|Home | About | Journals | Submit | Contact Us | Français|
Despite tremendous advances in basic and clinical oncology, the prognosis for most patients with neoplastic diseases is still dismal. Therefore, new therapeutic strategies must be sought. A novel drug for the treatment of malignant diseases should have a mechanism of action different from those known for established oncological therapeutics. In addition, it is desirable to attempt to relate the antitumor activity of a candidate compound with molecular effects on proteins relevant for the pathogenesis of malignant diseases.
Geldanamycin (GA; NSC 122750) has the potential to fulfill the aforementioned criteria. It was first purified in 1970 from the broth of Streptomyces hygroscopicus var geldanus var nova (DeBoer et al 1970). As a benzoquinone ansamycin (BA), it consists of a quinone ring and a hydrophobic ansa bridge (Fig 1; Rinehart and Shield 1976). The DNA sequences responsible for GA biosynthesis have been previously characterized (Allen and Ritchie 1994). The antineoplastic effect of GA was already noted in its first description (DeBoer et al 1970).
Molecular studies revealed the binding of GA to members of the heat shock protein 90 (Hsp90) family of molecular chaperones (Whitesell et al 1994). Interference with the function of these Hsps seems to be the major mechanism of action of GA.
The use of various model systems generated data on the in vitro and in vivo activity against malignant tumors. Table 1 summarizes these results. In Table 2, studies are listed that were undertaken to define the mechanism of action of GA.
Three concepts are of particular importance: (1) formation of free radicals, (2) assumed inhibition of tyrosine kinases, and (3) binding to and interference with the function of members of the Hsp90 family of proteins.
The fact that a quinone ring is part of the GA molecule led to the initiation of detailed studies on the generation of intracellular free radicals through redox cycling (Benchekroun et al 1994b). It was shown that the treatment of eukaryotic cells with GA indeed causes the formation of free radicals. However, this could only be observed at high concentrations of GA, eg, 100 μM, whereas the same investigators—as others later—found cytotoxic effects in the nanomolar range. It can be reasonably concluded that the antitumor activity of GA cannot be explained by the formation of free radicals.
Since GA inhibited the tyrosine kinase activity of v-src, it was classified as a tyrosine kinase inhibitor (Yamaki et al 1995). The assumption was that BAs directly act on sulfhydryl groups of v-src (Uehara et al 1989). Yet it was known that the tyrosine kinase activity of v-src could only be influenced by herbimycin A—an analog of GA—in vivo but not in vitro (Uehara et al 1986). This suggested that the target molecule of GA is either downstream of v-src or an interacting protein. A detailed analysis of the course of GA effects showed that there is no correlation between its cytocidal activity and the late onset of inhibition of kinases of the src-family (Whitesell et al 1992).
All these findings could not explain the multitude of effects of BAs. Finally, through chemical cross-linking of GA to Sepharose beads, it was possible to precipitate the target molecule of GA. Further analysis identified it as Hsp90 (Whitesell et al 1994). The discovery that GA binds to Hsp90 was a decisive breakthrough in understanding the mechanism of action of GA and structurally or functionally related substances. For the first time, a molecular chaperone was identified as the target molecule of an anticancer drug.
In transformed cells, v-src forms a heterocomplex with Hsp90 (Mimnaugh et al 1995). Thus, the described reversion of the malignant phenotype (Uehara et al 1986) could be attributed to GA-induced dissolution of the Hsp90–v-src heterocomplex (Whitesell et al 1994). Some experimental findings implicate phosphoserine and phosphothreonine contents of Hsp90 as crucial for the maintenance of the Hsp90–v-src complex (Mimnaugh et al 1995).
Further detailed studies revealed that GA binds to the amino terminus of Hsp90 (Grenert et al 1997). Crystallographic experiments confirmed this and found a pocketlike conformation of the binding site (Prodromou et al 1997; Stebbins et al 1997; Roe et al 1999). It was demonstrated that this region has an adenosine triphosphatase activity that is inhibited by GA through competitive binding with adenosine triphosphate (Grenert et al 1997; Obermann et al 1998).
The amino terminal end of Hsp90 was also described as being able to bind unfolded peptides and to prevent their aggregation in a GA-sensitive manner (Young et al 1997). The charged linker region connecting the amino and carboxy terminal domains of Hsp90 modulates the interaction of GA with its binding site (Scheibel et al 1999).
Further experiments showed that GA also binds to the glucose-regulated protein 94 (Grp94), another member of the Hsp90 family located in the endoplasmic reticulum (Chavany et al 1996; Schulte et al 1999).
Hsp90 like other Hsps is named for its increased synthesis after heat shock that is contrary to the reduced synthesis of most cellular proteins under these conditions. Members of the Hsp90 family of molecular chaperones include Hsp90α (Hickey et al 1989), Hsp90β (Rebbe et al 1987), Grp94 (Little et al 1994), and Hsp75 (Chen et al 1996), also known as TNF receptor-associated protein-1 (TRAP-1) (Song et al 1995). So far no differential function has been assigned to the α and β subforms of Hsp90. Research on yeast cells shows that Hsp90 is essential (Picard et al 1990). Mutants with abrogated adenosine triphosphatase activity are not viable (Panaretou et al 1998). Since Hsp90 may represent 1% of the proteome even in the absence of heat shock, it must have functions beyond its role in the heat shock reaction. At least 3 functions of Hsp90 can be delineated. First, it is a molecular chaperone that prevents the aggregation and assists the refolding of proteins after various stresses, including heat shock (Nathan et al 1997). This finding is supported by experiments demonstrating the Hsp90-dependent renaturation of denatured enzymes (Bose et al 1996; Hartson et al 1996). In vitro, GA partially blocks protein renaturation (Schumacher et al 1996). Second, it exerts a “genomic buffer” function to suppress phenotypic traits that only become expressed after certain stresses (Rutherford and Lindquist 1998). Third, it stabilizes key regulatory proteins through the formation of heteroprotein complexes (Lindquist 1988) or keeps them in a defined functional state. Well-studied examples of the latter are the progesterone and the glucocorticosteroid receptors, which through interaction with Hsp90 and other accessory proteins are kept in a hormone-binding conformation (Sullivan and Toft 1993; Johnson and Toft 1995; Whitesell and Cook 1996). After hormone binding, Hsp90 is released from the complex.
Hsp90 is probably involved in either initiating or maintaining the transformed state: Hsp90 messenger RNA (mRNA) has been found to be overexpressed in human ovarian carcinoma cells, but not in normal ovarian cells or in those derived from benign ovarian tumors (Mileo et al 1990).
There is evidence that Hsp90 interacts with several proteins important to oncology. Interactions have been described for cdk4 (Chen et al 1997), cyclin B1 and Cdc2 (Kim et al 1999), heterotrimeric Gα subunits (Busconi et al 2000), telomerase (Holt et al 1999), and epidermal growth factor (EGF) and platelet-derived growth factor receptors (Sakagami et al 1999). In addition, it may influence fusion proteins, which are present only in certain neoplasias as evidenced from its effect on the bcr-abl protein found in most chronic myelogenous leukemias and some acute lymphatic leukemias (An et al 1999).
At least 4 proteins of major interest for oncology are being influenced by BAs: mutant p53, erbB2, raf-1, and focal adhesion kinase. Most cellular proteins are not affected by treatment with GA, as can be deduced from the comparison of radioactive methionine-labeled cell lysates before and after treatment (Ochel et al 1999).
As a transactivating protein, which acts as a tumor suppressor (Marx 1993), p53 is mutated or absent in approximately half of all human cancers and is the protein with the most commonly found molecular alteration in neoplasias (Nigro et al 1989; Levine 1992).
A role for Hsp90 in the pathophysiology of the malignant transformation was suggested by the fact that treatment of several breast cancer, prostate cancer, and leukemia cell lines with GA results in the destabilization of mutated p53 (Blagosklonny et al 1995). This was the first evidence of a specific pharmacological modulation of mutated p53. In this model system no influence on wildtype p53 levels was found, although in a different experimental setting accumulation of wildtype p53 accompanied by cell-cycle arrest was detected (McIlwrath et al 1996). These findings have been extended by studies using rabbit reticulocyte lysate and wheat germ extract. In these studies, the presence of Hsp90 was found to be necessary for p53 to acquire a mutated conformation (Blagosklonny et al 1996). Conversely, in vitro translation of mutated p53 in wheat germ extract, known to be functionally Hsp90 deficient, yielded p53 proteins only with wildtype but not with mutant conformation. This led to the hypothesis that mutant p53 stabilization, found in malignant cells, is not due to stabilization of the protein per se, but is caused by the formation of heterocomplexes of p53 with other proteins, such as Hsp90 (Blagosklonny 1997). The in vivo existence of a Hsp90–mutated p53 complex, which in some cases is located in the cytoplasm, has been demonstrated (Sepehrnia et al 1996). Further studies revealed that other chaperones, such as Hsp70 or Hop, can be coimmunoprecipitated with p53 and that these associations may be modulated by GA (Whitesell et al 1998). In some experiments, GA treatment induced the disruption of the Hsp90–mutated p53 complex (Whitesell et al 1998; Nagata et al 1999). In other studies, this was not found (Dasgupta and Momand 1997), pointing to different roles for this interaction in the pathophysiology of malignant diseases. All investigations showed a rapid destabilization of mutated p53.
It was already known that p53 may become ubiquitinated (Kubbutat and Vousden 1997). The ubiquitination increases with GA treatment (Nagata et al 1999), and this in turn stimulates p53 degradation by the proteasome (Whitesell et al 1997).
In summary, GA is the prototype of drugs that are able to selectively destabilize mutated p53. No restoration of wildtype transcriptional activity was detected under these conditions.
The product of the ERBB2 gene, also known as NEU or HER-2, is also a target of GA. The gene codes for the mRNA of a transmembrane protein, which belongs to the EGF receptor family (Bargmann et al 1986). ErbB2 is a receptor subunit that homodimerizes or heterodimerizes with other members of the EGF-receptor family and thus creates receptors of different function and ligand affinity (Goldman et al 1990). The significance for oncology stems from the finding that overexpression of this protein mainly by means of gene amplification (King et al 1985) in breast or ovarian cancers (Slamon et al 1989) is associated with a dismal prognosis (Slamon et al 1987). Hence, therapeutic, pharmacological modulation of erbB2 is desirable.
The original observation that erbB2 overexpressing SKBr-3 breast cancer cells treated with BAs exhibit a rapid and profound reduction of erbB2 protein was attributed to enhanced degradation (Miller et al 1994a). ErbB2 mRNA levels were found to be unchanged. A first hint of the underlying molecular mechanism came from experiments with photoaffinity-labeled ansamycins, which were shown to bind to an approximately 100-kDa protein (Miller et al 1994b). Competition of binding with GA proved that this was the target molecule of the BAs. By coimmunoprecipitation and Western blotting experiments, it was identified as Grp94 (Chavany et al 1996). Grp94 is a homologue of Hsp90 that is located in the endoplasmic reticulum (Little et al 1994). Treatment with GA results in the disruption of the Grp94-erbB2 complex, followed by polyubiquitination of the transmembrane protein and proteasomal degradation (Chavany et al 1996; Mimnaugh et al 1996; Hartmann et al 1997). A relation between BA structure and their respective erbB2 depletion activity was described (Schnur et al 1995a). ErbB2 depletion by GA could also be demonstrated in vivo in heterotransplanted malignant tumors (Schnur et al 1995b).
Raf-1 is part of a conserved signal transduction pathway that connects cytosolic and transmembrane tyrosine kinases with mitogen-activated protein kinases (Magnuson et al 1994). It is activated and recruited to the internal side of the plasma membrane through binding to the small, guanosine 5′-triphosphate binding protein ras (Marais et al 1995). Ras mutations leading to constitutive activation of this signal transduction pathway have been identified in different malignant tumors, but are especially prevalent in pancreatic cancers (Almoguera et al 1988). Raf-1 may also be oncogenically activated by mutation (Heidecker et al 1990). A physical interaction of raf-1 and Hsp90 in the form of a heterocomplex together with other accessory proteins was identified in vitro and in vivo (Stancato et al 1993). The treatment of MCF-7 breast cancer cells with GA results in the rapid destabilization of raf-1, mediated by the proteasome (Schulte et al 1997), and the dissolution of the association of raf-1 with ras (Schulte et al 1995). Upstream and downstream effectors of this signal transduction pathway are not affected by GA (Schulte et al 1996). Further analysis revealed that raf-1 concentrations declined concomitantly in different compartments of the cell, confirming the hypothesis that the association of raf-1 with Hsp90 may have significance for the transport of raf-1 to the cell membrane (Schulte et al 1995). These findings were confirmed with other models, although the disruption of the Hsp90–raf-1 complex was not always seen (Stancato et al 1997). The treatment of malignant cells with GA offers the option to abrogate the constitutive activation of raf-1 caused by ras mutations.
Focal adhesion kinase (fak) is a tyrosine kinase located predominantly in focal adhesions, ie, the cellular sites of attachment to the extracellular matrix (Schaller et al 1992). Fak is expressed in all human tissues. Fak mRNA and protein expression show a positive correlation with tumor invasiveness (Owens et al 1995). In metastases, more fak can be found than in premalignant lesions. They in turn contain more fak than homologous benign cells. This was described for colon, breast, and thyroid cancers and certain sarcomas (Weiner et al 1994). Fak−/− cells produced by gene knockout exhibit an increased number of focal adhesions, implicating fak in the turnover of these subcellular structures (Ilic et al 1995). Initially, it was not clear whether fak was a tumor-repressing or tumor-promoting protein, since no malignant transformation of cells was found in the usual tests. Recently, however, it was reported that activated fak is able to transform Madin Darby Canine Kidney (MDCK) cells (Frisch et al 1996). Integrin-dependent signal transduction via fak suppresses apoptosis (Frisch and Francis 1994). Taken together, fak is a promising target for pharmacological intervention.
Treatment with GA results in a time- and dose-dependent degradation of fak protein, which is mediated by the proteasome (Ochel et al 1999). Currently, GA is the only substance with the pharmacological ability to interdict fak function.
The relevance of the described molecular effects of GA to its cytotoxic property, at least in vitro, has recently been described. Twenty-eight different derivatives of GA were tested for their cytotoxicity. Loss of mutated p53, erbB2, and raf-1 occurred concomitantly and was only observed with those derivatives that displayed antiproliferative activity (An et al 1997).
MCF-7 breast cancer cells resistant to doxorubicin (MCF-7/ADRR) are cross-resistant to GA (Benchekroun et al 1994b). For the same cytotoxic effect, higher concentrations of doxorubicin or GA had to be used in resistant vs wildtype cells. Since verapamil, known to bind to and inhibit the multidrug resistance (MDR) pump, enhanced GA cytotoxicity, the conclusion was that GA is a substrate of the MDR pump and that overexpression of MDR may be a mechanism of GA resistance. In addition, GA-induced generation of free radicals was reduced in MCF-7/ADRR as opposed to the parental cells.
However, further studies with cells selected for GA resistance (MCF-7/GAR) showed that there is no difference in GA efflux between wildtype and MCF-7/GAR cells (Benchekroun et al 1994a). Verapamil had no cytotoxicity-enhancing effect, although MDR expression was twice as high in the resistant cells. In addition, free radical formation by GA was found not to contribute to this resistance.
Interestingly, Hsp90β can be found in direct physical association with the MDR protein that is only present in chemotherapy-resistant but not in wildtype cells (Bertram et al 1996).
Extensive pharmacological evaluation has shown that drug concentrations that have an antiproliferative effect in vitro are attainable in mice and dogs (Supko et al 1995). Three to 4 hours after application of the maximum tolerated dose, plasma concentrations fell into the subtherapeutic range. Liver toxic effects with elevations of transaminases were predominant. Further changes included elevations in creatinine phosphokinase, lactic dehydrogenase, and blood urea nitrogen as well as leukocytosis and reticulocytosis.
Because of the toxicity of GA, a search was initiated to characterize substances more suitable for clinical development. One of those 17-(allylamino)-17-demethoxygeldanamycin (NSC330507) is structurally related to GA (Schnur et al 1995b). It shares the ability to bind to Hsp90 and to deplete erbB2, raf-1, and mutated p53 (Schulte and Neckers 1998). This compound is currently undergoing a phase 1 clinical trial at 5 institutions in the United States and Great Britain.
Radicicol is a functional analog without structural similarity to GA (Mirrington et al 1965). It also binds to the amino terminal nucleotide binding domain of Hsp90, exerting effects comparable to GA on the aforementioned regulatory proteins (Schulte et al 1998). Radicicol derivatives have shown promising in vivo activity in various murine cancer models (Soga et al 1999).
Recent research revealed that the most effective drugs for the treatment of epithelial tumors, paclitaxel and cisplatin, both bind to Hsp90 (Byrd et al 1999; Itoh et al 1999). The binding site for cisplatin has been determined to be near the carboxy terminus. Hsp90-dependent inhibition of aggregation of citrate synthase was reversed by cisplatin (Byrd et al 1999). The region of Hsp90 bound by paclitaxel has not yet been determined. Some effects of paclitaxel, eg, the expression of tumor necrosis factor, are blocked by GA (Itoh et al 1999).
GA exerts antitumor activity in a multitude of preclinical models. It influences the cellular concentrations of several oncologically significant, signal-transducing proteins through the induction of their proteasome-dependent degradation. A correlation exists between GA cytotoxicity and this protein modulation. It is currently unknown to what extent GA-induced molecular alterations contribute to its anticancer activity in case such an effect is clinically demonstrable. BAs represent the first class of antitumor drugs for which binding to Hsp90 was identified as the major, novel mechanism of action. GA is thus a model compound for the development of structural or functional analogs with a more favorable toxicity profile. To prepare for the application of these compounds in multimodal therapy, studies are under way to determine their interaction with ionizing radiation. Similarly, GA is a formidable tool to elucidate the currently unknown significance of members of the Hsp90 family in the radiation response.
We thank Dr Neckers, Tumor Cell Biology Section, Medicine Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, who critically reviewed the manuscript. We also thank Mrs Lobe, Clinic for Radiation Therapy, Otto-von-Guericke-University, Magdeburg, Germany, for secretarial assistance.