Introduction

The mammalian Target of Rapamycin (mTOR) pathway is a central converging point of a diverse range of cellular growth signals, including hormones/growth factors, nutrients, energy status, and mechanical strain. The mTOR protein associates with other proteins to form mTOR complexes 1 and 2 (mTORC1 and mTORC2). mTORC1 is characterized by its association with Raptor (regulatory associated protein of mTOR), whilst mTORC2 is associated with Rictor (rapamycin insensitive companion of mTOR) [44]. Raptor and Rictor are both binding and scaffold proteins that direct mTOR activity towards specific substrates [49, 60, 80], therefore, providing specificity in determining cellular functionality. In the case of mTORC1, mTOR specificity is directed to p70 S6 Kinase (p70S6K) and 4E-BP1 [16, 37], which makes mTORC1 a central node regulating protein synthesis, cell size, and proliferation [36, 37, 98] that is highly responsive to nutrients and energy stress [7, 60]. On the other hand, mTORC2 directly phosphorylates AKT [81] regulating aspects of cell survival and the organization of actin cytoskeleton [80, 95].

4E-BP1 and p70S6K possess the conserved TOS (TOR signaling) motif that is essential for activation by mTOR [82, 83]; although mTOR exhibits differing kinase activity towards these targets [16, 20, 21]. Whilst mTOR-dependent phosphorylation of downstream effectors is a well-characterized signaling event [16, 82], there is frequent uncertainty regarding the functional significance of phosphorylation of the mTOR protein itself. The mTOR protein has multiple residues that can undergo phosphorylation. Phospho-specific antibodies have been developed and are frequently used in experimental and clinical studies. Antibodies that recognize the Ser2481 and Ser2448 phosphorylation residues of mTOR are widely reported in the skeletal muscle biology literature, from basic cell culture experiments to applied exercise/nutrition physiology studies in humans. Indeed, it has been fifteen years, since the first studies reported measurements of Ser2448 phosphorylation in muscle tissue [10, 76]. Since then, the biomedical and muscle research field has experienced a substantial growth of scientific papers reporting measurement of Ser2448 phosphorylation (Fig. 1) purportedly to assess mTOR kinase activation or as a proxy measure of induction of the mTOR pathway. However, the relevance of this phosphosite to the kinase activity of mTOR is frequently not considered. Thus, the role of the Ser2448 phosphorylation site in control of mTOR activity will be discussed herein and further how new insights in muscle physiology can be gained from revaluation of past studies utilizing this phosphorylation site.

Fig. 1
figure 1

Number of publications reporting Ser2448 per year. A timeline of key publications to understand the role of the Ser2448 residue of the mTOR protein is shown. Note that the first appearance of Ser2448 is in 1998; however, Bunn et al. studied the region, where mTOR Ser2448 is located, which is a key study for the understanding of the role of phosphorylation at this site

Full size image

Where Ser2448 is located within mTOR?

The (m)TOR protein is a highly evolutionarily conserved serine/threonine kinase protein from yeast to humans [26, 35, 46]. Multiple phosphorylation sites are present on the mTOR protein structure including Serine 1261, Serine 1415, Serine 2159, Threonine 2164, Threonine 2446, Serine 2448, and Serine 2481 (Fig. 2) [32, 99]. The Ser1261 residue lies within mTOR HEAT-repeat motif, is phosphorylatable on both mTORC1/mTORC2, and its phosphorylation appears to increase mTORC1 kinase activity [1]. The Ser1415 is a recent discovered phosphorylatable site by IKKα directly following activation by AKT that increases mTORC1 activity [25]. The Ser2159 and Thr2164 residues lie within the kinase domain and phosphorylation of these residues promotes activation of mTORC1 [32]. The others remaining phosphorylatable mTOR residues (Thr2446, Ser2448, and Ser2481) are found in the FIT domain (Found In TOR) spanning TOR residues 2427 to 2516 [88, 99], between the kinase domain and the FATC domain [51, 99]. Initially, it was thought that Ser2481 and Ser2448 were phosphorylation residues specific for mTORC2 and mTORC1 respectively [23]. However, Ser2481 has already been found to be phosphorylated in mTORC1 [87] and Ser2448 has also been found to be phosphorylated in mTORC2, despite a lack of an apparent correlation with its kinase activity [78]. Phosphorylation of Ser2481 residue has been shown to be an autophosphorylation residue as a proxy of mTOR’s own activity [87].

Fig. 2
figure 2

mTOR protein domains and localization of phosphorylatable residues. Phosphorylatable residues and key domains within the mTOR protein are highlighted. FRB FKBP12-rapamycin-binding domain, KD kinase domain, NRD negative regulatory domain, FIT found in TOR

Full size image

Although TOR proteins contain several domains that are conserved throughout eukaryotic evolution, the small FIT domain (where Ser2448 dwells) is a relatively recent acquisition in vertebrate evolution [88, 99] showing some divergences in taxonomic groups. The mTOR homologous proteins in yeast (TOR) and drosophila (dTOR) lack the residue Ser2448 [23, 98]. Within the FIT domain lies the Negative Regulatory Domain (NRD) which is defined as TOR residues 2430 to 2450 [50, 85]. Due to its recent acquisition during vertebrate evolution, the Ser2448 residue and its contiguous area are potentially a fine-tuning mechanism for control of mTOR’s own activity.

Mechanisms of mTOR Ser2448 phosphorylation

It was originally proposed that Ser2448 was an AKT consensus phosphorylation target [51, 71, 85]. However, in vitro phosphorylation analysis failed to demonstrate direct phosphorylation by AKT, thus an undescribed protein kinase theoretically lying between AKT and mTOR was hypothesized to exist, which would be called TOR Kinase, or TORK [84], which has never been shown to exist. Instead, AKT is capable of activating mTORC1 by inhibition of its negative regulators. AKT phosphorylates and inactivates TSC1/2 that regulates Rheb activity (by stimulating GTP hydrolysis), which in turn binds to mTORC1 [57, 64]. In addition, AKT can phosphorylate PRAS40 [79, 91, 95]. Recently, it was also shown that AKT could additionally modulate mTOR activity, indirectly, via IKKα [25].

Evidence to further support the lack of direct signaling between AKT and Ser2448 mTOR is found in the skeletal muscle biology literature. The anabolic pathways leading to mTOR activation in response to mechanical stimulation were originally proposed to use the upstream pathway as growth factors (PI3K-AKT-mTOR) [9]. However, this has now convincingly been shown to be an AKT independent event [40, 47, 68, 75]. Therefore, if mechanical stimulation increases the phosphorylation of mTOR at Ser2448 [76] and p70S6K at Thr389 by a PI3K/AKT independent mechanism [33, 40, 54, 55, 65, 72, 73], then it is unlikely that phosphorylation of the mTOR Ser2448 site is a direct target of AKT. Indeed, Atherthon et al. [5], using stretching as a model for mechanical stimulation of myotubes, were able to demonstrate that phosphorylation of the mTOR downstream targets, p70S6K1 and 4EBP1, actually precedes phosphorylation of mTOR at Ser2448.

Direct phosphorylation of mTOR by AKT was proven not to be the case by two elegant studies, published in 2005, which demonstrated that Ser2448 is actually phosphorylated by the mTOR downstream effector p70S6K [19, 53], as part of a feedback loop mechanism. Subsequently, the synthesis of the first specific p70S6K inhibitor (PF-4708671) enabled further confirmation that Ser2448 is indeed part of mTOR-p70S6k feedback loop as this inhibitor simultaneously reduces the phosphorylation of the p70S6K downstream effector rpS6 and mTOR Ser2448 [74]. More recently, von Walden et al. [92] has shown that this p70S6K inhibitor increases the phosphorylation of Thr389 on p70S6K, which demonstrate that inhibition of p70S6K activity increased mTOR activity in muscle cells, presumably by removing this negative feedback loop. The discovery that mTOR is phosphorylated at Ser2448 by p70S6K explains how mTOR has been previously observed to be phosphorylated at Ser2448 following stimulation with anabolic stimuli, including insulin, amino acids, and mechanical strain. The following question then is: does phosphorylation of Ser2448 by p70S6K increases or repress mTOR activity?

Role of NRD and Ser2448 phosphorylation in the regulation of mTOR kinase activation

Phosphorylation of mTOR Ser2448 is a common marker used in the cancer [43] and skeletal muscle literature [77] as an index of mTOR kinase activity and/or pathway signaling. However, analysis of mTOR pathway activation solely by analysis of Ser2448 phosphorylation status in cancer studies has already been withdrawn by some researchers due to its unknown role [67]. Whilst heightened Ser2448 phosphorylation may be associated with mTORC1 activation under certain conditions, it does not imply a cause and effect relationship [14, 19, 53, 78]. In fact, there is no evidence that phosphorylation of Ser2448 increases mTOR activity. Rather, there is evidence from cell culture experiments utilizing mTOR mutant protein versions to suggest that the NRD domain has a negative effect on mTOR activity. Deletion of this region (amino acids 2433–2450) increases mTOR activity [31, 66, 85], although recently, others have found that deletion of a slightly extended region (2443–2486) has the opposite effect to lower mTOR activity [97], and substitution of Ser2448 for the non-phosphorylatable alanine or the phosphomimetic glutamate had no clear effect on mTOR kinase activity [19, 85].

The residue Thr2446, such as Ser2448, is another phosphorylatable site within the NRD. Initially, it was hypothesized that Thr2446 was a target of AKT along with Ser2448, that would promote mTOR activity [85]. However, Thr2446 was subsequently been shown to be a major target of AMPK, with its phosphorylation suppressing mTOR activity [18]. These two close residues Thr2446 and Ser2448 (separated by only one amino acid) are unlikely to have divergent effects on mTOR function. At the time that Thr2446 was shown to be a target of AMPK (in 2004), the main concept was that Ser2448 was a downstream target of AKT. Therefore, it was hypothesized that these sites would have opposing effects on mTOR activity, since insulin treatment increased mTOR Ser2448 but not Thr2446, and starvation increased Thr2446 but not Ser2448. However, the discovery that Ser2448 is a p70S6K-dependent site explains how both residues function as negative regulators of mTOR activity. The deletion of the NRD increases mTOR activity not because it exerts an inhibitory effect on mTOR kinase activity as previously hypothesized, but rather, because mTOR becomes refractory to energy and nutritional cues. This explains why the mutant TOR protein-lacking NRD region [31] is protected from growth factor deprivation and why prevention of phosphorylation at Thr2446 and Ser2448 via treatment with an antibody raised against the region 2433–2450 increases mTOR kinase activity [13, 66]. Not surprisingly then, the loss of Ser2448 phosphorylation has been linked to cancer progression when currently accepted knowledge in the role of Ser2448 would expect otherwise [70]. This indicates that failure of p70S6K to feedback to mTOR to fine-tune its activity may lead to a constitutively active mTOR resistant to p70S6K feedback signals.

Since mTOR Ser2448 is a residue target of p70S6K, to the best of our knowledge, no studies so far have shown that p70S6K phosphorylation at Thr389 is further increased after the initial phosphorylation activity of mTORC1 towards p70S6K, which would indicate a positive feedback. Recently, however, it has been shown that a short isoform of p70S6K can also interact with mTOR protein and increase mTOR kinase activity [6]. Although this short isoform of p70S6K can bind to mTOR, it cannot phosphorylate mTOR Ser2448 as it lacks a kinase activity. The mechanism proposed for the enhanced activation of mTOR by the short isoform of p70S6K was actually that this short isoform competes with p70S6K for the mTOR Ser2448 residue, thereby preventing the phosphorylation of Ser2448 and mitigating the negative feedback loop. The evidence presented points to a role of a negative feedback loop initiated by the activation of p70S6K. p70S6K has been previously shown to be the effector of an additional negative feedback loop by phosphorylating and inhibiting IRS-1 [90, 98]. Hence, p70S6K regulates mTOR pathway at multiple levels.

Alternative markers of mTOR activation

Contrary to the domain, where Ser2448 dwells, (m)TOR-associated proteins known to regulate mTOR kinase activity are conserved in yeast TOR and dTOR, including TSC/Rheb, Raptor, and GβL amongst others [3, 26, 35]. PRAS40, besides not being conserved in some yeast, has a corresponding homology in drosophila [79]. Furthermore, the TOR complexes 1 and 2 seem also to be highly conserved in eukaryotic evolution throughout yeast to man [26, 46]. In addition, two independent research groups have identified the leucine sensor within cell, the enzyme Leucyl tRNA synthethase (LRS) [12, 48]. This enzyme activates mTOR through the Rag GTPase. This is another conserved mechanism both found in TOR [12] and mTOR pathways [48]. These data suggest that mTOR regulation and function are largely dependent on the proteins that mTOR docks or associate with, rather than the phosphorylation of mTOR protein itself. Nevertheless, new phosphorylatable residues on mTOR have also been discovered and their roles in regulation of mTOR kinase activity investigated [1, 25, 32]. The residue Ser1261 phosphorylation promotes mTORC1 activity. It is not found in yeast and plants, but contrary to Ser2448, is found in drosophila (d)TOR as a phosphorylatable Threonine, suggesting that this might be a more conserved mechanism regarding mTOR activity than Ser2448 [1, 28]. The residue Ser1415 also appears to be highly conserved [25]. Two other residues within the kinase domain, Ser2159 and Thr2164, have also been discovered recently with positive roles in the mTOR kinase activity. Ser2159 is only found in vertebrates, but Thr2164 is conserved in flies, plants, and yeasts [32]. Thus, these residues could be better markers than Ser2448 in the future, although the kinase responsible for Ser1261, Ser2159, and Thr2164 phosphorylation are not yet known [32, 99].

Therefore, to better understand mTOR protein activity, researchers should aim for mechanisms that are more conserved throughout species that potentially represent a main mechanism of its regulation. In practical terms, it is recommended that researchers willing to determine a meaningful snap-shot of mTORC1 activation using western blotting should focus on classical mTORC1 downstream targets: p70S6K and 4E-BP1. In addition, partner proteins may also add valuable data as PRAS40 phosphorylation on AKT-dependent site and Rag proteins. Moreover, whenever possible, immunoprecipitation of mTORC1 and mTORC2 protein partners, Raptor, and Rictor, respectively, may help determine on which mTOR, the phosphorylation of Ser2448, is complexed with.

Reinterpretation of exercise and muscle studies in light of a mTOR-p70S6K-mTOR negative feedback loop

Data supporting the role of mTOR in skeletal muscle growth appears overwhelming. To date, studies have focused on the mTORC1 downstream effectors, p70S6K and 4E-BP1, by measuring the mTOR-dependent phosphorylation sites at Thr389 on p70S6K, and Thr37/46, Ser65, and Thr70 on 4E-BP1 [82, 83]. Data on Thr389 p70S6K in particular seem quite convincing that mTORC1 activity is increased following exercise and that nutrition may play a role enhancing that response. However, the same is not true for mTOR’s own phosphorylation state. Many studies fail to demonstrate a consistent and reproducible Ser2448 phosphorylation response to a plethora of stimuli, whilst the mTORC1-dependent site Thr389 on p70S6K does. Titration of insulin [42] and IGF-1 [61] results in a clear dose-dependent phosphorylation curve for p70S6K at Thr389, whereas mTOR at Ser2448 phosphorylation response is disconnected in both cases. In rat skeletal muscle, phosphorylation of p70S6K at Thr389 was found to increase during feeding and decrease during fasting, whilst a corresponding effect for mTOR Ser2448 was not observed [86]. Furthermore, numerous past works measuring both mTOR Ser2448 and p70S6K Thr389 phosphorylation status have demonstrated that muscle load and/or nutritional interventions have a far greater and more obvious stimulatory effect on p70S6K Thr389 than on mTOR Ser2448 [2, 4, 8, 11, 17, 24, 27, 29, 30, 33, 34, 38, 39, 41, 45, 55, 59, 65, 69, 89, 94]. Since the discovery of the identity of the Ser2448 kinase by Holz and Blenis, and Chiang and Abraham [19, 53] in 2005, it has been established that p70S6K, and not AKT, regulates phosphorylation of the mTOR Ser2448 site as part of a feedback loop mechanism. Although some lines of evidence had already pointed to this regulation having a suppressive effect, it was only recently that the work from Ben-Hur [6] made it clear that a negative feedback loop exists between p70S6K and mTOR Ser2448.

Studies in the muscle literature in particular can have different interpretations in light of the real significance of Ser2448 phosphorylation. This is particularly true for studies, where mTOR Ser2448 is analysed in isolation or when the phosphorylation state of Ser2448 appears in opposition to other downstream targets of the mTOR pathway. For example, Leger et al. [63] showed that 4E-BP1 and p70S6K phosphorylation were unaltered by 8 weeks of resistance exercise training. At the same time, phosphorylation of mTOR Ser2448 was found to be significantly increased. After 8 weeks of detraining, accompanied by a reduction in muscle mass, the phosphorylation at Ser2448 remained elevated, whilst p70S6K and 4E-BP1 phosphorylation were once again unchanged. The authors concluded that activation of the mTOR pathway was increased throughout training based solely on mTOR Ser2448. This was despite the fact that heightened phosphorylation persisting throughout detraining and despite the lack of an effect of training/detraining on p70S6K and 4E-BP1. In addition, other researchers, expecting to observe phosphorylation of mTOR Ser2448 preceding Thr389 on p70S6K and having found the opposite, have suggested that phosphorylation at Thr389 could be a mTOR-independent event following resistance exercise [58]. With the knowledge that p70S6K and not AKT is the kinase of the mTOR Ser2448 site, such results may need to be reinterpreted.

Understanding the role of Ser2448 being a negative regulator of mTOR activity, subject to p70S6K activity in a negative feedback manner, aids in providing insight into the dynamic signaling events in muscle following either loading and/or protein ingestion. Phosphorylation of p70S6K Thr389 has been shown to peak at 15 [65] and 30 min post exercise in humans [52] and in a resistance training model in rodents [11]. Others have also found robust phosphorylation of p70S6K Thr389 at the same early time points with no apparent response on mTOR Ser2448 [56]. Phosphorylation at Thr389 appears to peak no later than the first hour post exercise and decreases in the following hours [15, 62]. Phosphorylation of mTOR Ser2448 should increase afterwards [5] and may thus contribute to the decrease in Thr389 of p70S6K. Thus, in studies that have taken biopsy timepoints exceeding 1 h or upwards, it is tempting to speculate that the feedback loop mechanism (p70S6K-mTOR-p70S6K) may have already taken place. Consistent with this notion, Churchward-Venne et al. [22] have shown recently that Thr389 was not increased compared to baseline at 1.5 h post resistance exercise when at the same timepoint, the other mTORC1 downstream 4E-BP1 was highly phosphorylated, as well as the p70S6K downstream effector S6rp at Ser240/244 also, demonstrating that mTOR/p70S6K was activated before this timepoint. Convincingly, mTOR Ser2448 was also highly phosphorylated; indicating that the decrease or return to baseline levels of Thr389 could have been due to the negative feedback mechanism. These results all are consistent with Ser2448 being a negative feedback mechanism dependent upon of p70S6K activation.

Concluding remarks

The Ser2448 residue of mTOR is part of a negative feedback loop mechanism in which activated p70S6K phosphorylates back onto mTOR to fine-tune its activity (Fig. 3). Whilst an increase in phosphorylation of this residue could potentially reflect an increase in the overall pathway, it does not apply for every cell and condition, as in fact it is rather a read-out of p70S6K activity and thus a late event in the pathway. The use of Ser2448 phosphorylation as the only marker of mTOR kinase activity or pathway activation should be discouraged. New phosphorylation residues have been discovered, which may have greater relevance towards mTOR activity. But as of yet their regulation is incompletely understood and their function should be better known to be widely used. mTOR seems to be mainly regulated as a complex, i.e., in conjunction with its partners (raptor, rictor, PRAS40, DEPTOR, Ragulator, Rheb, etc), rather than by its own phosphorylation. Moreover, because Ser2448 may also be phosphorylated on mTORC2, the significance of measuring global changes in Ser2448 phosphorylation is complicated further. On this basis, efforts to understand mTOR activity and/or activation in skeletal muscle should also focus on the mTOR partners and downstream targets.

Fig. 3
figure 3

p70S6K-mTORC1-p70S6k feedback loop mechanism. Cues that can modulate mTORC1 in the context of muscle cells are depicted. Red cross emphasizes that Ser2448 is not phosphorylated by Akt as previously thought, but by mTORC1’s downstream effector p70S6K. AKT although modulates mTOR activity, it does indirectly via TSC 1/2, and not directly via Ser2448 as previously though. Muscle contractions/exercise promote phosphorylation of mTOR targets (p70S6K and 4E-BP1), but also mTOR Ser2448. Since contractions can promote mTRO activity independently of AKT, Ser2448 cannot be a target of AKT, but as demonstrated by two independent studies (Chiang and Abraham, and Holz and Blenis) by p70S6K

Full size image

In conclusion, it is necessary to state that mTOR Ser2448 phosphorylation: (1) is a target of p70S6K and not AKT; (2) is not a cause of p70S6K activity but rather its consequence; (3) has a negative regulatory effect on mTOR kinase activity, and (4) it does not necessarily mirror an increase in mTOR kinase activity nor should be taken as a measure of mTOR activation. This last item is the most important for researches using Ser2448 phosphorylation, since an erroneous assumption could generate misleading interpretations if other more proper markers are not used in parallel. Based on the present analysis, the interpretation from many studies in the muscle research literature assessing phosphorylation levels of mTOR at Ser2448 may need to be revaluated.