Cell Reviews: Details on phosphorylation of mTOR substrates

Rapamycin, a secondary metabolite produced by bacteria, was initially identified as an antifungal agent and was later found to have immunosuppressive and anticancer activities. 

This property has aroused great interest among scientists, opening the search for the target of Rapamycin. 

The Target of rapamycin ( TOR ) is a highly conserved serine/threonine protein kinase that is best described in yeast, followed by mammalian cells. 

The second milestone in the field was the realization in the late 1990s that TOR plays a central role in regulating cell growth and metabolism.

 The third milestone was the discovery in the early 2000s that TOR exists in two structurally and functionally distinct complexes (TORC1 and TORC2) , which are well conserved from yeast to humans. 

Research over the years has shown that mTOR (mammalian TOR) dysregulation is associated with a variety of major diseases, such as diabetes, cancer, and neurological diseases. 

Given the importance of mTOR in basic biology and medicine, its research has become a large and complex area of ​​research.

On May 12, 2022, Michael N. Hall ‘s team (Lasker Prize winner) from the University of Basel in Switzerland published a review article entitled mTOR substrate phosphorylation in growth control in the journal Cell . 

Focusing on mammalian TOR (mTOR), the authors comprehensively reviewed all reported mTOR direct phosphorylation substrates, and based on recent structural information, we explored how mTORC1 and mTORC2 share the same catalytic subunit. How to select and phosphorylate different substrates.

1. mTORC1 and mTORC2

mTOR is a member of atypical protein kinases of the phosphatidylinositol (PI) kinase-related kinase (PIKK) family. 

Despite sharing a common ancestor with PI kinases, mTOR and other PIKKs have no known lipid kinase activity but instead have serine/threonine protein kinase activity. 

The N-terminus of mTOR contains multiple HEAT repeats, the middle part is the FRAP/ATM/TRRAP (FAT) domain, then the FKBP-rapamycin binding (FRB) domain, and the C-terminus contains the kinase domain (Figure 1) .

mTOR nucleates to form two functionally and structurally distinct complexes, mTORC1 and mTORC2. mTORC1 includes mTOR, mLST8 and RAPTOR (Figure 1) . 

mTORC1 is a dimer composed of the mTOR-mLST8-RAPTOR heterotrimer. RAPTOR is responsible for recognizing some mTORC1 substrates and targeting mTORC1 to the lysosomal surface. 

mTORC2 is a heterotetrameric dimer that includes mTOR, mLST8, RICTOR and mSIN1 (Figure 1) . 

mSIN1 intercalates into RICTOR via its N-terminus and then folds around mLST8. The conserved central region (CRIM) of mSIN1 is important for the recruitment of mTORC2 substrates. 

The mSIN1 C-terminal PH domain is required for the localization of mTORC2 on the membrane.

mLST8 is common to both complexes and binds close to the kinase site of mTOR (Figure 1) . Like RICTOR, embryos died at day 10.5 after mLST8 knockout, but at day 3.5 after mTOR or RAPTOR knockout, suggesting that mLST8 may not be important for mTORC1 function. 

mLST8 knockout MEFs retained the ability to phosphorylate mTORC1 substrates S6K1 and 4E-BP1, but failed to phosphorylate mTORC2 substrates AKT and PKCa. 

mLST8 stabilizes mTORC2 through direct interaction with mSIN1 and mTOR. 

Furthermore, the cryo-EM structure of mTORC2 revealed that mLST8 interacts with mSIN1 to localize the mSIN1 substrate-interacting CRIM domain. mLST8 does not have the same function in mTORC1.

Figure 1. Schematic diagram of mTOR and some substrates

2. Upstream regulation

mTORC1 integrates nutrients, growth factors, and energy input to promote anabolism and cell growth, while inhibiting catabolism (Figure 2) . 

mTORC1 senses nutrient availability through the small GTPases RAG-A, RAG-B, RAG-C and RAG-D. RAGs form heterodimers whose active conformations are RAG-A/RAG-B ^GTP and RAG-C/RAG-D ^GDP . 

This conformation recruits mTORC1 to the lysosome and is activated by the small GTPase RHEB. RAGs bind to regulators (a pentameric complex consisting of LAMTOR 1-5) on the surface of the lysosome .

 RAG-C/RAG-D is activated by folliculin complexes with GAP activity, including FLCN and FNIP1/FNIP2 . Inhibition of RAG-A/B upstream inhibitor GATOR1 can activate RAG-A/B. 

GATOR1 is inhibited by GATOR2, a complex of five proteins that is negatively regulated by amino acid sensors SESTRIN2 and CASTOR1.

 Recently, SAR1B was reported to cooperate with SESTRIN2 in leucine sensing and GATOR2 regulation.

Another interaction partner of GATOR1 is KICSTOR. KICSTOR anchors GATOR1 to the lysosome and ultimately inhibits RAG-A/B and mTORC1.

 In addition, the SAM sensor SAMTOR, upstream of mTORC1, indirectly senses the availability of methionine through SAM. 

Alpha-ketoglutarate produced by glutamate breakdown also modulates RAG activity by promoting RAG-B GTP loading and thus the localization of mTORC1 to the lysosome.

Growth factors activate mTORC1 through PI3K-AKT-TSC. PI3K is activated by tyrosine kinase receptors stimulated by cell surface insulin or other growth factors. PI3K converts PIP2 to PIP3. 

PDK1 binds to PIP3 in the cytoplasmic membrane and activates AKT by phosphorylating AKT-Thr308. Activated AKT phosphorylates and inhibits TSC complexes (including TSC1, TSC2, and TBC1D7) .

 The TSC complex is the GAP of the small GTPase RHEB. GTP-bound RHEB allosterically activates mTORC1 by binding to two HEAT repeats and the FAT domain in mTOR. Thus, PI3K-AKT-dependent TSC complex inhibition activates mTORC1 by locking RHEB in its activated GTP-bound state.

Energy stress inhibits mTORC1 via AMPK. Low ATP production causes increased intracellular AMP:ATP and ADP:ATP ratios, leading to allosteric activation of AMPK. 

In turn, AMPK restores ATP production by promoting glucose uptake and increasing b-oxidation. 

AMPK also activates autophagy by phosphorylating ULK1. Therefore, AMPK and mTORC1 are mutually antagonistic. AMPK inhibits mTORC1 under energy stress in two distinct ways:

(1) AMPK phosphorylates Thr1271 and Ser1387 of TSC2 to activate GAP activity of the TSC complex against RHEB, thereby preventing mTORC1 activation;

(2) AMPK directly phosphorylates RAPTOR Ser722 and Ser792 sites to inhibit mTORC1.

In contrast, mTORC1 directly phosphorylates the catalytic subunits a1 (Ser347) or a2 (Ser345 and Ser377) of AMPK, thereby reducing Thr172 phosphorylation in the AMPK activation loop, thereby limiting AMPK activity. 

Interestingly, AMPK specifically phosphorylates mTOR (Ser1261) at multiple sites in mTORC2 and RICTOR to increase mTORC2 activity independent of the mTORC1 negative feedback loop. 

It has been proposed that AMPK-dependent mTORC2 activation aims to activate anti-apoptotic AKT signaling to promote cell survival during acute energy stress.

Figure 2. mTOR signaling pathway

mTORC2 is activated by growth factors through PI3K. The core component of mTORC2, mSIN1, contains a PH domain that binds mTOR and thereby self-inhibits mTORC2. 

Upon activation of PI3K, the mSIN1 PH domain binds newly generated PIP3, thereby releasing mTOR to activate mTORC2.

 The PIP3-bound PH domain can also recruit mTORC2 to the plasma membrane, where it can phosphorylate membrane-bound substrates such as AKT. 

PI3K also promotes the association of mTORC2 with the ribosome, which is required for mTORC2 activation. Once activated, mTORC2 phosphorylates AKT Ser473 to fully activate AKT. 

mTORC2 activity can also be regulated through a negative feedback loop from mTORC1. mTORC1 and its direct effector S6K phosphorylate IRS1 at different sites to inhibit insulin signaling. 

mTORC1 also directly phosphorylates and activates GRB10. Thus, mTORC1 downregulates PI3K signaling to inhibit mTORC2 activity.

 S6K also phosphorylates Thr1135 and Thr86/Thr398 of RICRTOR and mSIN1, respectively, to destabilize mTORC2. The cellular localization of mTORC2 is more diverse than that of mTORC1.

 mTORC2 has been reported to be associated with plasma membrane, ribosome, mitochondria, Golgi apparatus, endosome, endoplasmic reticulum, and mitochondria-associated endoplasmic reticulum membrane. 

How mTORC2 is activated at these different bit sites is poorly understood.

3. mTOR inhibition

Rapamycin

mTOR is inhibited by rapamycin and its homologous analogs , called rapalogs . Rapamycin forms a complex with the endogenous protein FKBP, which binds and inhibits mTOR. 

FKBP-rapamycin binds to the FRB domain in mTOR adjacent to the catalytic cleft, thereby sterically hindering substrate access to the catalytic site. 

While mTORC1 is very sensitive to rapamycin, mTORC2 is not. 

RICTOR partially covers the FRB domain of mTOR. Thus, RICTOR shields the mTOR FRB domain in mTORC2, which in turn prevents FKBP-rapamycin binding. 

However, long-term rapamycin treatment can indirectly inhibit mTORC2. 

Rapalogs are rapamycin derivatives. They work in the same way as rapamycin but with improved drug-like properties. 

They are used clinically for immunosuppression and to fight cancer. 

ATP-competitive mTOR active site inhibitors have been developed that effectively inhibit both mTORC1 and mTORC2. However, to date no active site inhibitors have been approved for clinical use.

DEPTOR

mTORC1 and mTORC2 share the same endogenous inhibitor DEPTOR. Two recent studies have shown that DEPTOR regulates mTOR in an allosteric manner. 

DEPTOR is also a substrate of mTOR. Truncated mTOR directly phosphorylates DEPTOR, leading to its degradation by the proteasome.

 It is generally believed that DEPTOR is phosphorylated by mTORC1 and mTORC2, but this remains to be demonstrated. 

At high concentrations, DEPTOR binds to the mTOR FRB domain in a separate substrate-like mode through the junction region between the PDZ and DEPt domains. 

Interestingly, DEPTOR deletion only increased mTORC1 activity, whereas DEPTOR overexpression promoted mTORC2 activity. 

This paradoxical effect of DEPTOR on mTORC2 may be due to inhibition of the mTORC1 negative feedback loop.

PRAS40

mTORC1 has a specific endogenous inhibitor PRAS40. PRAS40 contains a TOS motif that is also present in some mTORC1 substrates. 

RAPTOR binds the TOS motif to recruit substrates to mTORC1. PRAS40 competes with other substrates for binding to mTORC1, thereby inhibiting downstream signaling. 

AKT phosphorylates PRAS40 Thr246, which induces its own release from mTORC1 and is sequestered by the cytoplasmic 14-3-3 protein. 

mTORC1 also phosphorylates PRAS40 Ser183, Ser212 and Ser221 to induce PRAS40 release from mTORC1.

4. mTORC1 substrates: S6K, 4E-BP, ULK1 and TFEB

The most well-described mTOR substrates are S6K and 4E-BP. 

They are widely used as indicators of mTORC1 activity because of the presence of antibodies that specifically recognize the phosphorylated forms of these two proteins.

S6K

There are two S6Ks in mammals, encoded by different genes. S6K belongs to the AGC kinase family. 

Members of the AGC kinase family are often substrates of mTORC2. S6K is the only AGC kinase and is an mTORC1 substrate. 

S6K1 is phosphorylated and activated by mTORC1 and PDK1. 

mTORC1 phosphorylates Thr389 in the HM of the linker region of S6K, and PDK1 phosphorylates Ser229 in the T-loop of the S6K kinase domain, phosphorylation of both sites is necessary for full activation of S6K1. 

S6K1 has an autoinhibitory C-terminus that is released by phosphorylation at Ser411, Ser418, Ser421 and Ser424 upon mitogen stimulation. 

Release of the autoinhibitory domain relaxes the S6K structure, allowing it to be phosphorylated by mTORC1 and PDK1. 

mTORC1 also phosphorylates Ser371 in S6K TM, which is also essential for S6K1 activity, however, there are few experiments to elucidate its specific function. 

S6K has a TOS motif that is required for mTORC1 recognition and phosphorylation.

4E-BP

There are three 4E-BP isoforms in mammals, each encoded by a different gene. 

All three are regulated by mTORC1. In the unphosphorylated state, 4E-BP binds and inhibits eIF4E to prevent translation initiation. 

4E-BP is in turn phosphorylated by mTORC1, and phosphorylation of Thr37/Thr46 reduces the binding affinity of 4E-BP-eIF4E by 100-fold, followed by phosphorylation of Ser65/Thr70 to reduce the affinity by another 40-fold.

 This releases 4E-BP, allowing eIF4E to bind eIF4G and initiate translation. 

A recent study elucidates how mTORC1 recognizes 4E-BP and phosphorylates it in a hierarchical manner. 

mTORC1 binds the TOS and RAIP motif of 4E-BP through RAPTOR (Arg-Ala-Ile-Pro, found only in 4E-BP) . 

Binding of these motifs by RAPTOR directs the phosphorylation site of 4E-BP toward the mTORC1 active site. 

eIF4E covers Ser65 and Thr70 sites, preventing them from being phosphorylated without prior phosphorylation of Thr37/Thr46. 

Interestingly, phosphorylation of Thr37/Thr46 was insensitive to rapamycin, whereas phosphorylation of Ser65/Thr70 was sensitive to rapamycin.

ULK1

ULK1 forms a complex with ATG13 and FIP200 to initiate autophagy. This step is inversely controlled by the opposing kinases mTORC1 and AMPK. 

Under nutrient-rich conditions, mTORC1 interacts with the ULK complex and phosphorylates ULK1 Ser757 and ATG13 Ser259, thereby inhibiting the complex ultimately blocking autophagy. 

Conversely, under nutrient-poor conditions, AMPK phosphorylates and activates ULK1 and ATG13, while inhibiting mTORC1 promotes autophagy.

 Once activated, ULK1 also promotes autophagy by inhibiting mTORC1 and hinders the interaction of substrates with RAPTOR through phosphorylation of RAPTOR at multiple sites.

TFEB

TFEB and TFE3 are homologous helix-loop-helix leucine zipper transcription factors that regulate involvement in lysosomal biogenesis, autophagy, and lipid metabolism. 

TFEB and TFE3 translocate into the nucleus to activate target genes upon starvation. 

When nutrients are abundant, they are phosphorylated by mTORC1 and thus retained in the cytoplasm by the 14-3-3 protein. Neither TFEB nor TFE3 contain a TOS motif and are recruited to mTORC1 by interacting with GDP-bound RAG-C/RAG-D. 

mTORC1 phosphorylates Ser122, Ser142 and Ser211 of TFEB. Ser211 phosphorylation mediates 14-3-3 binding and cytoplasmic sequestration. 

Mutation of TFEB-Ser142 or Ser211 to alanine results in constitutive nuclear localization of transcription factors. F

urthermore, in addition to controlling TFEB localization, Ser142 and Ser211 phosphorylation induces STUB1-mediated TFEB ubiquitination and degradation.

5. Substrates of mTORC2: AKT, PKC and SGK

The mTORC2 substrates AKT, PKC, and SGK all belong to the AGC kinase family. 

They have similar structures characterized by an N-terminal regulatory domain and a C-terminal catalytic domain consisting of a kinase domain and a C-tail. 

Key phosphorylation sites in AKT, PKC and SGK are located in the T-loop in the kinase domain and in the TM and HM (hydrophobic motifs) in the C-tail . 

Initial phosphorylation of the T-loop by PDK1 in response to upstream stimulation of PI3K followed by phosphorylation of HM by mTORC2.

 In the absence of PDK1 and T-loop phosphorylation, mTORC2 can still phosphorylate HM in AKT. 

Furthermore, unlike other AGC kinases, AKT also contains an N-terminal PH domain that is inhibited in the absence of PIP3.

 Therefore, PIP3-dependent membrane localization is also a prerequisite for T-loop (Thr308) and HM (Ser473) phosphorylation in AKT.

TM phosphorylation is distinct from T-loop and HM phosphorylation. In AKT and PKCa, the TM site is co-translationally phosphorylated by mTORC2. 

This phosphorylation is constitutive, independent of upstream inputs, and required for kinase stability.

 In AKT and PKC, phosphorylation of T-loop and HM determines kinase activity, while phosphorylation of TM determines protein stability. 

In AKT, T-loop phosphorylation is required for catalytic activity, whereas HM phosphorylation is required for maximal AKT activity. 

Less information is available on the phosphorylation of SGK, the SGK T-loop (Thr256) is phosphorylated by PDK1 and the HM (Ser422) is phosphorylated by mTORC2. 

The SGK TM kinase has not been identified, but TM phosphorylation is growth factor sensitive and required for full kinase activity.

Recently, a new model has been proposed that challenges the conventional view of HM phosphorylation in AGC kinases. 

Baffi et al. identified an evolutionarily conserved motif FXXXFT (F=phenylalanine, X=any amino acid, T=phosphoreceptor threonine) , called the TOR interaction motif (TIM) . 

TIM is the N-terminus of TM and is phosphorylated by mTORC2. Contrary to conventional wisdom, Baffi et al. propose that HM phosphorylation in AKT1 and PKCbII is not caused by mTORC2, but rather by AKT and PKCbII autophosphorylation.

 In this model, mTORC2 phosphorylates TIM, which in turn promotes PDK1 to phosphorylate T-loop, and then AKT1 and PKCbII to autophosphorylate HM.

6. Comprehensive review of all mTORC1 and mTORC2 substrates

mTORC1

The vast majority of substrates lack TOS, RAIP or any other motif. Substrates lacking known recognition motifs may bind mTORC1 in a specific way. 

Resolving such motif-free interactions may require structural analysis and docking simulations. 

Grouped according to function, these substrates can be clustered into translation, autophagy, and transcription. Translational regulation by 4E-BP is one of the first described roles of mTOR signaling. 

Direct mTORC1 phosphorylation by other translational or regulatory factors such as eIF2B, eIF4E, eEF2K, and S6K may be built-in functional redundancy to ensure fail-safe regulation of central functions of mTOR signaling. 

Similar reasoning can be applied to various autophagy factors (eg, AMBRA1, ATG13, ATG14, DAP1, NRBF2, PACER, TRPML1, ULK1, and WIPI2) , which are direct targets of mTORC1. 

For these autophagy factors, the result of mTORC1 phosphorylation is always towards the inhibition of autophagy.

 mTORC1 also phosphorylates and regulates components of signaling (eg, AMPK, IRS1, and GRB10) and metabolic pathways (eg, LIPIN1 and CRTC2) .

mTORC2

mTORC2 has fewer identified substrates than mTORC1. Most of the identified mTORC2 target proteins belong to the AGC kinase family (eg AKT, PKC and SGK) . mTOR autophosphorylation at Ser2481 in mTORC2 is conserved in vertebrates, but its significance is unclear. 

AMOTL2, MST1 and YAP belong to the Hippo pathway that controls organ size.

 The presence of three Hippo components among the few described mTORC2 substrates is highly suggestive of crosstalk between pathways.

 The mTORC2 site has a lower ratio of serine (27 sites) to threonine (20 sites) phosphoacceptor residues compared to the mTORC1 site .

 IGF1R and INS-R are reported to be phosphorylated on 2 tyrosine residues.

7. mTOR target protein phosphorylation MOTIF: S/TP

The authors found 104 and 51 phosphorylation sites in 56 mTORC1 and 26 mTORC2 substrates, respectively. 

Based on these in vivo substrates, a common recognition phosphorylation motif was constructed for mTORC1 and mTORC2. mTOR target sites favored serine (77%) as a phosphoacceptor residue compared to threonine (23%). 

Importantly, the mTOR target site mainly contains a proline at position +1. Leucine, glutamic acid, phenylalanine, tyrosine and glutamine are also favored at this position. 

Phosphorylation reactions may involve specific conformational adaptations or other effects beyond direct kinase-substrate interactions, such as post-translational modifications. 

Notably, there was no preference for any amino acid at position -1 when all substrates were considered. 

However, when focusing only on AGC kinase as a substrate, phenylalanine and leucine predominate at this position.

A 2011 study, based on in vitro phosphorylation of peptide libraries, proposed an mTOR co-phosphorylation motif. 

The putative motif consists of phosphorylated serine or threonine followed by proline, phenylalanine, leucine, tryptophan, tyrosine or valine in decreasing order of preference at the +1 position.

 Therefore, any sequence other than S/TP is limited in identifying putative mTOR sites.

8. mTOR substrate recognition: TOS and RAIP motif

mTORC1 and mTORC2 phosphorylate different substrates, although they share the same catalytic subunit.

 This is paradoxical because the kinase pockets in mTORC1 and mTORC2 are structurally similar and fully contained in mTOR without significant interactions with adjacent subunits. 

In addition, no mTORC1 or mTORC2-specific phosphorylation motifs were detected. 

The mechanisms explaining the phosphorylation of different substrates by mTORC1 and mTORC2 may involve subunits unique to each mTOR complex. 

There is evidence that subunits unique to each complex bind substrates to present to a common catalytic site.

In mTORC1, RAPTOR binds the TOS motif (FXF-(E/D)-F, F for hydrophobic residue, X for arbitrary residue) . 

The distance between the TOS-binding site in RAPTOR and the kinase site in mTOR is approximately 65 A˚, suggesting that at least a 20-amino acid separation is required between the TOS motif and the target phosphorylation site in the substrate. 

TOS motifs were first identified in S6K and 4E-BP and subsequently in other mTORC1 substrates.

 In addition, 4E-BP also possesses a RAIP motif, which can cooperate with the TOS motif to enhance the interaction of 4E-BP with RAPTOR. 

Other RAPTOR-binding motifs have also been reported, such as the SHC and IRS1 NPXY (SAIN) domains in IRS1 and the S/PXPXPP motif in eIF4E, ULK1, LARP1, and DAP1. 

Furthermore, some substrates can interact with mTORC1 in a RAPTOR-independent manner.

 Finally, the FRB domain in mTOR is considered a secondary substrate-binding site that can cooperate with the RAPTOR TOS-binding site in the S6K recruitment of mTORC1. 

In mTORC2, there is evidence that the CRIM domain in mSIN1 is responsible for substrate recruitment. 

The CRIM domain is highly flexible and is adjacent to the mLST8 and mTOR kinase sites. Whether there is a sequence motif in mTORC2 substrates, equivalent to the TOS or RAIP motif, recognized by the CRIM domain remains to be determined.

Conclusion and Outlook

mTOR can exert a wide range of effects by regulating targets such as 4E-BP and S6K, and through it regulates the levels of a variety of proteins. 

In addition, mTOR controls the level and activity of many transcription factors, thereby controlling gene expression.

 Therefore, direct phosphorylation of only a few mTOR targets is sufficient to have broad effects. 

Despite three decades of active research, many questions remain to be determined, such as how most substrates are recognized and recruited to mTORC1 or mTORC2 is unclear, and the intracellular localization of the TOR complex remains a huge problem. challenge.

Phosphorylated proteins are actively dephosphorylated by phosphatases. 

Thus, the activity of phosphorylated proteins and ultimately the signaling pathway is controlled by a balance of kinases and phosphatases.

 However, phosphatases are significantly fewer in number than kinases and are therefore less specific. This limited specificity has hindered progress in phosphatase research.

 An early indication of phosphatase involvement in mTOR signaling was the observation that PP2A binds to S6K under conditions of mTOR inactivation. 

mTOR phosphorylates PP2A, inactivating it to prevent S6K dephosphorylation. PP2A dephosphorylates MAP4K3, an upstream mTORC1 regulator required for amino acid stimulation. 

PP2A also binds RAPTOR to inactivate mTORC1 in Treg cells. 

Therefore, there is a functional interaction between mTOR and PP2A, so whether mTOR also interacts with other phosphatases is a question worth exploring in future studies.