Katanin controls mitotic and meiotic spindle length

Control of spindle length is important for different reasons during mitotic and meiotic cell divisions. Mitotic spindles of most animals and fungi lengthen in a process called anaphase B. This spindle lengthening provides a mechanism for moving sister chromatids apart that is redundant with anaphase A. The meiotic spindles of animal oocytes, however, are different because they mediate highly asymmetric cell divisions in which chromosomes are discarded in nonviable division products called polar bodies. The female meiotic spindles of mouse do not lengthen during anaphase, and a mutation causing anaphase B–like spindle elongation results in a large polar body and a correspondingly small egg (Verlhac et al., 2000). This result indicates that restriction of meiotic spindle length may be important for preserving egg volume and egg contents required by the embryo. Indeed, the meiotic spindles of Caenorhabditis elegans zygotes actually shorten in a coordinated, anaphase-promoting complex-dependent manner during meiosis I and II (Yang et al., 2003, 2005).

Results published during the last 10 yr indicate that a discrete set of spindle-lengthening and spindle-shortening mechanisms can control overall spindle length. In many organisms, the plus ends of kinetochore fiber microtubules polymerize at the same rate that the minus ends depolymerize (Mitchison, 1989). When plus-end polymerization is blocked with taxol (Waters et al., 1996) or by depletion of the kinetochore protein, CLASP (CLIP170-associated protein; Maiato et al., 2005), spindles shorten because of continued minus-end depolymerization. Conversely, when minus-end depolymerization is blocked experimentally, spindles can elongate continuously (Rogers et al., 2004). In addition, Drosophila melanogaster spindles elongate excessively in the absence of the kinetochore-associated microtubule depolymerase, KLP67A (Gandhi et al., 2004; Goshima et al., 2005). Another major lengthening mechanism is outward sliding of overlapping antiparallel microtubules mediated by kinesin-5 family members (Sharp et al., 1999; Kapitein et al., 2005). In unperturbed D. melanogaster embryonic spindles, a cell cycle–regulated cessation of minus-end depolymerization coincides with the initiation of anaphase B spindle elongation. In this case, the rate of kinesin-5–driven outward sliding is matched by the rate of minus-end depolymerization until this depolymerization stops (Brust-Mascher and Scholey, 2002). Evidence has also been presented for two other spindle-shortening mechanisms, inward sliding of overlapping antiparallel microtubules by kinesin-14 family members (Mountain et al., 1999; Sharp et al., 1999) and an undefined “tensile element” that squeezes inward and buckles microtubule bundles during nocodazole-induced spindle shortening (Mitchison et al., 2005).

None of these mechanisms appear to be universally used for spindle-length control. For example, cessation of minus-end depolymerization does not alter spindle length in human U2OS cells (Ganem et al., 2005) and minus-end depolymerization does not occur at all in some fungal cells (Mallavarapu et al., 1999; Maddox et al., 2000). Separation of spindle poles occurs in C. elegans mitotic embryos even after laser cutting of spindle microtubules (Grill et al., 2001). This outward movement of spindle poles is thought to be driven by a cortical motor protein pulling on astral microtubules that extend from the spindle pole toward the cortex. Thus, C. elegans mitotic spindles can “elongate” in the absence of an outward sliding mechanism. This may explain why the sole kinesin-5 in this species is not essential (Bishop et al., 2005). In contrast, budding yeast mitotic spindles can elongate in the complete absence of cortical pulling forces (Sullivan and Huffaker, 1992), and this species' kinesin-5 family members are essential (Hoyt et al., 1992).

The goal of this study was to determine whether the microtubule-severing protein, katanin, has a conserved role in spindle-length control. Katanin is a heterodimeric protein consisting of an AAA ATPase subunit called p60 and an accessory subunit called p80. The ATPase subunits from sea urchin (Hartman et al., 1998) and human (McNally et al., 2000) can sever microtubules on their own, but this activity is enhanced by the p80 subunit. Katanin's microtubule-severing activity and its conserved localization at the spindle poles of sea urchins (McNally et al., 1996) and vertebrates (McNally and Thomas, 1998) make it a strong candidate for playing a role in spindle-length control.

We previously reported that inhibition of katanin in mammalian cells causes a change in the distribution of γ-tubulin and a reduction in the rate of nocodazole-induced spindle microtubule depolymerization as assayed by decreasing fluorescence intensity of YFP-tubulin (Buster et al., 2002). Here, we report that katanin inhibition slows the rate of nocodazole-induced spindle shortening in mitotic mammalian cells and blocks the second of two meiotic spindle shortening phases in C. elegans.

Previous studies have demonstrated that inhibition of microtubule polymerization with taxol (Waters et al., 1996) or nocodazole (Mitchison et al., 2005) can induce mitotic spindle shortening and thus reveal spindle-shortening forces that were balanced by polymerization-dependent spindle-lengthening forces before drug treatment. To test whether katanin contributes to inward forces in the mitotic spindle, we measured the velocity of nocodazole-induced spindle shortening in cells expressing YFP-tubulin and CFP fusions to each of two katanin inhibitors. We previously found that these inhibitors slowed the rate of nocodazole-induced microtubule depolymerization as assayed by the rate of decrease in YFP-tubulin fluorescence intensity but did not quantify spindle-shortening velocities (Buster et al., 2002). Nocodazole-induced spindle shortening was biphasic with an early, fast phase and a later, slow phase. As shown in Table I, both con80, a p80 katanin fragment that displaces endogenous p60 katanin from spindle poles (McNally et al., 2000), and Ploop-p60, a point mutant of p60 katanin that inhibits the in vitro microtubule-severing activity of wild-type katanin (Buster et al., 2002), slowed both the early and late phases of nocodazole-induced spindle shortening relative to control cells expressing CFP alone. In contrast, a CFP fusion to a p60 katanin subunit that does not inhibit severing by wild-type katanin, ΔN-Ploop-p60 (Buster et al., 2002), did not slow the rate of spindle shortening relative to control cells expressing CFP alone. These results indicate that katanin-mediated microtubule severing promotes spindle shortening when polymerization is blocked. We propose that microtubule severing in untreated spindles generates arrays of short, overlapping microtubules near spindle poles. Nocodazole-mediated depolymerization would lead to a rapid loss of microtubule overlap that is required for the kinesin 5–mediated outward sliding that opposes spindle-shortening mechanisms. In the absence of microtubule-severing activity at spindle poles, microtubules would be longer, overlap zones would be longer, and more time would be required for nocodazole to cause a loss of overlap (see Discussion).

To test the generality of katanin's role in spindle-length control and to examine spindle shortening in the absence of drugs, we continued our analysis with C. elegans female meiotic spindles. These spindles were chosen because they are the only spindles that have well-documented and dramatic shortening phases (Yang et al., 2003) and they are the only spindles that have a demonstrated requirement for a katanin homologue, MEI-1–MEI-2 (Mains et al., 1990; Srayko et al., 2000).

In wild-type C. elegans, MI and MII spindles shorten from steady-state metaphase lengths of 7.7 and 6.2 μm, respectively, to minimum lengths of 2.8 and 2.4 μm (Yang et al., 2003; Table II and Fig. 1). After reaching minimal length, the spindles narrow and lengthen to form a midbody that extends into the polar body. As shown in Fig. 1 A, spindle shortening could occur by either of two general mechanisms, inward sliding of antiparallel microtubules or depolymerization of microtubule minus ends at spindle poles. In a sliding-only mechanism, microtubule density increases as the spindle shortens. If spindle shortening proceeded only by minus-end depolymerization or by a combination of inward sliding and plus-end depolymerization, microtubule density would not increase.

To distinguish between these possibilities, changes in microtubule density were monitored by obtaining the mean fluorescence intensity of GFP∷tubulin (mean pixel value) in the spindle throughout meiosis (see Materials and methods). In the example shown in Fig. 1 B, GFP∷tubulin fluorescence intensity remained constant when spindle length was constant (Fig. 1 B, 0–1.5 min) but began to increase dramatically as the spindle shortened (Fig. 1 B, 1.5–5 min). This increase in microtubule density abruptly switched to decreasing microtubule density when spindle length reached 56% of its starting metaphase length (Fig. 1 B, 5–7.5 min). The switch between increasing and decreasing microtubule density began within 1 min of spindle rotation from parallel to perpendicular relative to the cortex. We previously demonstrated that homologue separation initiates 0.7 min after spindle rotation (McNally and McNally, 2005). After the transition from increasing to decreasing density, the spindle continued to shorten to 33% of its starting length. Similar results, obtained in 12 out of 12 experiments (Table II) using wild-type C. elegans during either MI or MII, indicated that there are two sequential and mechanistically distinct phases of meiotic spindle shortening. The first phase most likely occurs by inward sliding of antiparallel microtubules that stops just before rotation, perhaps because at that point the microtubules are completely overlapped. Microtubule density increases uniformly in all parts of the spindle during the first phase (unpublished data), possibly because the microtubule overlap zones are heterogeneous in length and position within the spindle. The second shortening phase occurs after rotation, during chromosome segregation, and is accompanied by net microtubule depolymerization.

To test the role of microtubule severing in meiotic spindle shortening, we chose a partial-loss-of-function mutation in the p80 katanin homologue, mei-2, because null mutants of mei-2 or the p60 katanin homologue, mei-1, do not assemble bipolar spindles (Mains et al., 1990), and failure of these spindles to shrink (Yang et al., 2003) might be an indirect result of a lack of antiparallel microtubules. The missense mutant mei-2(ct98) has a reduced amount of MEI-2 protein (Srayko et al., 2000). We tested the in vitro microtubule-severing activity of purified MEI-1 for the first time (Fig. 2) and found that MEI-2 is absolutely required for the microtubule-severing activity of MEI-1. Thus, MEI-1 differs from katanin catalytic subunits from sea urchin (Hartman et al., 1998), human (McNally et al., 2000), or Arabidopsis thaliana (Stoppin-Mellet et al., 2002), each of which sever microtubules on their own. The MEI-2 dependence of MEI-1 suggests that mei-2(ct98) meiotic spindles should have a reduced amount of microtubule-severing activity. mei-2(ct98) worms were previously reported to lay viable embryos with large polar bodies (Mains et al., 1990), suggesting that these worms retain sufficient MEI-1–MEI-2 activity to assemble bipolar spindles but that these spindles might be unusually long at polar body induction.

Time-lapse imaging of a mei-2(ct98) strain expressing GFP∷tubulin revealed that both the maximum and the minimum meiotic spindle lengths were considerably greater than in wild type (Table II). Wild-type spindles maintain a constant metaphase length before initiating shortening (Yang et al., 2003, 2005). If this steady-state length were the result of a balance between katanin-dependent shortening forces and katanin-independent lengthening forces, one might expect to see a period of spindle lengthening during metaphase in mei-2(ct98) embryos. Instead, mei-2(ct98) spindles maintained a constant length before initiating shortening in nine out of nine embryos that remained parallel to the plane of focus for 3 or more minutes during metaphase. This suggests that the increased maximum length of mei-2(ct98) spindles is a result of a defect during spindle assembly rather than a change in a balance of forces during metaphase.

The abnormally long metaphase spindles of mei-2(ct98) embryos shortened at a wild-type velocity of −0.95 μm/min to 6.2 μm for MI or 5.8 μm for MII during a period of increasing microtubule density but maintained a constant length as microtubule density decreased (Fig. 3 and Table II). The percentage of spindle shortening in mei-2(ct98) mutants (MI, 64%; MII, 63%) is similar to the percentage of shortening that is observed for wild-type spindles at the switch to decreasing microtubule density (MI, 58%; MII, 63%; see Table II), a transition that occurs nearly concurrently with spindle rotation. At this point, the velocity in mei-2(ct98) drops to −0.01 μm/min compared with −0.74 μm/min for wild type, and consequently the mutant spindle length remains at 65% of maximum, whereas that of wild type decreases to 38%. These results suggests that mei-2(ct98) spindles undergo the initial, sliding-dependent phase of shortening but do not undergo the second phase of shortening, which occurs after rotation in wild-type embryos. Consistent with this interpretation, chromosome separation in mei-2(ct98) spindles initiated within 1 min after spindle shortening ceased or microtubule density began decreasing (n = 3), just as separation initiated within 1 min after rotation in wild type (n = 14; McNally and McNally, 2005).

In contrast to wild-type spindles, which rotated at the time that microtubule density switches from increasing to decreasing, mei-2(ct98) spindles did not rotate (Fig. 3). The spindle remained parallel to the cortex as the cortex invaginated and, in the example shown in Fig. 3, only later turned toward the cell interior by the invagination of the cortex. The greatest angle at which mei-2(ct98) spindles interacted with the cortex was highly variable in contrast to wild-type spindles, which rotated to a near perfect 90° angle (Table II). Although it is possible that the MEI-2 gene product is directly required for spindle rotation, a second hypothesis is that rotation can only occur when the spindle has shortened to a critical length. Polar bodies in mei-2(ct98) embryos were larger and more variable in size than in wild-type embryos (Table II), suggesting that the wild-type function of spindle shortening and rotation is to minimize polar body size and thus preserve embryo volume.

To more carefully compare changes in microtubule density during the late stage of spindle shortening, we used a strain expressing both GFP∷tubulin and mCherry∷histone to monitor spindle changes initiating at anaphase onset. Changes in the fluorescence intensity profiles down the pole–pole axis of a wild-type and a mei-2(ct98) spindle are shown in Fig. 4. In wild-type spindles, microtubule density in the spindle poles decreased, whereas microtubule density increased between the separating chromosomes. In contrast, mei-2(ct98) spindles exhibited a nearly uniform microtubule density down their pole–pole length and the relative density at the poles versus the midzone did not change over time. Similar results were obtained in 11 out of 11 wild-type embryos and 13 out of 13 mei-2(ct98) spindles. Thus, the microtubule-severing activity of MEI-1–MEI-2 katanin is required for the redistribution of microtubules from the spindle poles to the spindle midzone. Both wild-type (Fig. 1) and mei-2(ct98) spindles (Fig. 3) decrease in overall microtubule density after anaphase onset, indicating the additional action of katanin-independent microtubule depolymerizers in the reduction of overall microtubule density.

It could be that MEI-1–MEI-2 located at spindle poles is suddenly activated at anaphase onset, and its microtubule-severing activity promotes disassembly of spindle pole microtubules directly. It is also plausible that microtubule-severing activity during spindle assembly generates a wild-type spindle composed of overlapping short microtubules that are transported from the poles to the midzone after anaphase onset. In this case, mei-2(ct98) spindles would fail to undergo normal postanaphase morphogenesis simply because they were composed of excessively long microtubules. To test when MEI-1–MEI-2 katanin might be acting, we examined the localization of GFP∷MEI-1 by time-lapse imaging (Fig. 5 and Fig. S2). GFP∷MEI-1 was localized uniformly on spindle microtubules during early stages of MI spindle assembly (not depicted) and subsequently concentrated on both spindle poles and on chromosomes (Fig. 5). The metaphase localization pattern was similar to that previously shown by immunofluorescence for both MEI-1 (Clark-Maguire and Mains, 1994a) and MEI-2 (Srayko et al., 2000). GFP∷MEI-1 fluorescence intensity was the most intense and focused just after spindle rotation and when chromosomes had just reached the poles (Fig. 5). After this stage, GFP∷MEI-1 fluorescence decreased substantially as the midbody lengthened (Fig. 5 and Fig. S2) and was then associated predominantly with chromosomes. The presence of GFP∷MEI-1 at spindle poles just after rotation is consistent with a direct role of microtubule severing in the second phase of spindle shortening, and the decreasing intensity of GFP∷MEI-1 at the end of this phase is consistent with the decreasing density of microtubules at spindle poles (Fig. 4).

We previously reported that mei-1(null) spindles do not undergo timely translocation to the cortex (Yang et al., 2003). This translocation defect might indicate a direct role for microtubule severing in kinesin-dependent spindle translocation (Yang et al., 2005) or it might be an indirect consequence of the lack of antiparallel microtubule organization in mei-1(null) spindles. To test whether the spindle translocation defect in mei-1(null) mutants is simply a consequence of disorganized spindles, spindle translocation was examined in klp-18(RNAi) worms expressing either GFP∷tubulin or GFP∷histone. KLP-18 was chosen for this analysis because depletion of this motor was previously shown to have meiosis-specific spindle organization defects (Segbert et al., 2003). As shown in Table III, klp-18(RNAi) spindles, like wild-type spindles, translocated to the cortex either before the embryo exited the spermatheca or within 2 min of spermatheca exit. As previously reported (Yang et al., 2003), mei-1(null) spindles reached the cortex 9.9 ± 2.4 min after spermatheca exit. Surprisingly, this strong translocation defect in mei-1(null) mutants was completely suppressed by RNAi-mediated depletion of KLP-18 (Table III). These results indicate that (1) bipolar spindle organization is not required for timely spindle translocation, (2) the translocation defect in mei-1(null) mutants is a consequence of a KLP-18–dependent process, and (3) katanin is not directly required for spindle translocation. We suggest that in the absence of katanin, rigid, KLP-18–dependent microtubule bundles extend from the disorganized meiotic spindles and sterically inhibit movement toward the cortex.

Although the spindle translocation defect in mei-1(null) mutants was suppressed by klp-18(RNAi), bipolar spindles were still not assembled in mei-1(null) klp-18(RNAi) double mutants, indicating a direct role for katanin in spindle assembly. One possible role for katanin in spindle assembly was suggested by work on the plant cortical microtubule cytoskeleton, where katanin is required for formation of organized parallel/antiparallel microtubule bundles (Burk et al., 2001), just as it is in C. elegans meiosis. Plant cortical microtubules are nucleated from microtubule-associated γ-tubulin complexes at a discrete 42° angle from preexisting microtubules (Murata et al., 2005). It is intriguing to speculate that katanin-mediated severing might be required to release microtubules from these branched networks to allow parallel bundling. If this were the case with acentrosomal C. elegans meiotic spindles, then γ-tubulin (RNAi) might partially suppress the need for katanin in spindle assembly. An alternative hypothesis is based on the fact that microtubule severing and microtubule nucleation both have the same net effect of increasing microtubule number. In this hypothesis, katanin and γ-tubulin might have partially overlapping roles in increasing microtubule number early in spindle assembly. In this case, γ-tubulin (RNAi) would be expected to worsen the phenotype of a mei-1(null) mutant.

To test these hypotheses, GFP∷tubulin-expressing worms were fed bacteria expressing tbg-1 (γ-tubulin) or gip-1 (γ-tubulin ring complex [γ-TuRC] subunit) double-stranded RNA for 40–44 h. This amount of time is required to deplete maternal protein from the syncytial gonad. Longer feeding regimens resulted in oogenesis defects and failures in ovulation. Meiotic spindle morphology appeared wild type in tbg-1(RNAi) or gip-1(RNAi) worms (Fig. 6), consistent with previous reports of normal polar body size and number (Strome et al., 2001). Microtubule density, as determined from the ratio of mean GFP∷tubulin fluorescence in the spindle to fluorescence in the cytoplasm, was similar between wild-type, mei- 1(null), tbg-1(RNAi), and gip-1(RNAi) spindles but was dramatically reduced in mei-1(null); tbg-1(RNAi) or mei-1(null); gip-1(RNAi) double-mutant spindles (Fig. 6). This synthetic effect was not observed in mei-1(null); klp-18(soaking RNAi) double-mutant worms (Fig. S3, available), indicating that the enhancement was specific. Neither tbg-1(RNAi) nor gip-1(RNAi) suppressed the spindle morphology defect in mei-1(null) worms. Because tbg-1(RNAi) and gip-1(RNAi) enhanced rather than suppressed the spindle morphology defect in mei-1(null) worms, these results suggest that MEI-1 and γ-tubulin play redundant roles rather than antagonistic roles in promoting increasing microtubule density during spindle assembly. In addition, because the mei-1(null) mutant completely lacks function, we can conclude that MEI-1 and γ-tubulin are acting in parallel rather than sequentially.

If this hypothesis were correct, then MEI-1 and γ-tubulin should both be present at the time and place that meiotic spindle assembly initiates. As shown in Fig. 7 A, meiotic spindle assembly initiates at nuclear envelope breakdown, when a diffuse cloud of microtubules fills the volume of the perforated nucleus. These microtubules then organize into a much smaller assembly of dense microtubule bundles by the time the oocyte ovulates into the spermatheca (Fig. 7 A; Yang et al., 2003). By the time the embryo exits the spermatheca and enters the uterus, a bipolar, metaphase-length spindle is present within a vesicle-free zone visible by differential interference contrast (Fig. 7 C). Before NEBD, GFP∷TBG-1 became progressively more concentrated on the nuclear envelope as oocyte maturation progressed (Fig. 7 B). At nuclear envelope breakdown, GFP∷TBG-1 fluorescence entered the nuclear region, forming a diffuse cloud that might indicate association with either microtubules or vesicles derived from the nuclear envelope. This large, diffuse area of γ-tubulin enrichment did not contract in diameter nor increase in local intensity (Fig. 7 B), as did the α-tubulin (Fig. 7 A). By the time the embryo exited the spermatheca into the uterus, no discrete localization of GFP∷TBG-1 was observed in the vesicle-free zone that contains the metaphase I spindle (Fig. 7 D). GFP∷MEI-1 was also distributed throughout the volume of the perforated nucleus immediately after nuclear envelope breakdown (Fig. 7 F). GFP∷MEI-1 differed from GFP∷TBG-1 in that it became highly concentrated on chromosomes and spindle poles during spindle assembly and remained concentrated on these structures throughout meiosis (Fig. 7 F, Fig. 5, and Fig. S2). The transient presence of microtubules, MEI-1, and γ-tubulin within the perforated nucleus immediately after nuclear envelope breakdown is consistent with the hypothesis that microtubule severing and γ-TuRC–templated nucleation provide redundant mechanisms for increasing microtubule number during the initial stages of spindle assembly.

Our results demonstrate that inhibition of katanin inhibits spindle-shortening processes in fibroblast mitotic spindles and C. elegans meiotic spindles. Mitotic spindles of cultured mammalian cells do not normally shorten, but intrinsic shortening processes can be revealed when microtubule polymerization is blocked with nocodazole or taxol (Cassimeris and Salmon, 1991; Waters et al., 1996; Mitchison et al., 2005). These shortening processes are opposed by kinesin-5–dependent outward sliding forces (Brust-Mascher and Scholey, 2002), which depend on bundles of overlapping microtubules. We propose that untreated CV1 spindles are composed of discontinuous arrays of overlapping microtubules generated by a combination of microtubule severing near the poles and dynamic instability (Fig. 8 A; Buster et al., 2002). Short microtubules result in short overlap regions, which are lost rapidly during nocodazole treatment. The rapid loss of outward sliding forces allows rapid spindle shortening, which is possibly due to the inward squeezing “tensile element” (Mitchison et al., 2005). In this scenario, spindles are composed of more contiguous long microtubules after katanin inhibition, and these longer microtubules have long overlap zones that oppose spindle shortening forces for a longer time period during nocodazole treatment (Fig. 8 B).

C. elegans female meiotic spindles naturally undergo dramatic shortening cycles during each anaphase. Our results indicate that this shortening is initially driven by katanin-independent spindle shortening that is accompanied by increasing microtubule density and may be driven by inward sliding of antiparallel microtubules (Fig. 8 C). This phase is followed by a katanin-dependent phase in which microtubule density decreases at poles and increases in the midzone. We favor a model for late shortening in which katanin, localized at spindle poles, severs pole-proximal microtubules into short fragments that are then transported to the midzone (Fig. 8 E). The second phase of spindle shortening is also accompanied by a katanin-independent decrease in the overall density of spindle microtubules that may be caused by other microtubule-destabilizing proteins.

The increased steady-state metaphase spindle length in the partial-loss-of-function katanin mutant is consistent with a model in which wild-type meiotic spindles are assembled as an array of discontinuous, overlapping, short microtubules (Fig. 8 C), just as we propose for CV1 mitotic spindles (Fig. 8 A). A reduced rate of microtubule severing results in a longer spindle because mean microtubule length is longer. This increased microtubule length would limit the extent of spindle shortening by the early, inward sliding mechanism (Fig. 8, C and D; and Table II). Our model is qualitatively consistent with the EM tomography results of Srayko et al. (2006), who found many short microtubules in a wild-type spindle but only long microtubules in a mei-1(null) spindle. It is more difficult to resolve our model quantitatively with the EM tomography data, however, because more than half the microtubules in this tomogram extended beyond the tomogram and thus could not be measured.

The finding of numerous short microtubules in a wild-type meiotic spindle and the lack of short microtubules in a mei-1(null) spindle (Srayko et al., 2006) is also qualitatively consistent with our finding that katanin and γ-TuRC have partially redundant roles in promoting microtubule density increase during spindle assembly. Our results with GFP-tubulin fluorescence intensity measurements revealed microtubule densities that were similar between wild-type metaphase spindles and mei-1(null) spindles (Fig. 6). We propose that this difference is small because γ-TuRC–mediated nucleation can increase microtubule number in the absence of severing activity. This result is consistent with the small (25%) difference in polymer mass determined from fixed immunofluorescence analysis by Srayko et al. (2006).

Perhaps the most perplexing question remaining is why mei-1(null) mutants do not assemble extremely long bipolar spindles similar to what is seen with the partial-loss-of-function mei-2 mutant. We offer two hypotheses explaining the absolute requirement for katanin in C. elegans meiotic spindle assembly. In the first hypothesis, MEI-1–MEI-2 complexes have an additional, nonsevering function, such as cross-bridging antiparallel microtubules or recruiting other essential proteins to the spindle. In this scenario, mei-2(ct98) spindles actually have no microtubule-severing activity but can assemble bipolar spindles with a low concentration of MEI-1–MEI-2 cross-bridging activity. This cross-bridging activity may be more critical in noncentrosomal microtubule arrays. The nonlinear dependence of microtubule-severing activity on katanin concentration (Hartman and Vale, 1999) supports the notion that a reduced amount of MEI-2 could result in a complete loss of severing activity. In the second hypothesis, mean microtubule length increases as katanin concentration decreases until a critical stage is reached where microtubules are too long to allow spindle assembly. In the C. elegans meiotic embryo, the relative activities of other critical regulators of microtubule dynamics and motility would be optimized for assembling spindles from short microtubules. These other microtubule-binding proteins would then assemble aberrant structures from long microtubules.

The conserved localization of katanin at spindle poles of many species and the similarity of spindle-length phenotypes in CV1 cells and C. elegans meiotic spindles suggests that katanin regulates mitotic and meiotic spindle length in many species. The generality of katanin's relative role in the assembly of mitotic versus meiotic spindles remains to be elucidated. Our dominant-negative inhibitors did not prevent mitotic spindle assembly in CV1 cells (Buster et al., 2002); however, genome sequences reveal that human and mouse each have three distinct katanin catalytic subunits (unpublished data) as well as the related microtubule-severing ATPase, spastin. These different microtubule-severing proteins could easily have unique roles at different times in development.

CV1 cells with an integrated YFP-tubulin construct were transfected on 25-mm coverslips with Lipofectamine Plus (Invitrogen) using CFP-X plasmid constructs, which were described previously (Buster et al., 2002). 12–20 h after transfection, coverslips were assembled into perfusion chambers maintained at 37°C. Cells were imaged with a microscope (Microphot SA; Nikon) equipped with a 60× Plan Apo 1.4 objective, a charge-coupled device (CCD; Quantix KAF1400; Photometrics), and Ludl shutter controlled by IP Lab Spectrum software. Mitotic CFP-expressing cells were identified quickly using reduced illumination. A single CFP fluorescence image was captured using fixed neutral density filters, exposure, and gain so that the expression level could be estimated. 50 μl of culture medium containing 80 μM nocodazole was then added gently to the edge of the perfusion chamber and shuttered, and time-lapse acquisition of YFP fluorescence images was initiated. Diffusion of nocodazole to the imaged cell did not appear to be rate limiting, as the fastest changes in spindle length always occurred in the first 5 s of the image sequence.

A cDNA encoding the long isoform of MEI-1 (Clark-Maguire and Mains, 1994b) was cloned as a 6-his–tagged fusion along with an untagged MEI-2 cDNA into pFastBac Dual (Invitrogen), expressed in Sf9 cells using the Bac to Bac System (Invitrogen) and partially purified by Ni²⁺-chelate chromatography.

A 1:10 mixture of tetra-methy-rhodamine–labeled and unlabeled porcine brain tubulin was polymerized and stabilized with 20 μM taxol. Microtubules were immobilized on the surfaces of perfusion chambers at a final concentration of 0.1 μM tubulin dimer using an Escherichia coli–expressed G234A mutant of the first 560 amino acids of ubiquitous human kinesin heavy chain. Preparations of purified MEI-1 or MEI-1–MEI-2 complexes were perfused at final concentrations ranging from 0.5 to 5.7 μM in 20 mM K-Hepes, pH 7.5, 0.1 mM EGTA, and 2 mM MgSO₄ with ATP or ADP at 1.8 mM. Reactions were stopped at time intervals by perfusion with glutaraldehyde, and images of fixed microtubules were captured with a CCD.

In this study, wild type indicates one of several integrated transgenic strains. The integrated GFP∷tubulin strain WH204 (Strome et al., 2001) was used for the studies shown in Figs. 1 and 3 and Tables II and III. The integrated GFP∷tubulin strain AZ244 (Praitis et al., 2001) was consistently brighter than WH204 and was used for the studies shown in Figs. 5, 6, and 7. The integrated GFP∷histone H2b strain AZ212 (Praitis et al., 2001) was used in Fig. 5. The integrated GFP∷tubulin, mCherry∷histone strain, OD57, was the wild type used for Fig. 4. OD57 was generated by particle bombardment of a pie-1 promoter vector (Praitis et al., 2001) engineered to express a histone H2b fusion to a version of mCherry (Shaner et al., 2004) that included four C. elegans introns and C. elegans preferred codons. This modified mCherry DNA was synthesized by Molecular Cloning Laboratories. The mCherry∷histone strain was crossed with the GFP∷tubulin- expressing strain, OD4, to obtain OD57. mei-2(ct98) worms were obtained from P. Mains (University of Calgary, Calgary, Canada) and were crossed with WH204 to obtain mei-2(ct98) worms expressing GFP∷tubulin, with AZ212 to obtain mei-2(ct98) worms expressing GFP∷histone H2b and with OD57 to obtain mei-2(ct98) worms expressing both GFP∷tubulin and mCherry∷histone. The mei-1(null) allele used in this study was the nonsense mutant mei-1(ct46ct101) (Clark-Maguire and Mains, 1994b). Homozygous mei-1(ct46ct101) worms expressing GFP∷tubulin and used for the studies shown in Table III were described previously (Yang et al., 2003). For data shown in Fig. 6, homozygous mei-1(ct46ct101) worms expressing a higher level of GFP∷tubulin were derived from crosses between HR1069 (unc-13[e10910] daf-8[e1393] mei-1[ct46ct101]/hT2[bli-4{e937} let{h661}] I; hT2/+ III] and AZ244. The GFP∷MEI-1-expressing strain was EU1065 (Pintard et al., 2003). Western blots with anti–MEI-1 antibody indicated that this transgene is expressed at a much lower level than endogenous MEI-1 (not depicted), indicating that the localization in Fig. 5 is not due to overexpression. OD44, the GFP∷γ tubulin-expressing strain used in Fig. 7, was provided by A. Desai (University of California, San Diego, La Jolla, CA). mei-2(ct98) strains were maintained at 16°C and shifted to 20°C for 24 h before filming. EU1065 was maintained at 25°C to prevent germline silencing. All other strains were maintained at 20°C.

Adult hermaphrodites were anesthetized with Tricaine/tetramisole as described previously (Kirby et al., 1990; McCarter et al., 1999) and gently mounted between a coverslip and a thin agarose pad on a slide. Mineral oil or petroleum jelly was used to reduce evaporation at the edge of the coverslip. Images for Figs. 1, 2, and 3; histograms in Fig. 6; most of Fig. 7; and data for all tables were acquired on a microscope (Microphot SA; Nikon) as described for CV1 cell imaging. Images in Figs. 4 and 5, parts of Fig. 6, and Fig. S2 were acquired with a spinning-disk confocal microscope (Perkin-Elmer) equipped with an 100× Plan Apo 1.35 objective (Olympus), CCD (Orca ER; Hamamatsu), and Slidebook acquisition software. All quantitative analysis was performed with IP Lab Spectrum software (Scanalytics).

Spindle length in individual time-lapse images was determined using the Measure Length function of IP Lab Spectrum. The majority of spindle-length measurements were made from single focal plane, wide field images. Only spindles parallel to the plane of focus were used for this analysis. Either of two criteria was used to determine that a spindle was parallel to the plane of focus rather than rotating into the plane of focus to give the impression of shortening. First, the shape of both spindle poles appears identical only when the spindle is parallel to the plane of focus. Second, bundles of microtubules extending all the way between spindle poles are observed only when the spindle is parallel to the plane of focus. Note that the primary result of this paper is a lack of a late phase of spindle shortening that cannot be explained by any optical artifact.

Mean microtubule density in the entire spindle as shown in Figs. 1, 3, and 6 was determined in IP Lab Spectrum by highlighting the entire spindle as a “segment” and determining the mean pixel value within the entire segment. This measurement is proportional to microtubule density as opposed to microtubule mass, which would be proportional to the sum of the values of all of the pixels within a spindle. Three separate segments that encompassed most of one spindle pole, most of the middle quarter of the spindle, and a band between the spindle pole and the middle quarter of the spindle were used to repeat this analysis. During the period of overall density increase shown in Figs. 1 and 3, density (mean pixel value) increased uniformly in these three segments (not depicted). After spindle rotation, mean pixel values in these three segments behaved differently (not depicted), so line scan analysis (Fig. 4) was used to illustrate these more complex intensity changes.

Line scans of fluorescence intensity shown in Fig. 4 were generated in the following manner. GFP∷tubulin labeled discrete bundles of microtubules running from pole to pole. Exactly four of these bundles are visible in the wild-type spindle shown in Fig. 4. The ROI (region of interest) line tool in IP Lab Spectrum was used to draw a line down the length of one of these bundles. For wild type, a straight line was used. For mei-2(ct98), an irregular line was drawn to follow one of the irregular microtubule bundles. The ROI boundary function of IP Lab was then used to determine the pixel value at each point along this line, and these values were plotted as a function of spindle length in DeltaGraph. mCherry-histone–labeled chromosomes were between the microtubule bundles. Thus, a second line was drawn from pole to pole that passed over one chromosome pair to plot the anaphase movement of one chromosome pair.

Depletion of KLP-18 was accomplished by soaking worms in double-stranded RNA produced from cDNA clone yk745f12 provided by Y. Kohara (National Institute of Genetics, Mishima, Japan) using previously described methods (Yang et al., 2005). γ-TuRC subunits were depleted by RNAi by feeding using clones III-5K19 (tbg-1) and III-1F05 (gip-1) (MRC Gene Services; Kamath et al., 2003).

Fig. S1 shows purification of recombinant MEI-1–MEI-2. Fig. S2 shows that katanin localizes at spindle poles at the time of katanin-dependent microtubule redistribution. Fig. S3 shows that mei-1(null); klp-18(RNAi) double mutants do not show the synthetic effect on microtubule density observed in mei-1(null); tbg-1(RNAi) double mutants.

We thank the C. elegans Genetics Center, Paul Mains, Bruce Bowerman, Arshad Desai, and John White for strains; Yuji Kohara for cDNA clones; Roger Tsien for mCherry; and Jon Scholey and Ingrid Brust-Mascher for assistance with spinning-disk confocal microscopy. We also thank Lesilee Rose and Dan Starr for critical discussions and facilities.

This work was supported by a grant from the University of California Cancer Research Coordinating Committee.