Mechanistic target of rapamycin (mTOR) is the kinase subunit of two structurally and functionally distinct large multiprotein complexes, referred to as mTOR complex 1 (mTORC1) and mTORC2. mTORC1 and mTORC2 play key physiological roles as they control anabolic and catabolic processes in response to external cues in a variety of tissues and organs. However, mTORC1 and mTORC2 activities are deregulated in widespread human diseases, including cancer. Cancer cells take advantage of mTOR oncogenic signaling to drive their proliferation, survival, metabolic transformation, and metastatic potential. Therefore, mTOR lends itself very well as a therapeutic target for innovative cancer treatment. mTOR was initially identified as the target of the antibiotic rapamycin that displayed remarkable antitumor activity in vitro. Promising preclinical studies using rapamycin and its derivatives (rapalogs) demonstrated efficacy in many human cancer types, hence supporting the launch of numerous clinical trials aimed to evaluate the real effectiveness of mTOR-targeted therapies. However, rapamycin and rapalogs have shown very limited activity in most clinical contexts, also when combined with other drugs. Thus, novel classes of mTOR inhibitors with a stronger antineoplastic potency have been developed. Nevertheless, emerging clinical data suggest that also these novel mTOR-targeting drugs may have a weak antitumor activity. Here, we summarize the current status of available mTOR inhibitors and highlight the most relevant results from both preclinical and clinical studies that have provided valuable insights into both their efficacy and failure.
Mechanistic (formerly mammalian) target of rapamycin (mTOR) is the serine/threonine kinase subunit of two large and functionally distinct multiprotein complexes referred to as mTOR complex 1 (mTORC1) and mTORC2. These two complexes are highly conserved among eukaryotes  and control several anabolic and catabolic processes in response to environmental cues in organs/tissues including: liver, skeletal muscle, white and brown adipose tissue, lymphocytes, and the brain. Therefore, mTORC1 and mTORC2 are implicated in many physiological functions . However, it is now clear that their activities are deregulated in widespread human diseases such as: cancer , obesity , type 2 diabetes [5,6], epilepsy , and chronic neurodegenerative disorders , as well as during aging .
mTOR was discovered almost simultaneously by three independent groups in the mid-1990s and named rapamycin and FK506-binding protein-12 (FKBP-12) target 1 (RAFT1) , FKBP–rapamycin associated protein (FRAP) , and mTOR .
These names reflected the fact that mTOR was identified as the target of rapamycin, a macrolide antibiotic purified 20 years earlier from bacteria found in the soil collected on Easter Island (Rapa Nui in the local language) . Even prior to the discovery of mTOR, it had become apparent that rapamycin possessed remarkable immunosuppressant, anticancer, and antifungal properties, which were linked with its capacity to delay G1 phase progression of the cell cycle [14–16]. These effects were subsequently related to the inhibition of mTOR activity exerted by rapamycin.
Over the last 15 years, there has been an explosion of interest in the field of mTOR inhibition, as this therapeutic strategy has displayed a remarkable efficacy in preclinical models of several disorders, especially in cancer. Pharmaceutical companies have discovered impressive arrays of small molecules targeting mTOR and at present there are more than 400 registered trials (clinicaltrials.gov) testing the different classes of mTOR inhibitors as antitumor therapeutics. However, mTOR inhibition in cancer patients, especially as monotherapy, has generated extremely limited benefits, increasing survival by just a few months . In this review, after highlighting the regulation of mTORC1 and mTORC2 as well as the roles they play in the cell, we will outline the different classes of mTOR inhibitors currently available and give relevant findings from preclinical and clinical studies that have provided key insights into both their efficacy and failure.
mTOR, a 289-kDa protein encoded in humans by the MTOR gene mapping to chromosomal band 1p36.2 , is a member of the phosphatidylinositol 3-kinase-related kinase (PIKK) family of protein kinases . mTOR is composed of several domains (Figure 1A,B). The N-terminal portion of mTOR contains at least 20 Huntingtin, Elongation factor 3, a subunit of protein phosphatase-2A, and TOR1 (HEAT) repeats. The HEAT repeats constitute most of the N-terminal half of the protein and bind some critical mTOR regulators, including the regulatory-associated protein of TOR (Raptor), rapamycin-insensitive companion of TOR (Rictor), and the Tel2-interacting protein 1 (Tti1)/Tel2 complex, which is important for assembly and stability of both mTOR complexes .
Domains of mTOR and interactors of mTORC1 (A) and mTORC2 (B)
The FKBP-associated protein, ataxia-telengiectasia mutated (ATM), transactivation/transformation domain-associated protein (TRRAP), or FAT domain binds the mTOR regulatory subunit DEP domain-containing mTOR-interacting protein (Deptor) . The FKBP, rapamycin-binding (FRB) domain is the docking site for the FKBP-12-rapamycin complex. The kinase domain interacts with the scaffolding protein mammalian lethal with SEC13 protein 8 (mLST8) , while the FRAP, ATM, TRRAP, Carboxy terminal (FATC) domain is likely involved in substrate recognition .
Importantly, several mutation hotspots have been identified in the FAT, FATC, and kinase domains of mTOR. As we will see in this review, these mutations result in increased activity of mTORC1 and could predict sensitivity of neoplastic cells to rapamycin and its derivatives .
Apart from components that are common to mTORC2 (Tti1/Tel2 complex, Deptor, and mLST8), mTORC1 is defined by the association of mTOR with Raptor, which is fundamental for mTORC1 assembly, stability, substrate specificity and regulation , and with Proline-rich Akt substrate 1 40 kDa (PRAS40), a protein which blocks mTORC1 activity until growth factor receptor signaling unlocks PRAS40-mediated mTORC1 inhibition  (Figure 1A).
Activation of mTORC1 is achieved by growth factors, amino acids, and ATP levels through multiple mechanisms (Figure 2). Growth factors, such as insulin or insulin-like growth factor-1 (IGF-1), activate the phosphatidylinositol 3-kinase (PI3K). PI3K generates at the plasma membrane phosphatidylinositol 3,4,5 trisphosphate (PIP3) from phosphatidylinositol 4,5 bisphosphate (PIP2). PIP3 recruits to the plasma membrane phosphoinositide-dependent kinase 1 (PDK1) and Akt that then is phosphorylated by PDK1 at Thr308 .
Regulation and functions of mTORC1 and mTORC2
This phosphorylation event is sufficient for Akt to phosphorylate tuberous sclerosis complex 2 (TSC2) at Thr1462 . TSC2 is a GTP-ase activating protein (GAP) that functions in association with TSC1 to lock the small G-protein, Ras homolog enriched in brain (Rheb). Akt-driven TSC1/TSC2 complex inactivation allows Rheb to accumulate in a GTP-bound state. Rheb–GTP then can bind and activate mTORC1 . Moreover, Akt phosphorylates the mTORC1 inhibitor PRAS40 at Thr246. This phosphorylation causes PRAS40 to dissociate from Raptor, allowing mTORC1 activation . Also, the Rat sarcoma (Ras)/Raf/mitogen-activated protein kinase kinase (MEK)/extracellular signal-regulated kinase (ERK)/p90 ribosomal S6 kinase 1 (p90RSK1) module can activate mTORC1, as both ERK and p90RSK1 phosphorylate TSC2 (at Ser664 and Ser1798, respectively), thus inhibiting the TSC1/2 complex and triggering Rheb-dependent mTORC1 activation [30,31]. Furthermore, p90RSK1 can phosphorylate PRAS40 at Ser719/721/722, causing its dissociation from Raptor and promoting activation of mTORC1 .
Given that mTORC1 activates protein synthesis, it is of the outmost importance that cells do not proceed if the necessary nutrients (amino acids) and energy (ATP) are not available for supporting the high requirements of translation. Hence, a lack of amino acids prevents mTORC1 localization to lysosomal surfaces where Rheb activates mTORC1, thus overriding growth factor receptor-elicited proliferative signaling [33–38]. The Ragulator complex is a pentameric protein complex consisting of: p18, p14, the mitogen-activated protein complex kinase (MAPK) scaffold protein (MP1), the hepatitis B X-interacting protein (HBXIP), and C7orf57. In addition, the folliculin complex, Rag GTP-ases, and GAP activity toward Rags (GATOR) complex 1 and 2 are central to the amino acid-dependent regulation of mTORC1 activity. These complexes interact to finely tune mTORC1 activity . Under low amino acid conditions, Ragulator is found in an inhibitory state with the lysosomal v-ATPase, while GATOR1 exerts its GAP activity toward RagA, whereby keeping this GTPase in the inactive GDP-bound state that is not sufficient to recruit mTORC1. Upon amino acid stimulation, GATOR1 is likely inhibited by GATOR2, while Ragulator and v-ATPase undergo a conformational change that unleashes the GEF activity of Ragulator toward RagA, while the folliculin complex promotes RagC GTP hydrolysis. The active heterodimer, consisting of GTP-bound RagA and GDP-loaded RagC, is free to recruit mTORC1 to the lysosomal surface, where it interacts with and is activated by Rheb [40–42].
As to ATP, a low ATP/AMP ratio up-regulates the liver kinase B1 (LKB1)/AMP-activated protein kinase (AMPK) axis, which functions as an indirect mTORC1 inhibitor by promoting TSC1/2 complex formation, thereby increasing its GAP activity toward Rheb . This is achieved through AMPK-dependent TSC2 phosphorylation at Thr1227 and Ser1345 . Moreover, AMPK directly phosphorylates Raptor, leading to its inactivation through binding to 14-3-3 proteins with a consequent allosteric inhibition of mTORC1 . Therefore, high AMP levels can override growth factor signaling and block cell proliferation through mTORC1 in the absence of a sufficient amount of intracellular ATP .
Regarding other mechanisms that down-regulate mTORC1 activity, it should be emphasized that low oxygen levels can induce the expression of regulated in DNA damage and development 1 (REDD1) that inhibits mTORC1 function through TSC1/TSC2 . mTORC1 activity is also blunted following DNA damage, as REDD1 is encoded by one of the genes controlled by p63 , whereas p53 controls the expression of sestrins 1 and 2 that impinge on AMPK [49,50].
Once activated, mTORC1 controls several anabolic cell functions. Through phosphorylation of ribosomal protein p70 S6 kinase 1 (p70S6K1)  and eukaryotic translation initiation factor 4E-binding protein 1 (4E-BP1) , mTORC1 increases both cap-dependent and cap-independent translation . In particular, by phosphorylating 4E-BP1 on several residues (Thr37/46/70 and Ser 65), mTORC1 prevents the inhibitory action of eukaryotic translation initiation factor 4E (eIF4E) to allow the latter to initiate cap-dependent translation . Interestingly, it has been shown that the most common targets of mTORC1-dependent translation are those involved in survival, proliferation, invasion, metastasis, and ribosome biogenesis, evidence that supports the oncogenic role of mTORC1 [57–60]. As to p70S6K1, it controls protein synthesis through S6 ribosomal protein (S6RP) and eIF4B .
Growing cells, apart from increased protein requirements, need enough lipids for new membrane formation and expansion. Accordingly, mTORC1 increases de novo lipid synthesis through the sterol responsive element binding protein (SREBP) transcription factors, which regulate the expression of genes involved in both fatty acid and cholesterol biosynthesis . Furthermore, mTORC1 up-regulates the synthesis of nucleotides required for DNA replication [63,64] and shifts glucose metabolism from oxidative phosphorylation to glycolysis by increasing translation of the hypoxia-inducible factor 1α (HIF1α) transcription factor that then drives the expression of several glycolytic enzymes .
Last but not the least, mTORC1 promotes cell growth by suppressing protein catabolism. On one side, mTORC1 inhibits autophagy, the central degradative and recycling process of cellular components. This is achieved through the phosphorylation of Unc-51 like kinase 1 (ULK1), which drives autophagosome formation [66,67]. On the other side, recent findings has provided evidence that mTORC1 is a repressor of the ubiquitin-proteasome system, the second major pathway responsible for the degradation of long-lived cell proteins . A schematic cartoon of mTORC1 regulations/functions is presented in Figure 2. Because of all of the aforementioned effects, mTORC1 is key factor in cell growth, proliferation, and metabolism.
This complex is characterized by the interactions of mTOR with Rictor, mammalian stress-activated protein kinase interacting protein 1 (mSin1), and protein observed with Rictor (Protor) 1 or 2 (Figure 1B). Rictor is necessary for mTORC2 assembly, stability, and substrate interactions . mSin1 acts as a negative regulator of mTORC2 kinase activity . However, it also drives mTORC2 localization to the plasma membrane, where Sin1-mediated mTORC2 inhibition is relieved in response to growth factor receptor-dependent PI3K activation . As to Protor-1, it may play a role in enabling mTORC2 to efficiently phosphorylate serum and glucocorticoid-activated kinase 1 (SGK1) .
In contrast, the control of mTORC2 activity is not as well understood as the regulation of mTORC1 activity. However, plasma membrane localization is a key aspect of mTORC2 regulation. Indeed, it has been recently shown that the pleckstrin homology (PH) domain of mSin1 interacts with the mTOR kinase domain to restrain mTOR activity. PIP3, generated by PI3K at the cell membrane, binds mSin1-PH to release its inhibition on mTOR, thereby triggering mTORC2 activation . However, ribosomes could be involved also in mTORC2 activation in a PI3K-dependent manner .
As to the functions of mTORC2, this complex phosphorylates several members of the AGC family of protein kinases . These include protein kinase C (PKC) α, PKCδ, PKCγ, PKCε, and SGK1 . However, probably the most important and best known function of mTORC2 is phosphorylation of Akt at Ser473, which fully activates the kinase activity of Akt . Importantly, Ser473 p-Akt is absolutely required for phosphorylation of Forkhead box O 1/3a (FoxO1/3a) transcription factors, but not for that of other Akt targets, such as TSC2 and glycogen synthase kinase 3β (GSKβ3) [76–78]. A schematic cartoon of mTORC2 regulation/functions is presented in Figure 2.
By targeting AGC kinases, mTORC2 is mainly involved in the control of cell proliferation, survival, and migration, as well as of various aspects of cytoskeletal remodeling. Nevertheless, recent findings have highlighted how mTORC2 is also a repressor of chaperone-mediated autophagy , which is frequently deregulated in numerous age-related disorders, including Parkinson’s and Alzheimer’s diseases . Moreover, mTORC2 increases lipid synthesis, whereby promoting tumorigenesis . Therefore, it seems that some of the mTORC1 functions are regulated by mTORC2 as well.
Feedback mechanisms regulate the activity of both mTORC1 and mTORC2
The activity of the two complexes is finely and mutually tuned through some feedback circuits. Both mTORC1 and its substrate p70S6K1 provide a negative feedback to the insulin and IGF-1 signaling pathways through inhibitory serine phosphorylation of the insulin receptor substrate (IRS) 1 and 2, which target these adapter proteins for proteasomal degradation [82,83]. Hence, insulin-dependent, IRS-induced signaling is terminated, PI3K/Akt signals are inhibited and mTORC1 activity is restrained by such a negative feedback loop [84,85]. Another mTORC1 substrate that negatively impinges on PI3K/Akt is growth factor receptor-bound protein 10 (Grb10) . Indeed, Gbr10, which is directly activated and stabilized by mTORC1, interacts with IRS2, targeting it for ubiquitination and, most likely, proteasomal degradation  (Figure 3).
Feedback loops controlling the activity of mTORC1 and mTORC2
As to mTORC2, phosphorylation of mSin1 at Thr86 by Akt may regulate mTORC2 kinase activity, leading to a growth factor-dependent positive feedback loop that sustains mTORC2–Akt signaling . In contrast, mSin1 phosphorylation by p70S6K1 at both Thr86 and Thr389 residues may inhibit mTORC2 activity by dissociating mSin1 from mTORC2, thus providing a negative feedback mechanism downstream of mTORC1 in response to a variety of growth factors important for tumor cell growth, including insulin, IGF-1, platelet-derived growth factor (PDGF), and epidermal growth factor (EGF)  (Figure 3). Interestingly, it has been demonstrated that a Sin1 mutation (R81T), which had been previously identified in ovarian cancer patients, led to a nearly complete abolition of Thr86 p-Sin1 levels and to more sustained Ser473 p-Akt levels when expressed in OVCAR5 cells, in response to growth factor stimulation. Mechanistically, the R81T mutation disrupts the canonical the p70SK1 phosphorylation motif of Sin1 . Therefore, these findings highlighted the relevance of Sin1–R81T mutation in cancer patients, as it bypasses the mTORC1/p70S6K1-mediated negative feedback regulation on mTORC2/Akt activation.
mTOR was originally discovered as the target of the antibiotic rapamycin. We will now review the main classes of mTOR inhibitors: allosteric mTOR inhibitors which include rapamycin and its derivatives or rapalogs (ABT-578/zotarolimus; AP23573/ridaforolimus; RAD001/everolimus, CCI-779/temsirolimus); ATP-competive dual PI3K/mTOR inhibitors; ATP-competive mTOR kinase inhibitors (TORKIs); the recently described RapaLink-1, a molecule which combines the properties of rapamycin with those of TORKIs . In this review, we will mainly focus on the use of mTOR inhibitors as anticancer targeted agents . However, it should not be overlooked that some of these drugs are registered for treating other disorders. Rapamycin/rapalogs are used for preventing rejection of transplanted solid organs [91,92] and are widely and successfully employed in drug-eluting stents for treatment of coronary artery disease [93,94]. mTOR inhibitors are being considered for allogeneic hematopoietic stem cell transplantation to prevent and treat graft‐versus‐host disease graft-versus-host disease , rheumatoid arthritis , atherosclerosis , and a wide spectrum of neurological disorders, including Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, autism, dementia, epilepsy, stroke, and traumatic brain injury [98,99]. Furthermore, there is a growing interest in mTOR inhibition for preventing the detrimental effects of aging [100,101].
Allosteric mTOR inhibitors
The mechanisms by which these drugs inhibit mTOR is through the formation of a complex with the small immunophillin FKBP-12, with the resulting dimer binding the FRB domain of mTOR  (Figure 1A). This binding results in decreased interactions between mTOR and Raptor with a consequent down-regulation of mTORC1 activity . Not all of mTORC1 downstream targets are inhibited with equal potency by rapamycin/rapalogs. In particular, 4E-BP1 phosphorylation at Thr37/46 is usually resistant to treatment, with important consequences on cap-dependent translation [57,104,105]. Also, ULK1 phosphorylation at Ser758 is largely resistant to rapamycin ; hence, this class of molecules is considered as incomplete mTORC1 inhibitors.
The FRB domain of mTOR is exposed in mTORC1 but not in mTORC2, as Rictor blocks rapamycin–FKBP-12 access to the FRB domain in mTORC1 . Therefore, these drugs were initially thought to be mTORC1-selective therapeutic agents. However, there are studies that revealed how chronic treatment with rapamycin resulted in mTORC2 inhibition, by affecting the assembly of new mTORC2 complexes . Thus, this class of drugs has the potential for inhibiting Akt phosphorylation at Ser473 both in vitro and in vivo [109,110], although this phenomenon could be cell- and/or context-dependent .
As further evidence that rapamycin inhibits mTORC2 in vivo, it should be considered that a metabolic adverse event frequently observed in cancer patients chronically treated with rapalogs is hyperglycemia . In this connection, it should be highlighted that rapamycin impaired in vitro basal and insulin-stimulated glucose uptake in adipocytes from human donors, while decreasing formation of both mTORC1 and mTORC2 . Moreover, it has been demonstrated that rapamycin disassembled mTORC2 also in vivo and this prevented insulin-mediated suppression of hepatic gluconeogenesis .
Rapamycin/rapalogs have some serious drawbacks in terms of the desired molecular effects. By inhibiting p70S6K1 phosphorylation, they switch off the IRS-dependent negative feedback loop that prevents overactivation of PI3K/Akt signaling in response to insulin/IGF-1 [114–116]. This results in Akt/mTORC1 activity up-regulation which, importantly, has been reported to occur also in vivo in cancer patients treated with rapamycin [117,118]. Interestingly, loss of the activity of phosphatase and tensin deleted on chromosome 10 (PTEN) tumor suppressor, which dephosphorylates PIP3 to PIP2, thereby acting as a negative regulator of PI3K/Akt/mTORC1 signaling  (Figure 2), facilitated Akt hyperphosphorylation induced by rapamycin in bladder cancer cells. Remarkably, transitional cell carcinoma patients with a shorter progression-free survival under everolimus exhibited PTEN deficiency and increased Akt activation . In this context, it is important to highlight here that PTEN loss or functional inactivation is a very common event in a wide variety of human cancers . Inhibition of p70S6K1 by rapamycin could also positively impact on mTORC2, as Rictor is acutely phosphorylated in a growth factor-dependent manner by p70S6K1 at Thr1115 (Figure 3). This phosphorylation event stimulates binding of Rictor to 14-3-3 proteins that inactivate it . Although mTORC2 integrity or in vitro kinase activity were not affected by phosphorylation at Thr1115, expression of a phosphorylation site mutant of Rictor (T1135A) in either wild-type or Rictor-null cells caused an increase in the mTORC2-dependent Ser473 Akt phosphorylation .
In response to rapamycin treatment of both healthy and neoplastic cells, a compensatory activation of MEK/ERK signaling, which also impinges on mTORC1, is frequently observed. Such a phenomenon partly depends on the existence of a p70S6K1/IRS/Ras/Raf negative feedback loop control mechanism which is interrupted by rapamycin  (Figure 3), although other mechanisms have been more recently proposed, including one which involves Grb10 [124,125]. Interestingly, a comprehensive meta-analysis of published microarray data revealed that Grb10 expression was decreased in many tumor types compared with that in normal tissue counterparts, while GGrb10 knockdown cells displayed a lower apoptotic response to either staurosporine or etoposide. These findings suggested that Grb10 could indeed function as a tumor suppressor also because of its capacity to switch-off insulin signaling . However, it is now clear that PI3K/Akt/mTOR and Ras/Raf/MEK/ERK pathways are linked through multiple mechanisms (reviewed in ), therefore each pathway can inhibit or activate the other also depending on the genetic background of cancer cells [127,128]. As to the relevance of MEK/ERK overactivation, it should for example be considered that ERK activates mitogen-activated protein kinase (MAPK)-interacting kinase 1 (Mnk1), which phosphorylates eIF4E to provide a distinct signal to up-regulate cap-dependent mRNA translation  (Figure 2). Therefore, rapamycin treatment of cancer cells could lead to increased levels of oncogenic protein synthesis through the ERK/Mnk1 axis .
Compensatory signaling events that follow rapamycin treatment are not limited to PI3K/Akt and MEK/ERK overactivation, but, additionally, include up-regulation of several receptor tyrosine kinases (RTKs), which include IGF-1 receptor (IGF-1R), insulin receptor, and human EGF receptor 2 (HER2). Evidence suggests that overexpression of these RTKs may be dependent on up-regulated mTORC2/FoxO signaling elicited by rapamycin/rapalogs (see above and ).
Furthermore, it has emerged that FoxO transcription factors up-regulated the expression of Rictor, thereby creating an amplification loops which enhanced mTORC2 activity and Ser473 p-Akt levels . Accordingly, genetic downmodulation of Rictor through RNA interference diminished rapamycin-induced overexpression of IGF-IRβ mRNA and protein in gastrointestinal and pancreatic cancer cells .
Since mTORC1 is a repressor of autophagy, rapamycin and rapalogs usually promote autophagy which is a double-edged sword phenomenon, as in some cases it could protect cancer cells from mTORC1-targeted therapy whereas in others it is a death mechanism for neoplastic cells [134,135]. Preclinical studies indeed demonstrated that protective autophagy occurs in vitro and in vivo when mTOR inhibitors were used as antineoplastic therapeutics [136–140]. The first clinical trial in which the autophagy inhibitor hydroxychloroquine was combined with temsirolimus in melanoma patients demonstrated that the drug combination was safe and tolerable, and showed significant antitumor activity . Other studies have been subsequently published, however in these small scale trials rapamycin was used for mTOR inhibition, in combination with hydroxychloroquine [142,143]. Overall, these very preliminary findings have indicated further studies combining mTOR and autophagy inhibitors in cancer patients may be warranted. Nevertheless, critical steps in this direction will be represented by the development of more efficacious and selective autophagy inhibitors [144,145], as well as by the identification of reliable markers that will predict whether in a given neoplastic context autophagy could play either a protective or antitumor role .
Rapamycin/rapalogs as anticancer agents
Rapamycin was initially approved for use by the U.S. Food and Drug Administration (FDA) as an immunosuppressive agent to help prevent organ rejection in patients 13 years and older undergoing kidney transplantation, nevertheless its poor solubility in aqueous solutions prevented it to be fully developed as an anticancer agent. Although ridaforolimus is still being tested in clinical trials involving cancer patients with some encouraging results [147–149], its use has not been approved so far by either the FDA or the European Medicines Agency (EMA).
Everolimus and temsirolimus display improved pharmacokinetic and pharmacodymanic properties compared with rapamycin, as they are more stable and water soluble. While everolimus is orally biovalable, temsirolimus, a prodrug of rapamycin, is available as an injectable preparation.
Both everolimus and temsirolimus were approved by the FDA and the EMA for therapy of some types of cancer, which include advanced renal cell carcinoma (RCC) ; hormone receptor-positive, HER2-negative breast cancer in postmenopausal women [151–153]; pancreatic and other selected neuroendocrine tumors ; adult renal angiomyolipoma associated with tuberous sclerosis complex (TSC) disease ; pediatric or adult subependymal giant cell astrocytoma (SEGA) with TSC ; and relapsed/refractory mantle cell lymphoma (MCL) .
The rationale for targeting mTORC1 in both hereditary and sporadic RCC is that patients frequently display mutations in VHL gene which encodes the von Hippel-Lindau (VHL) tumor suppressor . This tumor suppressor, among other functions, targets HIF-1α for poly-ubiquitination and degradation. The impaired function of VHL results in increased transcription of hypoxia-inducible genes, leading to overproduction of vascular endothelial growth factor (VEGF), PDGF-β, transforming growth factor-α (TGF-α) and erythropoietin, thereby promoting growth of highly vascular tumors . In consideration of the critical role played by mTORC1 in HIF-1α transcription control, clinical studies on the efficacy of rapalogs in RCC began nearly 15 years ago.
The first rapalog to be evaluated was temsirolimus, which, when employed as monotherapy in refractory patients, displayed antitumor activity and encouraging survival and was generally well tolerated . In a subsequent phase 3 trial, 626 patients with previously untreated RCC were randomly assigned to be treated with intravenous temsirolimus, or subcutaneous interferon α, or combination therapy consisting of temsirolimus plus interferon α. Patients who received temsirolimus alone had longer overall survival (OS, median time 10.9 months) and progression-free survival (PFS) than did patients who received interferon α alone. OS in the combination-therapy group (median 8.4 months) did not differ significantly from that in the interferon α group (median 7.3 months) .
The RECORD (REnal cell Cancer treatment with Oral RAD-001) series of clinical trials analyzed the efficacy of everolimus in patients with advanced RCC.
In the first study, RECORD-1, everolimus monotherapy in previously treated patients was associated with prolonged PFS compared with placebo (4.6 months vs. 1.8 months). Nevertheless, this prolonged PFS did not translate into improved survival, likely because of patient crossover upon disease progression . In addition, overall response rates (ORRs) were consistently low, ranging from 1% to 5% of treated patients [150,162,163].
The phase 2 RECORD-2 trial compared efficacy and safety of first-line everolimus plus the humanized anti-VEGF monoclonal antibody bevacizumab (EVE/BEV) with interferon α-2a plus bevacizumab (IFN/BEV). However, both the efficacy and the safety of the two treatments appeared similar, with the exception of proteinuria which was detected in about one-fourth of the EVE/BEV cohort .
The aim of the phase 2 RECORD-3 study was the comparison of the sequence consisting of everolimus followed by multi-targeted RTK inhibitor sunitinib (EVE-SUN) at progression with the opposite (standard) sequence (SUN-EVE), as front-line therapy. No significant differences between the two sequences were observed regarding median combined PSF, which was 21.7 months in case of EVE-SUN and 22.2 months with SUN-EVE. In contrast, median OS was 22.4 months for EVE-SUN and 29.5 months for SUN-EVE, which supported the use of the standard sequence of treatment .
The Breast cancer trials of OraL EveROlimus (BOLERO) series of studies investigated the efficacy of everolimus in patients with advanced hormone receptor-positive, HER2-negative breast cancer previously treated with aromatase inhibitors. Indeed, it had been previously observed that resistance to endocrine therapy in breast cancer was associated with activation of mTORC1 signaling .
In the BOLERO-2 study, everolimus plus the aromatase inhibitor exemestane was associated with a higher ORR versus placebo plus exemestane (9.5% vs. 0.5%, P<0.001) . Adding everolimus to exemestane did not confer a statistically significant improvement in the secondary end point, OS, despite producing a clinically meaningful and statistically significant improvement in the primary end point, PFS (4.6-months prolongation in median PFS) .
In the BOLERO-3 randomized, double-blind, placebo-controlled phase 3 clinical trial, patients with HER2-positive, trastuzumab-resistant, advanced breast carcinoma who had previously received taxane therapy were treated with daily everolimus plus weekly trastuzumab and the chemotherapeutic vinorelbine, in 3-week cycles . The addition of everolimus significantly prolonged PFS compared with placebo. However, it should be highlighted that median PFS was 7.00 months with everolimus and 5.78 months with placebo .
The RAD-001 in advanced neuroendocrine tumor (RADIANT) series of clinical trials evaluated the efficacy of everolimus for treating patients suffering from gastropancreatic neuroendocrine tumors (NETs). NETs are frequently characterized by PTEN and TSC2 down-regulation that positively impact on mTORC1 activity .
The phase 2 RADIANT-1 study, documented that everolimus, with or without concomitant octeotride, displayed antitumor activity as measured by objective response rate and PFS and was well tolerated in patients with advanced pancreatic NETs after failure of prior systemic chemotherapy . RADIANT-2 was a randomized, placebo-controlled, phase 3 trial that demonstrated how the combination everolimus plus octeotride significantly increased median PFS with respect to placebo plus octeotride (16.4 months vs. 11.3 months) .
RADIANT-3 was a prospective, randomized, phase 3 trial where everolimus was evaluated as monotherapy for advanced, low-grade or intermediate-grade pancreatic NETs. Data analysis showed that median PFS was 11.0 months with everolimus as compared with 4.6 months with placebo . RADIANT-4 phase 3 study was aimed to assess the efficacy and safety of everolimus compared with placebo in 302 patients with advanced, progressive NETs of either the lung or the gastrointestinal tract. Median PFS was 11.0 months in the everolimus group and 3.9 months in the placebo group .
TSC is a disease characterized by the growth of benign tumors (hamartomas) in multiple organs, including the brain, kidneys, skin, lungs, and heart . Mutations in either TSC1 or TSC2 result in inappropriate mTORC1 signaling and this is considered to account for many of the TSC features .
Everolimus efficacy was evaluated in TSC patients with SEGA. The first trial was a prospective, open-label study where all 28 patients demonstrated either a reduction in tumor volume or cessation of growth. Overall, about 80% of patients had SEGAs that were reduced by a third, and more than 30% of patients had SEGAs that were reduced by more than 50% within 6 months of treatment. Side effects were largely mild to moderate in severity, and none led to treatment discontinuation . Further analysis after continuous treatment (median of 3 years) demonstrated that efficacy was sustained without encountering new or additional significant adverse effects . The second study (named EXamining everolimus In a Study of Tuberous sclerosis complex-1 or EXIST-1) was a randomized, placebo-controlled, double-blind, phase 3 trial involving 117 patients with SEGAs. EXIST-1 came to similar conclusions as Krueger and co-workers  as SEGA volume was reduced by more than 50% in 49% of patients treated for a median of 29 months [175, 176]. Subsequently, the EXIST-2 trial demonstrated the efficacy of everolimus in reducing also renal angiomyolipoma lesion volume in TSC patients with an acceptable safety profile . These studies led to approval of everolimus for treating both SEGA and angiomyolipoma in TSC patients . More recently, the EXIST-3 trial has demonstrated that adjunctive everolimus treatment significantly reduced seizure frequency with a tolerable safety profile compared with placebo in patients with TSC and treatment-resistant seizures .
As to relapsed/refractory MCL, in a phase 3 study 162 patients were randomly assigned to receive either intravenous temsirolimus or investigator’s choice therapy (mainly chemotherapy) from prospectively approved options. Temsirolimus 175 mg weekly for 3 weeks followed by 75 mg weekly significantly improved PFS and objective response rate compared with investigator’s choice therapy. These results led to EMA approval for temsirolimus treatment of relapsed/refractory MCL patients.
Overall, the findings that emerged from all the aforementioned clinical trials have clearly established that the benefits of rapalog treatment to sporadic cancer patients, although statistically significant with respect to control, are for the most extremely modest in absolute terms, even when the drugs are used in combination therapy.
Rapalogs are relatively well tolerated in cancer patients; however, clinical trials performed with either everolimus or temsirolimus have demonstrated that these drugs could cause several adverse events that include: rash, fatigue, stomatitis, mouth ulcers, pneumonitis, anemia, neutropenia, hyperglycemia and elevated cholesterol, and triglyceride levels [178–180]. For example, grade 2 or worse stomatitis was seen in 33% of patients of the BOLERO-2 study and in several cases led to discontinuation of the treatment. Overall, grade 3/4 stomatitis occurs in up to 9% of patients, as reported across multiple mTOR inhibitor clinical trials . However, prophylactic use of a dexamethasone oral solution substantially reduced the incidence and severity of stomatitis . In general, these adverse effects respond well to a lower dosage of the drugs although in some cases discontinuation of the therapy is required. It is likely that lower dosage contributed to some extent to the lack of therapeutic efficacy observed in many trials .
One of the most feared, although rare, adverse events of rapalogs is interstitial, noninfectious pneumonitis, whose pathophysiology still remains elusive. Interestingly, a very recent report has highlighted that pneumonitis is associated with a better outcome in terms of PFS, OS, and clinical benefit rate in patients with metastatic RCC and might therefore be used as a marker of everolimus efficacy .
Long-term responders to rapalog treatment
As emphasized above, the results of rapalog treatment in sporadic cancer patients are quite disappointing with a reported median PFS of less than 6 months.
One reason why rapalogs have had limited success to treat cancer could be that their drug action is cytostatic rather than cytotoxic, i.e. these drugs arrest cancer cell growth instead of killing them. Moreover, as we have discussed, they activate several prosurvival oncogenic pathways which further dampen the anticancer efficacy of allosteric mTOR inhibitors. However, there are in the literature occasional cases of patients who showed a dramatic clinical response to either everolimus or temsirolimus.
The first tumor where this was seen are Perivascular Epithelioid Cell tumors (PEComas), a rare sarcoma subtype which commonly displays mutations in either TSC1 or TSC2 . Quite a few PEComa patients, including some with massive tumors, showed a complete response (CR) to rapalogs lasting over a year [185–190].
By analyzing the genome of a metastatic urothelial bladder carcinoma patient who had achieved a sustained CR (at least 4 years, as far as it is known) to everolimus, Iyer and co-workers  identified a 2-bp deletion in TSC1 resulting in a frameshift truncation (c.1907_1908del, p.Glu636fs) and a nonsense mutation in NF2 creating a premature stop codon (c.863C>G, p.Ser288). These loss-of-function mutations are noteworthy, as alterations in these genes had been previously associated with mTORC1 signaling-dependence in preclinical models . Importantly, knockdown of NF2 expression in TSC1-null bladder cancer cells resulted in enhanced sensitivity in vitro to rapamycin . Therefore, it is conceivable that the NF2 mutation also contributed to the observed exquisite sensitivity to everolimus.
Another long-term responder belonged to a cohort of patients enrolled in a phase 1 study of pazopanib (a multitargeted RTK inhibitor) and everolimus in advanced solid tumors . This patient, with metastatic urothelial carcinoma, showed a complete radiographic response that lasted 14 months. Whole exome sequencing revealed the presence of two previously unknown activating mutations in MTOR. One mutation (E2419K) was located in the kinase domain, while the other (E2014K) was in the FRB domain. Each of these mutations activated mTORC1 signaling, being additive when both were present. The occurrence of these mutations may contribute to the strong dependency of cancer cells on mTORC1 signaling and the extremely good response to everolimus . The third long-term (18-month) responder belonged to a group of patients enrolled in a phase 2 trial where everolimus was used to treat thyroid cancer (NCT00936858). Whole-exome sequencing performed on pretreatment tumor, revealed a somatic nonsense truncating mutation in TSC2 (Q1178*) . This mutation is known to inactivate TSC2 by eliminating the GTP-binding protein domain, near the C-terminal, which is essential for inhibiting mTORC1 activity [28,195]. It should be also pointed out, however, that the patient displayed an N-terminal frame shift in FLCN (R17fs), a gene which encodes for the tumor suppressor, folliculin. FLCN is the causative gene for Birt–Hogg–Dubé syndrome  and its inactivation has been shown to result in increased mTORC1 activity [197,198]. This raises the possibility that both mutations contributed to everolimus sensitivity, although there are conflicting reports regarding the effects of folliculin on mTORC1 activity . Remarkably, when the patient became resistant to everolimus, a mutation (F2018L) was detected in the FRB domain of mTOR, which could explain the occurrence of drug-resistance, as it is predicted that substitution of a leucine for a phenyalanine would prevent rapalog binding through steric hindrance .
Five long-term responders were identified among patients treated with temsirolimus for advanced RCC, and one of these was on rapalog treatment for over 45 months at time of publication . Of these five patients, three had mutations in TSC1 or MTOR or both, whereas two had no genetic alterations in PI3K/Akt/mTORC1 pathway components. What is very interesting is that different regions of the primary tumor or tumor tissue from metastatic sites displayed different arrays of TSC1/mTOR mutations, although all of these alterations were predicted to up-regulate mTORC1 signaling in cancer cells, as they resulted in either functional loss of TSC1 or functional gain of mTOR kinase activity .
However, the most exceptional example of long-term responder to rapalogs reported so far is a patient with metastatic RCC who, at the time of publication, was disease-free 8 years after starting temsirolimus treatment . This patient displayed a novel missense mutation in MTOR (Y1974H) that, in vitro, resulted in increased mTORC1 activity and might explain the extremely high tumor sensitivity to temsirolimus.
Overall, these findings suggested that rapalogs might be highly effective in selected patients whose cancer cells harbor inactivating mutations in TSC1/TSC2 or activating MTOR alterations (and possibly also NF2 inactivating mutations), although not all long-term responders to this class of drugs displayed this kind of genetic anomalies. Interestingly, to date, none of known mutations of PI3K/ Akt/PTEN that are frequently detected in cancer [83,201] has been associated with increased response to rapalogs in patients. This might well reflect the many other functions of these signaling effectors in addition to mTORC1 activation .
Dual PI3K/mTOR inhibitors
As highlighted above, rapalogs only partially inhibit mTORC1-dependent translation and cause feedback activation of oncogenic pathways, including PI3K/Akt. These observations, coupled to the structural similarities between the catalytic domains of PI3K and mTOR , provided impetus for the development of ATP-competitive dual PI3K/mTOR inhibitors , a class of drugs that target PI3K and both mTOR complexes. The leading molecule in this class is PI-103 which did not overactivate Akt and was a more powerful cytostastic agent than rapamycin in a glioblastoma cell model, either when employed alone  or in combination with the EGF receptor (EGFR) antagonist, erlotinib . PI-103 also displayed a more potent proapoptotic activity than either everolimus or rapamycin in acute leukemia cells [105,207] where it inhibited phosphorylation of both p70S6K1 and 4E-BP1 . PI-103 was never translated into the clinic, mainly because of its rapid in vivo metabolism . However, over the next few years several other dual PI3K/mTOR inhibitors were discovered by pharmaceutical companies and tested in phase 1/2 clinical trials. These included NVP-BEZ235 (Dactolisib), XL765/SAR254409 (Voxtalisib), GDC-0980 (Apitolisib), GDC-0084, GSK1059615, GSK2126458, PQR309, VS-5584, PKI-587/PF-05212384 (Gedatolisib), and PF-04691502. Nevertheless, the clinical development of most of these drugs is no longer being pursued and, at present, there only are four active trials recruiting cancer patients for treatment with PKI-587/PF-05212384 and six for PQR309 (clinicaltrials.gov). Contrary to the expectations, clinical studies have revealed quite a limited efficacy of this class of drugs as cancer therapeutics even when they were administered in combination with other drugs [209–213]. Moreover, severe adverse effects were observed, which included nausea, diarrhea, vomiting, decreased appetite, hyperglycemia, mucosatis, cutaneous rash, elevated liver enzyme levels, renal failure, and hypertension [209,213–215]. In metastatic RCC patients treated with NVP-BEZ235, class 3 or 4 adverse events were more commonly observed than with everolimus and the therapeutic efficacy was even lower . Also in mTOR inhibitor-naïve patients with advanced pancreatic cancer, NVP-BEZ235 treatment did not demonstrate increased efficacy compared with everolimus and could be associated with a poorer tolerability profile .
Regarding PKI-587/PF-05212384, it is now being evaluated in phase 1 studies in combination with other anticancer agents (both targeted and classical chemotherapeutics) in patients with a variety of solid tumors (breast, pancreatic, head and neck, ovary, endometrial; see NCT02069158, NCT01920061, NCT03065062, NCT02626507). In a previous study performed in patients with recurrent endometrial carcinoma, PKI-587/PF-05212384, when given intravenously, was well tolerated . Furthermore, preliminary evidence of clinical activity was observed with PKI-587/PF-05212384 plus the MEK inhibitor PD-0325901 in patients with either ovarian or endometrial cancer and Ras mutations .
PQR309 is an orally bioavailable, blood–brain barrier-penetrating dual PI3K/mTOR inhibitor. In light of this important feature, PQR309 is being evaluated, among other neoplasias, in two phase 2 studies of relapsed/refractory primary central nervous system lymphomas (see NCT02669511 and NCT03120000).
Some mechanisms of in vitro resistance to this class of therapeutics have begun to emerge which could at least partially explains why dual PI3K/mTOR inhibitors have displayed limited anticancer efficacy in vivo.
Similarly to allosteric mTOR inhibitors, dual PI3K/mTOR inhibitors have the potential for up-regulating the expression of TKRs. Indeed, prolonged treatment with NVP-BEZ235 led to inhibitor-resistance that was accompanied by increased HER2 expression and phosphorylation in genetically engineered mammary mouse tumors in vivo and in 3-D, but not 2-D, BT474 cell cultures in vitro . Notably, HER2 mRNA levels were not significantly increased in response to NVP-BEZ235, an observation which suggested that increased HER2 expression was mainly due to protein stabilization.
In pancreatic ductal adenocarcinoma cell lines, dual PI3K/mTOR inhibitors (NVP-BEZ235, PKI-587/PF-05212384 and GDC-0980), at concentrations that resulted in complete dephosphorylation of Ser473 p-Akt and 4E-BP1, markedly and rapidly enhanced the MEK/ERK pathway. ERK overactivation could be prevented by cotreatment with MEK inhibitors such as PD-0325901 . Inhibitors of EGFR, HER2, insulin receptor, and IGF-1R could not prevent NVP-BEZ235-induced MEK/ERK signaling up-regulation that, therefore, could not be ascribed to enhanced expression of RTKs. However, Rictor knockdown through siRNA transfection markedly attenuated the enhancing effects of NVP-BEZ235 on ERK phosphorylation. Therefore, it was proposed that dual PI3K/mTOR inhibitors suppressed an as yet unidentified negative feedback loop mediated by mTORC2, thereby leading to overactivated MEK/ERK pathway . The pancreatic ductal adenocarcinoma cell lines used in this study (PANC-1 and MiaPaCa-2) display Ras mutations and a combination consisting of NVP-BEZ235 plus PD-0325901 caused a more pronounced inhibition of cell growth than that produced by each inhibitor individually . Therefore, these findings, obtained in vitro, are consistent with the recent results of the clinical trial in which PKI-587/PF-05212384 combined with the MEK inhibitor PD-0325901, displayed some efficacy in Ras-mutated ovarian or endometrial cancer patients .
mTOR kinase inhibitors (TORKIs)
This class of ATP-competitive molecules, which block only the mTOR catalytic domain, was designed to reduce toxicity due to the use of dual PI3K/mTOR inhibitors. The protoype of TORKIs was PP242 that inhibited both mTORC1 and mTORC2 but not PI3K . When compared with rapamycin, PP242 blocked in vitro phosphorylation of Ser473 p-Akt and Thr37/46 p-4E-BP1, inhibited cell proliferation more completely, and negatively affected cap-dependent translation under conditions where rapamycin had no effects. PP42 blocked rapamycin resistant outputs of mTORC1 and mTORC2 also in vivo . Numerous other TORKIs have been subsequently discovered, which include Torin 1, Torin 2, Ku-0063794, AZD8055, AZD2014 (Vistusertib), CZ415, INK128/MNL0128 (now referred to as TAK-228), OSI-027, WYE354, WYE312, WYE687, WAY600, Palomid 529, GDC-0349, CC223, and XL388. Some of these drugs, which included AZD8055, AZD2014, TAK-228, CC-223, and OSI-027, entered phase 1/2 clinical trials. However, at present, AZD8055 and OSI-027 have been discontinued, while there still are active trials for AZD2014 (26 trials), TAK-228 (27 trials), and CC-223 (1 trial) in patients with a wide variety of neoplastic disorders (clinicaltrials.gov). However, in a phase 2 study in patients with VEGF-refractory metastatic RCC, AZD2014 proved to be significantly inferior to everolimus in terms of both PFS and OS, while the occurrence of adverse effects was not significantly different between the two drugs . As to TAK-228, the results of two phase 1 clinical studies have been recently released. When the inhibitor was employed alone in 33 patients with refractory/relapsed hematological malignancies (multiple myeloma/MM, non-Hodgkin lymphoma/NHL, and Waldenström’s macroglobulinemia/WM), one MM patient had a minimal response, one WM patient achieved partial response, one WM patient had a minor response, and 18 patients (14 MM, two NHL, and two WM) had stable disease. The most commonly observed grade ≥3 adverse effects were thrombocytopenia (15%), fatigue (10%), and neutropenia (5%). It was therefore concluded that further studies including drug combination strategies could be warranted for possible clinical use of TAK-228 . The results of the other trial, where TAK-228 was administered plus paclitaxel, with/without trastuzumab, in 67 patients with advanced solid malignancies (breast, esophageal, lung, endometrial, and ovarian carcinomas), demonstrated that of 54 response-evaluable patients, eight achieved partial response, and six had stable disease lasting ≥6 months . Overall, the most common grade ≥3 drug-related toxicities were neutropenia (21%), diarrhea (12%), and hyperglycemia (12%). Also in this case it was concluded that further investigation of TAK-228 in combination with other therapeutic agents including paclitaxel, with/without trastuzumab, could be performed in solid tumor patients.
Remarkably, TORKIs were effective in terms of growth inhibition in cell models that either displayed intrinsic insensitivity to allosteric mTOR inhibitors or became resistant because, following a long-term exposure to rapamycin, they acquired an mTOR mutation (S2035F) in the FRB domain [225,226].
However, mechanisms of primary or acquired resistance to TORKIs have been reported as well, some of them being similar to those discovered for other classes of mTOR targeting agents. These include RTK overexpression/activation, in some cases due to the release of the FoxO1/3a transcription factors that are downsteam of Ser473 p-Akt [227–230]; MEK/ERK signaling overactivation [230–232]; alterations in the eIF4E/4E-BP1 ratio (high eIF4E/low or absent 4E-BP1) [233–236]; and compensatory glutamine metabolism [237,238].
Most of these findings are further evidence of the many adaptive capabilities of the PI3K/Akt/mTOR signaling network that are activated when one of its critical components is targeted.
The newest generation of mTOR inhibitors consists of RapaLink-1. This drug simultaneously associates with and allosterically inhibits mTORC1 via the FRB domain while blocking the catalytic activity of this mTOR complex by binding to the ATP-binding pocket of mTOR itself . The rationale behind the design of Rapalink-1 stems from the observation that resistant clones emerged from the breast cancer cell line MCF-7 exposed for several weeks to either rapamycin or the TORKI, AZD8055. While AZD8055-resistant cells harbored an mTOR mutation located in the kinase domain at the M2327I position, two rapamycin-resistant clones displayed mutations located in the FRB domain at positions A2034V and F2108L . Interestingly, the F2108L mutation had been previously reported in a long-term responder cancer patient who had become resistant to everolimus treatment and relapsed, as we have described earlier in this article , while the M2371 mutation had been previously observed in a total of five patients with different types of solid tumors . In cells with FRB domain mutations, phosphorylation levels of the normally rapamycin sensitive residues on p70S6K1 (Thr389) and S6RP (Ser235/236) were unaffected even at high everolimus concentrations (100 nM). The M237I mutation resulted in an increase in mTOR kinase activity and rendered cells resistant to a variety of TORKIs (PP242, WY354, and KU-0063794) in addition to AZD8055, as documented by lower sensitivity of 4E-BP1 phosphorylation to this class of drugs. These observations led to the synthesis of novel bivalent mTOR inhibitors, collectively referred to as RapaLink, consisting of a rapamycin-FRB binding element appropriately linked to a TORKi. The most effective of these new molecules is RapaLink-1 that could indeed reverse in vitro and in vivo resistance of breast cancer cells due to either mTOR FRB or kinase domain mutations . More recently, a study where RapaLink-1 was tested in a glioblastoma cell model has demonstrated that this drug crosses the blood–brain barrier and is more powerful than earlier mTOR inhibitors, as it dephosphorylates Ser473 p-Akt and Thr 37/46 p-4E-BP1 in vitro more efficiently than rapamycin and it induces a more durable mTOR inhibition than TORKIs. Indeed, association of RapaLink-1 with FKB-P12 enables accumulation of the inhibitor within the cells, whereas the TORKI TAK-228 shows a much shorter intracellular residence time. In mice, RapaLink did not display significant toxicities when given intraperitoneally. The drug markedly reduced growth of intracranial glioblastoma xenografts and achieved initial regression of the tumors, whereas both rapamycin and TAK-228 were only modestly effective .
Overall, these two recent articles suggest that this new class of mTOR-targeted agents holds promise for future therapy of cancer patients, although it should be highlighted that, after initial regression, the glioblastoma xenograts showed tumor regrowth. Therefore, it is likely that, after a while, even with this class of drugs acquired mechanisms of resistance will develop.
Challenges and future perspectives
mTOR inhibitors, as several other types of targeted drugs, have met with a limited success as anticancer therapeutics. The reason for this failure has been initially ascribed to the fact mTOR inhibitors do not display a powerful cytotoxic activity and unleash a series of compensatory signaling pathways that dampen their therapeutic efficacy. However, as we have discussed in this review, other reasons have become apparent, including the tumor cell genetic background and the existence of mTOR mutations that can explain the occurrence of drug-resistance. In this connection, the observations made through the study of extraordinary responders to rapalog treatment suggest that routine screening of cancer patients for genetic alterations in the mTOR pathway may be helpful for identifying a subset of patients who are more likely to respond to mTOR-targeted therapies much better than other patients. One major challenge of currently ongoing clinical trials is that the list of cancer genes and mutations that influence to sensitivity to mTOR inhibitors is still in its infancy. For example, preclinical studies have shown that PIK3CA mutant cancer cells which also contain RAS mutations are resistant to everolimus . However, there are no convincing clinical data to support this conclusion . Therefore, large-scale studies are needed to identify biomarkers of efficacy and resistance to mTOR inhibitors, as the ones performed so far have provided somehow inconclusive results [242–245].
However, it is probable that alternative mechanisms contribute to clinical response, including epigenetic events affecting one or more of the genes encoding proteins involved in controlling mTORC1 or mTORC2 [246,247]. Moreover, in case of combination therapy, we are just beginning to realize how mTOR inhibitors, associated with traditional chemotherapeutics that rely on cell-cycle progression to kill cancer cells, could result in antagonistic effects . Therefore, at present we cannot predict who will and who will not respond to therapy based on mTOR inhibition.
Another daunting hurdle to successful development of anticancer strategies, including targeted therapy, is represented by spatial and temporal intratumor/intertumor heterogeneity, which fosters tumor branched evolution and drug-resistance [249–252]. This issue is even more critical in light of the findings that targeted agents themselves may be the driving force that leads to the selection and emergence of multiple drug-resistant clones [253,254]. However, tumor evolution could also provide opportunities for developing new strategies for targeted therapy, as recently demonstrated in a study which highlighted that brain metastases displayed alterations associated with sensitivity to PI3K/AkT/mTOR and HER2 inhibitors not detected in the matched primary-tumor sample. Therefore, these findings suggest that sequencing of primary biopsies alone could lead to missing a substantial number of opportunities for targeted therapy . Although multiple tumor sampling from different regions of the primary cancer and from metastases is feasible, such a strategy is highly invasive and harmful to patients, as well as logistically challenging and costly. For overcoming these issues, liquid biopsy appears to be a valuable and promising alternative strategy to multiple tumor sampling. This innovative approach is capable of providing useful information with regards to the mutational status of EGFR and efficacy of EGFR-targeting agents in colorectal or lung cancer patients [256,257]. Therefore, it may be used also for detecting mutations of mTOR and TSC2.
In any case, we have now at our disposal a growing array of sophisticated techniques that could circumvent, at least in part, the issue of tumor heterogeneity. Single-cell whole genome and whole exome sequencing, single-cell RNA sequencing, single-cell epigenetics, single-cell proteomics/phosphoproteomics, high-throughput drug screening, and kinase inhibition data coupled with bioinformatics and computational biology are all emerging informative platforms that have the potential to enable the design of more effective and durable personalized anticancer therapies [258–260]. RNA sequencing should probably include also microRNA (MiR) analysis, as it is now emerging that MiR expression is associated with acquired resistance to targeted therapy in lung cancer  and could up-regulate mTORC1 signaling by increasing the levels of TSC1 in ovarian cancer cells .
It is worth emphasizing here the findings obtained through single-cell whole-exome sequencing and phosphoproteomics on a patient-derived in vivo glioblastoma model of TORKI resistance . It was shown that compensatory rewiring of heterogeneous signaling pathways during the responsive state of the treatment acted as a dominant mechanism of resistance, while sequencing ruled out the selection of a TORKI-resistant genotype. Hence, it was possible to identify combination therapies that resulted in complete and sustained tumor suppression in vivo. Importantly, alterations in the protein signaling cross-talks were detectable as early as 2.5 days after treatment, whereby anticipating drug-resistance long before it was clinically manifest . Overall, these results highlight the power of single-cell quantitative phosphoproteomics for guiding the choice of targeted therapies and therapy combinations.
A promising field for exploiting mTOR inhibition as an anticancer strategy is that of immunotherapy. Immune checkpoint blockade therapy for programmed cell death-1 and its ligand (PD-1/PD-L1) or cytotoxic T-lymphocyte-associated protein-4 (CTLA-4) has shown unprecedented response rates and has provided unparalleled clinical benefits in the treatment of several human malignancies, which include melanoma, RCC, urothelial cancer, and non–-small cell lung carcinoma . Indeed, immune checkpoint inhibitors have proven effective in about 20% of patients who are refractory to conventional treatments and molecular targeting therapy. Therefore, FDA and EMA have so far approved a total of six monoclonal antibodies targeting PD-1 or PD-L1 or CLTLA-4 as therapeutics for some types of advanced cancer . However, there are cancer types where ORRs to immune checkpoint inhibitors display intrinsically low efficacy, while multiple tumor- or immune-driven resistance mechanisms are being identified . Therefore, one of the challenges ahead is to rationally combine drugs able to make the tumor microenvironment more permissive to immunotherapy in order to potentiate its clinical activity . In this context, it is now emerging that concomitant mTOR inhibition could be very important to increase the efficacy of immune checkpoint inhibitors in cancer [267–272].
Last but not the least, it should not be overlooked that mTOR-generated signals have been identified as major compensatory pathways allowing tumors to escape several types of targeted cancer therapies, as mTORC1 and mTORC2 are central signaling hubs functionally related to several other oncogenic pathways . Hence, mTOR signaling should probably be monitored routinely in tumors and mTOR inhibition should be considered as a cotherapy of other types of treatment, such as MEK/ERK inhibitors, cyclin-dependent kinase inhibitors or EGFR .
Our knowledge about mTOR has considerably increased over the last 10 years. However, even with improved understanding of the mechanism of actions of mTOR inhibitors, targeted therapy has not as yet translated into improved response rates. Nevertheless, future studies aimed to determine additional functions and regulation of mTORC1/mTORC2, coupled with a deeper knowledge of the effects of mTOR inhibitors and of mechanisms of resistance, will of fundamental importance to design more efficacious treatments based on mTOR inhibition for improving cancer patient outcome.
All authors were involved in writing the paper and approved the final submitted and published versions.
The authors declare that there are no competing interests associated with the manuscript.
J.A.M. was supported in part by grants from East Carolina University Grants (#111104 and #111110-668715-0000).
eukaryotic translation initiation factor 4E-binding protein 1
AMP-activated protein kinase
Breast cancer trials of OraL EveROlimus
cytotoxic T-lymphocyte-associated protein-4
DEP domain-containing mTOR-interacting protein
epidermal growth factor
epidermal growth factor receptor
eukaryotic translation initiation factor 4E
European Medicines Agency
extracellular signal-regulated kinase
EXamining everolimus In a Study of Tuberous sclerosis complex
Forkhead box O 1/3a
FKBP–rapamycin associated protein
GTP-ase activating protein
GAP activity toward Rags
guanine nucleotide exchange factor
Growth factor receptor-bound protein 10
hepatitis B X-interacting protein
Huntingtin/Elongation factor 3/A subunit of Protein phosphatase-2A/ TOR1
human EGF receptor 2
hypoxia-inducible factor 1α
insulin-like growth factor-1
insulin receptor substrate
liver kinase B1
mantle cell lymphoma
mitogen-activated protein kinase kinase
mammalian lethal with SEC13 protein 8
MAPK-interacting kinase 1
MAPK scaffold protein 1
mammalian stress-activated protein kinase interacting protein 1
mechanistic target of rapamycin
mTOR complex 1
mTOR complex 2
overall response rate
ribosomal protein p70 S6 kinase 1
p90 ribosomal S6 kinase 1
programmed cell death-1
programmed cell death ligand 1
platelet-derived growth factor
phosphoinositide-dependent kinase 1
Perivascular Epithelioid Cell tumors
progression free survival
protein kinase C
phosphatidylinositol 3-kinase-related kinase
phosphatidylinositol 4,5 bisphosphate
phosphatidylinositol 3,4,5 trisphosphate
Proline-rich Akt substrate 1 40 kDa
protein observed with Rictor
phosphatase and tensin deleted on chromosome 10
RAD-001 In Advanced Neuroendocrine Tumor
rapamycin and FKBP-12 target 1
regulatory-associated protein of TOR
renal cell carcinoma
REnal cell Cancer treatment with Oral RAD-001
regulated in DNA damage and development 1
Ras homolog enriched in brain
rapamycin-insensitive companion of TOR
reactive oxygen species
receptor tyrosine kinase
subependymal giant cell astrocytoma
serum and glucocorticoid-activated kinase 1
sterol responsive element binding protein
ATP-competive mTOR kinase inhibitors;TGF-α, transforming growth factor-α
transactivation/transformation domain-associated protein
tuberous sclerosis complex
Tel2-interacting protein 1
Unc-51 like kinase 1
vascular endothelial growth factor