The high protein expression of FOXO3, but not that of FOXO1, is associated with markers of good prognosis

To better define the role of FOXO1 and FOXO3 transcriptional factors in breast carcinogenesis, we performed a comparative study of their expression at both the RNA and protein levels in a series of human breast tumors. We used qRT-PCR assay to quantify mRNA expression and Reverse Phase Protein Arrays (RPPA) to quantify protein expression in 218 breast tumors from patients with known clinical/pathological status and outcome. Weak correlations were observed between mRNA and protein expressions for both FOXO1 and FOXO3 genes. High expression of FOXO3 protein, but not FOXO1 protein, was a good prognostic marker, negatively correlated with KI67 and markers of activity of the PI3K/AKT/mTOR oncogenic pathway, and positively correlated with p53, a marker of apoptosis. Moreover, FOXO3 protein expression, but not FOXO1 protein expression, was also negatively correlated with various proteins involved in different DNA repair mechanisms. FOXO3 protein, but not FOXO1 protein, appears to be a tumor suppressor that inhibits breast cancer by altering DNA damage response (DDR), thereby inducing p53-dependent apoptosis. This antitumor effect appears to be suppressed by excessive activity of the PI3K/AKT/mTOR pathway. High FOXO3 protein expression could be a biomarker of deficient DDR in breast tumors.

cellular pathway induces the phosphorylation of FOXO proteins by AKT. This phosphorylation leads to the exclusion of FOXO proteins from the nucleus, inhibiting therefore their capacity to modulate the transcription of their target genes 4,10,11 . This phosphorylation has also been shown to promote the degradation of FOXO3 protein via the proteasome 12 . In human tumors, the PI3K/AKT/mTOR cellular pathway is frequently found over activated leading therefore to inhibition of the transcriptional activity of the FOXO proteins 5,13 .
Various studies strongly suggest that FOXO proteins are tumor suppressors. The conditional deletion of all FOXO1/3/4 alleles in adult mouse tissues induces the development of lymphoblastic thymic lymphomas and hemangiomas 14 . Overexpression of FOXO1 and FOXO3 proteins in breast cancer has been shown to inhibit the growth of breast cancer cells [15][16][17][18] . IκB kinase and ERk promote breast carcinogenesis via inhibition of FOXO3 protein expression [9]. The cytoplasmic expression of the FOXO3 protein is positively correlated with poor survival in breast cancer 15 . Low expression of FOXO1 or FOXO3 protein in breast tumors is correlated with poor clinical outcome 19,20 . Altogether, these results strongly suggest that FOXO1 and FOXO3 proteins act as tumor suppressors in breast cancer. However, other studies have described unexpected functions for these two FOXO proteins in resistance to breast cancer treatment and breast cancer promotion [21][22][23][24] . Notably, FOXO1 and FOXO3 proteins have been implicated in the promotion of breast tumor cell invasion 21,22 . FOXO3 protein expression has also been associated with poor survival in breast cancer 23,24 .
In order to better define the role of FOXO1 and FOXO3 proteins in breast cancer, we performed a comparative study of their RNA and protein expressions in 218 breast tumors by using real-time quantitative reverse-transcription polymerase chain reaction (qRT-PCR) and Reverse Phase Protein Arrays (RPPA) methods, respectively. We also determined the correlations between FOXO1 and FOXO3 protein expressions and classical clinical biological parameters, as well as the expression of proteins involved in the PI3K/AKT/mTOR pathway, DNA damage response (DDR), apoptosis, cell cycle, and/or cell proliferation.

FOXO1 and FOXO3 RNA and protein expressions in breast cancer.
To better define the role of FOXO genes in breast cancer, we analysed their RNA and protein expression in a series of 218 breast tumors (clinical parameters presented in Table S1).
In keeping with our previous work 25 , we found that the expressions of FOXO1, FOXO4, and FOXO6 at the RNA level varied widely in breast tumors compared to normal breast samples (Table S2). In general, FOXO1 and FOXO4 mRNA were significantly underexpressed (expression values < 0.33), whereas FOXO6 mRNA was overexpressed (expression values ≥ 3), compared to normal breast tissue (see Material and Methods page 11). FOXO4 expression in the HR + ERBB2 + subtype and FOXO6 expression in the HR-ERBB2-subtype were similar in tumors and normal breast tissue. FOXO3 gene expression was similar in tumors and normal breast tissue except in the HR + ERBB2 + subtype, in which it was slightly, but significantly overexpressed.
The RPPA method demonstrated a wide range of FOXO1 protein expression (ranging from 0.07 to 1.91) and FOXO3 protein expression (ranging from 0.19 to 3.98) in breast tumors (Table 1). FOXO4 and FOXO6 protein expression was not studied due to the lack of an appropriate specific antibody for these two proteins.
We examined the correlation between protein and mRNA expressions for FOXO1 and FOXO3 genes. We observed weak positive correlation for the FOXO1 gene (r = +0.342, p < 10 −4 ) and a weak positive correlation for the FOXO3 gene (r = +0.150, p = 0.027), strongly suggesting that the expression of these two genes in breast tumors is regulated by molecular mechanisms independent of transcription and RNA stability (Fig. 1). These results also highlight the importance of considering protein expression when studying the role of FOXO1 and FOXO3 genes in breast cancer.
Relationship between FOXO1 and FOXO3 protein expressions and clinical biological parameters of breast cancer. We investigated the relationships between FOXO1 and FOXO3 protein expressions and several classical clinical biological parameters (Tables 2 and 3). Marked differences were observed between these two FOXO proteins, as high FOXO1 protein expression was associated with negative estrogen receptor α (ERα) and progesterone receptor (PR) status, while low FOXO3 protein expression was weakly associated with these two biological parameters. We also showed that high FOXO3 protein expression was associated with low SBR histological grade, and, surprisingly, with high level of lymph node status. However, no association was observed between FOXO1 protein expression and these two clinical parameters.
To investigate in more detail the role of FOXO1 and FOXO3 genes in breast cancer, we also performed a log rank test to analyse the relationship between FOXO1 and FOXO3 protein expressions and metastasis-free survival (MFS). Patients with breast tumors expressing high levels of FOXO3 protein had better MFS than patients with breast tumors expressing lower levels of FOXO3 protein (p = 4.1.10 −2 ), which is consistent with a tumor suppressor role ( Fig. 2A). Such an association was not observed for FOXO1 protein. Multivariate analysis using a Cox proportional hazards model assessed the predictive value for MFS of the parameters found to be significant in univariate analysis (Table S1), i.e. lymph node status and macroscopic tumor size, and FOXO3 protein expression (p = 5.5.10 −3 ) ( Table S3). The prognostic significance of these three parameters persisted in multivariate analysis, indicating that FOXO3 protein expression is an independent prognostic factor in breast cancer. Interestingly, no correlation was demonstrated between FOXO1 or FOXO3 mRNA expression and MFS in this series of 218 breast tumors (Fig. 2B,C), highlighting once again the importance of studying FOXO1 and FOXO3 protein expressions in order to assess their role in breast cancer.
Altogether, these observations suggest that FOXO3 protein, but not FOXO1 protein, may act as a tumor suppressor in breast cancer. the role of FOXO1 and FOXO3 proteins in breast cancer, we also used the RPPA approach to study other proteins involved in the PI3K/AKT/mTOR oncogenic pathway, which regulates the activities of FOXO proteins, as well as various cellular processes involved in cancer and regulated by FOXO proteins: DDR, apoptosis, cell cycle, and cell proliferation (Table 4).
To evaluate the PI3K/AKT/mTOR pathway activity in our tumors, we studied the phosphorylation status of the S6K and S6, two proteins specifically phosphorylated by this pathway 26 . The PI3K/AKT/mTOR pathway-dependent phosphorylation of AKT to its threonine 308 (pAKT-T308) and, to a lesser extent, its serine 473, is essential for its kinase activity 27 . However, because of the lack of appropriate pAKT-T308 antibody for RPPA method, we did not use the phosphorylated forms of AKT to evaluate the PI3K/AKT/mTOR pathway activity. We found negative correlations between FOXO3 protein expression and protein expressions of Phospho-S6K (Thr-421/Ser-424) (r = −0.252, p = 2.10 −4 ), Phospho-S6K (Thr-389) (r = −0.163, p = 1.6.10 −2 ), Phospho-S6 (Ser-235/Ser-236) (r = −0.186, p = 5.8.10 −3 ), and Phospho-S6 (Ser-24) (r = −0.301, p < 10 −4 ), suggesting that the low expression of FOXO3 protein is associated to a weak activity of the PI3K/AKT/mTOR pathway in breast tumors. We observed that FOXO3 protein expression was also negatively correlated with the protein expressions of various components of this pathway: 29 . We did not detect significant correlation between FOXO3 protein expression and the protein expression of one of the most important negative regulators of the PI3K/AKT/mTOR pathway, PTEN. The weak PI3K/AKT/mTOR pathway activity found in the breast tumors expressing low level of FOXO3 protein would be therefore due to a low level of various components of this pathway but not to a high level of PTEN. Regarding FOXO1, we detected a negative correlation only with PDK1 (r = −0.186, p = 6.10 −3 ). Akt activation has been shown to promote degradation of FOXO3 protein by proteasomes 12 . Therefore, our results suggest that high activity of the PI3K/AKT/mTOR pathway in breast tumors would induce degradation of FOXO3 protein, but not FOXO1 protein.
In order to check that FOXO3 is functional in breast tumors, we performed a western blot analysis to visualize Phospho-FOXO3 (pSer-253), the inactive form of FOXO3, of several breast tumors of our series 10 . We detected at least 8 samples with negative or low levels of phospho-FOXO3 (pSer-253) expression among 12 breast tumor samples expressing high levels of FOXO3, suggesting that this FOXO3 protein may be functional in a majority of these tumors (Fig. S1).
Overall, our results suggest that FOXO3 protein, but not FOXO1 protein, acts as a tumor suppressor in breast cancer, at least in part by DDR inhibition and subsequent induction of p53-dependent apoptosis. They also suggest that the antitumor effect of FOXO3 is abolished by high activity of the PI3K/AKT/mTOR pathway.

Discussion
Many studies designed to examine the role of genes in carcinogenesis determine the correlations between their RNA expression and classical clinical biological parameters, survival, and the expression of others genes linked to cancer. However, due to post-transcriptional regulations, weak correlations are commonly observed between RNA expression and protein expression 33 . In our study, we found weak correlations between FOXO1 and FOXO3 RNA and protein expressions in breast cancer (Fig. 1). The absence of correlation between protein and mRNA expressions for FOXO1 and FOXO3 genes can be fully explain by the fact that these FOXO proteins undergo posttranslational modifications, such as acetylation, ubiquitination, and phosphorylation, modulating their subcellular localization and stability. To investigate the role of FOXO1 and FOXO3 genes in breast carcinogenesis, we therefore used the RPPA method to perform a comparative study of the protein expression of these two FOXO genes in a series of 218 breast tumors.
Our results strongly suggest that FOXO3 protein, but not FOXO1 protein, acts as a tumor suppressor in breast cancer. In particular, we found that patients with breast tumors expressing high levels of FOXO3 protein had better survival rates than patients with breast tumors expressing lower levels of this protein. We also showed that FOXO3 protein expression, but not FOXO1 protein expression, was negatively correlated with the expression of the KI67 marker of proliferation (Table 4). We recently showed that FOXO6 protein has an oncogenic effect in breast cancer 25 . Therefore, despite their homologies, the FOXO proteins appear to have different and specific effects on breast cancer development.
In line with the results of various studies, we provide experimental arguments suggesting a tumor suppressor activity of FOXO3 protein in breast cancer [15][16][17]20 . However, other studies have shown that high FOXO3 protein expression is associated with poor disease-free survival in TN breast cancer, and promotes proliferation, migration and invasion of TN breast cancer cell lines 22,23 . In addition, Sisci et al. reported that FOXO3 protein inhibits breast carcinogenesis in ERα-positive cells, and tends to promote breast carcinogenesis in ERα-negative cells 34 . The role of FOXO3 protein in breast carcinogenesis may therefore depend on the subtype of breast cancer and the stage of disease. Further protein expression studies based on larger series of breast cancers are necessary to determine the precise role of FOXO3 protein in the various subtypes of breast cancer. www.nature.com/scientificreports www.nature.com/scientificreports/ Several studies suggest that FOXO3 protein acts as a tumor suppressor in breast cancer by inducing the expression of cyclin-dependent kinase inhibitors (CDK inhibitors) and proapoptotic proteins 16,35 . In line with these findings, we showed that FOXO3 protein expression was positively correlated with expression of the p15 CDK inhibitor, p53 and two active forms of p53 phosphorylated at position S15 and S392 (Table 4). Surprisingly, we also demonstrated negative correlations between FOXO3 protein expression and the expression of numerous proteins involved in various DNA repair mechanisms, suggesting that high FOXO3 protein expression in breast tumors impairs DDR. Inhibition of DDR by FOXO3 protein could induce accumulation of DNA damage, thereby inducing p53-dependent apoptosis. FOXO3 was recently shown to negatively regulate the expression and activity of FOXM1, a forkhead protein activating the transcription of numerous genes involved in various DNA repair mechanisms and genotoxic agent resistance 36 . FOXO3 protein competes with FOXM1 for the binding to the same DNA motifs in target promoters and produces opposing transcriptional outputs. Therefore, one of the mechanisms by which FOXO3 protein could inhibit DDR in breast cancer, would be the inhibition of the transcription of DDR-genes induced by FOXM1.
PARP inhibitors have been shown to be highly lethal to tumor cells with a defect in DNA repair by homologous recombination called "BRCAness". The activity of these inhibitors is based on the principle of synthetic lethality, which consists of targeting two separate molecular pathways that are nonlethal when disrupted individually, but are lethal when inhibited simultaneously. We found negative correlations between the expression of FOXO3 protein and that of the three components of the MRN complex (NBS1, RAD50, and Mre11) crucial for DNA repair by homologous recombination (Table 4). High expression of FOXO3 protein could therefore be an attractive predictive biomarker of favourable response to treatment with PARP inhibitors in breast tumors.

Materials and Methods
Patients and samples. The conditions of patient's selection and sample collection were as previously described [19].
Real-time qRT-PCR. The theoretical basis, RNA extraction, cDNA synthesis, design of primers and qRT-PCR conditions have been previously described in detail [33]. The FOXO1 and FOXO3 expression values of the samples were normalized such that the median value for the 15 normal breast tissues was 1. Variation in   for TBP gene (132 bp PCR product); FOXO1-U (5′-GTCAAGAGCGTGCCCTACTTCA-3′) and FOXO1-L (5′-TGAACTTGCTGTGTAGGGACAGATTAT-3′) for FOXO1 gene (101 bp PCR product); FOXO3-U (5′-CCTACTTCAAGGATAAGGGCGACAG-3′) and FOXO3-L (5′-GTGCCGGATGGAGTTCTTCCAG-3′) for FOXO3 gene (62 bp PCR product); FOXO4-U (5′-TGGTCCGTACTGTACCCTACTTCA-3′). Over-and under-expressions were defined as threefold variations of expression relative to the median expression of normal samples. We have previously used the same approach to determine cut-off points for tumor gene altered expression 38-40 . RPPA. RPPA technology was used for quantifying the relative abundance of total protein expression as previously described 41 . Antibody references are available in Table S4. Low and high protein expressions were defined as twofold variations of expression relative to the median expression of the series of 218 breast tumors.
Statistical analysis. Statistical analyses were done as previously described [19]. The Cox proportional hazards regression model was used to assess prognostic significance and the results are presented as hazard ratios (HR) and 95% confidence intervals (CIs).
Compliance with ethical standards. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All patients who entered  Table 4. Spearman rank correlation coefficients (r) and p-values between FOXO1 and FOXO3 protein levels and other proteins of different pathways in the series of 218 breast tumours.