THE JOURNAL or BPXOGKXL CHEMWTRY

Vol. 253, No. 8, Issue of April 25, pp. 2846-2951, 1978 Printed in U.S.A.

Electron Microscopy of Metal-shadowed and Negatively Stained Microtubule Protein STRUCTURE OF THE 30 S OLIGOMER*

(Received for publication, July 27, 1977)

ROBERT B. SCHEELE$ AND GARY G. BORISY

From the Laboratory of Molecular Biology, University of Wisconsin, Madison, Wisconsin 53706

Microtubule protein purified from porcine brain was fixed at low protein concentration with glutaraldehyde under conditions which maximize the relative concentration of the ring-shaped 30 S oligomer. Fixed oligomer was separated from glutaraldehyde and other protein species by column chromatography. The fixed, isolated oligomer was deposited on electron microscopy grids, dehydrated, and then critical point-dried before shadow-coating with carbon/platinum alloy at a fixed angle. Analysis of the shadow lengths observed by electron microscopy revealed that the height of the 30 S oligomer is 15 nm. Microtubule protein deposited on electron microscope grids at high protein concentrations was examined by the negative stain technique and found to contain apparent stacks of oligomer from which the number of tubulin dimers per turn of the ring and the distance between turns could be determined. The number of subunits per turn was determined as 13.8. The distance between turns was found to be 7.4 nm, indicating that the 15 nm high, shadowed oligomers consisted of two turns. Additional in- formation from the literature is considered and a model is presented for the oligomer. The model is a helix of 29 tubulin dimers and five high molecular weight protein molecules arranged so as to preserve intersubunit bonding patterns found in microtubules.

The two preceding papers (Marcum and Borisy, 1978; Vallee and Borisy, 1978) have dealt with the hydrodynamic

properties and composition of the 18 S and 30 S oligomers of microtubule protein and have developed a model of the 30 S oligomer. This model is a double-layered ring structure with 13 to 15 tubulin dimers in each layer and several high molecular weight protein molecules bound to and projecting from its surface.

The conclusion that the 30 S ring is a double-layered species was based on indirect evidence. In order to establish directly whether the 30 S species is a double-layered particle, it is necessary to know its height. Electron microscopy of metal- shadowed samples allows the determination of particle

* This work was supported by National Institutes of Health Grant GM-21963. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “‘uduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

$ To whom correspondence should be addressed.

heights, and the method of critical point drying provides a means of drying samples without the destructive effects of surface tension encountered in air drying. Therefore, we have fixed samples of 30 S microtubule protein oligomer with glutaraldehyde to preserve the oligomeric structure, critical point-dried the samples, and shadowed them with carbon/ platinum alloy. In combination with the results of negative staining of apparent stacks of the 30 S oligomers, we have used the results of the shadowing experiments to refine the model for the 30 S species. The refined model is a helix of 29 tubulin dimers and 5 HMW’ molecules which retains bonding patterns between the tubulin and the HMW that we believe are likely to exist in microtubules.

MATERIALS AND METHODS

Preparation of Microtubule Protein - Microtubule protein for shadowing was prepared from porcine brain by the assembly-disas- sembly procedure (Borisy et al., 1975). Protein was stored as a pellet of microtubules at -80” and given an additional assembly-disassem- bly purification step in the experimental buffer, 0.05 M l,l-pipera- zinediethanesulfonic acid, 0.1 rnM MgCl,, and 1.0 mM GTP, pH 6.3 (ionic strength, 0.076) just prior to use. This buffer was chosen to eliminate the 18 S oligomer and to increase the concentration of the 30 S ring (Marcum and Borisy, 1978). Protein concentration was determined by the method of Lowry et al. (1951) as modified by Schacterle and Pollack (1973).

Fin&ion and Isolation of Rings -Purified microtubule protein in the pH 6.3 buffer was incubated at 15” for 10 min and mixed with I/XI volume of freshly prepared 2% glutaraldehyde in the same buffer. The mixture was incubated at 15” for 5 min and the reaction was slowed by immersion of the tube in ice water. The fixed protein was then isolated from unreacted glutaraldehyde by molecular sieve chromatography at 4”. A 2-ml sample containing 20 mg/ml of sucrose was applied to a lo-ml bed of Sephadex G-25 by underlayering. The protein was then fractionated on a column of (1.5 x 23 cm) of Bio- Gel A-15m to separate fixed aggregates from tubulin dimers. Col- umns were equilibrated and eluted in the pH 6.3 buffer without GTP or in 0.10 M NaCl.

The optical density of the column effluent was monitored at 280 nm by an Altex ultraviolet column monitor model 152 equipped with a cell of 2 mm path length. The Bio-Gel A-15m column was calibrated with tobacco mosaic virus to determine the void volume. The included volume was determined on similar columns, for which the ratio of void volume to included volume was the same, by measuring the elution volume of 1.0 M NaCl as detected refractometrically.

Ultracentrifugation -Fixed or unfixed microtubule protein was examined for the presence of specific oligomers by centrifugation at

1 HMW designates the class of high molecular weight proteins which are known to bind to and project from microtubules of neuronal origin (Murphy and Borisy, 1975).

2846

Electron Microscopy of Microtubule Protein 30 S Oligomer 2847

40,000 rpm in a Spinco model E analytical ultracentrifuge. Schlieren optics and double channel cells were used throughout.

Electron Microscopy - Fenestrated Formvar films (FernBndez- Moran and Finean, 1957) were applied to 200-mesh copper grids and vacuum-coated with a heavy layer of carbon for reinforcement. Very thin carbon films vacuum-coated on freshly cleaved mica were floated on distilled water and applied as windows over the holes in the carbon-reinforced Formvar films (Fernandez-Moran et al., 1966).

Fixed protein fractions from the Bio-Gel A-15m column were diluted as needed with the pH 6.3 buffer and a small volume of bromegrass mosaic virus in 1.0 mM ethylenediaminetetraacetic acid, pH 4.9, was added to aid in determination of the shadowing angle. A drop (10 ~1) of this mixture was applied to a grid for 1 min, rinsed with 3 or 4 drops of distilled water, and immersed in an acetone:water mixture (1:4). Grids prepared in this manner were then carried through a graded acetone:water series to 100% acetone and critical point-dried from CO, in a Samdri PVT-3 critical point drier (Tousimis Research Corp., Rockville, Md.).

Grids were shadowed by evaporation of a 4- to 5-mg pellet of carbon/platinum alloy (Ladd Research Industries, Burlington, Vt.) at an initial pressure of less than 1O-6 torr in a model E12E vacuum evaporator (Edwards High Vacuum Ltd., Crowley, England). The shadowing angle was nominally tan-’ 0.2; (length = 8 cm, height = 1.6 cm).

For negative staining, samples of microtubule protein prepared by the assembly-disassembly procedure in 0.10 M 1,4-piperazinedi- ethanesulfonic acid, 0.1 mM MgC&, and 1.0 mM GTP, pH 6.9 were applied at h’ h ig protem concentration to carbon-coated Formvar grids. The grid was rinsed first with 3 to 4 drops of 1% cytochrome c and then with 3 or 4 drops of distilled water and then 3 to 4 drops of 1% uranyl acetate. Excess uranyl acetate was pulled away from the grid by touching it with the edge of a piece of filter paper. Samples were then air-dried.

Shadowed and negatively stained samples were observed and photographed in a Philips model 300 electron microscope at a beam potential of 80 kV. The magnification factors for the electron micrographs were determined through the use of a shadowed carbon replica of a waMe grid ruled with 21,600 lines/cm (Ladd Research Industries, Burlington, Vt.).

RESULTS

Fixation and Isolation of Fixed Rings-In our initial at- tempts to determine the height of the rings, unfixed samples of microtubules depolymerized at low temperature were ex- amined. Microtubule protein was prepared under conditions known to produce a 30 S centrifuge peak and visible rings by negative staining and was applied to grids and shadowed. However, no regular structures of any kind were observed. The reasons for this failure were not clear but two possibilities were that the rings were not stable under the conditions required for critical point drying and that the presence of 6 S dimer obscured the images of the rings. We therefore decided to attempt to fix the rings in solution with glutaraldehyde and then to separate the fixed rings from glutaraldehyde and 6 S microtubule protein. Several variations of the fixation proce- dure were tried in efforts to optimize the conditions for obtaining fixed rings which resembled unfixed 30 S rings as closely as possible. The most successful procedure developed was to fix rings at 15” by the addition of r/lo volume of 2% glutaraldehyde for 5 min. Under these conditions rings were stable to dissociation by the addition of NaCl to a concentra- tion of 0.3 M, as judged by electron microscopy of negatively stained samples, indicating adequate fixation (Marcum and Borisy, 1978). Shorter fixation times resulted in reduced numbers of visible rings.

Samples fixed at several concentrations from 1.8 to 7.1 mg/ ml were subjected to analytical ultracentrifugation either after separation from glutaraldehyde on the Sephadex G-25 column or after separation of the fixed oligomers from mate- rial retarded by the Bio-Gel A-15m column. Fig. 1 shows a schlieren pattern of microtubule protein from the leading peak

FIG. 1. Ultracentrifuge schlieren pattern of glutaraldehyde-fixed microtuble protein, 1.6 mg/ml in 0.10 M NaCl. Protein was fixed at 3.8 mg/ml under conditions producing only 6 S protein and the 30 S oligomer (see “Materials and Methods”). Before centrifugation, the fixed oligomer was separated from 6 S protein and unreacted glutaraldehyde on a column of Bio-Gel A-15m in 0.10 M NaCl. Rotor speed, 40 x lo3 rpm; phaseplate angle, 46”; srO,w = 41.3 S.

of the Bio-Gel A-15m column after fixation at 3.8 mglml. The absence of any 6 S peak shows that the oligomers have been sufficiently well fixed to prevent depolymerization. The oligo- mer peak (s~,,~ = 41.3 S) is somewhat skewed toward the leading edge, but its relative lack of spreading indicates some degree of homogeneity in the sedimenting particles.

In addition to fixing the rings, it was important to avoid overfixing the preparation and inducing the formation of particles not originally present in solution. It was observed that the sedimentation coefficient and mass fraction of the oligomer peak increased with increasing concentration of protein during fixation (data not shown), indicating that fixation at high protein concentration increased the mass of the oligomer. The fixation procedure was therefore applied to microbubule protein at very low concentrations to minimize transfer of 6 S material to rings. Fixed protein, when fraction- ated on the Bio-Gel A-15m column, gave rise to two well defined peaks with little intervening material. When fixed at 0.5 mg/ml, protein carried through this procedure showed an early peak with Kd = 0.06 and a later peak 2.5 times as large with K, = 0.8. The K, values of these two peaks were similar to those of microtubule protein 30 S rings and 6 S tubulin dimer, respectively (Vallee and Borisy, 1978). The relative areas of the peaks were similar to those in observations of the mass fractions of dimer and oligomer in unfixed microtubule protein at low concentrations (Marcum and Borisy, 1978), indicating that at this low protein concentration very little 6 S material became transferred to the oligomeric particles during fixation.

Electron Microscopy of Fixed and Shadowed Rings -Sam- ples of fixed microtubule protein from various fractions through the early peak of the Bio-Gel A-15m column were prepared for shadowing and electron microscopy. Brome mo- saic virus was added to the fixed preparations to serve as an internal standard for size and shadowing angle. Micrographs were taken at a magnification of 81,000 or 83,000 of fields containing at least one virus particle and one ring. Fig. 2 shows some micrographs of a sample taken from the trailing edge of the early peak of the Bio-Gel A-15m column (K, = 0.18). The brome mosaic virus was identified as spherical particles of approximately 30 nm diameter (Incardona and Kaesberg, 19641. Measurements of numerous particles showed that virus diameters were fairly uniform as were their shadow lengths within a given field. Ring-shaped particles were also visible although they were not always precisely circular in

2848 Electron Microscopy of Microtubule Protein 30 S Oligomer

FIG. 2. Shadowed rings from the trailing edge of the leading peak eluted from the Bio-Gel A-Em column. Microtubule protein 30 S ring oligomers were fixed with glutaraldehyde, separated from glutaraldehyde on a Sephadex G-25 column, and size fractionated on a column of Bio-Gel A-15m. Brome mosaic virus was then added as an internal standard for size and shadow angle (light spheres) and the sample applied to thin carbon films over fenestrated Formvar- coated grids,.critical point-dried, and shadowed with carbonlplati- num. Magnification, x 81,000. Apparent virus diameter, 30.6 f 0.2 nm. Panels a and d show the shadows with rounded ends expected of flat rings. All panels show pointed, bilobed shadows expected for helical structures.

outline. In contrast to the virus shadows, the ring shadow lengths were not uniform, indicating some variability in the ring heights.

The shadows of the rings had several types of shapes. In addition to the rounded end (Fig. 2, a and d) expected for the shadow of a flat ring, pointed (Fig. 2a) and bilobed shadows (Fig. 2, a, b, c, e, and f, were also seen. These shapes imply that the ring structure may be other than that of a flat ring, a point which will be pursued under “Discussion.” There are two other features of the ring images which require comment. First, in spite of the evidence that the 30 S rings contain several HMW molecules per ring with portions which project from the ring surface (Vallee and Borisy, 19781, such projec- tions were not seen in the shadowed samples. In addition, the outside diameters of the shadowed rings was 46 + 3 nm which is close to the range observed in negative staining (Vallee and Borisy, 19781, but the wall thicknesses were somewhat larger (17 ? 7 nm) and the inside diameters consequently smaller than those observed by Vallee and Borisy (19781. This feature may be related to our failure to observe the expected projec- tions and may be the result of fixation of initially rather flexible long molecules against the surface of the ring by the glutaraldehyde treatment. Such collapse of the projecting portion of the high molecular weight molecules has been observed by Amos (1977) in negatively stained samples of

microtubules. Some of the increased wall thickness of the shadowed rings is undoubtedly due to the presence of the carbon/platinum shadow coat.

The conditions for shadowing the samples were developed by trial and error. The nominal shadowing angle was kept approximately constant at tan-’ 0.2 while the source-to-sam- ple distance and the size of the carbon/platinum pellet were varied. If distances less than about 8 cm or pellet weights greater than 5 mg were used shadow coats were obviously too heavy. This condition was most easily judged by the presence of columnar, rather than elliptical shadows for the virus particles. If conditions resulting in the deposition of a lighter coat were used the shadows of the particles were indistinct, making it impossible to locate their ends for length measure- ments. Under the conditions used, the shadows were just distinct enough to measure on the films. Subsequent printing of the images resulted in the higher contrast seen in Fig. 2.

In order to determine the height of a given particle the following parameters were measured directly on the films with a scale graduated in 0.1 mm: virus particle diameter, d,.; virus shadow length, l,.; and ring shadow length, l,.. To a first approximation the height, h, of a ring can be calculated from the relation h = d,..Z,/Z,.. This method does not require that the diameter of the virus particle be known beforehand and thus avoids the difficulty that would be introduced by assum- ing a particle diameter for the virus which is known to vary with solution conditions (Incardona and Kaesberg, 1964). However, the method does depend on the assumption that the virus particles are not flattened during drying.

The effect of shadow-coat thickness on the effective shadow- casting heights of virus and rings must also be considered. To this end, we have shown analytically2 that the shadow-coat thickness on the background field is of the order of 5% of the diameter of the virus and that the effective error in determi- nation of the height of an object whose height is approximately the same is of the order of 2.5% if the maximum width of the virus particle along a line perpendicular to the shadow direc- tion is taken as the virus diameter. The error is even less for determination of the height of shorter particles. The ring heights were therefore calculated according to the relation h = d,:l,/l,. without application of any correction factors.

A frequency distribution of ring heights as measured on micrographs from the trailing edge fraction of the early peak of the Bio-Gel A-Em column is shown in Fig. 3. The greatest number of rings had a height of 15 nm, and the mean and standard deviation of the distribution were 18 ‘-c 4 nm. Approximately 55% of the rings were 13 to 17 nm high. Fractions from the leading edge or center of the peak, or pooled fractions from samples fixed at higher protein concen- trations, had distributions which included the 15-nm class of particle heights but also showed a larger amount of the higher particles than is seen in Fig. 3. But even in the leading edge and center fractions of samples fixed at 0.5 mglml, the proportions of rings with heights of 13 to 17 nm were 20 and 21% respectively. Thus we infer that, although under the fixation conditions used some 6 S dimers became fixed to rings, the proportion of fundamental 15-nm rings escaping such modification is significant.

Negative Staining of Unfixed Microtubule Protein -It had been previously noted (Borisy et al., 19751 in negatively stained preparations of rings that apparent stacks of rings were sometimes seen in side view. The width of these struc- tures was the same as the diameter of rings, but the lengths

z Unpublished calculations.

Electron Microscopy of Microtubule Protein 30 S Oligomer

0.3

2 9 0.2 0 .r

5 ‘g 0.1

z IL

0 1 Ik II I5 I9 23 27 31

Ring Heights (nm)

FIG. 3. Ring height frequency distribution in shadowed rings from the trailing edge (K,, = 0.18) of the leading peak (Kd = 0.06) from the size fractionation column, n = 62. Conditions as noted for Fig. 2.

varied from about 30 nm to more than 250 nm. End views were also seen and both views showed some structural features which may shed light on the structure of rings.

Samples of microtubule protein at high concentrations in 0.10 M 1,4-piperazinediethanesulfonic acid, 0.1 mM MgCl*, and 1.0 mM GTP, pH 6.94, were applied to grids and negatively stained according to the procedure outlined under “Materials and Methods.” As shown in Fig. 4 (a, b, and c) apparent stacks were seen with transverse striations which appear to be individual rings seen from the side. The spacing of these striations was measured in micrographs taken by Dr. J. B. Olmsted and compared to the ring height measurement made in the shadowing experiments. Measurement of the spacing in 17 stacks representing 100 apparent layers of rings gave an average spacing of 7.38 ? 0.03 nm. This result leads to the conclusion that the 15 nm mode seen in the distribution of ring heights represents two layers of protein subunits, con- firming a prediction by Vallee and Borisy (1978) on the basis of hydrodynamic information.

Also seen in the micrographs where side views of stacks were seen are circular images of very high contrast but otherwise similar in appearance to the rings seen in the surrounding field and also at low protein concentrations (see Fig. 4, d, e, and f). These high contrast, end-on views of stacks, when found in heavily stained areas of the grid, are sufficiently clear to distinguish subunits in part of each circle. Although in no case have subunits been counted unambigu- ously all the way around the circle, the number of subunits in a circle can be estimated from the number of subunits in a given fraction of a cricle, divided by that fraction. Some of the micrographs used were taken by Dr. D. B. Murphy. Counts of eight circles and a total of 36 subunits gave an average number of subunits per circle of 13.8 * 0.1.

DISCUSSION

It was found that light fixation with glutaraldehyde of preparations of microtubule protein containing 30 S rings results in the stabilization of the rings against depolymeriza- tion to 6 S microtubule protein and against dissociation by salt. Some 6 S material was fixed to the rings but this was minimized by working at low protein concentrations. Under these conditions the major reaction was apparently the cross- linking of tubulin subunits contained within the ring as opposed to cross-linking of free tubulin dimers to each other or to the ring. However, the sedimentation coefficient of the fixed oligomer at low concentration was higher than that for un-

FIG. 4. Stacks of microtubule protein rings, side views (a to c) and end-on views (d to fl. High concentrations of microtubule protein were applied to carbon-reinforced, Formvar-coated grids and negatively stained. Note that the end striations of the side views sometimes do not extend all the way across the particle, suggesting a helical structure (a and 5). Subunits can be counted around only one side of each end-on view (d to f), indicating that subunits in adjacent layers of the stack lie along lines not parallel to the axis of the stack (see text). Note also that rings are seen with much lower contrast (f, arrow), distinguishing them from the stacks. Magnifi- cation, x -250,000.

fixed oligomer which suggests that another intraoligomer cross-linking reaction occurred. Vallee and Borisy (1978) have shown that inclusion of several high molecular weight mole- cules projecting from the ring surface leads to lower sedimen- tation coefficients than would result without them. If our failure to observe these projecting molecules is the result of their having been futed to the ring surface in a collapsed position, as we have suggested (see “Results”), then their contribution to frictional drag would be greatly reduced and the sedimentation coefficient correspondingly increased. Thus, although there is a 35% increase in sedimentation coefficient upon fixation of the rings, the basic subunit struc- ture is probably not much changed by fixation. In addition, chromatography on Bio-Gel A-15m was used to separate aggregation products from the fixed 30 S rings. Fractions from the oligomer peak isolated by sieve chromatography were subjected to the carbon/platinum shadowing procedure. The trailing edge fraction contained ring-shaped particles of the lowest heights seen. Thus we conclude that the trailing edge fraction of lightly fixed rings represents the state of microtu- bule protein rings existing in solution. Under these conditions more than half of the fixed rings were revealed by electron

2850 Electron Microscopy of Microtubule Protein 30 S Oligomer

microscopy of carbon/platinum-shadowed samples to be 13 to 17 nm in height. The height of one layer of stacked rings seen from the side in negatively stained preparations was 7.4 nm, leading to the conclusion that the rings fixed by glutaralde- hyde were largely double-layered species. Counts of subunits seen in end-on views of these stacks gave 13.8 subunits/layer.

Parenthetically, it should be noted that the views of stacks failing to show distinct subunits all the way around the ring are just what is to be expected from a multilayered structure except for the special case in which the subunits in all layers are situated along lines parallel to the axis of the stack. In general, it is to be expected that the stack is being viewed from a position not exactly on the axis. In the general case the subunits along only one side of the stack will lie along the line of sight and their outlines will reinforce one another. At other positions around the circle the subunits will not be in line, and the outline of subunits in different layers will be visible at small intervals around the circle with no clear outline of individual subunits distinguishable.

Vallee and Borisy (1978) have shown that the outside diameter of 30 S rings is 39 nm and that the wall thickness is 4.5 nm. Thus the centers of subunits would be situated at a radius of 17.25 nm from the center of the ring. The distance between the centers of two adjacent subunits would then be 7.6 to 7.7 nm for 14 or 13.8 subunits/turn, respectively. Since the (Y, /3 dimer of tubulin is known to be of the order of 8 nm in length (Amos and Klug, 1974), the subunits seen in the negatively stained rings or stacks must be dimers rather than cy or /3 protomers. A working hypothesis, then, for the struc- ture of the 30 S oligomer is that it is a two-layered structure containing about 28 tubulin dimers. This hypothesis is a minimal one and is subject to refinement as will be discussed below.

One potential refinement of the hypothesis assumes that the structure of the 30 S ring is helical. I f this is the case, then the number of subunits per turn need not be integral. Thus a measurement of 13.8 subunits/turn does not need to be consid- ered an approximation to 14/turn.

The suggestion that the structure may be helical is sup- ported by the electron micrographs of the shadowed samples and by the negatively stained side views of the stacks. Fig. 2 shows some bilobed and pointed shadows which are to be expected in shadowing short helical sections in certain rota- tional positions relative to the source. The rounded shadows to be expected from other rotational positions are also seen. Also, the side views of the stacks are sometimes seen to have one more protrusion on one side than on the other (Fig. 3, a, and b). This feature is also expected from helices in some orientations on the grid. Although these features are to be expected of helices the possibility that they are artifacts has not been excluded. As such, these aspects of the micrographs do not rule out a model of the 30 S species as a stacked pair of flat rings, and of the stacks as extensions of this same kind of interlayer relationship.

On the other hand, there are other reasons for preferring a helical model for both the 30 S species and the stacks. To this point, the contribution of the HMW proteins to the structure of the 30 S species has been neglected, although HMW is known to be an essential component of its structure. The HMW content of the 30 S rings was found by Vallee and Borisy (1978) to be 1 mol of HMW/ 5 or 6 mol of tubulin dimer. Because of the usual high specificity of protein-protein inter- actions, the bonding pattern that exists in microtubules be- tween HMW and tubulin might be expected to be retained in

the 30 S ring. Thus, we shall consider a helical model with this feature.

Although the precise interactions between HMW and tubu- lin are not known even in microtubules, the distribution of binding sites for HMW over the microtubule lattice is known. Amos (1977) has shown, using optical diffraction of electron micrographs of microtubules that the binding sites are distrib- uted over the tubulin lattice in a superlattice with a repeat distance along a protofilament equal to the length of 12 dimers (96 nm). The microtubule lattice can be alternatively de- scribed in terms of 8-start left-handed or &start right-handed helical families of tubulin dimers (Amos and Klug, 1974; Johnson and Borisy, 1975). In these terms the superlattice of HMW molecules can be described as having binding sites distributed at intervals of three dimers along two of the 8- start helices which are separated by a four-dimer distance in the axial direction.

If it is assumed that the long HMW molecule binds to a domain which extends along the groove between two 8-start helices of the microtubule (Vallee and Borisy, 1978), such a bonding arrangement can be preserved in a model for the 30 S ring which has the features demanded by the known physical and chemical properties of the ring. In this model the 30 S ring consists of 29 tubulin dimers and 5 high molecular weight molecules in a one-start helical arrangement of slightly more than two turns with 14.1 dimers/turn, as shown in Fig. 5. Dimers are positioned relative to each other along the helix in nearly the same way as they are positioned relative to each other along the 8-start helix in microtubules. In addition, if one considers the ring model as a 14-start helix, the dimers are positioned relative to each other along the 14-start helices with nearly the same relationship they have along the proto- filament in microtubules, i.e. end to end. If, as proposed above, the HMW molecule binds in the groove between 8-start helices in microtubules a binding site would consist of three pairs of dimers such that the HMW molecule binds to the (Y protomers of three dimers and to the p protomers of three more dimers in the adjacent 8-start helix. The same bonding pattern is preserved in the helical model for the 30 S ring which provides exactly live complete binding sites if 29 dimers are used with 14.1 dimers/turn. If all of these sites are filled the molar ratio of high molecular weight protein to tubulin dimer is 15.8 which is very close to the value of 1:5 or 1:6 reported by Vallee and Borisy (1978). The high molecular weight molecules would project from the surface of the ring in the same way in which they project from the microtubule surface (Murphy and Borisy, 1975; Vallee and Borisy, 1977) as suggested by the trypsinization experiments of Vallee and Borisy (1978).

Other properties of the 30 S ring are also matched by this model. Using the measured outer diameter of 39 nm and wall thickness of 4.5 nm (Vallee and Borisy, 19781, the distance between two 14-start helices is calculated to be 5.5 nm which is to be compared with the distance between protofilaments in microtubules of 5.0 to 5.4 nm (Erickson, 1974a; Amos and Klug, 1974; Cohen et al., 1975). In addition, the numbers of subunits per turn (14.1) corresponds quite closely to the number counted in end-on views of the stacks (13.8). This model would also give rise to a distribution of shadow lengths closely resembling Fig. 3 if it were rotated through a large number of orientations relative to a stationary light source.

The bonding relation of the high molecular weight protein to the tubulin dimers in the model of Fig. 5 may explain why single-layered rings are infrequently or never seen in the

Electron Microscopy of Microtubule Protein 30 S Oligomer 2851

5.5 n m -, 7.6 nm+ Id A

-39nm

FIG. 5. A schematic drawing of the arrangement of tubulin di- mers in the helical model of the 30 S microtubule protein oligomer. Each ovoid represents a tubulin dimer. There are 29 dimers arranged such that those in one turn of the helix meet those in an adjacent turn end-to-end, corresponding to the interdimer relationship ob- served in the protofilaments of microtubules. There are 14 such prototilament-like tracks situated 5.5 nm apart and tilted such that the pitch of the helix is 7.4 nm. The dimers are distributed along the helix at intervals of 7.6 nm, corresponding to the interdimer rela- tionships occurring in the g-start helices of microtubules. In this arrangement the contact zone between turns of the helix (cross- hatched areas) corresponds to similar zones between g-start helices in microtubules where high molecular weight microtubule-associ- ated proteins are presumed to bind. One end of each of six dimers constitutes one such binding domain. Accordingly the asymmetric structural unit for the helix is 6 tubulin dimers and 1 HMW molecule. The drawing is not made to scale.

shadowing experiments. Ordinarily the additional stability conferred by completion of the first turn of a helical polymer would be expected to lead to a single turn species as the smallest stable unit. The number of bonds per added subunit does not increase beyond that stage. However, HMW is known to promote the formation of 30 S rings (Vallee and Borisy, 1978) and to stabilize microtubules (Murphy et al., 1977). Thus it is’to be expected that the smallest helical structure in which HMW can participate in closure would be the smallest stable structure. For the helical lattice shown in Fig. 5 an outside diameter of 39 nm requires 29 tubulin dimers and 5 HMW molecules for this smallest stable species. This species repre- sents only the initial member of a potentially infinite series of species formed by the addition of more HMW molecules and tubulin dimers in a molar ratio of 1:3. Such a series may correspond to the stacks seen in negatively stained samples of microtubule protein at high concentrations and may be repre- seted by some of the particles seen to cast longer shadows.

Helical aggregates of tubulin and microtubule protein which are different from microtubules have been reported previously (Marantz and Shelanski, 1970; Fujiwara and Til- ney, 1975; Erickson, 1975). These aggregates form in the presence of vinblastine and appear to be X-start open coils of tubulin dimers. The diameter of the vinblastine-induced struc- tures is similar to that of the 30 S rings and the stacks we have studied, but the relationships among these structures are currently unknown.

There are two features of the model shown in Fig. 5 which are based mainly on unproven assumptions and which there- fore allow the existence of alternative models. These two features are that the model is helical and that dimers along a

turn of the ring are related to each other in the way dimers are related in the &tart helix of microtubules. There are no data which show conclusively that the ring is not composed of two flat layers of subunits, nor is it clear that the turns or layers of the ring do not correspond to protofilaments of a microbubule in the internal positioning of tubulin dimers (Kirschner et al., 1974; Erickson, 1974b). However, the known distribution of the binding sites for HMW on the tubulin lattice of microtubules (Amos, 1977) admits only one such binding site for every 12 tubulin dimers along a given protofi- lament, which would then suggest only 1 HMW moelcule/30 S ring, between the two layers. Parts of three binding sites for HMW could be accommodated by 26 to 29 dimers if a protofi- lament were wrapped in a helix of two turns, but no two-turn configuration of a protofilament exists which would give rise to a ring 15 nm in height or which would preserve the HMW bonding pattern observed in microtubules. Consequently, we prefer the model shown in Fig. 5. Finally, it should be noted that the helical structure for the 30 S oligomer is not incom- patible with the existence of a flat-layer ring structure and it is possible that both forms exist, perhaps in equilibrium with each other.

Acknowledgments-We are grateful to Professor Paul Kaes- berg for a gift of the brome mosaic virus and to Professor Hans Ris for the use of his critical point drying apparatus. We thank both of them for advice and assistance throughout this study. We also thank Mr. Steven T. Limbach and Mrs. Jalaine Limbach for help with the figures.

REFERENCES

Amos, L. A. (1977) J. Cell Biol. 72, 642-654 Amos, L. A., and Klug, A. (1974) J. Cell Sci. 14, 523-549 Borisy, G. G., Marcum, J. M., Olmsted, J. B., Murphy, D. B., and

Johnson, K. A. (1975lAnn. N. Y. Acad. Sci. 253, 107-132 Cohen, C., DeRosier, D., Harrison, S. C., Stephens, R. E., and

Thomas, J. (19751 Ann. N. Y. Acad. Sci. 253, 53-59 Erickson, H. P. (1974a) J. Cell Biol. 60, 153-167 Erickson, H. P. (1974b) J. Suprumol. Struct. 2, 393-411 Erickson, H. P. (1975) Ann. N. Y. Acad. Sci. 253, 51-52 Fern&ides-Moran, H., and Finean, J. P. (1957) J. Biophys. Biochem.

Cytol. 3, 725-748 Fernandez-Moran, H., van Bruggen, E. F. J., and Ohtsuki, M.

(1966) J. Mol. Biol. 16, 191-207 Fujiwara, K., and Tilney, L. G. (1975) Ann. N. Y. Acad. Sci. 253,

27-50 Incardona, N. L., and Kaesberg, P. (1964) Biophys. J. 4, 11-21 Johnson, K. A., and Borisy, G. G. (1975) in Molecules and Cell

Mouement (Inoue, S., and Stephens, R. E., eds) pp. 119-141, Raven Press, New York

Kirschner, M. W., Williams, R. C., Weingarten, M., and Gerhart, J. C. (1974) Proc. Natl. Acad. Sci. U. S. A. 71, 1159-1163

Lowry, 0. H., Rosebrough, N. J., Farr, A. L., and Randall, R. J. (1951) J. Biol. Chem. 193, 265-275

Marantz, R., and Shelanski, M. L. (1970) J. Cell. Biol. 44, 234-238 Marcum, J. M., and Borisy, G. G. (1978) J. Biol. Chem. 253, 2825-

2833 Murphy, D. B., and Borisy, G. G. (1975) Proc. Natl. Acad. Sci. U. S.

A. 72, 2696-2700 Murphy, D. B., Johnson, K. A., and Borisy, G. G. (1977) J. Mol.

Biol., 117, 33-52 Schacterle, G. R., and Pollack, R. L. (1973) Anal. Biochem. 51, 654-

655 Vallee, R. B., and Borisy, G. G. (1977) J. Biol. Chem. 252, 377-382 Vallee, R. B., and Rorisy, G. G. (19781 J. Biol. Chem. 253.2834-2845