The level of AKT phosphorylation on threonine 308 but not on serine 473 is associated with high-risk cytogenetics and predicts poor overall survival in acute myeloid leukaemia

The phosphoinositide 3-kinase/Akt pathway is an important signalling pathway governing cell survival and proliferation in acute myeloid leukaemia (AML). As full activation of Akt requires phosphorylation on both threonine 308 (Thr308) and serine 473 (Ser473) residues, we studied the level of phosphory- lation on the both sites in 58 AML samples by flow cytometry. The ratio of the mean fluorescence intensity of Thr308 and Ser473 represented a continuum ranging from 0.3 to 5.6 and from 0.4 to 2.87, respectively. There were no significant correlations between age, gender, French-American-British classification, leukocytosis, FLT3-ITD and Akt phosphorylation. However, the level of phosphorylation on Thr308, but not on Ser473, was significantly correlated with high-risk karyotype. Thr308high patients had significantly shorter overall survival (11 vs 47 months; P 0.01), event-free survival (9 vs 26 months; P 0.005) and relapse-free survival (10 months vs not reached; P 0.02) than Thr308low patients. Neither screening for AKT1 E17K mutation nor changes in the level of PTEN expression and phosphorylation could be linked to increased phosphorylation on Thr308 in high-risk cytogenetic AML cells. However, PP2A activity was significantly reduced in high-risk samples com- pared with intermediate-risk samples. Moreover, the specific Akt inhibitor, Akti-1/2, inhibited cell proliferation and clono- genic properties, and induced apoptosis in AML cells with high-risk cytogenetics, suggesting that Akt may represent a therapeutic target in high-risk AML.

Keywords: acute myeloid leukaemia; PI3K/Akt pathway; phosphorylation; Akt inhibitor; therapeutic target; complex karyotype

Introduction

The phosphoinositide 3-kinase (PI3K)/Akt pathway plays a key role in many cancer cells by controlling apoptosis, cell proliferation, differentiation and angiogenesis.1 Once activated by oncogenes, tyrosine kinase receptor, growth factors or even mutation, class I PI3K generates D3-polyphosphoinositides, which interact with the pleckstrin homology domain (PHD) of serine threonine kinase Akt (a member of the AGC kinase family), leading to membrane targeting. Akt activation is then dependent on phosphorylation of two residues located at the catalytic site (threonine 308, Thr308) and the C-terminal hydrophobic motif site (serine 473, Ser473). Thr308 phosphorylation is mediated by phosphoinositide-dependent kinase-1 (PDK1), whereas Ser473 is phosphorylated by PDK2, which has been recently identified as the mTOR complex 2.2,3 Several studies have shown that the PI3K/Akt pathway is deregulated in acute myeloid leukaemia (AML), contributing to inhibition of apoptosis, cell proliferation and drug resistance.4 However, the prognostic impact of Akt phosphorylation (phospho-Akt) in AML is unclear. In a retrospective study, constitutive phospho-Akt on Ser473 seemed to be an adverse prognostic factor for survival.5 Recently, the simultaneous activation of PKCa, ERK2, pERK2 and phospho-Akt pathways was shown to predict poor outcome in a large series of AML patients.6 In this study, although phospho-Akt was an adverse prognostic factor per se, the site of phospho-Akt determined by western blotting was not indicated. In another study, phospho- Akt on Ser473 also assessed by western blotting after serum starvation of leukaemic cells was correlated with better relapse- free survival (RFS) and overall survival (OS) in 92 young AML patients treated homogenously.7

As phosphorylation of both Thr308 and Ser473 is thought to be required for full kinase activity, examination of these two phosphorylation sites would allow a more complete analysis of the level of Akt activation in AML cells. Phospho-Akt has been mainly studied by western blotting, a technique that is not applicable in the routine setting. Recently, high levels of phospho- Akt on Ser473 were detected in AML patients by flow cytometry, correlating with Akt activity and western blot analysis.8,9 In this study, we used the same approach to analyse the phospho-Akt on both Thr308 and Ser473 in 58 young adults with AML. The level of phospho-Akt on Thr308, but not on Ser473, was significantly associated with high-risk cytogenetics and poor outcome. Enhanced Thr308 phosphorylation could be related, at least in part, to decreased activity of phosphatase PP2A observed in high- risk cytogenetic samples. Accordingly, high-risk cytogenetic samples were more sensitive to the Akt inhibitor, Akti-1/2, compared with intermediate-risk cytogenetic samples, suggesting that targeting Akt could be beneficial in this subgroup of AML patients with a poor prognosis on conventional chemotherapy.

Materials and methods

Cells and reagents

Samples were obtained from patients diagnosed with AML at the Department of Haematology, Hoˆpital Purpan, Toulouse, France,after obtaining their informed consent. The study was approved by the institutional review board. AML cells were isolated from the bone marrow at diagnosis by Ficoll–Hypaque density- gradient centrifugation and were cryopreserved in Iscoves modified Dulbecco medium (IMDM) containing dimethyl sulphoxide (10% final concentration) and foetal calf serum (FCS; 50% final concentration). Leukaemias were characterized according to their morphology (French-American-British (the FAB) classification), karyotype, immunophenotyping and FLT3 mutations (FLT3-ITD). Karyotype risk groups were defined according to the US Southwest Oncology Group criteria, which define the following chromosomal abnormalities as high-risk cytogenetic status: del(5q)/5, 7/del(7q), abn 3q, 9q, 11q, 20q, 21q, 17p, t(6;9), t(9;22) and complex karyotypes (X3 unrelated abnormalities).10 The human leukaemia cell line, KG1, was purchased from the American Type Culture Collection (Rock- ville, MD, USA) and cultured in IMDM containing 20% FCS. The clinical characteristics of the AML patients from whom samples were used for in vitro studies are shown in the Supplementary data (Supplementary Table S1). Antibodies were obtained from Cell Signalling (Beverly, MA, USA): phospho-Akt Thr308, phospho-Akt Ser473, PTEN, phospho-PTEN Ser380/ Thr382/383, anti-PP2A catalytic subunit (PP2Ac), phospho- FOXO3A (Thr32), phospho-GSK3b (Ser21/9) and GSK3b. Daunorubicin (DNR) was obtained from Sanofi-Aventis (Paris, France). The allosteric Akt kinase inhibitor targeting Akt1 and Akt2 (Akti-1/2) was kindly provided by Dr Claire Chaussade.11

Treatment

Between December 2000 and February 2005, 58 patients with de novo AML were treated with DNR (60 mg/m2, days 1–3) or idarubicin (IDA) (8 mg/m2, days 1–5) with cytosine arabinoside (AraC) (200 mg/m2, days 1–7), and a second course IDA (8 mg/m2 or DNR 35 mg/m2, days 17–18) AraC (1 g/m2/12 h, days 17–20) if 45% marrow blasts persisted on day 15. Complete response patients with an HLA-identical sibling received an allograft (directly
if they were p50 years of age; after consolidation and reduced- intensity conditioning if they were X51-year-old). Other patients
received one consolidation (IDA 12 mg/m2 or DNR 60 mg/m2, days 1–2 and AraC 3 g/m2/12 h, days 1–4), and then autologous transplantation or high-dose AraC (favourable karyotype).

Flow cytometry analysis

Leukaemic cells (1 105) were permeabilized with an Intraprep kit according to the manufacturer’s recommendations (Beckman Coulter, Miami, FL) and then incubated for 1 h at 4 1C in the dark with anti-phospho-Akt (Ser473 or Thr308) (Cell Signalling), or with an isotypic control. After two washes, the cells were incubated for 30 min in the dark with FITC-conjugated goat anti- rabbit antibody (BD Biosciences, Franklin Lakes, NJ). Samples were analysed with an Epics XL cytometer (Beckman Coulter) and a minimum of 10 000 events were recorded for each sample. The results were expressed as the ratio of the mean fluorescence intensity (rMFI) of the stained sample compared with the non-specific IgG control. Flow cytometric detection of primitive leukaemia cells (CD34 þ CD38— CD123 þ cells) is described elsewhere.9

AKT1 E17K mutational screen in AML with TaqMan single nucleotide polymorphism genotyping assays

DNA from 148 AML samples was subjected to a TaqMan PCR- single nucleotide polymorphism genotyping assay as described earlier.12 Each assay contained 6.25 ml of TaqMan Universal PCR Master Mix, 0.3125 ml of 40 × assay mix and 5 ml of DNA (2 ng/ml) in a final volume of 12.5 ml. 40 × assay mix contained forward primer, reverse primer, 6FAM dye–MGB labelled probe and VIC dye–MGB labelled probe (Applied Biosystems, Foster City, CA, USA). (See Supplementary Table S2 for primer design for the G49A AKT1 mutation.) Three reactive lymph nodes served as negative controls. Double-stranded oligonucleotides, 62 bp in length (with or without the mutation), were synthesized to serve as a control. (See Supplementary Table S2 for oligo- nucleotide sequences) These allowed us to determine the theoretical limit of detection (5% of mutated allele). Serial dilutions were tested: 0% (negative control), 5, 10, 25, 50, 75, 90, 95 and 100% (positive control).

Western blotting

Leukaemic cells (1 106), resuspended and washed twice in cell culture medium without serum, were denatured in Laemmli sample buffer for 5 min at 100 1C. Proteins were resolved by polyacryla- mide-sodium dodecyl sulphate gel electrophoresis (SDS-PAGE) and then transferred onto nitrocellulose membrane (Hybond-C super; GE Healthcare, Chalfont St Giles, UK) using a liquid transfer apparatus (GE Healthcare). The membrane was blocked overnight at 4 1C in Tris-buffered saline (TBS) containing 1% fat-free milk and 1% bovine serum albumin (BSA). The proteins were detected by blotting with the appropriate monoclonal or polyclonal antibodies in TBS, 0.1% Tween, 1% fat-free milk and 1% BSA. For PTEN quantification, membranes were incubated in the dark with Alexa Fluor 680-conjugated secondary antibody (Invitrogen, Carlsbad, CA, USA) and read with an Odyssey infrared scanner (Li-Cor Biosciences, Lincoln, NE, USA). The signal was quantified using Odyssey 2.1 software (Li-Cor Biosciences).

PP2A assays

PP2A assays were carried out using a PP2A immunoprecipita- tion phosphatase assay kit (Millipore, Bedford, MA, USA). Briefly, protein lysates (2 × 106 cells) in 50 mM Tris/HCl pH 7.4,150 mM NaCl, 1 mM EDTA and 1% NP40 were incubated for 2 h at 4 1C with 4 mg of PP2Ac antibody (clone 1D6) and Protein A-agarose. After three washes, immunoprecipitates were used in a phosphatase reaction according to the manufacturer0s instruc- tions. As an internal control, the specificity of the reaction was assessed by inhibiting PP2A activity with 1 nM okadaic acid before titration. The amount of immunoprecipitated PP2A was also assessed using anti-PP2Ac western blots.

Apoptosis

AML cells were incubated with or without 0.2 mM DNR for 24 h in the presence or absence of 10 mM Akti-1/2. After incubation, cells were washed twice in PBS and stained with annexin- V-FITC (BD Biosciences). Apoptosis was quantified by flow cytometry as the percentage of annexin-V-positive cells. For some patients’ samples, cells were also stained with CD34-PE Cy7, CD38-APC and CD123-PE antibodies (BD Biosciences) to identify the primitive CD34 þ CD38— CD123 þ leukaemic subpopulation.

Clonogenic assays

The median percentage of leukaemic cells after Ficoll separation was 80% (range, 34–96%). AML cells were adjusted to a final concentration of 1 105 cells/ml and grown in H4230 methyl cellulose medium (Stem Cell Technologies, Vancouver, BC, USA) supplemented with 10% 5637-conditioned medium as a stimulant (5637 is a bladder carcinoma cell line secreting GM-CSF, G-CSF, IL-1b, M-CSF and SCF)13 and increasing concentrations of Akti-1/2 (0, 1, 10 and 100 nM). The cells were then plated in 35 mm petri dishes in duplicate and incubated for 7 days in a humidified CO2 incubator (5% CO2, 37 1C). Leukaemia colony forming units (L-CFUs) were counted under an inverted microscope on day 7.14 and expressed as the rMFI of the stained AML sample and the isotypic control. The clinical characteristics of the AML patients are shown in Table 1. The numerical values obtained for both Thr308 (n ¼ 53) and Ser473 (n ¼ 52) represented a continuum ranging from 0.3 to 5.6 and 0.4 to 2.87, respectively. The median level of phosphorylation was 2.3 (±0.89) and 1.3 (±0.27) for Thr308 and Ser473, respectively. The level of total Akt assessed by western blotting was stable, indicating that the elevated levels of phospho-Akt were not due to changes in protein expression (data not shown). As described earlier, normal CD34 þ cells from healthy donors displayed a low level of phospho-Akt (data not shown).

Statistical analysis

Continuous variables were transformed to categorical variables using their median as a cut-off value. Pairwise comparisons between continuous variable distributions were carried out with the Kruskal–Wallis test and Fisher’s exact test for categorical variables. Kaplan–Meier curves were plotted for OS, event-free survival (EFS) and RFS and compared using the two-sided log- rank test. OS end-points, measured from time of diagnosis, were death (failure) or survival at the last follow-up (censored). EFS end-points, measured from time of diagnosis, were death or relapse (failure) and survival at the last follow-up (censored). RFS end-points, measured from time of complete remission, were death or relapse (failure) and survival at the last follow-up
(censored). Cox regression analysis related OS, EFS and RFS to analysed parameters. Results were significant at Po0.05. Statistical analysis was carried out using SAS V8.02 (SASInstitute Inc. Cary, NC, USA). For in vitro studies, different groups were compared using a parametric paired t-test.

To analyse the characteristics of the AML patients according to the level of phospho-Akt, patients were stratified according to their median rMFI (±2.3 for Thr308 and ±1.3 for Ser473). Patients with an rMFI value 42.3 or 41.3 were referred to as Thr308high or Ser473high, respectively, whereas the others were referred as Thr308low or Ser473low. There were no significant correlations between age, gender, FAB classification, white blood cell (WBC) count, FLT3-ITD and phospho-Akt on either residue. Interestingly, the intensity of phosphorylation of Thr308, but not of Ser473, significantly correlated with cytogenetic risk. Indeed, the median rMFIs on Thr308 were 1.95 (±0.37), 2.2 (±1) and 3.1 (±0.49) for low-risk, intermediate-risk and high-risk karyotypes, respectively (P ¼ 0.0069) (Figure 1 and Table 1).

Results

Correlation between Akt Thr 308 and cytogenetic risk in AML patients

The levels of phospho-Akt on Thr308 and Ser473 were evaluated by flow cytometry in pretreatment marrow samples high-risk cytogenetics has never been reported earlier. To increase the number of high-risk patients in this series, we evaluated six additional AML samples with high-risk cyto- genetics (n 17) (Table 2). The median rMFI on Thr308 for the 17 samples tested was 3.1 (±0.8). When compared with other cytogenetic subgroups (n ¼ 47) in which the median rMFI was Impact of Thr308 phosphorylation on clinical outcome All patients were treated by induction chemotherapy with anthracyclins and AraC. Patients in complete remission were then referred for consolidation chemotherapy, autologous or allogeneic stem cell transplantation according to the cytogenetic risk. The complete response rate was 83% with no significant difference between Thr308high, Thr308low, Ser473high and Ser473low subgroups. OS, EFS and RFS were 22.1, 14.1 and 15.6 months, respectively. In univariate analysis, WBC count and Ser473 phosphorylation did not influence outcome, whereas unfavourable karyotype, FLT3-ITD and Thr308high correlated with poor OS, EFS and RFS, and age only with OS and RFS. The impact of Thr308high on prognosis was validated using two analytical strategies. Indeed, the phosphorylation level was studied either as a categorical or a continuous variable. Thr308high patients had significantly shorter OS (11 vs 47 months; P 0.01), EFS (9 vs 26 months; P 0.005) and RFS (10 months vs not reached; P 0.02) than Thr308low patients (Figure 2). Age, FLT3-ITD, WBC count, Thr308high and cytogenetic risk were included in the Cox model for multivariate analysis. Only cytogenetic risk independently predicted worse OS (hazard ratio (HR) ¼ 2.58; P ¼ 0.001), EFS (2.18; P ¼ 0.003).

Figure 1 Phosphorylation of Akt (phospho-Akt) on Thr308 observed by flow cytometry according to cytogenetic status. After permeabiliza- tion, leukaemic cells were incubated with anti-phospho-Akt (Thr308) or non-relevant IgG and then with FITC-conjugated goat anti-rabbit antibody. A representative histogram of three AML samples according to their cytogenetic background is shown: low-risk t(8;21), inter- mediate-risk (normal karyotype) and high-risk (complex karyotype).

Absence of Akt mutation in AML with high-risk cytogenetics

We and others have earlier shown that no somatic mutations of class IA PI3K isoforms (PIK3CA and PI3KD) can be found in AML samples.15,16 A recurrent novel AKT1 PHD mutation (AKT1 E17K) leading to membrane translocation, constitutive AKT activation and leukaemia development in mice has been recently described.17 To explain the differences in phospho-Akt according to cytogenetic risk, we sequenced specimens from an independent series of 148 AML patients. Samples bearing t(8;21) (n ¼ 14), inv16 or t(16;16) (n ¼ 21), normal karyotype with (n ¼ 26) or without (n ¼ 49) FLT3-ITD mutations or complex karyotype (n 38) were assessed for the presence of AKT1 E17K.No mutations could be detected in these different AML subgroups. This result is consistent with data reported recently from smaller series.18,19

Status of PP2A, PTEN and PDK1 in AML with high-risk cytogenetics

To explore the mechanisms of Thr308 phosphorylation in high-risk cytogenetic AML, we assessed the activation status of its main regulators. The tumour suppressor, PTEN, is a dual-specificity phosphatase acting as a negative regulator of PI3K/Akt signalling by dephosphorylating phosphatidylinositol-3,4-bis- phosphate (PI(3,4)P2) and phosphatidylinositol-3,4,5-trisphos- phate (PI(3,4,5)P3) at the D3 position of the inositol ring.20 Although neither inactivating mutations nor loss of heterozygo- sity have been shown in AML, the level of protein expression varies between studies but has never been correlated with cytogenetic sub-classes of AML.4,21 PTEN phosphorylation at the C-terminal regulatory domain is known to induce a decrease in PTEN activity towards its lipid substrates.22,23 It has been shown that PTEN phosphorylation on Ser380/Thr382/Thr383 residues is associated with an increase in phospho-Akt and poor OS in AML.24 Therefore, we checked both the level of PTEN protein expression and its phosphorylation on Ser380/Thr382/ Thr383 residues using the Odyssey infrared imaging system to accurately quantify protein and protein phosphorylation by western blotting. As shown in Figure 3a, PTEN was expressed and phosphorylated in both high-risk and non-high-risk cytogenetic samples, with no significant differences between these two groups (phospho-PTEN quantification of 83.7±6 for high- risk vs 88.2±6.9 for other risk groups, n 7).

Figure 2 Prognostic value of Thr308high in AML. AML patients were stratified into two groups, Thr308high or Thr308low, according to the value of phospho-Akt Thr308 (rMFI 42.3, Thr308high and rMFI p2.3, Thr308low). Thr308high patients had significantly shorter overall survival (11 vs 47 months; P 0.01), event-free survival (9 vs 26 months; P 0.005) and relapse-free survival (10 months vs not reached; P 0.02) than Thr308low patients.

The phosphorylation/activation status of the kinase responsible for Thr308 phosphorylation, PDK1, was assessed in 16 AML samples (eight high-risk and eight intermediate-/low-risk) using the phospho-Ser241 antibody. PDK1 phosphorylation did not differ significantly between these two groups (data not shown).

PP2A is a major serine threonine/phosphatase in eukaryotic cells, which is known to downregulate the effects of activated protein kinases. Akt is inactivated by PP2A in vitro and its activity increases in cells treated with the serine threonine phosphatase inhibitor, okadaic acid.25 Recently, it has been shown that functional loss of PP2A activity is important for chronic myeloid leukaemia (CML) blastic transformation.26 In both CML blast crisis and Ph þ acute lymphoblastic leukaemia cells, restauration of PP2A activity is associated with inhibition of phospho-Akt. These results prompted us to assess the activity of PP2A in AML cells. PP2Ac was immunoprecipitated from both complex and intermediate-risk karyotype samples and its phosphatase activity was tested using a specific assay. As shown in Figure 3b, amounts of PP2A immunoprecipitates were comparable among the tested samples. Interestingly, PP2A activity was significantly reduced by 40% in complex karyo- types (550±80 pmol of released phosphate in 15 min, n 7) compared with intermediate-risk karyotypes (932±147 pmol of released phosphate in 15 min, n ¼ 6) (P ¼ 0.03). The specificity of PP2A activity was confirmed by its complete inhibition in the presence of 1 nM okadaic acid. Overall, these results indicate that the expression of PTEN and its phosphorylation do not change significantly in the cytogenetic subtypes studied, whereas PP2A activity was reduced in AML with high-risk cytogenetics.

Preferential activity of the Akt inhibitor, Akti-1/2, in AML cells, including primitive CD34 þ CD38— CD123 þ cells, with high-risk cytogenetics
As high-risk cytogenetic samples display high levels of Akt activation, it can be hypothesized that these cells use this pathway to survive and to escape genotoxic stress induced by chemotherapeutic agents. To specifically assess the role of Akt in high-risk cytogenetic AML cell survival and proliferation, we used the specific Akt inhibitor targeting Akt1 and Akt2 (Akti-1/2). In contrast to the phosphoinositide analogue perifosine, another Akt inhibitor whose activity has been assessed in AML cells, this novel allosteric inhibitor inhibits the kinase activity of Akt and blocks its phosphorylation and activation by PDK1.27,28 At a concentration of 1 mM, it inhibits full-length AKT1 or CaMK1 by 80%, whereas other kinases are also inhibited when used at higher concentrations.29 We first tested this compound on the CD34 þ CD38—PgP þ KG1 cell line bearing complex chromo- somal abnormalities frequently found in high-risk AML such as —5,del(7)(q21;q35) and 17p—. In KG1, the level of phosphory- lation on Thr308 was remarkably high when compared with Ser473 (Figure 4a). Incubation of KG1 cells with 10 mM Akti-1/2 for 4 h potently inhibited phospho-Akt on both sites. Two downstream Akt targets, GSK3b and FOXO3A, were also dephosphorylated by Akti-1/2 (Figure 4b). KG1 cells were incubated in the presence of increasing concentrations of the drug, and cell proliferation was evaluated by methyl thiazolyl tetrazolium (MTT) assays. As shown in Figure 4c, Akti-1/2 potently inhibited KG1 cell proliferation in a dose-dependent manner.

In leukaemic cells from high-risk cytogenetic AML patients (mainly with -5/5q—; -7/7q— and 17p— chromosomal abnorm- alities), Akti-1/2 also inhibited the phospho-Akt (Figure 5a).

Figure 3 Assessment of phosphatases, PTEN and PP2A, in high-risk cytogenetic AML. (a) Levels of PTEN and phospho-PTEN (Ser380/Thr382/ Thr383) were analysed by western blotting using specific antibodies and quantified according to Odyssey technology. A representative figure of high-risk vs low-risk samples is shown. (b) PP2A assay on untreated (black bars) and okadaic acid-treated (white bars) AML cells. Results are expressed as pmol of released phosphate. The difference between intermediate-risk and high-risk cytogenetics is P 0.037 (*). As shown for patients #18 and #4 (bottom panel), the amount of immunoprecipitated PP2A was monitored by western blotting using an anti-PP2Ac antibody.

As fresh AML cells hardly proliferate in liquid culture, we used a clonogenic assay in methylcellulose conditioned with haema- topoietic growth factors to study the effect of Akti-1/2 on samples from 12 AML patients (seven with high-risk and five with intermediate-risk cytogenetics). As shown in Figure 5b, Akti-1/2 inhibited the capacity of AML cells to generate L-CFU in a dose-dependent manner in all samples tested. However, 100 nM Akti-1/2 reduced the clonogenicity of 5/7 high-risk cytogenetic samples by 450%, whereas the magnitude of this reduction did not reach 50% in 5/5 intermediate-risk cyto- genetic samples. Mean percentages of L-CFU recovery with 10 nM and 100 nM Akti-1/2 were 51.6±3 and 44±3% in high- risk cytogenetic samples vs 76±4 and 67±4% in intermediate-risk samples, respectively (P 0.001 for both concentrations). We then investigated whether decreased proliferation induced by Akti-1/2 was due to apoptosis in four fresh leukaemic samples with high-risk cytogenetics. After 24 h incubation with Akti-1/2 (10 mM), features of apoptosis were detected by annexin-V staining (25±3% of annexin-V-positive cells vs 14±2% for controls) (Figure 5c). The capacity of Akti-1/2 to enhance the activity of conventional chemotherapeutic agents was also assessed in the same samples. As shown in Figure 5c,the cytotoxic effect of 0.2 mM DNR was increased when cells were incubated with 10 mM Akti-1/2 (37±2.9%, Akt-i þ DNR vs 25±3.5%, DNR alone; P ¼ 0.05).

As shown earlier, phospho-Akt can be detected in phenoty- pically primitive AML stem cells expressing the CD34 þ CD38— CD123 þ phenotype (Figure 6a).9 This compartment is generally resistant to genotoxic agents, such as DNR and AraC, compared with non-stem-cell AML fractions.31,32 Accordingly, 0.2 mM DNR did not induce significant cell death in CD34 þ CD38— CD123 þ leukaemic cells from high-risk cytogenetic samples (n 3) (Figure 6b and c). Interestingly, Akti-1/2 alone induced an apoptotic response in this primitive leukaemic subpopulation (13.5±1.4% treated cells vs 4.6±0.9% un-
treated cells; P 0.01). Moreover, the combination of Akti-1/2 and 0.2 mM DNR induced an additive effect on apoptosis (21.5±0.3%, Akti-1/2 þ DNR vs. 13.5±1.4%, Akti-1/2; P ¼ 0.0126) (Figure 6b and c).

Discussion

Flow cytometry is the optimal method to evaluate the activation state of cell signalling pathways involved in leukaemic cell proliferation and survival. This technique requires a limited amount of cells that can easily be collected, can be performed rapidly and can detect target cells by double immunostaining in samples with a low percentage of blast cells. Flow cytometry can therefore be used to assess the prognostic impact of the activation status of several kinases. It also allows a rapid analysis of potential therapeutic intracellular targets in blast cells at diagnosis to select patients for clinical trials assessing new kinase inhibitors. Moreover, four-colour flow cytometry can be used to detect the phosphorylation status of kinases in primitive leukaemic cells with CD34 þ CD38— CD123 þ phenotype, which are enriched in SCID leukaemia-initiating cells.31 In AML, two earlier studies using this technique have shown that the levels of Akt and ERK phosphorylation are easily detectable in different blast cell populations and are not affected by cryopreservation.

Figure 4 Activity of Akti-1/2 on KG1 proliferation. (a) Phospho-Akt on Thr308 and Ser473 was assessed in KG1 cells by western blotting (right panel) and flow cytometry (left panel) using the antibodies indicated. Cells were cultured for 4 h alone (full line) or in the presence of 10 mM Akti-1/ 2 (hatched line) and then subjected to the flow cytometry analysis. (b) KG1 cells were incubated in the presence of increasing concentrations of Akti-1/2 for 4 h and then lysed and analysed by western blotting with the indicated antibodies. (c) KG1 cells were cultured in liquid medium alone or with increasing doses of Akti-1/2 for 3 days. Cell viability was quantified by MTT assay. Results are the mean±s.e.m. of three independent experiments performed in triplicate (*Po0.05).

Figure 5 Activity of Akti-1/2 on high-risk cytogenetic AML cells. (a) Cells from a patient with complex karyotype AML were cultured for 4 h alone (control) or in the presence of 10 mM Akti-1/2 and then assessed for phospho-Akt (Thr308 and Ser473) by flow cytometry using the indicated antibodies. (b) Cells from seven patients with high-risk (white bars) and five with intermediate-risk (black bars) AML were grown in clonogenic assays in the presence of increasing doses of Akti-1/2 (0, 1, 10 and 100 nM). Results are presented as a percentage of the control and are the mean±s.e.m. of duplicate determinations. (c) Fresh leukaemic cells from four patients with high-risk cytogenetics were incubated for 24 h with or
without 10 mM Akti-1/2 and/or 0.2 mM daunorubicin and then processed for apoptosis studies using annexin-V staining. Results are presented as the mean percentages (±s.e.m.) of annexin-V-positive cells in treated and untreated cells.

Figure 6 Activity of Akti-1/2 on primitive CD34 þ CD38— CD123 þ leukaemic cells. (a) Flow cytometry detection of phospho-Akt on Thr308 in CD34 þ CD38— CD123 þ cells. (b, c) AML cells (n 3) were cultured for 24 h alone or in the presence of 10 mM Akti-1/2 and/or 0.2 mM daunorubicin and then processed using four-colour immunostaining. Induction of apoptosis in primitive AML cells was assessed using annexin-V staining after gating on CD34 þ CD38— CD123 þ cells. A representative AML sample is shown in (b). Results are presented as mean percentages (±s.e.m.) of annexin-V positive cells in treated and untreated cells.

In this study, we report for the first time that the intensity of phospho-Akt on Thr308 is significantly correlated with cyto- genetic risk in AML, particularly with a complex karyotype. Conversely, this phosphorylation is quite low in core binding factor AML (i.e., inv16 and t(8;21)), as shown earlier in a few samples assessed by western blotting. Indeed, in Tazzari’s study, samples bearing inv16 chromosomal abnormalities displayed a low level of phosphorylation on Thr308 and low kinase activity.8 The mechanisms involved in the phosphorylation of Thr308 in the high-risk cytogenetic subgroup are unclear. Although a direct relationship has not been firmly established, the data presented here show a correlation between decreased PP2A activity and Thr308 phosphorylation. Regulation of Akt activation is a process tightly controlled at different levels by the activity of lipid phosphatases, such as PTEN, acting upstream of Akt, and protein phosphatasesFincluding PP2A and PHD leucine-rich repeat protein phosphatase (PHLPP)Fwhich directly regulate phospho-Akt.33 Interestingly, it has recently been shown that PHLPP, a PP2C-like phosphatase, specifically dephosphorylates Ser473, whereas PP2A preferentially dephos- phorylates Thr308.34 PP2A, a tumour suppressor with a crucial role in the regulation of cell cycle progression, survival and differentiation, is required for the development and mainte- nance of other myeloid malignancies such as blast-crisis CML cells.26 However, the mechanisms leading to this reduced PP2A activity in high-risk cytogenetic AML cells remain to be determined. One potential scenario could involve SET, a potent physiological PP2A inhibitor, which is deregulated in acute leukaemia through chromosomal abnormalities, such as t(6;9)(p23;q34) or 11q23 rearrangements.

However, as residual okadaic acid-sensitive PP2A activity is observed in high-risk cytogenetic AML cells, other upstream regulators of Akt may also contribute to the regulation of Thr308 phosphorylation. The PDK1 is the upstream kinase for Thr308 and phosphorylation of this residue is dependent on the conformation of Akt.37 Indeed, engagement of the Akt PH domain on the membrane by binding to D3-polyphosphoinosi- tides relieves auto-inhibition of the active site allowing PDK1 to access Thr308 in the activation loop.38 Although PDK1 is over- expressed in 45% of AML patients, its expression has not been reported to be associated with high-risk cytogenetics to date.39 Furthermore, phosphorylation of PDK1 on Ser241 did not significantly differ according to cytogenetic subgroup in our study. Another upstream regulator, PTEN, counteracts PI3K activity upstream of Akt and thus, regulates its phosphorylation on both Thr308 and Ser473. In AML cells, the role of PTEN in Akt activation is unclear. Although PTEN expression has not been studied extensively in AML, it has been shown that its expression is reduced or even absent in some AML patients.40,41 Complex AML karyotypes are characterized by genomic instability and loss of p53 function. p53 transcriptionally regulates PTEN expression, which in turn inhibits PI3K/Akt and subsequently MDM2, a negative regulator of p53, in a positive feedback loop.42 Thus, loss of p53 may contribute to the aberrant activation of Akt in these samples. However, we found that the level of PTEN expression was similar in high-risk compared with low-/intermediate-risk samples. The regulation of PTEN by phosphorylation at the C-terminal regulatory domain stabilises the molecule and leads to a decrease in PTEN activity towards its substrate.22,23 We found that PTEN is phosphorylated in AML cells with no difference between high-risk and intermediate-risk samples. The impact of PTEN phosphorylation is not clear but it seems to be a common phenomenon in all subtypes of AML that could contribute to increase the levels of phospho-Akt on both Thr308 and Ser473. However, this mechanism in cooperation with PP2A loss of function could lead to enhanced Thr308 phosphorylation in high-risk samples. Overall, it is conceivable that deregulation of the PI3K/Akt pathway in AML relates to phosphatase defects as no activating mutations of PI3K or Akt have been shown to date.

In our study, the level of phosphorylation on Thr308 was a strong prognostic factor for OS, EFS and RFS, butFin contrast with earlier studiesFSer473 was not associated with either a poor or a favourable prognosis.5,7 In non-small-cell lung cancer,it has also been reported that Thr308, but not Ser473, has prognostic value.43 These discrepancies may reflect the differ- ence in number of cases analysed, chemotherapy regimens, differences in the preparation of leukaemic cells, or technical differences in staining and scoring, thus warranting further well- designed prospective studies in larger series of AML patients. This study is ongoing in our institution, as phosphorylation of the two Akt sites is part of the immunophenotypic panel, which is currently being used in the diagnosis of all patients. Interest- ingly, the difference in the prognostic value of phosphorylated Thr308 compared with Ser473 may also reflect the fact that the specific downstream effectors of Akt when highly phosphory- lated on Thr308 may be the most relevant in AML cell biology and chemoresistance. Indeed, a recent study has shown that dual phosphorylation of Thr308 and Ser473 is not necessarily required for all Akt functions. Thr308 phosphorylation alone is sufficient to regulate downstream effectors such as GSK3, TSC1/ TSC2 and mTOR complex 1, while Ser473 phosphorylation is crucial for FoxO1/3a activation.44

Although further studies are needed to fully assess the expression, activity and mutational status of regulators of Akt, our findings suggest that Akt inhibitors should be tested in clinical trials, particularly in AML with complex karyotype, which have a very poor prognosis even with intensive chemotherapy and allogeneic stem cell transplantation. It has been shown that the Akt inhibitor perifosine, an alkylphos- pholipid, which prevents binding of Akt to the plasma membrane, inhibits cell proliferation, induces apoptosis in THP-1 and MV 4-11 leukaemic cell lines and some cytogeneti- cally undefined AML patient samples.27 Promising preclinical anti-leukaemic activity has also been shown with the dual class IA PI3K/mTOR inhibitor, PI-103.45 Here, we show that the potent and specific inhibitor Akti-1/2 also inhibits cell prolifera- tion and clonogenic properties, and induces apoptosis in AML cells with high-risk cytogenetics. Moreover, this compound also enhances the cytotoxic activity of DNR. Interestingly, Akti-1/2
could target primitive CD34 þ CD38— CD123 þ cells, suggesting potential activity on leukaemic stem cells (LSC), although this remains to be shown in non-obese diabetic/SCID xenograft models. Furthermore, Akti-1/2 activity in this cellular fraction was enhanced by low-dose DNR, in line with earlier studies, suggesting that the mechanism of LSC death involves the combined inhibition of survival pathways and activation of stress pathways.

In conclusion, this study shows that AML cells with high-risk cytogenetics display a high level of phospho-Akt Thr308, correlating with a decrease in PP2A activity and a potent effect of the Akt inhibitor, Akti-1/2, on AML cell proliferation, survival and clonogenicity. This suggests that this anti-apoptotic pathway may be a valuable therapeutic target, particularly in high-risk cytogenetic AML. Thus, clinical trials are warranted to determine if PI3K/Akt inhibitors could improve therapeutic results specifically in this subgroup of very high-risk patients.