The Role of mTOR Signaling as a Therapeutic Target in Cancer

The aim of this review was to summarize current available information about the role of phosphatidylinositol-3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) signaling in cancer as a potential target for new therapy options. The mTOR and PI3K/AKT/mTORC1 (mTOR complex 1) signaling are critical for the regulation of many fundamental cell processes including protein synthesis, cell growth, metabolism, survival, catabolism, and autophagy, and deregulated mTOR signaling is implicated in cancer, metabolic dysregulation, and the aging process. In this review, we summarize the information about the structure and function of the mTOR pathway and discuss the mechanisms of its deregulation in human cancers including genetic alterations of PI3K/AKT/mTOR pathway components. We also present recent data regarding the PI3K/AKT/mTOR inhibitors in clinical studies and the treatment of cancer, as well the attendant problems of resistance and adverse effects.

Activated mTOR further phosphorylates downstream substrates, thus implicating a variety of cellular processes. mTORC1 directly regulates protein synthesis; plays a central role in lipid and nucleotide synthesis and energetic homeostasis, ribosome biogenesis, nucleotide metabolism, and cell cycle progression; and also negatively regulates catabolic processes such as autophagy, therefore controlling the balance between anabolism and catabolism in response to environmental conditions [53]. mTORC2 mainly regulates cell proliferation, survival, cytoskeletal remodeling, and cell migration [54]. Activated mTOR further phosphorylates downstream substrates, thus implicati variety of cellular processes. mTORC1 directly regulates protein synthesis; plays a cen role in lipid and nucleotide synthesis and energetic homeostasis, ribosome biogen nucleotide metabolism, and cell cycle progression; and also negatively regulates catab processes such as autophagy, therefore controlling the balance between anabolism catabolism in response to environmental conditions [53]. mTORC2 mainly regulates proliferation, survival, cytoskeletal remodeling, and cell migration [54]. mTORC1 regulates protein synthesis through the phosphorylation of two molecules, S6K1 (p70 S6 kinase) and 4E-BP1 (eukaryotic initiation factor 4E-bind protein), that promote translation and protein synthesis [55]. Unphosphorylated 4Esuppresses the initiation of translation by binding and sequestering translation initia factor 4E (eIF4E). Upon phosphorylation by mTORC1, 4E-BP1 dissociates from eI allowing for 5′cap-dependent mRNA translation to occur [56][57][58].
The activation of PI3K-mTOR signaling is normally controlled at numerous le Activated RTKs are dephosphorylated by protein tyrosine phosphatases. To contro extent of mTORC1 activation and restore TSC regulation after this stimulus, the mTO mTORC1 regulates protein synthesis through the phosphorylation of two key molecules, S6K1 (p70 S6 kinase) and 4E-BP1 (eukaryotic initiation factor 4E-binding protein), that promote translation and protein synthesis [55]. Unphosphorylated 4E-BP1 suppresses the initiation of translation by binding and sequestering translation initiation factor 4E (eIF4E). Upon phosphorylation by mTORC1, 4E-BP1 dissociates from eIF4E, allowing for 5 cap-dependent mRNA translation to occur [56][57][58].
S6K1 and mTORC1 upregulate the transcription of rRNA by enhancing the activity of RNA polymerase I and RNA polymerase III through the phosphorylation of the upstream binding factor (UBF) [59], transcription initiation factor 1A (TIF-1A) [60], and MAF1 regulatory factors [61,62]. S6K1 also enhances protein synthesis by activating eIF4B [63], a positive regulator of cap-dependent translation, and by degrading the eIF4A inhibitor PDCD4 (programmed cell death protein 4) [64].

Regulation of mTORC1 by Other Factors
The activity of mTORC1 is also regulated by amino acids, energy status, phosphatidic acid, and oxidative stress (reviewed in [72,73]).
Unlike growth factors, amino acids activate mTORC1 through a PI3K-, AKT-, and TSC-independent mechanism [74,75]. The amino acid-dependent activation of mTORC1 requires the Rag subfamily of Ras small GTPases [76,77]. Amino acids stimulate the loading of RagA or RagB with GTP, which enables RagA/B to interact with the RAPTOR component of mTORC1 [77]. This interaction results in the translocation of mTORC1 to the lysosomal surface, where the Rag GTPases dock on a multi-subunit complex called Ragulator, and lysosomal Rheb stimulates mTORC1 kinase activity [49].
Another important regulator of mTORC1 signaling is the AMP-activated protein kinase (AMPK), the principal energy sensor in most eukaryotic cells [78]. In contrast to growth factors, the depletion of cellular energy inhibits mTORC1 activity and cell growth. During periods of energy or oxygen deprivation, AMP levels rise and promote the activation of AMPK [78,79]. Activated AMPK directly inhibits mTORC1 by phosphorylating RAPTOR [80], and indirectly inhibits mTORC1 by phosphorylating TSC2 on T1227 and S1345, as well as enhancing the ability of TSC2 to inhibit mTOR signaling [81].
An additional regulatory mechanism of mTORC1 signaling involves the lipid second messenger phosphatidic acid (PA) [72]. The potential role of PA in the regulation of mTORC1 signaling was revealed in a study that demonstrated that PA can directly bind to the FRB domain of mTOR and stimulate the activation of the mTOR substrate S6K1 and the phosphorylation of 4E-BP1 in HEK293 cells [82]. It was further demonstrated that phospholipase D (PLD), an enzyme that promotes the hydrolysis of phosphatidylcholine to yield choline and PA [83,84], is a critical regulator in the activation of mTOR signaling by a variety of stimuli [85][86][87]. The elevated expression of either PLD1 or PLD2 activates mTORC1 signaling in various types of cells (reviewed in [72]).

Deregulation of the PI3K/AKT/mTOR Pathway in Cancer
According to the published data, the PI3K/AKT/mTOR pathway is activated in approximately 70% of ovarian or breast cancers [88,89]. For instance, the aberrant activation of the PI3K/AKT/mTOR pathway has been found in 90% of lung adenocarcinomas (ADCs) and 40% of squamous cell carcinomas (SCCs) [90].
In normal physiology, regulatory mechanisms tightly control the activity and homeostasis of the PI3K/AKT/mTOR pathway, but it can be constitutively activated in various cancers. These mechanisms include the amplification or mutation of genes encoding PI3K subunits, AKT, and other pathway members; the activation of receptor tyrosine kinases, mutation, or overexpression of growth factor receptors, e.g., epithelial growth factor receptor (EGFR) or human epidermal growth factor receptor 2 (HER2); the inactivating mutations in the genes encoding key tumor suppressors PTEN or INPP4B; the inactivating mutations in the genes encoding mTOR regulators such as TSC1 and TSC2; and the activating mutations in MTOR itself [91][92][93].

PI3K in Human Cancer
PI3Ks belong to a family of lipid kinases that phosphorylate the 3-OH group of phosphoinositides [94]. Based on their primary structures and in vitro lipid substrate specificity, PI3Ks are classified into three classes: class I PI3Ks, class II PI3Ks, and class III PI3Ks [28]. PI3K-related kinases, which are sometimes referred to as class IV PI3Ks, are protein serine/threonine kinases with a structure similar to the catalytic subunits of PI3Ks. Examples of PI3K-related kinases are mTOR and the DNA-dependent protein kinase (DNA-PK).
Class I PI3Ks are heterodimeric molecules composed of regulatory and catalytic subunits. Class I PI3Ks are further divided into two subclasses-IA and IB-based on their regulatory subunit and upstream activator.
Class IA PI3Ks are widely present in carcinogenic processes. The PIK3CA gene, encoding the p110α subunit, is frequently mutated or amplified in the most common human cancers, including those of the breast, colon, gastric, endometrial, cervical, prostate, and lung, as well as glioblastoma [98][99][100][101][102][103]. Most of these mutations occur at two hotspots-E545K and H1047R-located in the helical domain and the kinase domain of p110α, respectively [103].
Catalytic subunit p110α normally binds to p85, which stabilizes it and controls its enzymatic activity [104]. p110α with the E545K mutation gains the ability to associate with IRS-1 independently of the p85 regulatory subunit, resulting in the constitutive activation of the PI3K pathway and increased cell proliferation, survival, and motility [105]. Another highly recurrent PIK3CA mutation, H1047R, increases protein activity through alteration of its catalytic site, resulting in the upregulation of PI3K signaling. However, these two mutants have slightly different phenotypic impact and expression of p110α, as E545K produces a more severe metastatic phenotype than that induced by expressing p110α H1047R in a breast cancer cell line [106].
The PIK3CA gene was found to be mutated, on average, in 15% of human cancers, and cancers of the liver, breast, and colon harbor the most PIK3CA mutations, with average mutational frequencies of 36%, 26%, and 25%, respectively [99]. A detailed analysis of the COSMIC (Catalogue Of Somatic Mutations in Cancer, UK) database revealed that PIK3CA is indeed most frequently mutated in breast cancer (28.83%), especially in estrogen receptor (ER)-positive carcinomas (38.88%), as well as endometrium (27.39%) and urinary tract (20.2%) cancers (Table 1). By contrast, elevated PIK3CA expression has been found in about 32% of lung cancers (Table 1).

AKT in Cancers
AKT is an evolutionarily conserved serine/threonine kinase that belongs to the AGC kinase family. Three highly conserved AKT isoforms have been identified: AKT1 (i.e., AKT), AKT2, and AKT3. The alteration of AKT activity is associated with several human diseases, including cancer and diabetes [107].
AKT is composed of three conservative structure domains: an N-terminal PH domain, a central kinase domain, and a C-terminal regulatory domain. A single amino acid substitution, E17K, in the lipid-binding PH domain of AKT-1 is a recurrent somatic cell mutation that occurs in breast cancer, meningioma, colorectal, endometrial, and ovarian cancers, and the mutation results in constitutive AKT1 activation [108,109]. This mutation dramatically increases the affinity of E17K AKT to PtdIns(4,5)P2 [110], activates AKT1 by means of pathological localization to the plasma membrane, and stimulates downstream signaling [111,112]. According to the COSMIC database, the frequency of AKT1 mutations is about 4% in breast cancer and 15% in meningiomas, and elevated expression is more common for endometrium (8.14%) and lung (8.73%) cancers ( Table 2).
The overexpression of phosphorylated AKT (p-AKT) is also considered to be an indicator of poor prognosis in many malignancies. A meta-analysis conducted to evaluate the association of p-AKT overexpression with breast cancer prognosis showed that high p-AKT expression was significantly associated with a higher risk of death and disease recurrence [113]. In meta-analyses, p-AKT overexpression was also associated with worse survival in NSCLC (non-small cell lung carcinoma) [114,115], lymph node metastasis and poor prognosis in patients with gastric cancer [116], and poor overall survival and progression-free survival in patients with epithelial ovarian cancer [117]. Table 2. Genetic alterations of the AKT1 gene in human cancers.

Alterations of Receptor Tyrosine Kinases
The RTK-mediated activation of the PI3K/AKT/mTOR pathway is crucially important for its oncogenic activity and is clearly linked to the RTK signaling. Examples include PI3K activation by EGFR in lung cancers harboring somatic-activating mutations in EGFR [118] and HER2 mutations in breast cancers with HER2 amplification [119]. EGFR mutations have also been reported in colorectal cancers [120] and glioblastomas [121].
EGFR/HER1/ErbB1 and HER2/ErbB2 are the members of the ErbB receptor family. Activated receptors bind various signaling proteins and stimulate activation of many signaling pathways, including the PI3K/AKT/mTOR, Ras/RAF/MEK/ERK, phospholipase C (PLC)-γ1, and SRC pathways (reviewed in [122]). It has been shown that many cancer cells are characterized by EGFR hyperactivation, gene amplification leading to receptor overexpression, or mutants with dysregulated signaling [123]. In the EGFR gene, mutations occur within exons 18-21, which encode a portion of the EGFR kinase domain. For instance, approximately 80-90% of patients with EGFR-mutated, non-small cell lung cancer have either deletions in exon 19 or substitutions of leucine for arginine (L858R) in exon 21 of the EGFR gene [124].
An analysis of the COSMIC database displayed that the incidence of EGFR mutations significantly varies among the different types of cancers. The highest frequency is detected in lung cancers (26.58%) and especially in adenocarcinoma (30.61%), where AKT is also overexpressed ( Table 3). In contrast to lung cancer, the EGFR mutations are rare in breast cancer [125,126], with a frequency of no more than 5-6% (Table 3). However, the overexpression of EGFR is observed in 15-30% of breast carcinomas and is associated with a large tumor size and poor clinical outcomes [126].

PTEN
The PTEN protein consists of two major domains, the N-terminal phosphatase catalytic domain (residues 7-185) and a C-terminal domain (residues 186-351) [127]. The N-terminal phosphatase domain of PTEN contains a consensus PtdIns(4,5)P2-binding motif, and the C-terminal part contains the lipid-binding C2 domain. The C2 domain is believed to be required for the correct positioning of PTEN at the plasma membrane, the site of the lipid substrates of PTEN [128].
PTEN has been shown to be lost or inactivated by multiple mechanisms in a wide spectrum of human cancer types: a loss of heterozygosity of PTEN was found in 60-80% of patients with glioblastoma and in 45% of endometrioid carcinoma of the ovary; a loss of PTEN protein expression was found in 20-40% of patients with colorectal cancer; and a decreased PTEN expression was found in 30% of patients with head and neck cancer, PTEN mutations were found in up to 40% of patients with glioblastoma and in 15-50% of patients with prostate cancer [134]. Mutations of the human PTEN gene are also frequently observed in breast cancer, glioblastoma, endometrial cancer, malignant melanoma, and prostate cancer [135]. Deletions of PTEN have also been shown in lung cancer. The loss of PTEN expression assessed by immunohistochemistry has been demonstrated in up to 24% of 125 resected early stage NSCLC specimens [136]. Another series of early stage NSCLC specimens revealed that PTEN protein expression was reduced or lost in 74% of tumors [137]. Elevated levels of miR-21 in NSCLC, relative to adjacent non-tumor tissues, were correlated with decreased PTEN mRNA levels and advanced tumor stages [138].
An analysis of the COSMIC database revealed PTEN mutations in 55.71% of endometrioid carcinomas, in 39.68% of endometrium cancers, in 13.36% of central nervous system cancers, and in 8.4% of skin cancers (Table 4).

mTOR and Cancer Development
Receptor tyrosine kinase hyperactivation or overexpression, PTEN loss of function, and mutations in PIK3CA and AKT are not the only mechanisms that can lead to mTOR activation. mTOR can become activated via additional molecular mechanisms that include gene amplification and mutation (reviewed in [139]). It was reported that two different point mutations, S2215Y and R2505P, identified in the COSMIC database, confer the constitutive activation of mTOR signaling in cell cultures even under nutrient starvation conditions. S2215Y was identified in large intestine adenocarcinoma, whereas R2505P was identified in renal cell carcinoma [140]. In cell cultures, the association of the mTOR P2229R mutation with the activation of both the mTORC1 and mTORC2 pathways, increased cell proliferation, and cell survival was also demonstrated [141]. Grabiner et al. identified 33 MTOR mutations that occur in multiple cancer types and confer pathway hyperactivation [142]. It was also found that resistant colonies emerged after the exposure of the MCF-7 breast cancer cell line to high concentrations of a first generation mTORC1 inhibitor, rapamycin, or a second generation mTOR ATP competitive inhibitor (AZD8055) for three months. Deep sequencing revealed that the AZD8055-resistant clones harbored an mTOR mutation located in the kinase domain at the M2327I position, while two rapamycinresistant clones contained mutations located in the FKBP12-rapamycin binding domain at positions A2034V and F2108L [143].
An analysis of the COSMIC database revealed an overexpression of MTOR in ovary (9.77%), urinary tract (8.33%), and skin (8.25%) cancers and a downregulation in tumors of the central nervous system (13.06%). The highest incidence of MTOR mutations is in meninges (18.63%), endometrium cancer (10.43%), and, especially, endometrioid carcinoma (12.63%) ( Table 5). Many clinical reports have attempted to investigate the roles and the potential prognostic value of mTOR and p-mTOR in a variety of cancers. A meta-analysis of data from 915 patients with esophageal squamous cell carcinoma (ESCC) was conducted to evaluate the prognostic and clinicopathological significance of mTOR/p-mTOR expression. The pooled analysis identified that positive mTOR/p-mTOR expression was significantly correlated with the worse conditions of differentiation degree, depth of tumor invasion, and lymph node metastasis but had no relationship to gender. Moreover, mTOR/p-mTOR ex-pression was also significantly associated with worse overall survival, disease-free survival, and cancer-specific survival of patients with ESCC [144].
The prognostic impact of the PI3K/AKT/mTOR signaling pathway in advanced esophageal squamous cell carcinoma was also assessed in 145 tumor and 145 non-tumor samples of patients from China. The PI3K/AKT/mTOR signaling pathway was shown to be significantly upregulated and PTEN was largely downregulated in tumor tissue. The expression level of PTEN, mTOR, p-mTOR, and S6K1 was closely related to the presence of lymph node metastases. The expression of PTEN, mTOR, and S6K1 were also correlated to the TNM stage and overall survival [145].

Targeting PI3K/AKT/mTOR
The PI3K/AKT/mTOR signaling pathway plays an important role in cell growth, proliferation, and survival, and it is one of the most commonly deregulated pathways found in human cancers, which makes components of this pathway attractive targets for anticancer therapy.

PI3K Inhibitors
PI3K inhibitors can be subdivided into pan-PI3K inhibitors, isoform-selective PI3K inhibitors, and dual PI3K/mTOR inhibitors [146]. Dozens of PI3K inhibitors have been developed as potential chemotherapeutic drugs (reviewed in [146,147]). The first PI3K inhibitors wortmannin and LY294002 did not reach clinical trials due to the problems with stability, solubility, and toxicity [148], and the development of the PI3K inhibitor as antitumor agent has been a hotspot area since 2006 when ZSTK474, a new pan-class I PI3K inhibitor with less toxic effects, was first presented [149].
Pan-PI3K inhibitors act on each of the four catalytic isoforms of class I PI3K and have, as expected, broad inhibition potential in a number of tumors. However, this broad activity may lead to a higher risks of side effects and toxicities [150]. Until the end of 2020, two pan-PI3K inhibitors, i.e., copanlisib and duvelisib, and two isoform-selective PI3K inhibitors, i.e., idelalisib and alpelisib, were approved for cancer treatment by the Food and Drug Administration (FDA) (Figure 2). Copanlisib (Aliqopa), an inhibitor of PI3K, predominantly against the isoforms PI3Kα and PI3Kδ, was approved by the FDA in September 2017 for the treatment of adult patients with relapsed follicular lymphoma who have received at least two prior systemic therapies (https://www.fda.gov/news-events/press-announcements/fda-approves-new-tre atment-adults-relapsed-follicular-lymphoma; 14 September 2017).
Isoform-selective inhibitors are emerging as next-generation of PI3K inhibitors with improved, precise targeting and reduced toxicity. This class of inhibitors has been developed to target cancers that are connected to one of the PI3K isoforms. The selective inhibition of specific PI3K isoforms may allow for the administration of therapeutic doses of drugs and have fewer toxicities compared to pan-PI3K inhibitors. On the other hand, isoform-specific PI3K inhibitors have the narrowest profile and may require careful patient selection based on potential biomarkers of sensitivity and resistance [151]. It was also shown that the acquired amplification and mutation of PIK3CA cause resistance to selective PI3K inhibitors [152]. It was proposed that PTEN loss can lead to clinical PI3Kα inhibitor resistance [153].
Idelalisib (Zydelig), a PI3Kδ inhibitor, was approved in July 2014. Zydelig is currently approved by the FDA for the treatment of relapsed chronic lymphocytic leukemia in combi-nation with rituximab and for the treatment of relapsed follicular B-cell non-Hodgkin's lymphoma or relapsed small lymphocytic lymphoma in patients who have received at least two prior systemic therapies (https://www.fda.gov/drugs/drug-safety-and-availability/fda-a lerts-healthcare-professionals-about-clinical-trials-zydelig-idelalisib-combination-other; 14 March 2016).

AKT Inhibitors
AKT could be a promising target in PI3K/AKT/mTOR pathway-activated tumors as one of the key effector nodes. However, the number of AKT inhibitors that have been explored in clinical trials is less than that of PI3K inhibitors. Most AKT inhibitors in clinical development inhibit AKT 1, 2, and 3 and are therefore termed as pan-AKT inhibitors. There are two distinct classes of AKT inhibitors currently in clinical development: ATPcompetitive and allosteric inhibitors. Allosteric inhibitors can prevent the localization of AKT with its PH domain to the plasma membrane, thereby blocking AKT phosphorylation and activation [154]. ATP-competitive inhibitors bind to the active conformation in which the PH domain has swung out from the kinase domain and expose the ATP-binding pocket [155,156]. Allosteric inhibitors are generally better than ATP-competitive inhibitors because they show less toxicity, reduced side-effects, and greater specificity [157].

mTOR Inhibitors
mTORC1 functions as a downstream effector for many frequently mutated oncogenic pathways including the PI3K/AKT and Ras/RAF/MEK/ERK (MAPK) pathways, and mTOR signaling is hyperactive in a range of 40-90% in different tumor entities [158], which makes mTOR an attractive target for cancer therapy.

The First Generation: Allosteric mTOR Inhibitors
Rapamycin and its analogs (rapalogs) are the first generation of mTOR inhibitors, which selectively inhibit the activity of mTORC1 by binding to FKBP-12 and forming a ternary complex with mTOR. Rapamycin is a macrolide, that is produced by the microorganism Streptomyces hygroscopicus and has shown antifungal properties [159]. Shortly after its discovery, immunosuppressive properties were detected, which later led to the establishment of rapamycin as an immunosuppressant [160,161]. In the 1980s, rapamycin was also found to have anticancer activity [162].
Rapamycin is an allosteric inhibitor of mTOR, and it inhibits some of the functions of mTORC1, such as the phosphorylation of the protein kinase S6K1. The clinical use of rapamycin is limited due to its poor water solubility and stability. Thus, several pharmaceutical companies have developed rapamycin analogs with improved pharmacokinetic properties ( Figure 2) (Table 7). Rapalogs differ in their chemical properties in terms of drug solubility and metabolism. For example, temsirolimus, a prodrug of rapamycin, and ridaforolimus (MK-8669) are water soluble and may be administered intravenously, whereas rapamycin and everolimus display low solubility and are therefore only available for oral formulations [163]. Rapalogs have been undergoing clinical trials for various malignancies and have already been approved by the FDA for the treatment of specific types of cancers (Table 7).
Though rapamycin inhibits mTOR with a high specificity, its effectiveness is dosedependent in different contexts. Two mTOR complexes, mTORC1 and mTORC2, have different sensitivities to rapamycin; different doses of rapamycin are needed to suppress mTOR in different cell lines, as well as the phosphorylation of different mTOR substrates, and these properties of rapamycin dosage can be largely attributed to the competition between rapamycin and phosphatidic acid for mTOR (reviewed in [164]).
While rapamycin inhibits S6K1, it does not fully inhibit 4E-BP1 phosphorylation, thus making it ineffective in blocking cap-dependent translation in most cell types [165]. Phosphorylated 4E-BP1 inhibits pro-oncogenic eIF4E. eIF4E-mediated translation is upregulated in tumors, and blocking this pathway may be crucial to preventing tumor growth in specific cancers [166][167][168]. On the other hand, the inhibition of mTORC1 may lead to the feedback activation of IGF-IR and AKT, which compromise the anti-cancer effect of rapalogs [169]. Rapalogs have proved more cytostatic than cytotoxic, perhaps because they also only partially block 4E-BP-dependent translation and fail to inhibit the pro-survival pathways regulated by mTORC2-AKT [170,171].
The combination of rapamycin with 5-aminoimidazole-4-carboxamide-1-β-4-ribofuranoside (AICAR) was reported to make rapamycin cytotoxic rather than cytostatic at doses that are clinically tolerated [172]. AICAR is a compound that activates AMPK, and a reciprocal regulation of PLD by AMPK and AMPK by PLD was demonstrated, i.e., the suppression of AMPK activity led to an increase in PLD activity and the suppression of PLD activity resulted in elevated AMPK activity [173]. The authors reported that the suppression of PLD activity by AICAR can improve the efficacy of rapamycin for both mTORC1 and mTORC2, and tolerable doses of rapamycin in combination with AICAR suppress both 4E-BP1 and AKT phosphorylation, as well as inducing apoptosis in cancer cells [172]. To more completely inhibit mTOR and target both mTORC1 and mTORC2, a number of ATP-competitive mTOR inhibitors have been developed. A second generation of mTOR inhibitors are small molecular ATP analogues that compete with ATP to occupy the kinase active site of mTOR. Differently from rapalogs, these molecules-also called selective mTOR kinase inhibitors (TORKIs)-ensure a complete block of both mTORC1 and 2, thus preventing AKT phosphorylation due to mTORC2 and avoiding the resistance observed in rapalogs.
In the majority of in vitro studies, compared with rapalogs, the ATP-competitive inhibitors have shown significantly higher inhibitory effects. Large scale trials have not yet been conducted to show greater and convincing efficacy than the currently available best treatment options; hence, TORKIs are still not approved by the FDA.

The Third Generation: RapaLink
Third generation mTOR inhibitors were synthesized and investigated to address the treatment resistance issues found in the use of the rapalogs and TORKIs. The new compounds are called Rapalink because they are made by the conjugation of rapamycin and ATP-competitive mTOR inhibitors, and they can bind both to FRB and the mTOR kinase domain [143]. This dual binding may serve to increase affinity and stability, both of which likely contribute to efficacy. Rapalink-1 has shown increased and durable inhibitory action compared to the first and second-generation inhibitors in glioblastoma and follicular lymphoma in vitro and in vivo [174][175][176]. A resistance of cancer cell cultures to rapalogs and TORKIs was shown to be overcome by the use of Rapalink [143].
The effectiveness of Rapalink-1 was also demonstrated in reducing prostate cancer tumor growth using an in vitro organoids assay and ex vivo tumor slice drug assays. The exposure of LAPC9 to Rapalink-1 was found to block mTORC1/2 signaling and reduce the fraction of CD44+ in vitro. Mice treated with Rapalink-1 showed a significantly delayed tumor growth, and cells recovered from the tumors of treated animals showed a marked decrease of CD44 expression [177]. The in vitro and in vivo therapeutic efficacy of Rapalink-1 against renal cell carcinoma (RCC) was evaluated and compared to temsirolimus. Rapalink-1 showed significantly greater effects against proliferation, migration, invasion, and colony formation in RCC cells. RNA sequencing showed that Rapalink-1 suppressed not only the mTOR signaling pathway but also a part of the MAPK signaling pathway, the ErbB signaling pathway, and ATP-binding cassette (ABC) transporters [178]. No large-sample clinical data have been reported for Rapalink.

Dual PI3K/mTOR Inhibitors
Though the inhibition of mTORC1 and mTORC2 can downregulate AKT S473 phosphorylation, mTOR inhibitors may paradoxically enhance the PI3K/PDK1 axis. Thus, an inhibitor targeting both PI3K and mTOR may have better anti-cancer activity compared to targeting mTOR alone [179,180].
PI3K and mTOR both belong to the PI3K-related kinase (PIKK) superfamily, and the catalytic isoform of the p110 subunit and mTOR have structural similarities; as a consequence, certain inhibitory compounds target both kinases. Dual PI3K/mTOR inhibitors are active against PI3K isoforms and both mTORC1 and mTORC2, thus targeting three most critical nodes of the same pathway. AKT activation would also be blocked by these inhibitors since PI3K blocking would diminish the production of PtdIns(3,4,5)P3, which acts as a docking site for AKT and PDK1. The inhibition of mTORC2 would also block the feedback activation of AKT. There are some promising trials with dual PI3K/mTOR inhibitors but only in phase 1 or 2 (Table 8). Therefore, more studies are needed to determine if the dual PI3K/mTOR inhibitors would be more effective than mTOR inhibitors.
Another possibility may be the dual inhibition with mTOR and AKT inhibitors to prevent the feedback activation of PI3K/AKT signaling after mTOR inhibition. Indeed, our own studies with a combinatorial treatment of hepatocellular carcinoma and cholangiocarcinoma cells with RAD001 and MK2206 showed synergistic effects on tumor growth in vitro and in xenotransplantation mouse models in vivo [181,182].

Summary and Conclusions
The discovery of mTOR is a fundamental breakthrough in the understanding of cell growth, metabolism, and diseases. Studies to determine the regulators and effectors of mTOR signaling have revealed multiple networks that interact together to integrate growth factor, nutrient, and nucleotide signaling. The understanding of the critical role of the mTOR pathway in tumorigenesis has driven to the development of a growing list of PI3K/AKT/mTOR inhibitors, though major clinical success has not been achieved.
The long-term use of the same inhibitor in tumors can lead to drug resistance, which is a main challenge in cancer therapy. Resistance can be developed by several mechanisms including the incomplete inhibition of mTORC1 functions, resistance mutations in MTOR, the compensatory activation of different pathways because of the mTORC1 inhibition, and the suppression of negative feedback loops (reviewed in [183]). More research to understand the molecular basis of mTOR networks and potential resistance mechanisms in mTOR-targeted cancer therapy are necessary to rationally apply mTOR inhibitors for the effective treatment of cancer.
A second problem comprises the clinical adverse effects associated with PI3K/AKT/mTOR inhibition that include hyperglycemia, hyperlipidemia, bone marrow suppression, pneumonitis, stomatitis, and hepatotoxicity [184]. Severe toxicities associated with PI3K/AKT/ mTOR inhibitors may limit the clinical application and approval of these agents. It is necessary to analyze the mechanism leading to the toxicities of PI3K/AKT/mTOR inhibitors that may help to develop optimal prevention and treatment strategies.
High-throughput molecular profiling, including next-generation sequencing, can give insights into the mechanism(s) of intrinsic sensitivity/resistance and the mechanisms of acquired resistance to PI3K/AKT/mTOR inhibitors. Work must be continued to validate predictive biomarkers, which may help to identify the patients who will most likely benefit from treatment with mTOR inhibitors and allow for the better determination of rational combinations of anti-cancer agents.

Conflicts of Interest:
The authors declare no conflict of interest.