To directly observe the formation of a stable bundle midzone by just two MTs, we looked for rare events in which one MT was nucleated along a single MT. In Figure 1C, nucleation takes place 4.0 μm away from an existing minus end. The sliding of the novel minus end gradually slowed as the new MT elongated and completely stalled after sliding a total length of 1.2 μm. Meanwhile, its plus end continued growing and eventually extended beyond the minus end of the underlying MT (Figure 1C, yellow circle). The resultant bundle midzone was 2.8 μm long. Figure 1D (white circle) shows a second example for which nucleation occurs only 1.2 μm away from the existing minus end, and movement stops after 0.4 μm of sliding. The resultant bundle midzone of only 0.8 μm shows that midzone length varies significantly with the location of MT nucleation.

Cell:“分子发动机和刹车”作用机制的研究
Figure 1B, label 2). A kymograph of a complete bundle can be viewed as a superposition of several MT triangles that gives rise to areas with increased intensity. Most MT overlap was seen in the bundle midzone between antiparallel MT minus ends (Figure 1B, label 3), but additional overlap zones were occasionally created by two parallel MTs emanating from the bundle midzone (Figure 1B, label 5). Furthermore, new MTs were nucleated (Figure 1B, label 4) and transported along existing MTs toward the bundle midzone. One example is highlighted in green (Figure 1B), for which sliding can be observed relative to a static speckle on the underlying MT (Figure 1B, yellow dots). Nucleated MTs are oriented antiparallel with respect to the underlying MTs, such that once their plus ends cross the bundle midzone, they eventually grow with the correct polarity toward the closest cell end (Janson et al., 2005
). In contrast to the sliding of new MTs, overlapping MT minus ends within the bundle midzone did not slide relative to each other (Figure 1B, label 3), raising the question of how sliding is regulated.
To understand the forces involved in the generation of a bundle midzone, we investigated the distribution of motor proteins along MTs by constructing two-color kymographs of simultaneously imaged klp2-GFP and mRFP-tubulin (Figure 2A and S1A; Movies S3 and S5). Interestingly, klp2-GFP tracked plus ends of growing MTs and was notably absent from MT minus ends. Sliding forces between overlapping MTs are therefore generated solely at MT plus ends, explaining the abrupt stop and constant overlap of the nucleated MT in Figure 1D once its plus end passed the underlying MT. Within mature bundles with more than two MTs, we furthermore noticed that no MT minus ends were pushed out of the bundle midzone by parallel MT sliding, suggesting that—in contrast to antiparallel sliding of new MTs—klp2p is unable to bundle and slide parallel MTs (see schematic in Figure 2A). In support of this we noticed that (1) parallel MTs can splay apart within bundles (Figure 2B and Movie S4) and (2) the plus ends of parallel-growing MTs in kymographs did not slide relative to each other (Figure 2A and S1A). The MT binding properties of klp2p are thus tailored to arrange MT minus ends stably into a small bundle midzone.

