattern of gene expression compared with those of all other mice (Fig. 4a). Three genes, Dyrk1a, Ttc7b and Nlrp3 (previously linked with appetite, obesity and inflammation, respectively) showed strong positive correlations with high BCAA intake (Fig. 4b). Transgenic mice with increased hypothalamic Dyrk1a messenger RNA expression exhibit increased food intake26, which is mediated by forkhead box O-induced Npy expression, a key gene in the orexigenic pathway that regulates appetite. Ttc7b expression, a marker of fat cell function and lipid storage, was elevated with high BCAA:non-BCAA intakes, together with an increase in the inflammasome marker Nlrp3 (Fig. 4b)27,28. Functional annotation revealed that genes positively associated with high BCAA:non-BCAA intakes were enriched in several pathways related to cancer, while negatively associated pathways included p53 signalling and apoptotic pathways (Fig. 4c). These pathways showed significant EASE scores29; however, after adjusting for multiple testing, these pathways did not remain significant (Supplementary Table 4).

Fig. 4: The ratio of dietary BCAA to non-BCAA influences hypothalamic gene expression.

a, Heat map of the hypothalamic genes significantly correlated with BCAA intake (averaged over 12–15 months of age) measured using RNA-seq. The red and blue colours indicate higher and lower mean expression, respectively, as measured by row standardized z-scores (n = 6 biologically independent mice). b, Volcano plot of the Pearson correlation coefficient P versus the Pearson correlation coefficient. Positively associated genes are labelled with orange; negatively associated genes are labelled with blue. P = 0.05 is marked with a red dashed line (n = 6 biologically independent mice). c, Partial list of the top enriched pathways (KEGG) positively (red) and negatively (blue) correlated with BCAA intake. The data are from males and females combined, collected at 15 months of age (n = 6 biologically independent mice). a,b, Pearson correlation was used.

High BCAA:non-BCAA diets promote hepatosteatosis and de novo lipogenesis

Hyperphagic mice on BCAA200 diets developed hepatosteatosis (Fig. 5a). Hepatic triglyceride content, fat scores and levels of plasma α-keto-δ-(NGNG-dimethylguanidino)-valeric acid (DMGV, a human metabolite recently found to be a biomarker for non-alcoholic fatty liver disease30), were significantly elevated in BCAA200 mice (Fig. 5b–d and Supplementary Table 5) and supported by elevations in plasma alanine transaminase (ALT) and aspartate transaminase (AST) (Fig. 5e,f, Supplementary Fig. 3b,c and Supplementary Table 5). Markers of hepatic de novo lipogenesis were decreased in BCAA20 diets. These include ATP citrate lyase (ACLY), stearoyl-CoA desaturase 1 (SCD1), fatty acid synthase (FAS) and acetyl-CoA carboxylase (ACC) (Fig. 5g,h). Excess BCAAs may contribute directly to de novo lipogenesis31 or via increased ACLY phosphorylation by the branched-chain ketoacid dehydrogenase kinase (BCKDK)32.

Fig. 5: Dietary BCAA imbalance promotes hepatosteatosis and de novo lipogenesis.

a, Representative H&E stains of livers (n = 12 biologically independent mice, assessed once by four independent observers blinded to the dietary treatment groups). b, Liver triglycerides (n = 12 biologically independent mice for all groups). c, Fat score (200, 100 and 50%, n = 12; 20%, n = 11 biologically independent mice). d, Circulating levels of DMGV, a new metabolite marker for hepatosteatosis. bd, All are increased on high BCAA diets (200%, n = 17; 100%, n = 14; 50%, n = 16; 20%, n = 11 biologically independent mice). e, Liver function tests as indicated by plasma ALT levels (200%, n = 17; 100%, n = 23; 50%, n = 17; 20%, n = 15 biologically independent mice). f, Liver function tests as indicated by plasma AST levels (200%, n = 17; 100%, n = 14; 50%, n = 18; 20%, n = 18 biologically independent mice). g, Markers of de novo lipogenesis in liver quantified using western blot (n = 10 biologically independent mice). h, Representative images quantified once using 10 biologically independent mice. The data are from males and females combined, collected at 15 months of age. For all bar graphs, ANOVA for normal and log-normal data, and Kruskal–Wallis tests for non-normal data, were used to determine significant differences between groups. Pairwise comparisons among diets for normal and log-normal data were made using two-sided t-tests. For non-normal data, pairwise comparisons among diets were made using a Kruskal–Wallis test. All bars indicate the mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 based on post-hoc analysis following correction for multiple testing.

Dietary BCAA:non-BCAA imbalance alters whole-body metabolism

No differences in energy expenditure were detected among groups (Fig. 6a, Supplementary Fig. 3h and Supplementary Table 5), nor were there differences in mitochondrial brown fat uncoupling protein 1 (UCP1) expression (Fig. 6c). When UCP1 expression was normalized for brown adipose tissue (BAT) mass, BCAA200 mice showed reduced UCP1 expression (Fig. 6d). Preferential oxidation of carbohydrate fuels by BCAA20 mice is indicated by an increased respiratory quotient (Fig. 6b) and supported by elevated hepatic Pepck mRNA expression (Fig. 6e), a marker of gluconeogenesis. BCAA20 mice also showed increased pancreatic islet glucagon content (Fig. 6f,g), a hormone that signals hepatic conversion and release of glucose from glycogen, a process central to the regulation of blood glucose levels. Increased glucagon in BCAA20 mice corresponds to the observed reduced fasting insulin levels and pancreas mass (Fig. 6h and Supplementary Fig. 2f), which persisted despite increased islet area and no change in islet insulin content (Supplementary Fig. 2g,h). We found no differences in basal blood glucose levels or glucose tolerance, but found a significant positive relationship between plasma BCAAs and the product of fasting glucose and insulin, an index of insulin sensitivity (Fig. 6i–l). Additionally, plasma triglycerides and plasma IGF1 levels did not differ across groups (Fig. 6m,n). Fibroblast growth factor 21 (FGF21) levels were not significantly different across diets when males and females were combined (Supplementary Fig. 2j), but showed a significant sex-by-diet interaction, with diet significantly affecting plasma FGF21 in males (FGF21 being elevated on BCAA200) but not females (Supplementary Fig. 3a and Supplementary Table 5). FGF21 resistance is unlikely since mRNA expression of the downstream regulatory co-receptors fibroblast growth factor receptor 1 and Klotho beta-like protein in white adipose tissue, the proposed site of FGF21 resistance (Supplementary Figs. 2e and 3g), did not differ between groups. Rather, elevated FGF21 levels may reflect amino acid imbalance and/or limitation on high BCAA:non-BCAA diets, rather than a FGF21-resistant state.

Fig. 6: Dietary amino acid imbalance alters whole-body metabolism.

a,b, Energy expenditure (a) and respiratory quotient (b) (200, 50 and 20%, n = 12; 100%, n = 16 biologically independent mice). Energy expenditure and respiratory quotient were measured in individual animals over 2 d and 2-night cycles using metabolic cages. c, UCP1 protein expression in BAT (n = 12 biologically independent mice). d, UCP1 protein per g of BAT mass (n = 12 biologically independent mice). All western blots were standardized to contain 25 μg of protein. e, Pepck mRNA expression in the liver (200%, n = 17; 100%, n = 21; 50 and 20%, n = 18 biologically independent mice). f, Glucagon content in pancreatic islets (n = 3 mice). g, Representative images of BCAA200 and BCAA20 islets visualized by immunohistochemistry, conducted once with three independent mice). hk, Glucose metabolism is shown by fasting insulin levels (200%, n = 14; 100%, n = 19; 50%, n = 14; 20%, n = 8 mice) (h), basal glucose levels (200%, n = 14; 100%, n = 19; 50%, n = 15; 20%, n = 9 mice) (i) and the AUC from GTTs (200%, n = 33; 100%, n = 44; 50%, n = 33%; 20%, n = 27 mice) (j,k). l, Relationship between plasma BCAAs and the products of fasting glucose and insulin, an index of insulin sensitivity (P = 0.054; r = 0.115; Pearson correlation; n = 33 mice). The red lines show the 95% confidence interval. m,n, Plasma triglycerides (200%, n = 14, 100%, n = 20; 50%, n = 16; 20%, n = 12 mice) (m) and IGF1 (200%, n = 17; 100%, n = 21; 50 and 20%, n = 18 mice) (n) were measured. For all bar graphs, ANOVA for normal and log-normal data, and Kruskal–Wallis tests for non-normal data, were used to determine significant differences between groups. Pairwise comparisons among diets for normal and log-normal data were made using two-sided t-tests. For non-normal data, pairwise comparisons among diets were made using a Kruskal–Wallis test. All bars indicate the mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.001 based on post-hoc analysis following correction for multiple testing.

The health impacts of high BCAAs: non-BCAA intakes are a consequence of hyperphagia

Our results suggest that the negative health and lifespan costs of long-term high BCAA:non-BCAA diets are a consequence of hyperphagia and obesity, rather than any intrinsic toxicity. Obese mice on a BCAA200 diet have elevated circulating BCAAs and metabolites linked with poor cardiometabolic outcomes14,33,34,35,36. However, mice on BCAA100 diets with similar levels of circulating BCAAs (Supplementary Fig. 4h) did not exhibit hyperphagia, metabolic dysfunction or shortened lifespan, suggesting that circulating BCAA levels alone cannot predict metabolic health and lifespan. Therefore, the question arises as to how much of the negative impact of high BCAA intake on health and lifespan is the result of BCAAs per se, and how much is a consequence of hyperphagia.

To distinguish between these alternatives, we explored whether preventing hyperphagia by either 20% calorie restriction or pair-feeding would prevent the metabolic and lifespan costs observed in ad libitum-fed mice on BCAA-supplemented diets. A separate cohort of mice was fed across the life-course on BCAA200 and BCAA100 diets provided as a single aliquot of food per day (80% of the ad libitum intake of BCAA100 mice). Calorie restriction increased median lifespans by approximately 30% relative to ad libitum-fed controls on both BCAA200 and BCAA100 diets (Fig. 7a; 134.8 versus 92.8 weeks on BCAA200 and 143.7 versus 102.3 weeks on BCAA100, respectively). While ad libitum-fed BCAA200 animals lived for a period 10% shorter than ad libitum-fed BCAA100 mice (P = 0.01) under calorie restriction, this difference was reduced to 6% (Fig. 7a; calorie-restricted 200, 134.8 weeks versus calorie-restricted 100, 143.7 weeks; P = 0.68). Preventing hyperphagia on a BCAA200 diet by 20% calorie restriction resulted in a reduction in body weight and percentage body fat (Fig. 7b,c) and led to a reduction of liver triglyceride content, fasting insulin, plasma IGF1 and plasma BCAAs (Fig. 7d–g). Body composition and metabolic differences apparent between BCAA200 and BCAA100 animals under ad libitum feeding conditions were mostly normalized under calorie restriction (Supplementary Fig. 5a–i), including no difference in circulating BCAAs between the calorie-restricted 200 and 100 groups (Supplementary Fig. 5c). Preventing hyperphagia by pair-feeding also showed that despite dietary BCAA supplementation, mice did not differ in body weight, develop obesity or show a reduction in lifespan (Fig. 7h–j; median lifespans 94 weeks for 23% protein, 110 weeks for 6% protein and 115.5 weeks for 6% protein + BCAA). Together, these data show that the metabolic and lifespan consequences of a high BCAA:non-BCAA diet are not due to BCAAs per se, but are predominantly the result of hyperphagia.

Fig. 7: Preventing hyperphagia on high BCAA diets averts metabolic and lifespan costs.

a, Survival curves of ad libitum-fed and 20% calorically restricted mice. Data from ad libitum-fed animals are replotted from Fig. 1n and are shown for direct lifespan comparison to calorically restricted mice. b,c, Body weight (ad libitum-fed, n = 15; calorically restricted, n = 21 independent cages) (b) and percentage body fat (ad libitum-fed, n = 15; calorically restricted, n = 21 independent cages) (c) of 15-month-old mice fed a BCAA200 diet at either ad libitum or calorically restricted conditions. dg, Liver triglyceride content (n = 12 biologically independent mice) (d) and plasma analysis of fasting insulin (ad libitum-fed, n = 14; calorically restricted, n = 21 independent cages) (e), IGF1 (ad libitum-fed, n = 17; calorically restricted, n = 22 independent cages) (f) and BCAA (ad libitum-fed, n = 12; calorically restricted, n = 13 biologically independent mice) (g). h, Survival curves of mice pair-fed with an exome-matched diet of either 23% protein, 6% protein or 6% protein + BCAAs. i,j, Body weight (i) and percentage body fat (j) of animals at 12 months of age (23% protein, n = 13; 6% protein, n = 12; 6% protein + BCAA, n = 13 biologically independent mice). bg, Two-sided t-tests were used. i,j, ANOVA and Tukey’s multiple comparisons post-hoc test were used. All bars indicate the mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.001 based on post-hoc analysis following correction for multiple testing.


In this study, we show that long-term high BCAA:non-BCAA intakes led to hyperphagia, obesity, altered appetite signalling and reduced lifespan. Additionally, we show that elevated circulating BCAAs can occur in both metabolically healthy and unhealthy obese and insulin-resistant mice; therefore, increased circulating BCAAs alone cannot predict metabolic dysfunction. Given that elevated circulating levels of BCAAs in humans have been reported as biomarkers for obesity14,37, hepatosteatosis38,39, insulin resistance17,18,20,35 and type 2 diabetes19,33,40, the question arises whether this widely reported relationship between elevated levels of circulating BCAAs and adverse health outcomes reflects intrinsic BCAA toxicity, dietary intakes or metabolic dysfunction. Although some studies report no difference in circulating BCAAs after whey protein supplementation in humans18, others show that circulating BCAAs are increased in overfeeding and can be uncoupled from health41. Across different strains of mice, BCAAs alone were not identified as the top metabolites associated with insulin resistance42. Similarly, our data show that elevated intakes and corresponding circulating levels of BCAAs are not toxic or by themselves markers of insulin resistance; rather, the major deleterious effects of excess BCAAs under an isocaloric density, moderate-protein, high-carbohydrate, low-fat background, arise indirectly via hyperphagia and obesity.

We discovered that the metabolic burden of dietary BCAAs under high carbohydrate intake was not due to increased hepatic mTOR activation, but primarily driven by hyperphagia. Creating dietary imbalance by increasing the ratio of BCAAs:non-BCAAs (particularly Trp and Thr) resulted in compensatory feeding and increased energy intake in an attempt to meet specific non-BCAA intake targets. Recognizing dietary amino acid imbalance and deficiency has been shown to occur via the uncharged tRNAs that signal deficiency in forebrain areas such as the anterior piriform cortex, which activates the phosphorylation of general control non-derepressible protein 2 to trigger the amino acid response to promote catabolism and inhibit anabolism12,43. In rats, diets devoid of or deficient in one or more EAAs result in growth failure and rejection of the diet10,43, a finding probably linked to extreme deficiency of an indispensable amino acid22. However, under ad libitum feeding conditions with less severe amino acid imbalance, compensatory feeding mechanisms increase food intake to obtain target intakes of limiting amino acids, rather than cause food aversion—a result reported by Rose in early work in rats44.

Preventing hyperphagia by 20% calorie restriction averted obesity and the metabolic and lifespan costs associated with high BCAA:non-BCAA diets. Similarly, preventing hyperphagia on BCAA-supplemented diets by pair-feeding averted body fat gain and did not shorten lifespan relative to non-BCAA-supplemented diets with the same total protein content. Together, these results indicate that: (1) dietary BCAAs do not appear to be intrinsically harmful; rather, the adverse health effects can be explained by their interactions with other amino acids leading to extreme hyperphagia; and (2) high circulating BCAA concentrations can occur in both metabolically healthy and unhealthy mice, suggesting that BCAAs alone are not a sufficient marker of metabolic health.

When considered in light of the present results, it is difficult to attribute causality to the relationship between BCAAs and health from previous studies, given that some reports linking elevated levels of circulating BCAAs with metabolic dysfunction in humans show concomitant increases in energy, total protein and/or BCAA intake13,14,33,34,36. In rodents, circulating levels of BCAAs have also been shown to be positively associated with total protein intake5,15. In humans, fasting BCAA levels in the obese, insulin-resistant state are attributed to changes in BCAA catabolic enzymes and perturbations in amino acid homoeostasis45. For example, changes in mTOR signalling in muscle and alterations in branched-chain amino acid transaminase (the rate-controlling and irreversible step in BCAA metabolism) and branched-chain α-ketoacid dehydrogenase, the mitochondrial BCAA oxidation checkpoint17, have been shown to link circulating BCAAs and insulin resistance or type 2 diabetes. In this study, we did not find any differences in hepatic mTOR activation, BCAA metabolism (measured by hepatic and skeletal muscle mRNA expression of Bcat2, the gene encoding branched-chain amino acid transaminase), Bckdha, Bckdhb or related metabolites such as the branched-chain α-ketoacids, acylcarnitines or the amino acids phenylalanine (Phe), tyrosine (Tyr) or glutamine (Gln)14 (Supplementary Fig. 4). This lack of response may be attributed to the dietary macronutrient background on which BCAAs are manipulated, whereby alterations to amino acid balance were conducted under a moderate-protein, high-carbohydrate, low-fat macronutrient background. We have previously shown that maximal hepatic mTOR activation occurs under a combination of high protein and low carbohydrate intakes. Therefore, it follows that high levels of circulating BCAAs may be associated with poor health outcomes because they reflect long-term elevated protein intake, whose impacts are especially pronounced when coupled with low carbohydrate intakes5. Others have also shown that mTOR in rats is increased with BCAA supplementation only when combined with a diet high in fat14, and that the interaction between high BCAAs and excess dietary fat is central to the development of insulin resistance46. Therefore, overall, previous studies are consistent with the concept that circulating BCAAs reflect health, but the relationship may be more complex than originally postulated. We show that circulating BCAAs are biomarkers for diet, particularly protein and total food intake, and that elevations in circulating BCAAs occur in both metabolically healthy and unhealthy animals. Recent work by Elshorbagy et al.41 supports this conclusion, showing that a 28-day intervention of calorie overfeeding in healthy men and women with a body mass index of 26.6 ± 0.6 resulted in elevated fasting serum isoleucine (Ile) and Val by day 3 in the absence of obesity, insulin resistance or type 2 diabetes. However, it is important to note that elevated circulating BCAAs have been associated with improved health in some settings47,48,49,50. In the short term, high levels of BCAAs caused by a high protein intake may be associated with improved outcomes in overweight individuals secondary to protein leverage-induced weight loss. Additionally, the anabolic effects of elevated BCAA and protein intakes are clearly beneficial for supporting growth and reproduction at appropriate life stages.

The impact of dietary BCAAs on food intake was mediated via interactions with other dietary amino acids. We identified stable intakes of three other EAAs, Trp, Thr and Met, which remained consistent as a result of the different food intakes observed in diets varying in BCAA content. Supplementing only the metabolically essential amino acids Trp or Thr, thereby rebalancing the diet for these amino acids against BCAAs, significantly reduced the hyperphagia associated with the BCAA200 diet. Indeed, exome matching predicted both Trp and Thr to be limiting in the BCAA200 diet, while Met is predicted to be over threefold in excess. Trp limitation in the imbalanced BCAA200 diet resulted in a downregulation of plasma metabolites associated with Trp metabolism including 5-HT and 5-hydroxyindoleacetic acid, a key neuropeptide and direct metabolite, derived solely from Trp and involved in energy balance and appetite control. Although studies have shown that suppressed thermogenesis in the BAT of obese mice induced with a high-fat diet is due to increased peripheral 5-HT (ref. 51), the inverse relationship between peripheral BCAA and Trp in obese BCAA200 mice is probably due to direct manipulation of these amino acids in the diet.

To elicit its effect on appetite regulation, Trp must cross the blood–brain barrier for central 5-HT synthesis, where it competes directly with the other large neutral amino acids—Phe, Tyr and the BCAAs—for transport across the blood–brain barrier by the LAT1 amino acid transporter24. Indeed, reduced plasma ratios of Trp relative to the other large neutral amino acids (Trp ratio) have been associated with depression and obesity. Increased ingestion and circulating levels of BCAAs, as shown in BCAA200 mice, lowers brain Trp uptake and 5-HT synthesis52 as demonstrated by reduced plasma and cortex Trp:BCAA ratios in hyperphagic BCAA200 mice and reduced synaptic release of 5-HT from the serotonergic neurons of the dorsal raphe nuclei. Early work in rats showed that peripheral injection of serotonergic agonists reduced food intake, while inhibition of 5-HT activity restored feeding53. More recently, administration of SSRIs, which increases serotonin availability, was reported to not only exert effects on mood and anxiety, but also reduce appetite25; a response we demonstrated by administering the SSRI fluoxetine to hyperphagic mice on a BCAA200 diet. Therefore, we show that hyperphagia in BCAA200 mice is linked to central Trp-mediated 5-HT depletion, which is probably due to competition for LAT1 transport. An important next step will be to investigate the role of LAT1 in mediating Trp uptake and serotonin production in our model of BCAA-mediated hyperphagia and obesity.

Several limitations of our study are acknowledged. First, our experiments did not directly measure BCAA turnover and metabolism; therefore, the fate of excess dietary BCAAs and whether they contribute directly to adiposity remain unclear. Recent work using in vivo whole-body isotopic tracing has shown that in insulin-resistant mice, there is a shift in BCAA oxidation from liver and adipose tissue towards muscle54. Therefore, it will be important for future studies to determine if BCAA oxidation and perturbations to amino acid homoeostasis contribute to obesity in BCAA200 mice and whether amino acid uptake from the circulation, protein turnover or amino acid oxidation differ in this mouse model.

Second, the experimental diets used in this study were derived from a combination of casein and individual amino acids. The availability of free amino acids can influence gastric emptying, with potential secondary effects mediated by activating vagal-vagal, enteroendocrine and enteric nervous systems, and can also impact the rate of amino acids appearing in the bloodstream55,56. Although using a combination of casein and free amino acids allowed precise manipulation of protein quality, future studies examining amino acid balance could consider the use of whole proteins to ensure normal physiological responses.

Third, we did not directly test the palatability of our diets; amino acid manipulations may influence palatability, food intake and weight gain. Although palatability is important in acute feeding studies, issues around palatability are less relevant to explaining our results over long-term feeding. In insects1 and mice57, short-term palatability responses are typically overridden by prolonged exposure to a single diet, such that an initially unpalatable diet may ultimately be eaten in greater amounts than a more palatable one since animals compensate for nutritional deficiencies. In short, palatability is not fixed and a short-term assessment may not help explain long-term patterns of intake.

Finally, we acknowledge that the changes in hypothalamic gene expression and 5-HT synaptic response in dorsal raphe nuclei may be due to the balance of dietary BCAAs:non-BCAAs or the hyperphagic nature of mice before tissue collection. To understand if there is a BCAA-specific effect, future investigations should provide the same diets while controlling for energy intake. Any subsequent changes in gene expression or synaptic response can then be directly attributed to the dietary amino acid balance.

In conclusion, we show that long-term dietary BCAA manipulation influences healthspan and lifespan in mice by regulating food intake via a mechanism that involves their interaction with key non-BCAAs, including Trp and Thr. Dietary amino acid imbalance influences central and peripheral appetite signalling via Trp-mediated 5-HT depletion, resulting in hyperphagia, obesity, hepatosteatosis and reduced lifespan. These metabolic and lifespan costs occur in the absence of hepatic mTOR activation and are reversed by preventing hyperphagia using 20% calorie restriction or pair-feeding. The results indicate that the adverse effects on imbalanced high BCAA:non-BCAA diets are secondary to hyperphagia rather than any intrinsic BCAA toxicity. Food intake, health and lifespan are titrated against both the macronutrient composition, quantity and quality of dietary protein.


Animals and husbandry

BCAA longevity study

Three hundred and twelve male and female C57BL/6J mice were obtained from the Australian Resource Centre at 4 weeks old and housed at the Charles Perkins Centre at the University of Sydney under a 12 h light–dark cycle. Experiments using C57BL/6J mice were conducted under the approval of Sydney University’s Animal Ethics Committee (protocol no. 2014/752). Animals were housed four per cage in standard approved cages and were not exercised for the duration of the study (see Nature Research Reporting Summary for additional details of animals and study design). At 12 weeks of age, mice were allocated to one of four experimental ad libitum diet treatments. An additional 192 male and female mice were obtained and assigned to a 20% calorie-restricted BCAA100 or BCAA200 diet. Daily aliquots were estimated from the ad libitum intake of the BCAA100 mice. Food intake and body weight were measured fortnightly until 6 months of age, followed by monthly measurements thereafter. Mice were checked twice weekly; animals losing more than 15% body weight were killed. At 15 months of age, plasma and tissue were collected from one mouse from each cage. Animals were anaesthetized using a mixture of ketamine and xylazine (100 mg kg−1 ketamine and 16 mg kg−1 xylazine) and killed by exsanguination via cardiac bleed using a 23 G needle. Tissues collected were snap-frozen in liquid nitrogen or fixed for histology; they included liver, muscle, white and brown fat, spleen, pancreas, kidneys, heart and reproductive organs. Tissue collection was performed between 10:00 and 12:00 consistently, 4–6 h after the initiation of the light cycle and when animals had completed their normal overnight feeding period. The rationale behind this collection time was to provide a stable, postprandial sample several hours after the animals had completed their normal overnight feeding period. The collection time ensured that significant alterations to the normal diurnal pattern of feeding/fasting and active/sleeping were not introduced. This provides the best indication of their baseline response to the dietary interventions over 15 months of feeding, whereas overnight fasting introduces a significant fasting response in mice that would interfere with the normal hormonal and metabolite plasma profile due to innate nocturnal feeding patterns.

Short-term (6-week) dietary interventions

For short-term feeding experiments adding back Met, Thr or Trp to a BCAA200 diet (Fig. 2), 60 male C57BL/6J mice obtained from the Australian Resource Centre, housed 2 per cage at 12 weeks of age, were fed for 6 weeks on either a BCAA100, BCAA200, BCAA200 + Trp, BCAA200 + Thr or BCAA200 + Met diet (see Supplemental Tables 1 and 2 for a detailed breakdown). Food intake was measured every 2 d and body weight was recorded weekly. For studies investigating the effect of diet on serotonin (5-HT) and fluoxetine administration on food intake (Fig. 3), an additional 52, 6-week-old male C57BL/6J mice from the were purchased, housed 2 or 3 per cage and fed on either a BCAA200 or BCAA100 diet for 6 weeks. Electrophysiological experiments were conducted on five mice per diet. The remaining 42 mice were treated with either fluoxetine (n = 24 mice; 8 cages) or saline (n = 18 mice; 7 cages).

Exome-matching longevity study

C3B6F1 female hybrids were used for studies using exome-matched diets (Supplementary Table 4); female hybrids were generated by a cross between C3H female and C57BL/6J male mice. Parental strains were obtained from the Charles River Laboratories and experimental animals were bred in an in-house animal facility at the Max Planck Institute for Biology of Ageing. Four-week-old female mice were housed in groups of five, in individually ventilated cages under specific-pathogen-free conditions. Experiments were conducted under the approval of the State Office for Nature, Environment and Consumer Protection North Rhine-Westphalia (approval nos. 84-02.04.2012.A245 and 84-02.04.2017.A175).

Experimental diets

BCAA longevity study

All experimental diets, except for exome-matched diets, were custom-designed and manufactured in dry, pelleted form by Specialty Feeds. Diets were isocaloric (14.4 kJ g−1) and matched in the total calculated net metabolizable energy from protein (18%), carbohydrate (64%) and fat (18%) (Supplementary Table 1). BCAA200: twice the BCAA content of control diet AIN-93 G; BCAA100: standard content of BCAAs; BCAA50 and BCAA20: containing one-half and one-fifth of the standard content of BCAAs.

Short-term amino acid study

Diets where Thr, Trp or Met were added to a BCAA200 diet were matched for total energy content, BCAA content and macronutrient composition (Supplementary Table 2).

Exome-matching longevity study

Four-week-old C3B6F1 female hybrid mice were fed a 6% protein diet ad libitum for 3 months. At the age of 4 months, mice were assigned one of three experimental diets and fed ad libitum for 3 weeks, during which food intake was recorded twice per week. The group with the lowest food intake (23% protein) was taken as the baseline group to which other groups were pair-fed. Pair-feeding was done by measuring the food intake of mice fed the 23% protein diet with ad libitum access to food and adjusting the food intake of the rest of the mice accordingly. The food intake of control mice was measured twice a week and the average value per week was used to adjust the food intake for the rest of the mice for the following week. Food aliquots were prepared 1 week in advance and all mice were fed daily at 9:00. To minimize the differences in daily rhythms of feeding and activity between the control group and the rest of the groups, controls were also given daily aliquots of food that exceeded the amount normally eaten by this group.

To generate the theoretical requirement of mice for dietary amino acids, we used an in silico technique called exome matching23. Briefly, we used the complete set of protein-coding sequences in the mouse genome (Ensembl v.54, May 2009, downloaded 2 July 2009) to calculate the median proportional representation of each amino acid across all proteins. To determine the theoretically limiting amino acids, we divided the proportional representation of each dietary EAA (expressed in moles) by its proportional representation in the mouse exome and found the amino acid with the lowest value. Every other amino acid is then considered to be in excess. For each mole of cysteine (Cys) or Tyr that was considered undersupplied in the food, the availability of Met (for Cys) or Phe (for Tyr) was reduced by 1 mol. Non-essential amino acids are not considered because they can be generated de novo as long as the general supply of nitrogen from other amino acids is sufficient. The exome-matched diets used in the lifespan study were manufactured by ssniff and were isocaloric (16.6 kJ g−1; Supplementary Table 6). All amino acids were crystalline; 35–40% of BCAAs were added to the 6% protein diet (6% protein + BCAA) and all other amino acids were reduced to accommodate for BCAA supplementation.

Body composition

Body composition of C57BL/6J male and female mice was assessed using an EchoMRI 900 (EchoMRI) at 15 months of age. Additionally, dual-energy X-ray absorptiometry (DXA) scans were obtained using the UltraFocus 100 DXA (Faxitron) before tissue collection at 15 months. Animals were anaesthetized with 3% isoflurane, maintained at 1.5% and imaged. The region of interest was set to exclude the head and tail during scanning. Body composition of C3B6F1 female mice on exome-matched diets were measured with the minispec LF50H Body Analyzer (Bruker) at 12 months of age.

Glucose metabolism

Glucose tolerance tests (GTTs) were performed on C57BL/6J male and female mice at 15 months of age, by fasting mice for 4 h before testing. Basal blood samples were obtained by tail tip excision; blood glucose was measured using a clinical glucometer (Accu-Chek Performa; Roche Diagnostics). Glucose (2 g kg−1 lean mass) was then administered via oral gavage. Blood was collected at baseline, then 15, 30, 45, 60 and 90 min from the original tail wound and serial tail tip excision was not required. The incremental area under the curve (AUC) was calculated. The AUC indicates the time taken to clear a bolus dose of glucose from the bloodstream and return to basal levels.

Metabolic hormones

FGF21 was measured in plasma collected from 15-month-old C57BL/6J male and female mice using the mouse/rat FGF21 enzyme-linked immunosorbent assay (ELISA) Kit (BioVendor Laboratory Medicine) according to the manufacturer’s protocol. Mouse leptin, insulin and IGF1 were also measured by ELISA, according to the manufacturer’s instructions (Crystal Chem).

Plasma amino acids

Amino acids from C57BL/6J male and female 15-month-old mice were analysed at the Australian Proteome Analysis Facility, Macquarie University, using the AccQ-Tag Ultra Chemistry Kit (Waters).

Liver histology and triglyceride content

Paraffin-embedded liver tissue collected from 15-month-old C57BL/6J male and female mice was sectioned at 5 µm and stained with hematoxylin and eosin (H&E). The extent of steatosis and inflammation was assessed and scored (0 = 0% fat present, 1 = 1–33%, 2 = 34–66% and 3 ≥ 67%) by four independent observers blinded to the dietary treatment groups.

Liver triglyceride content was measured by homogenizing 30 mg of frozen tissue from C57BL/6J male and female 15-month-old mice in a 1:2 ratio of methanol and chloroform using a bead-based tissue lyser (TissueLyserLT, Qiagen). Lipids were extracted overnight on a roller and dried down using a nitrogen apparatus and a heating block at 37–45 °C. The dried sample was then resuspended in absolute ethanol (RNA grade), quantified by GPO-PAP method (catalogue no. 11730711 216, Roche/Hitachi) with absorbance measured using an Infinite M1000 PRO plate reader (Tecan) at 490 nm.

Blood lipids and biochemistry

Blood cholesterol, triglycerides and liver function tests (ALT and AST) analyses were performed on plasma collected from C57BL/6J male and female 15-month-old mice at the Diagnostic Pathology Unit, Concord Hospital, NSW Health using a cobas 8000, c702 Photometric Modular Analyzer (Hitachi).

Metabolic phenotyping

To determine the whole-animal metabolic rate, substrate use and activity, 12–16 C57BL/6J male and female mice per diet were housed individually and assessed by indirect calorimetry in a Promethion high-definition continuous respirometry system for mice (Sable Systems International) at 15 months of age. Oxygen consumption (VO2) and carbon dioxide production (VCO2) were measured over 48 h, following an 8 h acclimation period, and maintained at approximately 22 °C under a 12:12 h light–dark cycle. Energy expenditure is shown a kcal h−1 corrected for lean mass. Mice were not given access to a running wheel.


Following 6 weeks on either a BCAA200 or BCAA100 diet, coronal dorsal raphe brain slices (280 µM) from 12-week-old C57BL/6J male mice were cut with the VT-1200 S Vibratome (Leica Biosystems). Slices were initially incubated in recovery solution containing 93 mM N-methyl-d-glucamine chloride, 2.5 mM KCl, 1.2 mM NaH2PO4, 30 mM NaHCO3, 20 mM HEPES, 25 mM d-glucose, 5 mM sodium ascorbate, 2 mM thiourea, 3 mM sodium pyruvate, 10 mM MgCl2, 0.5 mM CaCl2, pH 7.3, 300–310 mOsm l−1 heated at 34 °C and saturated with carbogen for 10 min and then stored in physiological saline (artificial cerebrospinal fluid (ACSF)) for at least 1 h either with or without the addition of Trp (50 µM). Slices were transferred to a recording chamber and superfused continuously at 2.5 ml min−1 with 34 °C ACSF containing 125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4 × 2H2O, 1 mM MgCl2, 2 mM CaCl2, 25 mM NaHCO3 and 11 mM d-glucose saturated with carbogen. Dorsal raphe nuclei were visualized with an Olympus BX51 microscope using Dodt gradient contrast optics. Whole-cell patch-clamp recordings were made from the dorsal raphe nuclei using an internal solution containing 95 mM potassium gluconate, 30 mM KCl, 10 mM HEPES, 2 mM EGTA, 15 mM NaCl, 1 mM MgCl2, 2 mM MgATP and 0.3 mM NaGTP. Neurons were voltage-clamped at −60 mV and liquid junction potentials of −12 mV were not corrected. Electrically evoked IPSCs were elicited via bipolar tungsten stimulating electrodes (10 stimuli delivered at 166 Hz, 100 V, 100 µs). Membrane currents were recorded with a Multiclamp 700B Amplifier (Molecular Devices), digitized and then acquired and analysed with the Axograph Acquisition software (Axon Instruments). All recordings were made in the presence of 100 µM picrotoxin (Sigma-Aldrich), 50 µM DL-2-amino-5-phosphonopentanoic acid (Abcam), 500 nM prazosin (Sigma-Aldrich), 10 µM cyanquixaline (Abcam) and 1 µM CGP 55845 (Tocris). Stock solutions of drugs were made in distilled water except for NAN-190 (Sigma-Aldrich), which was dissolved in dimethylsulfoxide, then diluted to their final concentration in ACSF and applied by superfusion.

Fluoxetine administration

Following 6 weeks on a BCAA200 diet, 12-week-old male C57BL/6J mice were gavaged with either fluoxetine (20 mg kg−1; Selleck Chemicals) or saline for 4 d. Food intake was recorded daily.

Gene expression

Total RNA from the liver and hypothalamus of 15-month-old male and female C57BL/6J mice were extracted using the TRIzol method (Sigma-Aldrich) and quantified spectrophotometrically using a NanoDrop (2000c, Thermo Fisher Scientific). For RNA-seq, hypothalamic RNA was sequenced by the Australian Genome Research Facility using an Illumina HiSeq 2500 System. The sequence reads were analysed according to Australian Genome Research Facility quality control measures. The cleaned sequence reads were then aligned against the Mus musculus genome (GRCm38). The TopHat aligner (v.2.0.14) was used to map the reads to the genomic sequences. The transcripts were assembled with StringTie v.1.2.4, using the read alignment with gencode M14. The fragments per kilobase of transcript per million mapped reads (FPKM) were generated using StringTie based on the formula FPKM = 106 C/(NL/103). C is the number of reads uniquely aligned to a gene, N is the total number of reads that is uniquely aligned to all genes and L is the number of bases of a gene. Heat maps were generated to model the relationship of BCAA intake with gene expression as measured by standardized z-scores. The rows are organized by hierarchical clustering using agglomerative clustering with complete linkage and Euclidean distance metric.

For real-time PCR, complementary DNA was synthesized (iScript; Bio-Rad Laboratories) and mRNA expression analysed using the SYBR Green methodology (LightCycler 480; Roche Molecular Systems). Primer pairs were designed using the Roche Universal Probe Library and BLASTed against the National Center for Biotechnology Information mouse genomic sequence database. Reactions were performed in triplicate and target gene expression normalized with eukaryotic elongation factor 2 as the endogenous control. The fold change was calculated based on a pooled sample. Primers were 300 nM in concentration and sequences for each gene are described: Npy-forward CGACACTACATCAATCTCATC, Npy-reverse AAGTTTCATTTCCCATCACC; Agrp-forward CTTCTTCAATGCCTTTTGC, Agrp-reverse TTTTTAAACCGTCCCATCC; Lepr-forward TGCTGAATTATACGTGATCG, Lepr-reverse AGACGTAGGATGAATAGATGG; Socs3-forward ATTTCGCTTCGGGACTAGC, Socs3-reverse AACTTGCTGTGGGTGACCAT; Pepck-forward CCAACGTGGCCGAGACTAGCG, Pepck-reverse GGCACATGGTTCCGCGTCCT; Eef2-forward TGTCAGTCATCGCCCATGTG, Eef2-reverse CATCCTTGCGAGTGTCAGTGA; Fgfr1c-forward TCCTCTTCTGGGTGTGC, Fgfr1c-reverse CTCCACTTCCACAGGGACTC; Klb-forward GAGGATGATCAGATCCGAAAGT, Klb-reverse AGCCTTTGATTTTGACCTTGTC.

Protein quantification

For all western blots, protein concentration was determined with a bicinchoninic acid assay (Sigma-Aldrich) and lysates loaded and standardized to contain 25 µg of protein. Liver, white adipose tissue and BAT from 15-month-old C57BL/6J male and female mice were resolved in either a 4–12% Bis-Tris gel (Bio-Rad Laboratories) or 3–7% Tris-Acetate Gel (Bio-Rad Laboratories). Following electrophoresis, proteins were transferred to a nitrocellulose membrane, blocked with 5% BSA in Tris-buffered saline with Tween 20 and incubated with various primary antibodies (all Cell Signaling Technology): mTOR, catalogue no. 2972; phospho-mTOR (Ser2448), catalogue no. 2971; UCP1 (D9D6X), catalogue no. 14670; ACLY (D1X6P), catalogue no. 13390; SCD1 (M38), catalogue no. 2438; Fatty Acid Synthase (C20G5), catalogue no. 3180; Acetyl-CoA Carboxylase, catalogue no. 3662; β-actin (13E5), catalogue no. 4970; cyclophilin A, catalogue no. 2175 (see Nature Research Reporting Summary) overnight at 4 °C. Bands were imaged following incubation for 1 h at room temperature in anti-rabbit horseradish peroxidase-linked secondary antibody (catalogue no. 7074; Cell Signaling Technology) on a ChemiDoc Imaging System (Bio-Rad Laboratories) using the ImageLab software v.5.1 to quantify relative protein expression.


Targeted LC–QQQ–MS analysis was performed to detect a different set of water-soluble metabolites in the positive and negative modes using an liquid chromatography mass spectrometry (LC–MS) system comprising a 1260 Infinity II LC System (Agilent) coupled to a QTRAP 5500 Mass Spectrometer (AB Sciex). Samples for analysis were from C57BL/6J 15-month-old male and female mice. For the hydrophilic interaction LC (HILIC) analysis, plasma and cortex samples (10 µl) were prepared via protein precipitation with the addition of 9 volumes of 74.9:24.9:0.2 v/v/v acetonitrile/methanol/formic acid containing stable isotope-labelled internal standards (l-valine-d8 (Sigma-Aldrich) and l-phenylalanine-d8 (Cambridge Isotope Laboratories)). For the amide analysis, plasma samples (30 µl) were prepared via protein precipitation with the addition of 70 µl of 75:25 v/v of acetonitrile/methanol containing stable isotope-labelled internal standards (Thymine-d4 (Sigma-Aldrich) and l-phenylalanine-d8 (Cambridge Isotope Laboratories)). The samples were centrifuged (20 min, 14,000 r.p.m., 4 °C) and the supernatants (10 µl) were injected directly onto a 2.1 × 150 mm, 3 µm Atlantis Silica HILIC Column (Waters) and a 4.6 × 100 mm, 3.5 μm XBridge Amide (Waters), for the HILIC and amide analyses, respectively. Eighty metabolites from the 84 metabolites optimized for positive mode detection were detectable in the plasma extracts. Of 110 metabolites optimized for negative mode, 73 were detectable in the mice plasma and were included in the final multiple reaction monitoring method. Raw data files (Analyst Software v.1.6.2; AB Sciex) were imported into the MultiQuant v.3.0 analysis software for multiple reaction monitoring Q1/Q3 peak integration; data were normalized relative to the pooled plasma samples that were analysed in the sample queue after every 10 study samples.

Immunohistochemistry and islet analysis

Paraffin-embedded pancreatic tissue from 15-month-old C57BL/6J male and female mice was sectioned and then stained as follows: sections were deparaffinized and rehydrated through a xylene-ethanol series and washed twice in PBS containing 0.1% BSA and 0.01% sodium azide (wash buffer). Slides were incubated with blocking buffer (DAKO) for 1 h at room temperature, then incubated with guinea pig anti-insulin (catalogue no. A0564; DAKO) and monoclonal anti-glucagon antibody produced in mouse (catalogue no. G2654; Sigma-Aldrich) overnight at 4 °C in a humidified chamber (see Nature Research Reporting Summary for antibody details). The following day, slides were washed twice in wash buffer, incubated in goat anti-guinea pig immunoglobulin G (IgG) and donkey anti-mouse IgG secondary antibodies (Alexa Fluor 488 and 594, respectively; Themo Fisher Scientific), washed twice again and then mounted in ProLong Diamond Antifade Mountant with 4,6-diamidino-2-phenylindole (Thermo Fisher Scientific). Slides were imaged with Leica DM6000 widefield and SP8 confocal microscopes.

Images were analysed with Fiji Image J (v.1.52b). The total pancreas area was calculated by tile-stitching entire pancreas sections in each slide. Fluorescence levels for each stain (insulin and glucagon) were thresholded by using control samples to exclude/eliminate any background fluorescence or noise. The threshold values were then applied to all sections, and intensity values were calculated by using the remaining normalized fluorescence readings. The total insulin- and glucagon-positive area were calculated; the total islet area was measured as a sum of total insulin and glucagon staining. At least ten islets per pancreas were averaged to obtain the pancreas measurement of each individual mouse.

Statistical analyses

Data are presented as the mean ± s.e.m. and significance was reached when P < 0.05 using R (v.3.4.1) and Prism (v.7.02; GraphPad Software). Comparisons between dietary treatments on various responses were analysed with analysis of variance (ANOVA); non-parametric data were analysed with a Kruskal–Wallis test. Sex was added as a cofactor and responses showing a significant diet × sex interaction are shown in Supplementary Fig. 3 and Supplementary Table 5. Comparisons of metabolic responses between ad libitum-fed and calorie-restricted animals were performed using a two-sided unpaired t-test. Details of the statistical tests for each graph are described in each figure legend. See the Nature Research Reporting Summary for data analysis and software specifications.

Survival data for C57BL/6J male and female mice and C3B6F1 female mice were analysed with the ‘survival’ package in R, using Cox’s proportional hazards models (CPHMs) implemented using the ‘coxph’ function58. CPHMs were used because they allowed us to explore the interactions between diet and sex effects on survival, as well as, where required, time-dependent effects in expanded CPHMs. Including the additive or interactive effects of sex alongside diet did not improve model fits based on the Akaike information criterion (AIC; BCAA data, Fig. 1n: AICdiet = 1,935.9, AICdiet + sex = 1,935.7, AICdiet × sex = 1,936; calorie restriction data, Fig. 7a: AICdiet = 2,075.6, AICdiet + sex = 2,075.6, AICdiet × sex = 2,076.2), suggesting diet alone was the best predictor of survival in the experiments.

For BCAA data from C57BL/6J male and female mice (Fig. 1n), we explored an expanded, time-dependent CPHM (survival curves cross around week 60); however, this model detected no significant interactions between time (pre- versus post-week 60) and diet (CPHM BCAA × time estimated log hazard ratio (lnHR) = 0.95, s.e.m. = 0.94, P = 0.31). CPHMs detected significant differences in survival between animals on BCAA200 and all other diets (BCAA20 versus BCAA200 estimated lnHR = −0.58, s.e.m. = 0.21, P < 0.005; BCAA50 versus BCAA200 estimated lnHR = −0.48, s.e.m. = 0.20, P < 0.05; BCAA100 versus BCAA200 estimated lnHR = −0.49, s.e.m. = 0.19, P < 0.005). Ten BCAA20 animals were killed before 60 weeks of age because of weight loss (n = 5) or seizures (n = 5) and were excluded from the survival analysis. For data comparing ad libitum-fed and calorie-restricted mice (Fig. 7a), significant differences were detected (CPHM: calorie-restricted versus ad libitum-fed P < 0.001; 100 versus 200 ad libitum-fed P = 0.01; 100 versus 200 calorie-restricted P = 0.68). For the exome-matched feeding experiments using C3B6F1 female mice, we found that CPHMs detected significant differences in survival on 23% protein diets relative to 6% protein diets (estimated lnHR = 0.68, s.e.m. = 0.21, P < 0.005), but not between 6% protein diets and 6% protein diets + BCAAs (estimated lnHR = −0.24, s.e.m. = 0.23, P = 0.30).

Metabolites from C57BL/6J male and female 15-month-old mice were correlated with BCAA intake; those with a Pearson correlation coefficient r > 0.1 were analysed in MetaboAnalyst v.4 using the Kyoto Encyclopedia of Genes and Genomes (KEGG; Mus musculus) pathway library (v.4.0;,60,61. Both positively and negatively correlated metabolite sets were analysed using the integrated pathway analysis. Significant Holm-adjusted P values are plotted in Fig. 3.

For the hypothalamic data from male and female C57BL/6J 15-month-old mice analysed using RNA-seq, the correlation coefficients of BCAA intake and the FPKM value of each gene were plotted as heat maps and show genes with a significant P. For the volcano plots, the FPKM data were filtered with averaged FPKM for each gene >0.1. The Pearson correlation coefficient and corresponding P for each gene were calculated. Correlation coefficients with less than 0.5th percentile or greater than 99.5th percentile were considered as significant and marked with red dashed lines. FPKM data were correlated with BCAA intake; genes with moderate-to-strong correlation (r > 0.3) were analysed in the Database for Annotation, Visualization and Integrated Discovery (v.6.8;,62. Both positively and negatively correlated gene sets were analysed using the Functional Annotation Tool and pathways assessed using the KEGG (Mus musculus) pathway library and shown in Supplementary Table 4.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

RNA-seq data have been deposited with the Gene Expression Omnibus and are accessible through accession number GSE114855. The data that support the plots within this article and other findings of this study are available from the corresponding author upon reasonable request.


  1. 1.

    Simpson, S. J. & Raubenheimer, D. The Nature of Nutrition: a Unifying Framework from Animal Adaption to Human Obesity (Princeton University Press, 2012).

  2. 2.

    Gosby, A. K. et al. Testing protein leverage in lean humans: a randomised controlled experimental study. PLoS ONE 6, e25929 (2011).

    CAS  Article  Google Scholar 

  3. 3.

    Simpson, S. J. & Raubenheimer, D. Obesity: the protein leverage hypothesis. Obes. Rev. 6, 133–142 (2005).

    CAS  Article  Google Scholar 

  4. 4.

    Le Couteur, D. G. The impact of low-protein high-carbohydrate diets on aging and lifespan. Cell. Mol. Life Sci. 73, 1237–1252 (2016).

    CAS  Article  Google Scholar 

  5. 5.

    Solon-Biet, S. M. et al. The ratio of macronutrients, not caloric intake, dictates cardiometabolic health, aging, and longevity in ad libitum-fed mice. Cell Metab. 19, 418–430 (2014).

    CAS  Article  Google Scholar 

  6. 6.

    Solon-Biet, S. M. et al. Macronutrient balance, reproductive function, and lifespan in aging mice. Proc. Natl Acad. Sci. USA 112, 3481–3486 (2015).

    CAS  Article  Google Scholar 

  7. 7.

    Wahl, D. et al. Comparing the effects of low-protein and high-carbohydrate diets and caloric restriction on brain aging in mice. Cell Rep. 25, 2234–2243.e6 (2018).

    CAS  Article  Google Scholar 

  8. 8.

    Grandison, R. C., Piper, M. D. & Partridge, L. Amino-acid imbalance explains extension of lifespan by dietary restriction in Drosophila. Nature 462, 1061–1064 (2009).

    CAS  Article  Google Scholar 

  9. 9.

    Miller, R. A. et al. Methionine-deficient diet extends mouse lifespan, slows immune and lens aging, alters glucose, T4, IGF-I and insulin levels, and increases hepatocyte MIF levels and stress resistance. Aging Cell 4, 119–125 (2005).

    CAS  Article  Google Scholar 

  10. 10.

    Harper, A. E. & Rogers, Q. R. Amino acid imbalance. Proc. Nutr. Soc. 24, 173–190 (1965).

    CAS  Article  Google Scholar 

  11. 11.

    Hasek, B. E. et al. Dietary methionine restriction enhances metabolic flexibility and increases uncoupled respiration in both fed and fasted states. Am. J. Physiol. Regul. Integr. Comp. Physiol. 299, R728–R739 (2010).

    CAS  Article  Google Scholar 

  12. 12.

    Soultoukis, G. A. & Partridge, L. Dietary protein, metabolism, and aging. Annu. Rev. Biochem. 85, 5–34 (2016).

    CAS  Article  Google Scholar 

  13. 13.

    Fontana, L. et al. Decreased consumption of branched-chain amino acids improves metabolic health. Cell Rep. 16, 520–530 (2016).

    CAS  Article  Google Scholar 

  14. 14.

    Newgard, C. B. et al. A branched-chain amino acid-related metabolic signature that differentiates obese and lean humans and contributes to insulin resistance. Cell Metab. 9, 311–326 (2009).

    CAS  Article  Google Scholar 

  15. 15.

    Maida, A. et al. Repletion of branched chain amino acids reverses mTORC1 signaling but not improved metabolism during dietary protein dilution. Mol. Metab. 6, 873–881 (2017).

    CAS  Article  Google Scholar 

  16. 16.

    She, P. et al. Obesity-related elevations in plasma leucine are associated with alterations in enzymes involved in branched-chain amino acid metabolism. Am. J. Physiol. Endocrinol. Metab. 293, E1552–1563 (2007).

    CAS  Article  Google Scholar 

  17. 17.

    Lackey, D. E. et al. Regulation of adipose branched-chain amino acid catabolism enzyme expression and cross-adipose amino acid flux in human obesity. Am. J. Physiol. Endocrinol. Metab. 304, E1175–1187 (2013).

    CAS  Article  Google Scholar 

  18. 18.

    Piccolo, B. D. et al. Whey protein supplementation does not alter plasma branched-chained amino acid profiles but results in unique metabolomics patterns in obese women enrolled in an 8-week weight loss trial. J. Nutr. 145, 691–700 (2015).

    CAS  Article  Google Scholar 

  19. 19.

    Fiehn, O. et al. Plasma metabolomic profiles reflective of glucose homeostasis in non-diabetic and type 2 diabetic obese African-American women. PLoS ONE 5, e15234 (2010).

    Article  Google Scholar 

  20. 20.

    Huffman, K. M. et al. Relationships between circulating metabolic intermediates and insulin action in overweight to obese, inactive men and women. Diabetes Care 32, 1678–1683 (2009).

    CAS  Article  Google Scholar 

  21. 21.

    Rose, W. C. II. The sequence of events leading to the establishment of the amino acid needs of man. Am. J. Public Health Nations Health 58, 2020–2027 (1968).

    CAS  Article  Google Scholar 

  22. 22.

    Reeds, P. J. Dispensable and indispensable amino acids for humans. J. Nutr. 130, 1835S–1840S (2000).

    CAS  Article  Google Scholar 

  23. 23.

    Piper, M. D. W. et al. Matching dietary amino acid balance to the in silico-translated exome optimizes growth and reproduction without cost to lifespan. Cell Metab. 25, 1206 (2017).

    CAS  Article  Google Scholar 

  24. 24.

    Breum, L., Rasmussen, M. H., Hilsted, J. & Fernstrom, J. D. Twenty-four-hour plasma tryptophan concentrations and ratios are below normal in obese subjects and are not normalized by substantial weight reduction. Am. J. Clin. Nutr. 77, 1112–1118 (2003).

    CAS  Article  Google Scholar 

  25. 25.

    Halford, J. C., Harrold, J. A., Lawton, C. L. & Blundell, J. E. Serotonin (5-HT) drugs: effects on appetite expression and use for the treatment of obesity. Curr. Drug Targets 6, 201–213 (2005).

    CAS  Article  Google Scholar 

  26. 26.

    Hong, S.-H. et al. Minibrain/Dyrk1a regulates food intake through the Sir2-FOXO-sNPF/NPY pathway in Drosophila and mammals. PLoS Genet. 8, e1002857 (2012).

    CAS  Article  Google Scholar 

  27. 27.

    Morton, N. M. et al. A stratified transcriptomics analysis of polygenic fat and lean mouse adipose tissues identifies novel candidate obesity genes. PLoS ONE 6, e23944 (2011).

    CAS  Article  Google Scholar 

  28. 28.

    Cai, D. & Liu, T. Hypothalamic inflammation: a double-edged sword to nutritional diseases. Ann. NY Acad. Sci. 1243, E1–E39 (2011).

    Article  Google Scholar 

  29. 29.

    Huang da, W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc. 4, 44–57 (2009).

    Article  Google Scholar 

  30. 30.

    O’Sullivan, J. F. et al. Dimethylguanidino valeric acid is a marker of liver fat and predicts diabetes. J. Clin. Invest. 127, 4394–4402 (2017).

    Article  Google Scholar 

  31. 31.

    Green, C. R. et al. Branched-chain amino acid catabolism fuels adipocyte differentiation and lipogenesis. Nat. Chem. Biol. 12, 15–21 (2016).

    CAS  Article  Google Scholar 

  32. 32.

    White, P. J. et al. The BCKDH kinase and phosphatase integrate BCAA and lipid metabolism via regulation of ATP-citrate lyase. Cell Metab. 27, 1281–1293.e7 (2018).

    CAS  Article  Google Scholar 

  33. 33.

    Wang, T. J. et al. Metabolite profiles and the risk of developing diabetes. Nat. Med. 17, 448–453 (2011).

    Article  Google Scholar 

  34. 34.

    Shah, S. H. et al. Branched-chain amino acid levels are associated with improvement in insulin resistance with weight loss. Diabetologia 55, 321–330 (2012).

    CAS  Article  Google Scholar 

  35. 35.

    Connelly, M. A., Wolak-Dinsmore, J. & Dullaart, R. P. F. Branched chain amino acids are associated with insulin resistance independent of leptin and adiponectin in subjects with varying degrees of glucose tolerance. Metab. Syndr. Relat. Disord. 15, 183–186 (2017).

    CAS  Article  Google Scholar 

  36. 36.

    Zheng, Y. et al. Cumulative consumption of branched-chain amino acids and incidence of type 2 diabetes. Int. J. Epidemiol. 45, 1482–1492 (2016).

    Article  Google Scholar 

  37. 37.

    Felig, P., Marliss, E. & Cahill, G. F. Jr. Plasma amino acid levels and insulin secretion in obesity. N. Engl. J. Med. 281, 811–816 (1969).

    CAS  Article  Google Scholar 

  38. 38.

    Lake, A. D. et al. Branched chain amino acid metabolism profiles in progressive human nonalcoholic fatty liver disease. Amino Acids 47, 603–615 (2015).

    CAS  Article  Google Scholar 

  39. 39.

    Goffredo, M. et al. A branched-chain amino acid-related metabolic signature characterizes obese adolescents with non-alcoholic fatty liver disease. Nutrients 9, E642 (2017).

    Article  Google Scholar 

  40. 40.

    Isanejad, M. et al. Branched-chain amino acid, meat intake and risk of type 2 diabetes in the Women’s Health Initiative. Br. J. Nutr. 117, 1523–1530 (2017).

    CAS  Article  Google Scholar 

  41. 41.

    Elshorbagy, A. K. et al. Food overconsumption in healthy adults triggers early and sustained increases in serum branched-chain amino acids and changes in cysteine linked to fat gain. J. Nutr. 148, 1073–1080 (2018).

    PubMed  Google Scholar 

  42. 42.

    Stöckli, J. et al. Metabolomic analysis of insulin resistance across different mouse strains and diets. J. Biol. Chem. 292, 19135–19145 (2017).

    Article  Google Scholar 

  43. 43.

    Gietzen, D. W., Hao, S. & Anthony, T. G. Mechanisms of food intake repression in indispensable amino acid deficiency. Annu. Rev. Nutr. 27, 63–78 (2007).

    CAS  Article  Google Scholar 

  44. 44.

    Rose, W. C. Feeding experiments with mixtures of highly purified amino acids. I. The inadequacy of diets containing nineteen amino acids. J. Biol. Chem 94, 155–165 (1931).

    CAS  Google Scholar 

  45. 45.

    Lynch, C. J. & Adams, S. H. Branched-chain amino acids in metabolic signalling and insulin resistance. Nat. Rev. Endocrinol. 10, 723–736 (2014).

    CAS  Article  Google Scholar 

  46. 46.

    Newgard, C. B. Interplay between lipids and branched-chain amino acids in development of insulin resistance. Cell Metab. 15, 606–614 (2012).

    CAS  Article  Google Scholar 

  47. 47.

    She, P. et al. Disruption of BCATm in mice leads to increased energy expenditure associated with the activation of a futile protein turnover cycle. Cell Metab. 6, 181–194 (2007).

    CAS  Article  Google Scholar 

  48. 48.

    Zhang, Y. et al. Increasing dietary leucine intake reduces diet-induced obesity and improves glucose and cholesterol metabolism in mice via multimechanisms. Diabetes 56, 1647–1654 (2007).

    CAS  Article  Google Scholar 

  49. 49.

    Hiroshige, K., Sonta, T., Suda, T., Kanegae, K. & Ohtani, A. Oral supplementation of branched‐chain amino acid improves nutritional status in elderly patients on chronic haemodialysis. Nephrol. Dial. Transplant. 16, 1856–1862 (2001).

    CAS  Article  Google Scholar 

  50. 50.

    D’Antona, G. et al. Branched-chain amino acid supplementation promotes survival and supports cardiac and skeletal muscle mitochondrial biogenesis in middle-aged mice. Cell Metab. 12, 362–372 (2010).

    Article  Google Scholar 

  51. 51.

    Crane, J. D. et al. Inhibiting peripheral serotonin synthesis reduces obesity and metabolic dysfunction by promoting brown adipose tissue thermogenesis. Nat. Med. 21, 166–172 (2015).

    CAS  Article  Google Scholar 

  52. 52.

    Fernstrom, J. D. Branched-chain amino acids and brain function. J. Nutr. 135, 1539S–1546S (2005).

    CAS  Article  Google Scholar 

  53. 53.

    Gietzen, D. W., Rogers, Q. R., Leung, P. M., Semon, B. & Piechota, T. Serotonin and feeding responses of rats to amino acid imbalance: initial phase. Am. J. Physiol. 253, R763–R771 (1987).

    CAS  PubMed  Google Scholar 

  54. 54.

    Neinast, M. D. et al. Quantitative analysis of the whole-body metabolic fate of branched-chain amino acids. Cell Metab. 29, 417–429.e4 (2019).

    CAS  Article  Google Scholar 

  55. 55.

    Dangin, M. et al. The digestion rate of protein is an independent regulating factor of postprandial protein retention. Am. J. Physiol. Endocrinol. Metab. 280, E340–E348 (2001).

    CAS  Article  Google Scholar 

  56. 56.

    Taylor, I. L., Byrne, W. J., Christie, D. L., Ament, M. E. & Walsh, J. H. Effect of individual l-amino acids on gastric acid secretion and serum gastrin and pancreatic polypeptide release in humans. Gastroenterology 83, 273–278 (1982).

    CAS  PubMed  Google Scholar 

  57. 57.

    Tordoff, M. G., Pearson, J. A., Ellis, H. T. & Poole, R. L. Does eating good-tasting food influence body weight? Physiol. Behav. 170, 27–31 (2017).

    CAS  Article  Google Scholar 

  58. 58.

    Chong, J. et al. MetaboAnalyst 4.0: towards more transparent and integrative metabolomics analysis. Nucleic Acids Res. 46, W486–W494 (2018).

    CAS  Article  Google Scholar 

  59. 59.

    Xia, J. & Wishart, D. S. Web-based inference of biological patterns, functions and pathways from metabolomic data using MetaboAnalyst. Nat. Protoc. 6, 743–760 (2011).

    CAS  Article  Google Scholar 

  60. 60.

    Xia, J. & Wishart, D. S. Metabolomic data processing, analysis, and interpretation using MetaboAnalyst. Curr. Protoc. Bioinformatics 34, 14.10.1–14.10.48 (2011).

    Article  Google Scholar 

  61. 61.

    Huang da, W., Sherman, B. T. & Lempicki, R. A. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 37, 1–13 (2009).

    Article  Google Scholar 

  62. 62.

    Therneau, T. M. & Grambsch, P. M. Modeling Survival Data: Extending the Cox Model (Springer, 2000).

Download references


We thank F. Held and P. Telleria Teixeira for their technical and administrative support. We thank the Laboratory Animal Services at the University of Sydney, N. Sunn of Sydney Imaging, W. Potts from Specialty Feeds, L. McQuade from the Australian Proteome Analysis Facility and the Diagnostic Pathology Unit at Concord Hospital. This work is supported by a National Health and Medical Research Council (NHMRC) project grant (GNT1084267 to D.R., D.L.C. and S.J.S.), the Ageing and Alzheimers Institute and the Sydney Food and Nutrition Network. S.S.B. is supported by the NHMRC Peter Doherty Biomedical Fellowship (no. GNT1110098) and the University of Sydney SOAR fellowship. A.M.S. was supported by a Discovery Early Career Researcher Award from the Australian Research Council (DE180101520). V.C.C. is supported by a University of Sydney Equity Fellowship. L.P. and P.J. were supported by the Max Planck Society and acknowledge funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7, 2007–2013)/ERC grant agreement no. 268739 and the Wellcome Trust (098565/Z/12/).

Author information




D.L.C., S.J.S. and L.P. conceived the study. S.S.B., S.J.S. and D.L.C. wrote the manuscript. D.R., L.P. and A.J.R. reviewed and assisted in writing the manuscript. S.S.B., V.C.C. and T.P. ran the study. G.J.C., D.W., X.C., A.E.B., E.B., G.C.G., R.P., J.O.S., Y.C.K., M.K., B.Y., C.A., G.S., T.D., J.A.W. and P.J. ran the experiments. S.S.B., A.M.S., Q.-P.W., K.B.A., M.D.W.P. and P.J. analysed the data.

Corresponding authors

Correspondence to Samantha M. Solon-Biet or Stephen J. Simpson.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Figs. 1–6, Supplementary Tables 1–4 and Supplementary Table 6

Reporting Summary

Supplementary Table 5

Statistical summary for the effects of diet, sex or diet–sex interaction

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Solon-Biet, S.M., Cogger, V.C., Pulpitel, T. et al. Branched-chain amino acids impact health and lifespan indirectly via amino acid balance and appetite control. Nat Metab 1, 532–545 (2019).

Download citation

Further reading


Sign up for the Nature Briefing newsletter for a daily update on COVID-19 science.
Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing