ANDERSON AND VALLEE
Selective Cobalt Oxidation as a Means to Differentiate Metal-Binding Sites of Cobalt Alkaline Phosphatase? Richard A. Anderson1 and Bert L. Vallee*
ABSTRACT: The conversion of exchange-labile Co(I1) to exchange-inert Co(II1) was examined in Escherichia coli alkaline phosphatase to differentiate its catalytic from its structural and regulatory metal-binding sites. Oxidation with hydrogen peroxide of phosphatase, containing 2 g-atoms of cobalt, results in the specific oxidation only of the cobalt moieties but does not affect the activity of the corresponding zinc enzyme, the return of activity of the apoenzyme subsequent to the addition of zinc, or the amino acid composition of the oxidized cobalt enzyme. Apoalkaline phosphatase reconstituted with 6 g-atoms of cobalt is inactivated more rapidly than that substituted with 4 g-atoms, and this, in turn, more rapidly than that to which 2 g-atoms of cobalt(I1) has been added. Oxidation of phosphatases containing 2 to 6 g-atoms of cobalt results in oxidized enzymes with characteristic qualitatively similar absorption, natural circular dichroic (CD) and magnetic circular dichroic (MCD) spectra but which differ in intensities. The present data provide further evidence for the formation of Co(II1) phosphatase (Anderson, R. A., and Vallee, B. L. (1979, Proc. Natl. Acad. Sci. U.S.A. 72, 394-397) and demonstrate that, al-
though the enzyme binds 6 g-atoms of cobalt, only 2 g-atoms per mol of enzyme is susceptible to oxidation. Zinc, cobalt, and magnesium metal hybrid enzymes were employed to ascertain the sites of the oxidized cobalt ions. Cobalt bound at the magnesium or regulatory sites is not susceptible to oxidation, and only 2 g-atoms of cobalt bound at the remaining four sites can be converted to cobalt(II1). Further, oxidation of only 1 g-atom of cobalt per dimer of phosphatase prevents the return of activity observed upon the subsequent addition of zinc. Oxidation affects neither the amount of cobalt bound nor the binding of zinc at the remaining zinc-binding sites, but it does influence the number of cobalt ions that can be displaced by zinc. Dialysis of the oxidized enzymes containing 2 or 4 gatoms of cobalt, under conditions normally employed to generate the native apoenzyme, removes all but 0.4 g-atom of cobalt. Subsequent to the addition of zinc there is a return of 60% of the activity of these enzymes when compared with the control enzymes treated similarly, but not exposed to peroxide.
Substitution of the metal ions of metalloenzymes is one of the mildest and most specific procedures presently available for the chemical modification of enzymes. Generally, such substitutions do not alter protein structure significantly and often maintain or even enhance catalytic function (Vallee and Wacker, 1970). Zinc, a constituent of many metalloenzymes, is diamagnetic ( d l 0 ) and colorless, and therefore unsuitable for the study of its presumed environment by spectroscopic techniques. Hence, it is often replaced by chromophoric and paramagnetic metal ions. Co(I1) is particularly useful in this regard, since the ionic radii of cobalt and zinc are similar, and these ions share the capacity to accept distorted geometries (Vallee and Williams, 1968a,b). In alkaline phosphatase, moreover, Co(I1) is the only metal which substitutes for zinc to yield significant enzymatic activity when measured by steady-state kinetics (Plocke and Vallee, 1962). However, zinc, cobalt, and all other metal ions used to replace them in alkaline phosphatase are exchange labile (Taube, 1952), thus generating problems in localizing and delineating their interactions with the catalytic, structural, and regulatory metal-binding sites. Therefore, it is of particular interest that Co(I1) alkaline phosphatase can be oxidized to exchange inert Co(II1) phosphatase (Anderson and Vallee, 1975), thereby providing a potential means to render these
metal atoms relatively immobile at their respective binding sites. Oxidation abolishes catalytic activity, decreases the intensity of the EPR' signal, and induces characteristic changes in the absorption, CD, and MCD spectra, indicative of the formation of Co(II1). Oxidation does not alter the amount of bound cobalt, but it does affect the number of cobalt ions which are exchange labile.
~
From the Biophysics Research Laboratory, Department of Biological Chemistry, Harvard Medical School and the Division of Medical Biology, Peter Bent Brigham Hospital, Boston, Massachusetts 021 15. Receioed April 29, 1977. This work was supported by Grant-in-Aid GM-15003 from the National Institutes of Health of the Department of Health, Education, and Welfare. Present address: Agricultural Research Service, United States Department of Agriculture, Beltsville, Md. 20705.
*
4388
BIOCHEMISTRY, VOL.
16, NO. 20, 1977
Materials and Methods Alkaline phosphatase was released from the periplasmic space of E. coli cells treated with EDTA (Neu and Heppel, 1965) or Chelex 100 (Bosron et a]., 1975) and purified to homogeneity by methods which prevent both the loss of metals and contamination with extraneous metal ions (Simpson et al., 1968; Bosron et al., 1975). Phosphatase isolated using either Chelex 100 or EDTA was identical under all conditions examined (Bosron et al., 1975; Anderson et al., 1975; Bosron et al., 1977). Enzymatic activity and protein concentration were measured as described (Anderson et al., 1976). Enzymatic activity values measured in 1 M Tris exhibit both hydrolase and transferase activities (Wilson et al., 1964), while those in 0.4 M NaCI-20 mM Verona1 show hydrolase activities only. Peroxide affected both activities equally, and therefore only one activity is presented. Buffers, glassware, and substrate were freed from contaminating metals as described (Thiers, 1957). After extraction with dithizone, buffers were stored in the Abbreviations used are: EPR, electron paramagnetic resonance; CD, circular dichroic; MCD, magnetic circular dichroic; Tris, 2-amino-2hydroxymethyl-1,3-propanediol;EDTA, (ethy1enedinitrilo)tetraacetic acid.
METAL-BINDING SITES OF PHOSPHATASE
TABLE I: Restoration of Activity of Co(I1) and Co(II1)
“Apoenzymes”.a Restored sp act.b Tris (1) (2) (3) (4)
2-Co and 4 - c o enzymes (control) 2 CO H202 4 CO H202 4 c o Dhowhated f H707
+ + +
52 31 30 49
Cobalt Verona1 remainineC 26 15 14 25
0 0.4 0.4 0
Cobalt alkaline phosphatase, 0.224 mM, was incubated with 8 m M H202, for 45 min. At that time, the activity of the 2- and 4-co enzymes was zero but that of the 4-co enzyme plus phosphate was 1.4 units. These enzymes were then dialyzed for 5 days at room temperature against 10 m M hydroxyquinoline-5-sulfonic acid in 100 mM Tris, pH 8.0. This was then followed by dialysis for 7 days against 10 m M Tris, p H 8, to remove the HQSA (Simpson and Vallee, 1968). Samples were incubated with tenfold excess of zinc and a fivefold excess of magnesium for 5 h before assaying; assay conditions were described under Materials and Methods. Cobalt remaining after extended dialysis against hydroxyquinoline-5-sulfonic acid and buffer as described above. 0.48 m M phosphate was added to the 4 - c o enzvme before addine H707.
presence of Chelex 100 (lo%, v/v) (Anderson and Vallee, 1975). Analyses of zinc, cobalt, and magnesium were performed by atomic absorption spectrometry (Fuwa et al., 1964; Iida et al., 1967). Preparation of apophosphatase and its properties are described elsewhere (Bosron et al., 1977). Visible absorption spectra were obtained with a Cary 14 spectrometer equipped with a 0-0.1 absorption slide-wire and magnetic and natural circular dichroism (CD) spectra with a Cary Model 61 spectropolarimeter at a magnetic field of 40 kG, all at room temperature. Since MCD and CD signals are additive, all MCD spectra were corrected for the CD component. Molar absorptivities, E, are in units of M-’ cm-’ and molecular magnetic ellipticities, [e], are given in deg-cm2 dmol-’ kG-’. Electron paramagnetic resonance measurements were performed at 5 K with a Varian E-9 spectrometer (Kennedy et al., 1972). Results Effect of Hydrogen Peroxide on the Activity of Cobalt(ZZ) Phosphatases. Hydrogen peroxide, 8 mM, rapidly inactivates apoenzyme, 0.218 mM, in 10 mM Tris-HC1, pH 8.0, prepared by reconstitution with 6 g-atoms of Co(I1). This inactivation is less rapid than that of the reconstituted enzyme with 4 gatoms, and this, in turn, more rapid than that to which 2 gatoms of Co(I1) has been added (Figure 1A). Peroxide does not affect the activity of the 2 or 4-Zn enzymes; moreover, incubation of the apoenzyme, 0.218 mM, with hydrogen peroxide, 8 mM, for 1 h does not prevent the return of enzymatic activity subsequent to addition of zinc. Magnesium affects the catalytic properties of Co(I1) alkaline phosphatase (Anderson et al., 1975, 1976; Bosron et al., 1977); hence, its effect on the peroxide inactivation of Co(I1) phosphatase was examined. Its presence accelerates the rate of the inactivation of the enzymes to which 4 g-atoms of cobalt is added (Figure 1B) but retards the rate of inactivation of the 2-Co enzyme. Phosphate, 0.48 mM, a competitive inhibitor and substrate, prevents the peroxide inactivation of phosphatase, 0.218 mM, reconstituted with 2,4, or 6 g-atoms of Co(I1). There is no loss of activity of any of these enzymes after incubation for 20 min with 8 mM H202 and upon dilution into 1 M Tris, pH 8.0, containing 6 X mM zinc; all regain activity equal to that
0
4
8
12
4
8
12
MINUTES FIGURE 1: Hydrogen peroxide inactivation of cobalt phosphatase f magnesium. (A) Apophosphatase, 0.218 mM; in Tris-HCI. I O mM,reconstituted with 2 (A),4 (M), or 5 g-atoms ( 0 )of cobalt was incubated with 8 mM Hz02. Aliquots were removed at the times indicated and assayed in 20 mM Veronal-0.4 M NaCl as described under Materials and Methods. (B) Same, except apophosphatase was incubated overnight with a 2.5 molar excess of magnesium.
of their respective control enzyme-phosphate complexes similarly treated but not exposed to peroxide. A series of phosphatases containing 0.5-6 g-atoms of Co(I1) was incubated with hydrogen peroxide to determine the number of Co(I1) ions that must be oxidized to prevent the increase of activity which is characteristically observed on addition of zinc to Co(I1) phosphatases (Tait and Vallee, 1966). The activities of enzymes containing less than 2 g-atoms of Co(I1) are so low that they cannot be assayed directly (Simpson and Vallee, 1968; Anderson et al., 1976). Apophosphatase, 0.218 mM, reconstituted with 0,0.5, 1.0, and 2-6 g-atoms of Co(I1) in 10 mM Tris, pH 8.0, 23 OC, and then incubated with 8 mM hydrogen peroxide for 3 h, followed by the addition of a 7.4-fold molar excess of zinc, restores 21, 13, 3, and less than 2 units of hydrolase activity, respectively, compared to 21 units for the 4-Zn control enzyme, assayed under similar conditions. Thus, oxidation of phosphatase, reconstituted with 1 g-atom or more of Co(II), prevents the restoration of the full native activity of the zinc enzyme. Removal of Co(ZZ) from Co(ZZZ) Phosphatase. Under the conditions examined, treatment of the apoenzyme with hydrogen peroxide does not affect the capacity of Zn(I1) to restore the activity of the apoenzyme or the amino acid composition of phosphatase, nor does it affect the enzymatic activity of zinc phosphatase (Anderson and Vallee, 1975). Hence, removal of cobalt from oxidized cobalt phosphatase should yield an apoenzyme identical to that which results when zinc is removed from the native enzyme. Therefore, 2- and 4-Co(II) alkaline phosphatase as well as the 4-Co(II) phosphatasephosphate complex were incubated with 8 mM hydrogen peroxide for 45 min and then dialyzed under conditions known to remove all metals (see Methods). The products oxidized in this manner were reconstituted with zinc and magnesium and the activities compared with those of preparations treated identically but not exposed to H202 (Table I). The peroxidetreated Co(I1) enzymes retain only 0.4 g-atom of cobalt and regain approximately 60% (Table I, samples 2 and 3) of the activity of the controls (Table I, sample 1) that do not retain residual cobalt. The residual cobalt atoms likely are a fraction of those initially oxidized and may account for the failure of this preparation to regain 100% activity. Under identical conditions, the peroxide-treated 4-Co enzyme-phosphate complex does not contain any residual cobalt, and addition of zinc plus magnesium restores activity virtually completely BIOCHEMISTRY, VOL.
16, NO. 20, 1977
4389
ANDERSON
AND VALLEE
lm WAVELENGTH. nm
3: Effect of H202 on MCD, CD, and absorption spectra of 2Co(I1) 1.6 Mg(I1) phosphatase. Enzyme, peroxide, and buffer concentrations were as in Figure I . Dashed lines indicate the addition of H202. FIGURE
2: EPR spectra of 2-Co(II), 4-C0(11), and 2-c0(11) 1.6Mg(I1)-phosphatase in the presence (-) and absence (-) of hydrogen peroxide. Same conditions as in Figure 1. These and all succeeding spectra were recorded 60 min or more after the addition of H202.Microwave power, 10 m W ;modulation frequency, 100 kHz; microwave frequency, 9.39 kHz, 50 kG,5 K. FIGURE
(Table I, sample 4). Under these conditions used to remove Co(II1) from alkaline phosphatase, there may be a direct release of Co(II1) or Co(II1) may be reduced to Co(I1) and then released from the protein, since hydroquinone compounds have been shown to reduce oxidized cobalt (Bodek and Davies, 1976). Spectral Properties of Cobalt Phosphatase f H202. Oxidation transforms Co(I1) from the paramagnetic (d7) to the diamagnetic (d6) state. Figure 2 shows the effect of H202 on the EPR spectra of 2-Co (lower panel), 2420 1.6-Mg (middle panel), and 4-Co phosphatases (upper panel). When 2 g-atoms of cobalt or less is present, oxidation abolishes both the EPR signal and enzymatic activity. However, when cobalt is in excess of 2 g-atoms, as in 4 - c o phosphatase, the activity is abolished, but the EPR signal is only decreased, not abolished (Figure 2, under panel). Apparently, a fraction of those cobalt atoms which are not involved directly in catalysis is not susceptible to oxidation. Figure 3 illustrates the effect of H202 on the absorption, CD, and MCD spectra of 2-Co 1.6-Mg phosphatase. On oxidation a single maximum at 530 nm (t 210) replaces those at 510 ( t 1 lo), 555 (t 155), and 640 ( e 90). Simultaneously, the natural CD spectrum changes markedly also signaling alterations in the coordination geometry and/or vicinal factors. Upon addition of H202, the CD extremum above 600 nm inverts, accompanied by the appearance of a large positive extremum at 500 nm ( [ e ] 23' = +4500) and a broad negative extremum (400 nm ([e] 23' = -9OO), (Figure 3, middle panel). The corresponding MCD spectrum exhibits an extremum at 540 nm (e = -90) with shoulders at 570 and 495 nm (Figure 3, upper panel) in accord with the features of the absorption spectrum. Except for the intensity, the changes in the absorption spectrum of the 4-Co enzyme (Figure 4, lower panel) are similar to those of the 2-Co 1.6-Mg enzyme (Figure 3, lower panel). A single absorption maximum at 530 nm (t 275) replaces those of the control, the 4-Co(II) enzyme, at 5 10 (€ 185), 555 ( E 240), 605 ( e 120), and 640 nm ( t 135) (Figure4, lower
4390
BIOCHEMISTRY. VOL.
16, NO. 20, 1977
WAVELENGTH FIGURE 4: Effect of H202 on MCD, CD, and absorption spectra of 4Co(I1)-phosphatase. Conditions were as in Figure 3 .
panel); its MCD spectrum reflects similar changes with an extremum at 540 nm ( [ e ] 23' = -150) (Figure 4, upper panel). The natural CD spectrum exhibits a large positive band extending from 450 to 600 nm with an extremum at 520 nm ( t +4800) (Figure 4, middle panel). The results for the 4-co 1.6-Mg and 6-Co enzymes were similar. Effect of H202 on 2-Zinc-Cobalt Hybrid Phosphatases. Zo-Co hybrid enzymes were studied to assign or identify the loci of the oxidized cobalt atoms. Overall, the EPR spectrum of 2-Zn 2-Co phosphatase resembles that of octahedral cobalt complex ions (Kennedy et al., 1972). Oxidation reduces the specific activity of this enzyme, assayed in 1 M Tris, from 27 to 20 units, and brings about minor spectral changes, consistent with the minimal alterations in the EPR (Figure 5, lower panel), CD, MCD, and absorption spectra (not shown), indicating that when cobalt is in the structural and/or regulatory sites it is not susceptible to oxidation. In the 4-Zn 1.6-Co enzyme, zinc apparently occupies both the catalytic and structural binding sites, while cobalt seems to occupy the "regulatory" sites (Anderson et al., 1975), and oxidation of this enzyme with H202 neither changes its specific activity nor its EPR signal (Figure 5, middle panel), supporting the conclusion that cobalt is apparently not susceptible to oxidation in this location. It is conceivable that the majority of cobalt atoms could
METAL-BINDING SITES OF PHOSPHATASE ~~
Zinc and Magnesium Binding to Cobalt Alkaline Phosphatase f H ~ 0 2 . ~ TABLE 11:
Metal added (g-atom/mol) (1) 2 c o (2) 2 Co (3) 2 CO (4) 2 Co (5) 2 Co (6) 2 CO (7) 4 c o (8) 4 Co (9) 4 CO
+ 7.4 Zn + H202 + H202 + 7.4 Zn + 2.5 Mg + 2.5 Mg + H202 + 7.4 Zn + H202
(10)4Co+H202+7.4Zn
Metal bound (aatom/moI)'Zn Co Mg 0 2.1 0 3.7 1.0 0 0 2.1 0 2.0 2.1 0 0 2.1 1.6 0 2.1 1.6 0 3.7 0 1.6
3.6
1.1 3.7 3.7
0 0 0 0
Metal-free alkaline phosphatase, 0.218 mM, in 10 mM Tris, pH 8.0, was reconstituted with 2 or 4 g-atoms of cobalt; hydrogen peroxide, 8 mM, was added as indicated. Twenty minutes after adding hydrogen peroxide, a 7.4 molar excess of zinc and/or a 2.5 molar excess of magnesium were added to samples as indicated. After another 20 min had elapsed, 0.25-mL aliquots were passed over a 1.7 X 15 cm Sephadex G-25 column equilibrated with metal-free 10 mM Tris, pH 8.0. All samples were analyzed for protein, zinc, cobalt, and magnesium (see Materials and Methods). occupy the same sites in both the 2-Zn 2-Co and the 4-Zn 1.6-Co enzymes. Therefore, studies of the 2-Zn 3.5-Co phosphatase were performed.2 Oxidation of this enzyme alters the pentacoordinate-like component of the EPR spectrum and enzymatic activity decreases from 3 1 to 21 units. The residual EPR signal is largely octahedral-like and the loss of the spectrum indicative of pentacoordinate-like cobalt (but not of that which is octahedral-like) is accompanied by a simultaneous loss of enzymatic activity. Metal Binding to Phosphatase. The diminution of the EPR signal on oxidation of the 2-C0, 4-Co and 2-Co 1.6-Mg enzymes cannot be explained based on decreased cobalt binding (Figure 2). Oxidation does not alter the amount of cobalt bound (Table II), though it does affect the number of cobalt ions which exchange with zinc. The addition of a 7.4-fold molar excess of zinc (zinc to protein, M / M ) to the control 2-Co(II) enzyme results in the binding of 3.7 g-atoms of zinc and 1.1 g-atoms of cobalt (Table 11, sample 2) (Anderson et al., 1975). While addition of excess zinc to the 2-Co(III) enzyme does not alter the cobalt content, it additionally binds 2.0 g-atoms of zinc (Table 11, sample 4). Apparently, the exchange inert Co(II1) occupies two of the metal-binding sites of the hydrogen peroxide treated enzyme; while zinc readily binds to two vacant metal-binding sites, it does not displace Co(II1). Upon addition of excess zinc, the control 4-Co(II) enzyme binds 3.7 g-atoms of zinc and 1.1 g-atoms of cobalt (Table 11, sample 8), but the oxidized 4-Co enzyme contains only 1.6 g-atoms of zinc and 3.7 g-atoms of cobalt (Table 11, sample 10). Discussion Alkaline phosphatase contains three classes of metal ions which are involved in catalysis, structure stabilization (Simpson and Vallee, 1968), and regulation (Anderson et al., 1975) and is particularly suited for such studies. Metal atoms involved directly in catalysis are postulated to be in a tetra- or pentacoordinate-like geometry (Simpson and Vallee, 1968; Taylor et al., 1973; Taylor and Coleman, 1972; Anderson et al., 1976),while the structural and regulatory metal atoms are To minimize the concentration of free cobalt ions in solution, less than 4 g-atoms of cobalt was added.
500 gouss FIGURE 5: Effect of H202 on Zn-Co hybrid alkaline phosphatase. Enzyme concentration, peroxide, and buffer were as in Figure 1. Zinc was added and allowed to equilibrate for 15 min prior to the addition of cobalt. Conditions of the EPR are described in Figure 2.
thought to occupy distorted octahedral-like sites (Simpson and Vallee, 1968; Anderson et al., 1976). The apparent differences in the geometries of these classes of metals, conditioned by their binding sites, coincide with different relative affinities for metals, and the binding of one affects that of others (Bosron et al., 1977; Anderson et al., 1975). Selective oxidation of phosphatase containing more than 2 g-atoms of cobalt differentiates the classes of metals in E . coli alkaline phosphatase. Peroxide-treated 2-Co phosphatase loses all activity, and the amplitude of the EPR signal is abolished. However, while oxidation of 4-co phosphatase results in the complete loss of enzymatic activity, there is only a 50% decrease in amplitude of the EPR signal, indicating that not all of the cobalt ions are susceptible to oxidation (Figure 2, upper panel). Thus, this oxidation seems selective for only one class of cobalt atoms. It does not oxidize the protein, since peroxide does not affect Zn phosphatase, the return of activity of apophosphatase subsequent to the addition of zinc (Figure l ) , or the amino acid composition of cobalt phosphatase (Anderson and Vallee, 1975). Since peroxide does not affect the amount of metal bound (Table II), the decrease of the EPR signal and the simultaneous loss of enzymatic activity reflect the oxidation of Co(I1) to Co(III), not changes in metal binding. However, oxidation does affect the amount of cobalt that can be displaced by zinc. For example, addition of a 7.4-fold excess of zinc to the control 2-Co(II) enzyme results in the binding of 3.7 g-atoms of zinc, but addition of similar amounts of zinc to the 2-Co(III) enzyme allows the binding of only 2 g-atoms of zinc (Table 11). Thus, oxidation affects neither the amount of cobalt bound nor the binding of zinc to the remaining zinc sites, but it does transform some of the bound cobalt atoms from exchange-labile Co( 11) to exchange-inert Co(II1). There is considerable evidence, derived from stopped-flow (Trentham and Gutfreund, 1968; Halford, 197 1; Lazdunski et al., 1970, 1971; Chappelet-Tordo et al., 1974) and phosphate-binding studies (Reid et al., 1969; Simpson and Vallee, 1970; Applebury et al., 1970; Lazdunski et al., 1971), indicating that the two subunits of alkaline phosphatase are not equivalent functionally. Our data indicate that the binding (and/or) oxidation of cobalt in cobalt alkaline phosphatase is also nonequivalent, because the oxidation of only 1 g-atom of cobalt per dimer prevents the restoration of activity upon addition of Zn. If the binding sites were identical, some of the phosphatase molecules should contain more than 1 g-atom of oxidized cobalt. Assuming this to be the case, a significant quantity of apophosphatase dimers should be detected; these BIOCHEMISTRY, VOL.
16, N O . 20, 1 9 7 7
4391
ANDERSON AND VALLEE
would not contain any oxidized cobalt and would allow restoration of enzymatic activity upon addition of zinc. The data do not support the latter assumption; rather, they imply the nonequivalence of the metal-binding sites of dimeric E . coli alkaline phosphatase. The data do not definitively identify the cobalt atoms which are oxidized with respect to the type of site with which they are associated, though they strongly suggest that those bound to the regulatory (magnesium) site are not susceptible to oxidation, since peroxide does not affect the spectrum or activity of the 4-Zn 2-Co enzyme (Figure 5, middle panel). Further, cobalt, when at the structural sites, also is not susceptible to oxidation, since peroxide minimally affects the activity and EPR spectra of the 2-Zn 2-Co enzyme in which the cobalt atoms have been thought to be at the structural sites in an octahedral-like geometry. This hypothesis is supported by functional data. Thus, this enzyme exhibits 27 units of activity contraindicating involvement of the catalytic site and displays both hydrolase and transferase activities. If cobalt were to occupy the catalytic sites, this would result in an enzyme totally lacking in transferase activity and with a specific hydrolase activity of less than 2 units. The 2-Zn 3.5-Co enzyme, where cobalt occupies both the structural and regulatory sites, is also relatively resistant to peroxide treatment. The small decreases in enzymatic activity of the 2-Zn 3.5-Co and 2-Zn 2-Co phosphatases are accompanied upon oxidation by similar decreases in the amount of pentacoordinate-like cobalt present; conversely, the amount of octahedral-like cobalt, which is not associated with activity, remains constant. Thus, the data suggest that peroxide does affect the cobalt atoms at the catalytic sites which gives rise to pentacoordinate-like spectral properties but does not alter those in the structural or regulatory sites which display octahedral-like characteristics. The 2-Co enzyme contains a mixture of cobalt ions coordinated in more than one geometry (Figure 3) (Anderson et al., 1976); yet, the oxidative abolition of the EPR signal indicates that all of these cobalt ions are affected. Since only cobalt in pentacoordinate geometry is susceptible to oxidation, an additional concurrent process, e.g., migration of cobalt ions, must be postulated to account for the data: once the cobalt ions exchange to and occupy a site whose ligands are suitable for a geometry favorable for Co(III), oxidation could occur. On this basis, the rates of oxidation and inactivation of the 2420 enzyme would be expected to be biphasic (assuming that migration is rate limiting), while those of the 6-Co enzyme, where all sites are filled, would be expected to be monophasic, as is indeed evidenced from the semilog plot (not shown) of the data in Figure 1. Our data indicate that it is the cobalt ions in a pentacoordinate-like geometry which are oxidized preferentially; however, in the 2-Co l .6-Mg enzyme in which a large portion of the cobalt ions are in a pentacoordinate-like geometry the rate of oxidation is slower than that of the enzyme containing only 2 Co atoms which relatively contains a lower amount of pentacoordinate-like cobalt ions. A plausible explanation for this phenomenon may be that magnesium stabilizes the pentacoordinate-like geometry and inhibits the formation of an octahedral-like geometry which is essential for Co(II1) (Wood and Remeika, 1967). The data of Bosron et al. (1977) on tritium exchange of cobalt phosphatase also demonstrate that magnesium stabilizes the structure of 2-Co phosphatase. Minimally, the conversion of exchange-labile Co(I1) to exchange-inert Co(II1) in situ should be accompanied by the diminution or abolition of the EPR spectrum characteristic of d7 cobalt, and the cobalt must remain bound and become relatioely exchange inert. Co(III), though its exchange reac-
4392
BIOCHEMISTRY, VOL. 16, NO.
20, 1917
tions are much slower than those of the Co(I1) species, is not exchange inert in an absolute sense (Taube, 1952). While it is relatively easy to obtain data regarding the decreased rate of exchange involving the conversion of Co(I1) to Co(III), they are not sufficient to prove the formation of Co(II1); toward that end, spectra and metal-binding data must be obtained. Our data satisfy all of these criteria. They further provide evidence for the loci of the oxidized metal atoms and information regarding the structural and chemical properties of the catalytic, structural, and regulatory metal-binding sites of E . coli alkaline phosphatase. References Anderson, R. A., Bosron, W. F., Kennedy, F. S., and Vallee, B. L. (1975), Proc. Natl. Acad. Sci. U.S.A. 72, 29892993. Anderson, R. A., Kennedy, F. S., and Vallee, B. L. (1976), Biochemistry 15, 3710-3716. Anderson, R. A., and Vallee, B. L. (1975), Proc. Natl. Acad. Sci., U.S.A. 72, 394-397. Applebury, M. L., Johnson, B. P., and Coleman, J. D. (1970), J. Biol. Chem. 245, 4968-4976. Bodek, I., and Davies, G. (1976), Znorg. Chem. 15. 922926. Bosron, W. F., Anderson, R. A., Kennedy, F. S., and Vallee, B. L. (1977), Biochemistry (submitted). Bosron, W. F., Kennedy, F. S.,and Vallee, B. L. (1975), Biochemistry 14, 2275-2282. Chappelet-Tordo, D., Iwatsubo, M., and Lazdunski, M. (1974), Biochemistry 13, 3754-3672. Fuwa, K., Pulido, P., McKay, R., and Vallee, B. L. (1964), Anal. Chem. 26, 2407-241 1. Halford, S. E. (1971), Biochem. J. 125, 319-327. Iida, C., Fuwa, K., and Wacker, W. E. C. (1967), Anal. Biochem. 18, 18-26. Kennedy, F. S., Hill, H. A. O., Kaden, T. A., and Vallee, B. L. (1972), Biochem. Biophys. Res. Commun. 48, 15331539. Lazdunski, M., Petitclerc, C., Chappelet-Tordo, D., and Lazdunski, C. (197 l), Eur. J . Biochem. 20, 124- 139. Lazdunski, C., Petitclerc, C., Chappelet-Tordo, D., Letterier, F., Douzou, P., and Lazdunski, M. (1970), Biochem. Biophys. Res. Commun. 40, 589-593. Plocke, D. J., and Vallee, B. L. (1962), Biochemistry 1, 1039 -1043. Neu, H. C., and Heppel, L. A. (1965), J. Biol. Chem. 240, 3685-3692. Reid, T. W., Paolic, M., Sullivan, D. J., and Wilson, 1. B. (1969), Biochemistry 8, 2374-2380. Simpson, R. T., and Vallee, B. L. (1968), Biochemistry 7, 4343 -43 50. Simpson, R. T., and Vallee, B. L. (1970), Biochemistry 9, 953-958. Simpson, R. T., Vallee, B. L., and Tait, G. H . (1968), Biochemistry 7, 4336-4342. Tait, G. H., and Vallee, B. L. (1966), Proc. Natl. Acad. Sci. U.S.A. 56, 1247-1251. Taube, H. (1952), Chem. Reu. 50, 69-126. Taylor, J. S., and Coleman, J. E. (1972), Proc. Natl. Acad. Sci. U.S.A. 69, 859-862. Taylor, J. S., Lau, C. Y., Applebury, M. L., and Coleman, J . E. (1973), J. Biol. Chem. 248, 6216-6220. Thiers, R. E. (1957), Methods Biochem. Anal. 5 , 273-335. Trentham, D. R., and Gutfreud, H. (1968), Biochem. J. 106. 455-460.
ASYMMETRY IN A TRIOSEPHOSPHATE DEHYDROGENASE
Vallee, B. L., and Wacker, W. E. C. (1970), Proteins 2nd Ed. 5. Vallee, B. L., and Williams, R. J. P. (1968a), Proc. Nutl. Acad. Sci. U.S.A.59, 498-505. Vallee, B. L., and Williams, R. J. P. (1968b), Chem. Br. 4,
397-402. Wilson, I. B., Dayan, J., and Cyr, K. (1964), J . Biof. Chem. 239, 4182-4185. Wood, D. L., and Remeika, J. P. (1967), J . Chem. Phys. 46, 3595-3602.
Molecular Asymmetry in an Abortive Ternary Complex of Lobster Glyceraldehyde-3-phosphateDehydrogenase? R. Michael Garavito,: Denis Berger,§ and Michael G. Rossmann*
ABSTRACT: An abortive ternary complex of lobster glyceraldehyde-3-phosphate dehydrogenase was produced by the covalent attachment of 3,3,3-trifluoroacetone to Cys- 149 in each subunit. X-ray diffraction analysis of the glyceraldehyde-3phosphate dehydrogenase-trifluoroacetone-nicotinamide adenine dinucleotide complex showed asymmetry with respect to the active-site conformations of the trifluoroacetone sub-
strate analogue and some catalytic groups. These results are consistent with 19Fnuclear magnetic resonance observations of this complex (Bode, J., Blumenstein, M., and Raftery, M. A. (1975), Biochemistry 14, 1153-1 160). Different substrate conformations were found on opposite sides of the molecular diad relating subunits whose active centers are in close proximity (the R axis).
T h e interaction of ~-glyceraldehyde-3-phosphate dehydrogenase (EC 1.2.1.12) with a variety of ligands is frequently different within each subunit. For instance, the apo-holo enzyme transition can exhibit either negative (Conway and Koshland, 1968; De Vijlder et al., 1969; Boers and Slater, 1973; Schlessinger and Levitzki, 1974) or positive (Cook and Koshland, 1970; Kirschner, 1971; Kirschner et ai., 1971) cooperativity, and the covalent addition of substrate analogues to only two of the subunits can completely inactivate the enzyme ( MacQuarrie and Bernhard, 197 1; Stallcup and Koshland, 1973; Levitzki, 1973; Seydoux et al., 1973; Bernhard and MacQuarrie, 1973). On the other hand, the binding of NAD+ to acylated enzyme (Seydoux et al., 1976) and the formation of the Racker band (Racker and Krimsky, 1952; Krimsky and Racker, 1955) is apparently identical in all subunits. The two models (Figure 1) which have been proposed to explain the functional nonequivalence of active sites in terms of conformational asymmetry are based on either ligand-induced conformational transitions (Stallcup and Koshland, 1973; Levitzki, 1973) or preexisting molecular asymmetry (Bernhard and MacQuarrie, 1973). Conformational asymmetry has been observed for oligomeric proteins in insulin (Blundell et al., 1972; Hodgkin, 1974; Bentley et al., 1976), a-chymotrypsin dimers (Tulinsky et al., 1973), and tobacco mosaic virus (Champness et al., 1976). Asymmetry in the binding affinity of ligands to covalently identical subunits has been shown for NAD+ to soluble malate dehydrogenase (Hill et al., 1972) and for sugars to hexokinase
(Anderson and Steitz, 1975). Conformational asymmetry has been induced on binding of coenzyme to lactate dehydrogenase (Adams et al., 1970) and removal of calcium from concanavalin A (Jack et al., 1971). Crystals of human holo-GAPDHI (Watson et al., 1972; Mercer et al., 1976) show a crystallographic twofold axis coincident with the Q molecular axis (see Buehner et al. (1974a) for the nomenclature of the three orthogonal P,Q, R molecular axes). The three-dimensional structure of lobster holo-GAPDH exhibits 222 symmetry to within the accuracy of a 3.0-1( resolution electron-density map (Buehner et al., 1974a). Moras et al. (1975) subsequently showed that a more accurate map at 2.9-A resolution suggested asymmetry of the protein structure between active centers as well as asymmetry of bound NAD+. Their results indicate a conservation of the Q axis but not of the P and R axes, consistent with the observations on human holo-GAPDH and the anticipated cooperative interaction of the subunits across the R axis (Buehner et al., 1974a). Nevertheless, the validity of these conclusions was brought into question by observing symmetric binding of nicotinamide 8-bromoadenosine dinucleotide to lobster GAPDH (Olsen et al., 1976a) and symmetric displacement of sulfate by citrate ions from the active center (Olsen et al., 1976b). Bode et al. (1975a,b) studied the interaction of the substrate analogue 3,3,3-trifluorobromoacetonewith rabbit muscle GAPDH. The TFA label was of the approximate size and shape as the natural thioester intermediate (Figure 2) which, it was hoped, would mimic the true enzyme-substrate complex, although McCaul and Byers (1976) have shown that analogues without a terminal phosphate have decreased reactivity. The CF3 group acted as a probe in the subsequent 19F nuclear magnetic resonance study of Bode et al. (1975a). They observed both the reduction of the keto group by NADH and
t From the Department of Biological Sciences, Purdue University, West Lafayette, Indiana 47907. Received April 8. 1977. The work was supported by the National Institutes of Health (Grant GM 10704) and the National Science Foundation (Grant BMS74-23537). Supported by a National Institutes of Health Molecular and Cellular Biology Training Grant. 5 Supported by a postdoctoral fellowship from the Swiss National Science Foundation. Present address: Hopital Contonal, Institut Universitaire de Medicine Physique et de Reeducation, 121 1 Geneve 4, Switzerland.
*
I Abbreviations used are: GAPDH, ~-glyceraldehyde-3-phosphate dehydrogenase; TFA-Br, trifluorobromoacetone; TFA, trifluoroacetonyl; PGA, 3-phosphoglycerate; NAD+, nicotinamide adenine dinucleotide; NADH, reduced NAD+; EDTA, (ethylenedinitri1o)tetraacetic acid.
BIOCHEMISTRY, VOL. 16, NO. 20, 1977
4393
