This issue includes three reports of the cloning and characterization of kakapo, a Drosophila gene whose product may keep transmembrane signaling proteins in their place, link the actin cortex to structural microtubules, or both. The three groups discovered Kakapo after looking at an existing paralyzed mutant (Prokop et al., page 1283), screening for mutants with blistered wings (Gregory and Brown, page 1271), or doing an antibody screen against presumptive tendon cells (Strumpf and Volk, page 1259).

The phenotypes that the three groups uncover are equally diverse. The Brown and Volk groups observe that Kakapo localizes to the apical and basal surfaces of specific epidermal cells. These cells differentiate into tendon cells and link the fly embryo's muscles to the external cuticle. The protein is also found along the microtubules that run between these two sites, spanning the cell and presumably providing structural support.

In the mutant, Gregory and Brown find that the muscles detach from the epidermis; based on electron microscopy, Prokop et al. state that the muscles have ripped the epidermal cells in half, taking the basal membranes with them. There is less electron-dense material around the basal adherence sites, and microtubules are no longer connected to the sites. This suggests that Kakapo may form part of the link from the actin to the microtubule cytoskeleton in these cells. The plakin group of proteins form a similar link between actin and intermediate filaments; the amino terminus of Kakapo resembles the plakins and includes a putative actin-binding site, but the carboxyl terminus diverges where the plakins have a site for binding intermediate filaments (which have not been identified in Drosophila).

Kakapo is needed in other situations that require structural integrity: epidermal cells in the mutant sometimes break apart from each other during movements that remodel the embryo, and the wing blistering appears to be a failure in adhesion between the dorsal and ventral epidermis when the wing is filled, under pressure, with a fly's equivalent of blood.

The kakapo mutants studied by Prokop et al. are paralyzed not only because of problems in muscle–tendon attachment—they also have fewer synapses and smaller nerve terminals at the neuromuscular junction (NMJ). Although smaller, these NMJs do conduct current. Kakapo is present at the NMJ, although mosaic analysis will be needed to confirm that Kakapo is acting in neurons and not muscle cells.

Large-scale growth of neurons appears to be normal in kakapo mutants, but Prokop et al. observe a second defect in local neuronal growth in certain dendrites of the central nervous system. In these neurons, the transmembrane cell adhesion protein Fas II spreads from its normal peripheral location towards a central site. Deletion of the gene for Fas II does not, however, relieve the kakapo defect in neuron branching.

The neuronal phenotype may be explained by a failure in actin–microtubule attachment, and indeed a microtubule structure in kakapo sensory neurons is detached from the plasma membrane. But there is another possible explanation: Kakapo may keep transmembrane proteins other than Fas II in their rightful place.

Kakopo may have a similar function in the epidermis. The differentiation of tendon precursor cells into tendon cells requires Vein, a protein that is secreted by the muscle cell. Strumpf and Volk find that Vein is diffusely distributed in kakapo mutants. This diffusion probably explains the increased number of tendon precursor cells. None of these cells, however, complete the differentiation process, perhaps because Vein is a weak ligand of the fly epidermal growth factor receptor (EGF-R). Vein may not be able to deliver a strong enough signal without some concentration function provided by Kakapo. The distribution of the EGF-R in kakapo mutants is not known, but it would not be surprising if the mutation led to a receptor localization that was more diffuse.

The notochord is famous as an inducer of other tissues, such as the neural tube and the somites. But few people consider how it is disposed of once it has done its job. On page , Aszódi et al. report that collagen II is necessary for the removal of the notochord and suggest an intriguing theory about how this might work.

The spine starts as a continuous column of mesenchyme around the notochord. This column is then patterned into alternating regions of high and low cell density. The regions of low density give rise to the vertebrae: the cells differentiate into cartilage and produce collagen II, the notochord disappears, and bone replaces cartilage. The regions of high cell density become the intervertebral discs. Cartilage forms around the outside, but notochord cells remain in the center to form the nucleus pulposus, the gelatinous material that allows us to bend our vertebral column (and that pushes out onto the sciatic nerve in a disc prolapse).

In their collagen II knockout mice, Aszódi et al. find that collagens I and III are ectopically expressed in the poorly organized cartilage. The vertebral bodies swell up and fail to eliminate the notochord, bone formation is defective, and the nucleus pulposus does not develop in what should be the intervertebral discs. They suggest that the swelling results from osmotic forces generated as water is attracted to the negatively charged proteoglycans in the cartilage matrix. Collagens I and III are soluble (in this setting) and apparently cannot form a matrix strong enough to contain the osmotic force.

In wild-type animals, the notochord could theoretically disappear through cell death in the presumptive vertebrae, with cell growth in the intervertebral area generating the nucleus pulposus. But Aszódi et al. show that neither cell death nor proliferation varies between the two areas. Their alternative explanation is that the osmotic pressure that is normally contained by the collagen II of the vertebral bodies forces the migration of the notochord cells to form the nucleus pulposus.

“The problem is that we cannot prove this last step,” says senior author Reinhard Fässler. Cartilage is capable of constraining a pressure of over three atmospheres, but the mouse embryo is too small for such measurements. Tagging of the notochord cells with an expressed protein and looking for evidence of migration may be the most definitive experiment possible.

The first direct evidence of a role for a large protein aggregate in viral fusion is presented by Markovic et al. (page ). Such a complex had been expected based on the dependence of fusion on the density of fusion proteins, the decrease in lateral diffusion of fusion proteins during fusion, and the lack of lipid diffusion between bilayers early in fusion, which suggested the presence of a large protein “fence.”

Baculovirus fusion is triggered by low pH in the endosome. Markovic et al. find that fusion is blocked by the addition of the reducing agent DTT, although the fusion protein gp64 remains in a trimeric form in membranes. DTT does not appear to block the pH-dependent conformation change. The change occurs in DTT, so the protein becomes fusion-competent when oxidized, even if the pH is now neutral.

Instead, DTT may be blocking formation of higher order complexes. Cell surface cross-linking of low pH–triggered gp64 followed by a density gradient reveals a complex of ∼2 MD, which is composed of ∼10 gp64 trimers. As with fusion, complex formation is transient, inhibited by DTT, and requires both low pH and contacting membranes. The size of the complex is close to an earlier estimate, which was based on the apparent size of the fusion pore given its electrophysiological properties.

The PTEN gene is mutated in many cancers; its protein product is a phosphatase with homology to the cytoskeletal protein tensin. Recent results have shown that it can dephosphorylate the signaling lipid phosphatidylinositol-3,4,5-triphosphate, and thus either promote apoptosis or suppress growth pathways. But Kenneth Yamada's group has focused on the protein phosphatase activity of this protein, and they have recently shown that PTEN can dephosphorylate focal adhesion kinase (FAK) and inhibit integrin-mediated cell spreading. In this issue (Gu et al., page 1375), they find that activation of the MAP kinase pathway by two means (ligation of integrins by fibronectin, or activation of the epidermal growth factor receptor [EGF-R]) can be suppressed by PTEN. Constitutive activation of the MAP kinase pathway partially reverses PTEN's suppression of cell spreading.

For the EGF-R pathway, Gu et al. find that receptor autophosphorylation and Shc binding are intact, but Shc phosphorylation is reduced by PTEN. Phosphorylated Shc binds Grb2, which binds the Ras guanine nucleotide exchange factor SOS, so it is no surprise that Ras activation is defective in cells lacking PTEN. Failure to activate Ras would explain low activity in the MAP kinase pathway, which lies directly downstream.

The sequence of events after integrin ligation by fibronectin is more complicated. PTEN appears to cause the dephosphorylation of both FAK and Shc. Both proteins have been implicated in integrin signaling, and either one can bind Grb2 and potentially activate the Ras pathway.

The small GTPase Rac, and its guanine nucleotide exchange factor Tiam1, have been shown to both inhibit and induce motility, even in the same epithelial cell line. Sander et al. solve this paradox on page 1385. The nature of the effect depends on the identity of the extracellular matrix, which appears to control the subcellular location of Tiam1.

On fibronectin or laminin, Tiam1 or a constitutively active Rac1 increases E-cadherin–mediated adhesion between cells and reverts ras-transformed cells to a nonmotile state. Even a small amount of collagen, however, induces motility in these cells. The level of Rac1 activation does not vary on the different substrates, but the location of Tiam1 does: in cells plated on fibronectin, Tiam1 is at adherens junctions, whereas collagen induces Tiam1's relocation to ruffles and lamellae.

Phosphatidylinositol 3-kinase is needed for Tiam1 to induce either increased adhesion or increased motility. Tiam1 could be localized by binding to particular phosphatidylinositol lipids that are produced by complexes associated with localized integrins or transmembrane receptors. The localization of Tiam1 may determine the substrates available to Rac and thus the cellular outcome of Rac activation.

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