Glyceraldehyde Phosphate Dehydrogenase, Phosphoglycerate Kinase, and Phosphoglyceromutase of Escherichia coli SIMULTANEOUS PURIFICATION AND PHYSICAL PROPERTIES*

Abstract A relatively simple seven-step procedure is described whereby glyceraldehyde phosphate dehydrogenase, phosphoglyceromutase, and phosphoglycerate kinase can be purified from extracts of Escherichia coli. The purified preparations (50 to 100 mg) of the first two enzymes were homogeneous when examined by polyacrylamide gel electrophoresis at pH 9 and by sedimentation techniques. The purified fraction of phosphoglycerate kinase contained small amounts of contaminating proteins. The molecular weights and s20, w0 values of the purified proteins were, respectively: glyceraldehyde phosphate dehydrogenase, 144,000 and 7.5 S; phosphoglycerate kinase, 43,700 and 3.5 S; and phosphoglyceromutase, 56,300 and 4.8 S. The amounts of these enzymes in cell extracts varied with the carbon source which the cells utilized during growth, but the three activities were invariably present in constant proportion to one another. Calculation of the number of molecules of each enzyme present in such extracts indicated a nonequimolar distribution (ratio of glyceraldehyde phosphate dehydrogenase to phosphoglycerate kinase to phosphoglyceromutase, 1:2.4:3.6).

The purified preparations (50 to 100 mg) of the first two enzymes were homogeneous when examined by polyacrylamide gel electrophoresis at pH 9 and by sedimentation techniques. The purified fraction of phosphoglycerate kinase contained small amounts of contaminating proteins. The molecular weights and s&.~ values of the purified proteins were, respectively: glyceraldehyde phosphate dehydrogenase, 144,000 and 7.5 S; phosphoglycerate kinase, 43,700 and 3.5 S; and phosphoglyceromutase, 56,300 and 4.8 S.
The amounts of these enzymes in cell extracts varied with the carbon source which the cells utilized during growth, but the three activities were invariably present in constant proportion to one another. Calculation of the number of molecules of each enzyme present in such extracts indicated a nonequimolar distribution (ratio of glyceraldehyde phosphate dehydrogenase to phosphoglycerate kinase to phosphoglyceromutase, 1:2.4:3.6).
The Embden-Meyerhof pathway of glycolysis has been of interest for many years, and several of its enzymes have been extensively studied.
In the last decade the enzymology of bacterial cells has been of increasing interest because so many important biological processes can be investigated to advantage in these simpler systems.
The species Escherichia coli has proved to be especially useful, and a larger body of basic biological information has been accumulated on this organism than for virtually any other in nature.
A relatively simple procedure is reported here whereby these three enzymes of the Embden-Meyerhof pathway can be obtained in large amounts (50 to 100 mg) from a modest quantity of E. coli cells (450 g). The molecular weight and s!$~ values of the purified proteins are also reported. The separate purification of E. coli glyceraldehyde phosphate dehydrogenase has been described by Allison and Kaplan (6), and several properties of this protein have been compared with those of glyceraldehyde phosphate dehydrogenases from other sources (6,7). A preliminary report of the present work has also appeared (8 Assays for Enzymatic Activity-Enzymatic assays were carried out in l-ml incubation mixtures, following the absorbance of reduced pyridine nucleotide at 340 mp for 1 to 2 min at 23" in a Cary recording spectrophotometer. The assay for glyceraldehyde phosphate dehydrogenase was conducted according to the procedure of Allison and Kaplan (6), except that 2-mercaptoethanol (3 mM) was included in the mixture; the reaction was initiated by addition of glyceraldehyde-3-P after enzyme and other components had been previously incubated for 3 to 5 min. Phosphoglycerate kinase was assayed in the reverse direction of glycolysis in an incubation mixture described by Adam (II), substituting 20 mM Tris-chloride, pH 7.5, for triethanolamine and including 1 mM ATP; the reaction was initiated by addition of enzyme. Activity of phosphoglyceromutase was measured according to the method of Czok and Eckert (12), substituting 20 mM Tris-chloride, pH 7.5, for triethanolamine and including 1 mu glycerate-3-P and 2.5 times the amount of enolase recommended by those authors; the reaction was initiated by addition of enzyme.
All enzymatic activities are expressed in international units (micromoles of substrate converted per min under the given assay conditions), with E = 6.22 x 1Oa M-' cm-l at 340 rnp for NADH (13) High speed sedimentation equilibrium determinations of molecular weights were carried out as described by Yphantis (14). A symmetrical condensing lens mask with 0.5-mm slits was employed, and fringe photographs were taken on spectroscopic II-G plates. With application of appropriate baseline corrections, plates were examined both for vertical fringe displacement as a function of radius and for radius as a function of fringe count.
Points of lowest concentration with fringe displacements of <lOO p were omitted in constructing  (14). Agreement between the two types of plots was invariably better than =t2%. All plates were read on a Nikon microcomparator, according to the technique of Trautman (15). Other Methods-Protein concentration was measured according to Lowry et al. (16) and by a microbiuret method (17). Both techniques, when standardized with bovine albumin, yielded similar estimates of the purified E. coli proteins described in this paper.
Polyacrylamide gel electrophoreses were carried out as described by Davis (18). Radioactivity was measured in a low background gas flow counter.

Enzymatic Activities in Cell Extra.&
Earlier work by Vogell et al. (19) and by Pette, Luh, and Bticher (20) has established that the enzymes of the Embden-Meyerhof pathway from triose phosphate isomerase to enolase are members of a constant proportion group.
That is, in cell extracts from different species, ranging from yeast cells to insects to mammals, as well as from different organs of the same animal, these activities are present in relatively constant proportion to one another.
For example, in each extract examined the ratio of glyceraldehyde phosphate dehydrogenase to phosphoglycerate kinase to phosphoglyceromutase was in the range of 1:0.6-1.2:0.6-1.2 when these activities were expressed in terms of micromoles of substrate utilized per min per g of wet tissue (20).
We have preliminarily investigated this situation in a common bacterial species (E. coli) grown on different carbon sources.
Typical results are shown in Table I. Although there seems to be relatively more phosphoglycerate kinase and phosphoglyceromutase activity in E. coli extracts (ratio of glyceraldehyde phosphate dehydrogenase to phosphoglycerate kinase to phosphoglyceromutase, 1: 1.7: 5.4) t.han in those studied by Pette et al. (20), the three activities are nevertheless in approximately constant proportion to one another no matter which of the three carbon sources the cells utilized.
(The methods of enzyme assay used here differ somewhat from those employed by Pette et al. (20)) but these differences do not appear to account for the disparity in the glyceraldehyde phosphate dehydrogensse to phosphoglycerate kinase to phosphoglyceromutase ratio.) In other experiments the ratio of activities of these three enzymes did not change significantly during the cycle of growth on a given carbon source.
Cells grown to stationary phase in a glucose-containing medium were chosen as a source of these enzymes for purposes of purification.

PuriJication of Enzymes
All operations were conducted at 0 to 5' unless otherwise indicated; centrifugations were performed at 15,000 x g for 30 min.
Tris-EDTA buffer refers to 20 mM Tris-chloride, pH 8, containing 2 mM EDTA. Procedures are summarized in Table  II.
In all subsequent studies the most purified fractions were employed.
Growth of Bacteria and Preparation of Extract-E. coli (Hfr Hayes, strain K12-3000, from the collection of J. Monod, Pasteur Institute) were grown to stationary phase with vigorous aeration in a glucose-yeast extract medium described earlier (21). Growth and harvesting of cells were carried out by Grain Proc- Abbreviations are defined in Table I  The procedure was carried out in a l-gallon stainless steel blending container with an outer jacket through which polyethylene glycol at -5" was circulated; the temperature of the extract never exceeded 10". The mixture was stirred at slow speed to form a thick suspension and then blended for 20 min at 75% of maximum speed. Tris-EDTA buffer (1.35 liters) was added, and the mixture was stirred at slow speed for an additional 10 min. After the beads had settled for 10 min, the liquid was decanted, and the beads were washed by slow speed stirring with 1 liter of Tris-EDTA buffer. After this suspension had stood overnight (15 to 18 hours), it was centrifuged, and the precipitate was collected and dissolved in 120 ml of Tris-EDTA buffer.
The solution of concentrated protein was clarified by a final centrifugation (Fraction III). Sephudex G-160 Gel Filtration-Fraction III (120 ml) was gently applied to the top of a column (70 cm2 x 120 cm) of Sephadex G-150 gel which had been equilibrated with Tris-EDTA buffer.
The same buffer was then pumped through the column at a rate of 60 ml per hour, and fractions of 35 to 60 ml were collected.
After the void volume had passed through the column, glyceraldehyde phosphate dehydrogenase was eluted as a separate peak followed by phosphoglyceromutase and phos- (The higher ionic strength buffer was necessary to avoid swelling of the gel during loading of the column.) The gel was then washed with 5 gel bed volumes of the same buffer and eluted with a linear gradient of NaCl (0 to 0.2 M) in 50 mM Tris-chloride, pH 8, 2 mm EDTA. Total gradient volume was 1.5 liters, and average flow rate was 40 ml per hour.
The enzyme appeared in the middle of the gradient after a peak of inert protein had been eluted.
Active The same buffer was then pumped upward through the column at a rate of 10 ml per hour.
The enzyme was eluted as a sharp peak closely followed by a small, broad peak of inert protein.
Fractions constituting the first half to two-thirds of the enzyme peak (those which yielded a single band in polyacrylamide gel electrophoresis) were pooled and concentrated BS described in the preceding section. Neutral saturated ammonium sulfate containing 2 m EDTA (0.18 volume) was added to the concentrated protein solution (to make it 0.8 M in ammonium sulfate), and the pursed enzyme was stored at 2" in this medium (Fraction VI-DH, 5 to 10 ml). As shown in the following paper, the enzyme can be crystallized at this point, but no further purification is evident.

FIG. 3. Polyacrylamide gel electrophoreses of Escherichia coli enzyme fractions.
Electrophoreses of 59 to 190 pg of protein were carried out in gels at pH 9 according to the method of Davis (18) G. D'Alessio and J. Josse 4323   TABLE   III   I  I  I  I  I  I  I  I  I  I "Pi-ATP-y-"P exchange catalyzed by E. coli glyceraldehyde phosphate dehydrogenase, phosphoglycerate kinase, and phosphoglyceromutase After incubation for 15 min at 37", the mixture was heated at 100" for 2 min and then chilled to 0". Ice-cold 2 N HCl (0.1 ml) and Norit A (20 mg in 0.2 ml of water) were added. Nucleotide adsorption to the charcoal was complete within 5 min at O", and the Norit was collected and exhaustively washed with cold water on a Millipore microfiber glass prefilter (AP 20). The Norit was dried under an infrared lamp and assayed for radioactivity.
The millimicromoles of ATP-+P formed were calculated from the given specific radioactivity of "Pi. This treatment ignores exchange dilution of the "Pi by unlabeled Pi from ATP-7-P and therefore yields low estimates. 4. Formation of ATP-~-3zP in the exchange reaction catalyzed by Escherichia coli glyceraldehyde phosphate dehydrogenase, phosphoglycerate kinase, and phosphoglyceromutase. The reaction mixture was similar to that described in Table III but was scaled up lo-fold (3-ml volume) and contained only 0.04 mM a*Pr (108 cpm per pmole).
After incubation at 37" for 15 min and then at 100' for 2 min, the mixture was chilled and applied to a column (0.2 cm* X 13 cm) of Dowex l-chloride.
Stepwise elution with solutions as indicated was conducted at 2', and fractions of 2 ml were collected. The peak of ATP-+P was pooled, concentrated by barium precipitation in ethanol, and converted to potassium salt with Dowex 50-potassium (23). This material had spectral characteristics of ATP, and phosphorus analyses indicated that the ratio of adenine to total phosphorus was 1:2X Radioactivity was released from the nucleotide (>97Q/,) by treatment with acid (1 N perchloric acid, 8 min at loo'), E. coli alkaline phosphatase, or snake venom phosphodiesterase (24). When this compound was incubated with yeast hexokinase and glucose (25) and the products were resolved by chromatography on Dowex lchloride as above (26) a Identification of the Norit-adsorbable radioactivity as ATPy-32P is described in Fig. 4. b A similar result (22 mrmoles of ATP-+P formed) was obtained when commercial preparations of rabbit muscle glyceraldehyde phosphate dehydrogenase, yeast phosphoglycerate kinase, and rabbit muscle phosphoglyceromutase were substituted for the respective purified E. coli enzymes.
c There was contaminating phosphoglyceromutase activity in phosphoglycerate kinase Fraction V-K (Fig. 2).
kinase zone (Fig. 2). Fractions containing the respective activities were combined as shown in Fig. 2 and concentrated by pervaporation as described above. The concentrated protein solutions (Fraction V-K containing phosphoglycerate kinase, 6.5 ml, and Fraction V-M containing phosphoglyceromutase, 4.2 ml) were clarified by centrifugation if turbidity was present. Purity and, Xtabiility of Fractions-The final fractions of glyceraldehyde phosphate dehydrogenase and phosphoglyceromutase appeared homogeneous when examined by polyacrylamide gel electrophoresis, but phosphoglycerate kinase Fraction V-K was obviously impure (Fig. 3). Small amounts of phosphoglyceromutase were invariably present as shown in Fig. 2. Attempts to purify phosphoglycerate kinase further (gel filtration through Sephadex G-75 or G-200, chromatography with hydroxyl apatite, DEAE-Sephadex, or DEAE-cellulose) have not yielded homogeneous fractions.
Removal of some of the contaminating proteins and small increases in specific activity have been achieved, but recovery of activity after these procedures has been low, and the resultant fractions were unstable. Inclusion 42 i of sulfhydryl compounds, glycerol, or ammonium sulfate did not improve the stability of these fractions.
Homogeneous glyceraldehyde phosphate dehydrogenase was stable for greater than 6 months, when stored in 0.8 M ammonium sulfate as described (Fraction VI-DH). Fractions V-K and V-M of phosphoglycerate kinase and phosphoglyceromutase, respectively, were also stable for 6 months or more.
Intermediate fractions of glyceraldehyde phosphate dehydrogenase and phosphoglycerate kinase during the purification were of variable stability, especially in column eluates of low protein concentration, and it is recommended that the entire procedure be carried out in sequence. This requires 2 weeks or less if the columns have been prepared in advance.
All fractions of phosphoglyceromutase were quite stable.  (14) ; initial protein concentrations were between 0.4 and 1.6 mg per ml of buffer. Sedimentation velocity experiments were conducted at 5 to 25', and initial protein concentrations varied between 1 and 8 mg per ml. Single and symmetricrtl schlieren boundaries were invariably observed (e.g. see Fig. 5 Table I. * For glyceraldehyde phosphate dehydrogenase partial specific volume (@ = 0.734 cc per g) was calculated from the amino acid composition of the protein given in the following paper (28, 29). Three molecular weight analyses were carried out at 18,140 rpm for 24 hours. The SZO.,,, values reported by Allison and Kaplan (6) were included along with data of the present study in determining the .~:a,~ value. = For phosphoglycerate kinase partial specific volume was assumed to be B = 0.75 cc per g (30). Six molecular weight analyses at two different rotor speeds were carried out (40,410 rpm for 26 hours and 30,170 rpm for 25 hours).

Properties of Purified Enzyms
d For phosphoglyceromutase partial specific volume was assumed to be 0 = 0.74 cc per g (31). Three molecular weight analyses were carried out at 26,210 rpm for 22 hours. was complete dependence upon glycerate-3-P.
Nevertheless, it was desirable to confirm independently the catalytic functions of these proteins.
Stoichiometry studies of the individual reaction steps are difficult because of instability of some of the intermediates, notably glycerate-1 ,3-P*, and because of tediousness in quantitative isolation of such intermediates. Penefsky et al. have shown that azPi can be exchanged into the y position of ATP by means of the concerted action of glyceraldehyde phosphate dehydrogenase and phosphoglycerate kinase (22). In a modification of this procedure, with the use of purified E. coli enzymes, such an exchange can be demonstrated and shown to be dependent upon glycerate-2-P, provided phosphoglyceromutase is also present in the reaction mixture (Table III and Fig. 4). For equations, see Scheme 1.
Molecular Weights and Sedimentation Coeficients of Pur$ed E. coli Enzymes-The high speed technique of sedimentation equilibrium was employed to determine molecular weights of the purified proteins. Logarithmic plots of the interference fringe data versus (radius)2 yielded strictly linear relationships with all samples of both glyceraldehyde phosphate dehydrogenase and phosphoglyceromutase, as expected from the electrophoretic homogeneity of these proteins (Fig. 3). However, with the highest concentrations of phosphoglycerate kinase there was slightly increasing slope of the logarithmic plot toward the base of the cell, a finding indicative of heterogeneity (14). Molecular weights obtained from the slopes of the logarithmic plots are listed in Table IV. The values given for phosphoglycerate kinase were obtained from linear portions of the plots and are estimates of the smallest protein species present.
In view of gel electrophoreses which show that the major protein component in the phosphoglycerate kinase preparation is of smaller size than contaminating minor components (Fig. 3), we conclude that the molecular weight listed in Table IV is a valid size estimate for the main component in Fraction V-K and assume that this is E. coli phosphoglycerate kinase. Sedimentation coefficients of the purified enzymes have also been measured and are recorded in Table IV. The contaminants present in purified phosphoglycerate kinase (Fig. 3) were not apparent in the schlieren patterns of sedimenting protein (Fig. 5). DISCUSSION We hope that the information given here will contribute toward continuing elucidation of biochemical function as it is carried out, in E. coli. For example, with use of various carbon sources it should be possible to devise genetic techniques for isolation of mutant cells with altered glyceraldehyde phosphate dehydrogenase, phosphoglycerate kinase, or phosphoglyceromutase. Such mutants would be of considerable interest in studies of FIG. 5. Schlieren boundaries observed during sedimentation of on a solution of purified phosphoglycerate kinase (8.5 mg of Frac-Escherichia co.5 phosphoglycerate kinase. A valve-type syn-tion V-K per ml). Photographs were taken at 4, 12, and 20 min thetic boundary cell (32) was employed in this experiment in which after attaining a rotor speed of 52,350 rpm. The angle of the solvent (0.1 M NaCl, 0.01 M sodium phosphate, pH 7) was layered schlieren diaphragm was 70", and temperature was 8".
bacterial physiology and metabolic control, and the altered proteins would provide useful analogues for investigation of enzyme structure and mechanism.
Such projects at the present time would be overwhelmingly difficult in complex multicellular organisms but are quite feasible with a bacterial system such as E. coli.
One relatively simple analysis that can be made with the data at hand relates to an interesting hypothesis proposed by Mier and Cotton about the constant proportion group of Embden-Meyerhof pathway enzymes (triose phosphate isomerase to enolase) (19,20,33). These authors noted that the data compiled in earlier studies of Pette et al. (20)) taken together with recorded molecular weight and specific activity values of purified enzymes, could be interpreted in terms of an operon hypothesis. They calculated that approximately equimolar amounts of each of the five enzymes were present in the various cell extracts studied by Pette et al., and proposed that, if these proteins were all members of the same operon (34), a single long strand of polycistronic messenger RNA might be transcribed from the entire operon and subsequently give rise to equimolar amounts of each of the proteins coded by the operon.
A corresponding calculation with the presently available E. coli enzyme data (Tables I, II, and IV) shows that the ratio of molecules of glyceraldehyde phosphate dehydrogenase to phosphoglycerate kinase to phosphoglyceromutase is 1: 2.4:3.6. (This calculation is independent of the method of assay of the respective enzymes so long as the same technique is employed with both the extract and the purified protein.) Therefore, in this bacterial species such an operon hypothesis does not seem to hold unless there are additional extrachromosomal genes for one or more of these three proteins in E. coli strain K12-3000 or disproportionate amounts of protein are synthesized during the translation process. Location of the genes for these three proteins on the E. coli chromosome map (35) would be of much interest.
The measured size of E. coli glyceraldehyde phosphate dehydrogenase is very similar to the reported molecular weights of analogous proteins isolated from rabbit muscle (36, 37)) pig muscle (36)) and yeast (37)) and the sedimentation coefficient, in agreement with previous measurements by Allison and Kaplan (6)) is indistinguishable from those of corresponding proteins from a wide variety of sources (6).
There is only one recorded value of phosphoglycerate kinase molecular weight with which to compare that of the E. coli enzyme; the size of the yeast protein is reported as 34,000 daltons (30). Finally, the molecular weight of E. coli phosphoglyceromutase is near the values reported for the enzymes from rabbit skeletal muscle (64,000 (31)) 57,000 (38)) and from chicken breast muscle (66,000 (39)) but is about half that found for the yeast enzyme (112,000 (31)). AcZxwwledgment-We thank Mrs. Ann Hoyt for invaluable technical assistance.