Signaling Networks Associated with AKT Activation in Non-Small Cell Lung Cancer (NSCLC): New Insights on the Role of Phosphatydil-Inositol-3 kinase

Aberrant activation of PI3K/AKT signalling represents one of the most common molecular alterations in lung cancer, though the relative contribution of the single components of the cascade to the NSCLC development is still poorly defined. In this manuscript we have investigated the relationship between expression and genetic alterations of the components of the PI3K/AKT pathway [KRAS, the catalytic subunit of PI3K (p110α), PTEN, AKT1 and AKT2] and the activation of AKT in 107 surgically resected NSCLCs and have analyzed the existing relationships with clinico-pathologic features. Expression analysis was performed by immunohistochemistry on Tissue Micro Arrays (TMA); mutation analysis was performed by DNA sequencing; copy number variation was determined by FISH. We report that activation of PI3K/AKT pathway in Italian NSCLC patients is associated with high grade (G3–G4 compared with G1–G2; n = 83; p<0.05) and more advanced disease (TNM stage III vs. stages I and II; n = 26; p<0.05). In addition, we found that PTEN loss (41/104, 39%) and the overexpression of p110α (27/92, 29%) represent the most frequent aberration observed in NSCLCs. Less frequent molecular lesions comprised the overexpression of AKT2 (18/83, 22%) or AKT1 (17/96, 18%), and KRAS mutation (7/63, 11%). Our results indicate that, among all genes, only p110α overexpression was significantly associated to AKT activation in NSCLCs (p = 0.02). Manipulation of p110α expression in lung cancer cells carrying an active PI3K allele (NCI-H460) efficiently reduced proliferation of NSCLC cells in vitro and tumour growth in vivo. Finally, RNA profiling of lung epithelial cells (BEAS-2B) expressing a mutant allele of PIK3 (E545K) identified a network of transcription factors such as MYC, FOS and HMGA1, not previously recognised to be associated with aberrant PI3K signalling in lung cancer.


Introduction
Lung cancer is the leading cause of cancer deaths worldwide [1,2]. Epithelial lung cancer is classified into two main groups: small-cell lung cancer (SCLC) (about 15% of all lung cancers) and non-small-cell lung cancer (NSCLC) (about 85% of all lung cancers) [3]. NSCLC comprises squamous-cell carcinoma (SCC), adenocarcinoma (ADC), and large-cell lung cancer (LCC) [3]. Despite advances in early detection and standard treatment, NSCLC is often diagnosed at an advanced stage and patients often have poor prognosis, with five-year survival rate less than 15% [4,5]. For this reason a better understanding of the molecular origins of the disease will contribute to improve therapeutic treatment of lung cancer patients.
Recent studies have shown that the phosphatidylinositol 3kinase (PI3K) signalling cascade is frequently overactivated in human cancer [6][7][8] playing a critical role both in the initiation and progression of NSCLC [9,10]. The PI3K pathway regulates cellular functions such as proliferation, survival, motility and angiogenesis that are critical to the growth and/or maintenance of tumours [11,12]. The end-point of the PI3K pathway is AKT, a serine/threonine protein kinase that mediates most signals funnelled through the PI3K pathway. AKT is activated by recruitment to cell membrane via binding of its PH domain to 39phosphorylated phosphatidylinositols generated by PI3K and subsequent phosphorylation at T308 and S473 [12,13]. Conversely, the lipid phosphatase PTEN attenuates AKT activation by dephosphorylating the 39 position of phosphatidylinositols [14].
Aberrant AKT activation contributes to lung carcinogenesis [9,10]. Hyperactivation of AKT is detected in most NSCLC cell lines [15][16][17], and in 30-75% NSCLCs [18][19][20][21][22] and promotes resistance to chemotherapy and radiation therapy [16]. AKT activation in cancer is currently evaluated using phospho-specific antibodies against S473 in immunohistochemical analyses of tumour specimens. Although phosphorylation of AKT at S473 has been correlated with poor clinical outcomes in many tumour types, results in lung cancer are apparently inconsistent [7][8][9][10] having been associated with either poor or good prognosis [20][21][22]. AKT can be activated through several mechanisms, which result from distinct and often mutually exclusive events that include activating mutations (KRAS, PIK3CA or AKT1), increased expression (PIK3CA, AKT1, AKT2) or loss of PTEN [10]. However, the relative contribution of the single components within the PI3K pathway to AKT activation in NSCLCs is still unclear. In this manuscript we have investigated the relationship between the genetic alterations present in these genes and the activation of AKT in NSCLC.

Ethics Statement
Patient accrual was conducted according to internal Review Board of the INT Fondazione Pascale (Naples, Italy) (CEI 556/10 of 12/3/2010). The study was approved by the internal Review Board of the AOU Mater Domini/University Magna Graecia (Catanzaro, Italy) in the meeting of 16/3/2011. Written informed consent was obtained from all participants to the study. All animal work was conducted according to the relevant Italian guidelines and was approved by the Internal Committee for Animal Study (CESA) of the Institute for Genetic Research ''Gaetano Salvatore on April 7 th 2008 (CESA 10-08).

Patients
Archive material from 107 patients diagnosed of NSCLC [3] was obtained from INT Fondazione Pascale (Naples, Italy). Median age was 64 year old (range 28-82). Among patients with clinical data available, women were 18 and males 83. Stage was known for 81 patients: 67 patients had stage I-II disease and 14 had stage III-IV disease. Grade was known for 83 patients: 35 cases were G1-G2 and 48 were G3-G4. See Table S1, Table S2 and Table S3 for more detailed clinical characteristics of all patients.
TMA slides were deparaffinized, heated in a pressure cooker with 1 mM EDTA, pH 8.0 for 10 min, and incubated with pepsin at 37uC for 30 min. The slides were then dehydrated in increasing ethanol concentrations, and then air-dried. The probes were denatured at 96uC for 5 min, and hybridization solution was applied on each slide and incubated at 75uC for 1 min. After overnight incubation at 37uC in a humid chamber, slides were washed with 0.46 SSC and 0.3% NP40 for 2 min at 75uC, airdried in darkness, counterstained with DAPI, and a coverslip was applied.
The anti-Akt1 and anti-Akt2 have been shown to be isoformspecific antibodies in previous work [25]. In addition, by using NCI-H460 cells interfered for Akt1 or Akt2, respectively, we confirmed that the anti-Akt1 antibody recognizes only the Akt1 isoform and the anti-Akt2 antibody recognizes only the Akt2 isoform ( Figure S1A).
The immunohistochemical score of pAKT and PTEN used in this work was selected on the basis of the widely established criteria existing in the literature [28,30,31]: pAKT was scored as positive when .10% of tumour cells were positive with strong or diffuse immunopositivity. PTEN expression was classified as (+) when staining was detected in .50% of the cells, (+/2) when staining was detected in 25-50% of cells and (2) when staining was detected in 0-25% of cells. For statistical analysis PTEN expression was considered lost when samples were classified as (2).
Also for the immunostaining scores of AKT1, AKT2 and PIK3CA, we selected criteria described in previous reports [27,28,32]. Tumor specimens were divided into four groups according to the percentage of positive cells: (2) comprised completely negative samples; (+) comprised samples with up to 10% of positive cells; (++) comprised samples with 11-50% of positive cells; and (+++) comprised samples with .50% of positive cells, respectively. For statistical reasons, tumours were classified into a low expression group comprising (2) and (+) and a high expression group that comprises (++) and (+++).
For each one immunohistochemical round a negative control has been included, by replacing the primary antibody with solvent at the same volume of that with the primary antibody resuspended in it. All controls gave satisfactory results. Stained TMA sections were evaluated by two expert pathologists (RF, GB) using uniform criteria. Discrepancies were resolved through simultaneous inspection and discussion of the results. Discrepancies between two cores from the same case were resolved through a joint analysis of the two cores.
For evaluation of copy number of the genes encoding AKT1, AKT2 and PIK3CA, a gene-to-control ratio of 1.0 was classified as disomy; ratios between 1.0 and 2.0 were considered gene lowlevel gains; ratios .2.0 were considered as high polysomy and/or gene amplification [33,34].
Accordingly, tumours were divided into different classes: disomy, trisomy (3 copies of chromosomes in .40% of cells), low polysomy ($3 copies of chromosomes in .40% of cells), high polysomy ($4 copies of chromosomes in $40% of cells), and gene amplification (presence of gene clusters with a ratio of gene-tochromosome of $2 per cell in $40% of cells or presence of small or nonenumerable clusters of the gene signal). This allowed the classification of patients into two groups: FISH-negative (disomy and gains) and FISH-positive (high polysomy and/or gene amplification).

PCR, RT-PCR and mutation analysis
Total RNA and genomic DNA were prepared as described [35,36]. Q-RT-PCR and Q-PCR were performed using the Power SYBR Green PCR Master Mix in an ABI Prism 7300 thermocycler (Applied Biosystems, Foster City, CA, USA). cDNAs were synthesized from 1 mg of total RNA using QuantiTect Reverse Trascription (Qiagen, The Netherlands, Venlo). Normalization was performed to GAPDH mRNA content. The relative amounts of mRNA or DNA were calculated by the comparative cycle threshold (CT) method by Livak and Schmittgen [37]. Mutation analysis for PIK3CA using LightCycler was performed with DNA Master/Hybridization probes kit (Roche Molecular Biochemicals, Mannheim, Germany). Direct sequencing was performed using the BigDye v3.03 cycle sequencing kit (Applied Biosystems) in a capillary automatic sequencer (ABI PRISM 3100 Genetic Analyzer; Applied Biosystems). Protocols and primers for Q-PCR, Q-RT-PCR and sequencing KRAS (exons 2 and 3) and PIK3CA (exons 9 and 20) are reported in Appendix S1.

Virus generation and Infection
To generate p110a encoding lentivirus, the cDNA encoding human p110a (Addgene, Cambridge, MA, USA) was cloned in pENTR1A vector (Invitrogen) and recombined in pLenti6.2/C-Lumio TM /V5-DEST Vector by making use of the Gateway Technology (Invitrogen). pLenti vector was used to generate lentiviral particles in HEK293T packaging cells as described [40]. Transduced BEAS-2B cells underwent three rounds of infection and were selected in medium containing 5 mg/ml blasticidin (Invitrogen). The Human PIK3CA (NM_006218), AKT1 (NM_005163) and AKT2 (NM_001626) MISSION shRNA set (Sigma-Aldrich, St.Luis, MO) and the Mission non-target control transduction viruses (SHC002V) were used to generate lentiviral particles in HEK293T packaging cells [40]. After transfection, supernatants were collected at 8-hour intervals, filtered and used for three rounds of transduction of NCI-H460 cells in the presence of 8 mg/ml of polybrene (Sigma).

Tumourigenic assays
Cells (1610 6 ) were suspended in 100 ml 10% FBS and 100 ml Matrigel (BD Biosciences, NJ, USA) and subcutaneously injected into the right flank of 6-week-old athymic nude mice (Charles River, West Germany) in triplicates. Every 7 days tumour size was measured with a caliper.

RNA profiling analysis
RNA concentration was determined with a Nanodrop (Nano-Drop, Wilmington, Delaware, USA) spectrophotometer and its quality was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies, Milano, Italy). For each sample, 500 ng of total RNA were synthesized to biotinylated cRNA using the Illumina RNA Amplification Kit (Ambion, Inc., 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, technical replicates were produced and 750 ng cRNA were hybridized for 18 hrs to Human HT-12_V3_0_R1 Expression BeadChips (Illumina Inc., San Diego, CA, USA) according to the protocol provided by the manufacturer. Hybridized chips were washed and stained with streptavidin-conjugated Cy3 (GE Healthcare Milano, Italy). BeadChips were dried and scanned with an Illumina BeadArray Reader (Illumina Inc.).

Microarrays data analysis: RNA profiling, genes' characterization, enriched pathways and bibliographic networks discovery
Expression files were normalized and analyzed using Gene-Spring 10.1 (Agilent Technologies, Santa Clara, CA). Differentially expressed (DEGs) genes between BEAS-2B and BEAS-PI3K-CA cells were selected on the basis of the fold change (the ratio between the expression levels in the two conditions) and the statistical significance. We filtered the lists using fold change 1.5 and T-test (p-value (0.01) as threshold. The DEGs list (composed by 2126 probesets) was used to evaluate the functional behavior in terms of Biological Processes and Molecular Function, Development Function and Disease and Disorder terms. The degree of enrichment was statistically evaluated to determine whether an observed level of annotation for a group of genes is significant. In particular, for each term, a q-value was computed by the Hypergeometric test (p#0.05) and corrected using False Discovery Rate (FDR) [41]. The terms with a q-value exceeding the significance threshold were then selected as representative. Pathway and network analysis were performed using Ingenuity Pathway Analysis (IPA, Ingenuity Systems).
The dataset was mined for significant pathways with the IPA library of canonical pathways, and networks were generated by using IPA as graphical representation of the molecular relationships between genes and gene products. The significance of the association between the list of DEGs and the Canonical Pathway was measured using a Fisher's exact test to calculate a p-value (p#0.05). Fisher's exact test results were also corrected for multiple testing using FDR.
In networks, genes or gene products are represented as nodes, and the biological relationship between two nodes is represented as an edge (line). All edges are supported by at least one reference from the literature, from a textbook, or from canonical information stored in the IPA Knowledge Base. Human, mouse, and rat orthologs of a gene are stored as separate objects, but are represented as a single node in the network. The network building's algorithm determines a statistical score for each network. This is done by comparing the number of focus genes that contribute to a given network relative to the total number of occurrences of those genes in all networks or pathways stored in the IPA Knowledge Base. The intensity of genes (node) colour in the networks indicates the degree of downregulation (green) or upregulation (red) of gene expression. Nodes are displayed using various shapes that represent the functional class of gene products.

AKT activation in NSCLCs
As a read-out of PI3K/AKT signalling in NSCLC we determined the phosphorylation status of residue S473 of AKT1 (pAKT). pAKT was evaluated on TMAs containing duplicated core biopsies of 107 NSCLCs. As controls 45 matched normal samples were used. Patients' clinico-pathological characteristics are described in Materials and Methods and summarized in Tables S1, S2 and S3. The results obtained from pAKT staining in NSCLC are summarized in Table 1. pAKT staining was barely detectable in the epithelial cells from normal alveolar epithelium and from upper airways (39 out of 45 samples) (See Figure S1). In contrast, AKT activation was observed in 60 out of 97 of NSCLC analysed (Table 1). Positive pAKT staining was significantly higher in the carcinoma samples than either normal alveolar or bronchial epithelium (P,0.001; Chi square test). pAKT staining was observed in 23/37 SCCs and 30/44 ADCs ( Figure 1A and B, respectively). We found a significant association between pAKT staining and the grade or the stage of the disease ( Table 2): pAKT staining was significantly more represented in patients with grades G3-G4 compared with patients with grades G1-G2 (p,0.05) and in patients with TNM stage III compared with patients with stage II disease (p,0.05). See Tables S4 and S5 for distribution of patients into SCCs and ADCs. These results demonstrate that, in agreement with work in other populations, in Italian NSCLC patients AKT activation occurs in tumour tissue and correlates with a more advanced stage of disease [20][21][22]. See also Table S9 and S10 for a detailed, patient-by-patient, list of pAKT positivity.

Mechanisms of AKT activation in NSCLCs: immunohistochemistry
To investigate the molecular mechanisms leading to AKT activation in Italian patients affected by NSCLC we performed a comprehensive analysis of the expression and/or the genetic status of AKT1 and AKT2 and their closest regulators (KRAS, PIK3CA and PTEN). Of the 107 cases present on the TMAs 96 could be properly analysed for AKT1, 83 for AKT2, 104 for PTEN and 92 for PIK3CA.
Patients accrued for this study had already been characterised for PTEN expression [38]: complete loss occurred in 41 of 104 (39%) NSCLCs and partial down-regulation was observed in 41 additional cases. PTEN loss was more frequently observed in SCCs (22/40, 55%) than in ADCs (14/51, 27%) (See Figure S5). However, when correlated with AKT activation, the loss or the reduction of the levels of PTEN protein was not associated with AKT activation (n = 95; p = 0.832) ( Table 3).
Notably, from the integrated analysis of the TMAs we found that AKT activation was more frequently observed in tumours showing aberrant expression of more than a single gene within the PI3K pathway (PTEN loss, or overexpression of AKT1, AKT2, p110a respectively). In fact, AKT activation was detected in 15-64% of tumours showing aberration in a single gene, 44-89% of tumours with aberrant expression of two genes, 67-100% of   Table 4).

Mechanisms of protein overexpression: FISH analysis
FISH analysis in NSCLCs was performed for AKT1, AKT2 and PIK3CA to determine the molecular mechanisms of the overexpression of the corresponding proteins. See Materials and Methods for classification of tumours by FISH. We found 20/82 NSCLC (24%) with copy number gain of the AKT1 gene at chromosome 14, of which 16 were high polysomy (.4 copies) and 4 focal amplification (SCC-11, SCC-12, SCC-14 and SCC-21) ( Figure 2C). Expectedly, several AKT1 FISH-positive NSCLCs (12 out of 20 cases, 60%) showed moderate or high AKT1 expression. See Tables S9 and S10 for a detailed list of genetic alterations detected in single SCC and ADC patients. In the case of AKT2, we observed 24/73 NSCLCs (31%) with copy number gain of the gene at chromosome 19, of which 23 patients had high polysomy and 1 patient had focal amplification (SCC-11). See Figure 3C for a representative example. However, the significance of AKT2 amplification in lung cancer remains unclear, since 13/ 24 (54%) cases of AKT2 FISH-positive tumours did not show increased expression of the corresponding protein. However, not all FISH-positive NSCLCs resulted in the activation of AKT signalling. As shown in Tables S6, S7 and S8, 11/18, 10/19 and 14/23 cases that were FISH-positive for PIK3CA, AKT1 and AKT2 resulted positive for pAKT, respectively.

Activated PI3K contributes to cell proliferation and tumourigenicity of NSCLC cells
Given the importance of PI3K signalling in NSCLCs, we investigated the role of constitutive PIK3CA activation on the tumourigenic potential of human lung epithelial cells. To this aim, we made use of a mutant cell line (NCI-H460) that harbours a heterozygous activating mutation (E545K) in PIK3CA. Cells were transduced with a lentivirus expressing shRNA for PIK3CA ( Figure 6A). Silencing of p110a expression was assessed by immunoblot ( Figure 6B). Here we show that suppression of p110a expression in NCI-H460 cells markedly reduced in vitro anchorage-dependent and in vivo tumour growth of cells subcutaneously injected into immunodeficient mice (n = 6/group) ( Figure 6C and D, respectively), indicating that PI3K activation plays a significant role in the malignant behaviour of NSCLC cells.

Molecular profiling of PI3K activation in lung epithelial cells
To further characterize the role played by PIK3CA in development of NSCLC, we performed RNA profiling analysis of human lung epithelial cells expressing an active PI3KCA mutant (E545K) to identify cellular targets of constitutive PI3K signalling. Expression of exogenous PI3KCA allele was determined by immunoblot (Fig. 7A). Expression values obtained were filtered for fold change greater than 1.5 and subjected to t-test (pvalue cut-off of 0.01) with Benjamini-Hochberg (B-H) FDR correction [41], obtaining a total of 2126 differentially expressed probe sets, of which 1005 were down-regulated and 1121 were upregulated. The complete microarray data for all probe sets with the respective normalised values will be available at ArrayExpress and are provided in additional files (Table S11).
We used Ingenuity Pathway Analysis (IngenuityHSystems, http://www.ingenuity.com, IPA) to investigate the biological relevance of the PI3K-dependent expression changes by categorizing our dataset into biological pathways and/or functions and diseases ( Figure 7B, 7C, 7D, 7E; Figure S6; Table S11). The function ''Cancer'' was most frequent and associated with 466 genes, followed by ''Cell Death'' (392 genes), ''Cellular Growth and Proliferation'' (357 genes), ''Cellular Movement'' (196 genes), ''Cell Cycle'' (161 genes), ''Cell-to-cell Signalling and Interaction'' (112 genes), and ''Cellular Morphology'' (97 genes), respectively. We found that active PI3K regulates expression of most cell cycle molecules such as CCND1, CCND2, Cdk6 and Cdk inhibitors as well as of several apoptosis-related genes such as BAG3, IGFBP7, IGFBP3, TRADD and TRIB1. As to ''Cell movement'' function,     Figure 8B). However, it is to be noted that in both cell lines and tumours, data showed a trend that was not statistically significant given the low number of samples analysed and the huge heterogeneity of expression shown by HMGA1, FOS and MYC in tumors.

Discussion
We report a detailed analysis of the contribution of the different members of PI3K/AKT pathway to AKT deregulation in lung cancer. The most interesting findings of this study were that in Italian NSCLC patients activation of AKT was associated with advanced stage and higher grade and that, in these tumours, the major determinant of AKT activation was the over-expression of the catalytic subunit of phosphatidylinositol 3-kinase, p110a. Experimental evidence obtained by manipulation of PI3K signalling in NSCLC cells also indicated that p110a is required for in vitro and in vivo growth and disclosed a network of PI3Kregulated transcription factors that may be responsible for the oncogenic effects exerted by aberrant PI3K signalling in cancer [48].
To our knowledge this is the first comprehensive analysis aimed at determining the role of AKT signalling performed on a cohort of Italian NSCLC patients. So far, little information concerning AKT activation in Italian NSCLC patients was available. In the cohort of NSCLC patients studied here, AKT pathway is activated in 62% of cases, with significant S473 phosphorylation detected more frequently in patients with advanced disease (TNM stage III vs. stage II; n = 26; p,0.05) and higher grade (G3-G4 compared with G1-G2; n = 83; p,0.05). Several NSCLCs analysed in this study over-expressed PIK3CA, implying that the deregulated expression of wild type p110a might represent an oncogenic event during cancer development in the lung. Conversely, we found PIK3CA mutation in only one SCC patient, confirming that, although frequent in breast, gastric and hepatocellular cancers, PIK3CA mutations are rare in NSCLCs [49]. Other molecular lesions detected in NSCLC patients comprise PTEN loss (39%) and AKT1 or AKT2 over-expression (18% and 22%, respectively). It is of note that although PTEN loss in NSCLCs is more common than overexpression of p110a, our results indicate that the latter is the unique alteration that is significantly associated to AKT activation (p = 0.02).
Interestingly, simultaneous aberrant expression of two or more members within the PI3K pathway was relatively infrequent in unselected NSCLCs but was significantly more frequent in NSCLCs with activated AKT (see Table 4 for details). This observation suggests that p110a over-expression alone is not sufficient to activate AKT signalling and hence requires other alterations to be fully oncogenic in NSCLCs. Moreover, at difference with the significant AKT activation shown by NSCLCs with mutant KRAS or AKT1, the tumour that harboured mutant PIK3CA was negative for pAKT, suggesting that the type or the position of the alteration within the pathway may influence mechanisms and effects of pathway deregulation [45,[49][50][51]. Accordingly, KRAS mutations were mutually exclusive with other genetic alterations (except for ADC-23 who presented simultaneous presence of KRAS mutation and polysomy of AKT1 and AKT2) whereas copy number variations of PIK3CA, AKT1 and AKT2 were not [52]. These findings are reminiscent of breast or endometrial cancer, in which PIK3CA mutations are frequently detected in settings of low PTEN expression or mutations [53,54], and suggest that genetic alterations of the PI3K/AKT pathway in NSCLCs are not functionally redundant.
In addition, this manuscript provides novel experimental evidence to the observation that SCCs and ADCs develop by different genetic alterations: i) mutations in PIK3CA and AKT1 (3% altogether) were detected only in SCCs [this manuscript ; 24] whereas KRAS mutations were observed in ADCs (19%); ii) SCC patients (85%) presented at least one genetic alteration in PI3KCA, AKT1, AKT2 or PTEN more frequently than ADC patients (50%); iii) PIK3CA copy gains occurred more frequently in SCCs (25%) than in ADCs (9%) as described previously [15,49]; iv) coexistence of at least two alterations in the members of the PI3K pathway occurred more frequently in SCC patients (45%) than in ADC patients (8%).
FISH results indicated that gene amplification of the PIK3CA gene at 3p21 is responsible for ,20% of cases with enhanced p110a expression, in agreement with previous reports indicating that gains of part or of the entire long arm of chromosome 3, where the PIK3CA gene maps, are recurrent in NSCLCs [15,55,56]. Yet, since several NSCLCs overexpress p110a in the absence of gene amplification other mechanisms must be involved in the dysregulation of PIK3CA expression in NSCLCs.
The functional effects of mutant or amplified PIK3CA in lung cancer are unclear [49]. Our data indicated that in NSCLC cells, PIK3 signalling is required in vitro and in vivo since suppression of p110a expression inhibits the growth of xenografted cells carrying an activated PIK3CA allele. However, it is likely that PI3K might act in concert with other oncogenic hits to promote malignant transformation of lung epithelial cells since several NSCLCs present aberrant expression of AKT1, AKT2 or loss of PTEN in addition to PIK3CA overexpression (7%, 10% and 21%, respectively) and PIK3CA mutations are not mutually exclusive with EGFR and KRAS mutations in lung cancer [49][50][51]54].
Finally, RNA profiling experiments led to the identification of .2000 differentially regulated transcripts that likely contributes to the oncogenic effects of aberrant PI3K signalling in lung epithelial cells. Categorization of differentially expressed genes into biological pathways and/or functions identified gene expression changes induced by the constitutive activation of PI3K-dependent signalling in lung epithelial cells. Interestingly, analysis of DEGs retrieved several functions linked to lung cancer of both ADC and SCC histotypes, suggesting that the activation of PI3K signalling induces a transcriptional programme that is characteristic of lung cancer cells. In this sense, it is worth noting that IPA analysis identified a network of transcription factors that are linked to PI3K activation -such as MYC, JUN, JUN-B, FOS, HMGA1 and HES1-that are the central nodes of multiple molecular networks up-regulated by constitutive PI3K signalling. These findings suggest that part of the oncogenic activity exerted by PI3K in lung epithelial cells is dependent on the ability of PI3K to reprogram transcription. The existence of a correlation between PI3K signalling and the expression of oncogenic transcription factors is confirmed by the finding that cell lines and primary tumours with high AKT activation present, on average, consistently higher expression of MYC, FOS and HMGA1 than cell lines and tumours with low AKT activation.
It is of note that in agreement with its role in promoting cells cycle progression, active PI3K up-regulates the expression of several cell cycle promoting molecules -CCND1, CCND2, Cdk6 -as well as down-regulates Cdk inhibitors. To this regard, it is worth noting that PI3K-dependent regulation of CCND2 expression may occur indirectly through MYC [57].
In conclusion, the results reported in this manuscript indicate that PI3KCA over-expression occur at a much higher frequency in lung cancers than do activating mutations, apparently representing the major determinant of AKT activation in NSCLC. PI3KCA overexpression in NSCLCs occurs, at least in part, through gene copy gains, which occur more often in SCCs than in ADCs. Finally, it is of particular interest the identification of a network of transcription factors that are upregulated by constitutive PI3K signalling and may represent critical mediators of the oncogenic effects exerted by aberrant PI3K.        Table S9 Summary of the genetic alterations in the PI3K/AKT pathway in SCC patients. Copy number gains in AKT1, AKT2, PI3KCa genes were determined by FISH: high polysomy (HP) and gene amplification (A). Mutation analysis identified activating mutations of PI3KCA (E545K), KRAS (G12C, G12V, G12A, G13C) and AKT1(E17K). PTEN expression was classified as (+) when staining was detected in .50% of the cells, (+/2) when staining was detected in 25-50% of cells and (2) when staining was detected in 0-25% of cells. AKT activation was evaluated with phospho-specific antibodies (pS473), scored as negative (,10% of the tumour cells with weak, focal immunopositivity or absence of staining) and high (.10% of tumour cells with strong or diffuse immunopositivity).

(DOCX)
Table S10 Summary of the genetic alterations in the PI3K/AKT pathway in ADC patients. Copy number gains in AKT1, AKT2, PI3KCa genes were determined by FISH: high polysomy (HP) and gene amplification (A). Mutation analysis identified activating mutation of PI3KCA (E545K), KRAS (G12C, G12V, G12A, G13C) and AKT1(E17K). PTEN expression was classified as (+) when staining was detected in .50% of the cells, (+/2) when staining was detected in 25-50% of cells and (2) when staining was detected in 0-25% of cells. AKT activation was evaluated with phospho-specific antibodies (pS473), scored as negative (,10% of the tumour cells with weak, focal immunopositivity or absence of staining) and high (.10% of tumour cells with strong or diffuse immunopositivity. (DOCX) Table S11 Genes significantly increased or decreased in BEAS-2B vs BEAS-PI3KCA-E545K. Expression microarray (HT-12_V3_0_R1) data were prefiltered to remove genes changing less than 1.5 fold, and a t-test was run to determine significant (p,0.01) changers. A multiple testing correction using the algorithm of Benjamini and Hochberg was used to reduce the false discovery rate. See file attached. (XLS)