The phosphatidylinositol-3-kinase (PI3K)/mTOR pathway has been a major focus of attention for cancer researchers in the past decade. A preliminary and incomplete understanding of the molecular biology of this complex network has importantly conditioned not only the development of the first generation of mTOR inhibitors, but also the biomarker studies designed to identify the best responders to these agents. Most recently, research in this pathway has focused on the dual nature of mTOR that is integrated by the mTOR complex 1 and complex 2. These two complexes are formed and regulated by different proteins and are also driven by multiple different compensatory feedback loops. This deeper understanding has allowed the development of a promising second generation of inhibitors, which are able to block simultaneously both complexes due to their catalytic activity over mTOR. Moreover, some of them also exert an inhibitory effect over PI3K that is a key player in the feedback loops. This article reviews the newest insights in the signaling of the mTOR pathway and then focuses on the development of the new wave of mTOR inhibitors. Mol Cancer Ther; 10(3); 395–403. ©2011 AACR.
This article is featured in Highlights of This Issue, p. 373
Since the discovery of mTOR in the early 1990s, the volume of research done in this pathway has been substantial. These data have provided us with an increasingly detailed knowledge about the proteins and regulators involved in it, their different functions, and the genetic abnormalities that are present across different tumor types. Moreover, the interest among the scientific community for this pathway has been fostered by the development of a natural product derived from the bacterium Streptomyces hygroscopicus. This compound called rapamycin (sirolimus, Rapamune; Wyeth) has shown inhibitory activity against mTOR protein after coupling its intracellular receptor. Subsequently, several compounds have been synthesized with similar characteristics to rapamycin integrating the family of rapalogs.
However, the clinical results obtained by targeting this pathway have not been as straight forward as it was presumed at the beginning. Moreover, drug development against mTOR was started when the knowledge about its functions was very preliminary. Several key findings have changed the course of clinical research in this field. First, the fact that mTOR is constituted by two complexes—mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2) that have a very intricate network of feedback loops, protein partners, substrates, and regulators that are specific to each. Second, the discovery that rapamycin and rapalogs exert an incomplete inhibition of mTORC1 and also are inactive against mTORC2. Finally, mTORC2 was shown to be one of the major regulators of the feedback loops associated with this pathway, thus explaining the limited activity of rapalogs observed in clinical studies. Therefore, a closer analysis of the recent advances in the molecular biology of this pathway will help to correctly understand the results from previous in vitro studies and clinical trials.
In this article, we will review the data on the characterization of mTORC1 and mTORC2, their protein components, functions, and regulators emphasizing the role of the feedback loops recently described within this complex network. Then, the approved indications for the rapalogs will be summarized. Finally, the last section will be devoted to a new class of compounds that are able to inhibit both mTOR complexes, and the new dual inhibitors that are also adding activity against the phosphatidylinositol-3-kinase (PI3K), a key component of the main feedback loop involved in this pathway.
Molecular Biology of the mTOR Pathway: A Story of Two Complexes
The PI3K-AKT-mTOR pathway (Fig. 1) is commonly altered in human cancers. Deregulation can be secondary to amplification or mutations in PIK3CA, which encodes the p110α catalytic subunit of the kinase complex and have been extensively described in several tumors (1); mutations and amplification in AKT; inactivation or mutations in phosphatase and tensin homolog (PTEN); and other less frequent events such as mutations in the insulin-receptor substrates (IRS) and the Ras homolog enriched in brain (RHEB; refs. 2–4).
mTOR is a serine/threonine kinase formed by two signaling complexes called mTORC1 and mTORC2 that contain common and specific partner proteins. Both complexes share the following proteins: mTOR, mLST8/GβL, and the negative regulator deptor. On the contrary, they are integrated by distinct partner proteins and regulatory mechanisms acting on different substrates, and having specific effects on distinct cellular functions (5). mTORC1 is specifically composed by a regulatory associated protein of mTOR (raptor) and a proline-rich AKT substrate of 40 kDa (PRAS40). mTORC2 couples with the rapamycin-insensitive companion of mTOR (rictor), mSin1, and PRR5/Protor (Fig. 1).
mTORC1 enhances cell growth and proliferation by inducing protein and lipid synthesis, ribosome biogenesis, and reduction of autophagy (6–9). Growth factors and nutrients, such as energy and amino acids, promote mTORC1 signaling through the phosphorylation of eukaryotic initiation factor 4E-binding protein 1 (4E-BP1) and ribosomal S6 kinase 1 (S6K1) which are the best-known downstream effectors of mTOR (10).
The tuberous sclerosis complex 2 (TSC2) is an essential link between growth-factor signaling and the mTORC1 activation triggered via PI3K-dependent or -independent pathways (11–13). Growth-factor signaling phosphorylates and inhibits TSC2 (tuberin), avoiding its association with TSC1 (hamartin), thus activating the mTORC1 by releasing the inhibition of RHEB, a small guanosine triphosphatase (GTPase) necessary for the activation of mTORC1 (14). Likewise, inactivating mutations in the TSC1 or TSC2 genes cause hamartoma syndromes associated with elevated mTORC1 activity (15, 16). However, TSC2 is not required for the regulation of mTORC1 by amino acids; the Ras-related GTPase proteins, a family of small GTPases, are the key regulators for mTORC1 amino acid activation (17). Other regulators of mTORC1 are raptor that positively regulates mTORC1 and functions as a scaffold for recruiting mTORC1 substrates (18), and PRAS40 and deptor that act as negative regulators (19, 20). The function of mLST8/GβL yet remains unknown (21).
Much less is known about mTORC2 functions, substrates, and regulators. Unlike mTORC1, which is a direct target of rapamycin, mTORC2 was initially described as rapamycin insensitive (22), although it has been recently reported that continued exposure to rapamycin also leads to its inhibition (23). mTORC2 promotes cell survival and actin cytoskeleton organization. It is exclusively growth-factor responsive and AKT is its first recognized substrate protein. Full activation of AKT requires the phosphorylation of two residues: Ser473 by mTORC2 and Thr308 by phosphoinositide-dependent kinase-1 (PDK1; ref. 24). Other mTORC2 substrates are serum- and glucocorticoid-induced protein kinase-1 (SGK1) and protein kinase C-alpha (PKCα; refs. 25, 26).
The regulatory mechanisms of mTORC2 also remain partially unknown, although it has been shown that rictor and mSin1 enhance mTORC2 signaling whereas deptor seems to negatively regulate it. TSC1 and TSC2 have also been involved in promoting mTORC2 activation. However, the function of PRR5/Protor is still not well defined (21, 27).
Finally, the mTOR complex has also been suggested to play a crucial role integrating extracellular and intracellular signals that regulates cellular metabolism. This also includes the control of inflammatory and tolerance responses via regulation of T cell receptor (TCR) TCRζ (28) and TGF-β–induced Foxp3, respectively (29). Although these physiologic functions need further mechanistic elucidation, they also open new avenues for development of biomarkers of mTOR inhibition through other alternative effects.
The mTOR Pathway: An Intricate Network with Feedback Loops
Development of resistance to mTORC1 inhibitors has been related with the presence of different feedback loops described within this complex network. Moreover, a better understanding of these mechanisms may help to identify novel therapeutic strategies to overcome the relative lack of efficacy of these compounds (5).
It is postulated that mTORC1 activation causes a negative feedback through S6K1 that reduce the activity of PI3K. The phosphorylation of S6K1 inactivates IRS-1, which is required for insulin signaling through PI3K (30). Therefore, mTOR inhibition will induce IRS-1 activation releasing the inhibition mediated by S6K1 and provoking the activation of AKT via an insulin growth-factor receptor 1 (IGF-1R) dependent signaling process (31). O'Reilly and colleagues published supporting evidence for this negative feedback loop. They have observed in a panel of cancer cell lines from different tumor types that rapamycin was able to upregulate IRS-1 levels and promote AKT phosphorylation (32). According to these findings, the biomarker study developed in the context of the first phase I clinical trial with everolimus (RAD-001, Afinitor; Novartis) showed a dose- and schedule-dependent inhibition of mTOR and a subsequent upregulation of AKT. These effects were observed in 50% of the patients and were assessed in both tumor and skin biopsies, thus validating the in vitro observation (33). Moreover, Wan and colleagues showed in human rhabdomyosarcoma cell lines and xenografts that blockade of IGF-1R led to an inhibition of the rapamycin-induced AKT activation (31), providing evidence for a synergistic effect of mTOR and IGF-1R inhibition. This combination is currently under clinical evaluation in a phase I multiple-dose escalating study using dalotuzumab, (a monoclonal antibody against IGF-1R; MK-0646; Merck) and ridaforolimus (an mTORC1 small-molecule inhibitor analog of the rapamycin; MK-8669, deforolimus; Merck and ARIAD). Preliminary results have revealed important antitumor activity in estrogen receptor-positive and highly proliferative breast tumors, which frequently harbor PIK3CA mutations and IGF-1R overexpression (34). Other two studies of the combination of cixutumumab (IGF-1R monoclonal antibody inhibitor; IMC-A12; ImClone) plus the rapalog temsirolimus (CCI-779, Torisel; Wyeth), and figitumumab (IGF-1R monoclonal antibody inhibitor; CP-751871; Pfizer) plus everolimus are underway (35, 36).
Furthermore, preclinical data have shown that mTORC1 inhibition results in a hyperactivation of the PI3K pathway and simultaneous increase of the signaling through the mitogen-activated protein kinase kinase (MAPK) pathway (37), thus proving the existence of another feedback loop that connects the PI3K-AKT-mTOR with the MAPK pathway. This observation has provided rationale for several ongoing phase I clinical trials combining mTOR, PI3K, or AKT inhibitors with MAP/ERK kinase (MEK) inhibitors. However, the most optimal combination of inhibitors deserves careful consideration due to dense cross-talk interactions among protein components of these complex pathways. Sophisticated systems biology analyses have recently predicted adverse effects in terms of reduction of cytotoxicity with the combination of a MEK and a first generation mTOR inhibitor. Specifically, in vitro validation of this in silico data showed that rapamycin, which led to significant activation of AKT when combined with a MEK inhibitor (U0126), rendered an increase in cell viability. In contrast, simultaneous inhibition of PI3K-AKT and MAPK pathways decreased cell viability and points toward this combination as the most optimal way to effectively inhibit both pathways (38). On the other side, clinical studies have reported significant toxicities in a phase I trial which is testing the combination of an AKT inhibitor and a MEK inhibitor. Considering these preclinical and clinical results in conjunction, the combination of PI3K or second generation mTOR inhibitors with MEK inhibitors warrants further clinical validation.
First Generation of mTOR Inhibitors
The first generation inhibitors of mTOR are derivatives of rapamycin that specifically inhibit mTORC1. This group of drugs is integrated by rapamycin and its analogs also known as rapalogs: everolimus, temsirolimus, and ridaforolimus (previously known as deforolimus). Rapamycin has been clinically approved several years ago for prophylaxis of organ rejection for renal transplant patients (Table 1 and Fig. 2; ref. 39).
The mechanism of action of rapamycin has been very well described. This drug along with the FK506-binding protein (FKBP12) targets the FKBP12-rapamycin–binding domain adjacent to the catalytic site of the mTOR protein (40). Several studies have shown that mTORC2 is rapamycin insensitive (22, 41), although long-term exposure to rapamycin can also inhibit mTORC2 and then disrupt AKT signaling. Strikingly, this response has been shown to be tissue specific (23).
mTORC1-mediated 4E-BP1 phosphorylation induces the dissociation of 4E-BP1 from the eukaryotic initiation factor 4E (eIF4E), thus allowing the assembly of the eIF4F complex to initiate cap-dependent mRNA translation. 4E-BP1 is phosphorylated at multiple sites such as Thr36, Thr45, Ser64, Thr69, and Ser82 and needs to occur in a prespecified order (42). The activation of Thr36 and Thr45 are the leading events necessary for phosphorylation of Thr69 that will be followed by Ser82 (43). Except for Ser82, all phosphorylation sites are sensitive to rapamycin as shown by the complete inhibition of the initiation of cap-dependent mRNA translation by the treatment with rapamycin in specific cellular and histologic contexts (44). However, it has been recently observed that rapalogs may not fully block 4E-BP1 despite the complete inhibition of S6K1 (45). This fact could be due to different reasons such as a relative lack of effect on the phosphorylation of Thr36 and Thr45 (46, 47), the existence of unknown feedback loops, and the inability to inhibit mTORC2; and it explains the unpredictable antitumor effect of rapalogs across different cancer subtypes (48–52). Mechanistic details are discussed in the section devoted to second generation inhibitors.
Despite their limited cytotoxic activity, rapalogs have shown antiproliferative properties. Temsirolimus and everolimus have been approved by the Food and Drug Administration (FDA) and the European Medicines Agency (EMEA) for treatment of advanced renal cell carcinoma; temsirolimus has been authorized for treatment of relapsed or refractory mantle-cell lymphoma by the EMEA only (Table 1 and Fig. 2). The approval of temsirolimus for treatment of previously untreated metastatic renal cell carcinoma was based on the results from a phase III clinical trial in which 626 patients randomly received temsirolimus, IFN-α, or combination therapy with temsirolimus and IFN-α. Temsirolimus alone rendered longer overall survival (10.9 vs. 7.3 months; HR = 0.73; P = 0.008) and progression-free survival than IFN-α alone (5.5 vs. 3.8 months; P < 0.001). In addition, no differences between the combination therapy and the IFN group were observed in terms of overall survival (53). After that, everolimus was approved for the treatment of patients with advanced renal cell carcinoma who had progressed on sorafenib, sunitinib, or both. The authorization was supported by the data coming from a phase III clinical trial that randomized 410 patients to receive everolimus or placebo in a 2:1 ratio. Everolimus showed a significant improvement in progression-free survival with mild adverse effects (4 vs. 1.9 months; HR = 0.30; P < 0.001; ref. 54). Finally, temsirolimus showed improvement in progression-free survival and higher objective response rates compared with investigator's choice treatment in patients with relapsed or refractory mantle-cell lymphoma leading to the approval by the EMEA (55).
Therefore, the next step in the development of rapalogs will be the discovery of new biomarkers to predict specific molecular features and tumor subtypes are more likely to respond to mTOR inhibitors. In this regard, responses to PI3K-AKT-mTOR pathway inhibitors may be higher among those tumors harboring PIK3CA mutations (56) and also those with loss of PTEN (57). Another example of response to rapalogs in specific tumor subtypes is the case of microsatellite instable colorectal cancers. PI3K-AKT-mTOR pathway has been involved in the pathogenesis of colorectal cancer. In fact, PIK3CA mutations have been identified in approximately 20% to 30% of colorectal tumors and have been associated with shorter cancer-specific survival, poorer outcomes, and resistance to cetuximab (1, 58, 59). Although single-agent everolimus has not achieved objective responses in refractory metastatic colorectal cancer (48), in vitro studies have suggested that colorectal tumors displaying microsatellite instability could potentially respond better to therapies against the PI3K-AKT-mTOR pathway (60). According to these results, dual PI3K-mTOR inhibitors may represent an interesting option to be evaluated in this specific tumor subtype.
Second Generation of mTOR Inhibitors
Whereas rapamycin exerts its action almost exclusively through mTORC1 inhibition, a second generation of inhibitors targeting the ATP site of the kinase domain of mTOR has been developed. These compounds are able to block both mTORC1 and mTORC2. Theoretically, their most important advantages would be a significant decrease of AKT phosphorylation on mTORC2 blockade and a better mTORC1 inhibition. In addition, the preclinical data of these agents have contributed to a better understanding of the functions of mTORC2 and the limitations of rapalogs.
Due to the fact that the catalytic domain of mTOR and the p110α subunit of PI3K are structurally related, some of these second generation compounds have dual activity against both PI3K and mTOR. These drugs, compared with single-specific mTORC1 and PI3K inhibitors, have the potential benefit of inhibiting mTORC1, mTORC2, and all the catalytic isoforms of PI3K (61). Therefore, targeting both kinases simultaneously should reduce the upregulation of PI3K that typically produced on inhibition of mTORC1 (30).
The majority of dual PI3K-mTOR inhibitors have already entered into phase I and II clinical trials alone or in combination with other agents for different cancer subtypes (Table 2). NVP-BEZ235 (Novartis) is one of these dual kinase inhibitors and reversibly blocks the p110α catalytic subunit of PI3K and mTOR (62). Initial in vitro data analyzing pharmacodynamic endpoints in breast tumor xenografts treated with NVP-BEZ235 have shown a decrease in phosphorylation levels of AKT, 4E-BP1, and S6K1 following treatment with this drug and higher antiproliferative activity than everolimus (63). A phase I of NVP-BEZ235 has been recently presented with promising efficacy. Among 51 evaluable and heavily pretreated patients, 14 achieved stable disease longer than 4 months and partial responses were observed in breast and lung tumors. However, pharmacokinetic studies showed that the area under the curve increased nonproportionally with dose, so future studies will use a new formulation of the drug. No dose-limiting toxicities were reported and the maximum tolerated has not been reached (64). On the contrary, XL-765 (Exelixis) has exhibited potent pharmacodynamic effects on the inhibition of PI3K with a stable pharmacokinetic profile. In addition, durable disease stabilizations were observed in patients with different tumor types such as colorectal cancer, lung cancer, renal cell carcinoma, mesothelioma, and appendiceal cancer (65).
Regarding single-specific mTOR catalytic inhibitors, several small molecules have also been identified (Table 2; Fig. 2), and 3 of them have entered into phase I clinical development [AZD-8055 (AstraZeneca), INK-128 (Intellikine), and OSI-027 (OSI Pharmaceuticals)]. Preclinical data with INK-128 have shown a potent inhibition of the phosphorylation of S6K1, 4E-BP1, and AKT at Ser473 in vitro, and important antiproliferative activity against multiple xenograft models and cells lines resistant to rapamycin and pan-PI3K inhibitors (66). At the same time Feldman and colleagues have reported the activity of two compounds PP-242 and PP-30 (University of California) with activity against both mTORC1 and mTORC2. These compounds are able to completely suppress 4E-BP1 and S6K1 along with a reduction of phosphorylation of AKT at Ser473, thus leading to a higher antiproliferative effect compared with rapamycin. However, the inhibition of mTORC2 did not result in a total blockade of AKT, suggesting that additional mTORC1 inhibition by these compounds could be the basis for their superior antitumor activity (67). In this regard, Hsieh and colleagues suggested that the therapeutic benefit of PP-242 is mediated through the inhibition of mTORC1-dependent 4E-BP1-eIF4E hyperactivation (68). Other preclinical studies with these ATP-competitive and ATP-specific mTOR inhibitors have observed similar results and have confirmed its activity over those rapamycin-resistant functions of mTORC1. In addition, these drugs induce a stronger G1 cell cycle arrest in several cancer lines and formidable autophagy activation (69–73). Finally, a first-in-human phase I study exploring three schedules of OSI-027 has been recently presented with preliminary evidence of pharmacologic activity. The maximum tolerated dose has not yet been defined and dose escalation is ongoing. Left ventricular ejection fraction and fatigue have been reported as dose-limiting toxicities (74). In the following years, we will obtain more detailed data from phase I studies about the pharmacokinetic profile, optimal dose, toxicity, and preliminary activity of all of these compounds.
mTOR is one of the signaling pathways that has attracted more interest among basic and clinical researchers. Two main factors are responsible for this phenomenon: mTOR is a downstream central effector of multiple pathways thus making it a very attractive target, and the drug rapamycin which renders an incomplete inhibition of this protein complex became available in 1975. These facts have fostered the efforts of the pharmaceutical industry to synthesize newer and better compounds against it. In a relatively short period of time, several companies have launched development programs of different drugs blocking the same target, including clinical trials to examine the activity of these compounds in solid and hematologic malignancies. In parallel, basic scientists continued exploring and trying to fill the gaps in the knowledge of the molecular biology of this pathway. At some point, biomarkers studies and clinical trials were developed without having a final clear portrait of the biology behind mTOR. Therefore, several unexpected and initially unexplainable results came back as a consequence of these studies.
Initial disappointment about preliminary clinical results decreased the excitement for targeting mTOR. It was later known that the mTOR pathway is almost a duality constituted by two complexes with different functions and many feedback loops, thus changing the original simplistic view of it. Now, a second generation of smarter compounds developed taking into account the latest biologic data is currently being developed. On one side, these compounds are able to inhibit both mTORC1 and mTORC2, and on the other side also incorporate activity against PI3K. Initial data from phase I clinical trials with these drugs have recently shown significant clinical activity, particularly in patients with deregulation of the PI3K-AKT-mTOR pathway.
Therefore, it is important to learn the lessons from the development of rapamycin and rapalogs. A complete understanding of the molecular biology of the pathway and its actors is needed to appropriately develop its targeted drugs and to correctly interpret the results from clinical studies. Finally, identification of biomarkers on the basis of genetic, genomic, and systems biology approaches will allow defining what tumor subtypes may derive in a higher benefit with the use of mTOR inhibitors. These studies should be run parallel to early clinical trials, thus accelerating their implementation into phase III trials. Then, biomarkers will be validated and ready to be approved simultaneously with drug indication.
Disclosure of Potential Conflict of Interest
No potential conflicts of interest were disclosed.
This work was supported in part by University of Michigan Comprehensive Cancer Center Core Support grant (NIH 5P30CA46592) and Michigan Institute for Clinical and Research Health (UL1RR024986).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received September 28, 2010.
- Revision received December 23, 2010.
- Accepted December 28, 2010.
- ©2011 American Association for Cancer Research.