Pharmac. Ther. Vol. 52, pp. 127-147, 1991 0163-7258/91 $0.00 + 0.50 Printed in Great Britain. All rights reserved © 1992 Pergamon Press plc Specialist Subject Editor: E. HAMEL EFFECTS OF ANTIMITOTIC AGENTS ON TUBULIN-NUCLEOTIDE INTERACTIONS JOHN J. CORREIA Department of Biochemistry, University of Mississippi Medical Center, 2500 North St., Jackson, MS 39216, U.S.A. Abstract--The interaction of antimitotic drugs with guanine nucleotides in the tubulin-microtubule system is reviewed. Antimitotic agent-tubulin interactions can be covalent, entropic, allosteric or coupled to other equilibria (such as divalent cation binding, alternate polymer formation, or the stabilization of native tubulin structure). Antimitotics bind to tubulin at a few common sites and alter the ability of tubulin to form microtubules. Colchicine and podophyUotoxin compete for a common overlapping binding site but only colchicine induces GTPase activity and large conformational changes in the tubulin heterodimer. The vinca alkaloids, vinblastine and vincristine, the macrocyclic ansa macrolides, maytansine and ansamitocin P-3, and the fungal antimitotic, rhizoxin, share and compete for a different binding site near the exchangeable nucleotide binding site. The macrocyclic heptapeptide, phomopsin A, and the depsipeptide, dolastatin 10, bind to a site adjacent to the vinca alkaloid and nucleotide sites. Colchicine, vinca alkaloids, dolastatin 10 and phomopsin A induce alternate polymer formation (sheets for colchicine, spirals for vinblastine and vincristine and rings for dolastatin 10 and phomopsin A). Maytansine, ansamitocin P-3 and rhizoxin inhibit vinblastine-induced spiral formation. Taxol stoichiometrically induces microtubule formation and, in the presence of GTP, assembly-associated GTP hydrolysis. Analogs of guanine nucleotides also alter polymer morphology. Thus, sites on tubulin for drugs and nucleotides communicate allosterically with the interfaces that form longitudinal and lateral contacts within a microtubule. Microtubule associated proteins (MAPs), divalent cations, and buffer components can alter the surface interactions of tubulin and thus modulate the interactions between antimitotic drugs and guanine nucleotides. CONTENTS 1. Modes of Interaction 127 1.1. Mechanism of microtubule assembly 127 1.2. General features of antimitotic drugs-tubulin interactions 129 1.3. Thermodynamic linkage 130 2. The Colchicine Site 132 2.1. Colchicine 132 2.2. Podophyllotoxin 133 2.3. Estrogenic drugs 134 3. The Vinblastine/Maytansine Site 135 3.1. Vinca alkaloids 135 3.2. Maytansinoids 136 3.3. Phomopsin A and dolastatin 10 137 4. Taxol 138 5. The Third Nucleotide (ATP) Binding Site 140 6. N-site GTP 140 7. Microtubule Associated Proteins (MAPs) 141 Acknowledgements 142 References 142 1. MODES OF INTERACTION The interaction of antimitotic drugs with tubulin and microtubules has been extensively reviewed (Lacey, 1988; Hamel, 1990) and is a very active area of research. These drugs are useful as biochemical and cytological probes, and have proven or potential utility as antineoplastic agents. This review will high- light a particular aspect of their interaction with tubulin: the interplay between antimitotics and nucleotides. This interaction will be discussed first in terms of a number of general molecular aspects. Specific drugs and topics will be discussed later in this chapter. 1.1. MECHANISM OF MICROTUBULE ASSEMBLY The mechanism of microtubule assembly/disas- sembly has been reviewed elsewhere (Correia and Williams, 1983; Purich and Kristofferson, 1984; Dustin, 1984; Kirschner and Mitchison, 1986; 127
`
`IMMUNOGEN 2291, pg. 1
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`128 J.J. CORREIA Mandelkow and Mandelkow, 1989; Engelborghs, 1990), but, to facilitate the discussion of drug and nucleotide effects, a brief overview is presented here. Microtubules are composed of tubulin (an 7/~ heterodimer of 40% sequence-similar polypeptides, each of about 50kDa) and numerous microtubule associated proteins, MAPs (Wiche et al., 1991), that decorate the exterior walls of the hollow microtubule structure. Typical purification schemes generate microtubule protein (MTP), tubulin plus MAPs, or purified tubulin (Williams and Lee, 1982). Multiple tubulin genes, gene products, and post-translational modifications occur in a species specific manner (Cleveland and Sullivan, 1985; Sullivan, 1988). There are two guanine nucleotide binding sites in a tubulin heterodimer. One, the exchangeable (or E-site) on the /? chain, will rapidly exchange GTP for GDP in a Mg 2÷ dependent manner (Correia et al., 1987), and will hydrolyze GTP during microtubule formation. The other site (N-site) on the ~ chain is noncatalytic, is always occupied by GTP, and is nonexchangeable for nucleotides, although it will slowly exchange free Mn 2+ for bound Mg 2+ (Correia et al., 1988). In vivo, tubulin forms a helical rod composed of 13 proto- filaments that run the length of the polymer (Scheele et al., 1982). In vitro conditions often support struc- tures composed of 12-16 protofilaments (Scheele et al., 1982). It is not known how this variation affects ligand binding sites. Microtubules are 300 A in diam- eter, by electron density maps from fiber diffraction studies (Besse et al., 1987a,b). The molecular mechan- ism of microtubule assembly is believed to proceed from a nucleation event, a highly unlikely event that exhibits a critical tubulin concentration, Co, below which polymerization does not occur. Nucleation probably involves protofilament formation and lat- eral association of protofilaments into a sheet struc- ture that curves into a helical rod (Thompson et al., 1981; Detrich et al., 1985). Sheets are often observed by electron microscopy and the occurrence of a short sheet region with a microtubule is consistent with a helical lattice with a seam, referred to as the B-type lattice (Mandelkow et al., 1986; Linck, 1989; Mandelkow and Mandelkow, 1989). There is evi- dence that microtubules are cylindrical sheets, a two dimensional polymer constrained to a cylindrical surface, and that sheets and ribbons are overshoot products (Mandelkow and Mandelkow, 1989). Upon closure of the sheet, microtubule growth, called elongation, occurs in an endwise manner with het- erodimers adding to both ends in vitro, although at different rates (Fig. 1). The fast growing end is the plus end (Fig. 1). The minus end is typically anchored to microtubule organizing centers (MTOC) like basal bodies or centrosomes (Euteneuer and Mclntosh, 1981a,b), and thus growth in vivo typically occurs in the plus direction. A different critical concentration and rate of growth at the two ends is allowed thermodynamically because of the irreversible step of GTP hydrolysis that is coupled to subunit addition to the polymer (Kirschner and Mitchison, 1986). This hydrolysis step occurs concurrent with or soon after subunit addition and limits the size of the GTP cap, a region of tubulin heterodimers at both ends of the microtubule that contain GTP bound to their E-site. Subunit disassembly occurs by endwise loss of heterodimers, now containing GDP at the E-site (Melki et al., 1989), although at high Mg 2÷ concen- tration protofilament or ring structures may disas- semble, thus accelerating the rates (Mandelkow and Mandelkow, 1985; O'Brien et al., 1990; Fig. 1). The hydrolysis of GTP and the conformation switch to the GDP form of tubulin within the microtubule is believed to include the release of MgPO4. This is similar to the mechanism of ATP hydrolysis in actin and actomyosin filaments (Korn et al., 1987; Carlier, (-) or (+) (*) or (-) end end Q m I HPO4Mg I1 Mg=+ O FIG. 1. An idealized mechanism of microtubule assembly/disassembly. GTP-tubulin heterodimers with GTP at the E-site are represented by dark ovals; GDP-tubulin heterodimers are represented by light ovals. Microtubule growth occurs at both ends in vitro. In vivo the slow growing end, the minus end, may be anchored at MTOC's. GTP hydrolysis occurs upon assembly and, after the release of HPO4Mg, generates a microtubule that is stabilized by a GTP cap, a layer of GTP-tubulin subunits that prevent catastrophic disassembly. Disassembly may occur by the loss of oligomers. Subunits may also form nonmicrotubule polymers, typically double walled rings. Only the GDP-tubulin reaction is shown (Frigon and Timasheff, 1975a,b; Howard and Timasheff, 1986).
`
`IMMUNOGEN 2291, pg. 2
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`Tubulin-nucleotide interactions 129 1991a), although the rate of PO 4 release appears to not be rate limiting for microtubules (Carlier et al., 1989; Combeau and Carlier, 1989; Melki et aL, 1990). The addition of GTP-tubulin subunits to the ends of a growing microtubule, and the hydrolysis of GTP after addition to the microtubule, leads to a gradient of GTP within the microtubule and thus to two forms of polymers, the GDP- and the GTP-form. This has been described as a phase transition (Hill, 1983). It is now widely believed (however see discussion in Carlier, 1991 b) that only a few GTP-subunits (maybe as few as one GTP-tubulin layer) exist at the end of any given microtubule, consistent with a lateral cap model (Bayley et al., 1990) or a vectorial model (Carlier et al., 1984). This conclusion is strongly dependent upon the ability to rapidly fix and trap a population of microtubules during growth such that they retain y32p labeled GTP (Stewart et al., 1990). The temporal sequence of this event boils down to whether the delay after GTP-tubulin binding and before GTP hydrolysis, with the corresponding con- formational change(s), is ~sec or sec (see Carlier, 1991b for a discussion). Alternative theories endorse co-operative effects at the end of a cylindrical surface (Mandelkow and Mandelkow, 1989). Under steady state conditions, the removal of this cap allows for a conversion from an assembling or growing micro- tubule with GTP-tubulin at the ends to a disassem- bling or shrinking microtubule with GDP-tubulin at the ends. This process, referred to as dynamic insta- bility, was first observed with microtubules nucleated from isolated centrosomes (Mitchison and Kirschner, 1984a,b) and is known to occur in vitro and in vivo. The effect is due to the fact that the off rate for GDP-tubulin heterodimers is ca. 100 fold faster than GTP-tubulin heterodimers (Gal et al., 1988; O'Brien et al., 1990). This may in part be due to the increased ability of GDP-tubulin heterodimers to form oligomers in a Mg 2÷ dependent manner (Frigon and Timasheff, 1975a,b; Howard and Timasheff, 1986; see Fig. 1), and is consistent with different conformations of GDP- and GTP-tubulin. The consequences of this are that (1) microtubules are simultaneously growing and shrinking in the same solution although maintaining a constant polymer mass, (2) GTP is hydrolyzed in an initial burst with microtubule polymerization and at a steady state with dynamic instability, and (3) the mean length of microtubules increases with time as a few micro- tubules completely depolymerize (catastrophic dis- assembly) and the heterodimers released exchange bound GDP for free GTP and add onto existing microtubules. Recovery or rescue from disassembly (tempered disassembly) is typically observed (Horio *Sulfhydryl modifications are chemically reversable by reducing agents, however tubulin is often irreversibly changed by the modification of 1-2 SH groups/dimer, preventing complete recovery of polymerization activity. Sulfhydryl modification of tubulin has been reviewed in this series (Luduena and Roach, 1991). and Hotani, 1986; Bayley et al., 1990; O'Brien et al., 1990). The extent of microtubule dynamics is strongly dependent upon the buffer conditions. Glycerol (Kristofferson et al., 1986) and the presence of acces- sory proteins, MAPs (Kristofferson and Purich, 1981; Horio and Hotani, 1986; Keates, 1990), reduce dy- namics. There is some support for treadmilling (Weg- ner, 1976) under these conditions, a vectorial process where growth occurs at one end and disassembly occurs at the other end, leading to a flux of subunits through the microtubule (Farrell et al., 1987; Hotani and Horio, 1988; see discussion in Correia and Williams, 1983). Most of the above discussion pertains to in vitro experiments although the same concepts are often invoked in explaining in vivo results. For example, microtubules in nonneuronal cells are known to be dynamic (Schulze and Kirschner, 1986) and micro- tubule dynamics are believed to be involved in the formation of the mitotic spindle and chromosome movement (Gorbsky and Borisy, 1989; Mitchison, 1988). Thus, cellular requirements of a cytoskeleton are consistent with the dynamic instability model of microtubule assembly/disassembly (see Kirschner and Mitchison, 1986). 1.2. GENERAL FEATURES OF ANTIMITOTIC DRUGS-TuBuLIN INTERACTIONS As explained above, guanine nucleotides are involved in microtubule assembly and dynamics. Antimitotic agents that inhibit microtubule for- mation can be viewed as repressor molecules in this interaction. The activators are GTP and MAPs. The activity is dynamic assembly. In this context, then, our goal is to outline the various molecular means by which these effectors interact. Drugs that influence the interactions of GTP might reduce the affinity of its binding, alter the activity of the GTPase, or shift the conformation of GTP-tubulin to GDP-tubulin or to a nonmicrotubular form. The general principles of protein-protein interactions, enzyme kinetics and allosteric regulation pertain to this discussion (Nichol and Winzor, 1972; Oosawa and Asakura, 1975; Frieden and Nichol, 1981; Perutz, 1990; Wyman and Gill, 1990). Most of the interactions to be discussed here are reversible interactions. Most drug binding to tubulin is rapidly reversible.* Colchicine is the primary excep- tion, due to a large activation barrier of binding, and thus an extremely slow off rate (see below). Taxol binding to tubulin is often referred to in the literature as being strong and irreversible, but its affinity for microtubules is only 106 M -1 (Parness and Horwitz, 1981), and methods have been described for its recovery from microtubule solutions (Collins and Vallee, 1987; Collins, 1991). Antimitotic drugs pre- vent microtubule assembly (colchicine, podophyllo- toxin, vinca alkaloids, etc.) or prevent microtubule disassembly (taxol). Cell toxicity is thus believed to be
`
`IMMUNOGEN 2291, pg. 3
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`130 J.J. CORREIA due to preventing formation of the mitotic spindle, or, in the case of taxol, preventing disassembly of the cytoplasmic microtubules at metaphase, or pre- venting disassembly of the mitotic spindle and thus cell division during mitosis. Any disruption of micro- tubule dynamics may induce metaphase arrest (see Jordan and Wilson, 1990). Upon polymerization into a microtubule, tubulin becomes a GTPase and cleaves the terminal phosphate bond of GTP bound at the E-site. Thus, if a drug prevents microtubule formation or alters steady state dynamics, then it indirectly prevents or diminishes GTP hydrolysis. This is not in general believed to be a direct allosteric modification of the GTP binding site, but an indirect effect due to the inhibition of polymerization. There are many potential mechanisms of drug inhibition of microtubule assembly. There are two kinds of stoichiometric effects. A direct steric mechanism involves drug binding at a site on the heterodimer that forms an interface within the micro- tubule and thus blocks longitudinal or lateral con- tact and bond formation. Alternatively, an indirect effect occurs when binding induces a conformational change in tubulin that interferes with microtubule bond formation. Both of these mechanisms involve sequestering of a drug-tubulin complex in a nonpoly- merizable form. At stoichiometric concentrations, polymorphism, the formation of alternate nonmicro- tubular polymer forms, may be induced. As long as the free tubulin concentration is less than Co, polymerization is inhibited. For example, vinblastine and vincristine induce spiral protofilaments that are nonmicrotubular and lack GTPase activity. Vinca alkaloids also prevent microtubule assembly in a substoichiometric manner by means of a mechanism that is believed to cause poisoning of the ends of growing microtubules. A drug-tubulin complex adds to a growing protofilament, alters the conformation of that end and prevents the addition of tubulin subunits. This mechanism may involve copolymeriza- tion of tubulin and drug-tubulin molecules that alter the overall effÉciency or energetics of growth (Sternlicht and Ringel, 1979). To put this in quanti- tative terms, vinblastine will inhibit mitosis at 7.5 x 10 -8 M, but in vitro it binds to tubulin with an affinity of ca. 4 × 104M 2. This reflects the difference between inferring an effect due to substoichiometric poisoning by measuring microtubule assembly and a direct equilibrium binding measurement (see below). Since treadmilling may predominate under certain conditions, poisoning of endwise growth may be selective or preferential for one end of a microtubule. (It will probably not be exclusive. Treadmilling can be expressed quantitatively with S =0 meaning no treadmilling at steady state, and S = 1 meaning only association at one end and only disassembly at the other end. The experimental data puts S at ca. 0.1 (Correia and Williams, 1983). This means both ends grow, one just grows faster.) Alternatively, drugs may induce alternate polymer formation and yet retain the activation of GTPase activity. Colchicine forms sheets or ribbons that have GTPase activity (Andreu and Timasheff, 1982c). Colchicine also induces weak GTPase activity in the heterodimer. It is not clear if GTP hydrolysis is coupled to colchicine-tubulin polymer formation although the evidence is that the GTPase activity increases with polymerization. Possibly the weak GTPase activity observed with heterodimers is actually coupled to small oligomer formation, perhaps as small as dimers of heterodimers (Heusele and Carlier, 1981). There is no evidence that colchicine-tubulin polymers are dynamic structures that assemble and disassemble in a GTP-dependent manner. The molecular details of these interactions are not understood. Many studies of antimitotic analogs are beginning to unravel the structure-function requirements of these drugs (Lin et al., 1988; Batra et al., 1988; Liu et al., 1989; Andreu et al., 1991; Medrano et al., 1991; see Muzaffar and Brossi, 1991). However, the molecular structure of tubulin and microtubules is not likely to be known for some time (see Mandelkow and Mandelkow, 1989 for a discussion), and thus the topology of the binding sites remains a mystery. Below, attempts will be made to correlate binding at a particular site with conse- quences. The effects are not often additive, consistent or conclusive. In the future, upon obtaining the crystal structure of a dru~tubulin complex and viewing the molecular interaction involved, it will then be our task to explain how these contacts allosterically affect contacts within the rnicrotubule. Many workers persist in using a molecular weight of the heterodimer (110 kD instead of 100 kD) de- rived from SDS PAGE or denaturing sedimentation equilibrium experiments. The use of this erroneous value will systematically affect tubulin concentrations by ca. 9%, and will thus overestimate stoichiometries and binding constants determined for drug or nucleo- tide interactions with tubulin. Much of the tubulin literature uses the terms oligomer and aggregate interchangeably. Thus, there is no distinction made between reversible oligomer formation and irrevers- ible aggregate formation. Ignoring the irreversible GTPase step, microtubule formation is reversible in the sense that heterodimers can be cycled between functional monomers and polymers. This is more like recycling an enzyme. Most drug induced polymorphic forms are reversible polymers and are sensitive to concentration effects. Irreversible tubulin aggregates are often irregular, denatured forms and are not our concern here. 1.3. THERMODYNAMIC LINKAGE The theory of thermodynamic linkage as applied to macromolecular interactions was first developed by Wyman (1964) and has been applied extensively to biological macromolecules (Wyman and Gill, 1990). It is based upon the thermodynamic principle that a
`
`IMMUNOGEN 2291, pg. 4
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`Tubulin-nucleotide interactions 131 change in the activity of any species affects the activity of all other species in solution. The Wyman linkage relationship is (6 In K/6 In a3)r.pc, z = A~ where K is the equilibrium constant for some reaction to product at constant temperature, pressure, and protein concentration, a 3 is the activity of the ligand, and Af is the change in the apparent additional binding of component 3 to the protein, component 2, during the reaction. Note that it reflects a change in binding and not total amount bound, and is thus referred to as preferential interaction. This equation can be applied to protein stability, self-association, ligand binding, or solubility. In the context of micro- tubule assembly the binding of a ligand like a drug influences the binding of a heterodimer to the grow- ing end of a microtubule. A~ can be positive or negative and does not imply a specific binding site. For example, nonspecific electrostatic binding as described by Record et al. (1978) is an application of preferential interactions. The ligand need not bind uniquely to the polymer form, just more tightly. If a ligand preferentially binds to microtubules over heterodimers, then the addition of that ligand will shift the reaction to microtubule formation. For example, microtubule formation is accompanied by the binding of one H + (0.86) and one Mg 2+ (0.78) per heterodimer addition (Andreu and Timasheff, 1986). Thus microtubule formation is favored by lower pH and increasing Mg 2+ concentration. The sites of binding may be newly formed at the subunit interfaces in the polymer, but their location is irrele- vant to the consequence; the extent and free energy of microtubule formation is increased. Thermodynamic linkage applies to microtubule assembly in numerous ways. (1) Colchicine and a number of colchicine analogs induce large con- formational changes in tubulin. Some of these conformational states have GTPase activity not coupled to microtubule assembly. Using a bifunc- tional model and ligand-linked conformational equi- libria, a thermodynamic model has been proposed that is consistent with and explains the kinetics of colchicine binding (Andreu et al., 1991). (2) The Mg 2+ induced assembly of tubulin into rings involves the binding of one additional Mg 2+ per heterodimer and proceeds by isodesmic chain growth and a ring closure step (Frigon and Timasheff, 1975a,b). The vinblastine-induced assem- bly of tubulin into spirals (Na and Timasheff, 1986a,b) can be described by a ligand-mediated plus ligand-facilitated polymerization model. As described below, there is one assembly-linked vinblastine bind- ing site. The concentration dependence of micro- tubule and nonmicrotubule polymer formation magnifies the allosteric interactions between multiple ligands. This in turn is coupled to large changes in the solvent activity due to large changes in surface area upon assembly (Wyman and Gill, 1990). (3) The concept of preferential interactions has been applied to explain the influence of solvent components on the induction of microtubule assembly. Typical examples are glycerol or DMSO nonspecific stimulation of assembly at very high concentrations. The driving force is the preferential hydration of tubulin and the minimization of surface area to avoid interaction between the solute, e.g. glycerol, and the macromolecule (Lee and Timasheff. 1977). (4) In addition to nonspecific preferential inter- actions, there are a number of specifc chelation equilibria that are important to the tubulin system. The original discovery of in vitro microtubule assem- bly conditions involved the addition of EGTA as a Ca 2 + chelator. Ca 2 + inhibits microtubule assembly, in the mM range for MTP and in the #M range for pure tubulin (Berkowitz and Wolff, 1981). Since Mg 2+ is required for strong GTP binding to the E-site, chelation of divalent cations in general will also inhibit assembly. Tropolone (Andreu and Timasheff, 1982a) and daunomycin (Na and Timasheff, 1977) are weak microtubule inhibitors due to binding to tubulin. However, they also bind Mg 2+ and thus could influence assembly by altering the free divalent cation concentration (Andreu and Timasheff, 1982d; Dabrowiak, 1980). Mn 2 + will sub- stitute for Mg 2+ and in the early reports it was noted that Mn 2 ÷ was less effective than Mg 2 + in promoting assembly (Buttlaire et aL, 1980). However, those authors used lmM EGTA in their buffers, a weak chelator of Mg 2+ but a strong chelator of Mn 2+ (Correia et al., 1988). In addition drugs may interact with buffer components. Hinman and Cann (1976) reported that chlorpromazine complexes with sucrose and complicated the interpretations of their study of chlorpromazine interaction with mouse tubulin. Chlorpromazine competes with colchicine and in tissue culture resembles colcemide in arresting mitosis and disrupting organized microtubule structure (Appu Rao et aL, 1978). (5) It has been reported that vincristine stabilizes the colchicine binding activity of tubulin (Wilson, 1970) and that colchicine itself prolongs native con- formation of the protein (Garland, 1978). This is a thermodynamic linkage between drug binding and tubulin stabilization. The native tubulin confor- mation is stabilized against denaturation by the ad- ditional energy of drug binding or nucleotide binding (cf. Brandts and Lin, 1990). The induction of polymer formation may provide additional stabilization. This probably contributes to the influence of vinca alkaloids or taxol on tubulin stability. Monod et al. (1963) predicted that "no direct interaction need occur between the substrate of the protein and the regulatory metabolite which controls its activity". In the examples listed above the driving forces are free energy changes mediated by coupled equilibria and allosteric interactions. The thermo- dynamic linkage principle is sufficient to understand
`
`IMMUNOGEN 2291, pg. 5
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`132 J.J. CORRE1A this partial list, and to correlate structure, function and energy aspects of an antimitotic-microtubule system. 2. THE COLCHICINE SITE 2.1. COLCHICINE Colchicine (Fig. 2) is the classic antimitotic; in fact, tubulin was once known as the colchicine binding protein (Wilson, 1970). Colchicine has four major effects on tubulin: (1) It induces a change in secondary structure as measured by circular dichroism, (2) it induces a weak GTPase activity in the heterodimer, (3) it inhibits microtubule formation substoichiomet- rically and thus inhibits the microtubule-linked GTPase activity, and (4) at elevated temperatures it induces sheet-like polymer formation at stoichio- metric concentrations of drug. The kinetics (Lambeir and Engelborghs, 1981; Bane et al., 1984; Engelborghs and Fitzgerald, 1987; Hastie, 1989) and thermodynamics (Andreu et al., 1991; Medrano et al., 1989, 1991) of colchicine binding to tubulin have been extensively characterized. Roles for the A, B and C rings have been proposed. Colchicine is known to undergo a slow conformational change upon binding to tubulin and to induce a confor- mation change in tubulin. The slow kinetics of effectively irreversible colchicine binding are due to a large activation barrier (E, = 24 kcal/mol), in part involving distortion of the central or B ring (Garland, 1978; Lambeir and Engelborghs, 1981; Banerjee et al., 1987). Removing the B ring increases the overall rate and decreases the activation barrier with a slight reduction in the overall affinity (Bane et aL, 1984; Engelborghs and Fitzgerald, 1987). The specificity of binding is due to H-bonding and hydro- phobic interactions at the trimethoxybenzene A ring and to stacking interactions at the 7 membered tropolone C ring. The link between colchicine and tubulin conformational changes has been experimen- tally inferred from kinetic studies (Garland, 1978; Lambeir and Engelborghs, 1981), measured by CD (Detrich et aL, 1981; Andreu and Timasheff, 1982a) and quenched protein fluorescence (Andreu and Timasheff, 1982a), demonstrated by immunological CH=O~NH-R c..o" "T" ~ "~ cH,O ],~0 OCH3 OH o CH=O/~OCH: OCH~ Colchicine: R • COCHs Podophyllotoxin Colcemid: R = CH= F16.2. The structure of colchicine, colcemid and podophyl- lotoxin. methods (Morgan and Spooner, 1983), and inferred by the induction of GTPase activity in the hetero- dimer (David-Pfeuty et al., 1979; Andreu and Timasheff, 1981). The small CD changes are also induced by analogs of colchicine, MTC (2-methoxy- 5-(2,3,4-trimet hoxyphenyl)-2,4,6-cycloheptatrien- 1 - one) (Andreu et al., 1984), allocolchicine (Medrano et al., 1989) and tropolone methyl ether (Andreu and Timasheff, 1982b). Podophyllotoxin (Fig. 2) is known to compete with colchicine for binding to tubulin and is believed to share a site with colchicine through the trimethoxy- benzene moiety; yet podophyllotoxin does not induce a conformational change in tubulin or induce the GTPase activity. In fact, podophyllotoxin and tro- polone, the C ring alone, can bind to tubulin simul- taneously suggesting that only the A ring is shared in these binding sites (Cortese et al., 1977). Neither tropolone methyl ether nor vinblastine induces the GTPase activity (Andreu and Timasheff, 1981). An early report suggested that griseofulvin and dauno- mycin induce GTPase activity in heterodimers and do not compete with colchicine binding (David-Pfeuty et al., 1979), but this has not been followed up. A molecular understanding of this selectivity must await determination of a three dimensional structure of tubulin and a knowledge of the drug binding sites. Recent work suggests that the colchicine related site that induces GTPase activity requires binding to both the A ring and C ring subsites. Thus, GTPase activity is trimethoxybenzene-specific while the confor- mational change can be due to interactions, stacking or n stacking of the C ring alone (Lin and Hamel, 1981; Andreu et al., 1991; Medrano et al., 1991). This is also the mechanism of colchicine and colcemid dimerization (Engelb0rghs, 1981). The identity of the aromatic amino acid participating in this stacking is not known, but tryptophan has been suggested (Andreu and Timasheff, 1982a; Hastie, 1989). There is evidence that the B ring subdomain may also be involved in induction of the GTPase activity (Banerjee et al., 1987). An alternative model of the colchicine binding site has been proposed based upon the rapid and revers- ible binding of combretastatin A-4 and A-2, natural products that have replaced the tropolone ring with a phenyl ring at the C-ring position (Lin et al., 1989). This model suggests a symmetry to the colchicine site that appears to be inconsistent with the head-to-tail asymmetry of the microtubule (see a complete discus- sion in Andreu et al., 1991). It is not clear from this report if these compounds induce large confor- mational changes in tubulin or alternate polymer formation. GTPase activity was assessed in 1.0M glutamate, 4% DMSO under conditions where the drugs inhibit turbidity development but stimulated GTPase activity above the microtubule associated levels under otherwise identical (drug-free) con- ditions. The half-lives of colchicine displacement of these compounds from tubulin are much shorter
`
`IMMUNOGEN 2291, pg. 6
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`Tubulin-nucleotide interactions 133 than for colchicine (3.6 and 2.4 vs 405 min) but also longer than that for 2-methoxy-5-(2',3',4'- trimethoxy)tropone, MTPT, a tropolone analog lack- ing the B ring (12-17 sec). Direct measurements of binding constants and rates of binding will greatly assist in evaluating the implications of these unusual colchicine analogs. Since colchicine turns on the GTPase activity of the heterodimer, possibly mediated by small oligomer interactions (Heusele and Carlier, 1981), this will necessarily change the interactions between divalent cations, guanine nucleotides, and tubulin. The affinity of the GTPMg complex and of GDP may be altered. The conformation of the complex is clearly different, at least during hydrolysis where the triphosphate takes on the activated complex conformation. The tran- sition state probably inv