FLT3 Expressing Leukemias Are Selectively Sensitive to Inhibitors of the Molecular Chaperone Heat Shock Protein 90 through Destabilization of Signal Transduction-Associated Kinases

Acute leukemias are known to have overexpression or mutations of oncoproteins resulting from distinct cellular genetic alterations. Examples are the oncoproteins that result from gene fusions involving transcription factors, such as MLL² or signal transduction-associated kinases, such as FLT3, RAF, and AKT. These oncoproteins are of great interest as potential molecular targets for specific chemotherapy. Rarely, leukemia results from a single genetic alteration; an example is leukemia induced by the Bcr-Abl oncoprotein, where the early phase of the disease can be reversed by inhibition of the fusion oncoprotein. However, most leukemias have more than one genetic alteration. As a result, therapeutic agents that disrupt more than one oncoprotein in the leukemia cells are of potential interest for multitargeted molecular therapy.

Hsps (including Hsp90 and related molecules) are attractive molecular targets because they are known to act as chaperones that prevent the degradation of a number of important cellular oncoproteins including receptor and nonreceptor kinases (1, 2). Hsp90 client proteins are brought into a multiprotein Hsp90/p23/Hsp70/Hop/Hsp40 complex (3). Ansamycin antibiotics, such as geldanamycin, specifically inhibit Hsp90 function. The structure-function relationship of geldanamycin and Hsp90 is well established (4, 5). Geldanamycin treatment of cells results in a loss of p23 protein from Hsp90 complex (6) and induces a degradation of client proteins (7, 8).

FLT3⁺ leukemias, including MLL fusion gene leukemias, were chosen for this study because they are relatively well characterized at the molecular level, generally have more than one genetic abnormality (9), frequently have a poor prognosis using conventional chemotherapy (10), and because the role of FLT3 in Hsp90 inhibition is not well defined. At the molecular level, FLT3⁺ MLL fusion gene leukemia cells often express an MLL fusion oncoprotein, frequently express high levels of FLT3 (11), and occasionally have mutations of FLT3 (12). In this study, we evaluated the hypothesis that some or all of the FLT3⁺ leukemia cell lines, including MLL fusion gene leukemia cell lines, require the chaperone function of Hsp90 and, as a result, will be sensitive to Hsp90 inhibitors. One of these ansamycin family of Hsp90 inhibitors, 17-AAG, is already in clinical trials in adult solid tumors (13, 14, 15). Here we show that treatment of FLT3⁺ leukemias with 17-AAG results in: (a) the degradation of RAF, AKT, and both wt and mutant FLT3; and (b) cell growth arrest and apoptosis.

17-AAG was kindly provided by the Developmental Therapeutics Branch of the Cancer Therapy Evaluation Program/National Cancer Institute/NIH (Bethesda, MD). HMA was obtained from Sigma (St. Louis, MI). HMA and 17-AAG were stored in the dark at 4°C and reconstituted in DMSO before use.

Human leukemia cell lines are RS 4;11 (MLL-AF4, established previously in our laboratory), Kid92, SEMK2 (MLL-AF4, gifts from Dr. Finbarr E. Cotter, Department of Hematology and Oncology, Institute of Child Health, London, United Kingdom), 1E8, LAZ221, U937 (obtained from Dr. Tucker W. Lebien, University of Minnesota, Minneapolis, MN), Nalm 6 (obtained from Dr. Jun Minowada, Hayashibara Biochemical Laboratories, Inc., Okayama, Japan), MV 4;11 (MLL-AF4, obtained from Dr. Carolyn A. Felix, University of Pennsylvania), Molm 13 (MLL-AF9), KOPN-8, NALM20, KLM-2 and p30/OHKUBO (obtained from Dr. Yoshinobu Matsuo, Hayashibara Biochemical Laboratories, Inc., Okayama, Japan). All of the cell lines were grown in RPMI 1640 tissue culture medium (Life Technologies, Inc., Grand Island, NY) with 10% FCS, 100 IU penicillin/ml, and 100 units streptomycin/ml in 5% CO₂ at 37°C. Before HMA or 17-AAG treatment, cells were seeded in 96-well plates at a density of 10,000 cells/well. The cells were treated with increasing doses of HMA or 17-AAG as indicated. The IC₅₀ for cell growth was determined using Nonradioactive Cell Proliferation kit (Promega, Madison, WI) after 72 h of treatment.

Cells were incubated with phycoerythrin-conjugated CD135 (anti-FLT3; BD PharMingen, San Diego, CA) for 30 min at 4°C. The cells were washed twice in PBS/0.1% BSA and analyzed on a FACS Calibur using CellQuest-Pro software (Becton Dickinson, Mountain View, CA).

One million cells were suspended in 1-ml solution containing 50 mg/ml PI, 0.1% sodium citrate, and 0.1% Triton X-100. The PI-stained samples were analyzed within 24 h. The cell cycle distribution and apoptosis were determined by the analysis of nuclear DNA content using CellQuest-Pro software (Becton Dickinson).

Cells cultured under the various conditions were harvested and labeled using the CaspaTag caspase activity kit (Serologicals Corporation, Norcross, GA). Briefly, harvested cells were incubated with FAM-VAD-FMK that irreversibly binds to caspases-1, 2, 3, 4, 5, 6, 7, 8, and 9. After two washes, PI was added, and cells were analyzed on a FACS Calibur flow cytometer. Green fluorescence signal (Caspatag) was measured on the FL-1 channel, whereas red fluorescence (PI) was measured on the FL-3 channel. Dot-plots of log FL-1 versus log FL-3 were generated using the CellQuest-Pro software.

PCR was performed on genomic DNA as published previously (16).

Cells were lysed by lysis buffer [50 mm Tris (pH7.4), 100 mm NaCl, 1 mm EDTA, 1 mm DTT, and 1% SDS], and the cell lysate was clarified by centrifugation. Twenty μg of cell lysate of each sample were electrophoresed by 10% SDS-PAGE, transferred to a nitrocellulose membrane, and immunoblotted with different antibodies, such as anti-FLT3 polyclonal antibody (s-18; Santa Cruz Biotechnology Biotechnologies, Santa Cruz, CA), anti-Hsp70 monoclonal antibody, anti-RAF monoclonal antibody, anti-AKT polyclonal antibody (BD PharMingen), anti-actin monoclonal antibody (Sigma), and anti-AF4 monoclonal antibody. Antihuman AF4 monoclonal antibody was produced in our laboratory using AF4-GST fusion protein as immunogen. For detection, the blots were incubated with horseradish peroxidase-conjugated anti-IgG antibody (Promega) and developed using the enhanced chemiluminescence detection system (Amersham Pharmacia Biotech, Arlington Heights, IL).

To assess FLT3 phosphorylation, cells were washed three times with serum-free RPMI 1640 and placed in the same medium overnight. Cells were then stimulated with human FL (100 ng/ml) at 37°C for 10 min, or treated with 17-AAG (37°C; 6 h) before FL stimulation (37°C; 10 min). Cells were washed twice with 1× PBS, lysed in IP-1 lysis buffer [20 mm Tris (pH 7.4), 1% NP40, 1 mm EDTA, 50 mm NaCl, 25 mm NaF, 4 mm sodium orthovanadate, 1 mm phenylmethylsulfonyl fluoride, and a protein inhibitor mixture tablet prepared according to manufacture’s recommendation (Roche Applied Science, Indianapolis, IN)] at 4°C for 15 min and centrifuged at 15,000 × g at 4°C for 15 min. The protein concentration of the supernatants was determined using BCA protein assay kit (Pierce, Rockford, IL). Protein (1.5 mg) from cell lysates were precleared with 5 μl of a 50% slurry of Protein A-Sepharose CL-4B (Amersham Pharmacia Biosciences, Piscataway, NJ) for 1 h and then incubated with 5 μl of anti-FLT3 rabbit polyclonal antibody (s-18; Santa Cruz Biotechnology Biotechnologies) overnight at 4°C. Thirty μl of Protein A beads were added and incubated for 4 h. This complex was washed three times with IP-1 lysis buffer, and 30 μl of SDS sample buffer was added to each sample. The samples were then electrophoresed, transferred to nitrocellulose membranes, and immunoblotted with anti-FLT3 polyclonal antibody (s-18; Santa Cruz Biotechnology) or anti-phosphotyrosine monoclonal antibody (4G10; Upstate Biotechnology Inc., Lake Placid, NY).

Cells were lysed with IP-2 lysis buffer [20 mm Tris (pH 7.4), 0.02% NP40, 1 mm EDTA, 50 mm NaCl, 20 mm sodium molybdate, 4 mm sodium orthovanadate, and a protein inhibitor mixture tablet prepared according to manufacture’s recommendation (Roche Applied Science)]. Antibody resin was prepared by incubating anti-Hsp90 monoclonal antibody (H9010) or anti-p23 monoclonal antibody (JJ3) with a 50% slurry of Protein A-Sepharose CL-4B (Amersham Pharmacia Biosciences) in 1× Tris-buffered saline (pH 8.0) for 1 h at room temperature. The H9010- or JJ3-conjugated resin was washed three times in IP-2 lysis buffer. Clarified cell lysate was incubated with 30 μl antibody resin at 4°C for 2 h. The resins were washed with IP-2 lysis buffer five times at 4°C and immunoblotted with anti-FLT3 polyclonal antibody (s-18; Santa Cruz Biotechnology), anti-Hsp90 (H9010) monoclonal antibody, and anti-p23 (JJ3) monoclonal antibody.

The relationship between the expression of FLT3 and the IC₅₀ of 17-AAG was examined using the nonparametric Spearman’s rank correlation coefficient. This method gives an idea of the strength of the relationship between the two variables using rank of the data. Spearman’s rank correlation coefficient was deemed to be the most appropriate measure of the relationship between the two variables. Data were analyzed using SAS software.

Human acute leukemia cell lines with or without FLT3 expression and with or without an MLL fusion protein (MLL-AF4 or MLL-AF9) were chosen for study. Cell lines were tested for FLT3 expression and mutation as described in “Materials and Methods.” As shown in Table 1, leukemia cell lines were grouped by the presence of: (a) FLT3-ITD; (b) FLT3-wt; and (c) FLT3-negative. FLT3-negative leukemia cell lines were generally relatively mature with a CD10⁺ CD19⁺ phenotype, and none had an MLL fusion protein. FLT3-wt leukemia cell lines were either: (a) CD10⁻ with an MLL fusion protein; or (b) CD10⁺ with or without an MLL fusion protein. Both FLT3-ITD leukemia cell lines were CD10⁻ with an MLL fusion protein. FLT3 expression levels that demonstrated by flow cytometry and Hsp90 expression levels determined by immunoblotting are shown in Table 1.

We conducted cell growth experiments with 17-AAG; 17-AAG is an ansamycin that binds to and inhibits the function of Hsp90 and is currently undergoing a Phase I clinical trial in solid tumors (13, 14). One of three similar experiments for FLT3⁺ human leukemia cell lines (RS 4;11 and SEMK2) and FLT3-negative phenotype-matched (CD19⁺) B-lineage leukemia cell lines (Nalm 6 and 1E8) is shown in Fig. 1,A and Table 1. We found that the IC₅₀ values for cell proliferation at 72 h were 700 ± 52 nm (RS 4;11), 350 ± 25 nm (SEMK2), 2400 ± 287 nm (Nalm 6), and 2100 ± 196 nm (1E8; Table 1; Fig. 2). Therefore, the two FLT3⁺ leukemia cells were more sensitive than the leukemia cells without FLT3. Interestingly, monocytic-macrophage Molm 13 leukemia cells, which contain both FLT3-ITD and MLL-AF9, were significantly more sensitive than FLT3-negative CD10⁻/CD19⁻-matched U937 (monocytic-macrophage leukemia cells) and were the most sensitive of all of the lines tested (Fig. 1,B; Fig. 2). Molm-13 is known to have both an ITD of FLT3 and a wt FLT3 allele (16, 17). The IC₅₀ values for cell proliferation at 72 h were 31 ± 3 nm (Molm 13) and 4500 ± 321 nm (U937; Table 1; Fig. 2).

This raised the question as to whether FLT3 mutations were present in the other cell lines that were under study. We used PCR to detect ITD of the FLT3 gene. In the cell lines tested, Molm 13 has the ITD in one allele and MV 4;11 has the ITD in both alleles, whereas none of the other cell lines were found to have ITD in the FLT3 gene (Table 1). We also tested the cell lines for other FLT3 mutations and found that none have the Asp mutation in position 835 that is also known to result in constitutive activation of FLT3 (16).

To address whether the sensitivity to 17-AAG in different leukemia cell lines is determined by the relative expression levels of the FLT3-wt and constitutively active form of FLT3-ITD, additional leukemia cell lines were compared for cell growth and expression levels of cell surface FLT3. As shown in Fig. 2 and Table 1, Molm 13 and MV 4;11, which express FLT3-ITD, are the most sensitive to 17-AAG. KLM2, 1E8, NALM6, and U937, which do not express FLT3, are the least sensitive to 17-AAG. Intermediate sensitivity was observed with cells expressing FLT3-wt. In an analysis of the relationship between the level of FLT3 expression and IC₅₀ of 17-AAG, all of the cell lines except those with FLT3-ITD were compared using Spearman’s rank correlation coefficient. The value for the correlation coefficient was −0.5358 with a P of 0.02, demonstrating a significant correlation between the expression of FLT3-wt and sensitivity to 17-AAG.

To directly evaluate the effects of FLT3-wt and FLT3-ITD on 17-AAG sensitivity, cell growth experiments were performed on wt murine Baf3 cells, and cells transfected with FLT3-wt and FLT3-ITD. The IC₅₀ values for cell proliferation at 72 h were 310 ± 60 nm (Baf3) and 90 ± 11 nm (Baf3-FLT3-wt, Baf3-FLT3-ITD). These results provide direct demonstration that FLT3 increases sensitivity of cells to 17-AAG.

We also tested Hsp90 expression levels in most of FLT3-ITD, FLT3-wt, and FLT3-negative cell lines. As shown in Table 1, Hsp90 expression levels were similar in all of the cell lines, except U937. The reduced sensitivity of U937 to 17-AAG might be caused by its high Hsp90 expression level. However, the overall sensitivity to 17-AAG was not significantly determined by Hsp90 expression levels.

The cell lines were treated with a second ansamycin Hsp90 inhibitor, HMA. As described in “Materials and Methods” cells were treated with a range of HMA concentrations for 72 h. Results in Fig. 1 C represent one of three similar experiments demonstrating dose-response data for the four leukemia cell lines. The IC₅₀ values for cell proliferation at 72 h were 65 ± 5 nm (RS 4;11), 40 ± 3 nm (SEMK2), 873 ± 28 nm (Nalm 6), and 674 ± 57 nm (1E8). In summary, the two FLT3-wt leukemia cell lines, RS 4;11 and SEMK2, were significantly more sensitive than the FLT3-negative leukemia cell lines, Nalm 6 and 1E8.

To determine whether the HMA has effects on a leukemia cell line with FLT3-ITD, Molm-13 was compared with the phenotype-matched U937, which does not express FLT3-ITD or the MLL-AF9 fusion proteins. Results shown in Fig. 1 D demonstrate dose-response data for the cell lines. The IC₅₀ values for cell proliferation at 72 h were 25 ± 2 nm (Molm 13) and 7800 ± 512 nm (U937).

Therefore, FLT3-ITD leukemia cells (Molm 13) were the most sensitive, FLT3-wt leukemia cells (RS 4;11 and SEMK2) were the intermediate, and FLT3-negative (Nalm 6, 1E8 and U937) were the least sensitive to HMA. Thus, the sensitivity to HMA was FLT3 dependent, as seen with 17-AAG.

Inhibition of cell proliferation or/and induction of apoptosis were evaluated as possible mechanisms of the cell growth inhibition by 17-AAG. We evaluated cell cycle arrest in FLT3⁺ leukemia cells treated with varied concentrations of 17-AAG for RS 4;11, 1E8, Molm 13, and U937 by staining them with PI and analyzing the DNA content using the CellQuest-Pro program as described in “Materials and Methods.” A representative experiment (n = 3) is shown in Fig. 3. After 24 h, treatment with 17-AAG resulted in a significant accumulation of cells in G₁ phase and fewer cells in S phase in both FLT3-wt leukemia cells (RS 4;11) and FLT3-ITD leukemia cells (Molm 13) as compared with phenotype-matched cells 1E8 and U937, respectively (Fig. 3).

For the induction of apoptosis, activation of caspases is a key event required in this process. To evaluate the induction of apoptosis by 17-AAG treatment, we analyzed activation of caspases and degradation of PARP, a caspase substrate. RS 4;11 and Molm 13 cells were treated with varied concentrations of 17-AAG. Fig. 4 shows that both RS 4;11 and Molm 13 developed apoptosis by 17-AAG after 48 h of treatment by immunofluorescence using flow cytometry. Viable cells were decreased from 91% (0 nm) to 69% (5000 nm) for RS 4;11 and 90% (0 nm) to 12% (1000 nm) for Molm 13, respectively. Apoptotic and dead cells were increased from 7% to 23% for RS 4;11 and 8% to 75% for Molm 13, respectively. In Fig. 5, cleaved PARP was also observed in both RS 4;11 and Molm 13 cells in a dose-dependent fashion after 48 h of treatment with 17-AAG by immunoblotting. Both immunofluorescence and immunoblotting experiments showed consistently that Molm 13 (FLT3-ITD) cells were more sensitive to 17-AAG than RS 4;11 (FLT3-wt) cells.

One important question is whether the MLL fusion oncoproteins require Hsp90 as a molecular chaperone, and whether the MLL-AF4 and MLL-AF9 fusion oncoproteins will be sensitive to Hsp90 inhibitors. This question is raised in part because the cell lines that were most sensitive to 17-AAG express both FLT3 and MLL fusion proteins.

To directly study possible effects of 17-AAG on levels of the MLL-AF4 protein, immunoblotting was conducted using FLT3-wt cell lines, RS 4;11 and SEMK2, after the incubation with varied concentration of 17-AAG. As shown in Fig. 6, no change was noted in the mass of MLL-AF4 fusion protein in the presence of 17-AAG. Therefore, the immunoblotting experiment provides evidence that the MLL fusion proteins are not directly dependent on Hsp90 in their role in the pathogenesis of the leukemia.

Other evidence that MLL fusion proteins are not important targets for 17-AAG is found in the leukemia cell lines that express wt FLT3 but do or do not express MLL fusion proteins. As shown in Table 1, there is no relationship between 17-AAG sensitivity and expression of MLL fusion genes.

Previous studies have demonstrated that inhibitors of Hsp90 reduce the total cellular mass of several protein kinases in solid tumor cells. RAF and AKT have been found to be reduced in quantity after Hsp90 inhibition in several solid tumor lines (5, 18, 19). Apoptosis in a tandemly duplicated FLT3-transformed leukemia cell line was also induced by Hsp90 inhibitors (20). In our study, we evaluated the molecular effects of 17-AAG on FLT3, RAF, and AKT protein expression in 17-AAG-sensitive RS 4;11, HPB-Null, Molm 13, and MV 4;11 leukemia cell lines. The amount of FLT3, as well as RAF and AKT, decreased after 17-AAG treatment in a dose-dependent manner for all four of the cell lines after 24 h by immunoblotting (Fig. 7).

Another Hsp, Hsp70, is known to be a participant in the Hsp90 chaperone complexes (21). 17-AAG has been reported to increase the level of Hsp70 in human colon cells (18). We examined the effect of 17-AAG on the amount of Hsp70 in FLT3⁺ leukemia cells that are 17-AAG sensitive. As shown in Fig. 7, Hsp70 expression increased after 24 h treatment with 17-AAG in a dose-dependent manner for RS 4;11, HPB-Null, Molm 13, and MV 4;11 leukemia cell lines.

We tested FLT3 expression in all of the cell lines in the study by immunofluorescence using flow cytometry. We next evaluated whether the Hsp90 inhibitor, 17-AAG, inhibited the cell surface expression of FLT3 in FLT3⁺ leukemia cell lines. Fig. 8 shows that 17-AAG resulted in significant reduction in the level of expression of FLT3 on cell surface with a shift of the peak to the left, in a dose-dependent manner, in SEMK2, RS 4;11, and Molm 13.

To evaluate the level of FLT3 phosphorylation in leukemia cell lines, FLT3 basal phosphorylation was measured using antiphosphotyrosine immunoblotting. Ligand-independent FLT3 phosphorylation was observed in MV 4;11, Molm 13, SEMK2, and RS 4;11, whereas FLT3 phosphorylation in THP-1 and HPB-Null cells was only seen after FL stimulation (Fig. 9 A). Basal phosphorylation of FLT3 in MV 4;11 was not affected by FL stimulation.

To distinguish the effect of FL on FLT3-wt cell lines (RS 4;11 and HPB-Null) and FLT3-ITD cell line (Molm 13), cell growth of these cell lines was determined in the presence or absence of FBS and in the presence or absence of FL (Fig. 10). In the presence of FBS, the cell growth of both the FLT3-wt and the FLT3-ITD cell lines was not affected by FL stimulation. This is likely explained by the existence of the trace of cytokines, including FL, in FBS. In the absence of FBS, the cell growth of FLT3-wt cell lines (RS 4;11 and HPB-Null) was enhanced by FL, whereas the cell growth of the FLT3-ITD cell line (Molm 13) was not affected by FL. Therefore, FLT3 signal is required for cell growth of FLT3-wt cells, and the FLT3-ITD cells harboring constitutive activation of FLT3 are largely independent of FL.

The effect of 17-AAG on FL-stimulated and constitutive phosphorylation of FLT3 was additionally tested (Fig. 9 B). Pretreatment of 17-AAG inhibited ligand-induced phosphorylation in RS 4;11 and HPB-Null cells in a dose-dependent manner. Basal phosphorylation in MV 4;11 cells was also inhibited by 17-AAG in a dose-dependent manner.

To determine whether 17-AAG modulated the expression of FLT3-wt and FLT3-ITD proteins, the membranes were stripped and reprobed with anti-FLT3 antibody. There was a decrease in the expression of both FLT3-wt and FLT3-ITD proteins in cells treated with 17-AAG. Therefore, the inhibition of FLT3 phosphorylation by 17-AAG is largely caused by the decrease of total amount of FLT3 protein in cells.

Because we had found that exposure to 17-AAG results in a significant decrease in FLT3 protein detected in cell lysate by immunoblotting (Fig. 7) and on the cell surface by immunofluorescence (Fig. 8), we asked whether FLT3 has a direct interaction with Hsp90. The molecular interaction between Hsp90 and FLT3-wt was studied using SEMK2 leukemia cells as described in “Materials and Methods.” Fig. 11 A shows that the FLT3-wt molecule was immunoprecipitated by either anti-Hsp90 antibody or anti-p23 antibody. This result provides direct evidence that FLT3-wt, Hsp90, and p23 are in the same molecular complex.

Because the FLT3-ITD expression level in either MV 4;11 or Molm 13 is very low, we were unable to use immunoprecipitation methods to demonstrate the interaction of FLT3-ITD and Hsp90 in human leukemia cell lines. As an alternative approach we compared the effect of 17-AAG on the interaction of FLT3 with Hsp90 in FLT3-wt and FLT3-ITD transfected cells. Two murine cell lines, Baf3/FLT3-wt and Baf3/FLT3-ITD, were chosen because of abundantly expressed FLT3-wt and FLT3-ITD. Baf3/FLT3-wt and Baf3/FLT3-ITD cells were treated with 0, 100, 500, and 1000 nm 17-AAG for 6 h. Cell lysates were incubated with anti-Hsp90-conjugated Protein A resin to immunoprecipitate Hsp90. Fig. 11 B demonstrates that both FLT3-wt and FLT3-ITD are associated with Hsp90 and p23 complex. 17-AAG resulted in reduced cellular FLT3-wt and FLT3-ITD in a dose-dependent manner. Both FLT3 and p23 were also found to be released from the Hsp90 complex. Our results do not permit us to determine the extent to which reduction in band intensity is because of a decrease in total FLT3 and/or release of FLT3 from the Hsp90 complex.

The major observations in these studies are that FLT3 is a client protein of Hsp90, and FLT3⁺ human leukemias are selectively sensitive in vitro to the Hsp90 inhibitor, 17-AAG. 17-AAG was found to be effective at concentrations that are attainable with reportedly acceptable tolerance in vivo in humans (13, 14, 15). The leukemia cells that were most sensitive were those that express FLT3-ITD followed by those that express FLT3-wt.

Because there is a strong concordance between FLT3 and MLL fusion gene expression, and all of the MLL fusion gene leukemia cell lines examined to date also express FLT3 protein (Table 1), it was important to determine whether the MLL fusion oncoprotein is an Hsp90 client and a possible target for Hsp90 inhibitors. Our studies provide evidence that MLL fusion proteins are not sensitive to Hsp90 inhibitors. We found that: (a) the amount of MLL fusion proteins was not reduced after 17-AAG treatment as seen in immunoblotting assays; (b) FLT3-wt leukemia cell lines have similar sensitivity to 17-AAG whether they are MLL fusion gene positive or negative; and (c) FLT3-wt and FLT3-ITD transformed Baf3 cells were more sensitive to 17-AAG than nontransformed Baf3 cells. These results are consistent with a report from others; Dias et al. (22) found that FLT3-wt-transformed HL-60 cells were sensitive to geldanamycin, another ansamycin family member. Whereas it is possible that total mass as reflected in immunoblotting assays does not necessarily reflect functions of the MLL fusion proteins, it is likely that the functions of MLL fusion proteins are also not sensitive to Hsp90 inhibition.

A likely explanation for the association between MLL fusion protein expression and FLT3 expression is that cells with cell stage-specific tyrosine kinases undergo malignant transformation by the fusion gene. FLT3 is an excellent example of such a cell-stage specific tyrosine kinase. Examples of this association are: (a) FLT3 is expressed very early in the Pro B lymphocyte cells that are transformed by MLL-AF4; and (b) FLT3 expressing cells in the monocyte lineage that are transformed by MLL fusion genes. To date there is no evidence that the fusion gene alters FLT3 expression.

Recent studies with a variety of FLT3 kinase inhibitors have demonstrated effects in the presence of mutated but generally not wt FLT3 (23, 24, 25, 26, 27). Our results suggest that both FLT3-wt and FLT3-ITD are important targets for the Hsp90 inhibitor, 17-AAG. In all of the FLT3-wt cell lines we tested, only one, THP-1, is an acute myelogenous leukemia cell line, and it has the same sensitivity to 17-AAG as the acute lymphoblastic leukemia cell lines tested. This suggests that the type of leukemia (acute myelogenous leukemia or acute lymphoblastic leukemia) may not be important to the sensitivity to 17-AAG. Several lines of evidence indicate that in FLT3⁺ leukemias, the FLT3 oncoprotein is important in the response to 17-AAG. First, we found a direct correlation between the expression of FLT3 and the response to 17-AAG. Leukemias tested that were relatively resistant to 17-AAG (1E8, Nalm 6, and U937) were all FLT3 negative, whereas the sensitive cells (SEMK2, RS4;11, HPB-Null, Molm 13, and MV 4;11) were all FLT3-positive. Second, we directly demonstrated a decrease in both FLT3-wt and FLT3-ITD cell surface expression in SEMK2, RS 4;11, and Molm 13 cell lines after 48 h treatment of 17-AAG by immunofluorescence. Third, Molm 13 and MV 4;11 cell lines, which were reported previously (16, 17, 28) and confirmed by our studies to have ITD of FLT3 resulting in constitutive activation of FLT3, were found to be most sensitive to 17-AAG. Fourth, 17-AAG has an effect on both FLT3-wt and FLT3-ITD by inhibition of their phosphorylation and dissociation from Hsp90.

Previous studies demonstrated that the Hsp90 inhibitor, HMA, inhibited the growth of murine cells transfected with a mutant FLT3 (FLT3-ITD/32D; Ref. 29) and also that FLT3-ITD formed a complex with Hsp90 (20). We found that: (a) 17-AAG inhibited the growth of both FLT3-wt and FLT3-ITD leukemia cells; (b) FLT3-wt formed a complex with Hsp90 and p23 in human leukemia cells; (c) both FLT3-wt and FLT3-ITD formed complexes with Hsp90 and p23 in Baf3 cells transfected with FLT3-wt and FLT3-ITD; and (d) 17-AAG induced the dissociation of p23 and FLT3 from Hsp90 complex. Our results provide direct evidence that FLT3, like other kinases, RAF (30), AKT (31), and Bcr-Abl (32), is a client of Hsp90. Both FLT3-wt and FLT3-ITD proteins apparently require Hsp90 to act as a molecular chaperone to maintain their biological functions.

We found that 17-AAG resulted in a reduction of total amount of FLT3, RAF, and AKT. Whereas previous studies have demonstrated that ligand binding of FLT3 results in activation of the downstream signal transducers and activators of transcription, RAF/RAS/mitogen-activated protein kinase and phosphatidylinositol 3′-kinase/AKT kinase cascades (33, 34, 35), there is no evidence that FLT3 controls the total amount of RAF and AKT kinases. Most likely, 17-AAG inhibits the Hsp90 chaperone function on the kinases independently, with the possibility that the total cellular effects of 17-AAG become synergistic.

We found the effect of 17-AAG to be less in FLT3-wt cell lines (RS 4;11 and HPB-Null) than FLT3-ITD cell lines (Molm 13 and MV 4;11). FLT3-ITD cell lines were much more sensitive to 17-AAG. The reason for these differences is not clear and may be multiple. One possibility is that the downstream signals of FLT3-wt and FLT3-ITD are not identical. Previous studies have indicated that Flt3-ITD led to strong factor-independent activation of signal transducers and activators of transcription 5 (36) and up-regulation of the serine-threonine kinase Pim-2 (37).

Hsp90 inhibitors, such as 17-AAG, may be considered as multitarget therapy, because any cellular protein that uses Hsp90 as a molecular chaperone is likely to be a target for the inhibitor. This multitarget therapy is likely to be advantageous in the therapy of leukemias, which in most cases have more than one genetic mutation.

17-AAG is currently in Phase I clinical studies in breast, colon, and other solid tumors, and is reported to be well tolerated in these studies (13, 14, 15). Similar Phase I studies in leukemias expressing FLT3 are planned.

We thank Dr. David O. Toft for supplying mouse monoclonal antibodies of anti-Hsp90 (H9010) and anti-p23 (JJ3), and Dr. D. Gary Gilliland for providing Baf3/FLT3 and Baf3/FLT3-ITD cell lines. We also thank Susan Puumala for excellent support with statistic analysis.