The composition of the pyruvate dehydrogenase ... - Wiley Online Library

Eur. J. Biochem. 142, 541 - 549 (1984) 0 FEBS 1984

The composition of the pyruvate dehydrogenase complex from Azotobacter vinelandii Does a unifying model exist for the complexes from gram-negative bacteria? Hans J. BOSMA, Arie DE KOK, Adrie t i . WLSTPHAL, and Cees VEEGER Department of Biochemistry, Agricultural University, Wageningen (Received February 8IApril24, 1984) - EJB 84 0139

An improved purification procedure of the pyruvate dehydrogenase complex of Azotobucter vinelundii is described. This procedure minimizes losses of components and results in the isolation of the pure complex with a specific activity of 15 - 19 U/mg and an overall yield of 40 %. The chain ratio of the three components was determined by covalent modification of the lysine residues with trinitrobenzene sulfonic acid, followed by separation of the components on sodium dodecyl sulfate gels. These determinations yielded an average chain ratio of 1.3:3 :0.5 for El :E, :E, respectively. Based on Ez this corresponds with a minimum molecular mass of approximately 216 kDa. Because the molecular mass of the complex has been determined previously to be 800 50 kDa, it is concluded that the complex as isolated from A . vinelundii is based on a tetramer of E2 chains. The complex can be resolved into its individual components, which can be recombined to yield a fully active complex. Titration of E,E, subcomplexes with El resulted in maximum complex activity at an EJE, ratio of 1.5 - 1.6. Similar titrations of EIE, subcomplexes with E, resulted in maximum activity at an E,/E, ratio of 0.45 0.55. From these experiments it is concluded that the complex has maximum activity with a composition of three El dimers, one E, tetramer and one E, dimer. With excess of either El or E, a decrease in activity is observed which indicates competition between these components for binding sites on E,. Asshownbefore[Bosma, H. J., deKok,A.,Markwijk,B.W.,andVeeger,C.(1984)Eur. J . Biochem. 140,273-2801, the isolated E, component is composed of 32 peptide chains of 66 kDa each. Upon addition of El or E,, E, dissociates into tetramers. Dissociation is complete upon the addition of four El dimers of four E, dimers per E, tetramer. Addition of El to saturated EzE3 subcomplex or E, to saturated EIE, subcomplex did not result in extra binding but rather in displacement of bound E, or El respectively. It is therefore concluded that the binding sites of El and E, to the E, chains are either identical or so closely spaced that steric hindrance prevents simultaneous binding of both components. A model is presented based on the cubic structure of the isolated E, component. In this model the 32 E2 peptide chains are arranged in tetramers in the corners of the cube. This model is discussed in connection with the existing model for the Escherichia coli complex. Bacterial pyruvate dehydrogenase complexes (PDC’s) are multi-enzyme complexes composed of multiple copies of three enzymes [l]. These enzymes are, in sequence of their reaction, pyruvate dehydrogenase (El), lipoate acetyltransferase (E,), and lipoamide dehydrogenase (E,). The three enzymes catalyse the overall reaction : Pyruvate

+ NAD’ + CoASH-+Acetyl-S-CoA + NADH + H + + CO,.

The El and E, chains are non-covalently bound to a core of E, chains [2]. PDC’s of gram-positive bacteria seem to be based on a core of 60 E, chains with an icosahedral symmetry [3]. It is generally accepted that the PDC of Escherichia coli contains Abbreviations. PDC, pyruvate dehydrogenase complex; (NO,),PhSO,, trinitrobenzene sulfonic acid; PhMeSO,F, phenylmethanesulfonyl fluoride ; SDS, sodium dodecyl sulfate. Enzymes. Pyruvate dehydrogenase (El), pyruvate :lipoate oxidoreductase (EC 1.2.4.1); lipoate acetyltransferase (E,), acetyl-CoA :dihydrolipoate S-acetyltransferase (EC 2.3.1.12); lipoamide dehydrogenase (E,), NADH :lipoamide oxidoreductase (EC 1.8.1.4, previously numbered EC 1.6.4.3).

a core of 24 E, chains, arranged in a cube with 432 symmetry [l]. Estimates of the molecular mass of this complex vary within 3 -6 MDa [2,4 -71. There is a long-lasting discussion on the stoichiometry of the E. coli complex; chain ratios of 2:2:1 [6], 2 : l : l [8] and I : I :I [9] have been proposed. It has been recognized that the amounts of El and E, are not constant in the organism, that preparations of the complex show heterogeneity, and that the chain ratio may change during purification due to losses of these components [5,9 - 131. Nevertheless, the E2 core has an inborn capacity for the binding of El and E, dimers. The discrepancy between the reports of several investigators is therefore probably due to a different interpretation of results as obtained by various methods. The pyruvate dehydrogenase complex as isolated from the gram-negative bacterium Azotobucter vinelandii is much smaller than the E. coli complex [14]. Its sedimentation coefficient is 17 -19 S, that of E. coli PDC is 53 -60 S [2,4,7,10]. However, a 17 -20-S form of the E. coli PDC has been observed [7,10] and we have shown that the A . vinelandii complex can be converted into a 56-S form, resembling the E. coli PDC on electron micrographs [15].

542 We have shown recently [I61 that the 56-S form of the A . Ginelandii PDC is an octamer of a 750 - 850-kDa particle. The tendency to aggregate is especially pronounced in the isolated E, component; its appearance on electron micrographs is similar to that of E, from E. coli. By sedimentation and light scattering studies it wa5 shown that it is composed of 32 peptide chains. Combining these observations it seems likely that the monomeric PDC from A . vinelandii is based on a tetrameric E, ‘core’. From determinations of the amount of reactive lipoyl groups per dimer of E3 we have concluded previously [17] that the A . vinelundii PDC contains four E, chains. This conclusion required certain assumptions and direct chain stoichiometry determinations were lacking. In this paper such direct determinations are described on complexes obtained by an improved isolation procedure. Dissociation and reassembly of the components is studied to determine the ‘optimal’ stoichiometry. From these measurements we conclude that each E2 tetramer has four, probably equivalent, binding sites to which El or E3 dimers can bind in a competitive fashion to yield an ‘optimal’ stoichiometry of three El dimers and one E, dimer. A model is discussed in connection with the model proposed for the E. coli complex by Reed’s group [I].

MATERIALS AND METHODS

Materials Trinitrobenzene sulronic acid was obtained from Eastman Kodak (USA). [2-14C]Pyruvate and N-ethyl[2,3-“C]maleimide were from Amersham International (UK). [3H]Methylacetimidate was synthesized from [3H]acetonitrile (NEN, USA) as described by Bates et al. [IS]. Biochemicals were from Boehringer (FRG); all other chemicals were analytical grade.

Enzyme purification Pyruvate dehydrogenase complex was isolated from Azotobacter uinelandii cells (strain ATCC 478) grown on a large scale [14]. A revised method was used to minimize inactivation and loss of components, since we had indications that the chain stoichiometry may change during purification [ 131. ‘Harsh’ treatments like ammonium sulfate precipitation, isoelectric precipitation or ultracentrifugation at high speeds were avoided. Frozen cells (250g) were thawed in 600ml 50mM potassium phosphate buffer (pH 7.0) containing 1 mM EDTA and 50 pM PhMeS0,F (standard buffer). PhMeS0,F was always added freshly from a stock solution in 100% ethanol. All steps were performed at 4 “C. Cells were broken by passing the suspension once through a Manton-Gaulin laboratory homogenizer at 50 MPa. Unbroken cells and cell debtis were removed by centrifugation (20000 x g. 30 min). A 2 ,/n (w,’v) protamine sulfate solution (pH 7.0) was added dropwise to the cell-free extract until an extra addition (20 ml under the conditions described here) results in a detectable precipitation of PDC (as monitored by pilot experiments). After removal of the protamine-sulfate-complexed material by centrifugation (20000 x g, 20 min), 50 ml of a 2% (w/v) yeast RNA solution in standard buffer was added to precipitate exccss protamine sulfate. The supernatant was made 10% (wlv) in poly(ethy1ene glycol) 6000 by dropwise

addition of a 50% solution in standard buffer. 80 -90% ofthe 2-oxoglutarate dehydrogenase complex activity was precipitated by addition of 1.5 mM MgC1,; over 90% of the PDC activity was precipitated within 4 - 15 mM MgCI, , almost free of 2-oxoglutarate dehydrogenase activity. This PDC pellet was resuspended by gentle stirring overnight in 200 ml potassium phosphate buffer (pH 7.0) containing 3 mM EDTA and 50 pM PhMeS0,F. Undissolved material was removed by centrifugation (60000 x g, twice for 20 min). The supernatant was applied on a 150-ml ethanolamine-Sepharose column as described by Visser et al. [19]. The column was equilibrated with 10 mM potassium phosphate buffer (pH 7.0) containing 1 mM EDTA and 50 pM PhMeS0,F. The column was eluted with a 500-ml0 - 100 mM KCI gradient in this buffer. All PDC-containing fractions were pooled and concentrated by ultracentrifugation at moderate speed (40000 x g , 16 h). The pellet was resuspended by gentle rolling in a small volume of supernatant. The protein solution (7 -15 ml, protein concentration approx. 30 mg/ml) was applied on a column (2.6 x 95 cm) of Bio-Gel A-5.0 m, eluted with standard buffer at 20 ml/h. The fractions that appeared to be pure, as judged from SDS gels, were pooled. After concentration by centrifugation, the preparations were stored in liquid nitrogen at a protein concentration of 20 -40 mg/ml. This method results in a pyruvate dehydrogenase complex from A . uinelandii with a specific activity of 15 -19 pmol NADH produced min-’ mgprotein-’, with anoverall yield of 40%. 80% of the losses in enzyme activity could be accounted for in the form of discarded sidefractions. Met hods

Enzyme activities were assayed at 25’C as described previously [14]. Protein concentrations were measured according to the method of Lowry et al. [20].The flavin content of the complex was determined by the method or Wassink and Mayhew, using FMN standards [21]. The complex was resolved into its individual components by the thiol-Sepharose method of De Graaf-Hess and De Kok [22]. SDS gel electrophoresis was performed according to a modification of the procedure of Laemmli [23], as described by Dorssers et al. [24]. Sedimentation analysis was performed in standard buffer with an MSE Centriscan 75, equipped with an ultraviolet/visible monochromator. Partial specific volumes were calculated from the amino acid compositions of the proteins, according to Cohn and Edsall [25]. Amino acid analysis was carried out on a Jeol-JLC 5AH amino acid analyzer according to standard procedures [26]. Protein samples were hydrolyzed for 24 h and 96 h in 6 M HCI at 110 ”C in sealed evacuated tubes. L-Norleucine (2-aminohexanoic acid) and 1-(a-amino-/I-guanidin0)propionic acid were included as internal standards for the long and the short column respectively ; the former standard was added prior to and the latter immediately after the hydrolysis step. The analytical results with the hydrolysates were averaged with the following exceptions. Threonine, serine and tyrosine contents were derived from extrapolation to zero-time of hydrolysis, and the measures of the 96-h hydrolysates were taken for valine and isoleucine. Tryptophan content was determined fluorimetrically as described in [27]. For the lipoamide dehydrogenase component, determination of cysteine and cystine by an acid ninhydrin method [28] gave essentially the same results.

543 (N0,),PhS03 modification of the complex

The complex (0.2 - 1.5 mgjml) was modified with 3 mM (N0,)3PhS03 in 6 M urea, 1A (w/v) NaHCO, (pH 8.5) for 16 h at room temperature. All steps were performed in the dark to prevent photolysis of (N02)3PhS03and its products. The reaction was stopped by addition of an equal volume of SDS incubation buffer [24], containing 15 mM glycine. The sample was heated for 5 min at 10OoC, followed by electrophoresis on tubular gels ( 5 x 80 mm). On each gel 20 - 80 pg of protein was applied, and each sample was analysed on 4 -8 gels. To remove background absorption due to non-protein (N0J3PhS03products, the gels were washed for several hours in fixation solution. The gels were scanned at 345 nm with a Gilford spectrophotometer, equipped with a linear transporter. The areas of the peaks were determined by cutting the peaks from the graph paper and weighing.

Table 1. Amino acidcomposition of the three components of thepyruvute dehydrogennse complexes of Azotobacter vinelandii (A. v.) and Escherichia coli (E.c.) The amino composition is expressed as a molar percentage of the total number of amino acids. Those of the E. coli enzymes are from Stephens et al. [29-311

Amino acid composition and quaternary structure of the three components The amino acid compositions of the three components of Azotobacter vinelundii PDC have to be known in order to calculate the chain ratios from the results of lysine modification by (NO,),PhSO, or radioactive methylacetimidate [8]. Table 1 lists the amino acid compositions of the components of A . uinelandii PDC : those of the corresponding enzymes from Escherichia coli, as determined by Stephens et al. [29-311 from the elucidation of the DNA sequence, are tabulated for comparison. The compositions compare quite well, especially those of the E, components. The largest deviations are found in the E, components where the difference in the number of small side-chain residues and of proline may indicate differences in the secondary structure of these components. From the amino acid compositions we calculated partial specific volumes of 0.734, 0.749 and 0.750 ml/g for the E l , E, and E3 components respectively, according to the method of Cohn and Edsall [25]. These values were used to calculate the molecular mass of the components from sedimentation equilibrium experiments. The El component has a subunit molecular mass of 94 kDa as estimated by SDS gel electrophoresis. Its mobility is slightly higher than that of the E. coli El component. From sedimentation equilibrium measurements at pH 7.0 we calculated a molecular mass of 1 9 0 f 5 kDa. At this pH the component has a sedimentation coefficient of 9.8 f0.3 S; this decreases to 6.3k0.3 S at p H 10. These observations are in good agreement with the reports on E. coli El [32 -341 and show that the dimer of El dissociates into its monomers at higher pH values. The E, component has a subunit molecular mass of 56 kDa as estimated from SDS gel electrophoresis and its FAD content. Its mobility on SDS gels is slightly lower than the E3 component from E. coli. The E, component has a sedimentation coefficient of 5.4 S. From sedimentation equilibrium a molecular mass of 100 k 4 kDa was estimated, showing E, is a dimer. Comparison of the subunit molecular masses of the E2 components from the E. coli and A . vinelundii complexes showed no significant differences. From SDS gels a molecular mass of 83 kDa can be calculated, whereas sedimentation equilibrium experiments in 5 M guanidinium hydrochloride give a molecular mass of 61 kDa [16]. Because the DNA

-

A.U. ~

E.C.

E2

E3

__~___ A.V. E.C.

A.v.

EL.

~~

mo1/100 mol Asx Thr Ser Glx Pro GlY Ala

4c y s RESULTS

El

Amino acid

Val

10.0 4.1 4.4 12.5 4.5 8.7 8.4 0.5 6.2 3.0 5.1 8.4 4.0 4.3 5.2 2.6 6.9 1.4

10.2 4.7 4.9 12.1 4.0 8.9 8.1 0.7 5.8 2.4 6.3 1.9 4.7 3.8 5.4 2.6 5.6 3.2

5.6 3.3 6.0 11.2 7.5 8.5 19.5 0.6 9.3 1.8 5.1 8.9 0.8 1.6 5.5 1.0 3.4 0.4 ~-

7.9 4.3 4.6 11.6 5.9 8.1 15.3 0.2 10.8 2.5 7.2 5.2 0.5 3.0 8.4 0.8 3.2 0.5 __

7.5

8.1 5.5 2.9 9.5 9.9 3.5 4.4 11.0 10.7 13.9 10.5 0.6 1.0 11.6 9.5 1.6 2.3 5.9 8.2 8.3 7.2 1.5 1.7 2.9 2.9 7.3 8.2 2.2 2.7 2.4 3.2 0.4 ~_0.8 __

94000

99414

66000

65959

56000

50554

sequence indicates a mass of 66 kDa for the E. coli E, component, we have used the same value for the E, component of the A. cinelundii complex. The isolated E, component from the A . vinelandii complex has a sedimentation coefficient of 20.8 f 0 . 5 S. The molecular mass of this component has been determined to be 2.0 f 0.1 MDa by three different methods [16]. From these and other experiments [16] it was concluded that E, is a 32-mer. Chain ratio determination by 1-ysine modqication with trinitrobenzene sulfonic acid Bates and Perham [8] introduced a method for the determination of chain ratios based on the covalent modification of lysine residues in the three components. They used radioactively labelled methylacetimidate as a reagent, and the reaction was performed under denaturing conditions to ensure the modification of all lysine residues. In our hands, however, this method was not reproducible. For one complex preparation for instance, we found 0.92 (&0.09):1:0.88 (f0.08) and 1.91(&0.23):1:1.58(f0.06) for El :E2:E3 in two different experiments. In another experiment, a different PDC preparation with an almost identical FAD content, lipoyl content and specific activity yielded a ratio of 1..52(f0.16):1:0.59(~0.05) (prep. 1, Table 2). The gels sometimes showed appreciable degradation of protein bands, especially of E,. Possibly the E, component of A . vinelandii PDC is very sensitive to proteolysis during the long dialysis step at room temperature that is required in this procedure. We sometimes also observed material on top of the gel. We therefore used trinitrobenzene sulfonic acid as a reagent for the lysine residues. (N0,),PhS03 reacts specifically

544 The (NO,),PhSO, method is relatively cheap and fast. The protein sample is transferred onto the gel directly after modification, thus preventing proteolysis as observed with the radioamidination method, or other possible artefacts. Its reproducibility and accuracy is at least comparable to the latter method.

OOr

Stoichiometry of the isolated A. vinelandii PDC

1

0

I

I

02

04

I

06

I

08

1

10

RF

Fig. 1. Densitogrum qf a tubular SDS gel of’ a recombined pyruaate dehydrogenuse complex. El, E, and E, were mixed in a 1.45:1:0.51 ratio and were modified with (NO,),PhSO,. The chain ratio as estimated from this densitogram is 1.43:l :0.51. The small peaks at RF=0.47 and 0.68 probably originate from proteolytic breakdown of the El component and represent less than 3% of its trinitrophenyl content

with the &-amino group of lysine residues and the aminoterminal of the peptide chain, yielding a trinitrophenylated residue with a strong absorbance at 345 nm [35]. The chain ratios were calculated from scans of gels using the amino acid compositions as given in Table I . The number of reactive amino groups is 44, 34 and 40 for El, E, and E, respectively. Corrections were made for the amino-terminal residue and the two postulated lipoic acid residues (see below) that are covalently linked to E2 through the &-aminogroup of lysine residues. The results from these chain-ratio determinations did not vary significantly when the protein concentration (during modification) was varied within 0.1 -1.5 mg/ml, or whtn reaction time was varied between 30 min (lower than 25 4 modification) and 30 h (over 95% modification). At room temperature, more than 90% of the lysine residues were modified within 8 h. The modification can also be performed at 40 ‘C for 2 h as described by Habeeb [36].However, under these conditions more background due to reaction of (NO,),PhSO, with urea was observed. Omission of urea leads to erroneously low estimates of the E2 component at protein concentrations below 0.5 mg/ml; above 0.75 mg/ml no significant difference was found between modification in 6 M urea, in buffer without any addition and in 5 M guanidiiiium hydrochloride. We therefore routinely used a 16-h modification at room temperature in 6 M urea. It is surprising that the apparent chain ratio is not influenced by the extent of the modification, or by the presence of denaturing agents. It has been shown that the lysine residues do not react at the same rate with (NO,),PhSO, [37]. If there is a difference in reaction rate of the various lysine residues in the A . cinelandii PDC, their relative contribution must be comparable in each of the three components. We omitted 2-mercaptoethanol from our electrophoresis buffers since this compound can reduce the trinitrophenyl group, resulting in a decrease in absorption at 345 nm [38]. As a test, we mixed the isolated three components in a ratio of 1.47(f0.04) :I :0.51 (f0.03) and analysed this mixture with the (N02)3PhS03 method. It resulted in a ratio of 1.45(f0.05):1:0.51(~0.03), indicating that the method is reliable, at least under our conditions. A densitogram of one of the gels of this experiment is shown in Fig. 1.

The chain ratios of some preparations of A. vinelandii PDC are given in Table 2. The preparations were obtained by the isolation method as described in Materials and Methods. For preparation 1, chain ratios as determined by the (N0,),PhS03 method and the radioamidination method are given; it illustrates the underestimation of E, by the radioamidination method as described in the previous section. The (NO,),PhSO,-chain ratios do not vary more than lo”/; between the various complex preparations and they are comparable to those as obtained by Perham’s group for the E. coli complex by the radioamidination method [8]. Lipoyl and FAD contents of some of the complex preparations are also given, since these values have been used for the estimation of the stoichiometry of PDC [6,39,40]. The flavin contents, as calculated from our chain ratios, are within 10% of the measured values. The lipoyl content determined by incorporation of [14C]acetyl groups varied considerably between different preparations; however, we always found values of 10 nmol/mg or higher when fresh samples were used. For some unknown reasons, the amount of acetylatable groups decreases upon ageing, with no apparent effect on the overall enzyme activity. Loss of lipoyl groups without loss of activity has been observed for the E. coli complex [41,42J. For preparation 1 of Table 2, we were able to detect the higher amounts of lipoic acid residues by modification with N ethyl[2,314C]maleimide after reduction with NADH, under conditions where no modification of E, took place, as described in [43]. As a consequence we conclude that the E, chains of the A . vinelandii PDC contain at least two acetylatable groups, probably lipoyl groups. The molecular mass of the complex has been estimated to be 800 50 kDa from laser light-scattering measurements [16]. The complex preparation we used in these experiments had a chain ratio of 1.30:1 :0.47 (Table 2, prep. 5), corresponding to a minimum molecular mass based on E2 of 216 kDa. We therefore conclude that the PDC of A . uinelundii is based on a tetramer of E,. Reconstitution of the complex from isolated enzymes

The isolated components of the A . vinelandii PDC as obtained by the thiol-Sepharose method of De Graaf and De Kok [22] can be recombined to give a fully active complex. Fig. 2A (filled circles) shows the effect of recombination of complex activity with varying amounts of El ; a fixed amount of E3 was added to these E,E, subcomplexes to restore complex activity. Interpolation of the initial increase in activity and the plateau value at saturating amounts of El gives a ratio of 1.59(f0.16):3 for E, :E, (five experiments with different preparations, E, concentrations of 20 - 80 pg/ml, i. e. E,-chain concentration 0.3 - 1.2 pM). This interpolation is justified since inactivation studies with thiamin thiothiazolone diphosphate, a transition state analog of thiamin diphosphate [44], have indicated that the overall enzyme activity is linearly correlated with the amount of active El bound to the complex [45] (and results to be published). The same equivalence point

545 Table 2. Chain ratios of some preparations of pyruvate dehydrogenase complex from Azotobacter vinelandii Standard deviations are given between brackets. n.d. means not determined Preparation

Chain ratio measured

FAD content

Lipoyl content

~ _ _ _ _ _ _

El

E2

E,

~

measured

calculated

measured

~

calculated

~

mol/mol Ez

nmol/mg 1.52( j,0.16) :1 :0.59( +O.OS)a 1.43(f0.08) : l :0.50(&0.05)' 1.37(f0.08) :1 :0.58(:t0.05)b 1.27(f0.04) :1 :0.48($0.02)' 1.27(+0.04):1:0.51(+0.03)b 1.30(i 0.12) :1 :0.47( t 0.05)

1. I. 2. 3. 4. 5. a

2.0($_0.1)

2.5(+0.3) 2.2( i0.2) 2.6(10.3) 2.3(+0.1) 2.4(k0.2) 2.2( f0.2)

2.3( kO.1) 2.4( 10.1) 2.2(,0.1) 2.5(+0.1)

7.2(k0.2)' 9.8( 0.3)d n.d. 1 1.7(10.3)" n.d. 8.4( 0.)'

1.7"-2.4d(f0.2) 1.6' -2.2d( F0.1) 2 . q k0.2)

1.8(k0.2)

As determined by the radioamidination method of Bates et al. [8].

As determined by the (NO,),PhSO, method. As determined from incorporation of [2-'4C]pyruvate [17]. As determined from modification with N-ethyl[2,314C]maleimide in the presence of NADH [43]

LA-0.5

EI/EZ r a t l o lmol /mol 1

1

E+Ez

1.5

2

2.5

3

r a t i o lmol /mol 1

Fig. 2. Reconstitution of Azotobacter vinelandii PDCJrom its individua components. (A) Titration of E, with E, : varying amounts of ?, were added to a fixed amount of E,. After 20 rnin of incubation at room temperature, a fixed amount of E, was added to restore overall enzyme activity. After 20 min further incubation, 10 -25 p1 of the incubation mixture was assayed for complex activity (0, x ) or recombination was performed in the cuvette containing all the components of the assay mixture and the reaction was started by the addition ofpyruvate (0) Enzyme . activity is expressed as pmol NADH produced min-I. Enzyme preparations were in 50 mM potassium phosphate pH 7.0. ( 0 ) 1.17 pM E,, 1.05pM E3; ( x ) 1.17 pM E,, 4.15 pM E,; (0) 13.8 nM E,, 12 nM E,. (B) TitrationofE, with E,: sameconditions asin A. 10-25 pl ofthe incubation mixture was assayed for complex activity in a volume of 1 ml (O),or recombination was performed in the cuvette and the reaction was x). (0) 1.17 pM Ez,3.42 pM El ; (0)49 nM E,, 172 nM E, ; ( x ) 13.8 nM E,, 217 nM El. All started by the addition of pyruvate (0, Concentrations are expressed in chain molarities

was found when smaller amounts of E, were added, only the plateau value and the slope of the initial curve decreased (not shown). Fig. 2B (filled circles) shows reconstitution of the complex activity with varying amounts of E,. With the same range of E, concentrations as used above in the titration with E l , the titration curve is linear only at low E,/E2 ratios and no plateau value is reached. When the reconstitution is performed at E, concentrations of 10 -50 nm (open circles), the initial increase in overall activity is almost linear until a n optimal value is reached, whereafter a decrease in activity is observed. Extrapolation of both parts of the titration curve yields an E3/E2 chain ratio of 0.5( 0.05) (five experiments with different preparations). From these data it is concluded that in the optimal stoichiometry one E2 tetramer binds one E, dimer and three El dimers. When the El titration is performed at E, concentrations of 10 -50 n M (open circles in Fig. 2A) no sharp equivalence point is observed. It is concluded that the affinity of El for E2 is considerably lower than the affinity of E3 . Assuming three El binding sites, a Scatchard plot could be constructed using points that deviate from the stoickaiometric titration curve. From this linear plot a dissociation constant of 0.5 n M for the

binding of El to E, could be calculated. Due to the limited range over which the binding could be studied no conclusion could be drawn about possible interactions between binding sites. From the shape of the titration curves it is clear that in the presence of excess of either component competition is observed. This competition is particularly strong between El and E3 for El binding sites. Such a mutual binding inhibition has also been observed for the E. coli El and E, components [46]. From the El titration curve in the presence of excess E, (crosses in Fig. 2A) or the decrease in activity in the presence of excess E, in the E, titration curve (open and closed circles of Fig. 2B) it can be concluded that the affinities of El and E3 for El binding sites are of a comparable magnitude. Much less competition is observed between El and E3 for the E3 binding site, as can be concluded from the slight decrease in activity in the El titration curve (closed circles of Fig. 2A) with excess E l . A titration curve similar to the E, titration in the presence of excess E3 is obtained at low E, concentrations in the presence of a large excess of El (crosses in Fig. 2B). From the degree of inhibition and the concentration ratio of the free peripheral components, calculated for each point of this curve, it can be calculated that the affinity of E, for the E, binding site is two orders of magnitude higher than

546

0.1

0.8

E,/E,

1.2

1.6

2.0

(mol /mol 1

0.4

0.8 E,/E,

1

I

1.2

1.6

-0

2.0

2.1

l m o l /mol j

Fig. 3. Sedimentationunalysis of'E, E2 and E2E3subcomplexes of Azotobacter vinelandiipyruvate dehydrogenasecomplex. Varying amounts of the peripheral components were added to a fixed amount of Ez in standard buffer. After 30 min of incubation at 20 "C, sedimentation runs were started. (A) Titration of Ez with E l . The concentration of E, varied over 0.34 -0.61 mg/ml. Scans were performed at 280 -290 nm depending on the absorbance; ( 0 )percentage of slowly sedimenting material; (0) percentage free E, ; both expressed as a percentage of total ultraviolet absorbance. (B) Titration of Ez with E,. The concentration of E, varied over 0.45 -0.61 mg/ml. ( 0 )Percentage of slowly sedimentingmaterial as calculated from the scans of Fig. 4 at 280 nm; (m) percentage of slowly sedimenting material from scans at 456 nm; (0)percentage of free E,, expressed as a percentage of total 280-nm absorbance; (0) percentage of free E,, expressed as a percentage of total 456-nm absorbance. Other conditions: see Fig. 4

the affinity of El for this site. No absolute affinities could be estimated and it may well be that the affinity of El for this site is lower than for the other three sites. Thus it may be concluded that the E, tetramer contains four binding sites for the peripheral components. The reciprocal binding-inhibition relation leads to an optimal stoichiometry of three El dimers and one E, dimer. In these experiments the binding of one component is studied in the presence of a saturated amount of the other component. Thus the observed differences in affinity may result from mutual interaction between the peripheral components and need not reflect the presence of intrinsically different binding sites on the E, tetramer. In the recombination experiments, the resolution of the complex was performed under mild conditions to prevent inactivation of components and it therefore did not result in total removal of the E, component from E, , as can be seen in Fig. 2B. The El and E, preparations were essentially pure as judged from enzyme activities and SDS gels. The order of addition of the components did not affect the results. The maximal overall enzyme activity observed was 62 pmol NADH produced min-' mg ET1. This corresponds to about 17 Ulmg PDC for the reconstituted complex, a n activity slightly higher than that of the complex preparation used for the isolation of the components.

Sedimentation studies on El E, und E2E3 subcomplexes As described in a previous paper [16], the addition of El to E, causes dissociation of the large E, core. Upon centrifugation, two boundaries are observed, one based on a tetrameric E2 core and one based on a 32-mer of E,. Both species are in slow equilibrium and the separation of boundaries is only observed at high speed (45000 rpm). Addition of El causes a shift in the equilibrium towards the smaller species and at full saturation only one boundary of the E,E2 subcomplex based on the E2 tetramer is observed. Fig. 3A shows a titration curve in which the percentage of the smaller species (szo,u= 10 - 18 S) is plotted against the EI;E2 ratio. It is clear that full saturation is reached at a molar ratio close to 2. i. e. when one dimer of El is bound per E, chain. At higher ratio the boundary of unbound El is observed. The isolated pure E, contains a small amount of slowly sediment-

ing material (Fig. 4A). Whether this material is inert or reconstitutionally active could n t t be decided upon. Both cases would result in a n estimated 10/0 correction in the observed ratio. Fig. 4 B - F show sedimentation velocity scans at increasing ratios of E,/E,. The general picture is similar to the effect of El on E, [I 61. Two boundaries are observed, a slowly sedimenting boundary with s20,w increasing from 7 S to 12.5 S and a rapidly sedimenting boundary with szo,wincreasing from 21 S to about 30 S. Above a ratio of about 1.9 the boundary of free E, is detectable. In Fig. 3B the titration curve is plotted. The equivalence point is, within experimental error, identical with that obtained with E l . It is concluded that each E, chain is also capable of binding one E, dimer. At an E,/E, ratio of about 0.6 a break in the titration curve is observed. Apparently the first dimers of E3 that are bound to the large E, core haye a n appreciably smaller effect on the dissociation than further additions. This is in contrast to the binding of E l , which results in comparable effects on the dissociation at all ratios. With E3 it is possible to study the distribution of this component between the two species by performing scans a t 456 nm, the absorption maximum of the FAD. Because of the relatively low absorbance such scans were not possible below an E3/Ez ratio of 1.0. From these scans it appears that the slowly sedimenting material contains somewhat less E, than the rapidly sedimenting material. The effect is small however, and its significance is questionable. To test whether the binding of El and E, is competitive or additive, El was added to a saturated E2E3 subcomplex in a ratio El/E2=0.5. This resulted in the appearance of the boundary of the free components, most of which was E, as judged from the 456-nm absorbance (not shown). Similarly, addition of E, to a saturated E,E, subcomplex resulted in displacement of E l . From these experiments it is concluded that the binding of E, and El on thc four E, binding sites leads to mutual exclusion.

DISCUSSION From the data presented in this paper we conclude that the pyruvate dehydrogenase complex from Azotobucter ziinelundii is based on a tetramer of E, . The isolated components can be

547

A

n

B

! Y

Fig. 4. Sedimentation ve1ocit.v ultraviolet tzhsorption putterns o j the isolated E2 component of Azotobacter vinelandii pyruaate dehydrogenuse complex and ofsome E2E3suhcomplexes of varying composition. The E,/E, molar ratios are given in the figure. The concentration of E, varied within 0.45 -0.61 mglml in standard buffer. Temperature 20 "C; 3-min intervals between scans; rotor speed 45000 rpm

recombined to yield a fully active complex and from these reconstitution experiments we deduce a catalytically optimal EJE, chain ratio of 1.5 - 1.6. Similar estimations of the E3/E, chain ratio yield values close to 0.5. Since the pyruvate dehydrogenase complex from A . vinelandii is much smaller than the complex from Escherichiu coli, interpretation of the experimental data into integer stoichiometries is easier. Based on the reconstitution experiments, we conclude that in the optimal stoichiometry one dimer of E3 and three dimers of El are bound to the E, tetramer. The sedimentation experiments show a binding capacity of one Elor E, dimer per E, chain. Thus fully assembled complex based on an E, tetramer would contain a total of four El and E, dimers. Due to the high turnover number of the E3 component, maximal overall enzyme activity is expected for a complex particle containing three El dimers and one E3 dimer, as experimentally confirmed by the reconstitution experiments. In the case of a trimeric E, core, maximal activity would be expected upon binding of two El dimers and one E3 dimer. Such a composition is in conflict with the EI:E3ratios as obtained from the chain-ratio determinations and reconstitution experiments. The predicted molecular mass and flavin content of such a complex would also be in conflict with experimentally determined values. Therefore a trimeric core, considered by some groups as the morphological subunit of the E. colicomplex [l,71, seems highly unlikely for the A . vinelandii complex. The average chain ratio of the complex preparations isolated by the revised purification method is 1.3:1:0.5 for El :E, :E,. Addition of El causes a 20% increase of the overall PDC activity, whereas we never detected a measurable increase in activity upon addition of extra E,. These observations strongly support our 6 :4 :2 model and it also indicates that the

revised isolation procedure yields PDC preparations that are almost maximally active. This is also reflected in the very s?all loss in complex activity during its purification (less than 20 A). The question how a 3:l ratio of E, :E, is obtained on the basis of affinities is indicated by the competition experiments. Apparently the binding of one dimer of E3 in the presence of three dimers of bound El is preferred over the binding of a fourth El dimer. This could be due to the presence of a single high-affinity binding site for the E3 dimer on the E, core but it is also possible that the binding sites for the El and E, dimers on the E, tetramer are identical and the binding of three El dimers leads to increased affinity of the unoccupied site for an E, dimer. Because we can only deduce relative affinities nothing can be said at present about cooperativity in the binding of the El or E, dimers. With a tetrameric E, two quaternary structures need to be considered : square planar and tetrahedral. For the arrangement of the peripheral components the following considerations are important : (a) the mutually exclusive binding of four El or E, dimers; (b) the unique position of the E, dimer; (c) symmetry requirements and (d) the association behaviour. The square planar model can be excluded on the basis of these considerations. For the tetrahedron two arrangements seem possible. These are pictured in Fig. 5 as part of the cubic structure of the isolated E, component as observed in electron micrographs [16]. The positions of El and E, are considered topologically equivalent in the E, tetramer. Thus, there are only four binding sites to consider to which mutually exclusive binding of El and E3 can occur. The positions of El and E, in the cube are no longer equivalent. E, is positioned in the corner of the cube and the El dimers are either located on the edges or in the planes. It is clear that the position of the E, and E, dimers on a trigonal axis presents symmetry problems. In the model of

548

n

Fig. 5. TM'O arrangements of' the subunits of the E2 tetramer with its peripheral components that ,fit into the cubic structure of the isolated E2 component and aggregated,form o f t h e complex of A. vinelandii. E, is pictured as a monomer (shaded): El and E, are pictured as dimers

Fig. 5A three trigonally symmetric E,-binding domains on the E, dimer are required which seems very unlikely. For this reason we prefer the model in Fig. 5B in which each component is supposed to interact with one E, chain through a single, probably twofold symmetric, binding domain. The E, component might show (pseudo) threefold symmetry [30]. We have demonstrated before [16] that the intact PDC from A . tlinelundii participates in a rapid monomer-dimer equilibrium and that under forced conditions, such as the presence of 3 % poly(ethy1ene glycol) and I0 mM MgC1, , the equilibrium shifts to a well-defined octameric aggregation state, called the 56-S form of the complex, without loss of components. Thus, this form is based on a core of 32 E, chains. In agreement with this observation on the intact complex we have shown that the isolated E, has a molecular mass of 2.0 0.1 MDa and is therefore probably composed of 32 chains of 66 kDa. The appearance of E, in electron micrographs is that of a cubic structure, similar to that of E, from E. coli. The organization of the 56-S form of the complex is unclear. Electron micrographs show regular but cloudy appearances which lack structural details. Assuming that the extrapolation of our model for the monomeric PDC to the cubic structure of the E, component is allowed, what would be the features of such an organization? The single E, dimer is positioned on the trigonal axis on the outside of the cube. In this unique position the E3 dimer is far removed from the interfaces between E, tetramers and it is expected that the binding of E3 to this position has no influence on the dissociation. Such geometry could provide the explanation for the shape of the titration curve of Fig. 3B. The El component is close to the interface between E, tetramers, so that dissociation is expected to start upon the first additions of E l . In our model, the two active sites of E, are asymmetrically positioned with respect to the lipoyl groups. The high mobility of the lipoyl groups [47] and of part of the peptide chain of E2 [4S] may cause an averaging in time. We have previously observed that E, itself is highly mobile in the intact complex [13]. The unique position together with this mobility and its high turnover number makes E3 very suitable to function as a single output site for reducing equivalents. This brings us to the question of the number of lipoyl groups. In freshly isolated complex we have repeatedly found values between 2 and 2.5 per E, chain. Assuming that this deviation from an integer value is significant, we have to consider the possibility of a nonequal distribution of lipoyl groups. From the models pictured in Fig. 5 it is clear that the E, subunits are topologically different with respect to input of

acetyl groups (El) or output of reducing equivalents (E,). The E, subunit on the trigonal cubic axis may be assigned to fulfil a special function, e. g. the transfer of reducing equivalents from other E, subunits to E,. In this trigonal symmetric orientation it could contain three lipoyl groups. Stephens et al. [30] concluded from the DNA sequence that the E2 subunit from E. coli contains three homologous lipoyl binding sequences. The main feature of our model is that it is composed of morphological subunits. The small form of the complex is supposed to associate to the octameric form without reorienta tion or redistribution of the component enzymes. The association equilibria in poly(ethylene glycol) and those of partially reconstituted (sub)complexes are, however, slow processes with relaxation times in the order of 10 -30 min. This indicates a high energy of activation which could be caused by changes in the quaternary structure. Therefore the extrapolation from the small tetramer form to the cubic structure of E, and vice versa may not be allowed. This touches to the more general problem :why would a tetramer in the form of a regular tetrahedron associate to a cubic structure? This seems only possible when the symmetry is somehow distorted upon association, resulting in a cubical structure composed of two classes of E, chains: 24 of these chains would have identical environments different from the 8 chains in the corners of the cube (Fig. 5B). This arrangement is therefore less symmetrical than the 432 structure proposed for the E, core of the E. coli PDC. An important difference between the model proposed here and the model for the E. coli complex is that the latter is based on a single large entity which does not take into account dissociation into fully assembled smaller particles. However, partial dissociation of the E. coli complex into active 1 9 3 particles has been observed with conservation of the chain stoichiometry [7, lo]. In the 24 :24:12 model proposed by Reed [I, 61 dissociation into eight corner structures, based on an E, trimer will lead to asymmetric and incompletely assembled particles. Another possibility is the fragmentation of the E2 core into planar tetramers to which two El dimers and one dimer of E, are bound. Such a fragmentation also seems quite unlikely. Another problem presents the optimum number of El dimers which can be bound to the complex from E. coli. Direct determinations of chain stoichiometry of isolated complexes and reconstitution experiments of El with E2E3 subcomplexes performed by the group of Perham [8,33,49] are very close to the values reported here for the A . cinelandii complex, when they are corrected for the molecular mass of 66 kDa of E, . A value of 1.6 for the optimum El :E, ratio would indicate a binding capacity for about 18 El dimers on

549 the E2 core (in the presence of an optimum amount of E, on the core). This is clearly incompatible with the 24 model. Thus neither model can explain all observations in a satisfactory way. Because of the many similarities which exist between the complexes from both sources it seems attractive to propose a unifying model in which the A . vinelundii complex may represent the morphological subunit of the larger structure present in E. coli and perhaps other gram-negative bacteria. On the other hand, the unique association behaviour of the A . vinelundii PDC could be caused by a unique quaternary structure. It could represent a step in the development of the still larger assemblies found in gram-positive bacteria, apparently a try-out that failed in this case. The question whether association into these large assemblies in prokaryotic and eukaryotic systems only has a structural function or may also lead to an improvement in catalytical power remains as yet unanswered. Possibly the larger structures are more efficient under physiological conditions or can be more finely tuned by their regulators. The A . vinelundii complex offers an interesting object to study these questions. We are indebted to Dr S. Visser (Thr Netherlands Institute for Dairy Research, Ede, The Netherlands) for the amino acid analyses, to Mr M. M. Bouwmans for drawing the figures, and to Miss C. M. Verstege for typing the manuscript. This investigation was supported by the Netherlands Foundation for Chemical Research (SON) with financial aid from the Netherlands Organization for the Advancement of Pure Research (ZWO).

REFERENCES 1. Reed, L. J. & Oliver, R. M. (1968) Brookhaaen Symp. Bid. 21, 397 -411. 2. Koike, M., Reed, L. J. &Carroll, W. R. (1963) J . Biol. Chem. 238, 30 -39. 3. Henderson, C. E., Perham, R. N. & Finch, J. T. (1979) Cell 17, 85 -93. 4. Dennert, G. & Hoglund, S. (1970) Eui.. J. Biochem. 12, 502 -507. 5. Vogel, O., Hoehn, B. & Henning, U. (1972) Eur. J. Biochem. 30, 354-360. 6. Eley, M. H., Namihira, G., Hamilton, L., Munk, P. & Reed, L. J. (1972) Arch. Biochem. Biophys. 152, 655 -669. 7. Danson, M. J., Hale, G., Johnson, P., Perham, R. N., Smith, J. & Spragg, P. (1979) J . Mol. Biol. 129, 603-617. 8. Bates, D. L., Harrison, R. A. & Perham, R. N. (1975) FEBS Lett. 60, 427 -430. 9. Vogel, O., Hoehn, B. & Henning, U. (1972) Proc. Natl Acad. Sci. USA 69, 1615-1619. 10. Schmitt, B. (1976) Biochimie (Paris) 58, 1405 - 1407. 11. Schmitt, B. & Cohen, R. (1980) Biochem. Biophys. Res. Commun. 93, 709 -712. 12. Gilbert, G. A. & Gilbert, L. M. (1980) J . Mol. Biol. 144, 404 -408. 13. Bosma, H. J., de Graaf-Hess, A. C., de Kok, A,, Veeger, C., Visser, A. J. W. G . & Voordouw, G. (1982) Ann. New York Acad. Sci. 378, 265 -285. 14. Bresters, T. W., de Abreu, R. A,, de Kok, A,, Visser, J. & Veeger, C. (1975) Eur. J . Biochem. 59, 335 -345. 15. Bosma. H. J., Voordouw, G., de Kok, A. & Veeger, C. (1980) FEBS Lett. 120, 179 - 182.

16. Bosrna, H. J., de Kok, A,, Markwijk, B. W. & Veeger, C. (1984) Eur. J. Biochem. 140, 273 -280. 17. De Abreu, R. A,, de Kok, A,, de Graaf-Hess, A. C. & Veeger, C. (1977) Eur. J. Biochem. 81, 357-364. 18. Bates, D. L., Coggins, J. R. & Perham, R. N. (1975) Anal. Biochem. 68, 175 - 184. 19. Visser, J., van Dongen, W. & Strating, M. (1978) FEBS Lett. 85, 81 -85. 20. Lowry, 0. H., Rosebrough, N. J., Farr, A. L. & Randall, R. J. (1951) J . Biol. Chem. 193, 265 -275. 21. Wassink, J. H. & Mayhew, S. G. (1975) Anal. Biochem. 68, 609 -616. 22. De Graaf-Hess, A. C. & de Kok, A. (1982) FEBS Lett. 143, 261 -264. 23. Laemmli, U. K. (1970) Nature (Lond.) 227, 680 -685. 24. Dorssers, L., Zabel, P., vander Meer, J. &van Kammen, A. (1982) Virology 116, 236 -249. 25. Cohn, E. J. & Edsall, J. T. (1943) Proteins, Amino Acids and Peptides, pp. 370 -381, Rheinhold, New York. 26. Moore, S. & Stein, W. H. (1963) Methods Enzymol. 6, 819-831. 27. Pajot, P. (1976) Eur. J . Biochem. 63, 263-269. 28. De Koning, P. J. &van Rooijen, P. J. (1971) Milchwissenschaft 26, 1-6. 29. Stephens, P. E., Darlison, H. G., Lewis, H. M. & Guest, J. R. (1983) Eur. J . Biochem. 133, 155-162. 30. Stephens, P. E., Darlison, M. G., Lewis, H. M. & Guest, J. R. (1983) Eur. J . Biochem. 133, 481 -489. 31. Stephens, P. E., Lewis, H. M., Darlison, M. G. Guest, J. R. (1983) Eur. J. Biochem. 135, 519-527. 32. Dennert, G . & Eaker, D. (1970) FEBS Lett. 6, 257 -261. 33. Perham, R. N. &Hooper, E. A. (1977) FEBS Lett. 73,137-140. 34. Vogel, 0. (1977) Biochem. Biophys. Res. Commun. 74, 1235-1241. 35. Okuyama, T. & Satake, K . (1960) J . Biochem. (Tokyo) 47, 454 -466. 36. Habeeb, A. F. S. A. (1966) Anal. Biochem. 14, 328-336. 37. Goldfarb, A. R. (1966) Biochemistry 5, 2574-2578. 38. Means, G. E. & Feeney, R. E. (1971) Chemical Modijication of Proteins, Holden-Day, San Francisco. 39. Collins, J. H. & Reed, L. J. (1977) Proc. Nut1 Acad. Sci. USA 74, 4223 -4227. 40. Speckhard, D. C. & Frey, P. A. (1975) Biochem. Biophys. Rex Commun. 62, 614 -620. 41. Stepp, L. R., Bleile, D. H., McRorie, D. K., Pettit, F. H. & Reed, L. J. (1981) Biochemistry 20. 4555 -4560. 42. Berman, J. N., Chen Guan-Xiong, Hale, G. & Perham, R. N. (3981) Biochem. J. 199, 513-520. 43. De Kok, A., Visser, A. J. W. G. & de Graaf-Hess, A. C. (1982) in Flaains and Fluaoproteins (Massey, V. & Williams, C. H., eds) DD. 61 -65. Elsevier, Amsterdam. 44. &towski, J. A. & Lienhard, G. E. (1976) J . Biol. Chem. 251, 2863 -2866. 45. Stanley, C. J., Packman, L. C., Danson, M. J., Henderson, C. E. & Perham, R. N. (1981) Biochem. J . 195, 715-721. 46. Reed, L. J., Pettit, F. H., Eley, M. H., Hamilton, L., Collins, J. H. & Oliver, R. M. (1975) Proc. Nut1 Acad. Sci. USA 72, 3068 -3072. 47. Grande, H. J., van Telgen, H. J. & Veeger, C. (1976) Eur. J. Biochem. 71, 509 -518. 48. Wawrzynczak, E. J., Perham, R. N. & Roberts, G. C. K . (1981) FEBS Lett. 131, 151 -154. 49. Bates, D. L., Danson, M. J., Hale, G., Hooper, E. A. & Perham, R. N. (1977) Nature (Lond.) 268, 313-316.

H. J. Bosma, Naarden International B.V., Huizerstraatweg 28, NL-1411-GP Naarden, The Netherlands A. H. Westphal, A. de Kok, and C. Veeger, Laboratorium voor Biochemie der Landbouwhogeschool, De Dreijen 11, NL-6703-BC Wageningen, The Netherlands