Localization and dynamics of nonfilamentous actin in cultured cells [published erratum appears in J Cell Biol 1993 Nov;123(3):following 767]

Although the distribution of filamentous actin is well characterized in many cell types, the distribution of nonfilamentous actin remains poorly understood. To determine the relative distribution of filamentous and nonfilamentous actin in cultured NRK cells, we have used a number of labeling agents that differ with respect to their specificities toward the filamentous or nonfilamentous form, including monoclonal and polyclonal anti-actin antibodies, vitamin D-binding protein (DBP), and fluorescent phalloidin. Numerous punctate structures were identified that bind poorly to phalloidin but stain positively with several anti-actin antibodies. These bead structures also stain with DBP, suggesting that they are enriched in nonfilamentous actin. Similar punctate structures were observed after the microinjection of fluorescently labeled actin into living cells, allowing us to examine their dynamics in living cells. The actin-containing punctate structures were observed predominantly in the region behind lamellipodia, particularly in spreading cells induced by wounding confluent monolayers. Time-lapse recording of cells injected with fluorescent actin indicated that they form continuously near the leading edge and move centripetally toward the nucleus. Our results suggest that at least part of the unpolymerized actin molecules are localized at discrete sites, possibly as complexes with monomer sequestering proteins. These structures may represent transient storage sites of G-actin within the cell which can be transformed rapidly into actin filaments upon stimulation by specific signals.     

A 13-A map of the actin-scruin filament from the limulus acrosomal process [published erratum appears in J Cell Biol 1993 Dec;123(6 Pt 1):1621]

We have determined the structure of the actin-scruin filament to 13-A resolution using a combination of low-dose EM and image analysis. The three-dimensional map reveals four actin-actin contacts: two within each strand and two between strands. The conformation of the actin subunit is different from that in the Holmes et al. (1990) model as refined by Lorenz et al. (1993). In particular, subdomain II is tilted in a similar way to that seen by Orlova and Egelman (1993) in F-Mg2(+)- ADP actin filaments in the absence of Ca2+. Scruin appears to consist of two domains of approximately equal volume. Each scruin subunit cross- links the two strands in the actin filament. Domain I of scruin contacts subdomain I of actin and makes a second contact at the junction of subdomains III and IV. Domain II of scruin contacts actin at subdomains I and II of a neighboring actin subunit. The two scruin domains thus bind differently to actin.     

Cell and heparin binding in the distal long arm of laminin: identification of active and cryptic sites with recombinant and hybrid glycoprotein [published erratum appears in J Cell Biol 1993 Dec;123(6 Pt 1):1623]

The long arm of laminin, which binds heparin and cells, consists of three polypeptides (A, B1, and B2) joined in a coiled-coil rod attached to a terminal A chain globule (G). Previously, we found that recombinant globular domain (rG) supported heparin and myoblast binding (Yurchenco, P. D., U. Sung, M. D. Ward, Y. Yamada, and J. J. O'Rear. 1993. J. Biol. Chem. 268:8356-8365). To further analyze long arm functions, we expressed the distal moiety of the mouse laminin A chain extending from the middle of the rod to the carboxyl terminus (rAiG). This larger glycoprotein, secreted by Sf9 insect cells infected with recombinant baculovirus, was intercalated in vitro into the corresponding disulfide-linked B chain segments of laminin fragment E8 (distal long arm rod and proximal globule). The hybrid molecule (B- rAiG) possessed a structure similar to laminin long arm as judged by electron microscopy and limited proteolysis. By joining rAiG with E8-B chains, the affinity of G domain for heparin decreased from that observed with rAiG and rG to one similar to native protein. HT1080 cells adhered to E8, rAiG, and B-rAiG, less well to rG, and not to denatured E8/B-rAiG, the A and B chain moieties of E8, or to a mixture of rG and E8-B chains. Cell adhesion to E8 and B-rAiG, in contrast to rAiG, was inhibited with antibodies specific for alpha 6 and beta 1 integrin chains. Since intercalation (a) restored a conformationally dependent alpha 6 beta 1 integrin recognition site present in native protein, (b) inactivated a cryptic cell binding activity in the A chain, and (c) inhibited a heparin binding site present in proximal G domain, we conclude that biological activities of laminin are different from that of its isolated subunits.     

Aldolase exists in both the fluid and solid phases of cytoplasm [published erratum appears in J Cell Biol 1988 Dec;107(6 Pt 1):following 2463]

We have prepared a functional fluorescent analogue of the glycolytic enzyme aldolase (rhodamine [Rh]-aldolase), using the succinimidyl ester of carboxytetramethyl-rhodamine. Fluorescence redistribution after photobleaching measurements of the diffusion coefficient of Rh-aldolase in aqueous solutions gave a value of 4.7 x 10(-7) cm2/S, and no immobile fraction. In the presence of filamentous actin, there was a 4.5-fold reduction in diffusion coefficient, as well as a 36% immobile fraction, demonstrating binding of Rh-aldolase to actin. However, in the presence of a 100-fold molar excess of its substrate, fructose 1,6-diphosphate, both the mobile fraction and diffusion coefficient of Rh-aldolase returned to control levels, indicating competition between substrate binding and actin cross-linking. When Rh-aldolase was microinjected into Swiss 3T3 cells, a relatively uniform intracellular distribution of fluorescence was observed. However, there were significant spatial differences in the in vivo diffusion coefficient and mobile fraction of Rh-aldolase measured with fluorescence redistribution after photobleaching. In the perinuclear region, we measured an apparent cytoplasmic diffusion coefficient of 1.1 x 10(-7) cm2/s with a 23% immobile fraction; while measurements in the cell periphery gave a value of 5.7 x 10(-8) cm2/s, with no immobile fraction. Ratio imaging of Rh-aldolase and FITC-dextran indicated that FITC-dextran was relatively excluded excluded from stress fiber domains. We interpret these data as evidence for the partitioning of aldolase between a soluble fraction in the fluid phase and a fraction associated with the solid phase of cytoplasm. The partitioning of aldolase and other glycolytic enzymes between the fluid and solid phases of cytoplasm could play a fundamental role in the control of glycolysis, the organization of cytoplasm, and cell motility. The concepts and experimental approaches described in this study can be applied to other cellular biochemical processes.     

Two steps of insulin receptor internalization depend on different domains of the beta-subunit [published erratum appears in J Cell Biol 1993 Nov;123(4):1047]

The internalization of signaling receptors such as the insulin receptor is a complex, multi-step process. The aim of the present work was to determine the various steps in internalization of the insulin receptor and to establish which receptor domains are implicated in each of these by the use of receptors possessing in vitro mutations. We find that kinase activation and autophosphorylation of all three regulatory tyrosines 1146, 1150, and 1151, but not tyrosines 1316 and 1322 in the COOH-terminal domain, are required for the ligand-specific stage of the internalization process; i.e., the surface redistribution of the receptor from microvilli where initial binding occurs to the nonvillous domain of the cell. Early intracellular steps in insulin signal transduction involving the activation of phosphatidylinositol 3'-kinase are not required for this redistribution. The second step of internalization consists in the anchoring of the receptors in clathrin- coated pits. In contrast to the first ligand specific step, this step is common to many receptors including those for transport proteins and occurs in the absence of kinase activation and receptor autophosphorylation, but requires a juxta-membrane cytoplasmic segment of the beta-subunit of the receptor including a NPXY sequence. Thus, there are two independent mechanisms controlling insulin receptor internalization which depend on different domains of the beta-subunit.     

5''nucleotidase is sorted to the apical domain of hepatocytes via an indirect route [published erratum appears in J Cell Biol 1993 Nov;123(3):following 767]

In hepatocytes, all newly synthesized plasma membrane (PM) proteins so far studied arrive first at the basolateral domain; apically destined proteins are subsequently endocytosed and sorted to the apical domain via transcytosis. A mechanism for the sorting of newly synthesized glycophosphatidylinositol (GPI)-linked proteins has been proposed whereby they associate in lipid microdomains in the trans-Golgi network and then arrive at the apical domain directly. Such a mechanism poses a potential exception to the hepatocyte rule. We have used pulse-chase techniques in conjunction with subcellular fractionation to compare the trafficking of 5' nucleotidase (5NT), an endogenous GPI-anchored protein of hepatocytes, with two transmembrane proteins. Using a one-step fractionation technique to separate a highly enriched fraction of Golgi-derived membranes from ER and PM, we find that both 5NT and the polymeric IgA receptor (pIgAR) traverse the ER and Golgi apparatus with high efficiency. Using a method that resolves PM vesicles derived from the apical and basolateral domains, we find that 5NT first appears at the basolateral domain as early as 30 min of chase. However the subsequent redistribution to the apical domain requires > 3.5 h of chase to reach steady state. This rate of transcytosis is much slower than that observed for dipeptidylpeptidase IV, an apical protein anchored via a single transmembrane domain. We propose that the slow rate of transcytosis is related to the fact that GPI-linked proteins are excluded from clathrin-coated pits/vesicles, and instead must be endocytosed via a slower nonclathrin pathway.     

Midzone microtubule bundles are continuously required for cytokinesis in cultured epithelial cells [published erratum appears in J Cell Biol 1996 Dec;135(6 Pt 1):1679]

The current model of cytokinesis proposes that spindle poles and associated microtubules determine the cleavage plane, and, once the signal has been delivered to the cortex, the entire mitotic apparatus can be removed without affecting cell division. While supported by compelling data from Echinoderm embryos, recent observations suggest that the model may not be universally applicable. In this study, we have examined the relationship(s) among microtubules, chromosomes, and cleavage activity in living normal rat kidney (NRK) cells with multipolar mitotic figures. We found that cleavage activity correlated with the distribution of midzone microtubule bundles and Telophase Disc 60 protein (TD60) rather than the position of spindle poles. In addition, reduction of midzone microtubules near the cortex, by either nocodazole treatment or spontaneous reorganization in tripolar cells, caused inhibition or regression of furrowing. These results demonstrate that continuous interaction between midzone microtubule bundles and the cortex is required for successful cleavage in tissue culture cells.     

Microinjected centromere [corrected] kinetochore antibodies interfere with chromosome movement in meiotic and mitotic mouse oocytes [published erratum appears in J Cell Biol 1990 Dec;111(6 Pt 1):following 2800]

Kinetochores may perform several functions at mitosis and meiosis including: (a) directing anaphase chromosome separation, (b) regulating prometaphase alignment of the chromosomes at the spindle equator (congression), and/or (c) capturing and stabilizing microtubules. To explore these functions in vivo, autoimmune sera against the centromere/kinetochore complex are microinjected into mouse oocytes during specific phases of first or second meiosis, or first mitosis. Serum E.K. crossreacts with an 80-kD protein in mouse cells and detects the centromere/kinetochore complex in permeabilized cells or when microinjected into living oocytes. Chromosome separation at anaphase is not blocked when these antibodies are microinjected into unfertilized oocytes naturally arrested at second meiotic metaphase, into eggs at first mitotic metaphase, or into immature oocytes at first meiotic metaphase. Microtubule capture and spindle reformation occur normally in microinjected unfertilized oocytes recovering from cold or microtubule disrupting drugs; the chromosomes segregate correctly after parthenogenetic activation. Prometaphase congression is dramatically influenced when antikinetochore/centromere antibodies are introduced during interphase or in prometaphase-stage meiotic or mitotic eggs. At metaphase, these oocytes have unaligned chromosomes scattered throughout the spindle with several remaining at the poles; anaphase is aberrant and, after division, karyomeres are found in the polar body and oocyte or daughter blastomeres. Neither nonimmune sera, diffuse scleroderma sera, nor sham microinjections affect either meiosis or mitosis. These results suggest that antikinetochore/centromere antibodies produced by CREST patients interfere with chromosome congression at prometaphase in vivo.     

Signal transduction in Dictyostelium fgd A mutants with a defective interaction between surface cAMP receptors and a GTP-binding regulatory protein [published erratum appears in J Cell Biol 1988 Dec;107(6 Pt 1):following 2463]

Transmembrane signal transduction was investigated in four Dictyostelium discoideum mutants that belong to the fgd A complementation group. The results show the following. (a) Cell surface cAMP receptors are present in fgd A mutants, but cAMP does not induce any of the intracellular responses, including the activation of adenylate or guanylate cyclase and chemotaxis. (b) cAMP induces down-regulation and the covalent modification (presumably phosphorylation) of the cAMP receptor. (c) The inhibitory effects of GTP gamma S and GDP beta S on cAMP binding are reduced; the stimulatory effect of cAMP on GTP gamma S binding is lost in fgd A mutants. (d) Basal high-affinity GTPase activity is reduced 40% and the stimulatory effect of cAMP is decreased from 40% in wild type to 30% in fgd A. (e) GTP-mediated stimulation and inhibition of adenylate cyclase is normal in mutant membranes. The results suggest a defective interaction between cell surface cAMP receptors and a specific G-protein in fgd A mutants. This interaction appears to be essential for nearly all signal transduction pathways in Dictyostelium discoideum.     

Primary structure of the brain alpha-spectrin [published erratum appears in J Cell Biol 1989 Mar;108(3):following 1175]

We have determined the nucleotide sequence coding for the chicken brain alpha-spectrin. It is derived both from the cDNA and genomic sequences, comprises the entire coding frame, 5' and 3' untranslated sequences, and terminates in the poly(A)-tail. The deduced amino acid sequence was used to map the domain structure of the protein. The alpha-chain of brain spectrin contains 22 segments of which 20 correspond to the repeat of the human erythrocyte spectrin (Speicher, D. W., and V. T. Marchesi. 1984. Nature (Lond.). 311:177-180.), typically made of 106 residues. These homologous segments probably account for the flexible, rod-like structure of spectrin. Secondary structure prediction suggests predominantly alpha-helical structure for the entire chain. Parts of the primary structure are excluded from the repetitive pattern and they reside in the middle part of the sequence and in its COOH terminus. Search for homology in other proteins showed the presence of the following distinct structures in these nonrepetitive regions: (a) the COOH-terminal part of the molecule that shows homology with alpha-actinin, (b) two typical EF-hand (i.e., Ca2+-binding) structures in this region, (c) a sequence close to the EF-hand that fulfills the criteria for a calmodulin-binding site, and (d) a domain in the middle of the sequence that is homologous to a NH2-terminal segment of several src-tyrosine kinases and to a domain of phospholipase C. These regions are good candidates to carry some established as well as some yet unestablished functions of spectrin. Comparative analysis showed that alpha-spectrin is well conserved across the species boundaries from Xenopus to man, and that the human erythrocyte alpha-spectrin is divergent from the other spectrins.     

Differential trafficking and timed localization of two chitin synthase proteins, Chs2p and Chs3p [published erratum appears in J Cell Biol 1996 Dec;135(6 Pt 2):1925]

The deposition of the polysaccharide chitin in the Saccharomyces cerevisiae cell wall is temporally and spatially regulated. Chitin synthase III (Chs3p) synthesizes a ring of chitin at the onset of bud emergence, marking the base of the incipient bud. At the end of mitosis, chitin synthase II (Chs2p) deposits a disk of chitin in the mother-bud neck, forming the primary division septum. Using indirect immunofluorescence microscopy, we have found that these two integral membrane proteins localize to the mother-bud neck at distinct times during the cell cycle. Chs2p is found at the neck at the end of mitosis, whereas Chs3p localizes to a ring on the surface of cells about to undergo bud emergence and in the mother-bud neck of small- budded cells. Cell synchronization and pulse-chase experiments suggest that the timing of Chs2p localization results from cell cycle-specific synthesis coupled to rapid degradation. Chs2p degradation depends on the vacuolar protease encoded by PEP4, indicating that Chs2p is destroyed in the vacuole. Temperature-sensitive mutations that block either the late secretory pathway (sec1-1) or the internalization step of endocytosis (end4-1) also prevent Chs2p degradation. In contrast, Chs3p is synthesized constitutively and is metabolically stable, indicating that Chs2p and Chs3p are subject to different modes of regulation. Differential centrifugation experiments show that a significant proportion of Chs3p resides in an internal compartment that may correspond to a vesicular species called the chitosome (Leal- Morales, C.A., C.E. Bracker, and S. Bartnicki-Garcia. 1988, Proc. Natl. Acad. Sci. USA. 85:8516-8520; Flores Martinez, A., and J. Schwencke. 1988. Biochim. Biophys. Acta. 946:328-336). Fractionation of membranes prepared from mutants defective in internalization (end3-1 and end4-1) indicate that the Chs3p-containing vesicles are endocytically derived. Collectively, these data suggest that the trafficking of Chs2p and Chs3p diverges after endocytosis; Chs3p is not delivered to the vacuole, but instead may be recycled.     

Neurotrophin-3 induced by tri-iodothyronine in cerebellar granule cells promotes Purkinje cell differentiation [published erratum appears in J Cell Biol 1998 Dec 28;143(7):following 2090]

Thyroid hormones play an important role in brain development, but the mechanism(s) by which triiodothyronine (T3) mediates neuronal differentiation is poorly understood. Here we demonstrate that T3 regulates the neurotrophic factor, neurotrophin-3 (NT-3), in developing rat cerebellar granule cells both in cell culture and in vivo. In situ hybridization experiments showed that developing Purkinje cells do not express NT-3 mRNA but do express trkC, the putative neuronal receptor for NT-3. Addition of recombinant NT-3 to cerebellar cultures from embryonic rat brain induces hypertrophy and neurite sprouting of Purkinje cells, and upregulates the mRNA encoding the calcium-binding protein, calbindin-28 kD. The present study demonstrates a novel interaction between cerebellar granule neurons and developing Purkinje cells in which NT-3 induced by T3 in the granule cells promotes Purkinje cell differentiation.     

Inhibition of in vitro tumor cell invasion by Arg-Gly-Asp-containing synthetic peptides [published erratum appears in J Cell Biol 1989 Jun;108(6):following 2546]

The interaction of cells with extracellular matrix components such as fibronectin, vitronectin, and type I collagen has been shown to be mediated through a family of cell-surface receptors that specifically recognize an arginine-glycine-aspartic acid (RGD) amino acid sequence within each protein. Synthetic peptides containing the RGD sequence can inhibit these receptor-ligand interactions. Here, we use novel RGD-containing synthetic peptides with different inhibition properties to investigate the role of the various RGD receptors in tumor cell invasion. The RGD-containing peptides used include peptides that inhibit the attachment of cells to fibronectin and vitronectin, a peptide that inhibits attachment to fibronectin but not to vitronectin, a cyclic peptide with the opposite specificity, and a peptide, GRGDTP, that inhibits attachment to type I collagen in addition to inhibiting attachment to fibronectin and vitronectin. The penetration of two human melanoma cell lines and a glioblastoma cell line through the human amniotic basement membrane and its underlying stroma was inhibited by all of the RGD-containing peptides except for the one that inhibits only the vitronectin attachment. Various control peptides lacking RGD showed essentially no inhibition. This inhibitory effect on cell invasion was dose-dependent and nontoxic. A hexapeptide, GRGDTP, that inhibits the attachment of cells to type I collagen in addition to inhibiting fibronectin- and vitronectin-mediated attachment was more inhibitory than those RGD peptides that inhibit only fibronectin and vitronectin attachment. Analysis of the location of these cells that were prevented from invading indicated that they attached to the amniotic basement membrane but did not proceed further into the tissue. These results suggest that interactions between RGD-containing extracellular matrix adhesion proteins and cells are necessary for cell invasion through tissues and that fibronectin and type I collagen are important for this process.     

Regulation of fibronectin receptor distribution [published erratum appears in J Cell Biol 1992 Jul;118(2):491]

Studies on human osteoclast formation have been hampered by lack of a defined isolated progenitor cell population. We describe here the establishment of a human leukemic cell line (designated FLG 29.1) from bone marrow of a patient with acute monoblastic leukemia. The cultured cells are predominantly undifferentiated leukemic blasts, but addition of 12-o-tetradecanoylphorbol 13-acetate (TPA; 0.1 microM) induces irreversible differentiation into adherent, non-dividing, multinucleated cells. TPA-treated cells bear surface antigens typical of fetal osteoclasts, degrade 45Ca-labeled devitalized bone particles, display tartrate-resistant acid phosphatase in both mononuclear and multinuclear cells and receptors for calcitonin. Calcitonin increases intracellular cAMP accumulation in TPA-treated cells. TPA-treated cells show some ultrastructural features of osteoclasts as evidenced by transmission EM. These results indicate that FLG 29.1 cells may represent an osteoclast committed cell population, which upon induction with TPA acquire some morphological, phenotypical, and functional features of differentiated osteoclasts.     

Identification of an E-selectin region critical for carbohydrate recognition and cell adhesion [published erratum appears in J Cell Biol 1993 Feb;120(4):1071]

E-selectin elicits cell adhesion by binding to the cell surface carbohydrate, sialyl Lewis X (sLe(x)). We evaluated the effects of mutations in the E-selectin lectin domain on the binding of a panel of anti-E-selectin mAbs and on the recognition of immobilized sLe(x) glycolipid. Functional residues were then superimposed onto a three-dimensional model of the E-selectin lectin domain. This analysis demonstrated that the epitopes recognized by blocking mAbs map to a patch near the antiparallel beta sheet derived from the NH2 and COOH termini of the lectin domain and two adjacent loops. Mutations that affect sLe(x) binding map to this same region. These results thus define a small region of the E-selectin lectin domain that is critical for carbohydrate recognition.     

Interleukin-6 undergoes transition from paracrine growth inhibitor to autocrine stimulator during human melanoma progression [published erratum appears in J Cell Biol 1993 Apr;121(2):following 477]

The ability to penetrate the dermal basement membrane and subsequently proliferate in the underlying mesenchyme is one of the key steps in malignant progression of human melanomas. We previously undertook studies aimed at assessing how normal dermal fibroblasts (one of the main cellular components of mesenchyme) may affect the growth of human melanoma cells and facilitate the overgrowth of malignant subpopulations (Cornil, I., D. Theodorescu, S. Man, M. Herlyn, J. Jambrosic, and R. S. Kerbel. 1991. Proc. Natl. Acad. Sci. USA. 88:6028- 6032). We found that melanoma cell lines from early-stage (metastatically incompetent) lesions were growth inhibited whereas those from advanced-stage (metastatically competent) lesions were stimulated under the same conditions by co-culture with fibroblasts; conditioned medium from such cells gave the same result. Subsequent studies using biochemical purification and neutralizing antibodies revealed the inhibitory activity to be identical to interleukin-6 (IL- 6). We now report that addition of purified recombinant human IL-6 resulted in a growth inhibition in vitro by G1/G0 arrest of early, but not advanced stage melanoma cells. Despite this alteration in response there was no significant difference in melanoma cell lines of varying malignancy in respect to their expression of genes encoding the IL-6 receptor, or gp130, the IL-6 signal transducer. Scatchard analysis also revealed similar [125I]IL-6 binding activities in both IL-6 sensitive and resistant groups. However, studies of IL-6 production indicated that five out eight IL-6 melanoma cell lines known to be resistant to exogenous IL-6-mediated growth inhibition constitutively expressed mRNA for IL-6; they also secreted bioactive IL-6 into culture medium. To assess the possible role of this endogenous IL-6 in melanoma cell growth, antisense oligonucleotides to the IL-6 gene were added to cultures of melanoma cells. This resulted in a significant growth inhibition only in cell lines that produced endogenous IL-6. In contrast, neutralizing antibodies to IL-6 were ineffective in causing such growth inhibition. This indicates that endogenous IL-6 may behave as a growth stimulator by an intracellular ("private") autocrine mechanism. Thus, a single cytokine, IL-6, can switch from behaving as a paracrine growth inhibitor to an autocrine growth stimulator within the same cell lineage during malignant tumor progression. Such a switch may contribute to the growth advantage of metastatically competent melanoma cells at the primary or distant organ sites and thereby facilitate progression of disease.     

Expression of normal and mutant avian integrin subunits in rodent cells [published erratum appears in J Cell Biol 1989 Oct;109(4 Pt 1):1187]

We describe the expression of the beta 1 subunit of avian integrin in rodent cells with the purpose of examining the structure-function relationships of various domains within this subunit. The exogenous subunit is efficiently and stably expressed in 3T3 cells, and it forms hybrid heterodimers with endogenous murine alpha subunits, including alpha 3 and alpha 5. These heterodimers are exported to the cell surface and localize in focal contacts where both extracellular matrix and cytoskeleton associate with the plasma membrane. Hybrid heterodimers consisting of exogenous beta 1 and endogenous alpha subunits bind effectively and specifically to columns of cell-binding fragments of fibronectin. The exogenous avian beta 1 subunit appears to function as well as its endogenous murine equivalent, consistent with the high degree of conservation noted previously for integrins. In contrast, expression of a mutant form of avian integrin beta 1 subunit lacking the cytoplasmic domain produces hybrid heterodimers which, while efficiently exported to the cell surface and still capable of binding fibronectin, do not localize efficiently in focal contacts. This further implicates the cytoplasmic domain of the beta 1 subunit in interactions required for cytoskeletal organization.