Rapamycin and mTOR kinase inhibitors

Abstract

Mammalian target of rapamycin (mTOR) is a protein kinase that controls cell growth, proliferation, and survival. mTOR signaling is often upregulated in cancer and there is great interest in developing drugs that target this enzyme. Rapamycin and its analogs bind to a domain separate from the catalytic site to block a subset of mTOR functions. These drugs are extremely selective for mTOR and are already in clinical use for treating cancers, but they could potentially activate an mTOR-dependent survival pathway that could lead to treatment failure. By contrast, small molecules that compete with ATP in the catalytic site would inhibit all of the kinase-dependent functions of mTOR without activating the survival pathway. Several non-selective mTOR kinase inhibitors have been described and here we review their chemical and cellular properties. Further development of selective mTOR kinase inhibitors holds the promise of yielding potent anticancer drugs with a novel mechanism of action.

Phosphatidylinositol 3-kinase-related kinases

Mammalian target of rapamycin (mTOR) is a serine/threonine protein kinase whose catalytic domain most closely resembles the one found in lipid kinases of the phosphatidylinositol (PI) 3-kinase (PI3K) family. PI3Ks phosphorylate inositol phospholipids on the 3′-hydroxyl of the inositol ring and are divided into three classes based on structural features and substrate specificity (reviewed in Ref. [1]). Class I PI3Ks consist of four different catalytic subunits (p110α, p110β, p110δ and p110γ) that form heterodimers with a variety of regulatory subunits. These enzymes preferentially phosphorylate PI 4,5-bisphosphate (PI(4,5)P₂) in vivo. Crystal structures of p110γ bound to ATP and to various inhibitors have been instrumental in determining structure-activity relationships for PI3K inhibitors (discussed below). The structure of p110α bound to a fragment of one of its regulatory subunits was also solved recently [2]. Class II PI3Ks (PI3K C2α, C2β and C2γ) phosphorylate PI and PI(4)P in vitro and contain a C2 homology domain at the C terminus that is not found in other PI3Ks. The class III PI3K Vps34 only phosphorylates PI and may play a role in the regulation of mTOR by amino acids [3]. Type III PI 4-kinases (PI4Ks) phosphorylate the 4′-hydroxyl of the inositol ring and are closely related to PI3Ks [1].

mTOR belongs to the family of PI3K-related protein kinases (PIKKs) that includes ataxia–telangiectasia mutated (ATM), ATM- and Rad3-related (ATR), DNA-dependent protein kinase (DNA-PK) and suppressor of morphogenesis in genitalia-1 (SMG-1; reviewed in Refs. [1, 4]). These are large proteins (~300–500 kDa) that contain a conserved kinase catalytic domain together with other regions that can include HEAT repeats and FAT and FATC domains. mTOR also has an FRB (FKBP12/rapamycin-binding) domain that binds the drug rapamycin in complex with its intracellular receptor protein FKBP12. ATM, ATR, DNA-PK, and SMG-1 are involved in DNA and mRNA surveillance and repair pathways. By contrast, mTOR integrates growth factor-activated signals with permissive signals in the presence of sufficient nutrients and energy to promote cell growth, proliferation, and survival. mTOR signaling is upregulated in many cancers and some benign growth or proliferative disorders. Therefore, drugs that target mTOR activity are expected to be useful therapies for these conditions.

mTOR signal transduction and cellular functions

The discovery of mTOR and the understanding of its biological functions were greatly facilitated by the use of rapamycin, which inhibits some of the functions of mTOR. Until recently, rapamycin sensitivity was the major criterion used to identify mTOR-controlled events. However, it is now known that mTOR binds to different regulatory subunits to produce complexes with distinct signaling functions and rapamycin sensitivity [5, 6]. The mTORC1 complex (containing mTOR, Raptor and mLST8) phosphorylates ribosomal protein S6 kinase (S6K) at Thr389 and the translation repressor 4EBP1 and is rapamycin sensitive. Biological processes regulated by mTORC1 include translation, ribosome biogenesis, autophagy, glucose metabolism, and the cellular response to hypoxia [7–9]. The mTORC2 complex (containing mTOR, Rictor, mLST8, and mSin1) phosphorylates the protein kinase Akt at Ser473 and is insensitive to rapamycin. In comparison to mTORC1, the biological function of mTORC2 is less clear. However, available evidence suggests that this mTOR complex controls cell survival and organization of the actin cytoskeleton.

The physiological importance of the mTOR complexes is underscored by genetic ablation of their molecular components in mouse models. Mouse embryos lacking mTOR die at E5.5–6.5 days [10, 11]. Ablation of Raptor to disrupt mTORC1 is equally lethal at approximately E6.5 days [12]. Mouse knockouts of Rictor or mSIN1 leading to disruption of mTORC2 are also embryonic lethal [13, 14]. Not surprisingly, embryos lacking mLST8 also do not survive [12]. Although Akt Ser473 phosphorylation was blocked in cells isolated from Rictor^−/− and mSIN1^−/− embryos, phosphorylation of several Akt substrates was not inhibited, with the exception of FOXO transcription factors [12, 13]. Part of the prosurvival function of Akt is to phosphorylate and suppress the activity of FOXO proteins. These results suggest that mTORC2 is necessary for Akt-FOXO survival signaling. Additional studies using cells derived from these animals might shed more light on the specific cellular processes regulated by the distinct mTOR complexes. These studies will also be facilitated if small molecules that can specifically inhibit mTORC2 activity become available.

A number of different signaling pathways regulate mTORC1 activity, and the best characterized positive effector is the growth factor/PI3K/Akt pathway (Fig. 1). Akt plays an unusual role in mTOR signaling because it acts upstream of mTORC1 and downstream of mTORC2. Akt controls mTORC1 in part through tuberous sclerosis complex (TSC) 2, a protein that has GTPase-activating protein (GAP) activity toward Rheb, a small GTP-binding protein related to Ras [15]. TSC2 forms a tight complex with TSC1. The GAP activity of the TSC1/TSC2 heterodimer converts Rheb to an inactive GDP-bound state and thus suppresses mTOR activity [16]. In the presence of growth factors that activate Akt, Akt phosphorylates TSC2 at Ser939 and Thr1462 and inhibits its GAP activity [17]. This allows Rheb to maintain an active GTP-bound form that activates mTORC1 and increases signaling to S6K and 4EBP1.

Fig. 1.

mTOR signaling. Growth factor activation of PI3K causes increased production of PI 3,4,5-trisphosphate (PI(3,4,5)P₃). Akt is activated upon binding to PI(3,4,5)P₃ and phosphorylation of Thr308 by phosphoinositide-dependent protein kinase 1 (PDK1) and Ser473 by mTORC2. The TSC1/TSC2 complex inhibits mTORC1 by stimulating the GTPase activity of Rheb, converting it to the inactive GDP-bound state. Akt phosphorylation of TSC2 inhibits the TSC1/TSC2 complex and allows Rheb to remain in the active GTP-bound form. Phosphorylation by AMPK activates TSC2 and opposes the Akt signal. In energy-poor conditions, increased AMP levels allow LKB1 to phosphorylate and activate AMPK. Decreased signaling through Vps34 in the presence of low amino acid levels to inhibit mTOR might be independent of TSC1/TSC2 and Rheb. mTOR is inhibited during hypoxia due to activation of TSC1/TSC2 by REDD1. Rheb-GTP binds to and activates mTORC1. mTORC1 phosphorylates the translation repressor 4EBP1 and S6K at Thr389, leading to its activation. Inhibitors that target the mTOR kinase domain block both mTORC1 and mTORC2, whereas rapamycin only inhibits mTORC1. Upregulation of S6K in cells with hyperactive mTORC1 signaling has a feedback inhibitory effect on Akt. Therefore, treatment with rapamycin leading to inhibition of S6K can upregulate Akt and enhance cell survival. In contrast, mTOR kinase inhibitors block both S6K and Akt activation and should not enhance cell survival. It is not clear how TSC1/TSC2 and Rheb regulate the activity of mTORC2 [24]

An energy-sensing pathway also regulates mTOR activity. Inhibition of mTOR in response to low intracellular energy levels is mediated by AMP-activated kinase (AMPK) and its activator, the protein kinase LKB1 [18]. Under conditions in which intracellular ATP is depleted and the level of AMP is increased, AMP binds to AMPK and allows LKB1 to phosphorylate Thr172 on the catalytic α subunit to activate AMPK [19]. AMPK then phosphorylates TSC2 on Ser1345, which primes TSC2 for subsequent phosphorylation of Ser1341 and Ser1337 by glycogen synthase kinase 3 [20]. Together, these modifications are thought to enhance the GAP activity of TSC2, inactivate Rheb, and turn off mTORC1 signaling (Fig. 1).

Hypoxia and low amino acid levels also negatively regulate mTOR activity. Signaling pathways that mediate the mTOR response to these stimuli are less well characterized. The class III PI3K, hVps34, seems to play an important role in transducing the amino acid signal to mTOR [3, 21]. Evidence indicates that the hypoxia signal is mediated by a pathway from REDD1 (regulated in development and DNA damage responses) to TSC1/2 and then to mTOR [22].

Recent studies indicate that feedback from the mTORC1/S6K pathway inhibits growth factor signaling to PI3K (Fig. 1). TSC1^−/− or TSC2^−/− cells have abnormally low Akt activity due to hyperactivation of mTORC1/S6K [23–25]. Conversely, S6K1^−/− cells are hypersensitive to insulin activation of PI3K signaling [26, 27]. Treatment of some cancer cells with rapamycin upregulates Akt, which can enhance survival under conditions that usually induce apoptosis [23, 28–30]. As a result, there is concern that reactivation of Akt in tumors following rapamycin treatment could lead to resistance to other chemotherapeutic agents. It has been proposed that small molecule inhibitors that target the kinase domain of mTOR will display broader antitumor activity than rapamycin because they should not reactivate Akt [7, 31, 32]. Theoretically, a combination of rapamycin plus an inhibitor of Akt or PI3K would have the same effect.

Rapamycin and rapalogs

Rapamycin (sirolimus) is a macrocyclic antibiotic produced by the bacterium Streptomyces hygroscopicus found in the soil of Easter Island. Rapamycin was discovered as a potent antifungal agent, but it also exhibited what was at first considered to be an undesirable immunosuppressive effect, which subsequently led to its development as a clinically useful drug. The immunosuppressant FK506 (tacrolimus) and rapamycin have similar chemical structures and bind to the same intracellular receptor, FKBP12. Even though FK506 and rapamycin bind to the same protein, they have different mechanisms of action in cells. FK506 inhibits T cell proliferation by blocking the Ca²⁺/calcineurin-dependent transcriptional activation of genes required for growth, whereas rapamycin interferes with growth-promoting cytokine signaling [33]. Interleukin-2 (IL-2) induced S6K activation in a T cell line was extremely sensitive to rapamycin inhibition (IC₅₀ = 0.05 nM) [34]. By contrast, mTORC1 kinase activity in vitro was much less sensitive to FKBP12/rapamycin [6]. The reason for this apparent difference in mTORC1 sensitivity to FKBP12/rapamycin inhibition in vivo and in vitro is not understood.

In addition to rapamycin, three rapamycin analogs (rapalogs) are now in use in humans (Fig. 2). CCI779 (temsirolimus) is a dihydroxymethyl propionic acid ester prodrug of rapamycin. This modification makes the compound water soluble and thus it can be administered intravenously. Upon injection, CCI779 is rapidly converted to rapamycin, which is probably responsible for most, if not all, of its pharmacological effects. RAD001 (everolimus) has a O-(2-hydroxyethyl) chain substitution at position C-40 and AP23573 (deforolimus) has a phosphine oxide substitution at the same position of the lactone ring of rapamycin. Rapalogs have the same mechanism of action as rapamycin. They bind to FKBP12 and interfere with the FRB domain of mTOR.

Fig. 2.

Chemical structure of rapamycin, rapalogs and mTOR kinase inhibitors. Rapalogs have the indicated O-substitutions at the C-40 position of rapamycin (underline)

Unlike kinase inhibitors that target the catalytic ATP-binding site, these compounds are highly specific for mTORC1. However, the exact mechanism of how the interaction with the FRB domain leads to inhibition of mTORC1 remains unclear. FKBP12/rapamycin inhibits mTOR autophosphorylation and phosphorylation of 4EBP1 in vitro, suggesting that changes in the FRB domain might exert an allosteric influence on the catalytic domain [35]. It was initially thought that rapamycin interacts with the FRB domain only after binding to FKBP12. The X-ray crystal structure of the FKBP12-rapamycin-FRB complex shows that rapamycin occupies two different hydrophobic binding pockets in FKBP12 and the FRB domain simultaneously and brings the two proteins close together, but there are relatively few interactions between the proteins [36]. Subsequently, rapamycin was shown to bind to the FRB domain without FKBP12, although with a lower affinity [37]. Using this information and a solution NMR structure of the FRB domain, Leone et al. virtually screened a chemical library and discovered additional small molecules that bind to the FRB domain in the absence of FKBP12 [38]. However, these molecules were generally ineffective at inhibiting mTOR kinase activity [38]. Another group used high-throughput screening to identify 4-[6-{[(1S,2R)-2-(benzyloxy)cyclopentyl]acetyl}-4-(2-thienyl)pyridin-2-yl]-4-oxobutanoic acid (HTS-1) that binds to the FRB domain in the low micromolar range [39]. Unfortunately, they did not show if the compound inhibits mTOR kinase activity in vitro. Interestingly, these investigators also found that phosphatidic acid, which activates mTOR [40], interacts with the FRB domain in the same binding pocket as rapamycin. It remains unclear why binding of phosphatidic acid activates mTOR whereas rapamycin binding has the opposite effect. Interestingly, mTOR with a mutation in the FRB domain that renders it rapamycin resistant exhibits decreased catalytic activity [41], supporting the hypothesis that this domain modulates the catalytic domain. Future efforts may identify novel compounds that bind to this area and potently inhibit mTOR.

Rapamycin has also been proposed to inhibit mTOR by destabilizing the mTOR-Raptor complex [42]. Interestingly, mTOR has been shown to be inhibited by S-trans,trans-farnesyl thiosalicylic acid (FTS), a compound that resembles farnesylcysteine found in Ras family members and other proteins. FTS dislodges farnesylated Ras proteins from cell membranes by competing for their membrane binding sites, thus facilitating their degradation [43]. FTS inhibits mTORC1 activity in intact cells and cell extracts by promoting the dissociation of Raptor [44]. It should be noted that although rapamycin does not exert acute effects on mTORC2, long-term treatment of some cell types with the drug has been shown to prevent mTORC2 assembly, thus inhibiting Akt [45]. This mechanism was proposed to account for hyperlipidemia that is commonly seen in patients treated with the drug and may also account for hyperglycemia, which is another common side-effect.

A recent study provided evidence that the intracellular protein FKBP38 is an endogenous inhibitor of mTORC1 [46]. FKBP38 bound to the FRB domain of mTOR and inhibited its activity in vitro, similar to FKPB12/rapamycin. Furthermore, the FKBP12/rapamycin complex competed with FKBP38 for binding to mTOR in vitro, and cells treated with rapamycin had reduced association between FKBP38 and mTOR [46]. Upon growth factor stimulation, Rheb–GTP bound strongly to FKBP38 and displaced it from mTOR, thus activating the enzyme. These data raise the possibility that the FKBP12/rapamycin complex might inhibit mTOR because it cannot be displaced by GTP–Rheb.

Clinical uses of rapamycin and rapalogs

Rapamycin is an oral drug and its bioavailability is low. RAD001 is also an oral compound and an often-cited rationale for its development is that it has improved bioavailability. However, in a clinical pharmacokinetic study, RAD001 was found to also have a relatively low bioavailability (~15%) [47]. Rapamycin and RAD001 are both approved for use as immunosuppressants for organ transplant patients. Due to their pharmacokinetic properties, drug level monitoring of these drugs is necessary and they require daily dosing. Rapamycin inhibits IL-2-induced T cell proliferation [34] that is due at least in part to a specific down-regulation of ribosomal protein mRNA translation [48]. It is believed that the immunosuppressive effect of RAD001 is caused by a similar mechanism of action. Post-transplant malignancy is a serious complication of solid organ transplantation. In contrast with calcineurin inhibitors such as FK506, the use of mTOR inhibitors as immunosuppressants has the added benefit of being antitumorigenic. Indeed, rapamycin is now the established treatment for post-transplant Kaposi’s sarcoma in patients who were on other immunosuppressive regimens [49].

CCI779 and AP23573 can be given intravenously and are being used as anticancer drugs. CCI779 was recently approved by the U.S. Food and Drug Administration (FDA) for treatment of renal cell carcinoma. AP23573 is in clinical trials for treating a variety of cancers, including sarcomas. RAD001 and rapamycin are also being tested in clinical studies for treatment of both hematologic and solid malignancies. Aside from a possible direct antiproliferative effect on tumor cells, experiments in a mouse model of metastatic cancer suggest that rapamycin limits the growth of solid tumors by blocking angiogenesis [50]. Intravenous CCI779 and AP23573 are administered on intermittent weekly schedules. Given in this fashion, these drugs do not appear to cause immunosuppression. This is surprising because rapalogs have long half-lives (RAD001, 24–35 h; CCI779, 13–25 h; AP23573, 45–74 h). In a rat model, a single dose of RAD001 suppressed mTOR signaling (measured by S6K activity and 4EBP1 Thr70 phosphorylation) in peripheral blood mononuclear cells for longer than 72 h [51]. A possible explanation for this discrepancy is that the immunosuppressive effect of rapamycin is mediated by other effectors that are not inhibited by the drug concentrations achieved during these treatments. S6K activity in blood cells or tissue might not be a good indicator of pharmacodynamic effectiveness because this enzyme is exquisitely sensitive to rapamycin inhibition.

Rapamycin-eluting coronary artery stents were approved for use by the U.S. FDA in 2003. Rapamycin inhibits vascular smooth muscle cell migration and proliferation and attenuates reocclusion of coronary arteries following angioplasty. In a randomized, double-blind trial with 1058 patients, patients treated with rapamycin-eluting stents had a restenosis rate of 8.6% vs. 21% in patients treated with a standard non-drug-eluting stent [52]. It is conceivable that the antiproliferative effect of mTOR inhibitors might be utilized to treat other non-malignant proliferative disorders such as polycystic kidney disease [53].

mTOR kinase inhibitors

A small molecule designed to compete with ATP in the catalytic site of mTOR would be expected to inhibit all of the kinase-dependent functions of mTORC1 and mTORC2, unlike rapalogs that only target mTORC1. Most if not all of the non-rapalog mTOR inhibitors described to date in the scientific literature were developed to inhibit other enzymes, especially class I PI3Ks. Because PI3K regulates mTOR activity (see Fig. 1), inhibitors that target both enzymes are generally not useful as research tools to study mTOR regulation or function. However, drugs that are dual PI3K/mTOR inhibitors might have a therapeutic advantage over single-target inhibitors in certain disease settings.

Wortmannin (Fig. 2) is a toxic steroidal furan that is produced by various fungi including Penicillium wortmanni [54]. The first kinase shown to be targeted by wortmannin was smooth muscle myosin light chain kinase (IC₅₀ = 0.17 μM). Inhibition was irreversible and the enzyme was partially protected by incubation in the presence of ATP [55]. Later studies exploring wortmannin inhibition of agonist-induced responses in neutrophils and basophils led to the discovery that the compound is a potent inhibitor of class I PI3Ks (IC₅₀ ~ 1–3 nM) [56, 57]. Biochemical analyses and the X-ray crystallographic structure of wortmannin bound to p110γ showed that the mechanism of inhibition is two-fold. First, wortmannin binds with high affinity to the ATP-binding site to block substrate binding. Second, the ε-amino group of a lysine in the active site forms a covalent bond with carbon-20 of the wortmannin furan ring to irreversibly inhibit the enzyme [58, 59]. The active site lysine (Lys802 in p110α and Lys833 in p110γ) is critical for catalysis and is highly conserved in lipid and protein kinases. Wortmannin also targets class II PI3K C2β (IC₅₀ = 1.6 nM), the human class III PI3K hVps34p (IC₅₀ = 2.5 nM), type III PI4Ks (IC₅₀ = 50–300 nM) and polo-like protein kinases that regulate mitosis (IC₅₀ = 24–48 nM) [60–64]. Not surprisingly, considering the similarity between the PI3K and PIKK catalytic domains, wortmannin also irreversibly inhibits PIKK family members. DNA-PK is the most sensitive (IC₅₀ = 16 nM), followed by SMG-1 (IC₅₀ ∼ 60 nM), ATM (IC₅₀ ~ 100–150 nM), mTOR (IC₅₀ ~ 200 nM) and ATR (IC₅₀ = 1.8 μM) [65–68]. Wortmannin forms a covalent bond with mTOR, presumably at Lys2187 in the ATP-binding site [65]. Based on its potency and selectivity, wortmannin at 100 nM has been recommended for use in cells to assess the role of PI3K [69], although effects on mTOR at this concentration should be considered. The use of wortmannin as a therapeutic agent is limited by its toxicity and instability in biological solutions. The wortmannin derivative PX-866 is more stable and less toxic than the parent compound and exhibits antitumor activity in mice, but it was reported to be inactive against mTOR at concentrations up to 10 μM [70].

Theophylline and caffeine are naturally occurring methylxanthines that have pleiotropic effects on the nervous, respiratory, cardiovascular, and renal systems. Caffeine is the world’s most widely consumed stimulant and reaches plasma concentrations of ~50 μM in moderate coffee drinkers. Plasma levels above 200 μM are toxic in humans. Caffeine is used medically to treat apnea of prematurity and is a component of various headache and pain remedies [71]. Theophylline causes bronchodilator and antiinflammatory responses and has long been used clinically for the treatment of asthma and other respiratory diseases. The therapeutic serum levels range from 55–111 μM, while concentrations >111 μM are considered to be toxic [72]. Many biochemical actions of methylxanthines have been identified, including antagonism of adenosine receptors, inhibition of cyclic nucleotide phosphodiesterases, and increased Ca²⁺ release from the sarcoplasmic reticulum. These compounds also inhibit mTOR and related kinases, most likely by acting as low affinity ATP analogs. Theophylline (5 mM) strongly inhibits mTOR kinase activity in vitro and blocked insulin activation of Akt in 3T3-L1 adipocytes [73]. However, since theophylline and caffeine also target class I PI3Ks with IC₅₀ values ranging from 75 μM to 1 mM [74], effects on Akt cannot be attributed solely to mTOR inhibition. Caffeine has been used extensively in cell-based experiments to investigate cell cycle checkpoints in G1/S and G2/M that are induced by DNA damage. These studies led to the discovery that caffeine inhibits ATM (IC₅₀ ~ 0.2 mM ), ATR (IC₅₀ ~ 1.1 mM), SMG-1 (IC₅₀ = 0.3 mM), and mTOR (IC₅₀ ~ 0.2–0.4 mM ). There is disagreement about whether DNA-PK is also inhibited by caffeine [41, 68, 75–77]. Interestingly, TORC1 signaling is an important target for caffeine in yeast, and several caffeine-resistant mutants of the TOR1 protein have been identified [78]. These proteins contain two mutations: one in the kinase domain that probably impairs binding to both caffeine and ATP, resulting in greatly reduced kinase activity, and a second mutation in the FRB that increases the kinase activity. The double mutants exhibit strong caffeine resistance and kinase activity that is high enough to support TORC1 function in vivo. By contrast, mutations in the FRB that cause rapamycin resistance decrease TOR1 kinase activity, again highlighting the idea that this domain regulates TOR1 activity in multiple ways. Caffeine and theophylline also block the activity of class II PI3K C2α (IC₅₀ ~ 0.4 mM) [74].

The field of PI3K signaling was revolutionized by the introduction of LY294002 (2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4-one) (Fig. 2), a synthetic PI3K inhibitor whose structure is based on the naturally occurring flavonoid quercetin [79]. LY294002 behaves as a competitive inhibitor for the ATP-binding site and is much less potent against class I PI3Ks than wortmannin (IC₅₀ ~ 0.3–4 μM) [80]. However, its superior chemical stability in solution has led to its widespread use as a PI3K inhibitor in cell-based experiments, where it is usually used at concentrations of 10–50 μM. Despite its frequent description as a “specific” PI3K inhibitor, LY294004 blocks the activity of a number of different protein kinases unrelated to PI3K [81]. In addition, use of immobilized LY294002 showed that the drug binds tightly to numerous non-kinase proteins with diverse functions, with as yet unknown biological consequences [80]. Class II PI3Ks (IC₅₀ = 6.9–19 μM) and PI4Ks are relatively resistant to LY294002 as compared with the class I enzymes [60, 61]. By contrast, the PIKK family members SMG-1, DNA-PK, and mTOR are targeted at low micromolar concentrations that also inhibit class I PI3Ks [41, 65, 82, 83]. The X-ray crystallographic structure of p110γ bound to LY294002 shows that the morpholino oxygen forms a hydrogen bond with the backbone amide of Val882 in the pocket that can be occupied by adenine in ATP [58]. Substitution of this oxygen with nitrogen yields a compound (LY303511) that is essentially inactive against PI3K [79]. Treatment of cells with LY303511 inhibited agonist-induced S6K Thr389 phosphorylation and autophosphorylation of mTOR at Ser2481, suggesting that LY303511 might be an mTOR inhibitor. However, effects on mTOR activity were not directly tested using in vitro assays, so whether or not LY303511 is a bona fide mTOR inhibitor remains an open question [84].

Since the advent of LY294002, there has been an explosion of interest in the synthesis of small molecule PI3K inhibitors that show isoform selectivity and improved potency and specificity [1]. Unfortunately, so far there have been no reports of an analogous effort to systematically produce mTOR inhibitors based on a structural or functional understanding of its catalytic site. However, some of the new compounds developed as PI3K inhibitors have also been evaluated for activity against mTOR. Knight et al. [85] determined the specificity profile of PI3K inhibitors from nine different chemical classes by measuring IC₅₀ values in vitro against 55 purified kinases, including 15 PI3K family members. Six compounds were reported to inhibit mTORC1 activity with IC₅₀s ≤ 10 μM. Interestingly, two of the compounds were less potent against mTORC2 than mTORC1, suggesting that the catalytic site of the common mTOR subunit might be subtly altered by its association with the distinct binding partners in the two complexes. The most potent mTOR inhibitor in this group was PI-103 (Fig. 2; IC₅₀ = 0.02 μM and 0.083 μM against mTORC1 and mTORC2, respectively). A model of the three-dimensional structure of PI-103 bound to the active site of p110γ showed that the inhibitor forms a hydrogen bond with Val882 in the “adenine pocket” and stretches deep into an “affinity pocket” not accessed by ATP that contains the side chain of Ile879 [85]. The residues in mTOR at equivalent positions are Val2240 and Ile2237, respectively, suggesting that binding of PI-103 to the catalytic site of mTOR involves similar interactions in this region.

Although PI-103 represents the first potent synthetic inhibitor of mTOR, it is essentially equipotent against class I PI3Ks (IC₅₀ = 0.008–0.15 μM). Although this dual specificity limits the use of PI-103 as a probe to study mTOR function, PI-103 has proved to be useful as an experimental compound in cancer research. PI-103 was found to be the most active compound in a group of ten isoform-selective PI3K inhibitors that were evaluated for the ability to block the proliferation of glioma cells in vitro [86]. The cytostatic property of the compound was attributed to its ability to inhibit both p110α and mTOR. PI-103 treatment of mice also reduced the size of established tumor xenografts at doses that produced no observable toxicity [86]. These results suggest that dual-specificity PI3K/mTOR inhibitors, or use of a PI3K inhibitor in combination with rapamycin, might be a viable therapeutic option for the treatment of certain cancers.

Programs to develop potent and specific inhibitors of DNA-PK have also yielded some compounds that inhibit mTOR [83, 87]. In general, these compounds inhibit DNA-PK (IC₅₀s in the sub-millimolar range) and p110α and mTOR (low millimolar IC₅₀s), but not ATM or ATR. One exception to this rule is 2-(morpholin-1-yl)pyrimido[2,1-α]isoquinolin-4-one (compound 13 in Ref. [83]), which does not inhibit PI3K. We used this compound (named 401 in Ref. [88]) to examine the cellular effect of mTOR inhibition without complicating side effects on PI3K. The chemical structure of 401 is shown Fig. 2. First we performed in vitro kinase assays to confirm that 401 is selective for mTOR over p110α and p110β. The specificity profile was further widened by assaying 40 different protein kinases in the presence of 5 μM 401. As expected for an mTOR inhibitor, treatment of cells with 401 blocked growth factor-induced phosphorylation of S6K Thr389 (an mTORC1 site) and Akt Ser473 (an mTORC2 site). By contrast, phosphorylation of Erk was not affected. These effects on cell signaling were not due to inhibition of DNA-PK, as incubation of cells that lack DNA-PK with 401 also decreased S6K and Akt phosphorylation. Finally, we examined the effect of 401 on the growth and survival of mouse embryo fibroblasts (MEFs) deficient for TSC1. These cells have abnormally high mTORC1/S6K signaling and abnormally low Akt activity due to feedback inhibition from the hyperactivated mTORC1/S6K pathway. Long term treatment of these cells with rapamycin turns off the negative pathway and upregulates Akt, which can provide a survival signal [23]. By contrast, we found that Akt phosphorylation remained low in TSC1^−/− MEFs cultured in the presence of 401. Furthermore, 401-treated cells exhibited strong growth inhibition and increased apoptosis as compared with MEFs treated with rapamycin [88]. Treatment of TSC1^−/− MEFs with an Akt inhibitor also increased apoptosis, suggesting that the cytotoxic effect of 401 might be due to suppression of mTORC2/Akt signaling. These results suggest that inhibition of mTOR kinase activity by a small molecule inhibitor such as 401 might be more effective than rapamycin at killing cancer cells that exhibit hyperactivated mTOR signaling.

Conclusions and outlook

Upregulation of the PI3K/mTOR/Akt pathway is a common feature of many proliferative disorders, including cancer. Up to now, rapamycin and rapalogs are the most well studied mTOR inhibitors and they are now clinically used as cancer treatments. However, the possibility that inhibition of mTORC1 with these drugs might lead to Akt upregulation and outgrowth of more aggressive lesions is a concern. Use of rapamycin plus an inhibitor of Akt or PI3K is one way to circumvent this problem, and another is to use a dual-specificity agent such as PI-103 that targets both PI3K and mTOR. Another strategy is to develop drugs that selectively target the mTOR kinase domain, which should inhibit both mTORC1 and mTORC2. A treatment combining an mTORC2 inhibitor plus rapalogs would have a similar effect. One caveat regarding these strategies is that each drug class is expected to cause a distinct spectrum of toxicities. For example, mTORC2 inhibition might be less toxic than PI3K or Akt inhibitors because it would affect mainly FOXO signaling and not other downstream effectors of Akt.

Considerable progress has been made in the last several years toward elucidating the structure and regulation of the two mTOR complexes, and it is now possible to assay each one in vitro. Determination of the three-dimensional structure of the mTOR kinase catalytic domain would be a major breakthrough that would aid in the design and analysis of new inhibitors. Development of drugs specific for the mTOR kinase domain or that disrupt the mTORC2 complex should narrow the gaps in our knowledge about the importance of mTORC1 and mTORC2 in health and disease.