KF25706, a Novel Oxime Derivative of Radicicol, Exhibits in Vivo Antitumor Activity via Selective Depletion of Hsp90 Binding Signaling Molecules

Radicicol, a macrocyclic antifungal antibiotic that was originally isolated from the fungus Monosporium bonorden (1), was shown to suppress the transformed phenotype caused by various oncogenes such as src, ras, raf, mos, and fos (2, 3, 4). Radicicol has also been reported to inhibit v-src tyrosine kinase activity and its downstream effector molecules at the cellular level (5, 6). Recently, we revealed that radicicol could inhibit K-ras-activated aberrant signaling pathway through the selective depletion of Raf-1 protein and subsequent inhibition of MAPK² pathway in K-ras-transformed rat cells (7).

Such a selective destabilization of Raf-1 protein was also reported for the benzoquinone ansamycin family antibiotic GA (8, 9, 10). In addition, geldanamycin was shown to directly bind to the NH₂-terminal domain of Hsp90 (11, 12) and to destabilize the Raf-Hsp90 multimolecular complex in the cells (8, 9, 10). Under these circumstances, we have assessed whether radicicol would bind to Hsp90 using several criteria. (a) Radicicol was shown to compete with the binding of Hsp90 to geldanamycin affinity column (13). (b) Biotinylated radicicol oxime was shown to selectively bind to cellular and recombinant Hsp90 (14). (c) Direct binding of Hsp90 family of proteins to radicicol was confirmed by the method using immobilized radicicol oxime beads, and this binding was competed with soluble GA.³ These results strongly suggested that radicicol represents a structurally unique antibiotic that is capable of binding to Hsp90 and interfering with its function.

Although radicicol exhibits interesting biological activities in vitro, the compound lacks in vivo antitumor activity in animal models⁴ (see also Fig. 5A and Table 2). Because it was reported that the inhibitory effect of radicicol against tyrosine kinases was abolished by reducing agents such as DTT (5), we have designed novel derivatives of radicicol that are stable in the presence of thiol.⁴ Our novel oxime derivatives were shown to be stable in mouse serum as well as in DTT, and several of these compounds exhibited significant antitumor activity in vivo in xenografted human tumor models that are completely resistant to radicicol by consecutive s.c. injections.⁴

Although the precise function of the Hsp90 family of proteins is still not fully elucidated, these proteins have been receiving much attention because of their unique property of participating in multiple signal transduction pathways. Recent studies have indicated that Hsp90 or Grp94 is required for: the proper function of steroid hormone receptors (15, 16); the activities of several tyrosine kinases such as EGFR, p185^erbB2, and v-src family kinases (17, 18, 19, 20, 21, 22); the activities of serine/threonine kinases such as Raf-1 (8, 9, 10, 23) and Cdk4 (24); and the activity of the mutated p53 tumor suppressor gene product (25, 26, 27, 28). Hsp90 binding agents, such as benzoquinone ansamycines (GA and herbimycin A) and radicicol, have been proven to be very valuable tools for understanding the multiple functions of Hsp90 in vitro; however, in vivo antitumor activity of the Hsp90 targeting agents and the proper role of Hsp90 in in vivo tumor growth are not fully elucidated, except in the case of several GA derivatives (Refs. 29, 30, 31; Ref. 32 and references therein).

Our goals in this study were as follows: (a) to validate whether KF25706, a novel oxime derivative of radicicol, would inhibit K-ras and v-src signaling pathway like radicicol; (b) to see whether KF25706 would deplete Hsp90-associated molecules such as Raf-1 and p185^erbB2; (c) to examine whether KF25706 would bind to Hsp90 using GA beads; (d) to examine whether the antitumor activity of KF25706 via clinically relevant administration route of i.v. injection; and (e) to validate whether the drug would deplete Hsp90-associated molecules in animal models.

Our results suggest that KF25706 can bind to Hsp90 as radicicol and thus destabilize Hsp90-associated molecules in vitro and inhibit their downstream signaling. The drug was also shown to exhibit antitumor effect in vivo via possible interaction with Hsp90, suggesting that the drug could be a candidate for further preclinical development.

Radicicol (Fig. 1A) was produced by fermentation, and its derivatives (KF25706 and KF29163; Fig. 1, B and C, respectively) were chemically synthesized at our institute according to methods to be published elsewhere.⁴

HEPES, PMSF, leupeptin, sodium o-vanadate (Na₃VO₄), sodium fluoride (NaF), and β-glycerophosphate were purchased from Sigma Chemical Co. (St. Louis, MO). NaCl, EDTA, and DTT were from Wako Pure Chemical Industries, Ltd. (Osaka, Japan), and NP40 was from Nacalai Tesque (Kyoto, Japan).

Anti-Raf-1(C-12) and anti-Cdk4(C-22) rabbit polyclonal antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-K-ras monoclonal antibody (clone 234-4.2), anti-v-src monoclonal antibody (clone 327), and anti-c-neu (ErbB2) monoclonal antibody (clone 3B5) were purchased from Oncogene Research Products (Cambridge, MA). We also used phosphospecific MAPK antibody (New England Biolabs, Beverly, MA), anti-Erk2 monoclonal antibody (clone 1B3B9; Upstate Biotechnology, Lake Placid, NY), antiphosphotyrosine monoclonal antibody (clone 25.2G4; Kyowa Medex, Co., Ltd. Tokyo, Japan), anti-p53 monoclonal antibody (clone 80; Transduction Laboratories, Lexington, KY), and anti-Hsp90 monoclonal antibody (clone MA3-011; Affinity BioReagents, Neshanic Station, NJ).

SK-BR-3, DLD-1, HCT-116, A549, PC-3, DU145, and A431 cells were purchased from the American Type Culture Collection through Dainippon Pharmaceutical Co., Ltd. (Osaka, Japan). MCF-7 and MCF-7/ADM were obtained from the National Cancer Institute (Bethesda, MD). NRK and KNRK5.2 cell lines were obtained from Dr. David A. Johnson (Eli Lilly and Co., Indianapolis, IN). 3Y1-B and SR-3Y1 cell lines were obtained from Riken Cell Bank (Saitama, Japan).

The cell cultures were performed at 37°C in a humidified atmosphere of 5% CO₂.

Human breast carcinoma MX-1 cell lines were obtained from Central Institute for Experimental Animals (Kanagawa, Japan), and MCF-7, DLD-1, and A431 xenograft cell lines were established by inoculation of the cultured cells into the flank of adult BALB/c nu/nu mice (Nippon Clea Co., Tokyo, Japan). These tumors were passaged in vivo using a troacar.

The cells were precultured in appropriate medium for 24 h in 96-well microwell plates (Nunc, Roskilde, Denmark). Drugs were added to the plate in serial 3-fold dilutions (n = 3), and the plates were incubated for another 72 h. To determine the time dependency of the antiproliferative activity, we exposed the cells to drugs for the indicated time periods (1, 2, 4, 8, 24, 48, and 72 h) and transferred them to drug-free medium. After 72 h of treatment, the cell viability was determined using a microculture tetrazolium (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; Sigma Chemical Co.) assay that was described previously (33, 34). The antiproliferative activities of the drugs are shown in terms of IC₅₀s.

Generally, cells plated in 24-well tissue culture plates (Nunc) were used for each sample. After preculture for 8 h, radicicol and its derivatives were added to each well without any fresh medium, and the cells were cultured further. After the drug treatment, the cells were washed once with PBS without calcium [PBS(−); ICN Biochemicals, Aurora, OH] and lysed by the addition of 20 μl per well of ice-cold lysis buffer [50 mm HEPES-NaOH (pH 7.4), 250 mm NaCl, 1 mm EDTA, 1% NP40, 1 mm DTT, 1 mm PMSF, 5 μg/ml leupeptin, 2 mm Na₃VO₄, 1 mm NaF, and 10 mm β-glycerophosphate]. The cells were lysed for 10 min on ice and clarified by centrifugation. Whole cell lysate was heated in 1 × SDS sample buffer for 5 min at 95°C and subjected to SDS-PAGE. The ex vivo tumor specimens were minced with a razor and strained with a Falcon cell strainer (Becton Dickinson, Lincoln Park, NJ). The cells were lysed, and SDS-PAGE samples were prepared in the same way as in vitro cultured cell samples. After SDS-PAGE, the protein was transferred to polyvinylidene difuluoride membranes (Immobilon-P; Millipore, Tokyo, Japan) and immunoblotted with appropriate primary antibodies. For detection, the blots were incubated with the appropriate secondary antibodies conjugated with horseradish peroxidase (Amersham Life Sciences, Buckinghamshire, England) and developed using the enhanced chemiluminescence detection system (Amersham), according to the instructions of the manufacturer.

GA was derivatized and immobilized as reported previously (22). Briefly, 1,6-hexanediamine was added to GA (10 mm in CHCl₃) at a 10-fold molar excess and allowed to react for 2 h at room temperature. After aqueous extraction, 17-hexamethylenediamine-17-demethoxygeldanamycin was dried, redissolved in DMSO, and reacted with AffiGel 10 resin (Bio-Rad, Hercules, CA). The resulting beads were washed in TNES buffer [50 mm TrisHCl (pH 7.5), 1% NP40, 2 mm EDTA, and 100 mm NaCl] and blocked in 1% BSA before use. SK-BR-3 cells were lysed in TNES buffer containing 1 mm sodium orthovanadate, 20 μg/ml aprotinin, 20 μg/ml leupeptin, and 1 mm PMSF. One hundred μg of total protein were incubated with DMSO (control) or KF25706 or KF29163 at the concentrations indicated in the text. After 30 min of the end-over-end mixing, resins were washed three times in TNES buffer and boiled in Laemmli gel loading buffer. Affinity-purified proteins were separated on 8% SDS-polyacrylamide gels and visualized by silver stain. Beads were displaced and quantified using Adobe Photoshop and NIH image software.

All human tumors were inoculated s.c. in the flanks of nude mice. For the evaluation of antitumor activity, the tumor volume was calculated using the following formula, according to the method of the National Cancer Institute (Ref. 35; length and width of the tumors measured in mm):

Drug efficacy was expressed as the ratio of the mean experimental V/V₀ value to that of the control group (T/C ratio), where V is the tumor volume at the day of evaluation and V₀ is the tumor volume at the day of the initial treatment with the drug.

Radicicol and its derivatives (KF25706 and KF29163) were administered twice daily or daily by i.v. injections for 5 days from day 0 to day 4. Human breast carcinoma MX-1 cells were transplanted on day −14. Human breast carcinoma MCF-7 were transplanted on day −19, and E.P. Hormone Depot (Teikoku Zouki Pharmaceutical Co., Ltd., Tokyo, Japan) was injected i.m. on days −18 and −5. Human colon carcinoma DLD-1 and human epidermoid carcinoma A431 were transplanted on days −20 and −18, respectively.

Because radicicol was shown to suppress the transformed phenotype of various oncogene-transfected cell lines and to inhibit aberrant signal transduction caused by ras and v-src (2, 3, 4, 5, 6, 7), we have assessed whether KF25706 would also inhibit the K-ras and v-src signaling pathway, using v-src- and K-ras-transformed cell lines.

KF25706 showed slightly more potent growth-inhibitory activities against K-ras- and v-src-transformed rat cell lines than radicicol (Table 1) and also induced the reversal of the transformed phenotype of K-ras- as well as v-src-transformed rat cell lines, namely, KNRK5.2 and SR-3Y1 (data not shown).

Previously, we revealed that radicicol could deplete Raf-1 protein, thus inhibiting its downstream effector proteins (7). To see whether KF25706 would inhibit the Ras-Raf pathway, KNRK5.2 cells were treated with varying concentrations of KF25706 for 40 h, and the cell lysates were subjected to Western blotting analysis. KF25706 apparently depleted the Raf-1 protein in a concentration-dependent manner and inhibited K-ras-induced phosphorylation of MAPK with little, if any, effect on Erk2 and K-ras protein expression levels in KNRK5.2 cells (Fig. 2A).

We also tested whether KF25706 would inhibit v-src-induced tyrosine phosphorylation in SR-3Y1 cells. SR-3Y1 cells exhibited a markedly elevated level of total tyrosine phosphorylation of the cellular proteins, as compared to normal 3Y1 cells (Fig. 2B and data not shown). After 48 h of treatment with KF25706, the level of total tyrosine phosphorylation of cellular proteins was decreased in a concentration-dependent manner in SR-3Y1 cells, as assessed by Western blotting using an antiphosphotyrosine antibody (Fig. 2B). We also found that v-src protein itself was depleted from the cells after KF25706 treatment (Fig. 2C).

These results suggested that KF25706 had activities to inhibit K-ras or v-src signaling through the depletion of Raf-1 or v-src protein expression. Because previous studies revealed that both Raf-1 and v-src proteins were associated with Hsp90, constituting a multimolecular complex containing Hsp90 (8, 9, 10, 20, 21, 22, 23), our result also suggests that KF25706 might affect Hsp90 function and destabilize the Hsp90-associated signaling molecules.

The growth-inhibitory activities of radicicol, KF25706 (active derivative of radicicol oxime),⁴ and KF29163 (less active derivative of radicicol oxime)⁴ (Fig. 1) were compared for various cultured human tumor cell lines (Table 1). Radicicol and KF25706 showed potent antiproliferative activity against a wide variety of human tumor cell lines from various tissue origin (breast, colon, lung, prostate, and vulva) and with various molecular characteristics (ErbB family receptor tyrosine kinase overexpression, activated mutation in ras oncogene, mutation in p53 tumor suppressor gene, and estrogen receptor expression status). KF29163, a less active analogue, showed weaker growth-inhibitory activities against these cell lines as well as rat 3Y1 cell lines. On average, KF25706 was slightly more potent than radicicol against these human tumor cell lines.

A large number of oncogenic and/or signaling molecules such as Raf-1, Cdk4, erbB1, erbB2, and mutated p53 molecules have been reported to exist in multimolecular complexes containing Hsp90 family chaperone proteins and to be dependent on the association with the chaperone for their stability and function. To assess the relationship between the effect on Hsp90-associated signaling molecules and antiproliferative activity, we examined the effect of KF25706 on protein expression levels of Hsp90-associated signaling molecules using the human breast carcinoma SK-BR-3 cell line, which was the most sensitive to KF25706 among the various human tumor cell lines tested (Table 1; IC₅₀ = 29 nm). This cell line was known to express high levels of p185^erbB2 and mutant p53. As shown in Fig. 3A, KF25706 completely depleted the expression of p185^erbB2 protein at 0.1 μm, whereas radicicol showed only partial depleting activity at 0.1 μm. KF29163, a less active analogue, was completely inactive at a concentration of 0.1 μm (Fig. 3A), As shown in Fig. 3B, KF25706 depleted p185^erbB2 in a concentration-dependent manner. In addition, the drug also depleted Raf-1, Cdk4, and mutant p53 proteins from the cells, whereas the level of Erk2 protein remained unchanged even at the presence of 1 μm KF25706 (Fig. 3B). To further clarify the relationship between the depletion of p185^erbB2 and the antiproliferative activity of KF25706, we carried out a kinetic analysis of these activities in SK-BR-3 cells. As shown in Fig. 3, C and D, it took 16 h for 0.3 μm KF25706 to deplete p185^erbB2 protein to basal levels. Other Hsp90-associated signaling molecules, such as Raf-1 and Cdk4, were also depleted in the course of 24 h of treatment with KF25706 (data not shown). On the other hand, cell growth inhibition (IC₅₀ < 100 μm) was observed after 8 h of treatment with KF25706. These results showed that the protein depletion occurred at the earlier time point and that, after that, cell growth was inhibited.

Taken together, these results suggest that the ability of KF25706 to deplete the Hsp90-associated signaling molecules contributes to the antiproliferative activity of the drug against human breast carcinoma SK-BR-3 cell line in vitro.

Previously, we reported that radicicol could bind to Hsp90 using a GA-affinity column (13). To assess binding of KF25706 to Hsp90, we again used the solid-phase GA competition assay.

SK-BR-3 cells were lysed in TNES buffer, and proteins were affinity-purified from lysate using GA beads. Preincubation of the lysates with KF25706 inhibited binding of cellular Hsp90 to immobilized GA with an EC₅₀ of 5.5 nm (Fig. 4A and B). This concentration was comparable to an IC₅₀ of 29 nm for cell growth inhibition (Table 1). KF29163, a less active analogue of radicicol, failed to compete with GA for binding to Hsp90 (Fig. 4C). These results suggest that the binding of KF25706 to Hsp90 could be very important for its growth-inhibitory activity.

Although radicicol shows potent growth-inhibitory activity in vitro, the compound lacks antitumor activity in vivo because of its instability when exposed to thiol molecules.⁴ Previously, we generated novel oxime derivatives of radicicol and revealed that several derivatives, including KF25706 but not KF29163, showed significant growth-inhibitory activity in vivo using xenografted human tumor models by consecutive s.c. injections.⁴

To see whether KF25706 could also exhibit in vivo antitumor activity via a clinically relevant route of administration, we have tested the antitumor effect of KF25706 and radicicol giving five consecutive intravenous injections. As shown in Fig. 5A, KF25706, at a dose of 100 mg/kg twice daily applied by five consecutive i.v. injections, showed a significant and potent growth-inhibitory activity in the ER(−) MX-1 breast carcinoma xenograft model. However, radicicol, at a dose of 50 mg/kg twice daily given by five consecutive i.v. injections (which was the maximal tolerated dose of radicicol; data not shown) did not show any significant antitumor effect. KF25706 also showed significant growth delay of MX-1 cells in vivo at a dose of 100 mg/kg once daily (five consecutive intravenous injections; Fig. 5A). As shown in Fig. 5, B–D, KF25706 at a dose of 100 mg/kg twice daily (five consecutive i.v. injections) showed a significant growth-inhibitory activity against ER(+) MCF-7 breast carcinoma (Fig. 5B), DLD-1 colon carcinoma (Fig. 5C), and A431 epidermoid carcinoma (Fig. 5D) xenograft models.

As summarized in Table 2, MX-1 breast carcinoma was most sensitive to KF25706, and MCF-7 breast carcinoma was also sensitive to the drug. The drug also showed significant but marginal activity against DLD-1 colon and A431 epidermoid carcinoma models. However, radicicol at a dose of 50 mg/kg twice daily for five consecutive i.v. injections (which was maximum tolerable dose of radicicol) did not show any significant antitumor effect in any of the models tested (Table 2).

To investigate whether KF25706 would affect Hsp90 and its associated signaling molecules in in vivo xenograft models, we examined the protein expression levels of Hsp90-associated signaling proteins by Western blot analysis using the human breast carcinoma MX-1 xenograft model, which was the most sensitive to KF25706 treatment (Table 2). An effective dose of KF25706 (100 mg/kg twice daily for five consecutive i.v. injections) was administered into the MX-1 human breast carcinoma-bearing mice (on days 0–4). On day 5, we removed MX-1 tumor fragments from the KF25706-treated mice and analyzed the Hsp90-associated protein levels (Raf-1 and Cdk4). As shown in Fig. 6, both Raf-1 and Cdk4 proteins were significantly decreased in KF25706-treated MX-1 tumor specimens, whereas Erk2 protein levels had declined slightly. With the same treatment schedule, KF29163 and radicicol did not show any effect against Hsp90-associated molecules in vivo. These results, together with the results of antitumor activity of these compounds, suggest that KF25706 might also bind to Hsp90 in vivo and that this effect might contribute to the antitumor activity of this compound.

Radicicol has been shown to exhibit various unique biological activities in vitro, including suppression of the transformed phenotype in a variety of oncogene-transformed cells (2, 3, 4, 6), v-src tyrosine kinase inhibition (5, 6), inhibition of the Ras signaling pathway (7), and inhibition of Hsp90 family chaperone function (7, 13). Because radicicol lacks antitumor activity in vivo in animal models, we recently synthesized a series of novel oxime derivatives of radicicol that showed significant antitumor activity in vivo, given by consecutive s.c. injections.⁴ In this study, we investigated the biological activities of KF25706, one of these radicicol oxime derivatives, in detail. KF25706 was shown to inhibit the aberrant signaling pathways of activated K-ras and v-src (Fig. 2) and deplete Hsp90-associated signaling molecules such as Raf-1, v-src, p185^erbB-2, Cdk4, and mutated p53 (Figs. 2 and 3, A and B). In addition to these effects, it was also demonstrated that KF25706 competed with GA for binding to Hsp90 using GA-affinity beads. These results suggested that KF25706 has the same mechanism(s) of action as radicicol, at least in vitro.

In human breast carcinoma SK-BR-3 cells, KF25706 inhibited the growth of the cells and also depleted Hsp90-associated molecules in the same range of concentration and time course (Fig. 3, C and D). These results suggested that KF25706 inhibited tumor cell growth through the putative inhibition of Hsp90 chaperone function in target cells. Because KF29163, a less active analogue of radicicol oxime, showed less potent activity in growth inhibition (Table 1), p185^erbB2 depletion (Fig. 3A), and binding to cellular Hsp90 (Fig. 4C), it was suggested that the affinity to Hsp90 protein might be important for the biological activities of radicicol derivatives.

KF25706 showed significant and potent antitumor activity in vivo against various human tumor xenograft models via clinically relevant i.v. injections (Fig. 5 and Table 2). Among these tumor cell lines, MX-1 cells are most sensitive to KF25706. Using this cell line, we further analyzed the effect of KF25706 on Hsp90-associated signaling molecules (Raf-1 and Cdk4) expression level after the treatment with the drug in vivo. Our results confirmed that KF25706 could deplete Raf-1 and Cdk4 proteins in drug-treated MX-1 tumor cells (Fig. 6), suggesting that the drug could affect Hsp90 function in human tumor xenograft in vivo as well as in vitro. To further clarify the relationship between the inhibition of Hsp90 function and in vivo antitumor activity, we also examined the effects of KF29163 and radicicol on Raf-1 and Cdk4 in vivo. As shown in Fig. 6, KF29163 and radicicol showed no effects on protein expression levels of Raf-1 and Cdk4 after five consecutive twice daily i.v. injections. These results suggest that in vitro affinity to Hsp90 as well as the stability in vivo could be the determinant for in vivo antitumor activity. Taken together, these results also suggest that a possible target of KF25706 in antitumor activity in vivo is Hsp90.

Hsp90 family chaperone proteins have been reported to associate with several important signaling molecules, including proto-oncogenes (Raf-1, EGFR, and p185^erbB2), a tumor suppressor gene (mutant p53), a cell cycle regulator (Cdk4), and steroid receptors such as ER and glucocorticoid receptor. Because cancer cells have multiple genetic alterations, it would be desirable to identify an agent capable of affecting multiple targets in the signal transduction pathway. In this context, Hsp90-binding agents (such as benzoquinone ansamycins and radicicol) might affect several important signaling pathway simultaneously through the down-regulation of several Hsp90-associated signaling molecules. However, because Hsp90 is thought to be a ubiquitous chaperone protein, there might be a possibility that inhibition of Hsp90 function might cause adverse effect(s). Because benzoquinone ansamycines were reported to exert liver and renal toxicity (36, 37), we assessed whether KF25706 would exhibit the same type of toxicity in mice. Our results showed that KF25706 showed no liver and renal toxicity at a dose 100 mg/kg twice daily for five consecutive injections (effective dose for tumor growth inhibition), which was determined by serum parameters (GPT and biliary urea nitrogen) and pathological analysis (data not shown). These results suggested that liver and renal toxicity of benzoquinone ansamycines might not be mediated through the inhibition of Hsp90 function. Further studies are needed to clarify this hypothesis.

Several previous reports revealed that Hsp90 family chaperones were up-regulated in tumor cell lines (38, 39, 40, 41) and some kinds of cancer specimens (42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52). In addition to this, poor survival in breast cancer patients overexpressing Hsp90 was also reported (53, 54). These results might be the rationale for Hsp90-binding agents as tumor-selective signaling blockers; however, further basic studies should be performed.

In this report, we assessed the effect of KF25706 against Hsp90 by the depletion of Hsp90-associated signaling molecules such as Raf-1, p185^erbB-2, Cdk4, and so on. However, there is a possibility that KF25706 exerts its antitumor effects through the interaction with nuclear steroid hormone receptor family proteins such as ER, progesterone receptor, or glucocorticoid receptor, which play an important role in hormone-dependent tumor cell growth. Previously, we showed that radicicol prevented the maturation of progesterone receptor complex and, thus, inhibited the binding of the ligand to the receptor (13), which was also shown for GA (55). In addition to these effects, radicicol was also reported to show various unique biological activities such as the inhibition of angiogenesis (56), the inhibition of cyclooxygenase 2 expression (57, 58), and the enhancement of gelsolin expression (4). The relationship between these biological activities and the function of Hsp90 family protein is currently unclear. However, there is a possibility that these effects could contribute to the antitumor activity of KF25706 in vivo.

In summary, we have revealed that KF25706, a novel oxime derivative of radicicol, could exhibit both in vitro and in vivo antitumor activity through inhibition of Hsp90 family chaperone function in the cells. These results also indicated that KF25706 could be a useful compound for investigating the Hsp90 function both in vitro and in vivo. Because the drug could be administrated via i.v. injections and showed no liver and renal toxicity in vivo, KF25706 could be a candidate drug for further clinical development.

We thank Miyoko Asano, Yuka Watanabe, and Izumi Omori for excellent technical assistance.