Zhenhai Gao, David M. Duhl and Stephen D. Harrison Chiron Corporation, Emeryville, CA 94608
Introduction - The ideal target for small molecule drug discovery is one that is essential for the disease state, yet non-essential for normal tissues. From a practical perspective, the ideal class of target has a proven track record of drugability, i.e. a history of the identification of agonists or antagonists with acceptable pharmaceutical properties. One recently emerged class of drugable targets is the protein kinases (1), exemplified by bcr-abl, the target of the highly successful new CML drug Gleevec (2). The common drugable feature of the kinase class is the ability to bind ATP. The ATP purine moiety is bound in a hydrophobic environment of the active site and stabilized by key hydrogen bonds (3). Such a site favors binding of flat, aromatic heterocycles, a class of compound that has proven amenable to optimization and drug development, thus making kinases eminently drugable. However, kinases are not the only class of enzyme that binds purines through key hydrogen bonds and hydrophobic stacking interactions; other such purine-binding enzymes include ATPases, GTPases, sulfotransferases and others. Thus, these enzymes offer additional possibilities for drug discovery. To illustrate the potential of this emerging class of drug targets we will focus our discussion on oncology and specifically on ATPases (Hsp90 and the kinesin KSP), GTPases and Sulfotransferases.
Heat-shock Protein 90 (Hsp90) - Hsp90s are ATP-dependent molecular chaperones that are preferentially over-expressed (2-10 fold) in some cancer cells (4). Crystallographic studies have revealed the existence of a non-conventional low affinity ATP binding cleft at their N-terminal domain that is well-conserved among the four family members: Hsp90 a and p, Grp94 and Trap-1 (5). The occupancy of the ATP binding site by the ansamycin antibiotics geldanamycin (GM), 1_, and herbimycin A (HA), 2, as well as the structurally unrelated fungal metabolite radicicol, 3, inhibits the intrinsic weak ATPase of Hsp90, and results in simultaneous destabilization and eventual ubiquitin-dependent degradation of its client proteins (69). Uniquely, many of the Hsp90 client proteins are well-known oncoproteins that are frequently mutated or over-expressed in cancer cells, such as the mutated p53, Bcr-Abl, Raf-1, Akt, ErbB2, and steroid receptors (9). The association with Hsp90 allows these otherwise unstable oncoproteins to function properly in multiple oncogenic signaling pathways, which are essential in maintaining the uncontrolled proliferation and the malignant phenotypes of tumor cells. Depending on cellular contexts, Hsp90 inhibitors effectively cause growth arrest, differentiation, or apoptosis of tumor cells both in vitro and in vivo (10-12). It is also conceivable that Hsp90s may be the key components of the very machinery that allows certain cancer cells to evoke alternative or overlapping signaling, and to efficiently develop resistance to a specific drug treatment (13). Hence, inhibitors of Hsp90, by concurrently disrupting a wide range of oncogenic pathways, may prove to be a very effective approach to combat a variety of hard-to-treat tumors.
17AAG, 4, a derivative of GM, is the first-in-class Hsp90 inhibitor that has entered human clinical trials in patients with solid tumors. It has recently completed phase I trials and is progressing to phase II (5). In the extensive preclinical studies, 17AAG has been shown to be very effective in blocking tumor growth in various human tumor xenograft models (5, 10). Although the efficacy of 17AAG was not the primary point of evaluation in the Phase I trial, very encouraging disease stabilization was observed in some patients with advanced solid tumors (14). In addition, 17AAG
also caused the expected reduction in the expression level of several "surrogate markers" such as Raf-1, Lck, and cdk4 in treated patients (14) Nonetheless, 17AAG is generally being viewed only as a valuable proof-of-concept anti-tumor agent, and may not eventually become a marketed drug due to several salient drawbacks: 1) poor solubility and difficulty in formulation; 2) lack of an easy synthetic route and dependence on fermentation; 3) hepatotoxicity associated with benzoquinone moiety in its structure; 4) lack of selectivity among the four Hsp90 family members, and 5) ineffectiveness on a subset of cancer cells (15). Likewise, the clinical development of radicicol derivatives has also been impeded by the fact that these compounds cause cataracts, an off-target effect in animals (16). Several GM analogs with pharmaceutical properties (i.e. potency and solubility) superior to those of 17AAG have recently been developed, e.g. the water-soluble 17-(dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG, 5) (17). The clinical advantages and benefits of these improved GM derivatives remain to be determined.
The published high-resolution crystal structures of Hsp90, in complex with adenosine nucleotides, GM or radicicol have raised the exciting possibility of developing selective Hsp90 inhibitors using structure-based rationale drug design (6, 8). ATP binds to Hsp90 in an unusual kinked conformation that distinguishes Hsp90 from other chaperone proteins or kinases. By computer-modeling the structure requirements, the first purine-based small molecule Hsp90 inhibitor, PU3, 6, was designed and synthesized (15). The relative binding affinity of PU3 for Hsp90 is 15 to 20|xM. Subsequent optimization and screening of a small biased library of compounds with purine scaffolds, has led to the identification a compound, PU24FCI, 7, that displays Hsp90 binding affinity, potency and a physiological profile comparable to those of 17AAG (16). In addition, a homogenous fluorescence-based conformational assay suitable for high-throughput screening (HTS) has recently been designed, which may help accelerate the pace of identification of potent and selective small molecule Hsp90 inhibitors with favorable drug-like properties (18).
While current drug development efforts have been focusing on the N-terminal ATP binding site, recent studies have suggested the existence of a cryptic ATP binding pocket located at the C-terminus of Hsp90 (19). More importantly, occupancy of this site by the gyrase inhibitor novobiocin, has also been shown to impair the chaperone activity of Hsp90, and cause cellular depletion of multiple client proteins (19). Therefore, targeting the C-terminal ATP binding site may represent a novel strategy to fully harness the anti-cancer potential of Hsp90 inhibitors.
Kinesin Spindle Protein (KSP) - Pharmaceutical agents that disrupt the function of the mitotic spindle and arrest cells in mitosis have proven to be remarkably effective and powerful chemotherapies for a broad spectrum of advanced tumors. Currently, all clinical anti-mitotic drugs act upon p-tubulin and affect the dynamics of microtubule (MT) polymerization and depolymerization, leading to cell cycle arrest and programmed cell death (20). One such agent, paclitaxel (Taxol®) has been widely recognized as one of the most clinically and commercially successful antitumor therapeutics developed in the past decades. Nonetheless, peripheral neuropathy, a notable side effect that is predictable from the critical role of MT in axonal transport in neuronal cells, has severely limited long-term use of these tubulin-binding compounds (20).
The recent advance in understanding the complex process of mitosis has led to the identification of several drugable targets that are not only essential but also more specific to the mitotic spindle apparatus. Of particular interest is KSP (also called Eg5). KSP is a member of kinesin superfamily of MT motors, which convert the energy released from ATP hydrolysis into mechanical force for transport along microtubules in the cell (21). A functional KSP is required for the establishment and maintenance of mitotic spindle bipolarity, and the proper segregation of replicated DNA into two daughter cells (22, 23). Moreover, KSP is predominantly localized to the mitotic spindle and appears to assume no function outside mitosis. Thus, specific inhibitors of KSP could have superior therapeutic value compared to tubulin-binding drugs in that they could effectively cause mitotic arrest and apoptosis in proliferating cancer cells, but may not incur the liability of producing significant neurotoxicity.
Both ATP hydrolysis and MT binding occur at the N-terminal motor domain of KSP (24). The nucleotide binding state regulates the KSP affinity for MT, and binding of MT in turn promotes ADP release and enhances ATP turnover rate. The nonhydrolysable nucleotide analogs (AMPPNP and AMPPCP) and adocia-sulfate-2, 8, a compound isolated from a marine sponge, are well known and characterized KSP inhibitors that are competitive with ATP and MT binding, respectively (25, 26). They are not, however, specific to KSP or cell-permeable, precluding the possible clinical use of these compounds. To date, development of potent and selective ATP- or MT-competitive KSP inhibitors has not been successful, presumably due to the fact that N-terminal motor domain shares significant structural and sequence similarity with other kinesins. An early attempt at structure-based rationale drug design using K560 (another member of kinesin superfamily) has led to the identification of several organic compounds that appear to be competitive for MT binding but not for ATP binding, and showed somewhat low selectivity (2-5 fold) for members of kinesin superfamily (27). As more and more crystal structures of kinesin proteins are becoming available, it would be interesting to determine if any of the observed subtle structural differences can be translated into the design of specific KSP inhibitors that target either ATP or MT binding pockets.
An alternative and appealing strategy to develop a KSP-specific small molecule inhibitor is to target an allosteric binding site that would interfere with the MT-
dependent ATPase activity. This approach appears to be feasible as demonstrated by recent identification of the first KSP-selective allosteric inhibitor, monasterol (IC5o=14-20 |iM), 9, by a phenotype-based screening of a 16,320-compound library (28). Subsequent high-throughput screening has resulted in the discovery of more potent and selective allosteric inhibitors, such as quinazolinone-based compounds CK-0106023, 10, and CK-0238273 (SB-715992) with binding affinity values (K,) below 20 nM (29). Detailed kinetic studies with monasterol and CK-010623 have shown that both compounds drastically reduce the MT-stimulated and basal rates of ADP release, but do not compete with either ATP or MT binding to KSP, indicating a novel allosteric inhibition mechanism (30, 31). Although all these inhibitors appear to interact with the N-terminal motor domain, the exact binding location and structural determinants remain elusive. Thus, the x-ray crystal structure of KSP in complex with an allosteric inhibitor will be extremely valuable for future design and optimization of KSP-selective small molecule inhibitors.
CK-0238273 has recently entered into phase I clinical trial. In preclinical studies, CK-0238273 (K,=0.6 nM) is very potent (IC5o, 1.2-9.8 nM) in causing mitotic arrest in various cultured human cancer cells (e.g. Colo205 and HT-29) with the appearance of a monopolar spindle, a hallmark of KSP malfunction during mitosis (32). In a broad panel of murine and human tumor xenograft models, CK-0238278 shows striking anti-tumor activities, with complete tumor regression observed with mice implanted with Colo205, Colo201, and Madison 109 lung carcinoma, and significant growth inhibition with HT29 and LL/2 Lewis lung carcinoma (32-34).
An important observation is that in a mouse model, paclitaxel clearly produces peripheral neuropathy, whereas this side effect is absent from CK-0238278 treated mice (32). This result is consistent with the notion that KSP function is restricted to the mitotic spindle in rapidly dividing cells, and bolsters confidence that selective KSP inhibitors could be superior anti-mitotic chemotherapeutic agents with a novel mechanism of action for advanced tumors.
GTPase proteins regulate intracellular signaling pathways by cycling between an active (GTP bound) conformation and an inactive (GDP bound) conformation (3537). When GTP is bound, the GTPase interacts with a diverse population of effector proteins to propagate the desired signal. The signaling cascade thus initiated by the activated GTPase will continue until the GTP is hydrolyzed into GDP and the protein assumes the inactive (GDP bound) conformation. The GTPases are themselves regulated by guanine nucleotide exchange factors (GEFs) and GTPase-activation proteins (GAPs). The GEFs act to accelerate the exchange of GTP for GDP, and therefore facilitate the transition of the GTPase to the active state. The GAPs accelerate the hydrolysis of GTP to GDP, and therefore facilitate the transition of the GTPase to the inactive state. Thus far, interest in GTPases as anti-cancer drug targets has focused on two families of proteins: the RAS/RHO family of small monomeric GTPases (20-30 kilo Dalton), and the heterotrimeric family of guanine binding proteins (commonly known as G proteins)(38, 39).
The high intercellular concentrations of GTP and the high binding affinity of the RAS/RHO family for GTP and GDP have been seen by many to be an overwhelming hurdle to targeting GTPases. Nevertheless, two promising approaches to cancer drug discovery have been identified: small molecules that affect the exchange of GTP for GDP by binding outside of the guanine binding pocket, and modifications of GTP which can facilitate GTP hydrolysis by mutated forms of RAS (substrate assisted catalysis).
GTPases are known to have a highly flexible and mobile active site, and this flexibility has been seen as the reason for their relatively poor catalytic efficiency (40). GEFs and GAPs are thought to stabilize conformations of the protein that facilitate the exchange of GTP for GDP and the hydrolysis of GTP, respectively (41, 42). In theory, it should be possible to stabilize the inactive state. One possibility might be to identify GDP analogs that lock the enzyme in its inactive conformations. An alternative would be to identify small molecules that bind to the GTPase in the inactive (GDP bound) state and inhibit the exchange of the GDP for GTP, thus potentially inhibiting the oncogenic nature of GTPases. Taveras and co-workers, in an attempt to develop compounds that bind RAS, have identified such an inhibitor (43). The authors use NMR, mass spectrometry, and molecular modeling to demonstrate that SCH-54292,1±, is bound to RAS-GDP in a major cleft of the switch-2 region. The highly flexible switch-2 region is known to change its conformation depending on whether GTP or GDP is bound to RAS. This result suggests that, with the appropriately designed screen, it should be possible to identify compounds that bind the switch-2 region, and stabilize the inactive conformation.
Substrate assisted catalysis (SAC) has been reviewed in detail elsewhere (44). In short, SAC envisions that both the enzyme and the substrate (GTP in the case of GTPases) play a role in the rate of catalysis. The crystal structures of RAS in both the GDP bound state and a transition state form of the enzyme (GDP-aluminum fluoride bound), have identified two amino acids in the binding pocket of all GTPases that are crucial for catalysis of GTP (Glncat & Argcat) (41, 42, 45). In both the small monomeric and heterotrimeric GTPases, Glncat is thought to function by polarizing the nucleophile and by orienting it for attack on the y phosphate (46). Mutations of Glncat decrease the intrinsic rate of hydrolysis by up to two orders of magnitude. The second essential catalytic amino acid is Argcat, which is thought to stabilize the transition state of hydrolysis. In small monomeric GTPases, Argcat is provided by insertion of an amino acid finger from a GAP (45), while in heterotrimeric GTPases Argcat is intrinsic to the protein itself. Similar to Glncat mutations, mutations of Argcat decrease the intrinsic rate of hydrolysis by up to two orders of magnitude. 3,4-Diaminobenzophenone-phosphoroamidate-GTP (DABP-GTP, (47)), 12, has been shown to restore the activity of Glncat mutants to near wild-type rates, but does not restore the activity when Argcat is inactivated with cholera toxin.
DABP-GTP has been shown to restore GTP hydrolysis to oncogenic GTPase deficient RAS mutated at either Gln61 or Gln12 (the most common mutation in tumors).
The Ga subunits of heterotrimeric GTPases contain the GTP catalytic domain. By grouping subunits with sequence identity of 50% or greater, Ga subunits are subdivided into four classifications (as, an0n, aq, 012/13) (48). Of the 16 mammalian Ga-subunits identified to date, Gas, Gae, Ga0, Goçq, Ga,L.Gaia,.GaZl Gai2.and Gai3 have been identified as having oncogenic potential in vitro, with Gai2 and Gai3 being the most oncogenic (49). Also mutations of Gas have been found in tumors
(pituitary tumors, thyroid adenomas and thyroid carcinomas). In addition to their oncogenic potential, Ga subunits can also be seen as targets for anti-cancer drug discovery by virtue of their role in downstream signaling from numerous oncogenic G-protein coupled receptors (50). Specifically, Gai2.ancj Gai3 have been shown to directly regulate potentially oncogenic RHO GTPases through the interaction of RGS containing GEFs. Since unmutated RHO has been shown to be over-expressed in breast, colon, and lung tumors, inhibition of Gai2, and. Gai3 may lead to the down regulation of this potent oncogene (51-54).
In support of the concept that binding outside of the guanine pocket may stabilize GTPase inactive conformations (outlined above), results similar to those with RAS and SCH-54292 have been found with heterotrimeric GTPases. Suramin, a symmetric polysulphonated napthylamine-benzamide-derivative used to treat African sleeping sickness, has been shown to inhibit the release of GDP from purified Ga subunits (55). As discussed for SCH-54292, Suramin is thought to bind the GTPase outside of the guanine pocket. Synthesis of Suramin-like compounds has lead to the identification of two compounds (NF503, 13, and NF449) with increased specificity for Gas over the other Ga subunits (56).
Through the convergence of nascent medicinal chemical efforts and an increased understanding of the biology, structure, and enzymology of GTPases, anti-cancer drug discovery within this area seems poised to release its untapped potential.
Sulfotransferase (ST) enzymes catalyze the transfer of sulfate from the donor molecule, purine-based PAPS (3'phosphoadenosine 5'phosphosulfate, 14) to a variety of intracellular substrates, ranging from small xenobiotics to macromolecules. These enzymes fall into two classes. Cytosolic ST enzymes, also referred to as SULTs, detoxify xenobiotics and modify small endogenous molecules, such as hormones and neurotransmitters. The ST enzymes from the second class, numbering over 30 individual types, are found within the Golgi membrane and catalyze the sulfation of carbohydrates (including glycoproteins, proteoglycans, and glycolipids) and proteins (57).
ST enzymes are being considered as therapeutic targets for a number of disease indications, including inflammation and cancer. The former application has previously been reviewed and we will focus on the potential for these enzymes as anticancer targets (58). It has been argued that members of both the cytosolic (SULTS) and Golgi ST classes have potential for the treatment of cancer. Because of the role of SULTs in xenobiotic detoxification, inhibitors of these enzymes may have use in potentiating the effect of cytotoxic chemotherapeutics. Estrogen sulfotransferase (EST) has also been suggested as a target because unusually high levels of its product, estrogen sulfate, have been found in breast cancer cells (59).
Amongst the Golgi enzymes, much interest has been focused on the carbohydrate sulfating enzymes. Carbohydrate sulfation modulates a number of processes that are important in cancer, such as cell-cell interaction and cell-matrix association and it is established that growth factor activity, such as that of fibroblast growth factor, is dependent on the presence of heparan sulfate. Chondroitin sulfation is associated with cancer progression, poor prognosis, tumor adhesion and metastasis (60-62). Thus inhibition of chondroitin ST enzymes may have therapeutic benefit. The protein tyrosine targeting ST enzymes may also have a role in tumorigenesis, as protein tyrosine sulfation, modifies extracellular protein/protein associations and hence potentially cell activation and environmental interactions (63).
The mechanisms of action of ST enzymes suggest that this class may be susceptible to inhibition by drug-like small molecules. The sulfate group from the purine PAPS is transferred to a hydroxyl or amine, in a manner analogous to the phosphate transfer from the purine ATP to amino-acid hydroxyls during kinase-catalyzed phosphorylations. The precedent of identifying selective purine site inhibitors for kinases raises hope that similar selectivity will be achieved for ST enzymes. Indeed apart from the conserved PAPS phosphate binding sequences (see below), there is considerable sequence diversity between ST enzymes, reflecting a broad array of substrates. For example, even amongst the subset of carbohydrate-targeting Golgi enzymes there are a variety of sulfate acceptor sites, including either the 4' or 6' hydroxyl of N-acetylglucosamine (as in chondroitin 4-0-ST and chondroitin 6-ST respectively) and the 3' hydroxyl of glucoronic acid (as in HNK-ST) (64).
The design and optimization of chemical inhibitors of ST enzymes will be greatly aided by the availability of crystal structures of four such proteins (EST, hydroxysteroid ST, catecholamine ST and heparan sulfate N-deacetylase/N-ST) (65). As mentioned, the adenine-binding sites of these enzymes are related to those of kinases. Despite this similarity, the ST class possesses two characteristic nucleotide-binding motifs not seen in kinases. The PSB loop contains a P-loop such as found in ATPases (66) that interacts with the 5'phosphate of PAPS and the 3'PB domain associates with the 3'-phosphate that is found in PAPS, but not ATP (67).
Several inhibitors of different ST enzymes have been identified using a variety of approaches. The screening of in vitro enzyme assays has yielded several low potency inhibitors (68, 69). By monitoring inhibition of the Nacetylglucosamine-6-ST, NodH, from Rhizobium meliloti, several inhibitors were identified from a purine-based library (e.g. 15: IC5o, 20nM) in addition to a single tyrphostin (16: IC5o, approximately 50|iM). The purines showed some selectivity against other ST enzymes, including N-acetylglucosamine-6-O-ST from human high endothelium venules and keratin sulfate ST. In an alternative approach, a tyrosylprotein ST (hTPST-2) assay was screened against a library of small aromatic O-methyl oximes at high concentration (200nM) (63). Weak enzyme binders were then dimerized via a short linker to yield bidentate compounds in an attempt to promote synergistic binding. Two compounds (17a: IC5o, 30|iM and 17b: IC6o, 40jiM) were identified, although strikingly these compounds turned out to be much more potent antagonists of EST (IC5oS, 250nM and 3|iM respectively) and may form the basis for an optimization approach to generate EST inhibitors. A more directed
approach to bidentate inhibition has been suggested by analysis of ST crystal structures. Bertozzi and coworkers have attempted to generate potent EST ligands from a library of compounds based on a PAPS mimetic fused to potential analogs of the estrogen substrate, 18 (70). This approach yielded several weak binders (18a. 18b and 18c: 80-90% EST inhibition at 200nM). In a subsequent study, a PAPS/estrogen hybrid with closer similarity to the presumed transition state of EST catalysis was generated and displayed very high affinity binding (19, IC50, 3-4nM). Although given that the two substrates have Km values comparable to this binding affinity, it is unlikely that synergistic association was achieved.
This promising inhibitor screening data, coupled with the possibility of structure-based drug design and increasingly attractive rationales for targeting ST enzymes, suggests that this class of enzymes may soon become as popular as kinases as targets for cancer chemotherapeutics.
Conclusions - The recent success in bringing Gleevec to market and the advancement of other kinase inhibitors into late stage clinical trials has proven that kinases are drugable targets for cancer therapeutics. A deeper understanding of the ATP-enzyme binding interaction suggests that the utility of kinases as targets can be expanded to include many purine-binding enzymes. By increasing the diversity of drugable targets, there is the potential for broadening our repertoire of small molecules that can bind the purine pocket, as well as the opportunity to develop compounds that allosterically interfere with the binding interaction. While we have illustrated the potential of purine-binding enzyme targets with reference to a limited number of enzyme classes, many others exist, including topoisomerases, ion pumps, P-glycoprotein and other drug transporters. This wide array of potential drugable targets promises to be a fertile source of new drug leads over the coming years.
Was this article helpful?
Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.