Characterization and intracellular distribution of microtubule-associated protein 2 in differentiating human neuroblastoma cells.

The use of a panel of monoclonal antibodies (mAbs) directed against different determinants of microtubule-associated protein 2 (MAP2) enabled us to identify two distinct high-molecular-mass MAP2 species (270 and 250 kDa) and a substantial amount of MAP2c (70 kDa) in human neuroblastoma cells. The 250-kDa MAP2 species appears to be confined to the human neuroblastoma cells and was not observed in microtubules (MTs) from bovine and rat brain, mouse neuroblastoma, or MTs from human cerebellum. A new overlay method was developed, which demonstrates binding of tubulin to human neuroblastoma high-molecular-mass MAP2 by exposing nitrocellulose-bound MT proteins under polymerization conditions to tubulin. Bound tubulin was detected with a mAb directed against beta-tubulin. The binding of tubulin to MAP2 could be abolished by a peptide homologous to positions 426-445 of the C-terminal region of beta-tubulin. Immunological cross-reactivity with several mAbs directed against bovine brain MAP2, taxol-promoted coassembly into MTs, and immunocytochemical visualization within cells were further criteria utilized to characterize these proteins as true MAPs. Indirect immunofluorescence with anti-MAP2 and anti-beta-tubulin mAbs demonstrated that there is a change in the spatial organization of MTs during induced cell differentiation, as indicated by the appearance of MT bundles and the redistribution of MAP2.

Organization of microtubules in cochlear hair cells.

The organization of microtubules in hair cells of the guinea-pig cochlea has been investigated using transmission electron microscopy and correlated with the location of tubulin-associated immunofluorescence in surface preparations of the organ of Corti. Results from both techniques reveal consistent distributions of microtubules in inner and outer hair cells. In the inner hair cells, microtubules are most concentrated in the apex. Reconstruction from serial sections shows three main groups: firstly, in channels through the cuticular plate and in a discontinuous belt around its upper perimeter; secondly, forming a ring inside a rim extending down from the lower perimeter of the plate; and thirdly, in a meshwork underlying the main body of the plate. In the cell body, microtubules line the inner face of the subsurface cistern and extend longitudinally through a tubulo-vesicular track between the apex and base. In outer hair cells, the pattern of microtubules associated with the cuticular plate is similar, although there are fewer present than in inner hair cells. In outer hair cells from the apex of the cochlea, microtubules occur around an infracuticular protrusion of cuticular plate material. In the cell body, many more microtubules occur in the region below the nucleus compared with inner hair cells. The possible functions of microtubules in hair cells are discussed by comparison with those found in other systems. These include morphogenesis and maintenance of cell shape; intracellular transport, e.g., of neurotransmitter vesicles; providing a possible substrate for motility; mechanical support of structures associated with sensory transduction.

Identification of novel proteins unique to either transverse tubules (TS28) or the sarcolemma (SL50) in rabbit skeletal muscle.

Novel proteins unique to either transverse tubules (TS28) or the sarcolemma (SL50) have been identified and characterized, and their in situ distribution in rabbit skeletal muscle has been determined using monoclonal antibodies. TS28, defined by mAb IXE112, was shown to have an apparent relative molecular mass of 28,000 D. Biochemical studies showed that TS28 is a minor membrane protein in isolated transverse tubular vesicles. Immunofluorescence and immunoelectron microscopical studies showed that TS28 is localized to the transverse tubules and in some subsarcolemmal vesicles possibly corresponding to the subgroup of caveolae connecting the transverse tubules with the sarcolemma. In contrast, TS28 is absent from the lateral portion of the sarcolemma. Immunofluorescence studies also showed that TS28 is more densely distributed in type II (fast) than in type I (slow) myofibers. Although TS28 and the 1,4-dihydropyridine receptor are both localized to transverse tubules and subsarcolemmal vesicles, TS28 is not a wheat germ agglutinin (WGA)-binding glycoprotein and does not appear to copurify with the 1,4-dihydropyridine receptor after detergent solubilization of transverse tubular membranes. SL50, defined by mAb IVD31, was shown to have an apparent relative molecular mass of 50,000 D. Biochemical studies showed that SL50 is not related to the 52,000-D (beta subunit) of the dihydropyridine receptor but does bind to WGA-Sepharose. Immunofluorescence labeling imaged by standard and confocal microscopy showed that SL50 is associated with the sarcolemma but apparently absent from the transverse tubules. Immunofluorescence labeling also showed that the density of SL50 in type II (fast) myofibers is indistinguishable from that of type I (slow) myofibers. The functions of TS28 and SL50 are presently unknown. However, the distinct distribution of TS28 to the transverse tubules and subsarcolemmal vesicles as determined by immunocytochemical labeling suggests that TS28 may be directly involved in excitation-contraction coupling. Our results demonstrate that, although transverse tubules are continuous with the sarcolemma, each of these membranes contain one or more unique proteins, thus supporting the idea that they each have a distinct protein composition.

Nucleation of macrotubules on double-walled microtubule seeds.

It is known that histone H1 is able to cause the formation of double-walled microtubules from microtubule protein. Now, we demonstrate that in dependence on the mass ratio H1/microtubule protein upon addition of tubulin to short pieces of double-walled microtubules either their inner or their outer wall elongates resulting in normal microtubules or in macrotubules, respectively. Because of their genesis we suggest that macrotubules like double-walled microtubules (see Unger et al., Eur. J. Cell Biol. 46, 98-104 (1988)) expose those sides of tubulin dimers at their surface which usually form the lumen face of microtubules.

Spindles and centrosomes during male meiosis in Drosophila melanogaster.

We have studied the spatial distribution of chromosomes, spindle fibers and centrosomes throughout the first meiotic division in males of Drosophila melanogaster. There seem to be two different types of spindle fibers: those which connect the poles to the chromosomes, and others arranged as cup-shaped hemispheres that reach from the poles to an unstained area on the equator of the cell. These pole-equator fibers could be responsible for positioning the nucleus and distributing cytoplasmic organelles around the nucleus during prophase, so that after meiosis, the daughter cells are provided with equal amounts of preorganized cytoplasmic organelles. These fibers remain until after the daughter nuclei have formed during telophase. An antigen associated with the centrosomes of mitotic spindles appears during meiosis as dispersed particles surrounding the nucleus; these particles might provide the developing spermatids with microtubule-organizing centers.

Selective purification of microtubule-associated proteins 1 and 2 from rat brain using poly(L-aspartic acid).

A rapid and selective purification procedure for microtubule-associated protein (MAP) 1 and MAP 2 has been established. This procedure is based upon the fact that poly(L-aspartic acid) (PLAA) can specifically remove MAP 1 from microtubules polymerized by taxol (Nakamura et al., 1989, J. Biochem. 106, 93-97). MAP 1 released by PLAA was further purified by column chromatography on phosphocellulose and Bio-Gel A-15m. The purified MAP 1 contained MAPs 1A and 1 B. From microtubules devoid of MAP 1, MAP 2, consisting of MAPs 2A and 2B, could also be isolated by exposure to high ionic strength solutions in the presence of taxol without heat treatment. Both MAPs 1 and 2 cosedimented with microtubules consisting of purified tubulin.

Microtubule-associated protein 3 (MAP3) expression in non-neuronal tissues.

Microtubule-associated protein 3 (MAP3, Mr 180,000), which in previous studies has been shown to be associated with glial processes and neurofilament-rich axons in rat brain, was examined in various non-neuronal rat tissues. Immunoblots of adult rat tissues (brain, liver, heart, spleen, adrenal medulla and kidney) showed that MAP3 is present in all organs tested. In addition we demonstrated that MAP3 is a heat-stable protein. Using immunohistochemistry, we established the localisation of MAP3 in various cell types. MAP3-containing cells appeared to have in common an asymmetric morphology with long processes that need structural support. In kidney MAP3 is limited to epithelial podocytes and in liver to Kupffer cells. In the adrenal gland, the cells of the cortex are devoid of MAP3 compared to the cells of the medulla. High concentrations of MAP3 are also found in cardiac muscle along the Z-disc and in the smooth muscle cells of the digestive tract. In spleen MAP3 is found in cells of the white pulp surrounding central blood vessels. A co-distribution of MAP3 with microtubules and intermediate filaments but not with microfilaments was found in each cell type examined. The widespread distribution pattern of MAP3 together with its molecular size and heat-stability indicate that MAP3 might be a member of the recently postulated family of homologous 200,000 Mr mammalian tissue MAPs. Potential functions for MAP3 in specific cell types are discussed.

Motion of individual human spermatozoa, both normal and lacking the outer dynein arms, during a continuous temperature rise.

The effect of increasing temperature from 22-25 degrees C to 37 degrees C on various motion characteristics of individual normal human spermatozoa and spermatozoa lacking the outer dynein arms (LODA) was studied by using a new automatic microscopic tracking method. It was found that: 1) The curvilinear velocity (Vc, measured between 1-3 sec) of both normal and LODA spermatozoa, fluctuated more or less intensely between spermatozoa; this fluctuation was not thermodependent. 2) The average Vc in the two groups of spermatozoa increased with the rise in temperature at a similar rate (1 micron/sec/degrees C), but LODA spermatozoa had an initial Vc lower than that of normal spermatozoa (12.5 +/- 5.3 microns/sec and 34.2 +/- 8.2 microns/sec, respectively). 3) The profile of the Vc increase associated with the temperature rise was different for the two groups of spermatozoa: for LODA spermatozoa it was linear between 25-37 degrees C, whereas for normal spermatozoa a plateau was reached at about 31 degrees C. 4) Various patterns of trajectory were found for both normal and LODA spermatozoa; these patterns were unrelated to temperature. However, LODA spermatozoa had more linear trajectories than normal spermatozoa. 5) Plots derived from reaction rate theory showed that the activation enthalpy, delta H was a function of the increase of Vc for both normal and LODA spermatozoa, but that delta H was higher for LODA spermatozoa.

Nuclear and mitotically enhanced epitope.

Salt-extracted proteins of taxol-stabilized microtubules from Chinese hamster ovary cells arrested at mitosis were used to immunize mice for hybridoma production. From a group of related monoclonal antibodies (MAbs), one, C9, recognized an epitope on antigens localized by immunofluorescence microscopy to interphase centrosomes and nuclei. The availability of the nuclear antigen was cell cycle-dependent; however, permeabilization of cells before fixation revealed that the antigen was present throughout the cell cycle. The nuclear antigen was exposed during prophase and was released from the nucleus upon nuclear envelope breakdown filling the cytoplasm of the mitotic cell. Antigenic material re-accumulated at daughter nuclei and was concealed during G1 phase. Detergent extraction of the cytoplasmic antigen from mitotic cells enabled localization of antigens to centrosomes, kinetochores, and the furrowing region/midbody. Immunoblot analysis of cells of a variety of species of origin identified an approximate 250 kD polypeptide as corresponding to the nuclear antigen, whereas polypeptides of 107/117 kD as well as approximately 250 kD accounted for the mitotic cytoplasmic antigens. No polypeptides could be associated with antigens at centrosomes, kinetochores, or midbodies. This MAb joins the antibody preparations previously reported that describe nuclear antigens, or epitopes on antigens, enhanced at mitosis.

Outer-arm dynein from trout spermatozoa: substructural organization.

Outer-arm dynein purified from trout spermatozoa was disrupted by low-ionic-strength dialysis, and the resulting subunits were separated by sucrose density-gradient centrifugation. The intact 19 S dynein, containing the alpha- an beta-heavy chains, intermediate chains (ICs) 1-5 and light chains (LCs) 1-6, yielded several discrete particles: a 17.5 S adenosine triphosphatase (ATPase) composed of the alpha- and beta-chains ICs 3-5 and LC 1; a 9.5 S complex containing ICs 1 and 2 together with LCs 2, 3, 4, and 6; and a single light chain (LC 5), which sedimented at approximately 4 S. In some experiments, ICs 3-5 also separated from the heavy chain complex and were obtained as a distinct subunit. Further dissociation of the 17.5 S particle yielded a 13.1 S ATPase that contained the beta-heavy chain and ICs 3-5. The polypeptide compositions of the complexes provide new information on the intermolecular associations that occur within dynein. Substructural features of the trout dynein polypeptides also were examined. The heavy chains were subjected to vanadate-mediated photolysis at the V1 sites by irradiation at 365 nm in the presence of Mg2+, ATP, and vanadate. Fragment pairs of relative molecular mass (Mr) 245,000/185,000 and 245,000/170,000 were obtained from the alpha- and beta-heavy chains, respectively. Photolysis of these molecules at their V2 sites, by irradiation in the presence of vanadate and Mn2+, yielded fragments of Mr 160,000/270,000 and 165,000/250,000, respectively. These values confirm that the alpha- and beta-heavy chains have masses of 430,000 and 415,000 daltons, respectively. Immunological analysis using monoclonal antibodies revealed that one intermediate chain from trout dynein (IC 2) contains epitopes present in two different intermediate chains from Chlamydomonas dynein. This indicates that specific sequences within the dynein intermediate chains have been highly conserved throughout evolution.

Characterization of the aluminum and beryllium fluoride species bound to F-actin and microtubules at the site of the gamma-phosphate of the nucleotide.

Aluminum fluoride and beryllium fluoride complexes have previously been shown to bind tightly to F-ADP-actin and GDP-microtubules in competition with Pi and to mimic the XDP-Pi transient state of the polymerization. The structure of the bound complexes is investigated here in further detail. Using a fluoride ion-specific electrode, the number of fluoride atoms per aluminum or beryllium atom in the bound complex could be determined. The results indicate that AIF-4 and either BeF2(OH)-.H2O or BeF3-.H2O are the tightly bound species in both F-actin and microtubules. The dependences of the binding on pF and pH are consistent with this conclusion. The possible geometries of aluminum and beryllium fluorides in the gamma-phosphate subsite of the nucleotide are discussed in correlation with the catalytic mechanism of nucleotide hydrolysis.

Unusual properties of a cold-labile fraction of Atlantic cod (Gadus morhua) brain microtubules.

A cold-labile fraction of microtubules with unusual properties was isolated from the brain of the Atlantic cod (Gadus morhua). The yield was low, approximately six times lower than that for bovine brain microtubules. This was mainly caused by the presence of a large amount of cold-stable microtubules, which were not broken down during the disassembly step in the temperature-dependent assembly-disassembly isolation procedure and were therefore lost. The isolated cold-labile cod microtubules contained usually only a low amount of microtubule-associated proteins (MAPs). Three high molecular mass proteins were found, of which one was recognized as MAP2. Cod MAP2 differed from mammalian brain MAP2; it was not heat stable and had a slightly higher molecular mass. In contrast to mammalian MAPs, MAP1 was not found in the cold-labile fraction of microtubules. A new heat-labile MAP of higher molecular mass (400 kilodaltons) was however present, as well as a heat-stable protein of slightly lower molecular mass than MAP2. These MAPs showed similar tubulin-binding characteristics as bovine brain MAPs, since they coassembled with taxol-assembled bovine brain microtubules consisting of pure bovine tubulin. In spite of the fact that Ca2+ bound equally to cod and porcine tubulins, it did not inhibit cod microtubule assembly even at high concentrations (greater than 1 mM). In contrast, rings, spirals, and macrotubules were formed. The results show that there are major differences between this fraction of cod microtubules and microtubules from mammalian brain.

Immunofluorescence colocalization of the 90-kDa heat-shock protein and microtubules in interphase and mitotic mammalian cells.

A mouse monoclonal antibody (AC88) that was raised against the 88-kDa heat-shock protein of the water mold, Achlya ambisexualis, and that cross-reacts with the 90-kDa mammalian heat-shock protein (hsp90), and an antibody against tubulin were used to localize hsp90 and microtubules, respectively, in the same cultured rat endothelial and PtK1 epithelial cells by indirect immunofluorescence. AC88 and tubulin antibodies labeled the same structures in cells at all stages of the cell cycle, regardless of whether cells were permeabilized before or after fixation. Labeling of cell structures by both AC88 and anti-tubulin antibodies was identically affected by treating cells with colcemid. Double labeling with AC88 and anti-tubulin antibodies in interphase and mitotic cells is consistent with the conclusion that all microtubules are labeled and that no subclass of microtubules is preferentially labeled. Fluorescent labeling by AC88 was prevented by preabsorption of the antibody with purified rat hsp90 but was unaffected by preabsorption with purified 6S tubulin dimer. In contrast to AC88, fluorescent labeling by an anti-tubulin antibody was prevented by preabsorption with tubulin dimer but was unaffected by preabsorption with rat hsp90. Western-blot analysis demonstrated no cross-reactivity of AC88 for tubulin and no cross-reactivity of the anti-tubulin antibody for hsp90. A polyclonal antiserum fraction from a rabbit immunized with the 89-kDa heat-shock protein from chicken also labeled the mitotic apparatus in dividing cells and, somewhat less distinctly, fibrous structures in interphase cells. Labeling by hsp89 anti-serum was prevented by absorption with hsp90. AC88 also labeled microtubules in cultured mouse (L929 and 3T3), rat (endothelium and TRST), hamster (CHO) and primate (BSC, COS-1 and HeLa) cell lines. The demonstration of colocalization of hsp90 with microtubules should provide a valuable clue to eventual understanding of the cellular function of this ubiquitous, conserved and abundant stress-response protein.

Organization of oligodendroglial membrane sheets. I: Association of myelin basic protein and 2',3'-cyclic nucleotide 3'-phosphohydrolase with cytoskeleton.

Membrane sheets elaborated by cultured murine oligodendroglia provide a unique system for examining associations between myelin proteins and cytoskeletal elements. Interactions can be observed and manipulated more readily than in the multilamellar myelin membrane in vivo. Immunocytochemical staining of 2',3'-cyclic nucleotide 3'-phosphohydrolase (CNPase) shows that it is distributed diffusely in some regions of membrane sheets, but colocalized with tubulin in lacy networks and major veins in other regions. Staining with phalloidin also reveals two distributions of F-actin: 1) small aggregates within the diffuse CNPase regions and 2) filaments colocalized with tubulin and CNPase in the lacy networks and veins. Application of colchicine at 10 micrograms/ml for 4 hr disrupts microtubular structures in the lacy network, while those in major veins remain intact. This suggests that microtubules in the lacy network are treadmilling more rapidly than those in the major veins. The distribution of CNPase and F-actin is not altered under these conditions. In contrast, cytochalasin B disrupts F-actin, microtubules, and CNPase in the lacy networks, indicating that cross-linking between these three proteins is disrupted. Both colchicine and cytochalasin B cause fusion of myelin basic protein (MBP) domains in membrane sheets. This appears to be a consequence of disruption of microtubules in the lacy networks, which normally outline the MBP domains. In summary, these results provide evidence for 1) direct association of CNPase with F-actin and tubulin in cytoskeletal structures and 2) organization of MBP into domains via association with microtubules in the lacy networks.

A novel 58-kDa protein associates with the Golgi apparatus and microtubules.

With the aim of identifying proteins involved in linking microtubules to other cytoplasmic structures, microtubule-binding proteins were isolated from rat liver extracts by a taxol-dependent procedure. The major non-tubulin component, a 58-kDa protein (designated 58K), was purified to homogeneity by gel filtration chromatography. To aid further characterization of 58K, purified preparations of the protein were used as immunogen for the production of monoclonal antibodies. Five different monoclonals were obtained, and each of these reacted on immunoblots of liver homogenates with a single band that comigrated with 58K. Based on the results of immunochemical, peptide mapping, and microsequencing experiments, 58K was found to be unrelated structurally to similarly sized cytoskeleton-associated proteins, such as tubulin, tau, vimentin, or keratin, and to represent a new protein species. Several in vitro properties of 58K were found to be characteristic of microtubule-associated proteins. For instance, 58K cosedimented quantitatively with microtubules out of liver extracts, stimulated polymerization of tubulin, and bound to microtubules in a saturable manner. In contrast to traditional microtubule-associated proteins, however, 58K was not found to be distributed uniformly along microtubules in cells. Immunofluorescence microscopy of cultured hepatoma cells revealed, instead, that 58K is associated principally with the Golgi apparatus. Moreover, Golgi membranes isolated from rat liver were observed by immunoblotting to contain significant levels of 58K, which, upon subfractionation of the membranes, partitioned as if it were a peripheral membrane protein exposed to the cytoplasmic side of the Golgi. These collective results have been evaluated in terms of earlier evidence that the intracellular position and structural integrity of the Golgi relies on the presence and organization of microtubules. In that context, the observations reported here suggest that the in vivo function of 58K is to provide an anchorage site for microtubules on the outer surface of the Golgi.

Differential utilization of beta-tubulin isotypes in differentiating neurites.

beta-Tubulin is encoded in vertebrate genomes by a family of six to seven functional genes that produce six different polypeptide isotypes. We now document that although rat PC-12 cells express five of these isotypes, only two (classes II and III) accumulate significantly as a consequence of nerve growth factor-stimulated neurite outgrowth. In contrast to previous efforts that have failed to detect in vivo distinctions among different beta-tubulin isotypes, we demonstrate using immunoblotting with isotype-specific antibodies that three beta-tubulin polypeptides (classes I, II, and IV) are used preferentially for assembly of neurite microtubules (with approximately 70% of types I and II assembled but only approximately 50% of type III in polymer). Immunofluorescence localization shows that an additional isotype (V) is partially excluded from neurites. Distinctions in in vivo localization of the neuron-specific, class III isotype have also been directly observed using immunofluorescence and immunogold electron microscopy. The sum of these efforts documents that some in vivo functional differences between tubulin isotypes do exist.

Identification and localization of a tau peptide to paired helical filaments of Alzheimer disease.

Amino acid sequencing of a CNBr digest of the tau protein isolated from bovine brain revealed an amino acid sequence of 17 residues, Pro-Gly-Leu-Lys-Glu-Ser-Pro-Leu-Gln-Ile-Gly-Ala-Ala-Pro-Gly-Leu-Lys, which we call peptide I, with heterogeneity at position 11 of glycine (peptide Ia) and proline (peptide Ib); peptide I showed no homology with the previously reported cDNA-derived mouse and human tau sequences. Antisera raised to synthetic peptides corresponding to peptides Ia and Ib labeled all the bovine tau polypeptides recognized by other monoclonal and polyclonal antibodies to bovine tau. Antisera to peptide Ib did not label any mouse tau polypeptides; however, an anti-Ia antiserum labeled two of the four mouse tau polypeptides. Antisera to both peptides labeled paired helical filaments (PHF) as neurofibrillary tangles, plaque neurites, and neuropil threads in Alzheimer disease brain and PHF polypeptides on immunoblots. Immunostaining with anti-Ia antisera of PHF in tissue sections and PHF polypeptides, but not bovine tau, on immunoblots was markedly increased when pretreated with alkaline phosphatase. These studies suggest that (i) the amino acid sequences of some isoforms of tau peptide might be different from that predicted from cDNAs, (ii) a tau peptide that is absent in the predicted sequences is present in PHF in Alzheimer disease, and (iii) tau in PHF is abnormally phosphorylated.

Albumin is a major protein component of transverse tubule vesicles isolated from skeletal muscle.

Despite multiple procedures used to isolate transverse tubule vesicles from rabbit skeletal muscle, few proteins have been identified and shown to be specific to transverse tubule vesicles. Markers for purified transverse tubules have included high affinity dihydropyridine binding, cholesterol content, Mg2+-ATPase activity, (Na+,K+)-ATPase activity, and [3H] ouabain binding. Despite these markers, few proteins from purified transverse tubules can be unequivocally identified using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). In this report we have biochemically and immunologically identified rabbit albumin as a major component of purified transverse tubule membranes from rabbit skeletal muscle. Albumin composed between 5.1 and 9.8% (n = 4) of the total protein in purified transverse tubules based on scans of SDS-PAGE. Furthermore, albumin and other serum proteins are present in preparations of transverse tubules and triads but not in light sarcoplasmic reticulum. Extraction of triads with low concentrations of saponin or sodium dodecyl sulfate completely removes albumin without removing intrinsic membrane proteins. Our results suggest that albumin and other serum proteins are present in the lumen of preparations of transverse tubules and albumin may be used as a marker for the transverse tubules when analyzed on SDS gels.

Tubulin isotype usage in vivo: a unique spatial distribution of the minor neuronal-specific beta-tubulin isotype in pheochromocytoma cells.

The neuronal cells of vertebrates express two beta-tubulin isotypes, called Class II and Class III, that are neuronal specific. In order to determine the distribution of the minor Class III isotype, site-directed antibodies were raised to synthetic peptides representing the carboxyl terminal, isotype-defining domains of the tubulins. These antibodies were applied to PC12 cells at various stages of differentiation. The Class III isotype was found to be expressed in undifferentiated PC12 cells as well as in cells at every stage of differentiation. The concentration of the Class III isotype, relative to the total beta-tubulin complement, did not change significantly. Indirect double immunofluorescence microscopy demonstrated that the Class III isotype was found in the soma and the neurites of differentiated PC12 cells; this spatial pattern of Class III expression paralleled the total beta-tubulin pattern. Although the anti-Class III antiserum could stain in vitro assembled neuronal microtubules in a filamentous pattern, a close examination of the Class III staining pattern in flattened PC12 cells revealed that this isotype was not incorporated into the nonaxoplasmic array of microtubules. Rather, the Class III isotype was localized in a nonfilamentous, granular pattern that was not readily extracted with nonionic detergent. Cells treated with taxol and then flattened and stained showed that the Class III isotype could be induced to assemble into microtubule bundles by taxol. Thus, the minor neuronal beta-tubulin isotype appears to be spatially specialized in its pattern of expression.