Glucose Catabolism in Brain INTRACELLULAR LOCALIZATION OF HEXOKINASE*

A major energy source in brain is glucose, which is committed to metabolism by hexokinase (Type I isozyme), an enzyme usually considered to be bound to the outer mitochondrial membrane. In this study, the subcellular location of hexokinase in brain has been rigorously investigated. Mitochondrial fractions containing hexokinase (greater than 500 milliunits/mg protein) were prepared by two different procedures, and then subjected to density gradient centrifugation before and after loading with barium phosphate, a technique designed to increase the density of the mitochondria. The gradient distribution patterns of both unloaded and loaded preparations show that brain hexokinase does not distribute exclusively with mitochondrial marker enzymes. This is particularly evident in the loaded preparations where there is a clear distinction between the peak activities of hexokinase and mitochondrial markers. The same observation was made when the mitochondrial fraction of either untreated or barium phosphate-loaded mitochondria was subjected to titration with digitonin. In fact, at concentrations of digitonin, which almost completely solubilize marker enzymes for both the inner and outer mitochondrial membranes, a significant fraction of the total hexokinase remains particulate bound. Electron microscopy confirmed that particulate material is still present under these conditions. Significantly, hexokinase is released from particulate material only at high concentrations of digitonin which solubilize the associated microsomal marker NADPH-cytochrome c reductase. Glucose 6-phosphate, which is known to release hexokinase from the brain "mitochondrial fraction" also releases hexokinase from this unidentified particulate component. These results on brain, a normal glucose utilizing tissue, differ from those obtained previously on highly glycolytic tumor cells where identical subfractionation procedures revealed a strictly outer mitochondrial membrane location for particulate hexokinase (Parry, D. M., and Pedersen, P. L. (1983) J. Biol. Chem. 258, 10904-10912). It is concluded that in brain, hexokinase has a greater propensity to localize at nonmitochondrial receptor sites than to those known to be associated with the outer mitochondrial membrane.

A major energy source in brain is glucose, which is committed to metabolism by hexokinase (Type I isozyme), an enzyme usually considered to be bound to the outer mitochondrial membrane.
In this study, the subcellular location of hexokinase in brain has been rigorously investigated. Glucose 6-phosphate, which is known to release hexokinase from the brain "mitochondrial fraction" also releases hexokinase from this unidentified particulate component. These results on brain, a normal glucose utilizing tissue, differ from those obtained previously on highly glycolytic tumor cells where identical subfractionation procedures revealed a strictly outer mitochondrial membrane location for particulate hexokinase (Parry, D. M., and Pedersen, P. L. (1983) J. Biol. Chem. 258, 10904-10912).
It is concluded that in brain, hexokinase has a greater propensity to localize at nonmitochondrial receptor sites than to those known to be associated with the outer mitochondrial membrane. Past work in this and other laboratories has shown that the hexokinase activity of highly glycolytic cancer cells is (a) markedly elevated (1, 2), (b) less resistant to glucose 6phosphate inhibition (3), and (c) bound in large amounts (50-90%) to the mitochondrial fraction (3-6) where it has preferred access to ATP generated by oxidative phosphorylation (7). Rigorous subfractionation studies, including digitonin fractionation and density gradient centrifugation, before and after loading the mitochondria with barium phosphate (4-6) have shown that hexokinase activity associated with the tumor mitochondrial fraction is indeed bound to the mitochondria and is not bound to contaminating membranes. Additional studies have shown that tumor hexokinase is bound to the outer mitochondrial membrane (3-5) where the protein "porin" appears to constitute at least one part of the receptor complex (8).
In contrast to highly glycolytic tumor cells, the hexokinase activity of normal tissues is quite low (1, 2). Subfractionation studies conducted in this laboratory on kidney (9) and liver (6) show that the particulate form of hexokinase distributes with microsomal markers rather than mitochondrial markers. These results on normal tissues differ from many reports on brain (for a review see Ref. 10) where the particulate location of the Type I enzyme is usually regarded as exclusively mitochondrial. However, evidence attributing this particulate activity to a mitochondrial location has not been rigorous. In fact, in submitochondrial fractionation studies with digitonin, only a few concentrations of digitonin were used (11) or the release of microsomal markers was not documented (12, 13). Although somewhat convincing, the immunochemical approach (13) has not been able to distinguish between hexokinase bound to mitochondria and that bound to closely associated structures (i.e. microsomes or cytoskeletal proteins). Finally, localization of hexokinase based solely on density gradient centrifugation has not eliminated all contaminating particles (14). Certainly, purified brain hexokinase has the capacity to bind to the outer mitochondrial membrane of rat liver mitochondria (15), but these findings alone cannot be taken as evidence for the enzyme's subcellular location in its tissue of origin.
As past work in this laboratory has revealed striking differences between the particulate distribution of hexokinase in normal and transformed tissues (5,9), it seemed important to examine rigorously the distribution of this important enzyme in brain which, similar to tumor cells, utilizes glucose as a major fuel source.

Methods
Preparation of Mitochondria-Animals were killed by cervical dislocation.
Skin was cut awav from the skull and reflected forward. The sharp point of the scissors was inserted into the foramen and the skull cut along its lateral sutures to completely expose the brain. The entire brain was scooped from the calvarium with a chilled spatula and placed into ( After centrifugation of the resuspended pellet at 600 X g for 8 min, the resulting pellet was discarded and the supernatant combined with the previous supernatant and centrifuged at 10,000 X g for 10 min. The pellet was saved on ice and the decanted supernatant centrifuged again at 10,000 X g for 10 min. The supernatant was discarded and the pellet resuspended along with the previous pellet in about 40 ml of isolation medium and centrifuged at 10,000 X g for 10 min. The supernatant was discarded and the pellet washed several times with isolation medium to remove the loosely packed white layer on top of the tightly packed brown layer. The washes were discarded and the pellet resuspended in about 40 ml of isolation medium and centrifuged at 600 X g for 8 min. The pellet was discarded and the decanted supernatant centrifuged at 10,000 X g for 10 min. The pellet was washed again with isolation medium and the remaining pellet resuspended in the final appropriate volume of isolation medium. Mitochondrial preparation II was obtained by the procedure described by Clark and Nicklas (17) Pedersen et al. (21). Briefly, the particulate components were fixed with 3% glutaraldehyde followed by 2% osmium tetroxide.
After dehydration, the specimens were embedded in an epoxy-based resin mixture, sectioned, and post-stained with uranyl acetate and lead citrate. Sections were examined in a Zeiss EM 10 electron microscope.

Comparison of Mitochondrial
Preparations I and II-Two procedures were employed to isolate brain mitochondria.
The major difference between the two procedures as emphasized in detail under "Experimental Procedures" is the use of density gradient centrifugation in mitochondrial preparation II. Consistent with a number of previous studies on brain hexokinase (For a review see Ref. lo), it can be seen in Table I that both methods result in a "mitochondrial fraction" which contains significant amounts of hexokinase. The values obtained of 520 milliunits/mg for preparation I and 679 milliunits/mg for preparation II are considerably higher than that characteristic of the mitochondrial fractions of other normal tissues like liver and kidney which exhibit values of -1 and 43 milliunits/mg, respectively (6, 9). Although preparation II, relative to preparation I, is somewhat enriched in both hexokinase and the mitochondrial marker enzymes, monoamine oxidase and succinate-cytochrome c reductase, it is important to note that this preparation still contains significant contamination by microsomes and cytoplasm. The microsomal enzyme, NADPH-cytochrome c reductase, and the cytoplasmic enzyme, lactic dehydrogenase, are still present in preparation II. As emphasized below, the latter enzyme arises from contamination of the mitochondrial fraction with synaptosomes. It seems clear from these studies that commonly used Hypotonic treatment was performed by centrifuging mitochondrial preparation I (3.8 mg/ml) at 5000 x g for 10 min and resuspending the pellet in water at the initial volume. After incubating on ice for 10 min, the mitochondrial suspension was diluted e-fold with Hmedium, centrifuged at 5000 X g for 10 min and then resuspended in H-medium at the initial volume. A control was treated identically except that H-medium was used in place of water. Detergent treatments were performed by incubating mitochondrial preparation I on ice for 15 min with different detergents. Controls were incubated in the absence of detergent. Hexokinase was assayed under isotonic conditions. schemes for brain mitochondria, whether based solely on differential centrifugation (preparation I) or on both differential centrifugation and density gradient centrifugation (preparation II) are not of sufficient purity to definitively ascribe the localization of a given enzyme to the mitochondria per se. Specifically, as it concerns the localization of hexokinase, these studies indicate that microsomes, mitochondria, and synaptosomes are all potential localization sites.

Latency of Hexokinase
Associated with the "Mitochondrial Fraction"-To test for latency, the hexokinase activity of mitochondrial preparation I was assayed in the presence of 0.23 M sucrose. About 50% of the total activity present is assayable under this condition.
The latency of hexokinase in brain mitochondrial preparations has been attributed to its entrapment in synaptosomes or pinched-off nerve endings formed during tissue disruption.
Assay of total hexokinase activity requires pretreatment of the mitochondrial preparation with membrane-disrupting techniques. Hypotonic treatment (i.e. osmotic shock) performed as described in the legend to Table II, exposed about 60% of the latent hexokinase activity present. Pretreatment of the mitochondrial preparation separately with three different detergents increased assayable activity slightly over 100%. These findings indicate that synaptosomes or pinched-off nerve endings containing entrapped hexokinase are present in the brain mitochondrial fraction.
Particulute Location of Herokinase Associated with the Mitochondrial Fraction-The particulate association of brain hexokinase within the mitochondrial fraction was investigated initially by subjecting preparation II to two sucrose gradient centrifugation procedures, based on those designed by Whittaker and Barker (22) for the subcellular fractionation of brain tissue. In the first procedure, a two-step sucrose gradient was used to resolve the mitochondrial preparation into a mitochondrially enriched fraction and into two other fractions enriched in the contaminants, synaptosomes and myelin. The percent distribution of hexokinase and lactate dehydrogenase (a cytoplasmic marker) in each of these fractions is reported in Table III. As expected, the proportion of lactate dehydrogenase is highest in the contaminating synaptosomal fraction. Hexokinase, on the other hand, is distributed in approximately equal proportions in the mitochondrial and synaptosomal fractions. An increase in the hexokinase to lactate dehydrogenase ratio in the mitochondrial fraction emphasizes that synaptosomal contamination alone cannot account for the particulate association of brain hexokinase in the mitochondrial preparation.
In the second procedure, a linear sucrose gradient was employed in an attempt to better resolve mitochondrial preparation I. As seen in Fig. 1, the distribution of hexokinase in the resultant fractions overlaps with that of mitochondria (monoamine oxidase and succinate-cytochrome c reductase), synaptosomes (lactate dehydrogenase), and microsomes (NADPH-cytochrome c reductase). It can be seen that almost 50% of the total hexokinase activity fails to sediment with the mitochondrial markers (compare Fig. 1, A and B) and the remaining activity overlaps significantly with the microsomal marker (compare Fig. 1, A and C). Also in Fig. 1, heterogeneity is observed in the distribution of the mitochondrial markers, an observation which is consistent with that reported by others (23-25).
Although the above results on mitochondrial preparations I and II show that a significant portion of the total hexokinase activity in brain does not distribute with mitochondrial markers, it cannot be concluded whether the remaining activity is of mitochondrial or nonmitochondrial origin.

Effect of Loading Mitochondria with Barium Phosphate on the Particulate
Location of Hexokinase-To further assess to what extent mitochondria serve as a particulate locus for Mitochondrial preparation I was fractionated by a density gradient centrifugation procedure based on that of Whittaker and Barker (22). In this procedure approximately 0.5 ml of the mitochondrial suspension was layered on top of a gradient of 2.0 ml steps of 0.8 and 1.2 M sucrose and centrifuged at 24,000 X g for 2 h. After centrifugation, 1.8 ml was removed from the top of the tube, stopping halfway between the two distinct turbid bands. This is the myelin fraction. The remaining 2.3 ml of gradient was removed as the synaptosomal fraction. The pellet at the bottom of the tube was resuspended in 0.5 ml H-medium yielding the mitochondrial fraction. Mitochondrial preparation I was fractionated by a density gradient centrifugation procedure based on that of Whittaker and Barker (22) as described under "Experimental Procedures." In this procedure approximately 0.5 ml of the mitochondrial suspension was layered on top of a linear gradient of 0.8-1.6 M sucrose and centrifuged at 24,000 X g for 2 h. After centrifugation, fractions were collected from the bottom of the tube and assayed for enzyme activities.
brain hexokinase, mitochondrial preparations were treated with barium phosphate immediately prior to density gradient centrifugation. Barium and phosphate are taken up by mitochondria when incubated, in vitro, in the presence of a metabolic energy source. Accumulation of these ions within the mitochondria results in the formation of insoluble complexes and thereby increases the density of these organelles (6,9). Two experimental approaches were taken. One was to restrict accumulation of barium phosphate by incubating mitochondria in the presence of a limiting amount of barium (intermediate or submaximal loading). The other approach was to permit maximal accumulation of barium phosphate by incubating mitochondria in the presence of a saturating amount of barium (maximal loading). Fig. 2 illustrates the density gradient distributions of succinic dehydrogenase and hexokinase after submaximal loading of mitochondria with barium phosphate. As expected under these conditions, two peaks of succinic dehydrogenase activity are obtained corresponding to "light" and "heavy" (submaximally loaded) mitochondria which peak in fractions 11 and 6, respectively. Although some hexokinase activity under these conditions moves to the more dense region of the gradient, it does not peak (i.e. in fraction 6) with the mitochondrial marker enzyme.
Reported in Table IV are the results obtained after maximal loading of mitochondria. In this experiment the entire sucrose gradient was separated as a whole from the pellet and both fractions were assayed for hexokinase and various marker enzymes. As expected, only the specific activities of the mitochondrial enzymes, monoamine oxidase and succinate-cytochrome c reductase were increased in the pellet, while the specific activity of the synaptosomal marker, lactic dehydrogenase, was markedly decreased. Significantly, the specific activity of hexokinase in the pellet remained essentially the same as that in the starting mitochondrial preparation, a Mitochondria were loaded submaximally with barium phosphate as described under "Experimental Procedures." Fractionation was carried out by layering 0.5 ml of the treated mitochondrial suspension on top of a gradient of 0.45-ml steps of 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, and 2.5 M sucrose and centrifuged at 115,000 x g for 2.5 h. Following centrifugation, fractions were collected from the bottom of the tube. Mitochondria (preparation II) were loaded maximally with barium phosphate as described under "Experimental Procedures." Fractionation was carried out by layering 0.5 ml of the treated mitochondrial suspension on top of 0.45-ml steps of 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, and 2.5 M sucrose and centrifuging at 115,000 X g for 3 h. Following centrifugation, the entire gradient was removed and the remaining pellet resuspended in 1 ml of 0.25 M sucrose. Values in parentheses indicate percentage of the corresponding specific activities of the starting mitochondrial preparation. Results are typical of two exoeriments. finding difficult to reconcile with a mitochondrial location for this enzyme. It is important to note also that 42.2% of the activity of the microsomal enzyme, NADPH-cytochrome c reductase was recovered in the pellet, and similar to hexokinase, its specific activity was not markedly altered.

Correlation of Hexokinnse Release with Microsomal Release on Digitonin
Fractionation of the Mitochondrial Preparations--At this point it seemed important to employ an alternative approach. Therefore, digitonin was used to fractionate the mitochondrial preparations. The results obtained with preparation I are illustrated in Fig. 3. As seen in panel A, synaptosomal membranes are relatively sensitive to disruption by digitonin as evidenced by the release of lactate dehydrogenase at the low digitonin concentration. The biphasic response in the release of adenylate kinase probably represents the disruption of synaptosomes at low digitonin concentration (releasing cytosolic adenylate kinase) and the disruption of the outer mitochondrial membrane at higher concentrations (releasing the adenylate kinase from the inter- Mitochondrial preparation I was treated with increasing concentrations of digitonin, as described under "Experimental Procedures," in the presence of the following conditions. Panels A and B, in the presence of H-medium plus 0.1% BSA and at a protein concentration of 2 mg/ml. Pan& C and D, in the presence of H-medium plus 1 mM EDTA and at a protein concentration of 1.3 mg/ml. Panels E and F, in the presence of 0.15 M KCl, 5 mM Hepes, and 0.5 mM EDTA and at a protein concentration of 1.4 mg/ml. The remaining particulate material was assayed for adenylate kinase (A), lactate dehydrogenase (0, hexokinase (x), monoamine oxidase (0), succinate cytochrome c reductase (A), NADH-cytochrome c reductase (0), and NADPH-cytochrome c reductase (W). Results are expressed as percentage of activity recovered in the pellet from control. membrane space). That hexokinase is highly resistant to cligitonin solubilization, even after the outer mitochondrial membrane enzyme monoamine oxidase has been almost completely released, is evident in panel B. (It will be noted also that no hexokinase is released on disruption of the synaptosomal membrane (i.e. along with lactate dehydrogenase release panels A and B) indicating that hexokinase within the synaptosomes is particulate.) To ensure that the resistance of hexokinase to solubilization by cligitonin is not simply a consequence of rebinding (after initial solubilization) to newly exposed particulate binding sites, digitonin fractionation was carried out also under conditions which are unfavorable for hexokinase binding. The digitonin fractionation profiles in panels C and D of Fig. 3 were obtained in the presence of EDTA and in the absence of Digitonin fractionation was carried out, as described under "Experimental Procedures," in the presence of H-medium plus 0.1% BSA and 0.5 mM EDTA. Panels A and B show fractionation of the mitochondrial preparation II at a protein concentration of 1.5 mg/ml. Pands C and D show fractionation of barium phosphate loaded mitochondria at a protein concentration of 3.8 mg/ml. Mitochondria were loaded maximally with barium phosphate as described under "Experimental Procedures." The remaining particulate material was assayed for adenylate kinase (A), monoamine oxidase (0), hexokinase (x) succinate-cytochrome c reductase (A) In panel A, the mitochondrial preparation was pretreated with a low concentration of digitonin (0.2 mg/mg protein) to first expose that fraction of the total hexokinase enclosed within synaptosomes. After centrifugation, the mitochondria were treated with glucose 6phosphate as described under "Experimental Procedures." After glucose B-phosphate treatment, the particulate material was assayed for hexokinase (X), monoamine oxidase (0), and succinate cytochrome c reductase (A). Results are expressed as percentage of activity recovered in control mitochondria treated identically but in the absence of glucose B-phosphate. In panel B, the mitochondria were treated with sufficient digitonin to remove also the outer membrane enzyme monoamine oxidase, prior to glucose 6-phosphate treatment.
chondrial marker enzymes. In fact, 70% of the hexokinase activity was recovered in the pellet even though almost all of the inner mitochondrial membrane enzyme succinate-cytochrome c reductase had been released (Fig. 3 The pellets obtained after centrifuging the digitonin-treated mitochondrial preparation and a control were assayed for hexokinase and monoamine oxidase activity and processed for electron microscopy as described previously. The ratio of the percentage activity of hexokmase to that of monoamine oxidase increased from 1 in control preparation (panel A) to 10 in the digitonin treated preparation (panel B). Magnification is X 13,200. , yielded similar results (Fig. 4, panels A and B) with hexokinase release correlating best with that of the microsomal enzyme. Even after further purification of the mitochondria (i.e. barium phosphate-loaded mitochondria) the release of hexokinase on digitonin fractionation still correlated better with the release of the microsomal enzyme than the mitochondrial enzymes (Fig. 4, panels C and D).
Site Specificity of Particulate Hexokinase after Treatment with Digitonin-Glucose 6-phosphate is known to solubilize particulate-bound hexokinase in a variety of tissues (4-6, 9). Failure of digitonin to solubilize hexokinase in the brain mitochondrial fraction described here might suggest that hexokinase had been released during cell fractionation procedures and then rebound very tightly to nonspecific sites. To test this possibility, the brain mitochondrial fraction was subjected to successive treatments with glucose 6-phosphate. Significantly, these treatments completely released bound hexokinase while leaving unaffected the binding of monoamine oxidase or succinate-cytochrome c reductase. When sufficient digitonin was used to first remove the outer membrane marker monoamine oxidase, glucose 6-phosphate addition still induced hexokinase release in soluble form (Fig. 5,  panel B).
These results with glucose 6-phosphate render it unlikely that the resistance of hexokinase to digitonin solubilization (Figs. 3 and 4) is due to its release and nonspecific rebinding during cell fractionation.
Electron Microscopic Analysis of the Brain Mitochondrial Fraction before and after Treatment with Digitonin-Electron microscopy was used to examine the membrane disrupting effect of digitonin.
Shown in panel A of Fig. 6 is an electron micrograph of mitochondrial preparation II. Evident are synaptosomes, some myelin material, and numerous mitochondria, the outer membranes of which are intact. The effects of digitonin can be observed in panels B-D. After incubation of the mitochondrial preparation with a low concentration of digitonin (0.20 mg/mg protein), most synaptosomes were ruptured and the myelin material somewhat disrupted, but mitochondria remained intact (panel B). At higher digitonin concentrations (1.4 and 2.0 mg/mg protein), essentially all structures are severely disrupted leaving membrane fragments which can still be pelleted at 12,000 x g (panels C and D). It is these fragments which retain substantial hexokinase activity but essentially no monoamine oxidase activity.