M4 and M5 acute myeloid leukaemias display a high sensitivity to Bortezomib-mediated apoptosis


Dr Ugo Testa, Department of Haematology, Oncology and Molecular Medicine, Istituto Superiore di Sanità, Viale Regina Elena 299, 00161 Rome, Italy. E-mail: ugo.testa@iss.it


The present study explored the sensitivity of leukaemic blasts derived from 30 acute myeloid leukaemia (AML) patients to Bortezomib. Bortezomib induced apoptosis of primary AML blasts: 18/30 AMLs were clearly sensitive to the proapoptotic effects of Bortezomib, while the remaining cases were moderately sensitive to this molecule. The addition of tumour necrosis factor-related-apoptosis-inducing ligand, when used alone, did not induce apoptosis of AML blasts and further potentiated the cytotoxic effects of Bortezomib. The majority of AMLs sensitive to Bortezomib showed immunophenotypic features of the M4 and M5 French–American–British classification subtypes and displayed myelomonocytic features. All AMLs with mutated FLT3 were in the Bortezomib-sensitive group. Biochemical studies showed that: (i) Bortezomib activated caspase-8 and caspase-3 and decreased cellular FLICE [Fas-associated death domain (FADD)-like interleukin-1β-converting enzyme]-inhibitory protein (c-FLIP) levels in AML blasts; (ii) high c-FLIP levels in AML blasts were associated with low Bortezomib sensitivity. Finally, analysis of the effects of Bortezomib on leukaemic cells displaying high aldehyde dehydrogenase activity suggested that this drug induced in vitro killing of leukaemic stem cells. The findings of the present study, further support the development of Bortezomib as an anti-leukaemic drug and provide simple tools to predict the sensitivity of AML cells to this drug.

Although the treatment of patients suffering from acute myeloid leukaemia (AML) has considerably improved during the last 40 years, new chemotherapeutic agents are urgently required to further improve treatment protocols and prolong overall survival rates. The majority of AML patients still suffer disease relapse after achievement of initial disease remission (Deschler & Lubbert, 2006). Relapsing patients usually develop resistance to standard anti-leukaemic drugs and therefore need to be treated with new chemotherapeutic agents that act through other cellular pathways, to reduce the risk of cross-resistance and to improve response to treatment. Among these new chemotherapeutic agents, proteasome inhibitors show particular promise (Vink et al, 2006; Nencioni et al, 2007); one of them, Bortezomib, also known as PS-341 or Velcade, has been registered for the treatment of multiple myeloma.

The ubiquitin-proteasome pathway is involved in the degradation of polyubiquitinilated proteins and represents the main cellular machinery involved in protein degradation in eukaryotic cells. Intracellular proteins are targeted for degradation by the conjugation of polyubiquitin chains to lysine residues of the protein, through an enzymatic process mediated by the ubiquitin-conjugating enzymes (Mani & Gelmann, 2005). The proteasome is a proteolytic enzymatic complex that recognizes ubiquitin-tagged proteins and catalyses their proteolytic degradation in an ATP-dependent manner (Adams, 2004a,b; Mani & Gelmann, 2005). The proteasome machinery is found both in the cytoplasm and in the nucleus of eukaryotic cells. The proteasome typically contains a 20S component that is associated to a 19S or a 11S regulator component (Adams, 2004a,b; Mani & Gelmann, 2005). The 20S component comprises both α and β subunits: β1, β2 and β5 subunits are responsible for the proteasome enzymatic activities, chimotryptic-like (β5), tryptic-like (β2) and caspase-like (β1). Bortezomib acts mostly by inhibiting β5 and, to a lesser extent, β1 and β2.

In vitro studies have provided clear evidence that some types of cancer cells are very prone to undergo apoptosis in response to inhibition of the proteasome pathway, through molecular mechanisms that have been only partly elucidated (Adams, 2004a,b; Mani & Gelmann, 2005). Leukaemic cells also display a high sensitivity to apoptosis induced by proteasome inhibitors. In this context, initial studies by Kumatori et al (1990) showed that human leukaemic cells expressed abnormally high levels of proteasomes, compared with their normal counterpart. A recent study provided evidence that the activity patterns of the various proteasome subunits varied in primary leukaemic cells and reflected Bortezomib sensitivity of leukaemic cells (Kraus et al, 2007). Importantly, it was shown that proteasome inhibitors may induce apoptosis of leukaemic stem cells and that these stem cells are more susceptible to proteasome inhibition than normal stem cells (Guzman et al, 2002). Furthermore, novel proteasome inhibitors, such as PR-171 (Stapnes et al, 2007) and NPI-0052 (Miller et al, 2007), exerted a potent anti-proliferative and proapoptotic effect on leukaemic cells. Finally, Bortezomib was shown to co-operate with anti-leukaemic drugs (Minderman et al, 2007; Pigneux et al, 2007) in killing of leukaemic cells, independently of multidrug resistance mechanisms and p53 status (Minderman et al, 2007).

The proapoptotic effects induced by Bortezomib on cancer cells, including murine leukaemia cell lines (Sayers et al, 2003) and chronic lymphocytic leukaemia (CLL) (Kabore et al, 2006), were potentiated in many tumours by tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) through undefined molecular mechanisms. In many instances the stimulatory effects of Bortezomib on TRAIL receptor-1 (TRAIL-R1)/TRAIL-R2 expression and the downmodulation of c-FLIP seem to be key events in mediating the enhancement of Bortezomib-mediated apoptosis by TRAIL (Koschny et al, 2007; Liu et al, 2007; Saulle et al, 2007).

The anti-leukaemic activity of Bortezomib was tested in some clinical studies. Two phase I studies based on the administration of Bortezomib alone in AML patients showed haematological improvements, but no complete remissions were reached (Orlowski et al, 2002; Cortes et al, 2004). Subsequently, phase I combination studies have been carried out, where Bortezomib was combined with AML induction chemotherapy achieving a high rate of complete remissions, associated with acceptable toxicities (Attar et al, 2005; Orlowski et al, 2005).

The present study explored the proapoptotic effects of Bortezomib on 30 primary AMLs; the majority of them (60%) were highly sensitive to this proteasome inhibitor in vitro. Resistance/low sensitivity to Bortezomib was associated with high cellular FLICE [Fas-associated death domain (FADD)-like interleukin-1β-converting enzyme]-inhibitory protein (c-FLIP) levels.

Materials and methods


Fresh leukaemic blasts from 30 patients with AML, obtained after informed consent, were isolated from either bone marrow or peripheral blood by Ficoll-Hypaque density gradient centrifugation and immediately processed. All patients were consecutively diagnosed at the Divisions of Haematology of the University ‘La Sapienza’, of the University ‘Tor Vergata’. leukaemias were classified by morphological and cytochemical criteria according to the French–American–British (FAB) classification. All analysed samples showed infiltration by more than 70% leukaemic blasts. Approval for these studies was obtained from the Institutional Review Board of the Istituto Superiore di Sanità, Rome, Italy. Informed consent was obtained in accordance with the Declaration of Helsinki.

Immunophenotypic analysis of leukaemic cells

Analysis of cell surface antigens was performed by flow cytometry using a FACScan Flow Cytometer [Becton Dickinson (BD), Bedford, MA, USA]. The following antibodies directed to membrane antigens were used for standard immunophenotypic analysis of AML: anti-CD3, -CD7, -CD11a, CD11b, -CD11c, -CD13, -CD14, -CD15, -CD18, -CD19, -CD33, -CD34, -CD36, -CD38, -CD41, -CD45, -CD61, -CD64, -CD71, -CD90, -CD116, -CD117, -CD123, -CD131w, -CD235 and -HLA-DR (all from BD Pharmingen, San Diego, CA, USA or from BD). In addition, this study used the following monoclonal antibodies to characterize AML blasts: anti-vascular endothelial growth factor receptor-1 (VEGFR1), -VEGFR2, VEGFR3, -Tie-2, – macrophage colony-stimulating factor receptor (M-CSFR), -c-met, -insulin-like growth factor I receptor (IGF-1R) (all purchased from R&D Systems, Minneapolis, MN, USA) and -CD133 (Milteny, Bergische Gladbach, Germany). Anti-Flt3 mAb was purchased from Serotec (Oxford, UK). Cells were labelled and analysed as previously reported (Testa et al, 2002).

Cell apoptosis

In some experiments leukaemic blasts (5 × 105 cells/ml) resuspended in Stem Pro 34 medium (Gibco, BRL, NY, USA) containing 10% Fetal Calf Serum (FCS; Gibco) were grown in 24-multiwell plates (Falcon, BD) for either 24 or 48 h in the absence (control) or the presence of Bortezomib at a final concentration corresponding to either 0·01, 0·1, 0·25, 0·5 μmol/l or of 50 ng/ml recombinant human superkiller TRAIL (Alexis Corporation, Lausen, Switzerland) and the proportion of apoptotic cells was evaluated by annexin V binding assay using a kit from BD Pharmingen. Fluorescein isothiocyanate (FITC)-conjugated annexin V binds to phosphatidylserine, which becomes exposed on the cell surface in the early process of apoptosis.

Western blot analysis

To prepare total extracts the cells were washed twice with cold phosphate-buffered saline (PBS) and lysed on ice for 30 min with 1% Nonidet P-40 lysis buffer ( 20 mmol/l Tris–HCl pH8·0, 137 mmol/l NaCl, 10% glycerol, 2 mmol/l EDTA) in the presence of 1 mmol/l phenylsulphonyl fluoride, 1 mmol/l dithiothreitol, 1 mmol/l sodium orthovanadate, 2 μg/ml leupeptin, 2 μg/ml aprotinin.

Cell debris was removed by centrifugation at 4400 g for 10 min at 4°C, and protein concentration of supernatants was determined using the Bio-Rad protein assay (Bio-Rad Laboratories, Richmond, CA, USA). Aliquots of cell extracts containing 100 μg of total protein were resolved by 10% SDS-PAGE under reducing and denaturing conditions and transferred onto Hybond-C extra nitrocellulose membrane (Amersham, Buckinghamshire, UK).

Filters were blocked for 1 h at room temperature in 5% non-fat dry milk dissolved in Tris Buffered Saline-Tween (TBS-T) (10 mmol/l Tris–HCl pH 8·0, 150 mmol/l NaCl, 0·2% Tween 20) followed by incubation with primary antibodies. After washing in TBS-T buffer, the filters were incubated for 1 h at room temperature in 5% non-fat dry milk dissolved in TBS-T containing 1:4000 dilution of corresponding peroxidase-coniugated secondary antibodies.

Proteins were visualized with the enhanced chemiluminescence technique according to the manufacturer's instructions (Super Signal West Pico, Pierce, Rockford, IL, USA). Anti-caspase-3, -8 and -9 were purchased from Upstate (Upstate Biotechnology Lake Placid NY, USA); anti-c-FLIP and anti-FADD were purchased from Biosource (Camarillo, CA, USA); anti-Bcl-XL was obtained from R&D Systems (R&D System Inc.); anti-Bax and anti-Bcl-2 from BD Pharmingen (BD Pharmingen); anti-actin was purchased from Oncogene (Oncogene Research Products, Cambridge, MA, USA) and used as loading control.

Flow cytometry analysis of TRAIL and TRAIL-Rs

Leukaemic cells were incubated with 5 μg/ml of phycoerythrin (PE)-conjugated anti-TRAIL-R1, -R2, -R3, -R4 or anti-TRAIL (R&D System, Minneapolis, MN, USA) for 1 h at 4°C. The isotypic control antibody was mouse IgG conjugated with PE (R&D System). After three washes with PBS, cells were immediately analysed for fluorescence using FAC-Scan (BD).

Aldefluor cell analysis

Aldehyde dehydrogenase (ALDH) activity was evaluated by flow cytometry using the Aldefluor assay according to the procedure recommended by the supplier (StemCoBio, Durham, NC, USA). Briefly, cells were resuspended in the Aldefluor assay buffer containing the ALDH substrate, BAAA (1 μmol/l). In each experiment, a sample of cells was stained under identical conditions with 50 nmol/l of specific ALDH inhibitor diethylaminobenzaldheyde (DAEB) as negative control. Flow cytometry analysis was carried out using a FACSSCAN (Becton Dickinson Immunocytometry System, Mountain View, CA, USA). Aldefluor fluorescence was excited at 488 nm, and fluorescence emission was detected using a standard FITC 530/30-nm band-pass filter.

Statistical analysis

The statistical significance of the data was evaluated by One Way analysis of variance (anova) test, with Bonferroni post-test analysis, comparing all pairs of data. P-value <0·02 was used to define statistical significance.


Sensitivity of AML blasts to Bortezomib

The aim of this study was to determine the sensitivity of AML cells to the proteasome inhibitor Bortezomib. The first set of experiments evaluated the capacity of Bortezomib to induce apoptosis of fresh AML blasts isolated from 30 AML patients, whose main clinical and biological characteristics are reported in Table I. Leukaemic blasts were incubated for 24 h either in the absence (control) or the presence of either Bortezomib or TRAIL or Bortezomib + TRAIL.

Table I.   Immunophenotypic features of acute myeloid leukaemia patients. The bold line separates the first 12 patients, exhibiting a low sensitivity to Bortezomib, from the last 18 patients showing a high sensitivity to Bortezomib.
PatientFAB WBC (×109/l) CD33 (%) CD34 (%) Ac133 (%)c- kit (%)CD123 (MFI) CD11b (%) CD14 (%) CD36 (%)M- CSFR (%)Flt3 (MFI)IGF- RI (%) VEGF-R2 (%) Tie-2 (%)CD31 (MFI)c- met (%)VLA -4 (MFI) CXCR4 (%) TRAIL (%)TRAIL- R1 (%)TRAIL- R2 (%)TRAIL- R3 (%)TRAIL R4 (%)
  1. FAB, French–American–British classification; WBC, white blood cell; MFI, mean fluorescence intensity; M-CSFR, macrophage colony-stimulating factor receptor; IGF-R1, insulin-like growth factor I receptor; VEGF-R2, anti-vascular endothelial growth factor receptor-2; VLA, very late antigen; CXCR4, chemokine (C-X-C motif) receptor 4; TRAIL, tumour necrosis factor-related apoptosis-inducing ligand; ND, not determined.

123M2 7615105651193501400067445410003025
135M2 38115342220116030173722961616130028816
170M2 8878627751151316161610119719612200000
129M5261 1739389276688766754766142809391185033871
176M1158878 852383662315 3 0596000044

Dose-response experiments showed that Bortezomib induced apoptosis of AML blasts in a dose-dependent manner, reaching plateau levels at concentrations corresponding to 0·1–0·25 μmol/l (Fig 1). The contemporaneous addition of Bortezomib + TRAIL only moderately increased the proportion of apoptotic cells compared with the values observed incubating the cells with Bortezomib alone (Fig 1). The analysis of individual cases showed that in some AMLs the contemporaneous addition of TRAIL and Bortezomib did not increase the proportion of apoptotic cells compared with the values observed with Bortezomib alone (one representative example in Fig 2, panel A), while in other cases the addition of TRAIL together with Bortezomib clearly increased the level of apoptotic cells induced by Bortezomib alone (Fig 2, panel B). In some instances (six cases), we had the opportunity to compare for the same patients the sensitivity of bone marrow and peripheral blood leukaemic blasts; these were found to be virtually identical (data not shown).

Figure 1.

 Effect of Bortezomib on the induction of apoptosis of acute myeloid leukaemic (AML) blasts. Top and middle panels: dose-response experiments carried out incubating AML blasts in the presence of increasing concentrations of Bortezomib (B) alone (0, 0·01, 0·05 0·1, 0·25 μmol/l) or of increasing concentrations of Bortezomib as above and in the presence of a fixed concentration of recombinant human tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) (50 ng/ml). Bottom panel: proportion of apoptotic cells observed in 30 AMLs grown for 24 h either in the absence (C) or in the presence of Bortezomib (B, 0·25 μmol/l) or TRAIL (50 ng/ml) or Bortezomib + TRAIL at the above concentrations. The proportion of apoptotic cells was measured using the annexin V binding assay. The bars compared with the mean values ±SEM.

Figure 2.

 Flow cytometric profile of the annexin V binding assay. Flow cytometric profile of the annexin V binding assay carried out in two cases of acute myeloid leukaemia (AML), one exhibiting a ‘high sensitivity’ (A) and the other ‘low sensitivity’ (B) to Bortezomib-mediated apoptosis. AML blasts have been incubated for 24 h in the absence (control) or in the presence of Bortezomib (either 0·05, 0·1 or 0·25 μmol/l) or in the presence of tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) (50 ng/ml) or of Bortezomib + TRAIL at the above concentrations and then annexin V binding assay was carried out as outlined in Materials and methods.

Subsequent experiments used Bortezomib at the 0·25 μmol/l dose. At this dose, the addition of Bortezomib exhibited a pronounced induction of apoptosis with >50% of apoptotic cells in 18/30 AML cases (Fig 1). The mean proportion of apoptotic cells increased from 8·2 ± 1·6 observed in control cells, to 58·3 ± 4·5 observed in cells incubated with Bortezomib alone, to 71·1 ± 4·1 in cells incubated with Bortezomib + TRAIL. The proportion of apoptotic cells observed between control and Bortezomib-treated cells or between Bortezomib and Bortezomib + TRAIL was statistically significant (P ≤ 0·05). It is of interest to note that after treatment with TRAIL + Bortezomib only 2/26 AMLs displayed <50% apoptotic cells. In line with a previous report (Riccioni et al, 2005), AML blasts exhibited a very low sensitivity to TRAIL-mediated apoptosis (11·2 ± 1·9 apoptotic cells).

The addition of a pan-caspase inhibitor, such as z-VAD, to Bortezomib-treated cells greatly inhibited the apoptosis induced by the proteasome inhibitor. In six AML cases the proportion of apoptotic cells was 12 ± 3 in control, 52 ± 8 in Bortezomib-treated cells, 10 ± 4 in cells treated with z-VAD and 21 ± 5 in cells treated with Bortezomib + TRAIL. The proportion of apoptotic cells observed in Bortezomib-treated cells was significantly different from that observed for cells treated with z-VAD + Bortezomib (P ≤ 0·01).

The sensitivity of AML blasts correlate, with their immunophenotypic features

The AMLs were subdivided into two subgroups according to their sensitivity to Bortezomib [low sensitivity group (LSG) with <50% apoptotic cells; high sensitivity group (HSG) with >50% apoptotic cells]. The analysis of the immunophenotypic features of these subgroups showed that the sensitivity to Bortezomib-mediated apoptosis correlated with the differentiation properties of AML blasts, in that the majority of the HSG cases corresponded to the M4 (6/18) and M5 (6/18) FAB subtypes (see Table IA and IB); the cellularity of the two subgroups was similar (LSG, 84 ± 27 × 109 WBC/l; HSG, 87 ± 18 × 109 WBC/l), while the majority of the LSG corresponded to M1 (5/12) and M2 (5/12) FAB subtypes. The comparative analysis of the immunophenotypic features of AMLs displaying high and low sensitivity to Bortezomib showed several interesting findings (Fig 3):

Figure 3.

 Immunophenotypic features of acute myeloid leukaemia (AML). AMLs were subdivided according to their sensitivity to Bortezomib into a ‘low sensitivity’ (exhibited <50% of apoptotic cells) and a ‘high sensitivity’ (exhibited >50% apoptotic cells) group and their reactivity with a panel of monoclonal antibodies was investigated. The positivity of CD34, Ac133, CD11b, CD14, M-CSFR, c-met, VEGF-R2 and Tie-2 was reported as the percentage of positive cells. The positivity of Flt3 was reported as the mean fluorescence intensity expressed in arbitrary units.

1 The expression of stem cell/progenitor cell membrane markers, such as CD34 and Ac133, was markedly higher in the LSG compared with the HSG. The mean percentage of CD34+ cells was 60 ± 8·6 in the LSG and 21·5 ± 6 in the HSG (P ≤ 0·001); the mean proportion of Ac133+ cells was 51 ± 7·6 in the LSG and 10·2 ± 3·2 in the HSG (P ≤ 0·001).

2 The expression of membrane granulo-monocytic markers was markedly higher in the HSG than in the LSG: the mean percentage of CD11b+ cells was 15·2 ± 3 in the LSG and 48·5 in the HSG (P ≤ 0·001); the mean proportion of CD14+ cells was 4·8 ± 1·4 in the LSG and 34 ± 6 in the HSG (P ≤ 0·001); the mean percentage of CD36+ cells was 13·6 ± 4·5 in the LSG and 41 ± 7 in the HSG (P ≤ 0·01).

3 The expression of membrane receptors abundantly expressed on myelomonocytic elements, such as M-CSFR, Flt3, IGF-1R and c-met, was higher in the HSG compared with the LSG: the mean percentage of M-CSFR positive cells was 10·7 ± 2·3 in the LSG and 44·8 in the HSG (P ≤ 0·001); the proportion of IGF-1R positive cells was 16·5 ± 6·4 in the LSG and 52·9 ± 6·5 in the HSG (P ≤ 0·01); the percentage of c-met positive cells was 11·2 ± 2·8 in the LSG and 49·5 ± 1·6 in the HSG (P ≤ 0·001); the mean fluorescence intensity of Flt3 expression was 24 ± 2·3 in the LSG and 41·8 ± 6·5 in the HSG (P = 0·05).

4 The receptors for the angiogenic growth factors, VEGF-R2 and Tie-2, were more expressed in the HSG than in the LSG: the proportion of VEGF-R2 positive cells was 5·2 ± 2·5 in the LSG and 31·2 ± 6 in the HSG (P ≤ 0·01); the percentage of Tie-2 positive cells was 5 ± 1·3 in the LSG and 39·4 ± 7·2 in the HSG (P ≤ 0·01).

Interestingly, the analysis of the incidence of FLT3 mutations [seven internal tandem duplications (ITD) and two D835 point mutations in a total of 30 patients] showed that all these mutations were found in the HSG and none in the LSG.

Recent studies have shown that the level of chemokine (C-X-C motif) receptor 4 (CXCR4) expression in AML cells represents an important prognostic factor, as AML cases associated with high CXCR4 expression have a poor prognosis (Dommange et al, 2006; Konoplev et al, 2007; Spon et al, 2007). Therefore, it was of interest to compare CXCR4 expression in the two subgroups. The mean percentage of CXCR4+ cells was 27·2 ± 7·2 in the LSG, compared with 48·7 ± 8·6 in the HSG (P = 0·05). It is of interest to note that in the HSG, 11/18 AML showed >50% CXCR4+ cells, while in the LSG 2/12 AML displayed >50% CXCR4+ cells (Fig 4).

Figure 4.

 CXCR4 levels. Acute myeloid leukaemias have been subdivided and their reactivity with a monoclonal antibody anti-CXCR4 was determined. The positivity of CXCR4 was reported as the percentage of positive cells.

Effect of Bortezomib on TRAIL-R1/TRAIL-R2 expression

The proapoptotic effects of TRAIL are mediated through its interaction with TRAIL-R1 and/or TRAIL-R2. Therefore, the expression of these two TRAIL-Rs was evaluated in AMLs. The expression of TRAIL-R1 was much higher in the HSG than in the LSG: the proportion of TRAIL-R1 positive cells was 8 ± 4·2 in the LSG and 32·8 ± 6 in the HSG; furthermore, in the HSG 12/18 AMLs were clearly TRAIL-R1+, while in the LSG only 2 of 12 AMLs were TRAIL-R1+ (Table I). Furthermore, TRAIL-R1 and TRAIL-R2 expression were measured after Bortezomib exposure, but there was no statistically significant difference when compared with untreated control cells (data not shown).

Bortezomib activates caspase-8 and caspase-3 and decreases c-FLIP levels in AML blasts

The analysis of caspase-8 and caspase-3 by Western blotting provided evidence about the marked activation of both these caspases by Bortezomib in the HSG, as shown by: (i) a marked decrease of procaspase-8 levels in Bortezomib-treated cells compared with controls; (ii) a decrease of procaspase-3 levels, associated with the appearance of a 12 kDa fragment, corresponding to active caspase-3 (Fig 5, panels A and B). It is of interest to note that in these AMLs the addition of TRAIL to Bortezomib in some AMLs potentiated the extent of caspase-8 and caspase-3 activation, compared with the effect elicited by Bortezomib alone (Fig 5, panels A and B). In addition to caspase-8 and caspase-3 activation, Bortezomib induced a marked decrease of c-FLIP levels (Fig 5, panels A and B). In contrast, in the HSG, the addition of Bortezomib elicited only a weak caspase-8 and caspase-3 activation and only a moderate decrease of c-FLIP (Fig 5, panels C and D).

Figure 5.

 Western blot analysis of caspases, Fas-associated death domain (FADD) and FADD-like interleukin-1β-converting enzyme-inhibitory protein (c-FLIP) in acute myeloid leukaemia (AML). Western blot analysis of caspase-8, caspase-3, FADD, cFLIP and β-actin in cell extracts derived from four AMLs grown for 24 h either in the absence (C) or in the presence of 0·25 μmol/l Bortezomib (B) or 50 ng/ml TRAIL (T) or both these reagents (B + T) at the above concentrations. For caspase-3 the full fragment (FF) corresponding to procaspase-3 and the cleaved fragment (CF) corresponding to active caspase-3 are shown. Panels A and B are related to two AMLs from the high sensitivity group; C and D to two AMLs from the low sensitivity group.

High c-FLIP levels in AML blasts were associated with low Bortezomib sensitivity

Bortezomib can act at multiple levels of the apoptotic machinery. In particular, previous studies have shown that c-FLIP could be a major target of Bortezomib, as c-FLIP levels are decreased by this proteasome inhibitor (Sayers et al, 2003; Kabore et al, 2006; Conticello et al, 2007; Saulle et al, 2007). It seemed therefore of interest to evaluate the expression of the various proteins involved in caspase-8 activation, such as c-FLIP and FADD, and of caspse-8 itself in AML blasts and to correlate it with the extent of apoptotic response to Bortezomib. The analysis of c-FLIP levels showed that this protein was heterogeneously expressed in various cases of AML (Fig 6A) and, importantly, its level of expression was clearly and significantly higher in the LSG than in the HSG (Fig 6B, P ≤ 0·001). Caspase-8 levels were very heterogeneous in various AML samples: particularly, some samples exhibited very low levels of caspase-8 (Fig 6A). The analysis of caspases-8 levels among the two AML groups subdivided according to Bortezomib sensitivity showed that the LSG displayed moderately higher caspases-8 levels that the HSG (Fig 6B, P = 0·05). Finally, FADD, caspase-3 and XIAP levels were also heterogeneous in the various AMLs, but their extent of expression was comparable in the LSG and in the HSG (Fig 6A and B).

Figure 6.

 Western blot analysis of caspases, Fas-associated death domain (FADD) and FADD-like interleukin-1β-converting enzyme-inhibitory protein (c-FLIP) in acute myeloid leukaemias (AML). Western blot analysis of caspase-8, caspase-3, FADD, cFLIP, XIAP in cell extracts derived from 24 representative AMLs. (B) Caspase-8, caspase-3, FADD, cFLIP and XIAP levels evaluated through densitometric analysis of Western blots and expressed as ratio with respect to β-actin (used to normalize for protein loading).

Bortezomib kills leukaemic stem cells

Recent studies have provided clear evidence that the majority of AMLs are initiated by the leukaemic transformation of cells that acquire the properties of leukaemic stem cells (Bonnet & Dick, 1997). The complete eradication of these cells is required to obtain a successful and durable therapy of AMLs (Bonnet & Dick, 1997). These cells are identified according to their peculiar membrane phenotype and functionally characterized through in vivo assays in immunodeficient mice (Bonnet & Dick, 1997). However, more recent studies have shown a more simple way to characterize these cells based on their capacity to express high levels of the enzyme ALDH (Pearce et al, 2005). The use of a ALDH fluorescent substrate enabled ALDH++ cells to be easily identified and isolated (Hess et al, 2004); these cells were also identified in AMLs (Hess et al, 2004).

It seemed therefore of interest to evaluate a possible effect of Bortezomib on leukaemic ALDH++ cells. To this purpose, four AML cases were selected that exhibited the ‘diffuse reactivity pattern’, according to Pearce et al (2005). In the ‘diffuse reactivity pattern’ the ALDH++ cells possessed a higher side scatter than normal stem/progenitor cells. The number of ALDH++ cells greatly varied in each of these four AMLs, but in all cases a very pronounced decrease in the number of these cells was observed after 24 h of incubation in the presence of 0·25 μmol/l Bortezomib and Bortezomib + TRAIL (from 5 ± 1·6% ALDH++ cells in the control to 0·5 ± 0·25% and 0·86 ± 0·7% ALDH++ cells in Bortezomib and Bortezomib + TRAIL-treated cells respectively) (see Fig 7A). The statistical analysis of these data showed that: control versus Bortezomib or versus Bortezomib + TRAIL P ≤ 0·001; Bortezomib versus Bortezomib + TRAIL P ≤ 0·003; control versus TRAIL P = 0·08. It was of interest to note that the ALDH++ cells present in untreated cells exhibited a wide side scatter, while cells that survived Bortezomib-mediated cell killing showed a scatter comparable to that observed for normal haemopoietic progenitors/stem cells (Fig 7B). Double labelling experiments showed that the large majority of ALDH++ cells were CD34+ (i.e. >90%) (data not shown).

Figure 7.

 Effect of Bortezomib on aldheyde dehydrogenase (ALDH) positive cells in acute myeloid leukaemia (AML). (A) Percentage of apoptotic cells (left panels) and ALDH++ cells (right panels) in three different AMLs grown for 24 h either in the absence (C) or in the presence of 0·25 μmol/l Bortezomib or 50 ng/ml tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) or Bortezomib + TRAIL at the above concentrations. (B) Flow cytometric analysis of ALDH-positive cells in one representative AML. AML cells have been incubated for 24 h either in the absence (C) or in the presence of 0·25 μmol/l Bortezomib alone or together with 50 ng/ml TRAIL and then have been labelled with the Aldefluor reagent either in the absence (right panel) or in the presence of diethylaminobenzaldheyde (DAEB), an ALDH inhibitor (left panel).

Importantly, using mononuclear cells isolated from normal cord blood, exhibiting 0·36 ± 0·04% ALDH++ cells, it was shown that neither Bortezomib (0·33 ± 0·05%), nor Bortezomib + TRAIL (0·31 ± 0·45%) significantly modified the number of these cells (data not shown).


We have shown that Bortezomib (a known potent anti-proliferative and proapoptotic agent) is a strong inducer of apoptosis of AML cells. Particularly, Bortezomib exerted a marked proapoptotic effect in 18/30 AMLs; furthermore, the proapoptotic effect of Bortezomib was significantly potentiated by TRAIL, which, when added alone, did not induce apoptosis of AML blasts. Our study provides evidence that virtually all AMLs display a clear sensitivity to a proapoptotic effect of Bortezomib. However, 60% of AMLs displayed a high sensitivity to Bortezomib, while the remaining 40% exhibited only a moderate/low sensitivity to Bortezomib. The high sensitivity of AMLs to Bortezomib seemed to be related to: (i) the differentiation properties of AMLs, as the large majority of these AMLs were M4 and M5 FAB subtypes; (ii) the level of c-FLIP, as the majority of these AMLs displayed low expression of this anti-apoptotic protein. Conversely, the large majority of AMLs that were barely sensitive to Bortezomib belonged to the M1 and M2 FAB subtypes and displayed high c-FLIP expression.

The c-FLIP acts as a natural inhibitor of caspase-8 activation. c-FLIP levels are controlled mainly through post-translational mechanisms (Kim et al, 2002). Although c-FLIP protein seems to be degraded into the cells through the ubiquitin-proteasome pathway (Fukazawa et al, 2001), studies carried out in several cancer cell types have provided evidence that Bortezomib induced a decrease of c-FLIP levels (Sayers et al, 2003; Kabore et al, 2006; Conticello et al, 2007; Saulle et al, 2007). Thus, studies performed in two murine leukaemia cell lines showed that Bortezomib reduced c-FLIP levels and, through this mechanism, sensitized leukaemia cells to TRAIL-mediated apoptosis (Sayers et al, 2003). Induction of apoptosis by Bortezomib in primary CLL cells was associated with downregulation of c-FLIP expression (Kabore et al, 2006). Furthermore, there was evidence that the decrease of c-FLIP expression induced by Bortezomib in ovarian cancer cells (Saulle et al, 2007) and in thyroid carcinoma cells (Conticello et al, 2007) could play a role in the mechanism of cell death elicited in these cells by the proteasome inhibitor. In line with these findings it was also observed that, in primary AML cells, Bortezomib decreased c-FLIP, a phenomenon associated with a clear induction of apoptosis. In addition to these observations, our results also indicated that c-FLIP levels could represent a major determinant of Bortezomib response of AMLs.

There are some interesting similarities between the effects induced by Bortezomib in primary CLL (Kabore et al, 2006) and AML cells (the present study). Bortezomib elicited variable levels of apoptosis and decreased c-FLIP, activated caspase-8 and caspase-3 in both cell types; however, in CLL cells, but not in AML cells, Bortezomib increased the expression of TRAIL, TRAIL-R1 and TRAIL-R2 (Kabore et al, 2006).

The sensitivity of AML blasts to Bortezomib was also related to their differentiation properties in that those AMLs that were highly sensitive to Bortezomib, mainly belonged to the M4 and M5 FAB subtypes. In line with this finding, HSG AMLs exhibited higher expression of CD11b, CD14, CD36, M-CSFR and of other membrane antigens preferentially expressed on myelomonocytic cells than LSG AMLs. This finding suggests that relapsing M4/M5 AMLs could represent the optimal candidates for future clinical trials based on the use of Bortezomib. Furthermore, the HSG also displayed a high expression of angiogenetic growth factor receptors (VEGF-R2 and Tie-2) on leukaemic blasts, a finding in line with a recent study showing the expression of these receptors on monocytic AMLs (Riccioni R et al, 2007).

It is also of interest to note that the majority of HSG AMLs exhibited high CXCR4 levels. This finding is not surprising in that the majority of HSG AMLs were classified as M4 and M5 FAB subtypes. In fact, previous studies have shown that CXCR4 expression was low in M0/M1 AMLs, but high in M4/M5 AMLs (Mohle et al, 2000). As before mentioned, several recent studies have shown that CXCR4 overexpression in AML was associated with a poor prognosis (Dommange et al, 2006; Konoplev et al, 2007; Spon et al, 2007). It seemed therefore of interest to use Bortezomib therapy for this type of AML, associated with a poor prognosis.

Particularly interesting was the observation the all the AMLs displaying FLT3 mutations, and particularly FLT3/ITD+ AMLs, were greatly sensitive to the proapoptotic effects of Bortezomib. Previous studies have shown that FLT3/ITD+ AMLs are associated with a poor prognosis (Kottaridis et al, 2001; Meshinchi et al, 2001). This observation suggests that Bortezomib could be used together with FLT3 inhibitors and other anti-leukaemic drugs for the treatment of these AMLs associated with a poor outcome.

Previous studies have shown that an efficacious and durable anti-leukaemic therapy must necessarily involve the eradication of leukaemic stem cells, the cells that initiate and maintain the leukaemic process (Bonnet & Dick, 1997). It seemed therefore of interest to assess the effect of Bortezomib on leukaemic stem cells. A simple method to identify and quantify leukaemic stem cells is based on the cytofluorimetric detection of ALDH. Particularly, in AMLs that exhibit a ‘numerous pattern’ of ALDH reactivity, ALDH++ cells correspond to leukaemic stem cells (Pearce et al, 2005). The study of three of these AMLs provided clear evidence that Bortezomib could induce the killing of the majority of leukaemic stem cells. Interestingly, the ALDH++ cells that survived Bortezomib treatment were, according to their number and scatter characteristics, comparable to those observed in normal haemopoietic tissue (bone marrow and peripheral blood), thus suggesting that they represent normal stem cells. Importantly, Bortezomib did not induce the killing of normal cord blood ALDH++ cells, corresponding to normal progenitor/stem cells.

In conclusion, the findings of the present study further support the clinical use of Bortezomib in the treatment of AML, and showed that the sensitivity of AML blasts to the proteasome inhibitor was inversely related to c-FLIP levels. Importantly, FLT3/ITD+ AMLs displayed high sensitivity to Bortezomib, suggesting a possible use of this proteasome inhibitor together with FLT3 inhibitors in future clinical studies. Therefore, our results strongly suggest that proteasome inhibition should be considered in AML therapy.