The standard pathway of membrane protein insertion involves a recognition and docking step, followed by insertion through a pore. The transfer is terminated by a hydrophobic stop-transfer sequence leading to the release of the partially translocated protein. On page , Schleiff et al. report that a pore is not needed for the correct integration of the voltage-dependent anion-selective channel, VDAC, into the mitochondrial outer membrane.

The recognition and pore complexes in the mitochondrial outer membrane are based around the proteins Tom20 and Tom40, respectively. Blockade of Tom40 function with unfoldable proteins or a temperature-sensitive Tom40 mutant does not prevent VDAC import, and Tom40 is not necessary for integration of functional VDAC into synthetic vesicles. Tom20 is necessary for VDAC integration into vesicles, however. Tom20 may function simply by increasing the local concentration of VDAC at the membrane, or it may have a more active role in releasing bound chaperones or inducing a conformational change that aids insertion.

A similar discovery in the 1970s, that an M13 coat protein could insert into membranes without the need for a channel, temporarily threatened the introduction of the signal hypothesis, but this case was subsequently recognized to be an exception. VDAC, however, is a cellular protein, and a member of a widespread group of proteins with a β-barrel conformation. Tom40 has some of the characteristics of a β-barrel protein. Perhaps the VDAC insertion mechanism is the answer to the chicken-and-egg evolutionary problem of how the first integral proteins, and the transport pores themselves, were inserted into membranes.

At the light-microscope level, membrane protrusion during cell movement is driven by actin polymerization at the leading edge. The actin is then turned over by depolymerization at variable distances away from the leading edge. The simplest interpretation of these observations is that motility is driven by the treadmilling growth of long actin filaments. Svitkina and Borisy (page ) take a closer look at actin filaments by electron microscopy and find support for another theory in which there is limited filament growth, but frequent nucleation, of a branched network that can adapt to rapid changes. In this theory, it is the network, rather than the individual filament, that treadmills.

The Borisy group has observed the branched network in fish keratinocytes, but needed to switch to frog cells so that they could use antibodies. Use of these antibodies now confirms that the Arp2/3 complex (composed of actin-related proteins 2 and 3 and five other proteins) is at the Y-junctions. Arp2/3 can nucleate actin polymerization, cap pointed ends, and bind those pointed ends to the side of existing filaments. Branching is frequent, so most of the distal (barbed) ends must be capped to prevent an exponential increase in the amount of actin. The emphasis on multiple nucleation events in this model provides plenty of scope for explosive actin growth when the cell needs to accelerate or change direction.

The dense branched network, or brush, extends back for only ∼1 μm, at which point it reverts to a looser, more readily depolymerized actin network. Cofilin could effect this transition through its severing or pointed-end depolymerization activities, but other factors must be limiting its action, as cofilin is present at the rear of the keratinocyte brush and throughout the fibroblast brush. The factor that spatially controls depolymerization may be displacing the Arp2/3 complex, thus exposing pointed ends.

Bub2 Branches Off

Six budding yeast mutants, mad1, mad2, mad3, bub1, bub2, and bub3, fail to arrest in mitosis when their spindles are disrupted with drugs such as nocodazole. Four research groups, including Fraschini et al. in this issue (page ), now show that these checkpoint proteins fall into two pathways. Five of the proteins appear to prevent anaphase onset until chromosomes are attached to the spindle, but Bub2p defines a separate pathway that may prevent cytokinesis onset until anaphase is either completed or well underway.

Bub2p has looked different in the past: it does not localize to the kinetochore (as do several mitotic checkpoint proteins), and the bub2 mutant cells still delay when confronted with chromosomes with defective kinetochores. Therefore, Fraschini et al. looked at mad2 bub2 double mutants, and found that they were more sensitive to microtubule-depolymerizing drugs than either single mutant. The double mutants rereplicated faster and more efficiently in nocodazole than the single mutants, acting as though no drug were present.

Fraschini et al. and Alexandru et al. (Alexandru, G., W. Zachariae, A. Schleiffer, and K. Nasmyth. 1999. EMBO (Eur. Mol. Biol. Organ.) J. 18:2707–2721) find that the Mad2p pathway is preventing the destruction of the anaphase inhibitor Pds1p and therefore chromosome disjunction. In contrast, the Bub2p pathway is preventing the destruction of mitotic kinase activity, and thus the final exit from mitosis, by controlling Clb2p cyclin levels and perhaps other regulators. Although a full mitosis checkpoint defect seems to require the removal of two systems, the isolation of the original single mutants was possible because the systems cover slightly different events, and because of leakiness between the two checkpoints.

Li (Li, R. 1999. Proc. Natl. Acad. Sci. USA. 96:4989– 4994) and Fesquet et al. (Fesquet, D., P.J. Fitzpatrick, A.L. Johnson, K.M. Kramer, J.H. Toyn, and L.H. Johnston. 1999. EMBO (Eur. Mol. Biol. Organ.) J. 18:2424–2434) use different proteins as starting points, but come to similar conclusions. Fesquet et al. find that Bub2p, but not the other checkpoint proteins, turns off Dbf2p protein kinase activity in the presence of nocodazole. Dbf2p is one of a group of proteins needed for mitotic exit. Alexandru et al. and Li both use sequence similarity to identify the budding yeast version of Byr4p (Li names this protein Bfa1p for Byr-four-alike). In fission yeast, Byr4p acts with the Bub2p homologue, Cdc16p, as a two-component GTPase activating protein and negative regulator of cytokinesis. Mutants in budding yeast byr4 act like bub2: they are sensitive to nocodazole, and show increased sensitivity when combined with mad2.

Both Bub2p (Li, 1999; Fraschini et al.) and Bfa1p/Byr4p (Li, 1999) localize to spindle poles. These proteins may sense the arrival of chromosomes at the poles at the end of anaphase and only then allow the initiation of cytokinesis. Controlling cytokinesis onset is very important in budding yeast, where the site of cell division (the bud neck) is formed very early in the cell cycle.

Checkpoints and Microtubule Dynamics

On page , Tirnauer et al. report that Bim1p is the first nonmotor budding yeast protein known to promote dynamic instability of microtubules. Bim1p promotes dynamics during G1, when dynamic microtubules are needed to position the newly formed spindle.

Yeast cells that lack Bim1p have shorter microtubules in G1 relative to wild-type cells. The microtubules have fewer catastrophe and rescue events, shrink more slowly, and spend far more time pausing. The sum of these effects is to reduce microtubule dynamics to below the level normally seen in mitosis.

The mammalian version of Bim1p is EB1. Taxol displaces EB1 from microtubules, which may in part explain why taxol suppresses microtubule dynamics.

EB1 binds the domain of the adenomatous polyposis coli (APC) tumor suppressor protein that is deleted in colon carcinoma. The cancer connection for EB1 may be checkpoints. Bim1p functions in a recently identified nuclear positioning checkpoint (which is not identical to, but may be connected to, the Bub2p checkpoint mentioned above). This checkpoint is important for yeast growing outside the laboratory, where frequent low temperatures cause depolymerization of astral microtubules and a failure of the nucleus to line up in the bud neck. Both EB1 and Bim1p bind the plus end of microtubules, so perhaps the yeast cell uses Bim1p's arrival at the cortex of the cell as the signal that astral microtubules have successfully found their way into both mother and daughter cells.

The splice products of three synapsin genes anchor synaptic vesicles to each other and to the cytoskeleton. Mice that lack synapsin I survive with few gross consequences, but are prone to epilepsy. On page , Terada et al. provide an explanation for the epilepsy.

When mutant inhibitory neurons in culture are first stimulated, they release a normal burst of neurotransmitter. Subsequent intense stimulation, however, elicits less than half the normal number of neurotransmitter quanta. After stimulation of mutant inhibitory neurons, there are fewer vesicles in regions away from the synaptic cleft. Based on these data and other measurements of neurotransmitter release, Terada et al. suggest that the mutant neurons fail to withhold a majority of vesicles in a reservoir pool, instead releasing most vesicles upon strong stimulation. The neurons are then temporarily unable to respond to further stimulation.

Repeated neurotransmitter release is normal in excitatory neurons, possibly because synapsin II can substitute for synapsin I in these cells. The imbalance between normal excitatory signals and insufficient inhibitory signals after strong stimulation, provides a plausible explanation for the epilepsy in the mutant mice.

Whereas most mice with epilepsy have learning problems, both the synapsin I mutant mice and most humans with epilepsy have no learning deficit. These mice may therefore provide a better model for human epilepsy.

By William A. Wells, 1095 Market St. #516, San Francisco, CA 94103. E-mail: