The PI3K pathway is a complex signaling network coordinating a number of direct upstream inputs from growth factors [epidermal growth factor (EGF), tumor growth factor (TGF), and others], tyrosine kinase receptors [insulin growth factor 1 receptor (IGF-1R), epidermal growth factor receptor (EGFR), HER2], or other membrane receptors such as Met as well as a RAS-mediated crosstalk with the Ras-Raf-Mek-Erk pathway (Figure 1 ).

PI3K proteins are heterodimers composed of a catalytic p110 subunit (PIK3CA) and a regulatory p85 subunit (PIK3R) that mediate receptor binding, activation, and localization of the enzyme.

The direct binding to phosphotyrosine residues of growth factor receptors or adaptors through pleckstrin homology domains leads to the allosteric activation of the catalytic subunit and the phosphorylation of phosphatidylinositol 4,5-bisphosphate (PIP2) to the active second messenger PIP3. As a result, the PI3K complex is recruited to the plasma membrane and activates the pyruvate dehydrogenase kinase 1 (PDK1) and Akt proteins. Akt is a serine threonine kinase that regulates a large number of downstream targets [7,8], ultimately controlling critical cellular survival and metabolic processes[9]. For example, Akt regulates downstream p53, nuclear factor-kB (NF-kB), and cAMP response element-binding protein (CREB), leading to the transcription of anti-apoptotic and proliferative genes such as X-linked inhibitor of apoptosis protein (XIAP), Bcl-2, and survivin. Akt also promotes cell cycle progression via cyclin D1 up-regulation[10]. In addition, downstream signaling via mTOR increases the translation of target genes involved in angiogenesis [e.g., vascular endothelial growth factor (VEGF)] or the cell cycle (e.g., cyclin D1 and c-Myc)[11].

mTOR is most likely the best described downstream target of Akt[12–14]. The mTOR complex is composed of two components: the mTORC1-Raptor complex and the mTORC2-Rictor complex, the function of which is less well described. Importantly, mTORC1 is sensitive to inhibition by rapamycin, whereas mTORC2 is not; in addition, mTORC2 exerts a positive feedback activation on Akt. As discussed below, this positive feedback loop may have important implications regarding the emergence of resistance to first-generation mTOR inhibitors (rapalogs) that exclusively target mTORC1, with no effect on mTORC2.

Importantly, the cellular machinery also contains a number of endogenous negative regulators of the PI3K pathway. The PTEN protein acts as a pathway repressor by dephosphorylating PIP3 back to PIP2. The p85 regulatory subunit of PI3K also functions as a tumor suppressor; its major function is to bind, stabilize, and inhibit the p110 catalytic subunit. The liver kidney kinase B1 (LKB1) acts as a tumor suppressor by negatively regulating mTOR through the activation of protein kinase, AMP-activated, alpha catalytic subunit (AMPK), and the tuberous sclerosis complex 2 (TSC2) is a GTPase-activating protein that negatively regulates mTOR through the small G-protein Rheb. Together, LKB1, AMPK, and TSC2 constitute a stress-inducible pathway that normally counteracts PI3K/Akt signaling[15,16].

In cancer, this pathway can be aberrantly activated via a number of mechanisms. In addition to activation via upstream input, the PI3K pathway can be “intrinsically” up-regulated due to 1) gain-of-function mutations or amplifications in PIK3CA; 2) mutations in PIK3R1/2; 3) mutations or amplifications in one of the Akt isoforms (AKT1, AKT2, and AKT3); 4) loss of or inactivating mutations in the tumor suppressors TSC or LKB1; or 5) loss of the negative regulator PTEN via inactivating mutations, copy number loss, or promoter hypermethylation.

The PI3K/Akt/mTOR pathway is frequently deregulated in OC. Array comparative genomic hybridization (aCGH) studies have identified this pathway as the most frequently altered in OC[17]. Copy number changes in the genes encoding both the p110a (PIK3CA) and p110β (PIK3CB) subunits of PI3K have been associated with a poor prognosis in patients with OC. The expression levels of both PIK3CA and phosphorylated Akt (pAkt) were analyzed in over 500 OC and found to be associated with decreased survival, and activation of the pathway, as measured by Akt or mTOR phosphorylation levels, was found to be an independent negative prognostic marker in OC[18–20].

Mutations are much more prevalent in the rare subtypes of OC: 20% of endometrioid and 35% of clear cell OCs have documented PIK3CA mutations, whereas PTEN loss-of-function mutations are well documented in 20% of endometrioid OC[21].

Importantly, intrinsic activation of the pathway, via PIK3CA mutations and PTEN loss, has been shown to initiate ovarian tumors in mice, and inhibition of PI3K/mTOR in these models was found to delay tumor growth and prolong survival, thus providing critical proof of concept for the oncogenic relevance of this pathway in OC and its potential as a therapeutic target[22,23].

The first inhibitors of the pathway to enter the clinic were rapamycin analogs that bind to the mTORC1 complex and prevent mTOR activity. Rapamycin was used for years as an immunosuppressor to prevent rejection in solid organ transplants and hematologic malignancies[24]. Rapamycin analogs with less immunosuppressive properties, such as temsirolimus, everolimus, and ridaforolimus, have shown activity in a number of tumor types and have been investigated in OC.

In a phase II trial of OC treated with weekly intravenous injection of temsirolimus at a flat dose of 25 mg, objective responses were observed in only 9.3% of patients (5 of 54), and the 6-month progression-free survival (PFS) rate was 24%, causing the study to fail to meet its efficacy endpoint[25]. Exploratory analyses were also conducted to identify potential predictive markers. pAkt, p-mTOR, p-p70-S6K, and cyclin D1 were measured in archival tumor samples as surrogates for activation of the PI3K pathway, but only cyclin D1 levels were weakly correlated with a PFS of >6 months (r = 0.28). The authors concluded that the observed activity was insufficient to justify a phase III trial of temsirolimus in unselected OC patients. The trial included mainly serous tumors and only a few endometrioid (4 of 54, 7%) or clear cell ovarian tumors (3 of 54, 6%), the two subtypes most likely to demonstrate PIK3CA mutations. Interestingly, 1 of the 3 clear cell ovarian tumors had an objective partial response (PR) to temsirolimus. Recently, other encouraging data were reported for 5 patients with clear cell OC treated with temsirolimus, with 1 objective response lasting 14 months and 1 stable disease (SD)[26]. Although the numbers are too small to draw conclusions, these data together with the frequent PIK3CA mutations in clear cell OC suggest that PI3K inhibition may be a promising strategy in treating this OC subtype[27].

Given the limited activity of mTOR inhibitors alone and evidence from preclinical studies suggesting an additive benefit of mTOR inhibitors with chemotherapy, a number of studies have investigated the effects of mTOR-cytotoxic combinations. Phase I studies of temsirolimus with either weekly topotecan or pegylated liposomal doxorubicin (PLD) have been conducted in gynecologic malignancies, but the studies were small, and activity was limited[28,29]. Three other phase I studies of rapalogs in combination with chemotherapy (temsirolimus plus carboplatin/paclitaxel[30], ridaforolimus plus carboplatin and paclitaxel[31], and everolimus plus weekly paclitaxel[32]) also included a small number of patients with advanced OC, and the responses were described (3 of 6 patients with OC had a PR to temsirolimus plus carboplatin and paclitaxel). Another phase II trial evaluating the efficacy of temsirolimus and trabectedin in patients with recurrent clear cell carcinoma was performed, with no treatment discontinuations due to unmanageable adverse events[33]. Among the 17 patients enrolled, the objective response rate (ORR) was 18%, including 1 patient with complete response (CR) and 5 (29%) with SD beyond 3 months, resulting in a clinical benefit rate (CBR; CR + PR + SD > 3 months) of 47%[33]. However, given the small number and the use of different chemotherapeutics in the combination studies, it is difficult to draw conclusions regarding the true added value of an mTOR inhibitor.

Several hypotheses have been proposed to explain the disappointing results observed to date with mTOR inhibitors. First, data from other tumor types such as renal cancer suggest that mTOR inhibitors may have more of a cytostatic, as opposed to a cytotoxic, effect, with a benefit mainly in terms of disease stabilization and PFS improvement rather than tumor shrinkage. The phase I and II trials conducted to date with mTOR inhibitors in OC have been single-arm studies in which the primary efficacy endpoint was usually ORR.

Other hypotheses have also been proposed: 1) rapalogs may be poor inhibitors of the pathway due to feedback loops and early escape mechanisms, better pathway inhibitors may be needed; 2) mTOR inhibitors may have limited activity in an unselected population and require predictive biomarkers to select for the subset of patients most likely to benefit; and 3) PI3K inhibition alone may be insufficient, and combination strategies may be required. Each of these hypotheses is addressed in turn below, and strategies to improve efficacy are proposed.

In light of the heterogeneity of OC, predictive as well as pharmacodynamic biomarkers are urgently needed to select the patients most likely to respond.

Early studies have demonstrated greater anti-proliferative activity of PI3K pathway inhibitors in PTEN-null or PIK3CA-mutated OC cell lines[43–45], and Di Nicolantonio et al.[46] showed that PIK3CA mutations sensitized cancer cells to everolimus. Janku et al. [21] conducted mutational analyses on 140 patients with breast and gynecologic malignancies (including 60 with OC) enrolled in phase I trials of PI3K/Akt/mTOR inhibitors and demonstrated that the RR was higher in patients with PIK3CA-mutated tumors than in those without PIK3CA mutations (30% vs. 10%). The authors concluded that PIK3CA mutations predicted a response to PI3K inhibition. However, all “responders” were actually treated in trials of temsirolimus, bevacizumab, and liposomal doxorubicin. Given the known activity of bevacizumab and liposomal doxorubicin in OC and the fact that half of the responding patients had never been previously exposed to liposomal doxorubicin, mutations may simply be a good prognostic marker and associate with a response to treatment in general rather than to PI3K inhibitors in particular.

More recent clinical and preclinical studies have reported contradictory associations between PI3K mutations or PTEN loss and response to inhibitors of the PI3K pathway[47,48]. In particular, a significant number of PI3K-mutated tumors failed to respond, whereas a proportion of tumors lacking PI3K and PTEN alterations benefited from the treatment. This is in contrast to the much stronger association between activating mutations and response to targeted agents such as EGFR or BRAF inhibitors. In a pooled analysis of 3 trials of mTOR inhibitors in endometrial cancers, MacKay et al.[49] found that PI3KCA or PTEN alterations were frequent but did not associate with a response or benefit from mTOR inhibitors. This could be explained by the fact that the mutation analyses were conducted on primary tissue samples and as such may not accurately capture the genomic profile of relapsed disease.

Most biomarker studies to date have focused on PIK3CA mutations. However, the most common type of OC, HGSOC, is unlikely to exhibit these mutations. HGSOC is more prone to copy number alteration. Whether PIK3CA or AKT1/2/3 amplifications might confer sensitivity to PI3K inhibition in HGSOC has never been investigated. Mechanistic studies in preclinical models as well as CGH analyses of HGSOC tumor samples from patients treated in ongoing phase I/II trials of PI3K inhibitors would be therefore useful in this regard.

Finally, another approach has been the use of gene expression signatures of pathway activation as predictive biomarkers. For example, a low-RAS signature has been derived in silico from 3 publically available datasets and was shown to associate with sensitivity to the AKT inhibitor MK2206 in vitro[50]. These scores, derived from multiple gene signatures, may offer a more comprehensive read-out of pathway activation than individual mutations and, as discussed below, are being incorporated into phase I studies to select patients for PI3K inhibitor combinations. For now, however, it remains unclear whether a signature derived from a gene expression profile can provide a better measure of the level of activation of the PI3K pathway than individual genomic alterations.

In conclusion, although biomarker-positive (PTEN-negative or PIK3CA-mutated) tumors may demonstrate a higher response to PI3K inhibitors, a proportion of biomarker-negative patients also respond. In light of the imperfect association between alterations of the PI3K pathway and response to PI3K pathway inhibitors, most ongoing trials are enrolling an unselected patient population; the key will be to conduct detailed molecular analyses to identify candidate predictive biomarkers.

The mTOR/Akt/PI3K pathway is promiscuous and interacts with a number of other intracellular signaling networks, such as the IGF-1R or RAS pathways, which can allow treatment escape. For example, IGF-1R can become activated by mTOR inhibition (Figure 2 )[51]. Indeed, insulin receptor substrate-1 (IRS-1) is normally under basal negative regulation via phosphorylation by mTOR. However, mTOR inhibition abolishes IRS-1 phosphorylation, thus allowing IRS-1 to complex with IGF-1R and promote Akt signaling[52], and thereby generating a mechanism for escape from selective mTOR inhibition.

Pleiotropic interactions with other parallel pathways may also allow escape from PI3K inhibition. Akt has been shown to be phosphorylated via crosstalk with RAS. Thus, in KRAS-mutated tumors primarily driven by a constitutively up-regulated RAS pathway, PI3K pathway inhibitors alone are unlikely to be effective. This hypothesis is supported by studies demonstrating that KRAS- or BRAF-mutated tumors are resistant to PI3K inhibition[53]. Using a panel of cell lines, including OC cell lines, PI3K-mutated tumors were shown to be sensitive to everolimus, whereas dual PIK3CA/KRASor PIK3CA/BRAF-mutated tumors were found to be resistant[46]. Importantly, these authors also demonstrated that the knock-down of mutated KRAS in these cells restored everolimus sensitivity in vitro and in vivo. In mouse models and human tumors, everolimus increased Erk1/2 activation in post-treatment tumor samples, providing further evidence for crosstalk between the PI3K/mTOR and Mek/Erk signal transduction cascades[54]. Overall, these observations suggest that the selective targeting of one pathway may simply result in compensatory up-regulation in another, and vice versa.