Report

Microtubule Disassembly by ATP-Dependent Oligomerization of the AAA Enzyme Katanin

See allHide authors and affiliations

Science  22 Oct 1999:
Vol. 286, Issue 5440, pp. 782-785
DOI: 10.1126/science.286.5440.782

Abstract

Katanin, a member of the AAA adenosine triphosphatase (ATPase) superfamily, uses nucleotide hydrolysis energy to sever and disassemble microtubules. Many AAA enzymes disassemble stable protein-protein complexes, but their mechanisms are not well understood. A fluorescence resonance energy transfer assay demonstrated that the p60 subunit of katanin oligomerized in an adenosine triphosphate (ATP)– and microtubule-dependent manner. Oligomerization increased the affinity of katanin for microtubules and stimulated its ATPase activity. After hydrolysis of ATP, microtubule-bound katanin oligomers disassembled microtubules and then dissociated into free katanin monomers. Coupling a nucleotide-dependent oligomerization cycle to the disassembly of a target protein complex may be a general feature of ATP-hydrolyzing AAA domains.

Microtubules, polymers of α- and β-tubulin subunits, form the mitotic spindle and organize membranous organelles in interphase cells. To accomplish these disparate functions, the microtubule cytoskeleton must rapidly reorganize into different configurations. Microtubules undergo spontaneous growth and shrinkage at their ends, even at steady state, which is important for the cellular rearrangements of these polymers (1, 2). In addition to end dynamics, the microtubule wall can be disrupted by the severing enzyme katanin (3). Potential in vivo roles for katanin-mediated microtubule severing include releasing microtubules from centrosomes (4), depolymerizing microtubule minus ends in the mitotic spindle as a component of poleward flux (5), and accelerating microtubule turnover at the G2/M transition of the cell cycle by creating unstable microtubule ends (6).

Katanin is a microtubule-stimulated ATPase, and ATP hydrolysis is required for it to sever and disassemble stable microtubules (3). Katanin is a heterodimer organized into a 60-kD enzymatic subunit (p60), which carries out the ATPase and severing reactions, and a targeting subunit (p80), which localizes katanin to the centrosome (7). The sequence of p60 reveals that it belongs to the AAA ATPase superfamily, members of which share one or two copies of a conserved 230-amino acid ATPase domain (8–10). AAA proteins have been implicated in a myriad of cellular processes as diverse as membrane targeting (NSF, VPS4, p97), organelle biogenesis (PAS1p), proteolysis (SUG1), and transcriptional regulation (TBP1) (11). AAA proteins have been proposed to act as nucleotide-dependent chaperones that can disassemble specific protein complexes or unfold polypeptides (8). However, little is known about how changes in the nucleotide state of the AAA domain are coupled to the disassembly of their protein targets.

Like the AAA protein NSF (12, 13), p60 katanin can form 14- to 16-nm rings, as shown by electron microscopy (7). However, p60 (14) and GFP-p60, a chimeric protein composed of green fluorescence protein and p60, migrated primarily as 4S monomers in 10% to 35% glycerol gradients in the presence of ATP, adenosine diphosphate (ADP), or adenosine-5′-(γ-thio)triphosphate (ATP-γ-S) (Fig. 1A) (15). GFP-p60 also migrated as a monomer by gel filtration (Stokes radius 66 Å). As a control, an NSF AAA domain (D2) migrated at 8S (Fig. 1A), the size expected for a hexamer of 30-kD AAA subunits (12,13). Therefore, in contrast to NSF, p60 does not form stable hexameric rings. Rather, the hydrodynamic and electron microscopy data taken together suggest that p60 monomers and oligomers exist in a reversible equilibrium and that p60 hexameric rings may not be stable to the time and dilution effects of sedimentation and gel filtration.

Figure 1

Oligomerization of p60 katanin. (A) Hydrodynamic analysis. Sedimentation profile of GFP-p60 katanin through 10% to 35% glycerol gradients (15) in the presence of 2 mM MgATP (closed circles) or MgADP (open circles). Both sediment as a single species of about 4S. GFP-p60E334Q(22), an active site mutant, was tested in the presence of 2 mM MgATP (closed triangles) and MgADP (open triangles). GFP-p60E334Q sediments as a mixture of 4S and 15S species in the presence of ATP but as a single species of 4S in the presence of ADP. As a control for AAA oligomerization, we sedimented an NSF AAA domain (D2) (12) through gradients containing 2 mM MgATP, and the sedimentation peak is indicated by an arrow. (B) FRET of a 1:9 mixture of CFP-p60E334Q (donor) and YFP-p60E334Q (acceptor) in the presence of 2 mM MgATP (closed circles) or 2 mM MgADP (open circles) (22). FRET is indicated by the decreased emission at 480 nm and the increased emission at 535 nm in the presence of MgATP. The MgADP emission profile is identical to that calculated for CFP-p60E334Q and YFP-p60E334Q measured separately (14).

We examined the dynamics of katanin ring formation in solution as well as in the presence of its microtubule substrate by using a fluorescence resonance energy transfer (FRET)–based assay. To achieve stoichiometric labeling at a defined location on the p60 molecule, we fused p60 to either cyan fluorescent protein (CFP) or yellow fluorescent protein (YFP) as a donor-acceptor pair (16,17). The half-maximal energy transfer distance,R 0, for the CFP and YFP pair is about 5 nm (18), which is similar to the intrasubunit distances within the AAA ring (9, 10). To test this FRET assay, we prepared an ATP active site mutant (E334Q) (19) of CFP-p60 and YFP-p60, which was designed to block nucleotide hydrolysis and trap the enzyme in the ATP-bound state (20). An equivalent mutation abolishes the ATPase and membrane fusion activities of NSF (21) and promotes oligomerization of VPS4, a single AAA domain protein involved in vacuolar targeting (22). As expected, p60E334Q had no detectable ATPase activity. When we mixed CFP-p60E334Q and YFP-p60E334Q fusion proteins in the presence of ADP, no energy transfer occurred and the emission of the CFP-YFP mixture was identical to that when the proteins were tested separately (14). However, in solutions that contained ATP, the mixture of CFPp60E334Q and YFP-p60E334Q showed a reduced CFP emission and correspondingly enhanced YFP emission, which is indicative of FRET (Fig. 1B) (23). This result indicates that p60E334Q subunits oligomerize when they are complexed with ATP.

To confirm the conclusion from the above FRET experiment, we determined the oligomeric state of CFP-p60E334Q by hydrodynamic analysis. In the presence of ATP, CFP-p60E334Q sedimented at 4S and 15S in glycerol gradients (Fig. 1A) (15). The 15S complex dissociated to 4S monomers when incubated with 2 mM ADP. To determine the oligomeric state of the 15S complex, we performed gel filtration of CFP-p60E334Q in the presence of ATP, which yielded a major peak with a Stokes radius of 8.6 nm (24). The Stokes radius and sedimentation coefficient predict a molecular mass of 520 kD, consistent with CFP-p60E334Q forming a hexamer of 90-kD subunits in the presence of ATP. These results agree with the FRET measurements, which also showed ATP-dependent oligomerization of p60E334Q.

We tested microtubules and ATP analogs for their ability to promote oligomerization of wild-type p60. We used the poorly hydrolyzable nucleotide, ATP-γ-S, to mimic the ATP state because it inhibits katanin ATPase activity (3) and because both ATP and ATP-γ-S supported similar amounts of FRET in p60E334Q(14). Little or no energy transfer occurred in the absence of microtubules regardless of the nucleotide present (Table 1). However, we observed a substantial increase in FRET when we incubated p60 with microtubules and ATP-γ-S but not ADP (Table 1) (23). When p60 was not bound to nucleotide (apyrase added), we observed a result similar to that with ADP (14). Hence, both nucleotide (ATP) and substrate (microtubules) cooperate in stimulating oligomerization of p60.

Table 1

Effect of nucleotides and microtubules on CFP-p60–YFP-p60 FRET. MgATP, MgADP, and MgATP-γ-S were present at 2 mM, and microtubules were included at 5 μM where indicated. The FRET signal (23) for ADP did not increase at higher (20 μM) microtubule concentrations. FRET values are normalized by using the ADP value as 100% (1.13 for p60E334Q, 0.46 for p60wt). The FRET signals for the p60E334Q and p60wt experiments cannot be directly compared because they were done with slightly different donor/acceptor ratios (p60E334Q, 0.13 μM total; 1:5 ratio of donor CFP to acceptor YFP; p60wt, 0.5 μM total; 1:2 ratio of donor CFP to acceptor YFP). The mean and standard deviation of two measurements (p60wt) or three measurements (p60E334Q) are shown.

View this table:

To determine the consequences of hexameric ring formation on the interaction of katanin with microtubules, we used a microtubule cosedimentation assay. GFP-p60 bound to microtubules with high affinity (K d ∼ 0.9 μM) in the presence of ATP-γ-S, whereas the affinity was reduced toK d ∼ 18 μM in the presence of ADP (Fig. 2) (25). High-affinity binding may require the microtubule polymer because oligomeric CFP-p60E334Q did not elute with monomeric tubulin on gel filtration columns (14). We also assessed the influence of the AAA domain of p60 on microtubule binding. When the AAA domain was deleted, the resultant protein (p60ΔAAA) (26) still cosedimented with microtubules, indicating that the NH2-terminus comprised a microtubule binding domain. Similarly, protein targeting by NSF also involves its NH2-terminal domain (27). However, p60ΔAAA binding was nucleotide-insensitive (14), and its affinity was in between GFP-p60 in its ATP- and ADP-bound states. Thus, the AAA domain of p60 affects the binding affinity of the adjacent microtubule binding domain, and tight binding occurs in nucleotide states (ATP or ATP analogs) that stabilize p60 rings.

Figure 2

Nucleotide-dependent binding of p60 katanin to microtubules. Cosedimentation of GFP-p60 and microtubules was tested in the presence of 2 mM MgATP-γ-S (closed circles) (K d ∼ 0.9 μM) or MgADP (open circles) (K d ∼ 18 μM) (25). p60ΔAAA is a truncated p60 that lacks the COOH-terminal AAA domain (closed triangles) (K d ∼ 6 μM) (26). Binding of this protein was not affected by nucleotide (14). Binding is expressed as the fraction of GFP-p60 that cosedimented with microtubules, and the best fit to a hyperbolic curve is shown.

Native as well as baculovirus-expressed katanin display an unusual microtubule-stimulated ATPase reaction in which the activity peaks at a microtubule concentration of 2 to 10 μM tubulin dimer and then decreases as the microtubule concentration is further increased (7). This differs from the expected Michaelis–Menten hyperbolic stimulation that, for example, is typical of microtubule motor proteins (28). One explanation for this behavior is that ATPase activation is driven by hexamer formation and the degree of oligomerization is determined by a competition between p60-p60 and p60 monomer–microtubule associations. To test this possibility, we determined the FRET and ATPase activities of p60 as the microtubule concentration was increased (29). Both ATPase activity and FRET increased together as the microtubule concentration was increased and then declined in a similar manner at higher microtubule concentrations (Fig. 3A) (30). In agreement with the ATPase measurements, microtubule disassembly by katanin was inhibited at a high microtubule-to-katanin ratio (Fig. 3B) (31). These results indicate that microtubules may stimulate the activity of p60 by facilitating p60-p60 interactions. Conversely, high concentrations of microtubules may reduce the ATPase and severing activities by preventing p60-p60 associations through the sequestration of p60 monomers at discontiguous, low-affinity binding sites on the microtubule. The data in Fig. 3 also revealed that release of a tubulin subunit from the microtubule wall requires the hydrolysis of, on average, about 50 ATP molecules. This coupling ratio is similar to that of the chaperone GroEL, which hydrolyzes 50 to 150 ATPs per renaturation of a misfolded protein (32).

Figure 3

Effect of microtubules on p60 oligomerization, ATPase, and microtubule severing activities. (A) Oligomerization and ATPase activity as a function of microtubule concentration. ATPase activity (closed circles) (29) and FRET (open circles) (23) were measured in a 0.2 μM, 1:5 mixture of CFP-p60 and YFP-p60. Values have been normalized to the percentage of the maximum observed FRET or ATPase signal. A curve fit is shown for two competing Michaelis–Menten reactions ({(A × [tubulin])/(B + [tubulin])} − {(C × [tubulin])/(D + [tubulin])}). (B) Comparison of ATPase (hatched bars) (29) and microtubule-disassembly activity (solid bars) (31) of 0.2 μM p60 at 5 and 10 μM microtubules. ATPase activity begins to decline above about 2 μM microtubules for untagged p60 (7) and above about 10 μM microtubules for CFP-p60–YFP-p60 (Fig. 3A). We used untagged p60 for this assay because microtubule concentrations > 10 μM are not compatible with the fluorescence microtubule disassembly assay. Activities have been normalized to activity at 5 μM, and error bars indicate standard deviation of two measurements. Maximum activity was 1.9 ATP p60−1 s−1 and 0.04 tubulin dimer p60−1 s−1, yielding a coupling ratio of about 50 ATP per tubulin dimer removed from the microtubule.

The above results suggest a model for how katanin disrupts tubulin contacts within a microtubule wall (Fig. 4). Katanin-ADP is monomeric, but nucleotide exchange for ATP enhances p60-p60 affinity. Oligomerization is most efficient, however, in the presence of its protein substrate, which suggests that microtubules act as a scaffold for promoting oligomerization. The p60 ring then binds to microtubules with high affinity, potentially as a result of forming multiple tubulin contacts. Once katanin oligomers assemble on the microtubule, ATPase activity is stimulated. Nucleotide hydrolysis and subsequent phosphate release could change the conformation of the katanin ring, leading to mechanical strain that destabilizes tubulin-tubulin contacts (Fig. 4). Consistent with this idea, large conformational changes have been observed for the NSF ring in its ATP and ADP states (12). A concerted conformational change also occurs for the chaperone GroEL, which binds misfolded polypeptides at multiple sites within its seven-membered ring and undergoes large interdomain motions that strain the bound polypeptide (33,34). Alternatively, as described for the microtubule-destabilizing kinesin XKCM1 (35), tight binding of katanin oligomers could strain tubulin-tubulin contacts, with ATP hydrolysis serving to dissociate katanin-tubulin dimers from this stable complex. In either scenario, ATP hydrolysis also serves a recycling function because p60-p60 and p60-tubulin interactions both weaken in the ADP-bound state, dissociating tubulin and releasing p60-ADP to begin a new round of disassembly. This proposed cycle has similarities to that of dynamin, which self-assembles into a spiral pattern on endocytic membrane tubules, changes conformation after hydrolysis of guanosine triphosphate in a manner that vesiculates the tubule, and then disassembles in the guanosine diphosphate state (36).

Figure 4

Model for microtubule severing by katanin. See text for detail of the mechanism. For simplicity, only a single protofilament of the microtubule is shown. T, DP, and D represent ATP, ADP + Pi, and ADP states, respectively. The relatively low affinity of katanin for nucleotide suggests that exchange of ATP for ADP would occur rapidly in solution. The conformational change is shown to occur with γ-phosphate bond cleavage, although this could also occur as a result of γ-phosphate release.

The oligomerization cycle described for katanin also may occur in many other AAA enzymes. ATP enhances oligomerization of VPS4, a single AAA domain protein, and it was proposed that this oligomerization could be further facilitated by an as yet unidentified membrane-associated target (22). Such reactions could be tested by the FRET-based assay described here. In contrast to katanin and VPS4, NSF is a constitutive hexamer (13) because of the presence of an additional nonhydrolytic AAA domain (D2) (21). Our results, however, raise the possibility that the ATP-hydrolyzing AAA domains (D1) may undergo cycles of tight and weak interactions while remaining tethered through their D2 domains. Thus, nonhydrolyzing AAA domains may serve as anchors to keep the enzymatic subunits in close proximity throughout the hydrolytic cycle.

  • * To whom correspondence should be addressed. E-mail: vale{at}phy.ucsf.edu

REFERENCES AND NOTES

View Abstract

Stay Connected to Science

Navigate This Article