Amino Acid Sensing in Metabolic Homeostasis and Health

Abstract

Sensing and responding to changes in nutrient levels, including those of glucose, lipids, and amino acids, by the body is necessary for survival. Accordingly, perturbations in nutrient sensing are tightly linked with human pathologies, particularly metabolic diseases such as obesity, type 2 diabetes mellitus, and other complications of metabolic syndromes. The conventional view is that amino acids are fundamental elements for protein and peptide synthesis, while recent studies have revealed that amino acids are also important bioactive molecules that play key roles in signaling pathways and metabolic regulation. Different pathways that sense intracellular and extracellular levels of amino acids are integrated and coordinated at the organismal level, and, together, these pathways maintain whole metabolic homeostasis. In this review, we discuss the studies describing how important sensing signals respond to amino acid availability and how these sensing mechanisms modulate metabolic processes, including energy, glucose, and lipid metabolism. We further discuss whether dysregulation of amino acid sensing signals can be targeted to promote metabolic disorders, and discuss how to translate these mechanisms to treat human diseases. This review will help to enhance our overall understanding of the correlation between amino acid sensing and metabolic homeostasis, which have important implications for human health.

Graphical Abstract

amino acids, sensing signals, metabolic homeostasis, mTORC1, GCN2

ESSENTIAL POINTS

• Increased levels of circulating branched-chain amino acids have been correlated with obesity, insulin resistance, type 2 diabetes, and other complications of metabolic syndrome in mice and humans

• AA sensing signaling pathways, in which mTORC1 senses AA abundance, while GCN2 senses AA deficiency, can regulate metabolic processes

• Amino acid sensing signals are integrated and coordinated at the organismal level, and then maintain whole metabolic homeostasis

• Human metabolic diseases and cancer associated with defects in amino acid sensing

• Understanding amino acid sensing mechanisms will enable our effective exploitation of strategies for the prevention or treatment of metabolic diseases and cancer

Introduction to Amino Acid Sensing

The 3 macronutrients glucose, lipids, and amino acids (AAs) are the main substances that maintain life. AAs, as the basic component of proteins, are also essential substrates for the synthesis of peptide hormones and other small molecules, such as serotonin and thyroid hormones, and, therefore, these have enormous physiological importance (1, 2).

According to the growth or nitrogen balance of animals, AAs have traditionally been classified as either essential amino acids (EAAs) or nonessential amino acids (NEAAs) (1). EAAs are defined as AAs whose carbon skeletons cannot be de novo synthesized or are insufficiently synthesized by cells relative to metabolic needs (2). Human proteins are assembled from 20 AAs, 9 of which are EAAs: leucine, isoleucine, valine, tryptophan, phenylalanine, lysine, methionine, threonine, and histidine. Among these are the following 3 branched-chain amino acids (BCAAs), leucine, isoleucine, and valine, which are thus named because they contain branched rather than linear aliphatic side chains (2). In addition, the oxidized hydrocarbon skeletons produced from AA breakdown during metabolism may be degraded to form amphibolic intermediates that can be converted into glucose or ketone bodies, which can finally be converted to fatty acids. In this regard, AAs can also be classified as glucogenic (glucose forming) or ketogenic (ketone body forming). Some AAs, including phenylalanine, isoleucine, threonine, tyrosine, and tryptophan, are both glucogenic and ketogenic (3).

The basic nutritional functions of AAs are well known; in recent years, the role of AAs as signaling molecules in regulating energy and metabolic homeostasis has attracted considerable attention. Thus, AAs not only participate in protein metabolism, but are also key molecular regulators of glucose and lipid metabolism. In this regard, AAs are primarily known for their function to increase protein synthesis and decrease proteolysis (4, 5). For example, AAs, specifically leucine as the primary AA, that activate the mechanistic target of rapamycin complex 1 (mTORC1) signaling pathway, which is essential for the initiation of protein synthesis through activating ribosomal protein S6 kinase 1 (S6K1) and inhibiting eukaryotic translation initiation factor 4E binding protein-1 (4EBP1) (6, 7). In fact, increasing AA levels enhance the expression of hepatic fatty acid biosynthetic gene in an mTORC1-dependent manner, while attenuating insulin-mediated inhibition of gluconeogenesis (8). In contrast, AA starvation inhibits liver gluconeogenesis (9) and adipocyte differentiation (10).

Since AA homeostasis is important in all living organisms, and for human metabolism and diseases in particular, we raise some questions: How does the body sense the availability of AAs? Will disturbance of the AA sensing mechanisms lead to metabolic disorders and diseases? The ability of an organism to modulate cellular processes based on nutrient availability is fundamental to life. These nutrient-dependent cellular processes, broadly termed “nutrient sensing,” contains a broad array of processes and pathways including nutrient transport, processing, and metabolic control (11). For example, the body’s sensing of a concentration increase in a particular nutrient may direct the binding of the sensed substance through the sensor onto the cell membrane or intracellularly, which initiates the transport process or signaling transduction. Alternatively, the body’s sensing of fluctuation in a nutrient may occur in an indirect manner through determining the surrogate molecule that reflects its abundance, such as the metabolites. Different pathways that sense intracellular and extracellular concentrations of AAs are integrated and coordinated at the organismal level and can maintain whole metabolic homeostasis. The ability of the body to respond to changes in AA availability is mediated by a complex regulatory network comprising various dynamic players, which are crucial for initiating downstream effectors at the cellular or molecular levels (12). In times of AA abundance, AA sensing signals engage anabolic and storage pathways, while scarcity triggers catabolism, such as degradation of muscle protein through autophagy. Considering the importance of AA sensing pathways for human health in particular, in this review we will discuss the literature describing how the important AA sensing signals respond to AA concentration fluctuations, and further how these sensing signals modulate metabolic processes, including energy, glucose, and lipid metabolism. We also provide updates on the importance of AA sensing signaling pathways in metabolic diseases, including obesity, type 2 diabetes mellitus (T2DM), and cancer, and whether dysregulation of these pathways can be targeted to promote metabolic disorders, as this has important implications for human health.

AA Sensing Signaling Pathways

mTORC1 signaling pathway

The mTORC1 signaling pathway senses and integrates AA signaling to regulate organismal growth and metabolic homeostasis. mTOR is a nonclassical serine/threonine protein kinase belonging to the phosphoinositide 3-kinase–related kinase family, and often forms the protein complexes mTORC1 or mTORC2 through interacting with distinct protein components (13). mTORC1 contains 3 core components: mTOR, regulatory protein associated with mTOR (Raptor), and mammalian lethal with Sec13 protein 8 (mLST8; also known as GβL) (13). The function of Raptor is facilitating substrate recruitment to mTORC1 by binding to the TOR signaling motif (14), and facilitating subcellular localization of mTORC1. In contrast, mLST8 is associated with the catalytic domain of mTORC1, which may have the role of stabilizing the kinase activation loop (15). Over the years, numerous other proteins have also been associated with AA sensing by mTORC1, including the mammalian vacuolar protein sorting 34 homolog (16), mitogen-activated protein kinase kinase kinase kinase (17) and inositol polyphosphate monokinase (18), but whether and how these proteins connect to the Rag-Ragulator system remains unknown. mTORC2, comprising mTOR, rapamycin-insensitive companion of mTOR (Rictor), mLST8, and mammalian stress-activated MAPK-interacting protein 1 (mSIN1; also known as MAPKAP1), is not sensitive to AAs and regulates a unique set of signaling pathways, including protein kinase C, AGC family kinases of protein kinase B/Akt, and serum- and glucocorticoid-inducible kinase 1 (19). Since numerous studies have demonstrated that the remarkable contribution of mTORC1 in metabolic regulation due to its significant sensitivity to AAs, hereafter we focus on the recent advances in the understanding of cellular AA sensing to mTORC1 activation.

Rag GTPase-dependent mTORC1 activation

Several studies have revealed that the lysosomal surface is the major site for mTORC1 activation, especially after the identification of the Rag GTPases in 2008 (20, 21). More recently, 2 publications almost at the same time reported building a 3-dimensional structural model of mTORC1 bound to Rag GTPases, small GTPase RAS homologue enriched in brain (Rheb), and lysosomal scaffold Ragulator on the lysosomal surface, which further provide extensive mechanistic details and substantially deepen our understanding of how AAs regulate mTORC1 activity at the lysosome (22, 23). AAs accumulated inside the lysosome generate a signal which is communicated to the Rag GTPases through the activation of the Ragulator–vacuolar H+-ATPase (v-ATPase) complex. Subsequently, mTORC1 is recruited to the lysosomal surface by active Rags, where it is activated by Rheb, and triggers mTORC1 signaling (24). During times of AA abundance, mTORC1 kinase activity is stimulated after translocation from the cytosol to the lysosomal membrane, where it interacts with Rheb-Rag GTPases and responds to AAs within the cell by altering their nucleotide state and their interaction with Ragulator (25) (Fig. 1).

Figure 1.

mTORC1 sensing signals in mammals. In the condition of amino acid (AA) abundance, mTORC1 kinase activity is stimulated following translocation from the cytosol to the lysosomal membrane, where it interacts with Rheb–Rag GTPases (RagA or RagB bound to RagC or RagD) (22, 23). Sestrin2 is reported to be a cytoplasmic leucine sensor, which is a negative regulator of both GATOR2 and RagA/B. Leucine binds to Sestrin2 and inhibits Sestrin2–GATOR2 interaction (26). Leucyl-tRNA synthetase (LRS) can also sense intracellular leucine and induce mTORC1 activation. This is achieved via direct binding of LRS to Rag GTPase, where it functions as a GTPase-activating protein for RagD (27). Similarly, arginine binds to sensor CASTOR1, and then disrupts the interaction of CASTOR1 with GATOR2, thereby leading to mTORC1 activation (28). SLC38A9 is an arginine transceptor in the lysosomal membrane that interacts with Rags and Ragulator to activate mTORC1 (29, 30). Methionine causes methyl donor S-adenosylmethionine (SAM) levels to increase and disrupts the SAMTOR–GATOR1 complex by binding directly to SAMTOR, thereby activating mTORC1 signaling (31). Glutamine and asparagine activate mTORC1 through Arf1 in the absence of Rag GTPases (32). Glutamine is also known to activate mTORC1 through the RagB-GTPase–dependent mechanism and through enhancing glutaminolysis and α-ketoglutarate production, which acts as a cofactor for prolyl hydroxylases that enhances GTP loading of RagB (33). In particular, AAs promote GTP loading of Rab1A, which in turn stimulates Rheb–mTORC1 interaction specifically in the Golgi (34) and AA transceptor PAT4 (SLC36A4), which has recently been shown to be predominantly localized on the trans-Golgi network interacting with Rab1A, Raptor, and mTOR (35). Arrows and bars represent activation and inhibition of downstream proteins, respectively. Dashed arrows represent translocation of AAs from cytoplasm to nucleus.

Sestrin2 and CASTOR1 have been identified as cytosolic sensors of leucine and arginine, respectively (26, 28), and the solute carrier (SLC) transporter SLC38A9 as a lysosomal arginine sensor (29, 30). Upstream of RagA/B is a 3-protein complex, known as GATOR1 that acts as a GTPase activating protein (GAP), which represses mTORC1 activation. Furthermore, GATOR1 activity is also inhibited by another 5-protein complex GATOR2, which acts as a positive regulator of mTORC1. A previous study has demonstrated that the transport of leucine into the cell in exchange for glutamine is facilitated by the SLC family protein complex SLC7A5–SLC3A2 (36). When the intracellular level of leucine increases, it disrupts the interaction of GATOR2 with Sestrin2, which ultimately allows activation of mTORC1 (26). Similarly, arginine binds to the sensor CASTOR1, and then disrupts interaction of CASTOR1 with GATOR2, thereby leading to mTORC1 activation (28). Therefore, both leucine and arginine bind to their sensors and stimulate mTORC1 activity through releasing their inhibition of GATOR2, suggesting GATOR2 as a central node during signal transduction of AAs to mTORC1. SLC38A9, a member of an AA transporter family, is found to be an arginine transceptor that interacts with Rags and Ragulator onto the lysosomal membrane to activate mTORC1 (29, 30). In addition, leucyl-tRNA synthetase (LRS) is also reported to sense intracellular leucine concentrations and induce mTORC1 activation (27). This is achieved by direct binding of LRS to the Rag GTPase, where it functions as a GAP for RagD (27). Additionally, a study by Gu et al. (31) identified SAMTOR, an S-adenosylmethionine (SAM) sensor that inhibits mTORC1 signaling by interacting with GATOR1; in contrast, methionine causes the methyl donor SAM levels to increase and disrupt the SAMTOR–GATOR1 complex through directly binding to SAMTOR, thereby activating mTORC1 signaling (Fig. 1).

The AA sensing mechanism of TORC1 is evolutionarily conserved from yeast to mammals. The yeast vacuole has a function similar to the lysosome in higher organisms, and yeast TORC1 localizes to the vacuolar surface. Similar to Rag proteins in mammals, Gtr1 (an ortholog of RagA and RagB) and Gtr2 (an ortholog of RagC and RagD) form a heterodimer in which the active conformation is Gtr1–GTP in complex with Gtr2–GDP (37, 38). Binda et al. (39) reported that Gtr1 and Gtr2 are also activators of TORC1 in response to leucine increase in yeast, and function as the Rag family proteins. Cdc60, a yeast ortholog of LRS, is necessary and sufficient to activate TORC1 by interacting with Gtr1 in response to leucine stimulation (40). Similar to mammals, upstream of Gtr1 is a GAP trimeric SEACIT, the yeast GATOR1 ortholog, which is inhibited by the yeast GATOR2 ortholog pentameric SEACAT (41, 42). In addition, another study indicated that the yeast Ego1–Ego2–Ego3 ternary EGO complex (Egoc) is an ortholog of the mammalian Ragulator complex, which plays a role similar to that of a scaffold (43). However, unlike in mammals, methionine can activate TORC1 in yeast following its use in the synthesis of SAM, which increases Ppm1-mediated methylation of Pph21/Pph22, the catalytic subunits of phosphatase PP2A (44). Furthermore, the methylated PP2A prevents assembly of the SEACIT complex through dephosphorylating Npr2, leading to Gtr1-mediated activation of TORC1 in yeast (44) (Fig. 2).

Figure 2.

TORC1 sensing signals in yeast. The yeast vacuole plays a similar function to the lysosome in mammals, and yeast TORC1 localizes to the vacuolar surface. Like Rag proteins in mammals, Gtr1 is an ortholog of RagA and RagB, while Gtr2 is an ortholog of RagC and RagD (37, 38). Cdc60 is a yeast ortholog of LRS, which is necessary and sufficient to activate TORC1 by interacting with Gtr1, in response to leucine stimulation (40). Similar to mammals, upstream of the Gtr1 is a GAP trimeric SEACIT, the yeast GATOR1 ortholog, which is inhibited by the pentameric SEACAT, the yeast GATOR2 ortholog (41, 42). The yeast Ego1–Ego2–Ego3 ternary EGO complex (Egoc) is a structural ortholog of the Ragulator complex, which plays a similar role to a scaffold (43). Methionine, as a methyl donor S-adenosylmethionine (SAM), can promote PP2A methylation, and thus dephosphorylates Npr2 to prevent assembly of the SEACIT complex, leading to Gtr1-mediated activation of TORC1 (44). Glutamine directly binds to yeast Pib2, promoting Pib2–TORC1 complex formation at the vacuolar membrane, an interaction which is required for TORC1 activation (45). Whi2 is a newly identified amino acid sensor that detects declining levels of leucine and negatively regulates TORC1 activity through the formation of a Whi2–Psr1/Psr2 complex (46). Egoc is separately sorted to vacuoles and perivacuolar endosomes, where it assembles Egocs that regulate spatially distinct pools of TORC1 and thereby play different roles; vacuolar TORC1 promotes protein synthesis through targeting Sch9, whereas endosomal TORC1 inhibits macroautophagy and ESCRT-driven microautophagy through phosphorylation of Atg13 and Vps27, respectively (47). Arrows and bars represent activation and inhibition of the downstream protein, respectively.

Rag GTPase-independent mTORC1 activation

Unlike leucine, arginine, and methionine, which signal to mTORC1 through the well-known Rag GTPase signal transduction, glutamine and asparagine activate mTORC1 through Rag GTPase-independent mechanisms (32,45,48). Jewell et al. (48) indicated that glutamine promoted mTORC1 translocation to the lysosome membrane in RagA and RagB knockout (KO) cells in a manner which required v-ATPase but not Ragulator. Moreover, they found that the adenosine diphosphate ribosylation factor-1 (Arf1) GTPase was necessary for mTORC1 activation and localization by glutamine. More recently, Meng et al. (32) found that, similar to glutamine, asparagine also signals to mTORC1 through the Arf1 GTPase in the absence of the Rag GTPases (Fig. 1). Similarly, glutamine can also directly bind to yeast Pib2, resulting in enhanced Pib2–TORC1 complex formation at the vacuolar membrane, which is required for TORC1 activation by glutamine (45). However, it should be noted that glutamine is also known to activate mTORC1 through the RagB-GTPase-dependent mechanism in cancer cells. Durán et al. (33) demonstrated that glutamine in combination with leucine activates mTORC1 through increasing glutaminolysis and α-ketoglutarate production, which acts as a cofactor for prolyl hydroxylases that enhance GTP loading of RagB in U2OS cells (Fig. 1). This reveals a difference in glutamine sensing between normal and cancer cells and may provide an explanation why glutamine is essential for cancer cells.

Recent study by Chen et al. (46) indicated that yeast Whi2 is necessary for inhibiting TORC1 activity under conditions of AA scarcity. Whi2 independently inhibits TORC1 activity through the SEACIT/Gtr pathway, similar to the GATOR1/Rag pathway in mammals. Moreover, they found that the plasma membrane-associated phosphatases, Psr1 and Psr2, which are necessary for Whi2 inhibition of TORC1, interact with Whi2; however, their function remains unclear (46, 49). Therefore, Whi2 is a newly identified AA sensor in yeast that detects declining levels of leucine and negatively regulates TORC1 activity through the formation of a Whi2–Psr1/Psr2 complex (Fig. 2).

Spatial localization of mTORC1 in AA sensing

Some AAs are the most potent component for mTORC1 activation and are responsible for its localization to the lysosomal membrane, although it is becoming increasingly clear that mTORC1 is controlled from other subcellular sites, such as the Golgi. Thomas et al. (34) identified that Rab1A (the yeast ortholog is Ypt1) is a conserved regulator of AA signaling to mTORC1 in a Rag-independent manner. Mechanistically, AAs increase GTP loading of Rab1A that in turn promotes Rheb–mTORC1 interaction specifically in the Golgi of HEK293E cells (Fig. 1). In particular, they found that Rab1A is overexpressed in colorectal cancer, which is positively correlated with mTORC1 activity, tumor proliferation and invasion, and progression (34). However, upstream of Rab1A and how the signal of AA availability is transmitted to Rab1A remain to be determined. AA transceptor PAT4 (SLC36A4) has recently been demonstrated to be predominantly localized on the trans-Golgi network that interacts with Rab1A, Raptor, and mTOR in HCT116 colorectal cancer cells (35) (Fig. 1). Together, these studies support the model that an mTORC1 signaling hub is assembled on the Golgi. The localization of mTORC1 on the Golgi as a result of Rab1A overexpression appears to be cancer cell specific and suggests that aberrant AA sensing signals are a common mechanism to stimulate oncogenic proliferation and transformation.

The finding reported by Song et al. (50) that mTOR colocalizes with the lysosome in basal conditions and that translocation of these complexes (mTOR, Rheb, and eIF3F) to the cell membrane occurs in human muscle in response to resistance exercise, which appears to be primarily influenced by muscle contraction, is of particular interest. Furthermore, Hodson et al. (51) found that mTORC1 location was redistributed within the cell in response to resistance exercise and protein–carbohydrate feeding, whereas localization of mTORC2 was not altered. These observations suggest that mTOR lysosome translocation in response to anabolic signals is driven primarily by mTORC1. Consistent with this, cellular trafficking of mTOR occurring in human muscle is a key event in the process of mTORC1 activation. Collectively, these studies further highlight the importance of spatial regulation of mTORC1 under conditions of anabolic stimulation. However, the relevance of mTORC1 colocalization with clinical metabolic diseases, such as obesity, T2DM, or cancer, needs further study.

In addition, in yeast, Gtr GTPases combine to form structurally conserved multiprotein complexes, such as Egoc, to activate TORC1 in response to AA stimulation. A recent study by Hatakeyama et al. (47) indicated that in yeast Egoc is membrane anchored after lipid modification mediated by Ego1, making it travel through the trans-Golgi network. Furthermore, Egoc is selectively anchored to vacuoles or perivacuolar endosomes, where it assembles Egocs that regulate distinct TORC1 pools and thereby play different roles; vacuolar TORC1 promotes protein synthesis through targeting Sch9, a mammalian ortholog of S6K, whereas endosomal TORC1 inhibits macroautophagy and ESCRT-driven microautophagy through phosphorylation of Atg13 and Vps27, respectively (47) (Fig. 2). Therefore, vacuolar and endosomal TORC1 spatial localization are key cues for protein synthesis and degradation, respectively. These observations collectively suggest that alternative mTORC1 signaling hubs and multiple layers of regulatory mechanisms on mTORC1 stimulation potentially respond to different types or concentrations of AA or different subcellular cues.

General control nonderepressible 2 signaling pathway

General control nonderepressible 2 (GCN2) belongs to the conserved serine/threonine kinase family that senses AA deficiency through binding of uncharged tRNAs (52). In mammals, GCN2 is expressed in almost all tissues but is predominant in brain, liver, and skeletal muscle (53). The mechanism of GCN2 activation has been well characterized in Saccharomyces cerevisiae, where GCN2 is the only kinase that phosphorylates eukaryotic initiation factor 2α(eIF2α). A previous study demonstrated that the interaction of GCN2 with GCN1 through its RING finger-containing proteins, WD-repeat-containing proteins, and yeast DEAD (DEXD)-like helicases domain is necessary for its activation; both proteins are tethered to the ribosome (54). Moreover, the key role of GCN1 is to transfer GCN2 uncharged tRNAs that enter the A site of the ribosomes, thereby activating GCN2 (54). IMPACT (imprinted and ancient) and its yeast homolog Yih1 harbor an RWD domain that competes with GCN2 for the binding of GCN1, thereby inhibiting GCN2 activity (55, 56) (Fig. 3). Studies have found that overexpression of IMPACT or Yih1 in mammalian or yeast cells, respectively, compromises AA starvation-induced eIF2α phosphorylation (56, 57). As with other components of the GCN2 signaling pathway, IMPACT/Yih1 is expressed in virtually all eukaryotic cells (58).

Figure 3.

GCN2 signaling pathways in amino acid (AA) sensing. GCN2 interacts with GCN1 through its RWD domain, which is necessary for its activation; both proteins are tethered to the ribosome (54). IMPACT harbors an RWD domain that competes with GCN2 for binding of GCN1, thereby inhibiting GCN2 activity (55, 56). Under AA starvation, binding of GCN2 to uncharged tRNA induces conformational changes that activate its kinase activity, which functions phosphorylation of eukaryotic initiation factor 2α (eiF2α), causing a global suppression of translation, but increased transcription of activating transcription factor 4 (ATF4) (59-61). ATF4 is a transcription factor that plays a critical role in adaptive cellular response through targeting stress response genes, such as Sestrin2, fibroblast growth factor 21 (FGF21), and SLC1A5 (62-65). Sestrin2 is transcriptionally induced upon prolonged AA starvation through ATF4 and leads to the inhibition of mTORC1 (66). FGF21 can repress insulin- or nutrient-stimulated activation of mTORC1 (67). Arrows and bars represent activation and inhibition of the downstream protein, respectively.

Under AA starvation, levels of uncharged tRNA are elevated and the resulting accumulation is sensed by GCN2 kinase. Binding of GCN2 to uncharged tRNA induces conformational transformation that activates its kinase activity, resulting in phosphorylation of eIF2α, causing a global suppression of translation; however, it also results in increased transcription of starvation-relevant transcripts, including activating transcription factor 4 (ATF4) (59-61). ATF4 is a transcription factor that plays a critical role in adaptive cellular response through targeting stress response genes (68). For example, fibroblast growth factor 21 (FGF21), a member of the FGF family, exhibits endocrine properties and is predominantly produced by the liver (69). FGF21 is a target gene of ATF4 and is critical to the adaptive metabolic response to AA deprivation (62, 63). Some SLC family factors of AA transporters, like SLC1A5, are also ATF4 targets under AA deprivation (64, 65) (Fig. 3).

GCN2 has also been connected to mTORC1 activation. Sestrin2 is transcriptionally induced by ATF4 upon prolonged AA starvation; therefore, activation of the GCN2/ATF4 signaling pathway in AA-deprived cells results in mTORC1 inhibition (66). In addition, GCN2 can also inhibit mTORC1 activity through an ATF4-independent mechanism under leucine or arginine deprivation (70). Although a study reported that FGF21 was a downstream factor that mediated mTORC1-regulated whole-body metabolism in muscle (71), FGF21 can repress insulin- or nutrient-stimulated activation of mTORC1 in the liver (67). Hence, activation of GCN2 and inhibition of mTORC1 at the organismal level act together to maintain metabolic homeostasis under conditions of AA deficiency.

Other AA sensing signals

Apart from GCN2 and mTORC1 signals, exploration of other AA sensing receptors and signaling pathways remains a current topic in the field of metabolic research. Recent advances have shown that AAs and peptides are sensed by transporters or specific “taste” receptors located in the cell membranes.

Enteroendocrine L cells are open-type enteroendocrine cells (EECs) that play an important role in AA sensing. The proton-coupled peptide transporter PepT1 (SLC15A1) was demonstrated to be a sensor that together with the Ca sensing receptor (CaSR), contributes to glucagon-like peptide-1 (GLP-1) release from enteroendocrine L cells (72, 73). Beyond this, CaSR plays a role in the sensing of primarily aromatic AAs (AAAs), including tryptophan and phenylalanine, and, in a more moderate fashion, the sensing of some aliphatic and polar AAs, such as alanine (74, 75). AAs and Ca, the main ligands of CaSR, have 2 different binding sites (76). CaSR becomes an AAA receptor up to 1 mM Ca²⁺ threshold concentrations (77). In particular, CaSR has been identified in cells producing somatostatin, gastrin, cholecystokinin, GLP-1, and gastric inhibitory polypeptide (GIP), and has been associated with hormone secretion (78, 79).

G-protein–coupled receptor 142 (GPR142), which is highly expressed in pancreatic islets, has recently been identified as a highly selective sensor of essential AAAs, in particular, tryptophan (80). In addition, the G-protein–coupled receptor family C subtype 6A (GPRC6A) that senses basic AAs, including arginine and lysine, and which is expressed in the gastric mucosa and pancreas, but not in the small intestine, could induce gastric acid and pepsinogen secretion as CaSR (81). Moreover, GPRC6A activity seems to be dependent on Ca²⁺ and has the same Ca binding site as CaSR (82). It therefore appears that CaSR and GPRC6A are very similar, except in the nature of AAs that they can sense. Other EECs, such as G cells, and a small proportion of D cells, also expressing GPRC6A could respond to basic AAs, including lysine, arginine, and ornithine (83).

Another AA “taste” receptor, named as the umami receptor, is expressed by gastric A cells and I cells in the small intestine and is activated by glutamate (84). In the intestinal lumen, monosodium glutamate is acknowledged to generate the “umami” taste through interaction with the taste receptor T1R1/T1R3 (85). The natural L stereoisoform can sense and bind to the heterodimer T1R1/T1R3, whereas D-AAs, which have a sweet taste in general, are sensed by the heterodimer receptor T1R2/T1R3, similar to artificial sweeteners and monosaccharides (86, 87). The distributions and functions of AA sensing receptors need further investigation.

Role of AA Sensing Signals in Metabolism

Under the condition of variable nutrient availability, most organisms have the ability to efficiently transition between anabolic and catabolic states through the evolved nutrient sensing mechanisms, allowing them to grow and survive. Particularly, AA sensing signaling pathways, in which mTORC1 senses AA abundance, while GCN2 senses AA deficiency, in the central hypothalamus or main peripheral metabolic organs (such as the intestine, liver, pancreas, muscle, and adipose tissue) can regulate metabolic processes, including energy, glucose, and lipid metabolism (Fig. 4).

Figure 4.

Amino acid (AA) sensing signals are integrated and coordinated at the organismal level, and then maintain whole metabolic homeostasis. The ability of a body to respond to changes in AA availability is mediated by a complex regulatory network composed of various dynamic players, which are crucial for initialing downstream effectors at cellular or molecular levels. Particularly, AA sensing signaling pathways, in which mTORC1 senses AA abundance and GCN2 senses AA deficiency, in the central hypothalamus or main peripheral metabolic organs (such as intestine, liver, pancreas, muscle, and adipose tissue) can regulate metabolic processes, including energy, glucose, and lipid metabolism.

mTORC1 and metabolic control

Energy metabolism

The central hypothalamus is the main brain area in which the central nervous system regulates peripheral metabolic signals (88). The hypothalamus can sense changes in AA concentrations through the mTORC1 signaling pathway and integrates peripheral signals to control energy metabolic processes, such as feeding and thermogenesis. The mediobasal hypothalamus (MBH), which contains arcuate and ventromedial nuclei, is a critical node in the neural network where central AA sensing contributes to maintain whole energy homeostasis. Initial studies by Cota et al. (89) demonstrated that mTORC1 is stimulated robustly and selectively within arcuate neurons expressing anorexigenic proopiomelanocortin (POMC) and orexigenic neuropeptide Y/agouti-related peptide (AgRP) during refeeding after a fast (89), suggesting that this hypothalamic AA sensing signal maintains organisms in a state of metabolic balance. Furthermore, they found that injection of leucine into the hypothalamus can also reduce appetite and weight in an mTORC1-dependent manner (89). Subsequent studies by Blouet and colleagues identified crucial functions of the MBH leucine and mTORC1/S6K1 signaling pathways in the control of energy balance, proposing the importance of leucine sensing abilities by the brainstem (90, 91). In addition, Su et al. (92) revealed that third intracerebroventricular (3ICV) injection of leucine rapidly activated S6K1 within the MBH, and thereby reduced food intake as early as 4 hours after administration. Hence, hypothalamic leucine is a potential nutrient signal to activate mTORC1 that may reduce food intake.

Hypothalamic mTORC1 sensing signals also play an important role in the regulation of thermogenesis. Studies conducted by our laboratory have shown that after 7 days of a leucine-deficient diet, abdominal fat loss and brown adipose tissue heat production increased rapidly in mice (93). Furthermore, we revealed that leucine deprivation stimulates fat loss by increasing expression of corticotrophin-releasing hormone (Crh) in the hypothalamus through the activation of the stimulatory G protein/cyclic adenosine monophosphate response (cAMP)/protein kinase A/cAMP response element-binding (CREB) protein pathway (93). We also indicated that leucine deprivation inhibited the mTORC1/S6K1 signaling pathway in the hypothalamus, subsequently regulating the expression of the melanocortin-4 receptor to mediate the upregulation of Crh (94). This effect is dependent on the hypothalamic mTORC1 signaling pathway, which cannot be replicated by injection of other BCAAs, such as valine; when rapamycin, an inhibitor of the mTORC1 signaling pathway, is injected into the ventricle, this effect disappears (94). Further research revealed a novel role for the thyrotropin-releasing hormone (TRH) in regulating energy expenditure through triiodothyronine during leucine deprivation. Trh expression is activated by CREB, which is phosphorylated by ERK1/2 and dephosphorylated by PPP1R3C-containing protein Ser/Thr phosphatase type 1 (PP1) under leucine deprivation (95) (Fig. 5). Collectively, our studies provide a novel link between hypothalamic AA sensing signaling and stimulation of energy expenditure under leucine deprivation.

Figure 5.

Central hypothalamic signals that sense leucine deprivation and regulate energy metabolism. The mediobasal hypothalamus (MBH), which contains arcuate (ARC) and ventromedial (VMN) nuclei, is a critical node in the neural network where central amino acid sensing contributes to whole body energy homeostasis. Leucine deprivation stimulates fat loss by increasing expression of corticotrophin-releasing hormone (CRH) in the hypothalamus via activation of the stimulatory G protein/cAMP/protein kinase A/cAMP response element-binding protein (CREB) pathway (93). Leucine deprivation inhibited the mTOR/S6K1 signaling pathway in the hypothalamus, subsequently regulating the expression of the melanocortin-4 receptor (MC4R) to mediate the upregulation of Crh (94). Leucine deprivation also increases ERK1/2 activity, while it inhibits dephosphorylation enzyme PP1 activity, and thus increases CREB phosphorylation. Phosphorylated CREB translocates into the cell nucleus in hypothalamic neurons and thereby transcriptionally activates the expression of Crh and Trh. Finally, CRH activates the sympathetic nervous system, and TRH stimulates mitochondrial uncoupling protein 1 (UCP1)–dependent thermogenesis via thymosin T3, thus increasing energy expenditure and fat loss (95). Arrows and bars represent activation and inhibition of the downstream protein, respectively. Dashed arrows represent translocation.

Glucose metabolism

Studies have indicated that MBH leucine is a potent stimulator of mTORC1 and S6K1 phosphorylation that contributes to the regulation of glucose metabolism. In fact, Su et al. (92) indicated that MBH leucine infusion alone has a hypoglycemic effect, and suppresses hepatic glucose production through a decrease in both gluconeogenesis and glycogenolysis.

Pancreatic β and α cells play essential roles in maintaining glucose homeostasis. The pancreatic β cells sense some circulating AAs, such as arginine and leucine, secrete insulin in response to the needs of the organism; this process is important to glucose metabolism (96, 97). The mTORC1 signaling plays an important role in pancreatic β cell function. Studies using mice with β cell-specific deletion of tuberous sclerosis 2 (β-Tsc2 KO) indicated that mTORC1 hyperactivation has a dual role on β cell function, which increased β cell mass and improved glucose tolerance in young β-Tsc2 KO mice but reduced β cell mass and hyperglycemia in older mice (98, 99). Therefore, mTORC1 hyperactivation in the pancreas initially improves glucose metabolism, but also results in a faster attenuation in β cell function and hyperglycemia over time. In contrast, glucagon is secreted from pancreatic α cells and plays an important role in blood glucose homeostasis under the fasting state. Using mice with α cell-specific deletion of Raptor, Bozadjieva et al. (100) indicated that inhibition of mTORC1 signaling impaired α cell development and reduced glucagon secretion during fasting, hypoglycemia, and glucoprivic signals. Thus, this study uncovered a potentially novel role of mTORC1 signaling in the regulation of α cell–glucagon secretion and glucose homeostasis.

Another study by Sengupta et al. (101) indicated that mice with constitutively activated liver mTORC1 signaling by liver-specific deletion of Tsc1 were unable to produce ketone bodies during fasting owing to suppression of peroxisome proliferator-activated receptor α (PPARα) in a mTORC1-dependent manner (101). Moreover, sustained activation of mTORC1 in the liver during the fasting period blocks the induction of autophagy, which is critical for providing free AAs for gluconeogenesis (102). Skeletal muscle is the major site of glycogen synthesis and glucose disposal in response to nutrient uptake and insulin secretion, and deregulation of its glucose metabolism links to T2DM. Studies have indicated that a high level of mTORC1 activation in the muscle of obese rodents fed with a high-fat diet (HFD) results in inhibition of insulin signaling, thereby reducing glucose uptake by the muscle and leading to insulin resistance (IR) (103, 104). Collectively, the hyperactivation of mTORC1 by genetic or dietary manipulation contributes to IR, thereby leading to speculation that the development of mTORC1 inhibitors (such as rapamycin) could improve glucose homeostasis and protect against T2DM.

Lipid metabolism

Numerous studies have revealed important roles for mTORC1 in regulation of adipocyte differentiation and lipogenesis in response to nutrient and insulin (105). Activating mTORC1 promotes adipocyte formation and enhances lipid synthesis both in vitro and in vivo, which is consistent with the phenotype observed in adipocyte-specific Raptor KO (Ad-Rap KO) mice that displayed progressive lipodystrophy and fatty liver disease (106). Furthermore, Ad-Rap KO mice also exhibited resistance to HFD-induced obesity owing to repressed adipogenesis, demonstrating that inhibition of mTORC1 in adipose tissue has both beneficial and adverse effects (107). In addition, the downstream effectors of the mTORC1 signaling pathway, such as S6K1 and 4E-BPs, play key roles in regulation of adipogenesis (108, 109). S6K1 regulates the early commitment of embryonic stem cells into adipogenic progenitors by promoting gene expression of early proadipogenic factors (108), and 4E-BP1 or 4E-BP2 controls adipocyte differentiation through translational regulation of the master proadipogenic factor PPARγ (109).

GCN2 and metabolic control

Energy metabolism

In contrast, the MBH has also been proposed as a primary sensing site of EAA deficiency and regulates food intake (110). Two independent studies indicated that the rapid determination of dietary EAA starvation within the anterior piriform cortex occurs in a GCN2-dependent manner in mice (58, 60). 3ICV administration of L-leucinol, a leucine analogue, increased eIF2α phosphorylation selectively in the MBH and was sufficient to reduce food intake in wildtype (WT) mice, but not in Gcn2 KO mice (111, 112).

Several studies conducted by our laboratory have revealed that ATF4, a downstream component of GCN2 (113), plays a key role in regulating energy metabolism. Using Atf4 KO mice, we found that these mice are lean, with increased thermogenesis and lipolysis (114). Furthermore, we reported that mice with an Atf4 KO specific to POMC neurons (Pomc-Atf4 KO) have lower body weight and fat mass, even though their food intake is unchanged (115). Furthermore, we found that energy expenditure and autophagy in POMC neurons of these mice were increased in an ATG5-dependent manner (115). More recently, our study found that leucine deficiency can induce white adipose tissue browning, which is unlikely to be caused by food intake, but is largely dependent on GCN2/ATF4 signal activation in amygdalar PKC-δ neurons (116). Collectively, these results indicate that the GCN2/ATF4 sensing signaling pathway is involved in the neuronal control of energy metabolism.

Glucose metabolism

A study by Xu et al. (117) showed that mice with either global or specific deletion of Gcn2 in the liver display impaired ability to maintain glucose homeostasis during fasting, whereas they exhibit normal plasma insulin level and sensitivity, suggesting that GCN2 may play a key role in maintaining glucose homeostasis by gluconeogenesis. They further found that the expression of the CCAAT enhancer-binding protein-β (C/EBPβ) in the liver fails to be induced in Gcn2-deficient mice after 24 hours of fasting, and found that liver-specific C/ebpβ-KO mice exhibit a phenotype similar to that of Gcn2-deficient mice with reduced gluconeogenesis under fasting (117). These results demonstrate that GCN2 is important in maintaining gluconeogenesis and glucose homeostasis in the liver, which may be mediated through the regulation of C/EBPβ (117). Studies by our group demonstrated that isoleucine or valine deprivation for 7 days significantly decreased blood glucose levels during fed status in WT and insulin-resistant mouse models, possibly because of the reduced expression of glucose-6-phosphatase, a rate-limiting enzyme in gluconeogenesis (9). Furthermore, we indicated that this effect is possibly mediated by decreased mTORC1/S6K1 and increased AMP-activated protein kinase (AMPK) signaling pathways in a GCN2-dependent manner (118). Moreover, we demonstrated that leucine deprivation improves hepatic insulin sensitivity by sequentially activating GCN2 and suppressing mTORC1/S6K1 signaling (118). Collectively, the liver cellular response to AA deficiency is controlled by GCN2, which then regulates hepatic glucose metabolism and insulin sensitivity.

Lipid metabolism

Armstrong et al. (119) reported that a reduction in AA levels, and thereby increase in uncharged tRNAs, can trigger activation of the GCN2-dependent AA sensing signaling pathway within adipocytes, causing increased rates of germline stem cell loss. Currently, GCN2 sensing signals in the regulation of lipid metabolism remain elusive. Nevertheless, our study found that AgRP neuron-specific Atf4 KO mice were lean and resistant to HFD-induced obesity; this may due to improved insulin and leptin sensitivity and decreased hepatic lipid accumulation (120). However, whether the GCN2/ATF4 sensing signal in the AgRP neuron regulates lipid metabolism in response to AA deprivation remains unclear.

FGF21 is a critical factor for the adaptive metabolic response to starvation, and is also induced under AA deficiency (62, 63). As expected, De Sousa-Coelho et al. (63) found that FGF21 expression was significantly increased in the liver of WT animals after 7 days of leucine deprivation, along with downregulation of lipogenic gene expression. In contrast, Fgf21-deficient mice exhibited enhanced lipogenesis and developed liver steatosis (63). In the case of ketogenic diets, methionine deficiency appears to contribute very specifically to the rise in FGF21, resulting in extremely high levels of circulating FGF21 (121, 122). In WT mice, consumption of a methionine/choline-deficient diet normally causes a fatty liver and eventually leads to a histological phenotype similar to that of nonalcoholic steatohepatitis. This is also associated with significant increases in hepatic expression and serum levels of FGF21 (123, 124). Moreover, hepatic FGF21 production was markedly increased in response to low-protein diets in mice (125), and, notably, Fgf21 KO mice were protected against the metabolic effects of low-protein diets (126). These data clearly indicate that hepatic FGF21 plays an important role in the regulation of lipid metabolism in response to protein or AA starvation.

Other AA sensing signals and metabolic control

Many AA sensing receptors are located on EECs; this sensing of AAs by EECs located in the intestinal epithelium triggers the release of gastrointestinal peptides, such as GLP-1 and cholecystokinin, as important players involved in metabolic regulation, including energy and glucose metabolism (127-130). Diakogiannaki et al. (72) indicated that in primary intestinal epithelial cells isolated from PepT1-deficient mice, the GLP-1 response is impaired when exposed to the specific nonmetabolizable PepT1 substrate glycyl-sarcosine (72). Another study conducted by Dranse et al. (131) indicated that acute administration of protein to the upper small intestine reduces glucose production and promotes glucose tolerance through PepT1- and GLP-1-mediated protein sensing mechanisms.

More recently, Rudenko et al. (80) revealed that Gpr142 KO mice exhibit a very limited metabolic phenotype. However, they were able to demonstrate that tryptophan-induced secretion of gut hormones is dependent on GPR142, such as a synthetic GPR142 agonist stimulated GIP, cholecystokinin, and GLP-1 secretion (80). Another study demonstrated that functional characteristics and cellular location of gut-expressed T1R1/T1R3 support its role as a luminal sensor for phenylalanine-, leucine-, and glutamate-induced cholecystokinin secretion (132). Nakamura et al. (133) found that L-glutamine induced an increase in intracellular cAMP concentration and GLP-1 secretion in a T1R3-dependent manner in the murine enteroendocrine L cell line (GLUTag cells). These results suggest that AA-specific sensors in EECs detect the changes in intestinal contents and cause gut hormone secretion upon activation and may contribute to glucose homeostasis.

Glutamate dehydrogenase (GDH), a glutamine and glutamate sensor, acts as a “‘glucokinase’” that triggers AA-induced secretion. The discovery that glutamate directly stimulates insulin exocytosis in permeabilized β cells prompted studies on the role of glutamate and GDH in fuel-stimulated insulin secretion (134). Furthermore, islet glucose-stimulated insulin secretion levels were reduced in β cell-specific Gdh KO mice (135). In addition, GPR142 is also a highly selective sensor of essential AAAs in pancreatic islets. Using the Gpr142 KO mice, Rudenko et al. (80) demonstrated that tryptophan-induced secretion of pancreatic hormones is mediated through GPR142 and activating GPR142 stimulated insulin and glucagon secretion.

AA Sensing Signals in Diseases and Clinical Relevance

AA and metabolic diseases in humans

In recent years, serum levels of AAs, most consistently of the BCAAs, have been associated with obesity, IR, T2DM, and fatty liver (136-140). The observation of higher levels of circulating BCAAs in obese individuals was first reported by Felig et al. (136) nearly half a century ago, and this association resurfaced later when She et al. (137) demonstrated that tissue-specific changes in BCAA metabolism, in the liver and adipose tissue but not in the muscle, may contribute to the increase in plasma BCAAs in obese individuals. A study conducted by Newgard et al. (138) in 2009 rekindled interest in this field by demonstrating that elevated BCAA levels contribute to a “metabolic signature” predicting IR in obese humans. Subsequently, Wang et al. (139) found that high levels of 5 BCAAs and AAAs (leucine, isoleucine, valine, phenylalanine, and tyrosine) were significantly associated with future progression to diabetes. Consistent with this, through the Young Finn’s Study, Würtz et al. (141) indicated that BCAAs, and to a lesser extent the AAAs phenylalanine and tyrosine, were associated with IR in young adults, whereas the gluconeogenic AAs alanine, glutamine, or glycine, and several other AAs (ie, histidine, arginine, and tryptophan) did not show an association with IR. Correspondingly, Shah et al. (142) found that the plasma BCAAs and AAAs phenylalanine and tyrosine decreased significantly in individuals with moderate weight loss. These findings represent the potential key role of AA metabolism in the early pathogenesis of obesity and diabetes, and suggest that AA profiles could contribute to metabolic disease risk assessment. Moreover, this adverse effect of excess BCAAs is consistent with findings from more recent studies by several groups indicating that restricting dietary protein or BCAA intake improves glucose homeostasis and other metabolic endpoints (93, 118, 143, 144). Mechanistically, another study proposed a possible explanation regarding how higher levels of circulating BCAAs caused IR by promoting lipid accumulation in the muscle through 3-hydroxyisobutyrate, a catabolic intermediate of the valine (145). However, whether there is a causal relation between the BCAA levels and type 2 diabetes risks in human remains unclear.

Lotta et al. (146) undertook large-scale human genetic analyses and indicated that human genetic variants associated with IR and dyslipidemia are strongly associated with increased circulating BCAAs; increases of 1 standard deviation in leucine, isoleucine, or valine in circulation were associated with a clear increase in risk for developing T2DM. In contrast, a study by Mahendran et al. (147) evaluated the causal relationship between fasting plasma BCAA levels and IR, and showed that higher levels of circulating BCAAs do not have a causal effect on IR, whereas increased IR drives higher circulating fasting BCAA levels. This is consistent with a previous study from Xu et al. (148) that reported that adult individuals with impaired fasting glucose and T2DM had elevated levels of BCAAs compared with healthy controls. The explanation for the discrepancies between these results may be the variations in participant characteristics, such as age, gender, ethnic background, degree of obesity, or diet. Collectively, more exact biological mechanisms, including the AA sensing pathways that may mediate the causal link between T2DM and BCAA profiles, need to be elucidated.

Involvement of AA sensing signal as drug targets

Although we are just starting to uncover the intricate relationships between AA sensing and some of the most prevalent chronic metabolic diseases, current findings do raise the possibility that they are linked by a similar putative mechanism (149). For example, the dysregulation of mTORC1 has been implicated in diverse metabolic disorders, including obesity, T2DM, and metabolic syndromes, as well as cancer (150-153). Here, we will discuss recent advances in the involvement of AA sensing signals in some of the most prevalent chronic diseases, including metabolic diseases and cancer, and discuss whether these mechanisms could be effective therapeutic targets for these diseases.

AA sensing signals in metabolic diseases

As mentioned earlier, mTORC1 is a core sensor of AA availability and plays key roles in metabolic homeostasis and disease. Under HFD-induced obesity, the central anorectic action of insulin and leptin is impaired (150). Mechanistically, HFD feeding disrupts the ability of leptin to activate hypothalamic mTORC1 and thereby blocks leptin’s ability to reduce food intake (151). These studies provide a possible hypothesis that inhibition of mTORC1 signaling may play a role in the development of obesity by favoring resistance to anorectic signals and by enhancing hyperphagia after feeding of a HFD. Moreover, several components of the mTORC1 signaling network have been implicated in a diverse array of metabolic disorders (97, 106, 108, 109). In fact, deletion of S6K1, a downstream effector of mTORC1, is sufficient to improve insulin sensitivity and to extend life in mice (154), which suggests that treatment with the mTORC1 inhibitor rapamycin might act in a similar manner. Unexpectedly, various rodents treated with rapamycin exhibited a profound deterioration in their metabolic profile. Prolonged treatment with rapamycin reduced β cell mass and function, causing severe IR and hyperlipidemia, and promoted hepatic gluconeogenesis (155-157), although treatment with rapamycin specifically to inhibit mTOR signaling in the small intestine can lower glucose production (158).

Why are contradictory results induced by rapamycin? Rapamycin is proposed to function by inhibiting mTORC1; however, Lamming et al. (159) found that chronic administration of rapamycin substantially disrupts mTORC2 in vivo. Moreover, inactivation of mTORC2 in the liver is sufficient to reduce hepatic insulin sensitivity and increase hepatic gluconeogenesis, which may be a major contributor to rapamycin-induced glucose intolerance (159). Nevertheless, these findings suggest that the specific inhibitors of mTORC1, including rapamycin, might have many benefits for metabolic diseases, while there is a need to avoid the side effects that currently limit their utility. In future, the effects of mTORC1 inhibitors on metabolic homeostasis, obesity, and T2DM need to be extensively tested in animals and clinically.

Alternatively, it has been demonstrated that GCN2 is necessary to maintain hepatic fatty acid homeostasis in conditions of AA deficiency, proposing a possible strategy that activation of GCN2 may reduce lipid synthesis and prevent obesity or fatty liver. Contrary to this hypothesis, Gcn2 KO mice were less obese than WT mice after HFD feeding, and exhibited significantly weakened HFD-induced hepatic steatosis and IR (160). Moreover, Gcn2 KO led to high expression of lipolysis genes in the liver, while lowering expression of genes related to fatty acid transport, synthesis, and lipogenesis (160). These findings suggest that strategies to inhibit GCN2 activity or expression in the liver may provide a novel approach to attenuate nonalcoholic fatty liver disease (NAFLD) development.

FGF21, a downstream effector of the GCN2/ATF4 signaling pathway, as a hepatokine appears more likely to be used as a drug target to treat metabolic diseases. Apart from the well-known positive functions of FGF21 in obesity and IR, several recent lines of evidence suggest a possible link between FGF21 and NAFLD (161). Unexpectedly, high levels of serum FGF21 are associated with NAFLD and its risk factors, such as involved in chronic inflammation and endoplasmic reticulum stress. In a previous study, a significant elevation in both serum FGF21 and Fgf21 mRNA in hepatic tissues was observed in subjects with NAFLD (162). The elevation of FGF21 in subjects with NAFLD has also been found in other cross-sectional and animal-based studies (163, 164). Furthermore, several studies have described a positive correlation between serum FGF21 and lipid metabolism parameters in patients who developed NAFLD (162, 165, 166), such as triacylglycerols and free fatty acids. Therefore, lipid metabolism risk factors in patients who have developed NAFLD will likely increase FGF21 expression. However, Fgf21 KO mice developed liver steatosis caused by upregulated expression of lipogenic genes (62), suggesting that the benefits of FGF21 in liver metabolism and disease remain controversial. Regardless, FGF21 is a potential drug target for treating metabolic diseases, such as dyslipidemia, NAFLD, and T2DM.

AA sensing signals in cancer

Cancer cells require high levels of AAs, either EAAs or NEAAs, for their growth, suggesting that modification of AA nutrients or their sensing mechanisms may alter the efficacy of cancer therapies. First, it was shown that AA restriction (such as methionine, serine, and glutamine) in humans is likely to inhibit tumor progression (167-169). More specifically, the current developments in antitumor drug targeting on glutamine metabolism of cancer cells mainly engage to induce glutamine depletion through inhibition of membrane glutamine transporter and glutaminase activity (170).

Second, hyperactive mTORC1 is a major factor in inducing diverse human tumors, making it is a desirable target for cancer therapy (171-173). In detail, mTOR mutations are observed in many cancers (167), and some components during mTORC1 signaling, like GATOR1, which negatively regulates AA sensing, act as tumor suppressors (174); also RagC mutations are associated with the development of follicular lymphoma (175). Moreover, as discussed above, Rab1A overexpression is widespread in human malignancies for promoting oncogenic transformation and malignant growth (34). Thus, the development of drugs targeting mTORC1 inhibition is a promising therapy for cancer.

To date, 2 generations of mTORC inhibitors have been tested in clinical trials. The first are rapamycin analogs (also named as rapalogs), including everolimus and temsirolimus, which are FDA-approved drugs for advanced breast and renal carcinomas (176), although the clinical benefit of rapalogs has been limited. The second-generation mTORC inhibitors are ATP-competitive catalytic inhibitors (mTOR kinase inhibitors, TORKi), such as Torin-1 and Torin-2 (177, 178). For example, Torin-2, a potent mTOR kinase inhibitor that is orally bioavailable and has significant selectivity against phosphatidylinositol-3 kinase-like kinase family kinases, has shown potent antitumor effects in preclinical studies (179). However, more clinical trials of these second-generation mTORC kinase inhibitors are required before marketing.

Conclusions and Outlook

Overall, AAs act as both nutrient substance and signaling factors in metabolic regulation; they will impair not only protein synthesis but also whole-body dynamic homeostasis when there is a deficiency of functional AAs. For example, although BCAAs are required and viewed to maintain metabolic health, an increase in the levels of circulating BCAAs is associated with a higher risk of metabolic disorder and future obesity, IR, T2DM, or fatty liver (136-140). In contrast, dietary BCAA restriction improves insulin sensitivity and glucose homeostasis in rodent models of IR or obesity (93, 118, 143, 144). This evidence supports the hypothesis that the efficiency of components of AA sensing networks may have important implications for therapy, and knowledge regarding these components may be crucial for the development of AA supplementation plans to counteract metabolism-related phenotypes. Therefore, research of the role and mechanism of AA sensing has become a current topic in the field of metabolic diseases, and, expectedly, dysregulation of components of AA sensing signals has been found to be connected with many pathological conditions. However, some questions regarding AAs and AA sensing in improving metabolic health need to be further explored.

As we know, cells can promptly detect the availability of different AAs via various sensors. Some AAs, such as leucine and arginine, are directly sensed, whereas others, such as methionine, might be sensed in the form of their metabolites. This brings a challenge in uncovering which metabolite signatures cause the metabolic aberrations in question and what are the consequences of these aberrations. For example, some ketogenic AAs could be catabolized to ketone bodies, including acetone, acetoacetate, and β-hydroxybutyrate (β-HB), which are alternative metabolic substrates used during periods of nutrient deficiency. However, ketone bodies also act as signaling molecules to regulate metabolism. For instance, β-HB binds to GPR109A and inhibits white adipose tissue lipolysis (180), which may serve as a negative feedback mechanism, given that lipolysis produces ketogenic substrates. β-HB also antagonizes GPR41 signaling in the sympathetic nervous system, which lowers sympathetic tone (181) and may be a physiological adaptation to starvation. Therefore, sometimes it is unclear whether metabolites other than AAs are sensed first by the organismal AA sensing signals, and thereby regulate metabolic processes. Although the characterization of cellular AA sensors is still at an early stage, elucidation of the mechanisms underlying how metabolites act as signaling molecules and engage specific sensors to regulate energy, glucose, and lipid metabolism, as well as hormone secretion, will be important in addressing these issues.

As discussed above, human diseases are associated with defects in AA sensing, including cancer. Undoubtedly, mTORC1 is the most potent sensor response to AA abundance; therefore, it would be an effective therapeutic target for metabolic diseases and cancer. However, treatment of rodent models with mTORC1 inhibitors, such as rapamycin, caused an unexpected effect of deteriorated metabolic state, suggesting that intensive studies are required to verify the physiologically crucial sensors and their mechanisms of action in response to mTORC1 activation. Moreover, further research is required to determine whether many of the proposed mechanisms in AA sensing are physiologically relevant.

Finally, a complete understanding of AA sensing mechanisms must clarify the potential crossregulation between different AA sensing pathways. For example, under the condition of leucine or arginine deprivation, activated GCN2 can inhibit mTORC1 activity in an ATF4-dependent or -independent manner (66, 70), and act together to maintain metabolic homeostasis. Moreover, in order to explain how cells sense AA availability to coordinate with cell metabolism and growth, the crosstalk of AA sensing signals from each metabolic organism needs to be more comprehensively explored. Although mTORC1, as a central cell growth controller, should integrate a wide range of extracellular and intracellular signals, signal transduction and integration and cross-talk between different organisms need to be further elucidated. As we know, one of the most successful nutritional interventions against the onset of metabolic diseases, such as obesity and T2DM, is limitation of nutrient intake or caloric restriction (182, 183). However, intermittent fasting or AA restriction is found to be a better intervention program for improving metabolic syndromes (184-186). Hence, understanding precise AA sensing mechanisms should contribute to designing better nutritional interventions or drugs against human chronic metabolic diseases.

Abbreviations:

Abbreviations:
 
• 3ICV

  third intracerebroventricular

 
• 4EBP1

  eukaryotic translation initiation factor 4E binding protein-1

 
• β-HB

  beta-hydroxybutyrate

 
• AAs

  amino acids

 
• AAAs

  aromatic amino acids

 
• AgRP

  agouti-related peptide

 
• AMPK

  AMP-activated protein kinase

 
• ARC

  arcuate nucleus

 
• Arf1

  adenosine diphosphate ribosylation factor-1

 
• ATF4

  activating transcription factor 4

 
• BCAA

  branched-chain amino acid

 
• C/EBPβ

  CCAAT enhancer binding protein-β

 
• cAMP

  cyclic adenosine monophosphate

 
• CaSR

  Ca sensing receptor

 
• CCK

  cholecystokinin

 
• CNS

  central nervous system

 
• CREB

  cAMP response element-binding protein

 
• CRH

  corticotropin-releasing hormone

 
• EAA

  essential amino acid

 
• EECs

  enteroendocrine cells

 
• eiF2α

  eukaryotic initiation factor 2α

 
• FGF21

  fibroblast growth factor 21

 
• GABA

  gamma-aminobutyric acid

 
• GAP

  GTPase activating protein

 
• GCN2

  general control nonderepressible 2

 
• GDH

  glutamate dehydrogenase

 
• GIP

  gastric inhibitory polypeptide

 
• GLP-1

  glucagon-like peptide-1

 
• GPR142

  G-protein–coupled receptor 142

 
• GPRC6A

  G-protein–coupled receptor family C subtype 6A

 
• HFD

  high-fat diet

 
• IECs

  intestinal epithelial cells

 
• IR

  insulin resistance

 
• KO

  knockout

 
• LRS

  leucyl-tRNA synthetase

 
• MBH

  mediobasal hypothalamus

 
• mTORC1

  mechanistic target of rapamycin complex 1

 
• NAFLD

  nonalcoholic fatty liver disease

 
• NEAA

  nonessential amino acid

 
• PepT1

  proton-coupled peptide-transporter 1

 
• PPAR

  peroxisome proliferator-activated receptor

 
• POMC

  proopiomelanocortin

 
• Raptor

  regulatory protein associated with mTOR

 
• Rheb

  small GTPase RAS homologue enriched in brain

 
• S6K1

  ribosomal protein S6 kinase 1

 
• SAM

  S-adenosylmethionine;

 
• SLC

  solute carrier

 
• T2DM

  type 2 diabetes mellitus

 
• TCA

  tricarboxylic acid cycle

 
• TRH

  thyrotropin-releasing hormone

 
• v-ATPase

  vacuolar H+-ATPase

 
• WT

  wildtype

Acknowledgments

Financial Support: This review was supported by grants from the National Key R&D Program of China (2018YFA0800600), the National Natural Science Foundation (91957207, 31830044, 81870592, 81970731, 81970742, 81770852, 81700761, and 81700750), and CAS Interdisciplinary Innovation Team, Novo Nordisk-Chinese Academy of Sciences Research Fund (NNCAS-2008-10). Xiaoming Hu was supported by Youth Innovation Promotion Association CAS.

Additional Information

Disclosure Summary: No potential conflict of interest relevant to this article was reported. The authors declare no competing interests.

Data Availability

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.