Dimeric ansamycins—A new class of antitumor Hsp90 modulators with prolonged inhibitory activity



The geldanamycin derivative 17-allyamino-17-demethoxygeldanamycin (17-AAG) is a clinical stage ATP-competitive HSP90 inhibitor that induces degradation of HSP90 client proteins. 17-AAG contains 1 ansamycin moiety and is highly potent in conventional cell killing assays. Since active Hsp90 exists as a dimer, we hypothesized that dimeric compounds containing 2 ansamycin pharmacophores might inhibit Hsp90 function more efficiently than 17-AAG. Here, we show that monomeric and dimeric ansamycins exert their activity in distinct ways. Under conditions of continuous exposure, 17-AAG induced client degradation and cell growth inhibition more readily than the dimeric drugs CF237 and CF483. By contrast, 24 hr treatment of various tumor cells with 17-AAG followed by drug washout caused temporary client degradation and cell cycle arrest but minimal cell death, whereas both dimers induced massive apoptosis. CF237 remained bound to Hsp90 for days after drug withdrawal and, while both monomeric and dimeric compounds caused accumulation of the inactive intermediate Hsp90 complex, this effect disappeared following washout of 17-AAG but not CF237. The dimer was also retained for longer in tumor xenografts and displayed superior antitumor activity in vivo. These results indicate that monomeric and dimeric Hsp90 inhibitors have distinct biological profiles and work differentially toward target inhibition. © 2006 Wiley-Liss, Inc.

Heat shock protein 90 (HSP90) is a conserved molecular chaperone that mediates the maturation and stability of a set of cancer-associated proteins, referred to as ‘clients’. These include steroid receptors, EGFR family members, MET, Raf-1 kinase, AKT, Bcr-abl, mutant p53, CDK4 and many other oncogenic molecules.1, 2, 3 Hsp90 functions as a super-chaperone complex in association with various cochaperone proteins. Hsp90 mainly exists in 2 types of multiprotein complexes, referred to as ‘intermediate’ and ‘mature’.3, 4 In the intermediate complex, which is the ADP-binding form, the major cochaperones are Hsp70, Hsp40, HOP and HIP. Upon ATP binding, cdc37, p23 and immunophilins replace the original cochaperones to assist the conformational maturation of the client proteins and maintain those proteins in an active state to exert their function.4 It is thought that Hsp90 inhibitors bind to and stabilize the intermediate complex, leading to recruitment of ubiquitin ligases and degradation of client proteins in the proteasome.5, 6

Ansamycin antibiotics such as geldanamycin (GM) are natural products that bind to the N-terminal ATP/ADP binding pocket of HSP90.5, 6, 7, 8 Exposure of cells to these compounds induces the degradation of a range of HSP90 clients and results in cell cycle arrest followed in some cases by apoptotic cell death.9 The GM derivative 17-allylaminogeldanamycin (17-AAG) was the first HSP90 inhibitor to enter clinical trials. The drug is well tolerated, despite the fact that it simultaneously targets many intracellular signaling proteins.10, 11

Although 17-AAG was found to be highly potent in most tumor cell lines examined in vitro, and in a variety of tumors in vivo,2, 12, 13, 14 we also noticed that 17-AAG showed limited longevity of its effects on target proteins and lost antiproliferative activity precipitously under conditions of brief cellular exposure. In a quest for ansamycin drugs with properties differing from 17-AAG, we synthesized over 400 novel semisynthetic analogues, including several dozen dimmers, and investigated their structure–activity relationships. Geldanamycin dimers were originally reported by Zheng et al.15 These early compounds were relatively weak Hsp90 inhibitors that primarily induced the degradation of the most sensitive Hsp90 client, HER-2. Since active Hsp90 is an obligate dimer and the 2 ATP binding sites are brought into close apposition during the conformationally driven ATPase cycle,16 we reasoned that 2 GM moieties separated by an optimized flexible linker might be able to engage both ATP sites simultaneously. In this case, the resulting avidity might render the binding to activated Hsp9017 operationally irreversible because both GM molecules would need to dissociate at the same instant. Here, we have characterized 2 highly active GM dimers, CF237 and CF483. As predicted from their structures, both drugs bind Hsp90 with high affinity. Furthermore, binding of CF237 to Hsp90 in cells was found to be extremely long-lived, and the compound locked Hsp90 into the nonproductive intermediate complex for days after drug withdrawal. By contrast, the biochemical and cellular effects of 17-AAG were transient. In cells, the advantages of stable binding to the target were revealed when cellular activities of monomeric and dimeric inhibitors were studied in a variety of tumor cells after continuous and brief exposure. In the continuous exposure paradigm, 17-AAG was more efficient than the dimers at causing degradation of client proteins, including HER family kinases, signal transducers involved in PI-3k/Akt and Raf/MEK signal transduction pathways and cell cycle regulators, leading ultimately to tumor cell death. In contrast, under the conditions of brief exposure CF237 and CF483 were more active in killing tumor cells. Dimer-treated cells displayed markedly reduced client protein levels for at least 48 hr after drug removal, but 17-AAG induced only transient suppression of the same clients and the pathways they control. Similarly, the potency of 17-AAG in growth inhibition and apoptosis assays was significantly reduced in a panel of tumor lines under brief exposure conditions, but that of CF237 and CF483 remained largely intact. These data indicate that the novel dimeric ansamycins CF237 and CF283 have distinct cellular properties compared to monomeric 17-AAG and may represent an alternative or complementary strategy to 17-AAG for cancer therapy. The dimer was also retained for longer in tumor xenografts and displayed superior antitumor activity in vivo.

Material and methods

Cell and reagents

Cell lines were obtained from Clonetics and ATCC unless otherwise indicated. Tumor cells used included BT474, MCF7 and MDA468 breast carcinoma, N87 stomach carcinoma, SKOV3 ovarian carcinoma, A431 epidermoid carcinoma, A549 and H460 lung carcinoma, HT29 colon carcinoma, U87 and SKMG3 (a gift from C. Thomas, University of Virginia) glioblastoma, K562 Bcr-Abl+ leukemia and HT1080 fibrosarcoma. Normal primary cells used were renal proximal epithelial cells (RPTEC), human microvascular endothelial cells (HMVEC), human lung fibroblast (HLF). Antibodies used were Hsp90, Hsp70, Hsp40, Cdc37 and Hop (Stressgen, CA), p23 (Alexis Biochemicals, CA). Phospho-HER2 (Y1248) and PI3-Kinase p85 (Upstate Biotechnology, NY), phospho-Raf (S259), phospho-44/42 MAP and phosphor-Rb (Cell Signaling Technology, MA), all the other antibodies were from Santa Cruz Biotechnology, CA.

Synthesis of Hsp90 inhibitors

17-AAG, CF237 and CF483 were synthesized from GM as previously described.18, 19

MTS assay

Cells were seeded in 96-well plates at 2,000 cells/well in a final volume of 200 μl overnight before the addition of increasing concentrations of 17-AAG, CF237 or CF483. For continuous exposure, the plates were incubated for 120 hr before assay. For brief exposure, plates were incubated for 24 hr then the wells were washed twice and reincubated in drug-free medium for 96 hr. Viable cell number was determined using the Celltiter-96 AQueous Cell Proliferation Assay (Promega). Percentage of viable cells = (A490 of treated sample/A490 untreated cells) × 100. The IC50 was defined as the concentration that reduced the viable cell number by 50%.

Flow cytometric analysis

For apoptotic study, treated cells were trypsinized and added to 96-well round-bottomed plate. Cell pellet was washed in BA buffer (0.2% BSA and 0.2% NaN3) and fixed with 200 μl ice-cold 70% ethanol for 30 min on ice. Cells were then permeabilized in 200 μl/well of PBS-0.2% Tween 20 at RT for 15–30 min, followed by PI staining (equal amounts of propidium iodide solution (0.01% propidium iodide, 0.001% Triton X-100, 0.0037% EDTA in PBS) and RNase A solution (0.01% RNase A in PBS)). The samples were analyzed using a FACSCalibur™ flow cytometer. Cells in the sub-G0/G1 area were considered as dead cells. For protein degradation, cells in 24-well plate were collected as described above and incubated with 100 μl PE-labeled Anti HER-2/neu or anti IGF-1R antibody diluted 1:20 in BA buffer for 15–30 min at RT. Anti-KLH antibody were used as background controls. Cells were analyzed in 500 μl BA buffer on a FACSCalibur™ flow cytometer.

Coimmunoprecipitation and Western blotting

Cell lysates were prepared in lysis buffer (20 mM HEPES, pH 7.3, 1 mM EDTA, 5 mM MgCl2, 100 mM KCl). Antibody to Hsp90 was added to the cell lysate (supernatant) and incubated with rotation for 1 hr at 4°C. Protein-A Sepharose beads (50 μl) (Zymed) was then added and incubated with rotation for another hour at 4°C. Bound beads were pelleted and washed 3 times in lysis buffer. SDS-sample buffer was added to the beads and the mixtures were heated for 5 min at 95°C. Samples were analyzed by SDS-PAGE. Western blotting was performed as described.20

Measurement of Hsp90-bound 17-AAG or CF237

BT474 cell lysate was prepared as described previously. Equal amount of proteins were immunoprecipitated with an excess of anti-Hsp90 antibody and Protein A beads. Drugs associated with Hsp90 were extracted with 2 × 800 μl of acetonitrile. The mixture was vortexed and sonicated for 5 min before centrifugation at 20,000g for 10 min. The organic supernatant was evaporated and reconstituted with 130 μl mobile phase. A standardized HPLC assay was performed on an Agilent model 1100 liquid chromatography system. 17-AAG and CF-237 were separated on an Agilent Zorbax SB300 C18 column (4.6 × 150 mm2, 3.5 μm) with Phenomenex Securityguard system, containing a C18 cartridge. The total run time was 25 min.

Measurement of drug-induced Hsp90 oligomerization

BT474 cell lysates were made with co-ip lysis buffer by douncing 15× with Potter-Elvejhm homogenizer. Five micrograms of protein was run on 4–12% 10-well Tris/Glycine ZOOM™ gel 1.5 mm (Invitrogen Cat. EC6038). The gel was developed using SilverXpress Silver Staining Kit (Invitrogen Cat. LC6100).

Distribution of 17-AAG and CF-237 in tumor and normal tissue

The tissue distribution of 17-AAG and CF237 following a single intraperitoneal administration was evaluated in nude mice bearing A549 tumors. Mice were administered with solutions of 17-AAG formulated in 10% DMSO, 0.05% Tween 80 in water or CF237 in a 10% aqueous solution of PEG400, ethanol, Tween 80 (60/30/10 v/v). Both compounds were administered at a dose of 50 mg/kg. At 1, 2, 4, 24 and 48 hr after dosing, small intestine and tumor were collected and frozen at −80°C prior to analysis. Drug concentrations were determined in small intestine and tumor using a standardized HPLC-UV assay. The assay limit of detection was 0.3 μg/ml for 17-AAG and CF237.

Antitumor activity

Animal care was carried out under institutional guidelines for Laboratory Animal Research (ILAR). A549 tumor-bearing mice were randomized into groups of 7 for study. 17-AAG and CF237 were administered intraperitoneally at 60 mg/kg and 40 mg/kg, respectively, to groups of 7 animals with established tumors (∼50 mm3). Animals were dosed 3 days/week (Monday, Wednesday and Friday) for 4 weeks. Tumor dimensions were measured using calipers and tumor volumes are calculated using the equation 0.5 (l × w2), where l and w refer to the larger and smaller dimensions collected at each measurement.


17-AAG exhibits higher potency than ansamycin dimers under continuous exposure conditions

We have synthesized ∼400 semisynthetic geldanamycin derivatives, including several dozen dimeric ansamycin compounds that contain 2 GM pharmacophores connected at the 17 position with distinct linkers. The potency of some of these compounds in a HER-2 degradation assay and the structures of the linkers are illustrated in Table I. Some linkers, such as those used in compounds CF488, CF489, CF670, CF679 and CF708, precluded potent Hsp90 inhibitory activity (IC50 > 400 nM) whereas others, such as those used in compounds CF208, CF237, CF483, CF508 and CF568 produced very potent compounds (IC50 < 20 nM). Interestingly, when one GM in any of several active dimers was replaced with a moiety other than the ansamycin pharmacophore—such as a steroid hormone—some or all Hsp90 inhibitory activity was lost, suggesting that the second GM molecule increased the overall binding activity of the linked compounds (Table I). Some early dimers with simple linkers, such as CF18 and CF24, caused efficient HER-2 degradation but had little effect on less sensitive clients as reported previously,15 but the highly active families, exemplified here by CF237 and CF483, displayed potent effects on a range of client proteins and malignant cells. These drugs were compared to 17-AAG in classical growth inhibition/cytotoxicity assays under 2 conditions of drug exposure. When the compounds were left in contact with the cells for 5 days, 17-AAG was 5- to 25-fold more potent than CF237 and CF483 in most of the tumor lines tested (Fig. 1a). The IC50 for 17-AAG was between 0.4 and 70 nM, while that for CF237 and CF483 fell in the range of 7–400 nM. Western blot analysis of the time-course and dose-response of client degradation in BT474 cells revealed that Hsp90 clients, such as HER-2, AKT, ERK and CDK4, were more sensitive to 17-AAG than to CF237. For example, in the time-course experiment, disappearance of HER-2, CDK4 and phosphorylated ERK by 17-AAG could be detected within 3 hr whereas this was not usually seen until 16 hr in CF237 treated samples (Fig. 1b). Furthermore, in the dose-response assay, 0.1 μM 17-AAG exhibited a more profound effect on phosphorylated HER-2 (pHER-2) and CDK4 than did CF237 at the same concentration. 17-AAG at 0.3 μM effectively depleted both Raf-1 and AKT while CF237 was only partially active, although the difference between the 2 compounds diminished when the concentration was increased to 1 μM (Fig. 1c). As expected from its greater pharmacodynamic potency, 17-AAG also induced G2/M cell cycle arrest (after 24 hr) and subsequent apoptosis (after 72 hr) more efficiently than the ansamycin dimer (Fig. 1d). Thus, under conditions of continuous drug exposure, potent monomeric ansamycins such as 17-AAG are more active Hsp90 inhibitors than their dimeric cousins, including CF237 and CF483.

Figure 1.

17-AAG exhibits potent cytotoxicity and induces client degradation and apoptosis more potently than dimeric ansamycins under continuous exposure conditions. (a) Cells were plated at 10–20% confluence and treated with increasing concentrations of 17-AAG, CF237 or CF483. Five days later, the surviving fraction was estimated by MTS assay. IC50: concentration of drug required to inhibit cell growth by 50%. (bd) BT474 cells were treated with DMSO, 0.3 μM 17-AAG or CF237 for the indicated times (b), or with indicated concentrations for 24 hr (c). Cells were lysed and the expression of HSP90 client proteins was measured by Western blot. (d) A549 cells were plated at 50–60% confluence and treated with DMSO, 17-AAG or CF237 at the indicated concentrations for 24 or 72 hr, stained with propidium iodide and analyzed by flow cytometry.

Table I. Linker Structure and Potency of Geldanamycin Dimers and Related Compounds
CF numberLinkerSecond moietyHER-2 degradation IC50 (nM)
  1. Hybrid ansamycins containing a GM pharmacophore linked at the 17-position to either another GM or a distinct moiety were incubated with MCF-7 cells for 16 hr and HER-2 expression was measured by direct immunofluorescence and FACS analysis.

508Trans[BOND]CH2C[DOUBLE BOND]CHCH2Geldanamycin19
125Trans[BOND]CH2C[DOUBLE BOND]CHCH2Steroid275

Ansamycin dimers exhibit higher potency than 17-AAG under brief exposure conditions

Since activated Hsp90 functions as dimer in the multichaperone complex, it seemed plausible that dimeric ansamycins might bind Hsp90 more tightly than their monomeric counterparts. Moreover, simultaneous occupation of both binding sites by 2 linked monomers could also radically reduce dissociation rates, as both arms must dissociate concomitantly from the binding pockets. Based on this, we sought to determine whether dimeric ansamycins would have more persistent effects on tumor cells under conditions in which the compound was removed from the culture medium after brief exposure. This paradigm is arguably more pharmacologically relevant, because following intravenous administration of ansamycins in animals and humans, serum levels peak within minutes and subsequently decrease rapidly to nanomolar levels as the drug is metabolized and excreted.21, 22 BT474 cells were treated with either 17-AAG or CF237 for 24 hr, the compounds were removed and the cells were cultured continuously in drug-free medium for up to 120 hr. Apoptotic cell death was measured by flow cytometry at daily intervals. As shown in Figure 2a, 1 μM 17-AAG caused a temporary increase in the G2 population 24 hr after treatment. Withdrawal of the compound allowed the cells to resume normal cell cycle progression. Although ∼10% of cells were found in the sub-G0 apoptotic fraction after 120 hr, the majority of the population survived unscathed. By contrast, CF237 caused a similar change in cell cycle distribution at 24 hr, but the arrested cells subsequently progressed into massive apoptosis (Fig. 2a). More than 80% cells were found in the sub-G0 population at 120 hr and no further cell proliferation was seen. To investigate whether this was an isolated phenomenon, we performed similar experiments in a variety of tumor lines, including breast carcinoma (BT474, MDA-MD-468), stomach carcinoma (N87), ovarian carcinoma (SKOV-3), glioblastoma (U87) and cervical carcinoma (A431) and in normal human endothelial (HMVEC), fibroblast (HLF) and epithelial (RPTEC) cultures. Consistent with the results in BT474, massive cell death was observed in all of the tumor lines, but not the normal cells, when they were briefly exposed to CF237 (Fig. 2b). On the other hand, 17-AAG under the same conditions failed to induce much cell death in either tumor or normal cultures (Fig. 2b). Taken together, these data suggest that in situations where the exposure of cells to drug is of limited duration, dimeric ansamycins might be expected to induce selective apoptosis of tumor cells more effectively than 17-AAG.

Figure 2.

Dimeric ansamycins induce apoptosis more potently than 17-AAG under brief exposure conditions. (a) BT474 cells were plated in 24 well plates at 50–60% confluence and treated with DMSO, 17-AAG or CF237 (1 μM). After 24 hr, the monolayers were washed twice, incubated in drug-free medium for a further 72–96 hr, stained with propidium iodide and analyzed by flow cytometry. The FACS histograms are shown in the left panel. The proportion of cells in Sub-G0 (apoptotic) G1/S and G2/M fractions were quantified (right panel). (b) A panel of tumor cell lines and normal human primary cells were treated as above and assayed by PI staining 96 hr after drug removal. The normal cells studied were kidney epithelium (RPTEC), capillary endothelium (HMVEC) and lung fibroblasts (HLF).

To determine whether ansamycin dimers were more effective than monomers after brief exposure across the pharmacologic range of drug concentrations, MTS assays were performed in a panel of tumor cells under both conditions. As would be expected from the earlier data, the IC50 of 17-AAG was markedly elevated in all 4 cell lines when drug was washed out after 1 day vs. continuous exposure. As shown in Figure 3a, the IC50s of CF237 or CF483 changed to a much lesser degree (1- to 6-fold) than that of 17-AAG (60- to 250-fold) when the conditions were made more stringent. The data from these experiments confirmed that dimeric Hsp90 inhibitors, unlike 17-AAG, retain substantial potency in terms of cell growth inhibition and induction of apoptosis even after drug withdrawal, suggesting that these second generation compounds interact with the target Hsp90 in a different way than ansamycin monomers.

Figure 3.

Dimeric ansamycins, but not 17-AAG, retain potent cytotoxic activity and pharmacodynamic activity under brief exposure conditions. (a) Four malignant cell lines were plated at 10–20% confluence and treated with increasing concentrations of 17-AAG, CF237 or CF483. The drugs were either left on for 5 days or were removed after 24 hr, followed by further 96 hr incubation in drug-free medium. The surviving fraction was estimated by MTS assay. The IC50 values in nM under the 2 conditions, and the ratio of the two, are shown. (b) MCF-7 cells were treated with 1 μM 17-AAG, DMAG, CF237 or CF483 for 24 hr. The cells were analyzed immediately or washed and incubated for another 24–48 hr in drug-free medium. Cell surface tyrosine kinase levels were detected with directly PE-labeled anti-HER-2 or anti-IGF-1R antibodies and quantified by flow cytometry. (c) BT474 cells were treated with DMSO, 1 μM 17-AAG or CF237 for 24 hr. The cells were then washed and incubated in drug-free medium for another 48 hr. Cells (including both floating and adherent cells) were collected at the indicated times, lysed and the expression of HSP90 client proteins was measured by Western blot. PI-3Kp85 (a nonclient protein) was used as a loading control.

Dimeric ansamycins have longer lived pharmacodynamic activity

To elucidate whether the differential effects of monomeric vs. dimeric ansamycins on cell growth and survival were linked to their regulation of Hsp90 clients, we tested the kinetics of client degradation after brief drug exposure by 2 methods. As shown in Figure 3b, FACS analysis of MCF-7 cells with PE-labeled anti-HER-2 and anti-IGF-1R antibodies demonstrated that 2 monomeric compounds, 17-AAG and another clinical-stage ansamycin, 17-DMAG,23 and 2 dimeric compounds, CF237 and CF483, caused greater than 80% degradation of both receptors after 24 hr of treatment. Washout of 17-AAG or 17-DMAG resulted in virtually complete recovery of HER-2 and IGF-1R, but CF237 and CF483 sustained extremely low levels of both clients even after the compounds had been removed for 48 hr. To extend the observation to intracellular clients and another cell line, Western blots were performed on monomer- and dimer-treated BT474 cells. As shown in Figure 3c, 24-hr exposure to either 17-AAG or CF237 caused a significant decrease of almost all of the clients examined, including HER-2, EGFR, Akt and Raf-1 as well as downstream targets regulated by them, including cyclin D and phosphorylated Rb (pRb). Forty eight hours after drug removal, most of these proteins had recovered to baseline levels in 17-AAG treated samples, but washout of CF237 did not permit reappearance of any clients. Notably, the ultimate downstream target of Hsp90-dependent cell cycle regulation, pRb, actually decreased in expression at the later time point (Fig. 3c). Thus, it appeared that the superior antitumor activity of dimeric Hsp90 inhibitors after brief drug exposure and washout was due to sustained effects of the dimers on Hsp90 function, perhaps reflecting sustained interaction with Hsp90 itself.

Ansamycin dimers stabilize the intermediate complex

Ansamycin Hsp90 inhibitors compete with ATP for binding to the N-terminal ATP-binding site of the chaperone, thereby preventing the ATP-dependent conformational change required for transition to the ‘mature’ p23 complex, effectively locking the ‘intermediate’ HOP complex.4 The intermediate complex cannot complete client folding, and stabilization of this complex apparently recruits ubiquitin ligases, leading to the proteasomal degradation of the client.24, 25 As shown in Figure 4a, both classes of Hsp90 inhibitor had little effect on the total cellular levels of the cochaperone members of Hsp90 multichaperone complexes, although Hsp90 levels were somewhat elevated. By contrast, treatment with Hsp90 inhibitors had a profound effect on the association of Hsp90 with its cochaperones, as evidenced by coimmunoprecitation experiments (Fig. 4b). In untreated cells, Hsp90 was primarily found in mature complexes, associated with p23 and cdc37 (Lane 1). As shown in Lanes 2 and 4, after 24 hr exposure to either 17-AAG or CF237, p23/cdc37-containing complexes were largely absent and intermediate complexes containing HOP, Hsp40 and Hsp70 predominated, as reported previously.12 After drug withdrawal, however, the fate of Hsp90 complexes diverged in cells treated with 17-AAG vs. CF237. Forty eight hours after drug washout, cochaperone association of Hsp90 in 17-AAG-treated cells resembled that of control cultures (compare Lanes 1 and 3). In stark contrast, cells exposed to CF237 contained only intermediate complexes, as they had prior to drug removal (compare lanes 4 and 5), suggesting that the biochemical influence of CF237 persisted long after the drug had been removed from the extracellular environment.

Figure 4.

Dimeric ansamycins have longer lived biochemical activity. (a) BT474 cells were treated and collected as described in Figure 3c and expression of HSP90 and cochaperones were measured by Western blot. (b) one milligram of protein from cell lysates treated as above was immunoprecipitated with anti-Hsp90 antibody and then blotted with antibodies for cochaperones as indicated.

The finding that drug-induced stable HOP complexes apparently persisted in dimer-treated cells after removal of ansamycin dimers suggested that these compounds interact with the target in a particularly stable manner. To measure the stability of the drug-Hsp90 interaction, it was necessary to perform the analysis in cell lysates, because of the physiological interactions with cochaperones that are required to bring the 2 ATP binding sites into close apposition during the conformationally driven ATPase cycle.16, 26 Cells were lysed at various times after the addition and removal of drug, the Hsp90 was immunoprecipitated and Hsp90-bound drug levels were measured by HPLC. As shown in Figure 5a, readily detectable levels of both 17-AAG and CF237 were bound to tumor cell-derived Hsp90 after 24 hr of incubation, but 1 or 2 days after drug withdrawal, only CF237 remained associated with the chaperone; 17-AAG was not detected at either time point. Taken together, these biochemical data strongly indicate that dimeric ansamycins bind dimeric activated Hsp90 with high avidity, resulting in prevention of ATP binding and inhibition of Hsp90 client folding activity that persists long after removal of the drug. The apparent avidity effect could arise either by intermolecular bridging of 2 Hsp90 dimers by a single dimeric ansamycin, or by intramolecular bridging between the 2 ATP sites of a single Hsp90 dimer. In either case, the function of bound Hsp90 would be blocked, because occupation of either ATP site by geldanamycin compromises the activity of the entire dimer and leads to client protein degradation. Reasoning that intermolecular bridging should lead to crosslinking of multiple Hsp90 dimers, we looked for Hsp90 species with higher apparent molecular weight (Mr) on nondenaturing gels of lysates from tumor cells treated with 17-AAG or CF237. As shown in Figure 5b, although several species of Hsp90 oligomer were observed, there was no clear difference between geldanamycin monomer- and dimer-treated cells, suggesting that CF237 does not efficiently bridge separate Hsp90 dimers in vivo. In addition, we were unable to generate Hsp90 oligomers in vitro by incubating recombinant Hsp90 with 17-AAG or CF237 at various ratios (data not shown).

Figure 5.

Dimeric ansamycins have longer lived binding activity, are selectively retained in tumors and possess improved antitumor activity. BT474 cells were treated and collected as described in Figure 3c. (a) Equal amounts of protein were precipitated with anti-Hsp90 antibody and brought down with protein A beads. The beads were washed thoroughly and bound drug was eluted and quantified by HPLC. (b) Equal amounts of cell lysate were loaded on nondenaturing gels and examined for the formation of Hsp90 oligomers by silver staining. (c) A549 tumor-bearing mice were dosed with 50 mg/kg 17-AAG or CF237. Drug concentrations in tumor and intestine were measured by HPLC. (d) A549 tumor bearing mice were treated with vehicle control, 60 mg/kg 17-AAG or 40 mg/kg CF237 IP at ×3/week for 4 weeks.

Ansamycin dimers show enhanced localization and antitumor activity in xenografts

To determine whether the differential biological effects of monomeric and dimeric ansamycins in vitro were reflected in vivo, we compared the pharmacokinetics and therapeutic efficacy of 17-AAG and CF237 in a lung cancer xenograft model. We first examined the level of 17-AAG and CF237 in tumor and a normal tissue, intestine, over 48 hr. As shown in Figure 5c, although both compounds were initially found at high levels in the tumors, the concentration of CF237 remained elevated for the duration of the experiment, whereas the levels of 17-AAG decreased substantially within 24 hr. In the gut, both drugs reached high levels in the first few hours but neither could be found by 6 hr post injection. These data suggest that both drugs are readily bioavailable and selectively retained in solid tumors, with intratumoral CF237 concentrations remaining elevated for longer than those of 17-AAG. Consistent with this finding, CF237 inhibited tumor growth by 63% on a thrice-weekly dosing schedule, while 17-AAG only produced 36% inhibition (Fig. 5d), indicating that the enhanced binding stability and sustained retention in tumors of dimeric ansamycins may confer improved antitumor activity. Both drugs were well tolerated at this dose and schedule. A somewhat lower dose of CF237 than 17-AAG was used, primarily because the solubility of the dimer in the early formulation used here was limiting.


It is becoming increasingly clear that targeting a single oncoprotein in cancer therapy may not be sufficient to obtain strong efficacy.27, 28, 29 This has brought Hsp90 center stage among experimental therapeutics because inhibition of Hsp90 causes simultaneous degradation of multiple oncogenic proteins and affects a number of signal transduction pathways that are important for cancer cell proliferation and survival.1, 2, 3 The semisynthetic geldanamycin derivative 17-AAG has completed Phase I clinical trials with encouraging results.10, 30 The compound is well tolerated and several long-lasting disease stabilizations were noted in heavily pretreated patients.10, 30 Despite this, 17-AAG has some limitations among ansamycin drugs. Although extremely potent (similar to the parent compound) in most cell lines,12, 13, 14 17-AAG, unlike geldanamycin, has limited activity against some clinically important subsets of tumor cells, including those defective in Rb signaling31, 32 and those overexpressing Bcl-2.33 Antiapoptotic Bcl-2 family proteins serve to retard commitment to apoptosis in stressed cells, raising the possibility that 17-AAG fails to inhibit Hsp90 for long enough to allow Bcl-2-overexpressing tumor cells to commit to apoptosis, even when continuously present. In that case, it would be predicted that the drug would rapidly lose apparent potency in all cells when the duration of drug exposure is limited. The data in Figures 2 and 3 show that the effects of 17-AAG on client protein expression decay rapidly after drug washout and the cytotoxic potency of the drug is markedly attenuated under brief exposure conditions relative to when the compound is present throughout the 5 day assay. One explanation for these results is that 17-AAG rapidly dissociates from its target, diffuses out of the cell and loses potency if it cannot be replenished from the extracellular environment. Since activated Hsp90 functions as dimer in the multichaperone complex,16 it seemed plausible that dimeric ansamycins might display radically reduced dissociation rates, as both arms would have to dissociate concomitantly from the binding pockets to release bound drug, assuming that either both ATP sites of a single dimer or one ATP site from 2 individual Hsp90 dimers could be bound by a single molecule of a dimeric drug. As predicted from this model, 2 dimeric Hsp90 inhibitors, CF237 and CF483, appeared to interact much more stably with Hsp90 than did 17-AAG. When cells were exposed to monomeric or dimeric ansamycins for 24 hr prior to drug washout and assays were performed 48 hr later, the dimeric compounds (but not 17-AAG) were found to (i) remain physically associated with the target, (ii) stabilize the nonproductive HOP complex, (iii) prevent reappearance of client proteins and (iv) retain potent growth inhibitory and proapoptotic activity. Client protein levels also rebounded quickly in cells treated with another ansamycin monomer, 17-DMAG, suggesting that 17-AAG is not unique in this regard. A recent study by Guo et al.34 indicates that monomeric ansamycins such as 17-AAG are reduced by the enzyme NQO1 to dihydro-17-AAG in cells and that this metabolite has increased binding affinity and intracellular retention. It is possible that NQO1-mediated activation of 17-AAG might underlie its superior activity at shorter exposure times. Although the structure of CF237 and CF483 does not preclude their similar activation, we have not observed diminished activity in NQO1-deficient lines. The effect of this metabolism may be masked by the increased affinity/avidity and consequent heightened intracellular retention of the dimeric compounds because of their extremely stable binding mode.

Not all ansamycin dimers were potent Hsp90 inhibitors. As shown in Table I, dimers with linkers of various lengths, structure and flexibility were synthesized, but no obvious features conferring potency were discernable. Among the compounds shown in Table I, CF18 and CF24 were originally described by Zheng et al.15 These dimers were quite potent against the sensitive client HER-2 (IC50 ∼ 60 nM) but had little effect on other Hsp90 clients. Since the dynamic, activated Hsp90 found in multichaperone complexes in cells cannot by crystallized by current methods, it was not possible to design an optimal linker by structural approaches or to confirm divalent binding of CF237 to a single activated Hsp90 dimer. However, the observation that the second moiety attached to a GM-linker intermediate profoundly affected potency is provocative. When smaller, cell permeable entities such as testosterone (CF124), estradiol (CF125) and raloxifene (CF691, CF692) were substituted for the second GM, activity in the HER-2 degradation assay decreased around 15-fold, indicating that the second GM most probably interacts with a high-affinity binding site on Hsp90.

Highly active dimers were superior to their monomeric counterparts under conditions of limited drug exposure, a situation that pertains in clinical cancer therapy, but 17-AAG may also possess some significant advantages in vivo. Monomers are smaller and more compact, likely increasing their ‘on’ rates when binding to Hsp90, which may explain the more rapid onset of client protein degradation and superior potency of monomers under conditions of continuous exposure (Figs. 1 and 2a). Furthermore, the greater mass of dimeric compounds (∼1,200 kDa) would likely retard their permeation through the tightly packed cells and fibrous stroma that characterize solid tumor masses.35 Similarly, the slow dissociation of dimers from their target would not favor penetration of the drug to distant extravascular sites. Thus, dimeric Hsp90 inhibitors might be most efficacious in leukemia, where limitations of drug mobility are less critical. Nonetheless, geldanamycin dimers are active against solid tumor xenografts. We have recently reported that a related dimeric ansamycin, termed EC5, shows good activity against head and neck cancer and retained potent activity in one particular cell line, JHU12, that was resistant to 17-AAG because of defects in Rb signaling.36

In this report, we have compared the biochemical, pharmacodynamic and cytotoxic properties of monomeric and dimeric ansamycin Hsp90 inhibitors in vitro and in vivo. The dimers have improved activity under conditions of brief drug exposure, probably because they form extremely stable, high avidity binding interactions with their dimeric target protein. To our knowledge, this is the first report of optimized bivalent drugs for cancer therapy, but a growing body of work suggests that bivalent and multivalent drugs can offer significant advantages over their monovalent analogues in other areas of medicine. Griffin et al. described optimization of a series of vancomycin dimers that show improved biological activity against susceptible and drug-resistant gram-positive bacteria.37 Both potency and duration of action were enhanced relative to the parent monomer.38 Although the most advanced studies with dimeric drugs involve antibiotics, bivalent modulators of mammalian targets have been reported, most notably longer-acting agonists of the dimerizing G-protein-coupled β2-adrenergic receptor.38 The binding of high-affinity dimeric Hsp90 inhibitors to their target may be essentially irreversible, but these compounds do not form covalent adducts with the Hsp90. Conventional, chemically reactive irreversible inhibitors have several inherent drawbacks, including poor solubility due to their reactivity with water,39 poor tolerability due to their reactivity with nontarget proteins and clinical management problems because their effects cannot be rapidly reversed by cessation of dosing. The approach described here, which may be applicable to other dimeric or oligomeric drug targets, does not suffer from the first 2 of these limitations. Indeed, the dimeric ansamycin was well tolerated at effective doses in xenograft studies.

In conclusion, we have shown that monomer and dimer Hsp90 inhibitors of the same structural class interact differently with the target and each possesses theoretical advantages and disadvantages for cancer treatment. The preferred approach will have to be determined clinically, but it is attractive to speculate that combination therapy with both classes of inhibitor might provide the most effective method of attacking the diversity of tumor cell populations.