Activation of the PI3K/mTOR/AKT Pathway and Survival in Solid Tumors: Systematic Review and Meta-Analysis

Background Aberrations in the phosphatidylinositol 3-kinase (PI3K)/mammalian target of rapamycin (mTOR)/AKT pathway are common in solid tumors. Numerous drugs have been developed to target different components of this pathway. However the prognostic value of these aberrations is unclear. Methods PubMed was searched for studies evaluating the association between activation of the PI3K/mTOR/AKT pathway (defined as PI3K mutation [PIK3CA], lack of phosphatase and tensin homolog [PTEN] expression by immunohistochemistry or western-blot or increased expression/activation of downstream components of the pathway by immunohistochemistry) with overall survival (OS) in solid tumors. Published data were extracted and computed into odds ratios (OR) for death at 5 years. Data were pooled using the Mantel-Haenszel random-effect model. Results Analysis included 17 studies. Activation of the PI3K/mTOR/AKT pathway was associated with significantly worse 5-year survival (OR:2.12, 95% confidence intervals 1.42–3.16, p<0.001). Loss of PTEN expression and increased expression/activation of downstream components were associated with worse survival. No association between PIK3CA mutations and survival was observed. Differences between methods for assessing activation of the PI3K/mTOR/AKT pathway were statistically significant (p = 0.04). There was no difference in the effect of up-regulation of the pathway on survival between different cancer sites (p = 0.13). Conclusion Activation of the PI3K/AKT/mTOR pathway, especially if measured by loss of PTEN expression or increased expression/activation of downstream components is associated with poor survival. PIK3CA mutational status is not associated with adverse outcome, challenging its value as a biomarker of patient outcome or as a stratification factor for patients treated with agents acting on the PI3K/AKT/mTOR pathway.


Introduction
Historically, the development of anti-neoplastic drugs has not focused on the targeting of specific molecular aberrations [1]. However, more recently, some targeted drugs have been developed against known oncogenes in selected patient populations. Examples of this are trastuzumab for HER2 over-expressing or amplified breast and gastric cancer [2,3], imatinib for chronic myeloid leukemia (CML) [4], vemurafenib for metastatic melanoma with (V600E) B-RAF mutations and crizotinib for non-small cell lung cancer patients with anaplastic lymphoma kinase (ALK) rearrangements [5,6]. In these examples, development of the drug was carried out in parallel with the identification of a biomarker that permitted the selection of patients with a higher chance of response.
The discovery and validation of biomarkers has become an integral part of successful drug development, with few drugs under development lacking associated biomarker programs. Such biomarkers are typically biological surrogates that can guide in the identification of patients with a higher probability of response. A biomarker can be a known oncogenic alteration like a gene mutation, an overexpressed protein or a protein that reflects the activation status of a signaling pathway [7]. Other types of biomarkers may help monitoring response to treatment or provide information about prognosis and outcome [7]. In this latter case, markers associated with worse outcome can be informative of a more aggressive phenotype potentially guiding the selection of more intensive treatment [8].
The phosphatidylinositol 3-kinase (PI3K)/mammalian target of rapamycin (mTOR)/AKT pathway has been linked to the pathophysiology of several neoplastic diseases [9,10]. Activation of this pathway can be a result of mutations in the PI3K or AKT genes, loss of phosphatase and tensin homolog (PTEN), or constitutive activation of upstream regulatory pathways such as receptor tyrosine kinases ( Figure 1) [9,10]. Given the prooncogenic role of the PI3K/AKT/mTOR pathway in cancer, it has become a target of interest for drug development. Inhibition of mTOR with rapalogs has shown clinical efficacy against some solid tumors, including everolimus for angiomyolipoma associated with tuberous sclerosis, metastatic renal cell carcinoma, breast cancer, or pancreatic neuroendocrine carcinomas and temsirolimus for renal cell carcinoma [11][12][13][14]. Many other agents in clinical development are designed to inhibit the PI3K/AKT/ mTOR pathway at different levels and include pure PI3K inhibitors, dual PI3K-mTOR inhibitors, AKT inhibitors or mTOR inhibitors [15,16]. Despite the approval of some drugs and the clinical development of other agents targeting the PI3K/ AKT/mTOR pathway, little is known about which patients are more likely to benefit from targeting this pathway. Similarly, the relationship between alterations of this pathway and a more aggressive phenotype is unclear.
Here we report a systematic review and meta-analysis of studies assessing the association of activation of the PI3K/AKT/mTOR pathway and clinical outcome in solid tumors.

Identification and selection of studies
This analysis was conducted in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [17]. Medline (Host: PubMed) was searched for studies published between January 2002 and December 2012, which evaluated the expression of components of the PI3K/AKT/ mTOR pathway and survival in solid tumors. We used the MeSH terms ''PIK3CA and cancer'' and ''PIK3CA-mTOR and cancer'' and ''PTEN loss and cancer'' adding the limitation of human studies. In addition we used the entry ''PIK3CA or mTOR'' and the name of each specific solid tumor (e.g. PIK3CA or mTOR and breast cancer) to recognize additional studies. The search was restricted to publications in English. Additional studies were identified through reviews of citation lists ( Figure S1 and figure  S2). Eligibility criteria were the availability of survival data for at least 5 years in relation to three types of pathway aberrations; mutations in the PI3K gene in any domain as measured by polymerase chain reaction (PCR) or other genomic techniques, the lack of PTEN expression by immunohistochemistry (IHC) or western-blot, or the evaluation of downstream components of the pathway like phospho-S6, mTOR, phospho-mTOR, AKT or phospho-4EBP1 by IHC. Studies reporting outcome of patients who had received a specific targeted agent against the PI3K/ AKT/mTOR pathway or related pathways were excluded as were studies reporting only disease free survival or cancer-specific survival.

Data Extraction
Two authors (VSC and AO) extracted information independently using pre-prepared data abstraction forms. The following details were extracted: tumor type, number of patients, duration of follow-up, mechanism for activation of PI3K/AKT/mTOR pathway (PI3K mutation, activation of mTOR/AKT or PTEN loss), methods used for the evaluation of pathway activation, and cut-off used for defining pathway activation. The outcome of interest was five-year overall survival (OS). In all cases, survival data were estimated from Kaplan-Meier curves independently by two authors (FV and MA).

Data Synthesis
The effect of any aberration in the pathway on overall survival was analyzed initially. Subsequently subgroup analyses were conducted to explore the relationship between survival and different components of the PI3K/AKT/mTOR pathway that were evaluated in each study. Group one was termed ''mTOR or AKT activation'' and included studies that evaluated downstream components of the pathway including phosphorylated proteins such as mTOR, AKT, S6, and others (see table 1). Group two was termed ''PIK3CA mutations'' and included those studies that

Statistical analysis
The proportion of patients surviving 5 years was estimated from the Kaplan-Meier curves for both normal (control group) and the presence of the molecular alteration (experimental group). The relative frequency of survival at 5 years between the control and experimental groups was expressed as an odds ratio (OR) and its 95% confidence interval (CI). Data were combined into a metaanalysis using RevMan 5.1 analysis software (Cochrane Collaboration, Copenhagen, Denmark). Estimates of ORs were weighted and pooled using the Mantel-Haenszel method. Cochran's Q (p, 0.10) and the I 2 index (.50%) were used to define inter-study heterogeneity. Due to significant heterogeneity, random effects modeling was used for all analyses. Analyses were conducted for all studies and differences between the subgroups were assessed using methods described by Deeks et al. [18]. All statistical tests were two sided, and statistical significance was defined as p,0.05. No corrections were made for multiple comparisons.

Description of studies
We identified 17 studies that evaluated activation of the PI3K/ mTOR/AKT pathway and survival in solid tumors. These studies comprised a total of 4746 patients with a median sample size of 279 patients. The characteristics of included studies are shown in table 1. Six studies evaluated the expression of the PI3K/AKT/ mTOR pathway in breast cancer, five in gastrointestinal tumors, three in gynecological cancers, and one each in prostate, non-small cell lung cancer (NSCLC) and oropharyngeal cancers. Six studies were included in the group called ''mTOR or AKT activation'', six in the ''PIK3CA mutation'' group and five in the ''PTEN loss'' group. The estimated median follow-up was 4.3 years (range = 2.6 to 16 years). The prevalence of pathway activation and the methods used for the analyses of molecular alterations of this pathway are shown in Table 2.

Association of pathway activation and survival by type of activation
The results of the subgroup analysis based on the specific part of the PI3K/AKT/mTOR pathway are shown in figure 3. For studies assessing PIK3CA gene mutations (n = 6), there was no association with worse 5-years survival (OR: 1.24; 95% CI 0.70-2.20, p = 0.46). In contrast, among studies evaluating activated  AKT (pAKTSer-473) pan-AKT(AbcamLid, Cambridge,UK); pAKT (Thr-308)(Cell Signallic Tecnology, beverly, USA); pAKT(Ser-473)(Biosource International Inc, CA,USA); Antibody specificity was checked by western blotting using a standard protocol The cut-off value for high and low levels of pAKT expression is defined as above and below the median histoscore.

Discussion
The identification of biomarkers that can inform the clinical behavior of a given tumor is important for patient education and treatment planning. In this study we explored the prognostic role of different components of the PI3K/AKT/mTOR pathway with the intention to identify tumors that rely on this molecular alteration. Such information may help to guide the clinical development of therapeutic strategies against this pathway.
Overall, there was evidence of an association between alterations in the PI3K/AKT/mTOR pathway and poor survival. When analyzing different components of the pathway we observed that those studies evaluating loss of PTEN and activated components of downstream proteins were linked with the poorest 5-years survival. Conversely, PIK3CA mutations were not linked with worse outcome in our analysis. The fact that PTEN is a major regulator of the activation of the PI3K pathway could explain its association with worse outcome [19,20] as those tumors with loss of PTEN expression could have activation of the different components of this pathway. In contrast, although PIK3CA mutations are considered driver mutations because they are linked to cell survival and increased proliferation, in our study there was no association with poor outcome [9,10]. PIK3CA mutations compromise numerous molecular lesions including those affecting both the catalytic and helical domains. Mutations in the helical domain may favor the oncogenic capability of PIK3CA by facilitating its interaction with certain signaling intermediates linked to the transmission of pro-oncogenic signals [21]. In the individual studies included in our analysis, mutations at different domains were pooled and this could decrease the statistical power needed to detect a worse outcome in this group for specific mutations. In addition, mutations can have different functional roles and different clinical behavior depending on the tumor type. For example, recent studies have shown that PIK3CA mutations are associated with different outcomes in breast cancer depending on whether the tumor is estrogen receptor positive or negative, and whether HER2 is over-expressed or amplified compared to HER2-normal [22,23]. Regardless of this, the data presented challenges the clinical relevance of PIK3CA mutations as unique measures of PI3K/AKT/mTOR pathway activation. This is relevant, as some ongoing clinical trials with agents that target this pathway are using such mutational analysis as an indication of pathway activity, with mutations being used as biomarkers for selection of patients undergoing experimental treatments with PI3K inhibitors. The analysis of studies evaluating downstream components by IHC showed a significant association with worse survival. These findings have substantial clinicopathological relevance, as evaluation of the activity of a protein appears more biologically relevant than the estimation of gene expression. Therefore, assessment of phosphorylated forms of signaling surrogates such as pS6 or pAKT may be more precise than evaluation of their total levels. However, results from these studies were heterogeneous as the markers evaluated belong to different components of the pathway (mTORC1 and mTORC2) [24].
These data may have relevance beyond prognostic value. Among solid tumors where targeted therapy has been developed against a known oncogene, presence of the oncogene has generally been associated with worse outcome. This is the case in HER2/ neu over-expressing or amplified breast or gastric cancers or BRAF-mutated melanoma. Consequently, it is possible that the effect of drugs targeting the PI3K/AKT/mTOR pathway will only be seen in patients where biomarkers consistently show a detrimental clinical outcome. Based on this hypothesis, it would be expected that PIK3CA mutations may not be associated with improvement in outcome from drugs targeting the PI3K/AKT/ mTOR pathway. This hypothesis is supported by data in breast cancer, which show little predictive value of PIK3CA with the mTOR inhibitor everolimus [15,16,25]. On the other hand, it is known that not all druggable molecular alterations in cancer are linked with worse outcome like the expression of estrogen receptors in breast cancer.
When analyzing the results by tumor type, alterations of the PI3K/AKT/mTOR pathway in breast cancer were not linked with worse outcome. However, most of these studies evaluated PIK3CA mutations. Conversely, studies in gynecological tumors and gastrointestinal cancers were more enriched in studies evaluating PTEN and protein markers of ''mTOR or AKT activation''; and these were linked with worse outcome. Only single studies in NSCLC, oropharyngeal and prostate cancers were available and these showed variable results. These studies were generally small and consequently reported wide confidence intervals which crossed the null boundary. Consequently, the relevance of these results in isolation remains unclear. The inconsistent measurement of pathway activation means that the independent effect of PI3K/mTOR/AKT activation in different tumor types cannot be evaluated with certainty Our study has limitations. This is a meta-analysis of the literature and is therefore more likely to be compromised by selection bias with enrichment for studies reporting positive results. In addition, there is also substantial intra-and inter-study heterogeneity including differences in biomarkers of interest and variability in PIK3CA domain mutations. Despite the use of statistical methods to reduce the effects of such heterogeneity, there remains uncertainty regarding the accuracy of the pooled estimates. Furthermore, hazard ratios were not reported by most studies and therefore we estimated the odds of death at 5 years instead. This is a less robust measure for survival, but was the only feasible method using the available data.
Finally, despite these limitations, results of this study do have some implications for both clinical and translational research. It is shown that the activation of the PI3K/AKT/mTOR pathway is related to poor outcome, and it is particularly relevant in gastrointestinal and gynecological cancers. In addition, the evaluation of PTEN levels ideally complemented with concomitant evaluation of the activation status of proteins such as pS6 and Figure 2. Odds ratio (OR) for 5-year overall survival (OS) in all studies. Forest plots of odds ratios for overall survival at 5 years based on activation of the PI3K/mTOR/AKT pathway. Odds ratios for each trial are represented by the squares, the size of the square represents the weight of the trial in the meta-analysis, and the horizontal line crossing the square represents the 95% confidence interval. The diamonds represent the estimated pooled effect based for each cohort individually (labeled subtotal) and for all cohorts together (labeled total). doi:10.1371/journal.pone.0095219.g002 AKT is linked with worse outcome probably identifying tumors that rely most on the PI3K/AKT/mTOR pathway. Checklist S1

Supporting Information
(DOC) Author Contributions Figure 3. Odds ratio (OR) for 5-year overall survival (OS) according to the expression of different components of the PI3K/mTOR pathway (group subtype). Forest plots of odds ratios for overall survival at 5 years split by subgroups defined by type of activation of the PI3K/ mTOR/AKT pathway. Odds ratios for each trial are represented by the squares, the size of the square represents the weight of the trial in the metaanalysis, and the horizontal line crossing the square represents the 95% confidence interval. The diamonds represent the estimated pooled effect based for each cohort individually (labeled subtotal) and for all cohorts together (labeled total). doi:10.1371/journal.pone.0095219.g003