Structure of the Vacuolar ATPase from Neurospora crassa as Determined by Electron Microscopy*

We have examined the structure of the vacuolar ATPase of Neurospora craesa using negatively stained preparations of vacuolar membranes and of detergent-solubilized and gradient-purified ATPase complexes. We also examined the peripheral sector (VI) of the enzyme after it had been removed and purified. Using different stains, vacuolar membranes displayed ball-and-stalk structures similar to those of the intact mi- tochondrial ATPase. However, the vacuolar ATPase was clearly different from the mitochondrial ATPase in both size and structural features. The vacuolar en- zyme had a much larger head domain with a distinct cleft down the middle of the complex. This domain was held above the membrane by a prominent stalk. Most intriguing was the presence of basal components. These structures appeared to project from the vacuolar membrane near the base of the stalks. Detergent-solubi- lized, gradient-purified ATPases displayed the same head, stalk, and basal features as those found with the intact enzyme on vacuolar membranes. The mitochondrial ATPase was significantly smaller, and no clefted head domains or basal components were observed. When and F1 particles were directly compared, a size and shape between these two soluble ATPase sectors was apparent. V1 retained all of the features seen in the globular head of the intact complex: V-shaped, triangular, and square forms around a stain-filled core.

The endomembrane system of eucaryotic cells contains several organelles: Golgi, lysosomes, vacuoles, secretory vesicles, and coated vesicles, which can be acidified by an ATPdriven proton pump called the vacuolar ATPase (1-5). This enzyme has been purified from many sources, and in recent years several different genes encoding subunits of the enzyme have been isolated. These data show that the vacuolar ATPase is a large, complex enzyme, composed of at least nine different subunits with an aggregate molecular mass of approximately 700 kDa.
Analysis of the sequences of the two ATP binding subunits of the vacuolar ATPase shows that the enzyme is distantly related to the F-type ATPases found in mitochondrial, chloroplast, and eubacterial membranes (6-9). The subunits of both F-type ATPases and vacuolar ATPases are organized into two sectors: a cluster of integral membrane subunits, the specific arrangement of which is not known, and a complex of peripheral subunits which can be removed from the membrane in a water-soluble form. Perhaps the most remarkable aspect of the structure of the F-type ATPase is that the binding site for ATP is in a peripheral, globular structure, held 4.5 nm above the proton-conducting subunits in the membrane by a thin (2-nm) stalk (10). The coupling between proton movement and ATP synthesis/hydrolysis is not understood. A preliminary model of the vacuolar ATPase, based on biochemical and molecular data, suggests that this complex has an overall structure similar to its mitochondrial counterpart (11).
We have used vacuoles from the filamentous fungus Neurospora c r a m to investigate the structure of the vacuolar ATPase. These vacuoles, which are the fungal equivalent of mammalian lysosomes, have relatively high levels of ATPase activity and appear to be thickly studded with vacuolar AT-Pases. In an earlier study we found that negatively staining vacuolar membranes with phosphotungstate revealed balland-stalk structures which superficially resembled the F-type ATPase of mitochondrial membranes but appeared to differ from the mitochondrial enzyme in size and in structural details (12). Treatment of the vacuolar membranes with nitrate removed the peripheral sector of the vacuolar ATPase from the membranes and caused the disappearance of the ball-and-stalk structures. Thus, the structures seen in the electron micrographs appeared to be the vacuolar ATPases. Vacuolar ATPases have also been seen by negative staining of several higher plant tonoplasts (13-15), bovine chromaffin granules (16), and Dictyostelium discoideum acidosomes (17).
In this paper we have examined in detail the structure of the vacuolar ATPase of N. crussa as seen by electron microscopy of negatively stained membranes. We have made a direct comparison with mitochondrial membranes from the same cells. In addition, we have examined the structure of the peripheral sector of the vacuolar ATPase after it has been removed from the membrane. We wished to see if this part of the enzyme, designated VI, is significantly different in size or structure from the F, sector of the mitochondrial ATPase. We also examined the structure of vacuolar ATPase that had been solubilized with detergent and purified on glycerol gradients.

EXPERIMENTAL PROCEDURES
Preparation of Vacuolar and Mitochondrial Membranes-Vacuoles and mitochondria were prepared as described previously (18,19). The purified organelles were lysed and washed in 1 mM EGTA,' pH 7.5, and centrifuged at 100,000 X g for 30 min. Vacuolar and mitochondrial membrane pellets were resuspended in 1 mM EGTA, pH 7.5, to a protein concentration of 2-5 mg/ml, frozen in liquid nitrogen, and stored at -70 "C. Both membrane preparations had specific ATPase Purification of Vacuolar and Mitochondrial ATPases-The vacuolar membrane ATPase was purified essentially as described by Uchida et al. (20). Briefly, EGTA-washed vacuolar membranes were solubilized in 0.5% Zwittergent 3-14 and centrifuged at 180,000 X g for 15 min. The supernatant was layered onto a linear 20-40% glycerol gradient and centrifuged at 200,000 X g for 5.5 h at 4 "C. The ATPase activity migrated halfway through the gradient, and these fractions were treated for electron microscopy as described below. The peripheral complex (VI) of the V-ATPase was prepared as described previously (12), except for one minor change. Prior to the final extraction in KN03 plus ATP, membranes were washed in 100 mM KN03 in the absence of ATP. The peripheral F, sector of the mitochondrial ATPase was prepared using a chloroform extraction method as described previously (21).
Other Methods-Gel electrophoresis was performed with samples dissolved in sodium dodecyl sulfate as described (22). Proteins were visualized by silver staining. Vacuolar and mitochondrial ATPase activities were assayed as described previously (18).
The molecular diameters of the soluble complexes were estimated by gel permeation chromatography on a 1 X 15-cm column of S-300 resin (Pharmacia LKB Biotechnology Inc.). Vacuolar and mitochondrial peripheral complexes were diluted 1:lO in vacuolar ATPase assay mix containing 5 mM Na2ATP, 5 mM Mg2S04, 10 mM NH4Cl, 10 mM PIPES (adjusted to pH 7.4) and concentrated to 50 pl (protein concentration of 1-2 mg/ml) in Centricon tubes spun at 10,000 x g for 20 min. Samples were chromatographed in ATPase assay mix at a flow rate of 5 ml/h. Thyroglobulin (669 kDa, 18-nm diameter), apoferritin (440 kDa, 12 nm), and catalase (232 kDa, 10.4 nm) were used as molecular mass standards. All procedures were performed at room temperature.
Transmission Electron Microscopy-Vacuolar and mitochondrial membranes were diluted to a protein concentration of 0.5-1.0 mg/ml in 1 mM EGTA, pH 7.5. One pl of a suspension was applied to a Formvar-coated 200-mesh grid. After 1 min, the droplet was drawn off and allowed to semidry. One pl of a 1% solution of phosphotungtate, uranyl oxalate, or ammonium molybdate (all at pH 6.5) was applied over the sample and immediately wicked off with filter paper.
The remaining stain was quickly blown dry (within a few seconds) to minimize exposure of the samples to the heavy metal solution. Gradient-and column-purified ATPase fractions were diluted 1:lO in 1 mM EGTA, pH 7.5, and concentrated in Centricon tubes. The ATPase samples were subsequently applied to grids as described above. Specimens were examined and photographed on a Jeol lOOB electron microscope at a primary magnification of 67,000 X with an accelerating voltage of 80 kV.
Image Processing and Analysis-Vacuolar and mitochondrial AT-Pases were digitally imaged and processed using the NIH Image 1.43 program integrated with a Pulnix TM-545i CCD camera/Macintosh I1 computer. To avoid subjective selection of images, we measured every apparent ATPase in Figs. 2 and 4 and 65 particles in Fig. 5, A and B. Briefly, ATPase projections were magnified 75 X to a final magnification of 5,000,000 X. Images were digitally processed using filtering functions to reduce background noise and sharpen edges of stain-protein boundaries. Measurements of the computer-enhanced images were made using digital calipers and mean values and standard deviations for dimensions determined. A negatively stained catalase standard was used to calibrate exact sizes.

Analysis of Vacuolar and Mitochondrial
Membranes-Vacuolar membranes were treated with three types of negative stains as described under "Experimental Procedures." As shown in Fig. 1, the vacuolar ATPase was clearly visible as a ball-and-stalk structure when the membranes were stained with phosphotungstate (panel A), uranyl oxalate (panel C), or ammonium molybdate (panel D). The highest density and best resolution of these structures were observed when phosphotungstate was used. Occasionally, ATPases were seen in surface view (as in panel B ) , but in most cases, the ball-andstalk structures were most prominent along the edge of the vesicles (panel A). For all subsequent experiments membranes were stained with phosphotungstate.
When mitochondria and vacuolar membranes were compared, both were thickly studded with ball-and-stalk struc-tures; however, the density, size, and shape of the vacuolar ATPase (Fig. 2, A-C) were clearly different from those of the mitochondrial ATPase (panel D). The particles were essentially uninterrupted along the mitochondrial membrane with 10-12 ATPases/100 nm of linear membrane segment. By contrast the vacuolar ATPases appeared in small clusters along the membrane. In regions in which the vacuolar particles were of highest density, with no apparent gaps, there were 8-9 ATPases/100 nm of membrane segment.
The vacuolar ATPase showed three characteristic features. First, the enzyme appeared to have a globular head with a prominent cleft or bifurcation visible in many of the particles (see structures labeled H in Fig. 2, A-C). In many views the stain penetrated to the center of the globular head, giving rise to a V-shaped or bilobed structure. In the case of the intact mitochondrial ATPase, penetration of stain into the head was rarely observed (see panel D ) . Second, the head domain of the enzyme was connected to the membrane by a thin stalk (see S in panels A X ) . Third, the vacuolar membranes appeared to have additional basal components associated with the ATPase (see B in panels A-C). In occasional views, these basal components appeared as projections from the membrane originating near the stalk of the ATPase (see B in panel A ) . These projections measured approximately 7-10 nm long (as measured from the center of stalk-membrane junction) x 2-4 nm wide. With the mitochondrial membranes no apparent basal components were found in the region between the membrane and the globular heads (panel D).
Negatively stained vacuolar ATPases appeared to be significantly larger than the mitochondrial enzymes. We measured several dimensions of both enzymes as summarized in Fig. 3. The globular head of the vacuolar enzyme was 11.5 nm wide x 9.3 nm high. The mitochondrial head sector measured 9.6 nm wide X 8.4 nm high. The stalk of the vacuolar ATPase was 6.6 x 3.1 nm, both longer and thicker than the mitochondrial counterpart, which measured 4.7 X 2.6 nm.
Examination of Detergent-solubilized Vacuolar ATPase-We examined detergent-solubilized ATPases that had been partially purified on a glycerol gradient (see "Experimental Procedures"). This preparation contained both integral membrane and peripheral subunits of the ATPase (18). As seen in Fig. 4, structures were visible which closely resembled those seen on intact vacuolar membranes. The solubilized ATPase had a roughly spherical sector separated from an apparent membraneous domain by a thin stalk. The dimensions of these structures ( n = 14) were very similar to those of the membrane-bound vacuolar ATPase. The head domain measured 12.0 f 0.9 nm wide X 10.5 f 1.3 nm high and the stalk, 6.4 f 1.1 nm long x 3.5 k 0.7 nm wide. The basal components seen on intact membranes (see B in Fig. 2, panel A ) were also apparent on the solubilized enzyme (Fig. 4, far left panel). The putative membrane sector was roughly the same size as the head domain but did not appear to have a regular structure. In some cases the enzymes were clustered, resembling small strips of membrane (Fig. 4, right panels).
Examination of the Detached VI Sector and Comparison with the Fl Sector from Mitochondria-We have shown previously that peripheral subunits of the vacuolar ATPase can be specifically released from the membrane (11). Incubation of membranes with 100 mM KNO, releases a complex of five subunits. When analyzed by size exclusion chromatography (see "Experimental Procedures"), the VI behaved like a particle of 12.5 nm in diameter with an apparent molecular mass of 480 kDa (data not shown). We examined negatively stained peripheral VI complexes in electron microscopy to see if they resembled the globular heads attached to the membranes (  Fig. 5, panel A, relatively homoge-had a diameter of 11.5 nm with an apparent molecular mass neous fields of particles were present with an average diameter of 380 kDa (data not shown). In the electron microscope the of 11.8 & 0.8 nm (n = 65). For comparison, we also examined F,-ATPase had a diameter of 9.2 & 0.7 nm (n = 65), exhibiting the F, sector of the mitochondrial ATPase (panel B ) . When a distinct ring structure with a hollow central region (panel analyzed on the same size exclusion column, the F,-ATPase B ) . F, particles have been shown to lie preferentially upright We then examined the vacuolar peripheral complexes immersed in deeper stain under higher magnification. The particles from the vacuolar membranes had a morphology significantly different from that of the F1 particles. In most projections stain penetrated into the molecules as if filling a cavity (Fig. 5, C and D). Particles did not appear in any preferential orientation. Many V-shaped or bilobed forms were observed, similar to the structures we saw on vacuolar membranes (see Vs in panefs C and D). In some projections substructure within the particles was distinguishable. Particles were found with three and four distinct densities in triangular and square forms (see Ts and Ss in panefs C and D). These latter forms were similar to the intact vacuolar ATPase projections seen in surface view (Fig. 1B). Surprisingly, no Fl-like hexagonal structures were observed.

DISCUSSION
The idea that the vacuolar-type ATPase may structurally resemble the F-type ATPase was suggested by early electron micrographs of negatively stained membranes from bovine chromaffin granules (26,27). These preparations showed balland-stalk structures which were suggested to be the vacuolar ATPases. A later investigation used rapidly frozen and freezedried tissue from specialized proton-secreting cells from rat kidney and toad and turtle urinary bladder (28). These cells appeared to have a V-type ATPase which formed stud-like projections, densely packed on the surface of the plasma membrane. More recent examinations of vacuolar-type membranes from N. crassa, higher plants, and bovine cells have confirmed that the vacuolar ATPase can be visualized in the electron microscope as ball-and-stalk structures which resemble the F-type ATPase of mitochondrial membranes (12)(13)(14)(15)(16)(17).
Vacuolar membranes from N. crassa have high levels of ATPase activity (2-4 pmol/min/mg) and seem to be quite amenable to examination by electron microscopy. When stained with phosphotungstate, a high density of ATPases can be seen with good resolution. Furthermore, the V1 sector of the ATPase is easily released from the membrane and can also be clearly resolved at high magnification after negative staining.
Analysis of the primary sequence of subunits of V-type and F-type ATPases, together with analysis of the subunit composition of the enzymes, strongly suggests that these two types of ATPases are derived from a common ancestor and have similar overall structures (4, 6-9). Therefore, it is somewhat surprising that mitochondrial and vacuolar ATPases can be readily distinguished from each other in electron micrographs. The sizes and shapes of the enzymes are clearly different. Some difference in size is expected because subunits of the vacuolar ATPases are larger than the corresponding subunits of the F-type enzymes. For N. crassa the predicted molecular mass of the V1 sector is 466 kDa (673573481301161 (12)), whereas the mitochondrial F1 sector is approximately 382 kDa (21,29,30). These molecular mass differences are proportional to the difference in the diameters of the head groups seen on vacuolar and mitochondrial membranes, 11.5 nm uersus 9.6 nm. However, if these were filled spheres, the VI would have 72% greater volume than the F1, and the 22% difference in molecular masses does not account for this. Thus, in addition to having larger subunits, the vacuolar ATPase may also have more open space between some of these components.
In the electron micrographs the most characteristic feature of the vacuolar ATPase is a prominent cleft often seen in the V1 sector. Presumably this is caused by a heavily stained cavity in the middle of the enzyme. This feature can also be seen in views of the detached V1, suggesting that the major subunits are held together at the base of the head component in the region attached to the stalk (see model in Fig. 3). When stained with phosphotungstate, most intact mitochondrial ATPases lacked a prominent cleft. However, Gogol et al. (10,25), using unstained F1-ATPase examined in amorphous ice, observed a central cavity in this enzyme. Furthermore, a recent crystal structure of rat mitochondrial F1-ATPase, at 3.6 A resolution, provides evidence that the a subunits are joined at the base of the F1, whereas the p subunits straddle the a subunits but do not contact each other (31). The vacuolar ATPase homologue of the F1 p subunit is the 67-kDa subunit. In N. crassa this polypeptide is 30% larger than the mitochondrial @ subunit. Thus, the V-type and F-type AT-Pases could have similar overall quaternary structures, but the V-type may look somewhat different because of larger subunits in a more open cluster in the V1 sector.
The other distinct features of the vacuolar ATPase are the prominent stalk and the additional basal components. The globular head of the enzyme, which contains the ATP-binding subunits, appears to be held 6.6 nm above the surface of the membrane by a stalk that is approximately 3.1 nm wide. One of the most intriguing aspects of the F-type and V-type ATPases is the question of how changes in the ATP binding sites on Fl and Vl can be coupled to proton movement through a membrane sector which, on a molecular scale, is very far away.
For the vacuolar ATPase we do not know which subunits make up the stalk and basal components. Vl detached from the membrane contains five different subunits but does not appear to have an extended stalk in the electron micrographs. Thus, the stalk may be composed, at least in part, of membrane-associated polypeptides. Membranes stripped of the V1 still have small protrusions, but they are too small and variable to be easily identified. The basal projections seen on the vacuolar membrane (Fig. 2) have no counterpart in F-type ATPases. In the future we plan to use antibodies raised against specific polypeptides to see if some of the distinctive components visible in the electron microscope can be specifically labeled.