Mitochondrial Citrate Synthase Is Immobilized in Vivo *

The enzymes of the tricarboxylic acid cycle in the mitochondrial matrix are proposed to form a multienzyme complex, in which there is channeling of sub-strates between enzyme active sites. However no direct evidence has been obtained in vivo for the involvement of these enzymes in such a complex. We have labeled the tricarboxylic acid cycle enzyme, citrate synthase 1, in the yeast Saccharomyces cerevisiae, by biosynthetic incorporation of 5-fluorotryptophan. Comparison of the 19 F NMR resonance intensities from the labeled enzyme in the intact cell and in cell-free lysates indicated that the enzyme is motionally restricted in vivo , consistent with its participation in a multienzyme complex. The protein concentration in the mitochondrial matrix is thought to be very high, between 270 and 560 mg/ml depending on the functional state of the organelle (1, 2). This very high protein concentration has been proposed to restrict the diffusion of enzymes and metabolites and to promote the association of matrix enzymes. This has led to the suggestion that the reactions of the tricarboxylic acid cycle may occur via channeling of metabolites between enzyme active sites in a multienzyme complex (3–6) or metabolon (7). associations acid cycle and and mitochondrial the cells were in stationary phase at a density of approximately 2 3 10 8 cells/ml. The cells were washed and resuspended in 500 ml of medium containing 2% galactose, 0.67% yeast nitrogen base, and a mixture of amino acids lacking tryptophan and leucine. After 2 h, 50 ml of a 0.2% solution of 5-FTryp was added and the culture incubated for a further 24 h prior to cell harvesting. Enzyme Assay— Cells were disrupted by vigorous agitation in extrac-* and 80% of the amount of enzyme in the original cell preparation. Where we have quoted concen- trations of fluorine label in lysates, these have been corrected for loss of enzyme during the extraction procedure. Therefore the enzyme was diluted by a factor of 7.5 or more, depending on the degree of cell extraction, compared with its concentration in the cell. The equations used to calculate the theoretical T 1 values and line widths for CIT1 have been described previously (26). The relaxation times were calculated assuming that the protein tumbles as a sphere and that the label is held rigidly within the molecule. The structure of pig citrate synthase has been used previously to model the structure of CIT1 (13, 31). Using this approach we calculated the hydrated radius of CIT1 to be 33.3 Å and the rotational correlation time to be 29.5 ns in a solution of aqueous viscosity.


F NMR resonance intensities from the labeled enzyme in the intact cell and in cell-free lysates indicated that the enzyme is motionally restricted in vivo, consistent with its participation in a multienzyme complex.
The protein concentration in the mitochondrial matrix is thought to be very high, between 270 and 560 mg/ml depending on the functional state of the organelle (1,2). This very high protein concentration has been proposed to restrict the diffusion of enzymes and metabolites and to promote the association of matrix enzymes. This has led to the suggestion that the reactions of the tricarboxylic acid cycle may occur via channeling of metabolites between enzyme active sites in a multienzyme complex (3)(4)(5)(6) or metabolon (7).
There is a large body of data that supports the idea of a tricarboxylic acid cycle metabolon. Specific associations have been demonstrated in vitro between tricarboxylic acid cycle enzymes and also between these enzymes and components of the respiratory chain and other proteins in the inner mitochondrial membrane (3). Complexes able to catalyze several consecutive steps of the tricarboxylic acid cycle have been isolated following gentle disruption of both liver mitochondria (8,9) and a number of different microorganisms (10,11). Coupled reactions within complexes of tricarboxylic acid cycle enzymes have been shown to have a kinetic advantage when compared with the completely solubilized systems (3,9). Molecular modeling studies on a citrate synthase-malate dehydrogenase fusion protein, for which there was experimental evidence of substrate channeling (12), showed that there could be a very efficient, electrostatically based, channeling mechanism for substrate transfer between the enzymes active sites (13). Evidence has also been obtained in vivo for a tricarboxylic acid cycle enzyme complex. Disruptions of the genes for citrate synthase and malate dehydrogenase in yeast resulted in cells that were unable to grow on acetate (14). This was despite the fact that there are isozymes in the cytosol which could, in principle, bypass the resulting blocks in the cycle. In the case of citrate synthase, introduction of a structurally similar but catalytically inactive mutant resulted in restoration of tricarboxylic acid cycle function and growth on acetate (15). These studies have been interpreted as indicating the presence of a complex of tricarboxylic acid cycle enzymes in which the enzymes have structural as well as catalytic roles. There is also evidence for substrate channeling in vivo. Analysis of 13 C labeling patterns in metabolites derived from tricarboxylic acid cycle intermediates indicated channeling of succinate and fumarate in the cycle (16,17).
Demonstration of these weak enzyme complexes in situ has been difficult, however, because many of them are dissociated during isolation due to dilution effects. There is also no direct evidence for their presence in an intact cell. We show here, using NMR measurements on a minimally derivatized enzyme of the tricarboxylic acid cycle, that it exists in a motionally restricted form in the yeast mitochondrial matrix in vivo, consistent with its participation in a multienzyme complex.
Plasmid Construction-The coding sequence for yeast mitochondrial citrate synthase 1 (CIT1) (20) was cloned into the Ecl136II site of the plasmid pYES2.0 (Invitrogen Corp.) to generate the plasmid pY-ESYCS1. This plasmid was linearized by digestion with KpnI, the resultant cohesive termini removed with T4 DNA polymerase and the blunt-ended molecule religated with BamHI linkers. This plasmid was digested with BamHI and the 1.5-kilobase pair fragment containing the CIT1 coding sequence was ligated into the unique BglII expression site of the LEU2-expressing plasmid pKV49 (21). These manipulations generated the plasmid pBF208, which expressed yeast mitochondrial citrate synthase 1 (CIT1, EC 4.1.3.7) under the control of a galactoseinducible version of the yeast phosphoglycerate kinase promoter.
Enzyme Labeling-Cells were transformed with the plasmid pBF208 by the method of Hinnen et al. (22). CIT1 was fluorine-labeled by inducing enzyme expression in stationary phase cells in the presence of 5-fluoro-DL-tryptophan (5-FTrp). 1 The labeling protocol employed was similar to that used previously to label phosphoglycerate kinase (23). Briefly, 2.5 ϫ 10 9 cells were used to inoculate a 500-ml culture containing 2% glucose, 2% bactopeptone, and 1% yeast extract. This culture was grown for 24 h, by which time the cells were in stationary phase at a density of approximately 2 ϫ 10 8 cells/ml. The cells were washed and resuspended in 500 ml of medium containing 2% galactose, 0.67% yeast nitrogen base, and a mixture of amino acids lacking tryptophan and leucine. After 2 h, 50 ml of a 0.2% solution of 5-FTryp was added and the culture incubated for a further 24 h prior to cell harvesting.
Enzyme Assay-Cells were disrupted by vigorous agitation in extrac-* This work was supported by the Wellcome Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  tion buffer (50 mM sodium phosphate, 5 mM EDTA, 1% Triton X-100, pH 7.0). Citrate synthase activity was assayed as described in (24). Enzyme activities are expressed per milliliter of cell water, assuming that 1.67 g of cells contain 1 ml of cell water (25). All means are quoted with their S.E.
Cell Immobilization and Perifusion-Cells were immobilized and perifused as described previously (26). The cells (6 g wet weight with 6 ml of 1.8% agarose) were perifused with an oxygenated buffer that had the same composition as that used for protein labeling, except that it was supplemented with 0.002% tryptophan instead of 5-FTrp.
Protein Purification-Mitochondria were isolated as described in Ref. 27. All procedures were performed at 4°C. Isolated mitochondria were disrupted by sonication, after the addition of Triton X-100 to 0.01% (v/v), and the debris and unlysed mitochondria removed by centrifugation (10 min, 15,000 ϫ g). The extract was brought to 85% saturation with (NH 4 ) 2 SO 4 and stirred for 30 min before centrifugation for 30 min at 27,000 ϫ g. The resulting pellet was dissolved in a buffer containing 1.7 M (NH 4 ) 2 SO 4 , 50 mM Tris-HCl, 5 mM EDTA, and 2 mM dithiothreitol, pH 8.0, and applied to a column of octyl-Sepharose (Amersham Pharmacia Biotech) pre-equilibrated with the same buffer. Citrate synthase was eluted from the column with a linear (NH 4 ) 2 SO 4 gradient ranging from 1.7 to 0 M. Fraction purity was assessed by SDS-polyacrylamide gel electrophoresis. Fractions that contained essentially pure CIT1 were pooled and concentrated using an Amicon stirred ultrafiltration cell (YM30 membrane) until the enzyme reached a concentration of greater than 1 mg/ml. CIT1 was stored at 4°C as an (NH 4 ) 2 SO 4 precipitate.
For NMR measurements the purified enzyme was desalted by gel filtration, using NAP-5 columns (Amersham Pharmacia Biotech), into NMR buffer (50 mM HEPES, 130 mM potassium acetate, and 2 mM dithiothreitol, pH 7.2) and then concentrated using Amicon centricon 10 microconcentrators. Samples contained 10% v/v 2 H 2 O for a field frequency lock. pH measurements were not corrected for any deuterium isotope effect. Sucrose was added to give solutions of higher viscosity where required, and the final protein concentration was typically 15-25 mg/ml. NMR Measurements-NMR experiments were performed at a 19 F resonance frequency of 376.29 MHz, as described previously (26). Fluorine-19 chemical shifts are quoted relative to p-fluorophenylalanine standards, either a 100 M internal standard in solutions of the purified enzyme or an external standard contained in a coaxial capillary with cell and lysate preparations. Longitudinal relaxation time (T 1 ) measurements on the purified protein were performed using an inversion recovery sequence. Peak integrals in these experiments were fit iteratively to a three parameter single exponential function.
Viscosity measurements on solutions of the purified enzyme were made on the same samples as used for the 19 F NMR experiments. The diffusion coefficient of water in the samples was measured from proton spectra by pulsed-gradient spin echo techniques (28). The diffusion coefficient was taken to be inversely and linearly proportional to the viscosity of the solution (29,30).
The NMR visibility of the labeled protein in the cell was assessed by comparing its signal intensity in the cell with that in diluted cell extracts. Extracts were prepared by disrupting 6 g of cells by vigorous agitation with glass beads in chilled extraction buffer. The lysates were then dialyzed against extraction buffer without Triton X-100 and their volumes adjusted to 27 ml, which was the sample volume used in cellular perifusions. Enzyme activity was assayed at this point in order to determine the degree of enzyme extraction. Typically the amount of enzyme in a dialyzed lysate was between 60 and 80% of the amount of enzyme in the original cell preparation. Where we have quoted concentrations of fluorine label in lysates, these have been corrected for loss of enzyme during the extraction procedure. Therefore the enzyme was diluted by a factor of 7.5 or more, depending on the degree of cell extraction, compared with its concentration in the cell.
The equations used to calculate the theoretical T 1 values and line widths for CIT1 have been described previously (26). The relaxation times were calculated assuming that the protein tumbles as a sphere and that the label is held rigidly within the molecule. The structure of pig citrate synthase has been used previously to model the structure of CIT1 (13,31). Using this approach we calculated the hydrated radius of CIT1 to be 33.3 Å and the rotational correlation time to be 29.5 ns in a solution of aqueous viscosity.

RESULTS
Protein Labeling-A mitochondrial isoform of citrate synthase (CIT1) was selectively fluorine-labeled in vivo by induc-ing its synthesis, in stationary phase cells, in the presence of 5-FTrp. The addition of labeled or unlabeled tryptophan, following galactose induction of enzyme expression, resulted in an approximately 10-fold increase in the activity of the enzyme. The increase was from 163 Ϯ 7 units/ml cell water (n ϭ 9) to 1630 Ϯ 60 units/ml cell water (n ϭ 10) in the presence of 5-FTryp and from 170 Ϯ 10 units/ml cell water (n ϭ 6) to 2000 Ϯ 100 units/ml cell water (n ϭ 3) in the presence of unlabeled tryptophan. In the absence of additional tryptophan there was only an approximately 1.5-fold increase in enzyme activity to 266 Ϯ 15 units/ml cell water (n ϭ 9). In the presence of 5-FTryp the enzyme concentration after induction was 166 Ϯ 6 M, assuming a specific activity for CIT1 of 100 units/mg and a molecular mass for the dimer of 98 kDa (32). The comparable increase in enzyme activity that occurred following induction of expression in the presence of 5-FTrp or unlabeled tryptophan indicated that the specific activity of the labeled enzyme was not significantly different from that of the unlabeled form. A similar observation has been made previously for the glycolytic enzymes phosphoglycerate kinase, hexokinase, and pyruvate kinase labeled with 5-FTrp (26).
S. cerevisiae contains three isoforms of citrate synthase, CIT1 (20) and CIT3 (33) are mitochondrial, whereas CIT2 is peroxisomal (34). Labeled CIT1 was localized predominantly in the mitochondria as mitochondria prepared from the induced cells contained 91.1 Ϯ 0.6% (n ϭ 6) of the total citrate synthase activity in the cell. The protein labeling procedure had little effect on cell viability. The number of cells that grew on agar plates at the end of the procedure was 81 Ϯ 2% of those that grew prior to the start of the procedure (n ϭ 4). The labeling procedure increases the enzyme concentration by a factor of 10. This level of overexpression of CIT1, however, has been shown to have no significant effect on mitochondrial function (35). 19 F NMR Measurements on the Purified Enzyme-Yeast CIT1 is a homodimer of molecular mass 98 kDa and has six tryptophans per subunit (20,32). Four of the six 5-FTrp resonances were resolved in the spectrum of the purified protein (Fig. 1). The longitudinal relaxation time constant (T 1 ) of the well resolved downfield resonance (resonance 1, Fig. 1), at near aqueous viscosity, was 1.1 Ϯ 0.2 s, in agreement with the value expected from theory of 1.0 s. The T 1 of the envelope of resonances (Fig. 1, resonances 2-6) was also near to 1 s, at 1.1 Ϯ 0.1 s. These data indicate that the tryptophans are relatively immobile and have a correlation time similar to that of the whole protein. The location of the tryptophans within the protein structure is consistent with this apparent lack of mobility. FIG. 1. 19 F NMR spectrum of purified 5-fluorotryptophan-labeled CIT1. The peak numbering is used for reference in the text. The spectrum was acquired with a spectral window of 12 kHz into 16,000 data points. The pulse flip angle was 90°and the interpulse delay 8 s. An exponential line broadening of 30 Hz was applied.
Mapping of the CIT1 sequence onto the structure of the bovine heart and chicken heart enzymes and investigation of the solvent accessibility of the tryptophan residues, with the programs MODELLER and NACCESS (36,37), indicate that all are buried or constrained within the protein structure. The T 1 of the envelope of resonances increased with solvent viscosity (Fig. 2), and this increase showed reasonable agreement with theoretically calculated values (26). The line widths also showed good agreement with theory. The line width of resonance 1 (Fig. 1) was measured at 60 Hz in a solution of nearaqueous viscosity, and the calculated value was 52 Hz. 19 F NMR Measurements in Vivo-Spectra were obtained from cells that had been immobilized in agarose gel threads and maintained in a metabolic steady state, during NMR data acquisition, by perifusion with oxygenated medium (Fig. 3). The metabolic status of the cells was confirmed by 31 P NMR spectra (38) acquired before and after the 19 F NMR experiments. There was no significant loss of citrate synthase activity in the cells during this period. The 19 F NMR data were acquired under fully relaxed conditions, assuming that the T 1 values of the CIT1 resonances were similar to those of the purified enzyme in a solution of near-aqueous viscosity i.e. approximately 1 s.
Unresolved 19 F resonances were observed in cells that had been induced to express CIT1 in the presence of 5-FTrp (Fig.  3C). However similar resonances were also observed in control cells, which had been transformed with the empty vector, pKV49, and taken through the same protein labeling protocol (Fig. 3A). The intensities of these unresolved resonances in the two sets of cells were similar. The concentration of detectable fluorine in the cells overexpressing citrate synthase was 300 Ϯ 30 M (mean Ϯ S.E., n ϭ 3) as compared with 180 Ϯ 20 M (mean Ϯ S.E., n ϭ 2) in the control cells. The concentration of detectable fluorine in dialyzed lysates prepared from the con-trol cells, when expressed on a cell water basis, was 285 Ϯ 45 M (mean Ϯ S.E., n ϭ 3) (Fig. 3B), showing that these resonances were due predominantly to nonspecific incorporation of 5-FTrp into cell proteins and that these proteins were fully visible in the 19 F spectra of the cells. However spectra of dialyzed lysates prepared from cells overexpressing CIT1 (Fig. 3D) were very similar to those obtained from the purified enzyme (Fig. 1), in particular the resolved downfield resonance (resonance 1, Fig. 1), which was never observed in spectra of cells, was clearly visible. Furthermore the concentration of detectable fluorine in these lysates, expressed on a cell water basis, was much higher at 850 Ϯ 110 M (mean Ϯ S.E., n ϭ 3). Thus the labeled enzyme, which was undetectable in 19 F spectra of the cells, was visible in the diluted extracts prepared from these same cells. DISCUSSION There is a considerable body of evidence for the organization of the enzymes of the tricarboxylic acid cycle in a multienzyme complex, in which there is channeling of cycle intermediates (reviewed in (3,4)). The concept of a tricarboxylic acid cycle metabolon, however, has remained controversial as it has been difficult to isolate an intact complex from the cell. The relatively weak interactions between the enzymes, which are favored in the cell by the very high protein concentrations in the mitochondrial matrix (39,40), are disrupted by the dilution that occurs during cell extraction.
In this study we have investigated the rotational mobility of a tricarboxylic acid cycle enzyme, CIT1, in the matrix of yeast mitochondria in vivo using 19 F NMR measurements on a fluorine-labeled enzyme. We have used this technique previously to measure the rotational correlation times of three glycolytic enzymes in yeast. Phosphoglycerate kinase and hexokinase were found to be tumbling in a cytoplasm with a viscosity approximately twice that of water (26), in good agreement with fluorescence measurements of cytoplasmic viscosity in mammalian cells (see for example (41,42)). Pyruvate kinase, however, yielded no detectable NMR signals in vivo, indicating that there was some degree of motional restriction of this enzyme in the cell. This was thought to be due to binding of the enzyme to other cellular macromolecules. CIT1, like pyruvate kinase, showed no detectable fluorine resonances in the intact cell, although it was readily detectable in diluted cell extracts. The 19 F signals that were observed in the cell could be assigned to nonspecific labeling of cell proteins (see Fig. 3). There are several possible explanations for this lack of NMR visibility of CIT1 in the cell.
NMR and fluorescence measurements have indicated a mitochondrial matrix viscosity of between 25 and 37 times that of water (43,44). Such a high viscosity would lead to substantial broadening of the 19 F resonances of labeled CIT1 and an increase in their T 1 relaxation times, resulting in signal saturation under the NMR acquisition conditions employed. For example at a viscosity of 25 centipoise (centipoise ϭ 10 Ϫ2 poise), the calculated line width of the 19 F resonances would be 1200 Hz and the T 1 22 s. With this T 1 the signals detected in the cell would be substantially saturated, and their intensities would be only 26% of the fully relaxed value. However recent time resolved fluorescence measurements on mitochondrially targeted GFP (45) showed the matrix viscosity to be close to that of an aqueous solution. The high apparent viscosity estimated from the earlier fluorescence anisotropy measurements (43) was shown to be explicable by probe binding.
The lack of NMR visibility of labeled CIT1 in vivo could be due to broadening of its resonances by paramagnetic ions present in the mitochondrial matrix. This, however, seems unlikely as the tryptophan residues are at least partially buried within the protein and should thus be inaccessible to paramagnetic ions. Furthermore there is no evidence for paramagnetic ions significantly affecting the relaxation rates of resonances from intramitochondrial metabolites, including the 31 P resonances of ATP (46 -49) and inorganic phosphate (48) and the 1 H resonance of water (44).
The most likely explanation, therefore, for the NMR invisibility of CIT1 is that its 19 F resonances are broadened by an increase in its correlation time, due to binding to other matrix proteins. A complex containing five tricarboxylic acid cycle enzymes has been isolated (8 -11). Assuming that there is one molecule of each enzyme in the complex then this would have a molecular mass of approximately 600 kDa. If the complex behaves as a hard sphere and CIT1 has the same correlation time as the whole complex, then the line width of the fluorine resonances of CIT1 would be increased to 210 Hz. Even with this degree of line broadening the protein could still be detectable in the cell. However since we have overexpressed the enzyme by a factor of 10, it is unlikely that much of the labeled enzyme could participate in such a stoichiometric complex. The enzyme has also been shown to bind, with other tricarboxylic acid cycle enzymes, to the mitochondrial membrane (3,4,14). This would lead to a much larger increase in the enzyme's correlation time and thus the 19 F resonance line widths. Under these circumstances the NMR signals would be broadened beyond detection and therefore membrane binding is a much better candidate as an explanation for the invisibility of CIT1. Immobilization of a GFP-tagged matrix enzyme was observed recently by Partikian et al. (45). By contrast, fluorescence recovery after photobleaching measurements on free GFP showed that its diffusion was relatively rapid, being only three to four times slower than in water (45). In order to explain the rapid diffusion of GFP it was proposed that the matrix proteins are organized peripherally in membrane-associated complexes, thus creating a central aque-ous region with relatively low protein density and low viscosity. Such a domain would allow the rapid and unrestricted diffusion of solutes. This model is consistent with the data presented here demonstrating immobilization of CIT1 in vivo and with previous studies showing that it binds with other tricarboxylic acid cycle enzymes to the mitochondrial membrane.