The phosphatidylinositol 3-kinase (PI3K) pathway has an important role in cell metabolism, growth, migration, survival and angiogenesis. Drug development aimed at targetable genetic aberrations in the PI3K/AKT/mTOR pathway has been fomented by observations that alterations in this pathway induce tumour formation and that inappropriate PI3K signalling is a frequent occurrence in human cancer. Many of the agents developed have been evaluated in early stage clinical trials. This Review focuses on early clinical and translational data related to inhibitors of the PI3K/AKT/mTOR pathway, as these data will likely guide the further clinical development of such agents. We review data from those trials, delineating the safety profile of the agents—whether observed sequelae could be mechanism-based or off-target effects—and drug efficacy. We describe predictive biomarkers explored in clinical trials and preclinical mechanisms of resistance. We also discuss key unresolved translational questions related to the clinical development of inhibitors of the PI3K/AKT/mTOR pathway and propose designs for biomarker-driven trials to address those issues.
Agents targeting different components of the PI3K/AKT/mTOR pathway have been shown to be safe and well tolerated
These agents inhibit the PI3K/AKT/mTOR pathway at recommended doses, and are effective in multiple tumour types
No clear correlation among tumour type, genotype and sensitivity to inhibitors of the PI3K/AKT/mTOR pathway has emerged in initial clinical trials of second-generation inhibitors
Some unsolved questions in the late development of inhibitors of the PI3K/AKT/mTOR pathway might benefit from a systems biology approach and from biomarker-driven studies
Activation of the phosphatidylinositol 3-kinase (PI3K) pathway is commonly observed in human cancer and is critical for tumour progression and resistance to antineoplastic drugs, including cytotoxic chemotherapy and targeted agents.1 In addition, PI3K signalling might have a role in non-cancerous cells within the tumour microenvironment.2 Several molecular aberrations that affect key components of this pathway, such as the mammalian target of rapamycin (mTOR) or AKT, have prompted the development of PI3K pathway-specific agents.1 Following the discovery of rapalogues—which inhibit mTOR through an allosteric reaction—and the demonstration of their antitumour activity,3,4,5,6 rationally designed agents that targeted PI3K, AKT and the catalytic domain of mTOR were developed. These second-generation PI3K pathway inhibitors have been evaluated in early clinical trials, and are transitioning to more-advanced clinical development. Partial responses and prolonged disease stabilization have been reported from their use in multiple tumour types, including breast, gynaecological, prostate and lung cancers, and mesothelioma, sarcoma and lymphoma, although at a lower than expected rate.7,8,9,10,11,12,13 Some of the major challenges in developing agents targeting this pathway have been to determine the pharmacodynamic effects of these inhibitors; to determine whether these agents can inhibit the pathway efficiently, or whether downregulating signalling might be sufficient to produce a clinical response. It is still not known whether downregulation of PI3K alone will demonstrate a similar benefit to other targeted therapies directed against kinases such as KIT,14 EGFR,15 or BRAF.16
Here, we review potential differences among the PI3K/AKT/mTOR pathway inhibitors (PAM inhibitors), and unresolved issues related to these agents that will require further study.
Molecular biology of PI3K in cancer
The PI3K/AKT/mTOR axis regulates essential cellular functions including cell metabolism, growth, migration, survival and angiogenesis (Figure 1). In addition, the role of members of the PI3K/AKT/mTOR pathway—such as PI3K and phosphatase and tensin homolog (PTEN)—in processes such as epithelial-to-mesenchymal transition, DNA repair, autophagy, senescence, and stemness is being investigated.17
PI3K are divided into three subclasses on the basis of structure, regulation, and lipid substrate specificity.1 Of these, four different catalytic paralogues (α, β, δ, and γ) compose the class I PI3K and are related to cancer. Class IA isoforms are heterodimeric proteins formed of a p110 catalytic subunit and a p85 regulatory subunit and are involved primarily in the pathogenesis of human cancer. Within class IA, three genes, PIK3CA, PIK3CB, and PIK3CD, encode the homologous p110α, p110β, and p110δ isozymes, respectively. Class IB consists of PIK3CG, which encodes p110γ. Of these proteins, p110α and p110β are ubiquitously expressed, whereas p110δ and p110γ are mainly found in immune and haematopoietic cells.17
The oncogenic potential of the PI3K pathway is explained by two key observations. First, alterations in the PI3K/AKT/mTOR pathway can induce cell line transformation and tumour formation in transgenic mice.18,19,20 Second, PI3K signalling activation frequently occurs following multiple molecular alterations in other components of the pathway downstream of PI3K, such as mutations (PIK3CA, AKT1, PTEN), gene amplifications (PIK3CA, AKT1, AKT2), loss of expression of the tumour suppressors PTEN and inositol polyphosphate-4-phosphatase type II, and other mechanisms21 that are frequently observed in numerous human tumours (Table 1).1 In addition, deregulation of the PI3K/AKT/mTOR axis is also observed when upstream oncogenes are mutated or amplified—such as EGFR, HER2, hepatocyte growth factor receptor (MET), fibroblast growth factor receptors (FGFR1–3), and KRAS—or deletions in other tumour-suppressor genes occur, such as Von Hippel-Lindau disease tumour suppressor (VHL), tuberous sclerosis genes TSC1 and TSC2, the threonine-serine protein kinase LKB1, neurofibromin (NF1) and TP53.1 These mechanisms suggest myriad opportunities for drug development in different types of cancer centred on targetable genetic aberrations in the PI3K/AKT/mTOR pathway.
Clinical development of PAM inhibitors
Drug development of PAM inhibitors has increased during the past decade, as indicated by the number of ongoing clinical trials assessing the numerous compounds targeting PI3K, AKT, and mTOR (Figure 2). Rapalogues were the first inhibitors of PI3K pathway downstream effectors to enter the clinic.22 Although allosteric inhibitors of mTORC1 have shown efficacy as monotherapies, these have only been effective in some neoplasms, such as renal cell carcinoma (everolimus and temsirolimus),3,5 neuroendocrine tumours (everolimus),6 and mantle-cell lymphoma (temsirolimus).23 Preclinical studies demonstrated several negative feedback regulatory mechanisms24 and other possible mechanisms of resistance triggered after blockade of mTORC1 by rapalogues, which potentially attenuate the activity of these therapeutic antagonists when used in the single-agent setting. The most impressive data on rapalogues have been generated by combination trials with endocrine therapy in breast cancer. Following the observation that PI3K activation might contribute to resistance to anti-oestrogen treatment,25 anti-oestrogen therapies combined with mTOR inhibitors demonstrated impressive efficacy,4,26 engendering the development of second-generation PAM inhibitors.
Current PAM inhibitors in early development (Supplementary Table 1 online) include reversible ATP-competitive inhibitors of the four isoforms of class I PI3K (known as class I PI3K selective inhibitors or pan-PI3K inhibitors, such as GDC-0941, or XL147); the irreversible pan-PI3K inhibitor PX-866; isoform-specific inhibitors (such as the p110δ-selective inhibitor CAL-101, and the p110α-selective INK1117 and BYL719); dual pan-class I–mTOR inhibitors (SF1126, BEZ235, XL765, and GSK1059615); AKT inhibitors (allosteric inhibitors, such as MK-2206, and catalytic inhibitors, such as GDC-0068 and GSK690693); and mTOR inhibitors (rapalogues and catalytic inhibitors such as INK128, AZD8055 and OSI-027).
Dual PI3K–mTOR inhibitors and mTOR inhibitors
The experience with rapalogues led to the development of small molecules designed to inhibit the catalytic site of mTOR. These compounds inhibit mTORC1 and mTORC2 while avoiding the feedback loop mediated by overactivation of AKT.27 The limited antitumour activity of rapalogues is suspected to be related to the fact that these agents only inhibit the mTORC1 complex. Although this limited inhibition could be solved with the use of catalytic inhibitors of mTOR (inhibiting both mTORC1 and mTORC2), expanding the spectrum of antitumour activity, these agents are being developed in the same tumour types in which rapalogues have demonstrated activity (endometrial and kidney cancer, and lymphomas). Other second-generation compounds are dual pan-class I PI3K–mTOR inhibitors. Their dual activity is based on the structural similarities of the catalytic domain of mTOR and the p110 subunit of PI3K, providing the potential advantage of targeting the pathway at two levels (suppressing mTOR in both the mTORC1 and mTORC2 complexes, and PI3K; Figure 1).28
PI3K selective inhibitors
An alternative strategy is the development of ATP-competitive pan-PI3K selective inhibitors. These could be directed at tumours with PIK3CA mutations that are 'addicted' to the PI3K pathway and whose growth and survival depend on mutations in this specific oncogenic pathway. Although isoform-specific PI3K inhibitors are expected to produce greater target inhibition with fewer adverse effects,29,30 the therapeutic window of pan-PI3K and isoform-specific PI3K inhibitors is likely to be limited as they lack selectivity for mutant isoforms. BRAF inhibitors have been developed with selectivity for mutant isoforms,31 but this has not yet occurred with PI3K inhibitors.
Some inhibitors of AKT are being tested clinically,13,32,33 although the development of AKT-specific and isozyme-selective inhibitors was predicted to be difficult. The anticipated hurdle was the high degree of homology in the ATP binding pocket between AKT, protein kinase A (PKA) and protein kinase C (PKC). GSK690693 and GDC-0068, for example, are ATP-competitive AKT inhibitors that inhibit the three AKT isoforms (AKT1, AKT2, and AKT3). By contrast, allosteric inhibitors seem to be more AKT-specific. For example, MK-2206 is a potent, allosteric non-catalytic pan-AKT inhibitor, and perifosine is an allosteric inhibitor that targets the pleckstrin homology domain of AKT, thereby preventing its translocation to the plasma membrane required for activation.34 The potential role of AKT inhibitors is being investigated. Although some cancers with AKT1 mutations and AKT1 and AKT2 amplifications might be expected to be among the most sensitive to AKT inhibitors, it seems that PIK3CA mutant cancer cell lines exhibit only minimal AKT activation and the role of these drugs in this setting should be considered carefully.35
The next wave of drugs targeting the PI3K/AKT/mTOR pathway soon to be tested in the clinic are isoform-specific PI3K inhibitors targeting PI3Kβ, inhibitors of ribosomal protein S6 kinase beta-1 (S6K), PDK1 inhibitors and isoform-selective AKT kinase inhibitors (thus far, no inhibitors have been developed to target protein kinase AKT1 versus protein kinase AKT2 independently). Figure 1 shows representative examples of early clinical and translational data related to PAM inhibitors and Figure 2 shows the status of their development.
Safety of PAM inhibitors
Overall, the profile of PI3K inhibitors regarding adverse events has been acceptable, with no unexpected toxic effects. Toxic effects have been primarily mild to moderate and manageable with supportive medication.7,13 Dose-limiting toxic effects reported with multiple agents include hyperglycaemia, maculopapular rash, gastrointestinal intolerance (anorexia, nausea, vomiting, dyspepsia, diarrhoea), and stomatitis.8,36 Although some of these toxic effects are 'off-target' effects, others may be related to target engagement and directly related to mechanisms of action.37 These mechanism-based toxic effects could be used as pharmacodynamic biomarkers and for clinical decision-making, including dose titration in cases in which there are no side effects; for example, dose escalation until the outbreak of rash in the development of EGFR inhibitors.38
Treatment with PAM inhibitors has resulted in hyperglycaemia, which has been manageable with metformin and, occasionally, with subcutaneous insulin.39 Hyperglycaemia is a mechanism-based toxicity of PAM inhibitors. Because PI3K is a key component in the insulin signalling pathway, inhibition of insulin signalling leads to insulin resistance and increased glucose levels in blood.39,40,41 However, a caveat for using hyperglycaemia as a pharmacodynamic biomarker is that there is a compensatory release of insulin (and C-peptide) and other factors, such as diet, that can influence glucose levels.36
Diarrhoea and skin rash have frequently been observed with PAM inhibitors. The skin rash observed in patients treated with PAM inhibitors is erythematous, nonblistering, maculopapular, and not acneiform, differing from the rash observed with EGFR–targeted agents.9,13 Interestingly, a similar type of rash—as well as diarrhoea—has been observed in patients treated with other kinase inhibitors, such as sunitinib.42 Dual PI3K–mTORC1/2 inhibitors can cause a wider range of off-target adverse effects, such as fatigue and elevated transaminases, but these do not seem to limit their clinical utility.8,12,36 Interestingly, some of the expected toxic effects that are frequently associated with rapalogues—such as hyperlipidaemia, pneumonitis and mucositis—are not associated with PAM inhibitors. For example, the prevalence of pneumonitis is approximately 20% in patients treated with rapalogues, but has rarely been described with dual PI3K–mTOR inhibitors.8,36 Research on the mechanisms responsible for rapalogue-associated toxic effects, whether they are associated with the immunosuppressant activity of rapalogues, mTORC1 inhibition or their biochemical structure, could help in elucidating the differences among classes of PAM inhibitors.
Biomarkers of cell signalling
The pharmacodynamic effects of PAM inhibitors on multiple PI3K pathway nodes have been assessed in surrogate tissues, such as peripheral blood mononuclear cells (PBMC), platelet-enriched plasma, skin, and hair, as well as in tumour biopsy specimens (Table 2). PBMC lysates or platelet-rich plasma have been used to investigate the extent of PI3K inhibition with GDC-0068, GDC-0941, GDC-0980, MK-2206 (weekly regimen) and PX-866.11,33,43,44 Blood markers are convenient because they facilitate the evaluation of dose–time relationships through serial sampling. Even so, technical and scientific challenges remain that could render these methods insufficiently robust; for example, drug concentrations in blood are expected to be higher than in tissue; therefore, the correlation with tumour effects is not clear. By contrast, human hair has provided a better outcome than skin, and could allow serial sampling for future dose, antitumour response, and duration of effect analyses.45,46
The readouts most frequently evaluated are phosphorylation of AKT at the residues Thr308 (PI3K activity readout) and Ser473 (mTORC2-dependent phospho-epitope); phosphorylation of the AKT substrate AKT1S1 (also known as PRAS40) at Thr246 (AKT activity readout), phosphorylation of eukaryotic translation initiation factor 4E-binding protein 1 (4EBP1) at Ser65 and Thr70 (mTORC1-activity readout), and phosphorylation of ribosomal protein S6 (RPS6) at Ser240 and Ser244 (mTORC1/S6K-activity readout; Figure 1 and Table 2). Quantification of a biomarker in proximity to PI3K—such as levels of PIP3—by mass spectroscopy or ELISA (enzyme-linked immunosorbent assay), does not seem feasible in the clinic. AKT assessment would be the next best option but, in our experience, measuring AKT in tissue requires very stringent conditions for sampling and handling to avoid variability in the context of multicentre clinical trials. Thus, assessment of phosphorylation of RPS6, 4EBP1 or pPRAS40 have been preferable in clinical biomarker development.7,11,36 Nevertheless, phosphorylation of RPS6 and 4EBP1 can also be regulated by enhanced RAS/RAF/ERK/mTORC1 activity, thereby potentially masking biomarker pharmacodynamic effects.47 Other studies have investigated the potential pharmacodynamic effects of PAM inhibitors on Ki-67, a marker of proliferation, and on the apoptotic marker TUNEL.9 Biomarker analysis from completed studies reflects a dose-dependent decrease (between 30% and 90%) in markers such as pRPS6, pAKT and 4EBP1 when used at the maximum tolerated dose (MTD) of the different compounds (Table 2), demonstrating target modulation.
Biomakers of metabolic effect
On the basis of the role of the PI3K pathway in physiological glucose metabolism and of Warburg's observations that tumour cells have a higher rate of glucose consumption, some studies with PAM inhibitors have included biomarkers of metabolic effect.41,48 Consequently, assessment of the dose–response relationship has been attempted by measuring fasting glucose, insulin and C-peptide levels in plasma, and glucose uptake using 18F-fluorodeoxyglucose (FDG)–PET scans.7 The assessment of plasma C-peptide has served as a surrogate marker to demonstrate dose-dependent PI3K inhibition, an effect that was observed after only a few days of treatment.7,36 Data from early clinical trials demonstrate that measuring C-peptide can be used to delineate a biologically active dose when comparing different dose levels, but diurnal fluctuations confound its use as a biomarker for decision-making for individual patients.36 A decrease in FDG uptake has been observed in several phase I trials following PAM inhibitor treatment.7,11,49 Whether the cause of this decrease is related to a pharmacodynamic effect of PI3K inhibition (the role of PI3K signalling in glucose uptake) or an antitumour effect is not known yet, and in some cases, both can have a role, as seen with mTOR inhibitors.50,51
Signs of activity and predictive biomarkers
Signs of antitumour activity (partial responses and prolonged disease stabilization with minor tumour shrinkage or tumour marker reduction) have been reported in the phase I trials of the first PAM inhibitors, including breast, non-small-cell lung, gynaecological, pancreatic, bladder, anal, and prostate cancer, renal cell carcinoma, melanoma, mesothelioma, sarcomas and lymphomas.52 With few exceptions, the clinical efficacy of many PI3K pathway inhibitors as a monotherapy is modest at best. However, the selective agent, CAL-101—a highly specific PI3Kδ inhibitor—has shown significant clinical activity in several lymphoid malignancies.10
One strategy to increase monotherapy response rates is to improve the selection of patients with known aberrations and mutations. Preclinical findings showing that tumours with PI3K pathway alterations are more sensitive to treatment with PI3K inhibitors than tumours without such alterations have encouraged the enrichment of trials with patients whose tumours harbour mutations in PIK3CA and PTEN, or have lost expression of PTEN.53,54,55,56 However, initial clinical trials with PAM inhibitors have shown no clear correlation between molecular alterations in the PI3K pathway and antitumour effect;7,13,44 although some pooled analyses have suggested that there is a correlation,57 others have suggested otherwise.58 Moreover, partial responses and significant tumour shrinkage were observed in KRAS mutant breast, ovarian and pancreatic cancer as well as BRAF-mutant melanoma patients following treatment with a PAM inhibitor,11,13,36,59 an unexpected finding as tumours with activating mutations in members of the MAPK pathway may potentially be resistant to PAM inhibitors.60,61,62,63
Once a recommended dose was established in some phase I studies (such as those studies assessing the inhibitors BKM120 and BEZ235), preselected patients with solid tumours and with PI3K pathway alterations were enrolled in the expansion phase of those studies to test the value of these alterations as predictive biomarkers.7,64 Although no data are yet available to show whether clinical responses are higher in this molecularly selected patient population, the response rates in selected versus unselected patients have not, in our experience, been markedly different. Despite the fact that selectivity towards a population harbouring mutations in PI3K has not been proven, several phase I studies assessing isoform-specific PI3K (such as BYL719 and INK1117) have restricted enrolment to only patients with PIK3CA-mutant tumours. Preliminary results of the first-in-human study of BYL719 have shown some objective responses and prolonged disease stabilization with tumour shrinkage in breast (among PIK3CA-mutant patients with oestrogen-positive breast cancer, 33% achieved a tumour shrinkage of more than 20%), colorectal (one partial response) and gynaecological malignancies (one partial response).59
The inconclusive predictive value of PIK3CA, PTEN, KRAS or BRAF mutations for delineating the clinical value of PAM inhibitors may be due to several reasons. First, tumours without PI3K alterations might have responded because early detection methods were based on a limited number of assays and were unable to detect other alterations that could be driving sensitivity such as alterations in AKT1/2, PIK3R1, LKB1, or NF1, or were using improper assays or thresholds (for example, detection of PTEN loss lacks a well-defined threshold for immunohistochemistry categorization and neither the antibodies used for immunohistochemistry nor the criteria for assessment are uniform among different laboratories). Second, tumours described as having PI3K alterations might have not responded because of the coexistence of mutations related to resistance, such as mutation in KRAS,61,62 intratumour heterogeneity,35 or the use of incorrectly standardized or unvalidated assays. Finally, there are conflicting preclinical data55,65,66 regarding the preferential activity of PAM inhibitors in specific molecular scenarios.
In summary, it seems that PAM inhibitors in monotherapy have not shown the same level of success as targeted therapies, such as kinase inhibitors directed towards ALK and BRAF. Clinical responses to PAM inhibitors as monotherapy are less frequent than had been expected and it seems that translating PAM alterations to the clinic as predictive markers for patient selection might not be as straightforward as originally hoped.
Stop and think: unresolved questions
After the first round of clinical trials with second-generation PI3K pathway inhibitors, several sets of important questions remain unanswered and may determine the future development of these agents (Box 1).
Because most phase I trials with these agents are considered 'completed' and these agents are now in the later stages of development, the pivotal unanswered questions are related to unsolved pharmacological issues. First, on the basis of clinical activity and toxicity profiles known to date, what remains to be defined is whether current PI3K, AKT or mTOR1/2 inhibitors are optimal drugs, and if there is a specific genetic context that prognosticates superior activity for each class. The magnitude of signalling inhibition required to produce biological and clinical effects has not been clearly delineated either. Whereas responses to selective BRAF inhibitors in melanoma were well correlated with an inhibition of intratumoural pERK pathway activity >80%, the same level of certainty is not available in the case of PI3K inhibitors.67 Based on the observed pathway inhibition (in the 30–90% range depending on the marker and the drug assessed), it is not known if these drugs when administered at the MTD are inhibiting the pathway sufficiently to have an antitumour effect. Off-target toxic effects—or an incorrect assessment and management of on-target, mechanism-based toxic effects—could preclude attaining profound pathway inhibition. Therefore, questions remain as to whether a therapeutic window can be achieved without inducing deleterious side effects, such as insulin resistance, and if this toxicity is acceptable in the context of the antitumour activity produced by PI3K inhibitors. Interestingly, some preclinical studies have shown that modulation of AKT is necessary, but not sufficient, to ensure tumour growth inhibition.68 These uncertainties pose unanswered questions regarding the therapeutic index of PAM inhibitors and their recommended phase II doses: should we select the MTD or the biologically active dose? Probably most important: would it be best to inhibit the pathway continuously or with a more-potent intermittent approach?
An opportunity for combinations
Because the PI3K pathway comprises a complex network of interactions with parallel cascades, its pharmacological inhibition releases negative feedback resulting in activation of compensatory signalling pathways. Among these compensatory mechanisms are a FOXO-dependent feedback reactivation of receptor tyrosine kinases (such as HER2,69 HER3,70 IGF1R, and insulin receptor71) and downstream kinases including ERK,69,72 MYC amplification, and NOTCH73,74,75 or Wnt–β-catenin pathway activation.76 Although these findings derive from in vitro experiments and indicate resistance to PI3K inhibition, further clinical validation is needed for more-definitive conclusions.
To overcome negative effects on PI3K pathway inhibition, in vivo combination therapy with receptor tyrosine kinase inhibitors (such as those directed against EGFR, HER2, or HER3) resulted in decreased proliferation, enhanced cell death and superior antitumour activity compared with single-agent PI3K inhibitors.69,70,71 Some of these combination strategies, such as trastuzumab or lapatinib in combination with PAM inhibitors, are being tested in ongoing clinical trials.77,78
Using horizontal combination approaches that include PAM inhibitors and MEK inhibitors has been proposed to overcome the parallel induction of the MAPK pathway by PAM inhibitors.79 These combinations of small-molecule kinase inhibitors might, however, be challenging, particularly because of the increased levels of toxicity associated with them (the rates of drug-related grade 3 or 4 adverse events are around 18.1% for monotherapy and approximately 53.9% when combining a MAPK and a PI3K inhibitor, with transaminase elevations, skin rash, and mucositis the most frequent ones),80 which could require using inadequate doses that do not provide sufficient dual pathway inhibition.81,82,83,84
Cell signalling changes resulting from PI3K alterations might not be limited to activation of receptor tyrosine kinases and MAPK pathway. In hormone-sensitive tumours, such as prostate and breast cancer, PI3K pathway alterations are triggered as resistance mechanisms to hormone therapies. In PTEN-deficient prostate cancer cells, PI3K and androgen receptor (AR) cross-regulate each other by a reciprocal feedback mechanism (inhibition of the AR activates AKT signalling, whereas inhibition of PI3K results in feedback signalling to HER2–HER3 and activation of AR), which is abrogated by combined pharmacological inhibition.18 Clinical trials combining androgen deprivation therapy with PAM inhibitors are currently recruiting patients.85,86,87 A similar phenomenon is observed in oestrogen receptor-positive breast cancer cells.88 Following the positive results of the BOLERO-2 trial of everolimus administered in combination with exemestane, many PAM inhibitors in combination with anti-oestrogen therapies are undergoing clinical evaluation in endocrine-sensitive breast cancer.89 Preclinical studies have also suggested that PI3K modulation could overcome resistance to chemotherapy agents, such as platinum-based drugs and taxanes, doxorubicin and 5-fluorouracil.90 In preclinical models, concomitant inhibition of the PI3K pathway with chemotherapy was shown to enhance the efficacy of some agents by decreasing the phosphorylation of AKT that can be raised in response to insult as a survival mechanism.91,92 However, this synergistic effect is not universally true. Inhibition of the PI3K pathway results in arrest in the G1 phase of cell cycle, which could decrease antitumour effect of agents whose activity induces arrest in later cell cycle phases.93,94,95 Approximately 30% of ongoing phase Ib–II trials using PAM inhibitors are investigating combination regimens that include chemotherapeutic agents (Figure 2). Preliminary results of ongoing trials with GDC-0941, PX-866 and GDC-0068 added to taxanes and platinum agents showed good overall tolerance at the MTD of both agents and responses were seen in multiple tumour types.96,97,98
Translational efforts to move forward
Given the moderate clinical benefit observed after exposure to PAM inhibitors as monotherapy, patients with demonstrated clinical responses and their tumour biopsies have become tremendously valuable as a vehicle to identify mechanisms of sensitivity. Additionally, on-treatment tumour biopsies would be useful for analysing mechanisms of action and defining an optimal biological dose. Therefore, we believe that including on-treatment biopsies in upcoming clinical trials is crucial for the further development of PAM inhibitors.
Other questions—such as what should be used as a predictive marker or how to match a drug target with a predictive marker—are also important and complex enough to warrant testing in specially designed proof-of-concept trials.99 Tissue-based novel phase II designs can provide information that can be used to design phase III trials. Biomarkers in different genetic contexts (such as the effect of the drug on kinase inactivation, gene expression, cell proliferation, apoptosis, and inhibition of the drug target in situ [by assessing pAKT, pPRAS40, pRPS6, and others]) can be studied in biomarker-driven trials (with paired biopsies) or in the neoadjuvant setting.100 If found, these answers could be used later for patient selection or exclusion from registration trials. This approach was feasible and useful in exploring the effect of the aromatase inhibitor letrozole combined with everolimus.4,101
Identifying reliable biomarkers for patient selection is of paramount importance. Tumour-based biomarkers could be expanded not only to target genomic aberrations (PIK3CA, PTEN, AKT1/2, PIK3R1, LKB1 and NF1), but also to include 'PI3Kness' status, which might serve as a readout of all activation alterations.102 In addition, non-tissue-based biomarkers are rapidly developing. One approach would be to determine noninvasively current mutation status in circulating DNA found in the plasma of patients with metastatic disease. This assessment is particularly important because the mutational pattern of tumours can change at the time of disease recurrence.103
Lacking a qualified and defined drug target-predictive marker relationship,100 one could opt for an 'all-comers' approach with a prespecified retrospective assessment of potential biomarkers. Advantages of a preselected patient population (enrichment approach) are, however, straightforward resulting in proof-of-concept, quick determination of responsive patient population and anticipated clinical benefit if the preclinical hypothesis holds true.104
In disease settings for which the benefit of first-generation mTOR inhibitors has been defined, direct comparison with second-generation agents (or combinations, based on solid preclinical data) is also appealing. Therefore, research efforts will benefit from addressing the differential effects, if any, of each class and the genetic contexts to which different drugs are best suited. An interesting approach for clinically testing PAM inhibitors in breast cancer, for example, is the 'add-on phase II trial design', in which patients on standard therapy (such as anti-HER2 or endocrine therapy) undergo biopsies at the time of disease progression and then receive different matched targeted agents that are expected to reverse resistance with each patient acting as her own control.
With the plethora of available agents, it seems daunting to address all related questions through clinical trial testing. This field will benefit greatly from a systems biology approach, the development of patient-derived xenografts that allow co-clinical trials, and current developments in bioinformatics and public databases.105,106,107
On the basis of the considerations described in this Review, it is reasonable to state that PAM inhibitors have the potential to inhibit the PI3K/AKT/mTOR pathway with reasonable efficacy and a favourable safety profile. The observed antitumour activity of mTOR inhibitors as monotherapy or in combination, and the potential indications that could be explored based on the frequency of alterations in the pathway clearly encourage further investigation with PAM inhibitors. To succeed in the late development of these agents, pharmacological, biological and translational issues need to be examined to better understand how to appropriately apply these agents.
Remarkably, each agent class and compound seems to be slightly different regarding its biological effect, but it is unknown whether mTOR kinase, AKT, pan-PI3K, or isoform-specific PI3K inhibitors will provide the greatest therapeutic index, or whether they will need to be combined with inhibitors of receptor tyrosine kinases or downstream targets. It is likely that the applicability of the different agent classes may depend on tumour type, genotype and therapeutic index of the combinations. Innovative, biomarker-driven studies are needed to answer some of these questions, avoiding as much as possible the serendipitous and empirical approaches that have been used in the development of many targeted therapies in the past.
The PubMed database was searched for articles published in English before October 2012. The search terms included: “PI3K”, “mTOR”, “AKT”, “cancer”, “clinical trial”. We considered articles published only in English and selected all the trials that included a PI3K/AKT/mTOR (PAM) pathway inhibitor. We also selected preclinical papers with the most robust and referenced data. Finally, we also considered abstracts released at the AACR and ASCO annual meetings and EORTC-NCI-AACR Symposium on Molecular Targets and Cancer Therapeutics in 2010, 2011, and 2012. The ClinicalTrials.gov website was searched for ongoing clinical trials with PAM inhibitors. This Review includes a summary of the authors' work and knowledge based on participating in more than 10 phase I trials with PAM inhibitors. Knowledge gained from reading the oncology literature and regular attendance at conferences, workshops, and other national and international meetings was also included.
The authors acknowledge José Baselga, Lew Cantley, and other members of the “Stand Up to Cancer Dream Team” Translational Research Grant for sharing their insights on targeting PI3K. In addition, Joann Aaron (The University of Texas MD Anderson Cancer Center) for editing this manuscript.
Development status of PI31/AKT/mTOR pathway inhibitors
About this article
Deoxyshikonin isolated from Arnebia euchroma inhibits colorectal cancer by down-regulating the PI3K/Akt/mTOR pathway
Pharmaceutical Biology (2019)