Specific gene expression signatures induced by the multiple oncogenic alterations that occur within the PTEN/PI3K/AKT pathway in lung cancer

Hyperactivation of the phosphatydil-inositol-3' phosphate kinase (PI3K)/AKT pathway is observed in most NSCLCs, promoting proliferation, migration, invasion and resistance to therapy. AKT can be activated through several mechanisms that include loss of the negative regulator PTEN, activating mutations of the catalytic subunit of PI3K (PIK3CA) and/or mutations of AKT1 itself. However, number and identity of downstream targets of activated PI3K/AKT pathway are poorly defined. To identify the genes that are targets of constitutive PI3K/AKT signalling in lung cancer cells, we performed a comparative transcriptomic analysis of human lung epithelial cells (BEAS-2B) expressing active mutant AKT1 (AKT1-E17K), active mutant PIK3CA (PIK3CA-E545K) or that are silenced for PTEN. We found that, altogether, aberrant PI3K/AKT signalling in lung epithelial cells regulated the expression of 1,960/20,436 genes (9%), though only 30 differentially expressed genes (DEGs) (15 up-regulated, 12 down-regulated and 3 discordant) out of 20,436 that were common among BEAS-AKT1-E17K, BEAS-PIK3CA-E545K and BEAS-shPTEN cells (0.1%). Conversely, DEGs specific for mutant AKT1 were 133 (85 up-regulated; 48 down-regulated), DEGs specific for mutant PIK3CA were 502 (280 up-regulated; 222 down-regulated) and DEGs specific for PTEN loss were 1549 (799 up-regulated, 750 down-regulated). The results obtained from array analysis were confirmed by quantitative RT-PCR on selected up- and down-regulated genes (n = 10). Treatment of BEAS-C cells and the corresponding derivatives with pharmacological inhibitors of AKT (MK2206) or PI3K (LY294002) further validated the significance of our findings. Moreover, mRNA expression of selected DEGs (SGK1, IGFBP3, PEG10, GDF15, PTGES, S100P, respectively) correlated with the activation status of the PI3K/AKT pathway assessed by S473 phosphorylation in NSCLC cell lines (n = 6). Finally, we made use of Ingenuity Pathway Analysis (IPA) to investigate the relevant BioFunctions enriched by the costitutive activation of AKT1-, PI3K- or PTEN-dependent signalling in lung epithelial cells. Expectedly, the analysis of the DEGs common to all three alterations highlighted a group of BioFunctions that included Cell Proliferation of tumor cell lines (14 DEGs), Invasion of cells (10 DEGs) and Migration of tumour cell lines (10 DEGs), with a common core of 5 genes (ATF3, CDKN1A, GDF15, HBEGF and LCN2) that likely represent downstream effectors of the pro-oncogenic activities of PI3K/AKT signalling. Conversely, IPA analysis of exclusive DEGs led to the identification of different downstream effectors that are modulated by mutant AKT1 (TGFBR2, CTSZ, EMP1), mutant PIK3CA (CCND2, CDK2, IGFBP2, TRIB1) and PTEN loss (ASNS, FHL2). These findings not only shed light on the molecular mechanisms that are activated by aberrant signalling through the PI3K/AKT pathway in lung epithelial cells, but also contribute to the identification of previously unrecognised molecules whose regulation takes part in the development of lung cancer.


Introduction
Lung cancer is the most frequent cause of cancer-related deaths worldwide [1,2]. Lung cancer comprises two main groups that include small-cell lung cancer (SCLC) and non-small-cell lung cancer (NSCLC) [1], of which the latter accounts for 80-85% of cases.
At present, five-year survival of lung cancer patients is low [3], because it is often detected in advanced stages [4]. For this reason a more complete understanding of the molecular origins of the disease may help contribute to improve therapeutic regimens.
The phosphatidylinositol 3-kinase (PI3K) signaling cascade plays a critical role in the initiation and/or progression of NSCLC [5][6][7][8][9][10][11]. This pathway regulates multiple cellular processes that are relevant in the growth and progression of lung cancer cells including cell proliferation, migration, apoptosis and angiogenesis [12]. The protein kinase B (PKB), also known as AKT, is an important mediator of the PI3K pathway, representing the end-point of signaling elicited by several growth factors and cytokines [13].
AKT is activated by recruitment to cell membrane through the binding of its PH domain to 3 0 -phosphorylated phosphatidyl-inositols generated by PI3K and subsequent phosphorylation at T308 and S473 [14][15][16]. Conversely, dephosphorylation of the 3 0 position of phosphatidylinositols exerted by the lipid phosphatase PTEN attenuates AKT activation [17].
In this manner, genes that are regulated by the PI3K/AKT pathway may contribute to mediate the multiple activities that are under the control of this pathway.
In NSCLC, AKT activation may result from distinct and often mutually exclusive events that include activating mutations or increased expression of one or more of AKT isoforms (AKT1, AKT2 or AKT3) or of their upstream regulators such as KRAS or PIK3CA or loss of negative regulators (i.e. PTEN) [8,31]. However, whereas the effects of AKT activation on the levels and/or the localization of specific substrates have been largely studied, so far a global analysis of the genes regulated by the different activated members of the PI3K/AKT signaling is missing.
In this manuscript we identified the panel of genes that are regulated by aberrant PI3K/ AKT signaling in lung epithelial cells, by analysis of gene profiles that are induced by specific alterations that include PIK3CA mutations, AKT1 mutations and PTEN loss. The findings reported here contribute to shed light on the molecular mechanisms that are activated by aberrant signalling through the PI3K/AKT pathway in lung epithelial cells, as well as to identify molecules, previously unrecognised, whose regulation contributes to the tumorigenic activity of this pathway in lung epithelial cells.

Materials and methods
Cell culture BEAS-2B, immortalized human bronchial epithelial cell line was purchased from Chambrex (Milan, Italy) and grown according to the manufacturer's protocol.

RNA extraction
Total RNA isolation from cells and clinical samples was performed with Trizol Reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. RNA integrity was checked by use of denaturing agarose gel electrophoresis.

RNA profiling analysis
RNA concentration was determined with Nanodrop spectrophotometer (Nano-Drop, Wilmington, Germany) and its quality was assessed with Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). For each sample, 500 ng of total RNA were used to synthesize biotinylated cRNA with Illumina RNA Amplification Kit (Ambion, Austin, TX). Synthesis was carried out according to the manufacturers' instructions. cRNA concentration and the quality were assessed out as described above. From each sample, three technical replicates were produced and 750 ng cRNA were hybridized for 18h to Human HT-12_V3_0_R1 Expression BeadChips (Illumina, San Diego, CA) as described earlier [32]. Hybridized chips were washed and stained with streptavidin-conjugated Cy3 (GE Healthcare, Milan, Italy). BeadChips were dried and scanned with an Illumina Bead Array Reader (Illumina).

Data analysis
Raw data from Array hybridization were loaded into and analyzed by Gene-Spring GX 14.5 (Agilent Technologies). DEGs were obtained performing ono-to-one comparison of the RNA profiles of BEAS-C derivatives (BEAS-AKT1-E17K, BEAS-PIK3CA-E545K and BEAS-shPTEN) compared to control BEAS-C cells, experiments were performed in triplicates. DEGs were selected on the basis of absolute fold-change (! 2) and statistical significance (p<0.01). Statistical analysis was performed by moderated t-test. P-values were corrected by Benjamini-Hochberg (B-H) FDR correction (q-value cut-off 0.01) [33]. Ingenuity Pathway Analysis (IPA, Ingenuity Systems) was used to evaluate the functional behaviour of DEGs in terms of Biological Processes and Molecular Function, Development Function, Disease and Disorder. The degrre of enrichment of the IPA analysis was performed by the Fisher's exact test (p 0.05).

Statistical analysis
RT-PCR data are expressed as means±SD of at least three independent experiments conducted in triplicates as indicated in the text. Statistical significance was evaluated by One-way, Twoway ANOVA or t-test tests as indicated in the figure legends. Statistical significance was indicated as follows: p 0.05 ( Ã ), p 0.01 ( ÃÃ ), p 0.001 ( ÃÃÃ ) and p 0.0001 ( ÃÃÃÃ ).

Results and discussion
Identification of PI3K/AKT-dependent gene expression signature in lung epithelial cells The purpose of this study was to identify genes that are modulated by aberrant PI3K/AKT signalling in lung epithelial cells, and for this reason, may represent novel determinants of the pro-tumorigenic activity of the PI3K/AKT pathway. To this aim, we have genetically modified bronchial epithelial cells that recapitulate the most frequent alterations in genes within the PI3K/AKT pathway. As model cell line we chose BEAS-2B cells, an immortalised bronchial epithelial cell line of human origin that do not support tumour growth [37].
The presence of exogenous mutant PIK3CA and AKT1 proteins or of the endogenous wildtype PTEN protein in transduced cells as well as the activation of PI3K/AKT signalling were already determined [38,39].
To identify genes that are targets of constitutive signalling of PI3K/AKT in lung cancer cells, we have analysed gene expression profiles of BEAS-C cells and derivatives. Expression values of the mRNAs obtained were filtered for log2 fold change !2. The transcriptome of BEAS-PIK3CA-E545K cells has already been analysed in a previous study, using a slightly lower log2 fold change threshold of 1.5 [8] and reanalysed with the more selective threshold used in this manuscript to be compared with BEAS-AKT1-E17K and BEAS-shPTEN cells.
Analysis of results allowed the definition of 3 lists of differentially expressed gene probes (See S1-S3 Files).
As shown, we identified 133 DEGs in cells expressing mutant AKT1 (85 of which were upregulated and 48 were down-regulated), 502 DEGs in cells expressing mutant PIK3CA (280 of which were up-regulated and 222 were down-regulated) and 1549 DEGs in cells silenced for PTEN (799 of which were up-regulated and 750 were down-regulated). See the heatmap in Fig  1A for graphic representation of the observed differences in gene expression. As shown the gene expression profile resulting by PTEN silencing was remarkably different from those induced by mutant AKT1 and PIK3CA, respectively. Microarray raw data for all probes have been deposited in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-5286.
Once we identified the genes that are regulated by aberrant PI3K/AKT signaling in human lung cells, we proceeded to compare DEG lists in order to identify genes whose expression was modified in cells expressing mutant AKT1, mutant PIK3CA, PTEN loss and/or a combination thereof. Being regulated by alterations in all three conditions, these genes are expected to be the main mediators of aberrant PI3K/AKT signalling in transformed lung epithelial cells, most of which have not apparently been recognised yet.

Validation of DEGs in BEAS-C cells and derivatives
We confirmed the results obtained from the array analysis by quantitative RT-PCR on a selected panel of up-regulated (PTGES, KRT81, S100P, DUSP5) and down-regulated genes (BMF, SGK1, IGFBP3, VWA5A, PEG10, MARCKS) (Fig 2). All DEGs validated by qRT-PCR were confimed to be statistically significant, as indicated in the  Subsequently, we used pharmacological inhibitors of AKT (MK2206) and of PI3K (LY294002) to investigate whether the identified DEGs responded to modulation of the activity of PI3K/AKT pathway. We found that exposure of BEAS-C cells and the corresponding derivatives to MK2206 or LY294002 induced a marked decrease in the mRNA levels of PTGES and S100P ( Fig 3A) and a marked increase in the mRNA levels of SGK1 ( Fig 3B) in BEAS-AKT1-E17K, BEAS-PIK3CA-E545K and BEAS-shPTEN cells. However, statistical analysis demonstrated that the modulation exerted by pharmacological inhibitors was statistically significant in all three modified cells in the case of S100P and SGK1. Conversely, for PTGES, PEG10 and IGFBP3 we observed an effect that was in line with the predicted modulation of the three DEGs as obtained by array analysis, which was not statistically significant. As control of the experiments we observed that both inhibitors significantly suppressed AKT phosphorylation in BEAS-C and derivative cells ( Fig 3C).

DOWN-REGULATED GENES
HS.570308 Pathways regulated by oncogenic PTEN/PI3K/AKT signalling in NSCLC    Expression of selected common DEGs in lung cancer cell lines Finally, we investigated the mRNA levels of representative common DEGs that had come out from the array analysis in 6 NSCLC cell lines. The NSCLC cell lines used were NCI-H23, NCI-H292, NCI-H522, NCI-H460, A549 and NCI-H226. To this aim, we analysed the mRNA Pathways regulated by oncogenic PTEN/PI3K/AKT signalling in NSCLC expression of SGK1, IGFBP3, PEG10, GDF15, PTGES and S100P by quantitative RT-PCR and correlated them with the activation status of the PI3K/AKT pathway as assessed by level of S473 phosphorylation. See Fig 4A. As shown in Fig 4B, NSCLC cells with low/absent levels of phosphorylated AKT (NCI-H522, NCI-H292, NCI-H23) generally showed lower mRNA levels of GDF15, PTGES and S100P (up-regulated DEGs) than cells with high levels of AKT activity (NCI-H226, NCI-H460, A549) (Fig 4B, p-value as indicated). Similarly, NSCLC cells with low/absent levels of phosphorylated AKT showed significantly high levels of SGK1, IGFBP3, PEG10 (down-regulated DEGs) at difference with what observed in cells with highly phosphorylated AKT (Fig 4C, pvalue as indicated).
Ingenuity Pathway Analysis (IPA) analysis of genes regulated by activation of the PI3K/AKT pathway in lung epithelial cells In order to search for novel common and/or exclusive effectors of the PI3K/AKT pathway, we used Ingenuity Pathway Analysis (Ingenuity1Systems, IPA, http://www.ingenuity.com) to functionally annotate the lists of DEGs obtained in the experiments described above and to investigate the biological relevance of the transcriptional changes induced by the gain-of-function mutations of AKT1, PIK3CA and/or by the loss of PTEN in lung epithelial cells. To this aim the datasets obtained from common and/or exclusive DEGs were categorized using IPA BioFunctions and we investigated how similar were the biological mechanisms and/or pathways modulated by the alteration of single members (PIK3CA, PTEN, AKT1) within the PI3K/AKT pathway.  Table 5).
The common DEGs enriched within the indicated BioFunctions are listed in Table 6. We observed that the three BioFunctions contained a common core of 7 genes, namely ATF3 (average log2 fold change of +2.1), CDKN1A (average log2 fold change of +2.5), GDF15 (average log2 fold change of +19.6), GJA1 (average log2 fold change of -3.4), HBEGF (average log2 fold change of +3.1), LCN2 (average log2 fold change of +5.1) and SDC4 (average log2 fold change of +2.9), which are potentially involved in the simultaneous regulation of proliferation, migration and invasion.
To increase the relevance of the common DEGs enriched in the above-mentioned BioFunctions, we validated by qRT-PCR the core of 7 common genes ( Fig 6A). Among these, only 5 genes (GDF15, LCN2, HBEGF, ATF3 and CDKN1A) showed in the qRT-PRC analysis significant p-values (One-way ANOVA test, p<0.05). As to GJA1 and SDC4, qRT-PCR analysis showed no difference between BEAS-C cells and the corresponding derivatives.
In addition, experiments performed with LY294002 indicated that in BEAS-C cells and derivatives the expression of GJA1, SDC4 and CDKN1A was not significantly regulated by modulation of PI3K/AKT activity (Fig 6B). Conversely, the remaining 4 genes GDF15, LCN2, HBEGF and ATF3 showed a significant modulation of their expression upon treatment with LY294002 (Two-way ANOVA test).
On the basis of these results, we propose that GDF15, LCN2, HBEGF, ATF3 and CDKN1A represent important effectors of PI3K/AKT signalling, whose aberrant modulation may contribute to the development of lung cancer, especially when transformation is driven by mutations of PIK3CA or AKT1 or by the loss of PTEN expression. It is important to underline that altought CDKN1A was not apparently modulated by LY294002 it has been recurrently associated to signalling through the PI3K/AKT pathway [40,41]. For this reason we are regarded CDKN1A as a part of common downstream effectors of PI3K/AKT signalling. CDKN1A encodes the cyclin-dependent kinase inhibitor p21, which promotes cell cycle arrest as an effector of multiple anti-proliferative signals [42,43]. By phosphorylating p21 at Thr145 and/ or Ser146, AKT increases the stability of the protein, induces its cytoplasmic retention [40], thus destroying its cell cycle inhibitory activity and stimulating migration and survival [44][45][46]. However, in certain conditions p21 may behave as an oncogene [47,48], with its expression frequently increased in cancer cell lines and tumors [49][50][51][52]. In agreement with a prooncogenic role of p21, the finding that all the genetic alterations that lead to the activation of the PI3K/AKT pathway induce an increase in the level of p21 mRNA in lung epithelial cells, suggests that p21 may mediate some of the pro-oncogenic activities exerted by aberrant PI3K/ AKT signalling in cancer cells such as to stimulate migration and/or invasion or to protect cells from apoptosis [53].
used to normalize AKT phosphorylation and protein loading, respectively. (B) The graph shows mRNA levels of up-regulated DEGs (GDF15, PTGES, S100P) in NSCLC cell lines. Statistical analysis was performed on 3 independent experiments conducted in triplicate and was performed by One-way ANOVA test for each gene relative to each cell line, p-values as indicated (C) The graph shows mRNA levels of three representative down-regulated DEGs (SGK1, IGFBP3, PEG10) in NSCLC cell lines. Bars are mean ± SD of a representative experiment. Statistical analysis was performed on 3 independent experiments conducted in triplicate and was performed by One-way ANOVA test for each cell line, p-values as indicated.
Heparin Binding EGF-Like Growth Factor (HBEGF) is an EGF-like ligand of the epidermal growth factor receptor (EGFR) that binds heparin sulfate proteoglycans [76].
As anticipated, the links of ATF3, GDF15, HBEGF and LCN2 with the PI3K/AKT pathway are contradictory and, as for ATF3 and GDF15, in most cases have failed to unambiguosly  identify these genes as downstream targets of PI3K/AKT signalling [68,[84][85][86]. The results reported in this manuscript are in line with the known functions of these 4 DEGs. Accordingly, the expression of ATF3, HBEGF, GDF15 and LCN2 is increased in all three derivatives of BEAS-C cells, suggesting that the coordinate changes of these genes may contribute to confer proliferative, pro-migratory, pro-invasive and/or anti-apoptotic properties elicited by aberrant PI3K or AKT in lung epithelial cells. Notably, these results also suggest that similar mechanisms may be operating to confer proliferative, pro-invasive and pro-migratory properties to lung epithelial cells transformed by aberrant PI3K signalling independently of the molecular alteration that have induced it. Subsequently, we analyzed the BioFunctions enriched by the three lists of exclusive DEGs to investigate whether the different genetic alterations that activate the PI3K/AKT pathway under study here signalled through multiple downstream effectors.
Finally, the BioFunctions enriched by exclusive DEGs in BEAS-shPTEN cells highlited genes regulating either cell growth such as Asparagine Synthetase (ASNS), wich predicts poor Notably, the significance of our findings was further strenghtened by two different observations. First, we observed that the pharmacological inhibition of PI3K/AKT signalling in lung epithelial cells modulated the mRNA expression of common DEGs in line with the array results. Second, we observed that in NSCLC cell lines the mRNA expression of these randomly selected down-regulated (SGK1) or up-regulated (GDF15, PTGES, S100P) DEGs correlated with the activation status of the PI3K/AKT pathway. These findings indicate that the identified DEGs may indeed represent critical downstream effectors of the PI3K/AKT signalling axis.
In conclusion, the results reported in this manuscript describe the mRNA profiles elicited in lung epithelial cells by the three most common alterations of the members within the PI3K/AKT pathway and have led to the identification of 5 novel downstream targets of PI3K/AKT signalling that may contribute to confer pro-mitogenic, pro-migratory and proinvasive properties to cells, once the PI3K/AKT pathway is activated, in a manner that is independent of the molecular alteration that may have caused it. On the other hand, this study led to the identification of genes that are exclusively regulated by mutant AKT1 (TGFBR2, CTSZ, EMP1), mutant PIK3CA (CCND2, CDK2, IGFBP2, TRIB1) and PTEN loss (ASNS, FHL2).
These findings not only shed additional light on the molecular mechanisms that are activated by signalling through the PI3K/AKT pathway in lung epithelial cells, but also may contribute to identify previously unrecognised therapeutic targets in lung cancer driven by aberrant PI3K/AKT signaling.