mTOR signaling and its roles in normal and abnormal brain development

1Mammalian target of rapamycin (mTOR) controls growth and metabolism by activating anabolic processes and suppressing catabolic processes. Dysregulation of mTOR signaling induces various disorders including cancer, diabetes, obesity, cardiovascular disease, inflammation, and neurodevelopmental and neurodegenerative disorders. Several inhibitors of mTOR have been developed and are now clinically approved. Recently, remarkable progress has been made in understanding the molecular nature of mTOR signaling and its biological significance. Here we overview the signaling network of mTOR cascades with special emphasis on translational control. We also focus on the normal and abnormal growth of neurons regulated by mTOR complex 1 (mTORC1).

2Mammalian target of rapamycin is a 289-kDa serine/threonine protein kinase that is, as indicated by its name, a target molecule of the immunosuppressant rapamycin. Rapamycin is an anti-fungal macrolide compound isolated from soil bacterium, Streptomyces hygroscopicus, in Rapa Nui (Easter Island; ). The 12-kDa FK506-binding protein (FKBP12) was first identified as a rapamycin-binding protein. However, yeast genetic screening revealed that FKBP12 was not a functional TOR because disruption of FRP1 (FKBP12 coding gene in the budding yeast Saccharomyces cerevisiae) did not cause growth inhibition/toxicity. This suggested that FKBP12 was likely a co-factor in executing the rapamycin action. Screening of rapamycin-resistant yeast mutants led to the identification of genes TOR1 and TOR2 as molecular targets of rapamycin-FKBP12 complex and functional targets of rapamycin. Subsequently, four groups identified rapamycin targets in mammalian cells: mTOR, FRAP (FKBP12 and rapamycin associated protein), RAFT1 (rapamycin and FKBP12 target 1), and RAPT1. Although recent database nomenclature defined the abbreviation MTOR as the “mechanistic TOR” in mammals, many researchers still use the name mTOR.

3Whereas yeast expresses two TOR molecules, TOR1 and TOR2, only one TOR homolog, mTOR, exists in mammals. mTOR nucleates two distinct complexes called mTORC 1 and 2 with several interacting proteins; these complexes have different functions. For example, mTORC1 governs cell growth, metabolism and cell cycle, whereas mTORC2 controls cell survival and cytoskeleton organization (see review by ). The composition of mTORC1 and mTORC2 is shown in Figure 1 and detailed below. The main characteristic component of mTORC1 is raptor. Raptor is a scaffold protein that regulates complex assembly in addition to substrate recognition, that is to say, it determines downstream signaling of mTORC1. Likewise, rictor is a crucial binding partner of mTOR to make it function as mTORC2. In fact, as discussed below, mTORC1 and mTORC2 functions are impaired by the knockout/knockdown of raptor or rictor, respectively. The mammalian lethal with sec13 protein 8 (mLST8) is thought to be a requisite for mTORCs activities and mSin1 is thought to be a scaffold for mTORC2. In addition, deptor and PRAS40 are inhibitor molecules for mTORCs. Several other molecules are reported to be involved in these complexes but their roles are still unclear. Rapamycin, in binding with FKBP-12, interacts with mTOR on its FKBP12-rapamycin binding (FRB) domain and inhibits mTORC1 activity. Many, but not all, of the substrates and biological processes controlled under mTORC1 are rapamycin-sensitive. mTORC2 activity, in contrast, has been thought to be rapamycin-insensitive. Recently structural analysis of mTOR gives us an important insight into the mechanism of rapamycin action. The work predicts that FKBP12-rapamycin complex bound FRB domain may come close to mLST8 thus reducing the access of the substrates to active site of mTOR. These structural properties of mTOR may contribute to the preferences of substrates and/or phosphorylation sites by rapamycin inhibition. In case of mTORC2, its components may already be bound proximately to FRB domain thus inhibits FRB. This notion is supported by the two findings. Prolonged treatment with rapamycin affects mTORC2 activity. Rapamycin does not liberate a pre-existing rictor-mTORC but inhibits the interaction of newly synthesized rictor with mTOR. Another finding is that micromolar order of rapamycin inhibits mTORC2 activity independently of FKBP12. High dose of rapamycin may allows free-rapamycin that can bind FRB domain and thus inhibits mTORC2. These findings advance our knowledge about the mechanism of the inhibitory action of rapamycin.

4In contrast to the dearth of knowledge about the signaling pathway of mTORC2, the upstream and downstream pathways of mTORC1 are much better understood. mTORC1 integrates several intracellular and extracellular cues, and transduces divergent downstream events. Growth factors and amino acids are the best-characterized extracellular stimuli that activate mTORC1. In addition, cellular energy status, oxygen/hypoxia and stressors regulate mTORC1 activity. In the central nervous system (CNS), neurotransmitters, neuromodulators, and hormones are reported to activate mTORC1.

5Growth factors bind to and activate receptor tyrosine kinases (RTKs). The RTK to Akt (aka PKB) pathway has been well investigated so far. Then, tuberous sclerosis complex 1 (TSC1; aka hamartin) and TSC2 (aka tuberin) have been identified as links between Akt and mTOR. Akt phosphorylates TSC2 and causes it to dissociate from TSC1. Dissociated TSC2 is captured by 14-3-3 and is prevented to form the complex. Phosphorylated and liberated TSC2 is degraded by ubiquitination, although the role of 14-3-3 in this process is unclear. The TSC1/2 plays a key role in the activation of mTOR. TSC1/2, together with a third component TBC1D7 functions as a GTPase-activating protein (GAP) for small G-protein Ras homology enriched in brain (Rheb;; ). GTP-bound Rheb, the active form, directly binds to mTORC1 and activates its kinase activity. The GAP accelerates GTP hydrolysis, which inactivates Rheb into its GDP-bound form. In steady state, without growth factor stimulation, unphosphorylated TSC2 maintains a heteromeric interaction with TSC1, and this complex is a suppressor of mTRC1. Until now, the guanine nucleotide exchange factor (GEF) that promotes GDP to GTP conversion of Rheb had not been identified. In addition to Akt, mitogen-activated protein kinase (MAPK), and Rsk inhibit TSC2 thus activating mTORC1. As discussed below, the mutations that causes loss of function of TSCs result in mTORC1 over activation and cause brain diseases.

6Tuberous sclerosis complex 2 directly receives many upstream signals by its phosphorylation. The phosphorylation of TSC2 induces both activation and inactivation of mTORC1 depending on the phospho-acceptor amino acid residues. For example, phosphorylation at Thr1462 of TSC2 by Akt and Ser664 by MAPK (aka Erk; ) inhibits TSC1/2 activity thus activating mTORC1, whereas Ser1345 phosphorylation by AMP-activated protein kinase (AMPK) enhances TSC1/2 activity. GSK3β (glycogen synthase kinase) also phosphorylates TSC2 at Ser 1337 and Ser1341 after priming phosphorylation at Ser1345 by AMPK and activates TSC1/2, thus inhibiting mTORC1.

7Insulin is the most-characterized growth factor that activates mTORCs through the PI3K/Akt/TSC/Rheb pathway. Indeed, insulin and insulin-like growth factor 1 (IGF1) enhance mRNA translation in neurons possibly through mTORC1. Among several growth factors tested, brain-derived neurotrophic factor (BDNF), the most prominent neurotrophic factor in the CNS, has been shown to activate mTORC1 signaling and enhance novel protein synthesis in cortical neurons. Interestingly, the effects of BDNF on translation were stronger than that of insulin likely because BDNF activates both Akt and MAPK, whereas insulin activates only Akt in neurons. This makes sense because TSC2 is phosphorylated by both Akt and MAPK. In addition to BDNF and insulin, several growth factors/RTKs and guidance molecules such as Eph, Slit, and Reelin (detailed in the axonal growth section) have been reported to activate mTORC1 signaling in neuronal cells.

8A list of activators of mTORC1 in neurons is shown in Table 1. For example, G-protein coupled receptors (GPCRs) such as μ-opioid, metabotropic glutamate, and cannabinoid receptors activate mTORC1 in neurons. Because GPCRs are known to transduce signals to Akt and/or MAPK, mTORC1 activation triggered by these ligands seems to inhibit TSC2. Other neurotransmitters were reported to possibly activate mTORC1 (meaning that the outputs are sensitive to rapamycin), such as serotonin (5-HT;;, ), may utilize this pathway.

9Neural activity-dependent regulation is another characteristic feature of mTORC1 activation in the CNS. For instance, the paradigms of synaptic plasticity such as long-term potentiation (LTP) and long-term depression (LTD) often induce mTORC1 activation. mTORC1 activation associates with spatial learning and fear conditioning as well. The direct activator molecules driving these plasticity-related or behavioral paradigms are likely the growth factors or neurotransmitters mentioned above. Indeed, the involvement of BDNF in LTP and metabotropic glutamate receptors (mGluRs) in LTD is well known. Interestingly, Rheb, an upstream activator of mTORC1, was originally found as a growth factor- and an activity-dependent transcript in the brain so that it named “enriched in brain”. This mechanism may also participate in mTORC1 activation by neural activity. The increase of blood flow during learning, as revealed by blood oxygenation level-dependent functional magnetic resonance imaging (BOLD fMRI) may supply enough nutrients such as glucose and amino acids. As mentioned below, these nutrients are other stimulants of mTORC1. More detailed discussions on mTORC1 in synaptic plasticity can be found in other reviews in this issue.

10Yeast TOR activity is controlled by nutrients. Similarly, amino acids, especially leucine, are other essential extracellular stimuli that activate mTORC1 in mammalian cells. Amino acids are not only the building blocks of proteins or intermediates of metabolism but are also evolutionarily conserved inter-cellular signaling molecules in both eukaryotes and prokaryotes. For example, glutamate and GABA are important neurotransmitters in the CNS. Recent reports have answered the long-lasting question, why and how leucine activates mTORC1. The analysis of raptor binding partner and siRNA screening of small GTPases revealed the link between Rag family small G-proteins (RagA–D) and mTORC1. RagA or RagB (~98% sequence similarity) forms a heterodimer with RagC or RagD (~87% sequence similarity). The active conformation is a heterodimer of RagA/B·GTP and RagC/D·GDP. When amino acid availability is sufficient, this complex directly binds to raptor and recruits mTORC1 to the lysosome with “Ragulator (consisting with three proteins, MP1/P14/P18),” which is a GEF for RagA/B. This makes mTORC1 close to and bind to Rheb, an activator of mTORC1 on the lysosomal membrane. A surprising molecule now comes into play – leucyl-tRNA synthetase (LeuRS) – both in mammalian cells and in yeast. LeuRS is an aminoacyl-tRNA synthetase that loads leucine to tRNA with high fidelity. LeuRS binds to RagD and acts as a GAP. Thus, RagD stays in a GDP-bound form, which is necessary for mTORC1 activation. This may be the mechanism that cells sense leucine and activates mTORC1. However, there are still unsolved questions, such as why among the 20 aminoacyl-tRNA synthetases only LeuRS works in the mTORC1 system. Because other amino acids also activate mTORC1, additional players must be involved.

11mTORC1 converges nutrients and growth factor signals, and there is cross-talk between them. Reports suggest that amino acid(s) sufficiency is essential for the insulin-induced activation of mTORC1 in several cell lines. Leucine activates mTORC1 in neurons after influx through system L amino acid transporter (LAT), consisting of LAT1 or LAT2 and 4F2hc (CD98; ). Uptake of arginine by cationic amino acid transporters CAT1 and CAT3 has also been reported to activate mTORC1 in neurons. However, contrary to the above-mentioned essential role of amino acids, starvation of amino acids is less effective on BDNF-induced mTORC1 activation in neurons. This discrepancy is unclear; however, cell-type specific amino acid contents or autophagy mechanisms may influence the system responses.

12mTORC1 signaling in neurons seems to be more sensitive to energy status. For instance, glucose starvation increases the ratio of AMP/ATP. Increased AMP activates AMPK. AMPK directly phosphorylates TSC2 and enhances TSC1/2 activity, leading to the inhibition of mTORC1. Another target to inhibit mTORC1 by AMPK is raptor. Raptor is directly phosphorylated by AMPK at Ser722/Ser792 and this leads to the inhibition of mTORC1 activity. In neurons, AMPK activation completely abrogated growth factor-induced mTORC1 activation and protein synthesis, with TSC2 and raptor phosphorylation. AMPK is phosphorylated and activated by LKB1 and calcium/calmodulin-dependent protein kinase kinase (CaMKK; ). Because CaMKK activity is dependent on calcium, both ionotropic and metabotropic receptor ligands (neurotransmitters) may participate in mTORC1 signaling through the Ca2+-CaMKK-AMPK pathway in neurons.

13Many substrates lead to diverse cellular responses downstream of mTOR. p70 ribosomal protein S6 kinase (p70S6K) and eukaryotic initiation factor 4E (eIF4E)-binding protein (4EBP) are the best-characterized substrates for mTORC1 that regulate translation. In contrast, ULK1 and ATG13 phosphorylation suppress autophagy. PRAS40, an inhibitory component of mTORC1, is a substrate of mTORC1 itself. All these mTORC1 substrates contain TOR signaling (TOS) motif. TOS motif is recognized by raptor and utilized for substrates recruitment to mTORC1 for optimal phosphorylation. As mTORC2 substrates, Akt mediates a variety of actions including mTORC1 activation and the inhibition of apoptosis, among others. Protein kinase C (PKC) also mediates a myriad of cellular responses, and serum- and glucocorticoid-induced protein kinase (SGK) affects transcription through FoxO. Recent phosphoproteomic analyses revealed many mTOR substrates that may link to the known cellular responses such as ribosome biogenesis, mitochondrial biogenesis, metabolism, mRNA splicing etc. The flow sheet of mTORCs cascades is summarized in Figure 2. The physiological roles of these events in neural functions are less understood. In contrast, translational control mediated by mTORC1 is well studied in the aspect of neural development, plasticity, and diseases. Thus we review this process in detail below.

14Upon activation, mTOR phosphorylates p70S6K at Thr389 and 4EBPs [there are three 4EBPs (4EBP1–3), and 4EBP2 is a major isoform in neurons] at Thr37/46 and Ser65 directly. These substrates bind to the scaffold protein raptor by a TOS motif and are phosphorylated by the kinase mTOR. Phosphorylation of 4EBPs liberates eIF4E and allows eIF4E to bind to eIF4G and form an eIF4F complex together with eIF4A, an RNA helicase. eIF4E recognizes the 7-methylguanosine 5-triphosphate cap structure of 5′-UTR and poly(A)-binding protein (PABP) binds to the poly(A) tail of mRNAs so that the eIF4F complex makes mRNA circular. Indeed, the circular form of mRNA is thought to stabilize it and facilitate translation. Phosphorylation and activation of p70S6K by mTORC1 induced by insulin facilitates the association of eIF3 (a large molecular complex consisting of 13 subunits) with eIF4G. The process is thought to be important for recruiting the 40S ribosome to the mRNA-eIF4F complex. eIF4B is phosphorylated at Ser422 by Akt directly and by Ser406 and Ser422 in MEK and mTOR/p79S6K-dependent manner. Phosphorylated eIF4B enhances eIF4A helicase activity, suggesting that mTORC2 also participates in translational control. eIF4G phosphorylation at Ser1108, Ser1148, and Ser1192 is reportedly rapamycin-sensitive (Figure 3, Upper panel). In addition, p70S6K phosphorylates eukaryotic elongation factor 2 kinase (eEF2K) and suppresses its activity, This causes the downregulation of eEF2 phosphorylation thus induces its activation (Figure 3, Lower panel).

15In neurons, BDNF, insulin, and IGF1 have been shown to enhance translation by activating initiation processes through mTORC1 signaling. In addition, BDNF enhances translation elongation processes through mTORC1-dependent downregulation of eEF2 phosphorylation. Enhancement of translation by dephosphorylation of eEF2 is also induced by 5-HT. A neuron-specific feature of mTORC1-dependent translational control is its spatial property. For example, the mTORC1 system is localized in dendrites as well as in cell bodies. BDNF and transmitters have been shown to induce “local” activation of mTORC1 and translation (see review in this issue).

16In summary, mTORC1 (and possibly mTORC2) modulates translational processes by the following: mTOR 1) phosphorylates 4EBP and induces cap-dependent translation; 2) phosphorylates p70S6K and allows eIF3 to bind eIF4G; 3) induces eIF4B phosphorylation through p70S6K and Akt; and 4) activates eEF2 through p-70S6K/eEF2K. Therefore, mTOR regulates translation both at initiation and elongation steps.

17A recent report suggested that mTORC1 specifically controls the translation of a certain subset of mRNAs. eIF4E preferentially binds to mRNAs that harbor 5′-terminal oligopyrimidine tract (5′-TOP) or related sequences. These mRNAs encode proteins of translation machinery such as ribosomal proteins and translation factors. It has been postulated that p70S6K/S6 regulates translation of 5′-TOP mRNAs. However, although these events occur in parallel, genetic and biochemical analysis have refuted a direct relation between them. This new insight can explain the direct interaction of mTORC1 and 5′-TOP mRNA translation. Indeed, acute signal(s) that activates mTORC1 enhances the production of proteins necessary for translation machinery and may prepare the succeeding and continuous activation of translation that leads to cellular growth. The authors concluded that 4EBP and eIF4G were the master regulators of mTORC1-dependent translation. However, further research must be conducted in postmitotic neurons, because the elongation process was the rate limiting step for translation in neurons.

18Many reports have suggested that mTORC1-dependent translational control is crucial in the development and plasticity of neurons. However, in the majority of the studies, conclusions are based on pharmacological approaches using inhibitors. Rapamycin, the most commonly used tool for analyzing mTORC1 signals, does not effectively inhibit all mTORC1-dependent translation. An ATP-competitive mTOR inhibitor such as Torin1 both inhibits rapamycin-resistant mTORC1 reactions and mTORC2 activity. Chronic rapamycin also inhibits mTORC2. The effectiveness is somewhat cell-type specific, and the efficacy of rapamycin on mTORC2 in neurons has not yet been verified. In any case, we suggest caution when making conclusions using simple pharmacological approaches. Although it is technically challenging, monitoring the actual translation processes together with analysis of translation machinery as well as mTORC1 signaling is desired in combined with certain biological/physiological responses.

19In addition to protein synthesis, mTORC1 also enhances lipid biosynthesis (see review by ). mTORC1 has been reported to activate SREBP (sterol regulatory element binding protein) 1 and 2, transcription factors responsible for the expression of many lipid metabolic enzymes. SREBPs are cleaved and the processed forms are translocated to the nuclei and promote gene transcription. Transcriptome and metabolome analysis have shown that SREBP-dependent lipid biosynthesis is rapamycin-sensitive. Rapamycin inhibited lipid biosynthesis and the expression of lipid synthetic enzymes including acetyl-CoA carboxylase, fatty acid synthase, and HMG-CoA reductase. Fatty acid oxidation and cholesterol synthesis have also been shown to be rapamycin-sensitive. A recent report showed that mTORC1-mediated activation of SREBP1 and 2 was mediated by p70S6K. Deletion of p70S6K induces the small body and small cell phenotype but does not affect translation; thus, the phenotype may be the result of inhibition of lipid synthesis, rather than protein synthesis. Because lipids are necessary to form plasma and organelle membrane component, as well as energy storage source and intracellular signaling, there is no doubt that lipid synthesis is essential to cell growth. Indeed, SREBP1 knockdown, as well as rapamycin treatment, can reduce cell size. Thus, both protein and lipid synthesis controlled by mTORC1 must be necessary for cell size regulation.

20In the CNS, BDNF, the most potent activator of mTORC1 in neurons, has been shown to enhance cholesterol synthesis, as well as protein synthesis. Thus, mTORC1 must contribute to BDNF-mediated neuronal cell growth and dendritic arborization (e.g.,, see review by ). Cancer cells that exhibit uncontrolled growth factor signaling often show activation of SREBPs and lipogenesis. Enough (or excess) lipids required for membrane synthesis are a critical process of cancer progression including migration and invasion along with membrane expansion. These cellular responses give us insight into the characteristics of neuritis extension and spine formation during development and synaptic plasticity. Thus, BDNF-mediated neuronal growth may be dependent, at least in part, on de novo lipid synthesis.

21mTOR plays a pivotal role in growth, proliferation, and migration of every cell; thus, it is thought to be essential for the development of an organism. Indeed, a knockout (KO) mouse study showed that mTOR was indispensable for normal development and viability. The first genetic evidence that mTOR plays an important role in brain development came from an ethyl-nitroso-urea-induced mouse mutation screening. This mutant, named flat-top, was a loss of function mutant of mTOR by missplicing; it showed a defect in telencephalon formation and died in mid-gestation. The milder phenotype compared to full KO may result from the incomplete loss of mTOR function. Indeed, the mutant still had about 10% p70S6K activity compared to the wild-type mouse.

22Complete ablation of mTORC components in mice resulted in embryonic lethality. In addition to mice lacking mTOR itself, raptor KO mice died very early in development, while mLST8 or rictor-null mice died at embryonic day 10.5. Therefore, a conditional KO (CKO) strategy was applied for analyzing the function of mTORCs in brain. Elimination of mTORC1 function in brain was achieved by crossing raptor-floxed mice and nestin-Cre mice. Nestin is a marker of neural progenitors. Although the raptor-CKO mice died soon after birth, they showed microcephaly via reductions in cell size and cell number. In addition, gliogenesis was affected in parallel with the downregulation of STAT3 phosphorylation at Ser727. This is plausible because STAT3 is pivotal for gliogenesis and is phosphorylated by mTORC1.

23The conditional deletion of rictor in several tissues shows a relatively mild phenotype compared with that of raptor. Mice with brain-specific deletion of rictor by crossing nestin-Cre mice have been shown to survive but with smaller brain size, like raptor-KO mice. Brain weight of this rictor-KO mouse was about 60% (nestin-Cre) of the wild-type mouse brain. Total dendritic lengths and soma sizes of pyramidal neurons of hippocampus and Purkinje neurons of cerebellum were reduced. Purkinje cell-specific deletion of rictor via an L7/Pcp-2-cre mouse showed impairments in synaptic function and morphology of these neurons, which are correlated with the ataxia-like phenotype of this mouse. Another report employed Emx1 promoter to drive Cre. Emx1 expresses only in dorsal neural progenitor cells that generate excitatory neurons and astrocytes in the dorsal cortex. The Emx1-Cre:rictor CKO mice also showed a small brain but to a lesser extent (about 80% of the wild-type mice), because the rictor-deficient cell population was more limited than that in the nesti-Cre:rictor mice. Neurons in cortical layers II–IV of the Emx1-Cre:rictor conditional mice were smaller than that of wild-type mice. These results implicate mTORC2 in the progression of neuronal size and morphology during brain development. To put it simply, because mTORC2 activates Akt, it will cause mTORC1 activation. If mTORC2 is inhibited, mTORC1 activity may be downregulated. However, mTORC1 activity looked normal in this rictor CKO mouse. In addition, in other cell types in other organs, for example muscle cells, mTORC2 inactivation has little effect on cell size compared to the mTORC1 inactivation. The precise mechanism of how mTORC2 regulates cell size is still unresolved.

24Detailed morphological analysis of mTOR on neurite formation comes from a culture study. Specifically, transfection of constitutively active or dominant negative forms of PI3K, Akt, and Ras revealed that PI3K-Akt increased the size of the soma and dendrites. Ras combined with PI3K-Akt enhanced the complexity of dendrites of hippocampal neurons. Similarly, knockdown of phosphatase and tensin homolog (PTEN), the phosphatase for Akt, by siRNA induced arborization of hippocampal dendrites. The dendritic growth induced by these manipulations was antagonized by chronic rapamycin treatment, siRNA-mediated knockdown of mTOR and p70S6K, and overexpression of phosphorylation-defective mutant 4EBP. These results indicate that mTOR, especially mTORC1, play pivotal roles in dendritic growth/maturation. Mice with neuron-specific deletion of PTEN showed macrocephaly and hypertrophy of neurons with enhanced mTORC1 signaling. Enhanced mTORC1 signaling and dendritic arborization by extracellular stimuli like BDNF and Reelin are both inhibited by rapamycin. In addition to mTORC1, mTORC2 is implicated in the dendritic growth of hippocampal neurons. The dendritic arbor was disturbed not only by raptor but also by rictor small hairpin (sh) RNA, as well as by mTOR inhibitor Ku-006379A, which inhibits both mTORC1 and 2.

25Axon guidance during development is regulated by balanced chemotactic cues by attractive and repulsive molecules. Semaphorin-3 and netrin-1 have been shown to induce growth cone collapse and repulsive turning in Xenopus retinal neurons. The response was abrogated by rapamycin, as well as by protein synthesis inhibitors, cycloheximide and anisomycin. Phosphorylation of 4EBP was observed in growth cones in response to semaphorin-3 and netrin-1. Slit2 has been shown to act on growth cone collapse and 4EBP phosphorylation similarly, but at later stage, in a rapamycin-sensitive manner. TSC2 haploinsufficiency in mice (TSC-/+) caused aberrant retinogenicular projection, suggesting the disruption of axon guidance. Since axon guidance of this tract is known to depend on ephrin-Eph signaling, the effect of ephrinA on mTOR pathway was investigated. In contrast to the case with semaphorin-3, the ephrinA-EphA signal that induced growth cone collapse suppressed the MAPK-TSC2-mTORC1 cascade and inhibited novel protein synthesis. TSC deficiency and constitutively active Rheb expression have been reported to counteract the actions of ephrinA-EphA. These results indicate that mTORC1-mediated translational control of growth cones plays pivotal roles in axon guidance during development.

26These studies reveal the universal and specific roles of mTORCs in neurons. mTORC1 activation induces protein and lipid synthesis so that it increases cellular mass with expansion of plasma membrane. Local protein (and possibly lipid) synthesis mediated by mTORC1 may participate in the extension of an axon and dendrites of a neuron. mTORC2 may be the a putative facilitator of growth cone motility, including neurite pathfinding and elongation because it is known to affect actin dynamics. Although the precise mechanisms are still not fully understood, regulated and coordinated activities of mTORC1 and 2 must be necessary for normal development of neurons and a brain.

27mTOR has been implicated as a cause of various diseases. Among them, most famous and directly related to mTOR disease is TSC. As mentioned previously, this disease is caused by the genetic mutation (including nonsense, missense, insertion, and deletion) of TSC1 or TSC2, the genes encoding TSC1 (hamartin) or TSC2 (tuberin), respectively. TSC was first described in late 19th century and the mutations have been identified in TSC1 (9q34) and TSC2 (16p13.3) in the 1990s. In 2002, several groups reported that TSC1/2 complex is a suppressor of mTOR. TSC1/2 complex is a GAP for Rheb so it inhibits mTORC1.

28Tuberous sclerosis complex is an autosomal dominant, multisystem disorder that affects brain, lung, heart, skin, and kidney. The neurological symptoms are intractable epilepsy, autism, and mental retardation. Pathological features are cortical tubers, subependymal nodules (SENs), and subependymal giant cell tumors (SGCTs) and subependymal giant cell astrocytomas (SEGAs). The characteristic features are abnormal cell proliferation and growth (see review; ). Because TSC is caused by a loss-of-function of TSC1 or TSC2, mTORC1 must be constitutively active in these cells. Indeed, highly phosphorylated p70S6K (Thr389), S6 (Ser235/236), and 4EBP (Thr37/46) have been observed immunohistochemically in pathologically abnormal, enlarged cells that may be of progenitor cell origin. Although the direct interaction of TSC1/2 and mTORC2 is still unclear, phosphorylation of mTORC2 substrates PKCa at Ser657, Akt at Ser473, and SGK1 at Ser422 have also been observed in the fetal TSC brain. The relationship between the activation of mTORC1 and cell (over)growth observed in human disease corroborates the experimental findings. In fact, the rapamycin analogue (called “Rapalog”) everolimus (RAD-001) is clinically approved for the medication of TSC (http://www.fda.gov/Drugs/InformationOnDrugs/ApprovedDrugs/ucm317490.html).

29There have been many reports about the KO of TSC1 or TSC2 genes. Homozygotes of these gene deficiency are lethal, and heterozygotes do not have tubers, although the mice show some abnormalities. Several CKO mice carrying biallelic deletion in certain cells have been generated, and they mimic some symptoms and pathology of human TSC. This may be because biallelic mutation is necessary to induce severe morphological alteration such as tubers or giant astrocytomas. In fact, so-called “two-hit” mechanisms of germline mutation and somatic mutation have been proposed for human tubers of TSC patients, although another report argued that the second hit was rare. The two-hit mechanism has been evaluated with homozygous deletion of TSC2 only in radial glia, and heterozygous deletion in all other cells by crossing Tsc2+/-, Tsc2flox, and GFAP-Cre mice. The mice showed cytomegaly, defects in lamination, astrogliosis, and hypomyelination. mTORC1 activation was confirmed by S6 phosphorylation. Another approach of double hit of TSC1 gene by in utero electroporation has been reported. The mice were made by crossing TSC1+/- mice and TSC1flox mice. TSC1 in the cells of certain brain regions were biallelically deleted via in utero electroporation of Cre plasmid. The offspring showed tuber-like lesions and cortical hyperexcitability. Other conditional homozygotic deletion of either TSC1 or 2 represented some of the features of TSC. All the animal models indicate that biallelic deletion (or dysfunction) is necessary to induce TSC pathology. We have learned a lot about mTOR biology from TSC including the signaling mechanism of mTORCs; however, there are still many unsolved questions like the above-mentioned two-hit hypothesis. Further comprehensive genetic analysis (not limited in TSC genes) in the TSC brain may address the pathological obscurity of genotype-phenotype interaction.

30There are two diseases with neurological symptoms and brain pathology similar to TSC – hemimegalencephaly and focal cortical dysplasia. Genetic mutations of these diseases have not been identified until now. Recently, somatic mutations of Akt, PI3K, and mTOR were reported in the hemimegalencephaly brain. Two of the cases are a trisomy of 1q that contains Akt3, and one is an activating mutation of Akt3 (G49A; ). This mutation encompasses the amino acid substitution E17K in a coding region, which converts Akt to a constitutively active form. Another group found same somatic mutation of Akt3 in one case, and constitutively active PI3K mutations (G1633A, E545K) in four cases. In addition, an mTOR somatic mutation at C4448T that causes C1483Y substitution was observed in hemimegalencephaly brain. In contrast to some mTOR active mutations developed in laboratories in mammalian cell, whether C1483Y mutation makes mTOR active or not is unknown. This is the first report of mTOR mutation in neurological disease so that the validation of activity is awaited. Dysmorphic cells in the brain of focal cortical dysplasia type IIB patient showed hyper-phosphorylation of S6. Although genetic analysis forthcoming, it is likely that there may be somatic mutations of unknown genes in mTORC1 signaling pathways that are responsible for cellular abnormality in focal cortical dysplasia.

31In this review, we summarize the updated molecular interaction of mTOR signaling. We focus on the growth and size regulation of neurons during development as a biological output of mTOR signaling. Diseases that are picked are limited to this aspect. TSC is caused by the mutation of TSC1 or 2 that are regulators of mTORC1. TSC and other pathologically related disease, hemimegalencephaly and forcal cortical dysplasia are characteristics of enlarged and dysmorphic neurons and glia. These are diseases of dysregulated cell growth.

32Although mTOR governs various processes, we describes much on the translational control and a bit about lipid biogenesis, because these two processes seem to be related with cell growth regulation directly. In addition, mTOR-mediated translational control is most (or almost only) studied in the field of neuroscience as a mechanism behind the neural plasticity, learning and memory.

33Pathology of TSC has given us a hit of link between mTORC1 activity and size control in the brain. Overactivation of mTORC1 (and potentially mTORC2) induces enlargement and dysmorphism of neurons and glia (or their progenitors). On the other hand, it has been well known that fetal and neonatal malnutrition causes reduced brain size. Insufficient intake of proteins and carbohydrates causes reduced amino acid and glucose. Because these nutrients are essential factors to induce mTORC1 function, it is likely that mTORC1 inhibition is related to dysgenesis of a brain. Beside these unusual situations, size differences are found in neurons. One characteristic feature of neurons is its high heterogeneity in function and morphology. Soma sizes vary from about 10 (cerebellar granule neurons) to 100 (Betz cell) micrometers in diameter. The cell size is thought to be determined genetically (and/or epigenetically). The concept includes timing and levels of expression of nutrients transporters and growth factor receptors that may affect mTORC1 activity. Although mTOR signaling components are ubiquitously expressed, rheb, for example, is an activity-inducible molecule in neurons. It suggests that neural activity may enhance mTORC1 through the upregulation of rheb level.

34mTOR is undoubtedly a master regulator of cell growth from yeast to human. However, how mTOR controls cell size is still not clear enough. Enhanced protein and lipid synthesis, and inhibition of protein degradation controlled by mTORC1 surely increase cellular mass that includes cell volume, axon elongation, and dendrite arborization. Conceptual scheme is shown in Figure 4. It must be clarify whether novel protein and/or lipid synthesis are necessary and sufficient for cell growth in the brain. Further studies will reveal the correspondence to mTOR downstream signaling pathway to phenotype (cell growth). It will shed light on the biology of cell size and may also contribute to the drug discovery for the TSC and related diseases more selective than rapamycin and rapalogs.

35The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.