The HSP90 inhibitor NVP-AUY922-AG inhibits the PI3K and IKK signalling pathways and synergizes with cytarabine in acute myeloid leukaemia cells


  • Elisabeth J. Walsby,

    Corresponding author
    • Cardiff Experimental Cancer Medicine Centre, Institute of Cancer and Genetics, Institute of Cancer and Genetics, School of Medicine, Cardiff University, ardiff, UK
    Search for more papers by this author
  • Michelle Lazenby,

    1. Cardiff Experimental Cancer Medicine Centre, Institute of Cancer and Genetics, Institute of Cancer and Genetics, School of Medicine, Cardiff University, ardiff, UK
    Search for more papers by this author
  • Chris J. Pepper,

    1. Department of Haematology, Institute of Cancer and Genetics, School of Medicine, Cardiff University, Cardiff, UK
    Search for more papers by this author
    • These authors contributed equally.
  • Steven Knapper,

    1. Cardiff Experimental Cancer Medicine Centre, Institute of Cancer and Genetics, Institute of Cancer and Genetics, School of Medicine, Cardiff University, ardiff, UK
    Search for more papers by this author
    • These authors contributed equally.
  • Alan K. Burnett

    1. Department of Haematology, Institute of Cancer and Genetics, School of Medicine, Cardiff University, Cardiff, UK
    Search for more papers by this author

Correspondence: Dr. Elisabeth J. Walsby, Institute of Cancer and Genetics, Cardiff University, Cancer and Genetics Building, Heath Park, Cardiff CF14 4XN, UK



Heat shock protein 90 (HSP90; HSP90AA1) is a molecular chaperone involved in signalling pathways for cell proliferation, survival, and cellular adaptation. Inhibitors of HSP90 are being examined as anti-cancer agents, but the critical molecular mechanism(s) of their activity remains unresolved. HSP90 inhibition potentially facilitates the simultaneous targeting of multiple molecules within tumour cells and represents an attractive therapeutic proposition. Here, we investigated HSP90 as a molecular target for acute myeloid leukaemia (AML) using the novel HSP90 inhibitor NVP-AUY922-AG. NVP-AUY922-AG induced dose-dependent killing in myeloid cell lines and primary AML blasts. In primary blasts, cell death in response to NVP-AUY922-AG was seen at concentrations almost 2 logs lower than cytarabine (Ara-C) (50% lethal dose = 0·12 μ mol/l ± 0·28). NVP-AUY922-AG was significantly less toxic to normal bone marrow (P = 0·02). In vitro response to NVP-AUY922-AG did not correlate with response to Ara-C (r2 = 0·0006). NVP-AUY922-AG was highly synergistic with Ara-C in cell lines and in 20/25 of the primary samples tested. NVP-AUY922-AG induced increases in HSP70 expression and depletion of total AKT, IKKα and IKKβ in cell lines and primary blasts. This study shows that the novel HSP90 inhibitor NVP-AUY922-AG has significant single agent activity in AML cells and is synergistic with Ara-C.


Heat shock protein 90 (HSP90; HSP90AA1) is a molecular chaperone that acts as part of a complex to maintain the conformational stability, shape and function of a range of important proteins required for normal cell homeostasis (Workman, 2003; Young et al, 2003). HSP90 expression is induced by heat shock and other cellular stresses (Workman, 2003) and enhances cell survival in tissues under stress, including conditions such as the nutrient-deprived conditions commonly found within tumours (Whitesell et al, 2003). HSP90α and β are found in the cytosol and nucleus while other members of the HSP90 family are found in the endoplasmic reticulum and the mitochondria (Chiosis et al, 2003; Banerji, 2009). Binding of over 100 client proteins by HSP90 ensures the correct folding and conformation, and also controls activation/deactivation or ubiquitination of the protein resulting in proteasomal degradation (Whitesell et al, 2003; Workman, 2003; Mosser & Morimoto, 2004; Smith & Workman, 2007).

There has been no evidence of mutation or aberrant amplification of the HSP90 family, suggesting that these proteins are not oncogenes in the conventional sense. HSP90 is relatively over-expressed, however, in many cancers, both solid and haematological, when compared to expression in normal tissues (Whitesell et al, 2003; Mosser & Morimoto, 2004). Expression of HSP90 is required for the stability and function of a large number of oncogenes that are themselves over-expressed, mutated or otherwise activated during malignant progression (Workman, 2003). HSP90 inhibitors therefore offer the potential to target multiple pathways; a number of which are frequently mutated in human cancers (Druker, 2002; Workman, 2003; Smith & Workman, 2007; Banerji, 2009).

The last decade has seen a move towards the development of molecular targeted therapies; the introduction of Gleevec, which specifically targets the BCR-ABL1 protein in chronic myeloid leukaemia, has shown that this approach can be successful (Druker, 2002). Given that most cancers do not express a unique fusion gene product that can be targeted, considerable effort has gone into defining potential molecular drug targets in cancer cells. HSP90 is over-expressed in acute myeloid leukaemia (AML) cells compared to normal cells (Yufu et al, 1992) and plays a role in cell survival and resistance to chemotherapy (Flandrin et al, 2008). Additionally, in cancer cells, detectable HSP90 is mostly in an activated form in the chaperone complex whilst only a small amount of the detectable HSP90 is activated in non-cancerous cells (Flandrin et al, 2008). This results in increased client protein binding in cancer cells and suggests that these cells are more dependent on HSP90 and potentially more susceptible to HSP90 inhibition strategies (Hertlein et al, 2010).

Most known HSP90 inhibitors block the chaperone cycle by competition with ATP binding at the unique N terminal nucleotide binding site that is not present on other HSP family members (Workman, 2003). Occupation of this site by a drug, such as geldanamycin, prevents dissociation of the proteins from the chaperone complex and consequently the trapped proteins do not achieve their mature functional conformation and are degraded by the proteasome (Chiosis et al, 2003). Radicicol was one of the first HSP90 inhibitors identified and was cytotoxic to HeLa cells (Uehara, 2003). 17-allylamino-17-desmethoxy-geldanamycin (17-AAG), a geldanamycin derivative, was the first HSP90 inhibitor to enter clinical trials and showed encouraging results in difficult-to-treat patient populations (Taldone et al, 2008). Second generation derivatives of 17-AAG have been developed to increase solubility and are showing activity in cancer cells (Hertlein et al, 2010; Lancet et al, 2010).

Synthetic HSP90 inhibitors are also in development and entering clinical trials, including CNF2024/BIIB021 (Lundgren et al, 2009) and the isoxazole resorcinol NVP-AUY922-AG/VER-52296. NVP-AUY922-AG binds to the N-terminal nucleotide binding pocket of HSP90 with the highest affinity of any synthetic small molecule HSP90 inhibitor yet described (Eccles et al, 2008). This present study was designed to investigate the in vitro activity and mechanism(s) of action of this HSP90 inhibitor in AML both alone and in combination with conventional chemotherapy.

Materials and methods

Cell culture

Bone marrow samples were collected in ethylenediaminetetraacetic acid (EDTA) from newly diagnosed AML patients entering the National Cancer Research Institute (NCRI) AML15, 16 and 17 trials, with the patients' informed consent using documentation approved by the Wales Multicentre Research Ethics Committee. The patient characteristics of the samples used in this study are shown in Table 1. Primary cells were enriched by density gradient centrifugation with Histopaque (Sigma, Poole, UK) and were maintained in Roswell Park Memorial Institute (RPMI) medium supplemented with 10% fetal bovine serum (FBS). All cultures were maintained at 37°C in a 5% CO2 humidified atmosphere. Normal samples from pre-annonymized non-leukaemic bone marrow donors were prepared in the same manner as the leukaemic samples. Cell viability was measured on a Vi-Cell XR (Beckman Coulter, High Wycombe, UK) and expansion of cultures was monitored by cell counts on the same machine. AML cell lines (HL-60, NB4) were maintained between 2 × 105 and 1 × 106 cells/ml in RPMI medium supplemented with 10% FBS.

Table 1. Patient Characteristics
patientAge (years)sexWBC (× 100/l)FAB groupKaryotypeDe novo/secondaryFLT3-ITD mutation statusFLT3-TKD mutation statusNPM1 mutation statusKIT mutation status
  1. FAB, French-American-British classification; ITD, internal tandem duplication; TKD, tyrosine kinase duplication.

162M82·2446,XY[20]de novowtwt
253M6·2246,XY[30]de novowtwtmutant
355M16·1146,XY[20]de novowtwtwtwt
443M2·2188,XXYY,−4,−13,−17,−19[4]/46,XY[21]de novoITDwtmutant
526F84·4247,XX,+?4[10].nuc ish 8q22(ETOx2),21q22(AML1 × 2)[100]de novoITD
647M5·7246,Y,add(X)(q23)[27]/47,idem,+8[7]de novoITDwtmutant
759F124446,XX,inv(16)(p13q22)[21]/47,idem,+22[4]de novoITDwtmutant
863M68·24de novowtwtwt
955M20·2145,X,Y,inv(16)(p13q22)[20]de novowt
1056M30445,X,-Y[20]de novoITDwtmutant
1144F38·8546,XX[20]de novowtwtmutant
1267F96·44e46,XX[20]de novoITDwtmutantwt
1342F21·6246,XX[20]de novowtwtmutantwt
1448M36·7046,XY[20]de novoITDwtmutantwt
1547F152146,XX[20]de novoITDwtAwt
1643M69·4446,XY[30]de novowtwtmutant
1748F2·6446,XX[20]de novo
1839F168·7146,XX[20]de novoITDwt
1922F66·6446,XX,inv(16)(p13q22)[9]/46,XX[1]de novowtwtwt
2058M50·6846,XY[20]de novowtwtmutant
2145M106·1446,XY[20]de novoITDwtwt
2256M25·8246,XY[20]de novowtmutantmutant
2358M21·7X46,XY[20]de novoITDwt
2480F45·7046,XX[50]de novowtmutantmutant
2844F78·8446,XX[20]de novoITDwtmutant
2962M285·85b46,XY[6]de novoITDwtmutant
3040F174·7246,XX[20]de novoITDmutantmutant
3160F68·5546,XX[20]de novoITDwtmutant
3222F24·1846,XX[20]de novowtmutantmutant
3362F49·7246,XX[20]de novowtmutantmutant
3445F7·5846,XX,t(3;5)(q25;q34)[3]/47,idem,+8[7]de novoITDwtwt
3545M38·2846,XY[20]de novoITDwtmutant
3645M38·2846,XY[20]de novoITDwtmutant
3724F27·9547,XX,+11[10]de novoITDwtwt
3865M84·7446,XY[20]de novo wtwt
3952M21·7846,XY[20]de novoITDwtmutant
4059M59·7846,XY[20]de novoITDwtwt
4126F18·2246,XX[21]de novoITDwtmutant
4235M33·7546,XY [30]secondarywtwtwt
4362F70·3846,XX[20]de novoITDwtmutant
4459M96·1046,Y,t(X;21)(p21;q22)[14] 46,idem,+1,der(1;7)(q10;p10)[7] 46,XY[4]de novowtwtwt
4524M11·8446,XY[25]de novoITDwtmutant

MTS assays

In vitro toxicity assays were performed on primary material over a 48-h period using a 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, (MTS) cell proliferation assay (Promega UK Ltd, Southampton, UK). Primary AML cells with a purity of over 80% and an untreated viability of >80% were analysed. Fifty percent lethal dose (LD50) values (the concentration of drug required to kill 50% of the cells) were calculated from the sigmoidal dose-response curves. Cells (1 × 105/well) were treated with NVP-AUY922-AG (0·008–1 μ mol/l) or cytarabine (Ara-C; 0·06–8 μ mol/l) by serial dilution and incubated for 48 h in a final volume of 90 μl. Following the culture period, 20 μl of MTS reagent was added and incubated for a further 4 h. The absorbance of each reaction after this time was read by spectrophotometry at 490 nm and the percentage of viable cells calculated relative to untreated control cells in the same assay. LD50 values were calculated using calcusyn software (Biosoft, Cambridge, UK).

Synergy between NVP-AUY922-AG and Ara-C was assessed in cell lines and primary AML samples (n = 25), using an experimentally determined fixed molar ratio of NVP-AUY922-AG with Ara-C (1:100). In order to establish the most effective concentration range for each agent, dose-response curves were generated and LD50 values were calculated using non-linear regression analysis. In subsequent experiments, cells were treated with serial dilutions of each drug individually and with both drugs simultaneously in the fixed molar ratio. It is worthy of note that exploration of the possible ratiometric combinations was constrained by the fact that we only considered clinically relevant doses of Ara-C in these experiments. calcusyn software was used to determine whether synergy existed between the agents using the Chou and Talalay method, based on the median effect method and which generates an combination index (CI) where a CI <1 indicates synergy, CI = 1 indicates additive interactions between the two drugs and CI >1 shows antagonism. The method is derived from the mass-action law principle that provides the common link between a single entity and multiple entities and first and higher order dynamics (Chou & Talalay, 1984) (Chou, 2010).

Apoptosis: Annexin V and Caspase-3 activation

AML cell lines were treated with NVP-AUY922-AG at concentrations of 0·05, 0·1 and 0·2 μ mol/l, and incubated for 48 h. Primary AML cells were incubated for 48 h with NVP-AUY922-AG at concentrations of 0·05–0·4 μ mol/l. Annexin V positivity was measured using the Annexin V Apoptosis Detection Kit (Axxora (UK) Ltd, Nottingham, UK) according to the manufacturer's instructions. Briefly, cells were washed in phosphate-buffered saline (PBS) and resuspended in the supplied binding buffer containing calcium chloride and incubated with fluorescein-labelled Annexin V in the dark for 10 min. Untreated samples were also prepared. Cells were rewashed with PBS and resuspended in binding buffer with 1 μg/ml propidium iodide (PI). Data on the Annexin V and PI positivity and forward and side scatter of the cells was collected on an Accuri C6 flow cytometer (Accuri Cytometers (Europe) Ltd, St Ives, UK). LD50 values were calculated using calcusyn software. All experiments were performed in triplicate.

AML cell lines were incubated in the presence of NVP-AUY922-AG at concentrations of 0·05, 0·1 and 0·2 μ mol/l, for 48 h. Cells were harvested by centrifugation and washed in PBS and incubated for 1 h at 37°C in the presence of PhiPhiLux G1D2 substrate (Calbiochem, Nottingham, UK). The substrate contains two fluorophores separated by a quenching linker sequence that was cleaved by active caspase-3. Once cleaved, the resulting products fluoresced green and were quantified using an Accuri C6 flow cytometer.


AML cell lines treated with 0, 0·01, 0·02 and 0·04 μ mol/l NVP-AUY922-AG for 4 h or 24 h, were washed 3 times in ice-cold PBS then lysed by resuspension in lysis buffer [20 m mol/l Tris (pH 7·4), 150 m mol/l NaCl, 1% Igepal (Sigma Aldrich, Poole, UK), 10% glycerol, 10 m mol/l EDTA, 20 m mol/l NaF, and 3 m mol/l NaVO4] plus complete protease inhibitors (MiniComplete EDTA-free; Roche, Burgess Hill, UK) for 30 min at 4°C followed by centrifugation at 16 000xg. The clarified protein lysates were quantified and equal aliquots subjected to electrophoresis using NuPage precast 4–12% Bis-Tris gels (Invitrogen, Paisley, UK) followed by transfer to nitrocellulose membranes (Invitrogen). Immunoblotting was performed with antibodies to HSP90, AKT, IKKα and IKKβ (Cell Signalling Technology Inc, New England Biolabs, Hitchin, UK), HSP70 (Millipore (UK) Ltd, Watford, UK) and Tubulin (Abcam, Cambridge, UK). Blots were visualized by chemiluminescence (ECL Advance; Amersham Biosciences, Piscataway, NJ), with the resultant images being analyzed using a LAS-3000 dark box (Fujifilm UK Ltd, Bedford, UK). Primary AML cells were also treated with the HSP90 inhibitors for 24 h and immunoblotted in the same manner as described.


NVP-AUY922-AG induces apoptosis in primary AML blasts and is more potent than Ara-C

Drug-induced cytotoxicity was assessed by MTS assay in AML cell lines NB4 and HL-60 and 45 primary AML samples exposed to NVP-AUY922-AG (0–1 μ mol/l) and Ara-C (0–8 μ mol/l) for 48 h. NVP-AUY922-AG was significantly more potent than Ara-C in both the cell lines and primary AML cells (P < 0·0001; Fig 1A). Indeed, NVP-AUY922-AG demonstrated LD50 values at least 59 times lower than Ara-C in both cell lines and primary AML blasts. The mean LD50 values for the cell lines, primary AML blasts and normal bone marrow are shown in Table 2. There was no correlation between response to the HSP90 inhibitor and the age or sex of the patients (P = 0·56 and 0·19 respectively and Figure S1A, B). Neither was there a significant difference in sensitivity to NVP-AUY922-AG between the French-American-British (FAB) classification sub-groups (P = 0·18, Figure S1C) or between samples from de novo and secondary AML patients (P = 0·47, Figure S1D). Patient FLT3 mutation [internal tandem duplication (ITD) or tyrosine kinase duplication (TKD)] or with an NPM1 mutation did not result in any statistically significant difference in response to NVP-AUY922-AG (P = 0·32, P = 0·78, P = 0·14, Figure S1E). As the effect of NVP-AUY922-AG on these samples was measured by MTS assay, this is consistent with previous reports that the anti-proliferative effects of HSP90 inhibition are not dependent on FLT3 mutational status (Reikvam et al, 2012). However there was correlation between presenting white blood cell (WBC) count and response to NVP-AUY922-AG (r2 = 0·13, P = 0·018, Figure S1F), where greater responses were seen in cases with a high WBC count. Four of the seven non-responding patients, with an LD50 of more than 0·2 μ mol/l, had low WBC counts. The non-responders were aged between 39 and 65 years and four of the five samples were FAB group 1, two were FAB group 2 and one FAB group 4. In the non-responding patients, FLT3 or NPM1 mutation status was not consistent between the samples. Importantly, non-leukaemic primary bone marrow was less susceptible to the effects of NVP-AUY922-AG as evidenced by significantly higher LD50 values when compared to the primary AML bone marrow blasts, although the therapeutic window was quite narrow (P = 0·02, Figure S1G).

Table 2. LD50 results
 NVP-AUY922-AG mean LD50 ± SD (μ mol/l)Ara-C mean LD50 ± SD (μ mol/l)

AML cell lines

NB4 and HL-60

0·009 ± 0·00030·42 ± 0·28

Primary AML samples

(n = 45)

0·12 ± 0·28

median 0·03

6·98 ± 7·24

median 4·42

Non-leukaemic primary bone marrow samples

(n = 4)

0·17 ± 0·04
Figure 1.

(A) The HSP90 inhibitor NVP-AUY922-AG is significantly more potent than cytarabine (Ara-C) in primary acute myeloid leukaemia (AML) samples (n = 45) following 48 h exposure to the agents. Apoptosis, as identified by annexin V and propidium iodide positivity, was measured by flow cytometry: (B) (i)untreated primary AML cells, (B) (ii) primary AML cells treated with 0·1 μ mol/l NVP-AUY922-AG, Biii) dose response of myeloid cell lines and (C) primary AML cells in vitro (n = 6). The proportion of apoptotic cells was significantly increased in a dose-dependent manner after 48 h treatment with NVP-AUY922-AG. (D) Apoptosis induction was associated with caspase-3 activation, also measured by flow cytometry, following treatment with NVP-AUY922-AG over the same timeframe. Inhibitor concentrations in all experiments are shown in μ mol/l and cell line experiments are the mean of three independent experiments. (E) Responses to Ara-C and NVP-AUY922-AG do not correlate (r2 = 0·0006, P = 0·88) suggesting that their primary mechanisms of action are distinct.

We subsequently investigated the mechanism by which NVP-AUY922-AG induced cell death. Cells were treated for 48 h prior to annexin V and PI staining to show whether the drugs caused apoptosis. In the NB4 and HL-60 AML cell lines, concentrations of NVP-AUY922-AG >0·05 μ mol/l resulted in a significant increase in annexin V/PI positivity (P ≤ 0·018; Fig 1B). Apoptosis was also induced in a dose-dependent manner in primary AML samples (n = 6) treated for 48 h with NVP-AUY922-AG (P = 0·001; Fig 1C). Furthermore, apoptosis induction was accompanied by an increase in caspase-3 activation in response to treatment with NVP-AUY922-AG (P = 0·04; Fig 1D). Importantly, the in vitro responses to NVP-AUY922-AG did not correlate with sensitivity to Ara-C in the primary AML samples (r2 = 0·0006, P = 0·88; Fig 1E), suggesting that the mechanisms of action of these agents are distinct. This raised the possibility that they may synergize when used in combination with one another.

NVP-AUY922-AG synergizes with Ara-C

We next combined NVP-AUY922-AG with Ara-C to determine whether these drugs could act synergistically with each other. The combination ratio used was determined experimentally and was based on the relative LD50 values of each drug. The combination of Ara-C with NVP-AUY922-AG at a fixed molar ratio of 100:1 proved to be optimal under these experimental conditions; only clinically-relevant concentrations of Ara-C were considered, so this set the limits of the assay once the fixed molar ratio was established. Combination indices (CI) were calculated where a value less than 1 indicates synergy and greater than 1 is indicative of antagonism between the drugs and are shown in Fig 2A (cell lines) and Fig 2B (primary AML samples). NVP-AUY922-AG and Ara-C showed synergistic interactions in both of the cell lines (mean CI = 0·70 ± 0·25). The combination of NVP-AUY922-AG and Ara-C was also synergistic in primary AML blasts with a mean CI of 0·85 ± 0·39 for all 25 samples tested. However, there was inter-patient variability in response, with 20/25 (80%) of the samples tested showing marked synergy (mean CI = 0·78 ± 0·39). Five samples appeared to show some degree of antagonism to the combination of NVP-AUY922-AG and Ara-C. There was no identifiable biological factor that was consistent between these samples that could account for this lack of synergy. The samples were both male and female, had an age range of 22–59 years, WBC count between 16 and 124 × 109/l, a range of FAB and World Health Organization (WHO) subtypes of AML were represented, two had normal karyotype, three were more complex, three were wild-type for FLT3 and NPM1 mutations, one had a FLT3-ITD mutation and one had both FLT3-ITD and NPM1 mutations. The response of the samples to NVP-AUY922-AG or Ara-C as single agents showed that two samples had a very low LD50 for NVP-AUY922-AG indicating remarkable sensitivity to this agent, two had very low LD50 for Ara-C and one had a very high LD50 for Ara-C. It seems most likely that, in these samples, the optimal fixed molar ratio for these samples may not have been used. Importantly, where synergy was evident, the synergistic interactions in primary AML samples were demonstrable over a wide range of concentrations (Fig 2C).

Figure 2.

(A) Cytarabine (Ara-C) and NVP-AUY922-AG showed synergistic interactions when combined at the experimentally-determined fixed molar ratio of 100:1. (B) 20/25 Primary AML samples showed in vitro synergy between Ara-C and NVP-AUY922-AG. The five samples that failed to demonstrate synergy showed marked sensitivity to one or other of the drugs used (C) Fractional effect plots of the interaction of Ara-C and NVP-AUY922-AG (100:1) show that the synergistic interactions between Ara-C and the HSP90 inhibitor is evident over a wide range of concentrations.

NVP-AUY922-AG inhibits the PI3K and NF-κB signalling pathways

In order to investigate the molecular basis of the response to NVP-AUY922-AG, the AML cell lines NB4 and HL-60 were treated with the drug at concentrations of 0·01, 0·02 and 0·04 μ mol/l and immunoblotting was then performed to determine the effect of this agent on known HSP90 client proteins (Fig 3). The median LD50 for primary AML samples for NVP-AUY922-AG was 0·03 μ mol/l (Table 2). Doses of NVP-AUY922-AG that were used to show the molecular basis of response to this drug were close to this concentration, showing that the down-regulation of proteins was not merely the result of cell death but a response to the effect of the drug in the cells. After 4 h of treatment, total HSP90 protein levels were not altered but HSP70 expression was increased in a dose-dependent manner in the NB4 cells. At this timepoint, the levels of AKT protein were not altered by NVP-AUY922-AG except at the highest concentration used (Fig 3A). IKKα was decreased at 4 h in both cell lines while IKKβ was induced in the NB4 cells but reduced in the HL-60 cells at 4 h. However, after 24 h of treatment with NVP-AUY922-AG the cell lines showed markedly increased levels of HSP70 and decreased levels of detectable AKT, IKKα and IKKβ (Fig 3B). In the NB4 cell line, the level of HSP90 itself was unaffected at 24 h but the HL-60 cells showed decreased HSP90 expression. Protein extracts from primary AML cells were also immunoblotted following 24 h treatment with NVP-AUY922-AG (Fig 4) and, consistent with the results demonstrated in the cell lines, showed increased HSP70 detection and decreased AKT, P-MAPK, IKKα and IKKβ. We have previously shown that there is synergy between NVP-AUY922-AG and another nucleoside analogue, fludarabine, which is commonly used in chronic lymphocytic leukaemia (CLL). In CLL, the nucleoside analogue as a single agent also results in decreases in AKT, MAPK, IKKα and IKKβ and that greater reduction of expression of these proteins is seen when both the nucleoside analogue and NVP-AUY922-AG are used in combination (Walsby et al, 2012).

Figure 3.

Immunoblotting was used to show the effects of HSP90 inhibitor NVP-AUY922-AG in myeloid cell lines, NB4 and HL-60, after (A) 4 h and (B) 24 h of treatment. HSP90 expression itself was not altered whilst HSP70 expression was increased following treatment for 4 h and persisted at 24 h. Total AKT was significantly decreased at 4 h at the highest concentration used and decreased in a dose-dependent manner by 24 h. Expression of IKKα and IKKβ were significantly reduced in HL-60 cells at 4 h and 24 h in a dose-dependent manner. In contrast, IKKβ was transiently induced at 4 h but then reduced at 24 h in NB4 cells.

Figure 4.

Primary acute myeloid leukaemia samples treated with NVP-AUY922-AG for 24 h in in vitro cell culture showed a dose-dependent increase in HSP70 and decreased expression in total AKT, P-MAPK, IKKα and IKKβ. In contrast, HSP90 itself remained unaltered by treatment with the HSP90 inhibitor in line with the results presented in the cell lines.


Development of targeted agents for the treatment of leukaemia has been the focus of much research in recent years (Gilliland, 2002; Aggarwal et al, 2007). Notable successes including the BCR-ABL1 kinase inhibitor imatinib mesylate (Gleevec) (Druker et al, 1996, 2001) and all-trans retinoic acid (ATRA) in APL patients with the t(15;17) translocation (Tallman et al, 1997; Randolph, 2000a,b). A potential drawback of drugs aimed at specific components of complex signal transduction pathways is that there is often redundancy and cross-talk between pathways leading to activation of multiple downstream transcription factors. Furthermore, cancers are frequently driven by several oncogenic abnormalities and ‘hijack’ multiple pathways that sustain the malignant phenotype. Therefore, incomplete or transient responses are likely if a single molecular target is modulated pharmacologically. A potential solution to this problem is the targeting of molecules with multiple downstream effects, for example, transcription factors or chaperone molecules. HSP90 is one such multi-client chaperone molecule and inhibition of this protein results in depletion of proteins important in signal transduction, invasion, cell cycle regulation, angiogenesis, metastasis and immortalization (Banerji et al, 2003). Inhibition of HSP90 function may simultaneously reduce the functional level of multiple proteins involved in leukaemic cell signalling (Reikvam et al, 2009). As critical clients of HSP90 chaperone function may vary from one patient's tumour to the next (Whitesell et al, 2003), the ability to modulate multiple proteins with a single inhibitor may be beneficial.

We report here that the HSP90 inhibitor, NVP-AUY922-AG, induced cytotoxicity in myeloid cell lines and primary AML cells. NVP-AUY922-AG was more potent than Ara-C in these in vitro model systems and appeared to be significantly more effective than the geldanamycin derivatives 17-AAG and 17-DMAG, based on previously reported LD50 values (George et al, 2004, 2005; Yao et al, 2007; Fiskus et al, 2009). Furthermore, we demonstrated that the cytotoxic profile of NVP-AUY922-AG is distinct from Ara-C (Fig 1E) suggesting that their primary mechanisms of action are different. Given these profiles, we considered the possibility of combining NVP-AUY922-AG with Ara-C as a therapeutic strategy in AML. In support of this, synergistic interactions between HSP90 inhibitors and conventional chemotherapeutics have previously been reported; 17-AAG synergizes with Ara-C in primary AML isolates (Mesa et al, 2005), and HSP90 inhibitors have also been shown to combine synergisitcally with standard chemotherapy in breast cancer (Banerji et al, 2003) and Hodgkin lymphoma (Boll et al, 2009). Here, we showed that NVP-AUY922-AG in combination with Ara-C was synergistic in both the AML cell lines and 20/25 primary AML samples. One rationale for the synergy demonstrated between HSP90 inhibitors and other drugs is that HSP90 stabilizes its client proteins, many of which have been implicated in cancer cell survival and drug resistance. Therefore, selective depletion of these client proteins within the cancer cells results in growth arrest or apoptosis and potentially sensitizes the tumour to the action of other chemotherapeutic agents (Neckers, 2002; Taldone et al, 2008; Reikvam et al, 2009).

HSP90 chaperones multiple proteins that affect many intracellular signalling pathways (Pratt & Toft, 2003). In this study we showed that selective inhibition of HSP90 with NVP-AUY992-AG reduced HSP90 protein expression in some samples but promoted an increase in HSP70 protein expression when cells were treated with increasing concentrations of the HSP90 inhibitor in all samples. Previous studies have shown that HSP70 expression is induced following treatment with HSP90 inhibitors in ex vivo peripheral blood lymphocytes (Clarke et al, 2000) and in other cell types (Whitesell et al, 2003; Mesa et al, 2005; Waza et al, 2005). These studies showed variable HSP90 levels following treatment with an HSP90 inhibitor with some cell types showing induction of HSP90AA1 gene expression and others not (Clarke et al, 2000). Induction of expression of HSP70 has previously been shown to accompany HSP90 inhibition (Solit & Chiosis, 2008; Usmani et al, 2009) and may be a potential mechanism of resistance to HSP90 inhibitors as HSP70 is a cytoprotective protein capable of blocking both caspase- dependent and caspase-independent apoptosis, autophagic cell death and necrosis (Jego et al, 2010).

Phosphatidylinositol-3 kinase (PI(3)K) inhibits apoptosis in haematopoietic cells (Scheid et al, 1995) by signalling though the serine/threonine kinase AKT pathway to mediate its anti-apoptotic effects (Franke et al, 1997). AKT is bound by HSP90 and inactivation of HSP90 leads to inactivation of AKT and increased sensitivity to apoptotic stimuli (Sato et al, 2000; Mesa et al, 2005). Here we demonstrated that treatment with NVP-AUY922-AG resulted in a decrease in the level of total AKT detectable in AML cells. This is consistent with results previously described in which other HSP90 inhibitors result in AKT down regulation (Basso et al, 2002; Solit et al, 2003). The resulting down regulation of AKT has been identified as sensitizing the cells to other chemotherapeutic agents that result in induction of apoptosis and may contribute to the synergy that we observed with Ara-C (Solit et al, 2003).

Nuclear factor (NF)-κB has been reported to be constitutively active in a number of malignancies including AML and CLL and its elevated expression is correlated with anti-apoptotic effects (Ozes et al, 1999; Grandage et al, 2005; Hertlein et al, 2010). NF-κB is held in an inactive form in the cytoplasm by IkB. Degradation of IκB by IKK releases NF-κB, allowing it to translocate to the nucleus and initiate transcription (Ozes et al, 1999). NF-κB regulates a variety of anti-apoptotic proteins and oncogenes, including BCL2, XIAP, CFLAR (c-FLIP), and MCL1. These genes are important in initiating or enhancing cell survival, suggesting that targeting NF-κB through depletion of IKK may be an attractive target for treatment (Hertlein et al, 2010). Additionally, AKT has been reported to associate with IKK and to induce IKK activation contributing to NF-κB activation (Ozes et al, 1999; Romashkova & Makarov, 1999). Combining the data presented in our study, we show NVP-AUY922-AG-induced down regulation of AKT and both IKKα and IKKβ. Therefore, inhibition of HSP90 results in the down regulation of multiple signalling pathways in AML and this precedes evidence of apoptosis in the treated cells. The inhibition of these survival pathways seems likely to contribute to the synergy observed when NVP-AUY922-AG was combined with Ara-C.

HSP90 is present in greater abundance in cancer cells compared to non-cancerous cells (Whitesell et al, 2003). A dependence of the tumour cells on this activated, high-affinity chaperone could make HSP90 a potential target in tumour cells (Kamal et al, 2003). HSP90 inhibitors display preferential cytotoxic activity in transformed cells compared to normal cells (Whitesell et al, 2003) and this was confirmed in the present study. The reported selectivity of cancer cells to HSP90 inhibitors may be explained by dependence of transformed cells on a particular HSP90 client protein, as previously discussed (Solit & Chiosis, 2008). However, the relatively modest therapeutic index between AML cells and normal bone marrow suggests that NVP-AUY922-AG may be most effective when used at doses significantly below the maximum tolerated dose when given in combination with Ara-C or possibly other chemotherapeutic agents. The fact that synergy was observed over a wide concentration range, even when using low concentrations of NVP-AUY922-AG and Ara-C, supports this argument.

In summary, NVP-AUY922-AG shows significant activity as a single agent in both AML cell lines and primary AML samples. When used in combination with Ara-C, NVP-AUY922-AG shows synergistic interactions that appear to be dependent on the inhibition of multiple intracellular signalling pathways including PI3K and NF-κB. Taken together, our data suggest that the use of HSP90 inhibitors, particularly in combination with standard chemotherapy, represents a promising therapeutic strategy in AML.

Supplementary information is available at the British Journal of Haematology website.


EJW performed the research, designed the study, analysed the data and wrote the paper, ML performed research and analysed data, CJP designed the study and wrote the paper, SK wrote the paper, AKB wrote the paper and provided essential materials.

Conflict of interest

No authors have any conflicts of interest to declare.