Mammalian Alkaline Phosphatases Are Allosteric Enzymes*

Mammalian alkaline phosphatases (APs) are zinc-containing metalloenzymes encoded by a multigene family and functional as dimeric molecules. Using human placental AP (PLAP) as a paradigm, we have investigated whether the monomers in a given PLAP dimer are subject to cooperativity during catalysis following an allosteric model or act via a half-of-sites model, in which at any time only one single monomer is operative. Wild type and mutant PLAP homodimers and heterodimers were produced by stably transfecting Chinese hamster ovary cells with mutagenized PLAP cDNAs followed by enzyme extraction, purification, and characterization. [Gly429]PLAP manifested negative cooperativity when partially metalated as a consequence of the reduced affinity of the incompletely metalated AP monomers for the substrate. Upon full metalation with Zn2+, however, the negative cooperativity disappeared. To distinguish between an allosteric and a half-of-sites model, a [Gly429]PLAP-[Ser84]PLAP heterodimer was produced by combining monomers displaying high and low sensitivity to the uncompetitive inhibitor l-Leu as well as a [Gly429]PLAP-[Ala92]PLAP heterodimer combining a catalytically active and inactive monomer, respectively. The l-Leu inhibition profile of the [Gly429]PLAP-[Ser84]PLAP heterodimer was intermediate to that for each homodimer as predicted by the allosteric model. Likewise, the [Gly429]PLAP-[Ala92]PLAP heterodimer was catalytically active, confirming that AP monomers act independently of each other. Although heterodimers are structurally asymmetrical, they migrate in starch gels with a smaller than expected weighted electrophoretic mobility, are more stable to heat denaturation than expected, and are more sensitive to l-Leu inhibition than predicted by a strict noncooperative model. We conclude that fully metalated mammalian APs are noncooperative allosteric enzymes but that the stability and catalytic properties of each monomer are controlled by the conformation of the second AP subunit.

Alkaline phosphatases (AP) 1 are ubiquitous enzymes found in most species from bacteria to man (1). Human APs are encoded by four genes (reviewed in Refs. 2 and 3), i.e. the placental (PLAP), germ cell (GCAP), intestinal, and tissuenonspecific AP isozyme, respectively. APs are dimeric metalloenzymes that catalyze the hydrolytic transfer of phosphate to water or its transphosphorylation to amino alcohols (4), but when separated the monomeric subunits fail to display enzyme activity. Three metal ions (two Zn 2ϩ an one Mg 2ϩ ) in the active site (5) are essential for enzymatic activity. However, these metal ions also contribute substantially to the conformation of the AP monomer and indirectly regulate subunit-subunit interactions (6).
Fully metalated Escherichia coli AP dimers are symmetrical, both by crystallographic measurements and spectroscopic methods (7), but partially metalated dimers manifest structural asymmetry. As a result of such molecular asymmetry, APs have been claimed to be capable of accepting only one single-substrate molecule in a half-of-sites reactivity mechanism and to display negative cooperativity (8,9). The existence of cooperativity for the interaction between AP subunits has been investigated intensively in the E. coli enzyme, and evidence for both the existence of positive and negative cooperativity has been presented (10,11). During early studies, hybrid E. coli AP dimers were formed using mixtures of native and partially modified enzyme forms during a random reconstitution step following denaturation of the dimers with chaotropic agents (12,13). More recently, heterodimers were generated upon controlled proteolysis of E. coli AP, not necessitating any AP denaturation (14). The latter study of the resulting hybrid APs clearly indicated asymmetry in these hybrids both in terms of structure and function, pointing to catalytically relevant subunit communication in the AP enzyme.
Mammalian APs have a unique surface loop not present in the E. coli enzyme that extends from amino acids 400 -430 (15). This loop has been shown to play an important role in defining the conformation and stability of the AP molecule (16). The loop is also partially responsible for the interaction of APs with extracellular matrix proteins, such as collagen (17). We have also shown that this loop is responsible for the unique property of mammalian APs of being uncompetitively inhibited by a number of amino acids and small peptides (17,18). We and others have found that a single amino acid substitution, E429G, was primarily responsible for the differential inhibition of PLAP and GCAP by L-Leu (19,20). The degree of inhibition was further modulated by a second substitution, N84S, in GCAP, causing a conformational change in the molecule accompanied by a 50% drop in the degree of inhibition by L-Leu (16). Because the [Gly 429 ]PLAP 2 and the [Ser 84 ]PLAP mutants dis-play a 100-fold difference in K i for L-Leu, heterodimers of these mutants provided an optimal test system to study AP subunit interactions and negative cooperativity for mammalian APs. In this paper we report that AP dimers display negative cooperativity when the AP monomers are partially demetalated, but both AP monomers function essentially independently when both subunits are properly metalated. This behavior of mammalian APs fits the definition of an allosteric enzyme. Site-directed mutagenesis was performed as described (21). The sequence of the mutagenesis primer pairs were as follows (underlined bases indicate changes): S92A, 5Ј-CTCTTCCGCTGGAGCCACAGC-CACGGC-3Ј and 5Ј-CTCTTCCAGCGTCTGGCACATGTTTGTC-3Ј; P209R, 5Ј-CTCTTCATGTTTCGCATGGGGACCC-3Ј and 5Ј-CTCT-TCAACATGTACTTTCGGCCTCC-3Ј; P479R, 5Ј-CTCTTCTGGCGC-CCGCCGCCGGCACC-3Ј and 5Ј-CTCTTCCGCCAGGTCGCAGGCGGT-GTAG-3Ј. Finally, the sequence TESESGSPE at positions 410 -418 in [S]PLAP was replaced by SMDVYAHNN to produce the heat-labile mutant [S]PLAP-t (17). All mutant PLAP cDNAs were subcloned into the pSVT7 vector (22) and transfected into Chinese hamster ovary cells (23), and transfected cells were selected (19). At confluency, the double transfectants were washed with 20 mM Tris-HCl buffer, pH 7.5, containing 140 mM NaCl. Cultured cells were extracted with a 1:1 mixture of butanol and 50 mM sodium acetate buffer, pH 5.5, containing 100 mM NaCl, 20 M ZnCl 2 , 0.5 mM MgCl 2 , and 0.05% merthiolate. The pH in these extracts was then titrated to 7.5 with 1.0 M Tris, and butanol was removed by evaporation. JEG3-choriocarcinoma cells were grown in the presence of increasing concentrations (0 -4 mM) of sodium butyrate to progressively induce GCAP (24). Following overnight dialysis and neuraminidase treatment (1 unit/ml), [S]PLAP-[RGR]PLAP heterodimers were separated from their respective parent homodimers by cation-exchange chromatography on Mono-S Sepharose in 20 mM acetate buffer, pH 5.8, and elution with a linear NaCl gradient (0 -250 M).
Enzyme Kinetics-AP kinetic determinations were performed as described previously (18) using p-nitrophenylphosphate (pNPP) as substrate in 1.0 M diethanolamine (DEA) buffer, pH 9.8, containing 20 M ZnCl 2 and 0.5 mM MgCl 2 . Experiments in the absence of transphosphorylation acceptors were performed in 10 mM Tris-HCl buffer, pH 7.5, containing 20 M ZnCl 2 and 0.5 mM MgCl 2 . To investigate the role of metal ions on the existence of enzyme cooperativity, [G]PLAP was treated with Chelex for 24 h in 50 mM Tris-HCl buffer, pH 7.5, containing 150 mM NaCl, after which Michaelis-Menten kinetics were studied for various degrees of metalation using buffered substrate treated with Chelex. Isolated homo-and heterodimeric APs were tested for their time-dependent inactivation at 56°C as described (25), and enzyme catalysis by the dimers in the presence of 0 -1.0 M guanidinium chloride was tested in 10 mM Tris-HCl buffer, pH 7.5, containing 20 M ZnCl 2 and 0.5 mM MgCl 2 .
Statistical Validation-Double reciprocal plots were constructed via linear regression analysis by correlating initial rates of pNPP conversion with substrate concentration, using the GRAFIT version 3.0 Erithacus software. For experiments performed in the presence of EDTA, inhibition constants (K i ) were derived from the experimental inhibitor concentrations and the calculated slopes of the fitted regression lines by introducing both parameters in the proper expression for the initial rate. The K i values were then calculated for several concentrations of inhibitor as the means Ϯ S.D. To confirm the validity of residual enzyme level predictions during AP inhibition by L-Leu, the experimental and the predicted residual AP activities of [S]PLAP and [G]PLAP homodimers were correlated via regression analysis, for inhibitions at equal concentrations of L-Leu. To statistically identify the kinetic model dictating AP enzyme catalysis (see below), residual experimental AP levels for [S]PLAP-[G]PLAP heterodimers were also correlated via linear regression analysis with residual enzyme levels at equal concentrations of L-Leu, as predicted by each separate kinetic model. Statistically weighted numerical values for K m were calculated as described by Wilkinson, including determination of standard errors of the mean (26).
Kinetic Mechanism of Subunit Interactions-The general scheme for the AP-catalyzed hydrolysis of phosphate substrates can be summa-rized as shown in Scheme I (4). Whereas at low pH E-P accumulates, at alkaline pH the dissociation of E⅐P is rate-limiting. However, in the presence of a transphosphorylating amino alcohol (R 2 OH), the enzyme is readily regenerated. Certain amino acids inhibit AP by blocking both hydrolysis and transphosphorylation of the E-P complex, thus explaining why this inhibition is uncompetitive (18,27).
The AP catalysis can be described by simple Michaelis-Menten kinetics, although both K m and k cat are complex functions of the rate constants, as depicted in Scheme I. However, to study subunit interactions and cooperativity and to comply with the dimeric nature of the AP enzyme, AP catalysis can better be represented by the generalized model applicable for allosteric enzymes (28) (Scheme II). In this model, K 1 and K 2 represent the Michaelis constants for both subunits respectively, whereas k 1 and k 2 stand for the catalytic rate constants for each monomer. For a classical noncooperative allosteric dimeric enzyme, K 1 ϭ K 2 and k 1 ϭ k 2 , the kinetic analysis of these enzymes being described accurately by linear Lineweaver-Burk plots, with k cat ϭ 2k 1 and K m ϭ K 1 (28).
However, because PLAP, respectively [S]PLAP, and GCAP, respectively [RGR]PLAP, have slightly different kinetic constants, in the case of [S]PLAP-[RGR]PLAP heterodimers Scheme II would be more accurately described by the following rate equation, at least for enzyme heterodimers in which both monomers act independently: with [E 1 E 2 ] o being the total AP concentration and k 1 , K 1 respectively k 2 , K 2 being the rate constants and Michaelis constants for the individual AP subunits. Linear Lineweaver-Burk kinetics can be anticipated from Equation 1, based on previously determined kinetic constants for [S]PLAP and [RGR]PLAP, respectively, with k cat ϭ k 1 ϩ k 2 . The K m corresponds to the positive solution of the following second order function: This analysis assumes that in the heterodimers each monomer will catalyze phosphate substrates with similar catalytic efficiencies as in the parent homodimers. Whereas no linear kinetics can arise when the SE 1 E 2 S intermediate would be the only active enzyme-substrate complex metabolized (i.e. SE 1 E 2 and E 1 E 2 S are inactive), linear kinetics are also predicted when only one of both subunits (i.e. only SE 1 E 2 and not E 1 E 2 S) participates in the catalysis (i.e. k 2 ϭ 0), because then Equation 1 reduces to a simple Michaelis-Menten form and k cat ϭ k 1 and K m ϭ K 1 . Such a situation would arise if as a result of structural cross-talk between both AP monomers; the second AP monomer is shut off for substrate positioning as a consequence of substrate binding to the first subunit, in agreement with a half-of-sites model. The reaction rate would then reduce to: and describe a mechanism that results in linear double reciprocal plots of enzyme activity versus substrate over a wide range of substrate concentrations. Formally, under those conditions, enzyme kinetics for AP heterodimers match those for equal mixtures of both homodimers, with comparable degrees of saturation. In this model homodimeric APs are described kinetically by k cat ϭ k 1 and K m ϭ K 1 /2. Formally this model also describes the kinetics of a heterodimer composed of one active and one inactive monomer, both in an allosteric and in a half-ofsites model. When SE 1 E 2 S is only formed at much higher substrate concentrations than required to form either SE 1 E 2 or E 1 E 2 S, nonlinear double reciprocal plots of enzyme activity versus substrate concentration will be found even for homodimers, typical of negative cooperativity, as can be substantiated from Equation 1, because both AP monomers would be nonequivalent.
It is clear that the above equations do not enable an easy distinction between the allosteric and the half-of-sites model. The uncompetitive amino acid inhibitor L-Leu on the contrary enabled us to further distinguish between Equations 1 and 3. Uncompetitive inhibition of AP homo-and heterodimers was carried out at saturating concentrations of pNPP (10 mM) in the presence of increasing concentrations of L-Leu (0 -50 mM), and the residual AP activity was measured. Because at high [pNPP] in the allosteric model (Equation 1) only the substrate intermediate SE 1 E 2 S is metabolized, schematically the inhibition of AP heterodimers can be represented as in Scheme III and be described by the following rate equation: On the contrary, according to the half-site model at saturating [pNPP], the inhibition is represented as in Scheme IV with the following rate equation: Based on the known values of k 1 , k 2 , K 1 , K 2 , K i1 , and K i2 (18) and assuming no asymmetry-dependent cross-talk between both monomers in a nonsymmetrical heterodimer, it is possible to predict residual AP activities in the presence of L-Leu, according to both models (Equation 5, respectively 4) and to compare these predicted enzyme levels with actual data collected for heterodimers between [S]PLAP and [G]PLAP forms.

RESULTS AND DISCUSSION
Active Site Zn 2ϩ Stability-Double reciprocal plots of enzyme activity versus the concentration of pNPP were linear for various EDTA concentrations tested (r ranging from 0.988 to 0.999) and intersected on the y axis (mean Ϯ S.D. of the intersection equalled A 405 nm Ϫ1 ϭ 1.63 Ϯ 0.26), compatible with competitive inhibition (Fig. 1A). Thus [S]PLAP activity was inhibited by EDTA with a K i equal to 0.51 Ϯ 0.06 mM, indicative of the high stability of active site bound Zn 2ϩ ions; wt PLAP was likewise inhibited with a K i equal to 0.87 Ϯ 0.3 mM. In agreement with these data, [S]PLAP could not be demetalated by Chelex, not even after prolonged incubations (not shown). On the contrary, [G]PLAP was also inhibited competitively by EDTA ( Fig. 1B; r of the regression lines ranging from 0.994 to 0.998; intersection at A 405 nm When [G]PLAP, fully loaded with Zn 2ϩ , is diluted in Chelextreated buffered substrate solutions, a progressive loss of the AP activity is observed over time, independently of the substrate concentration, as shown by the decreasing slope of AP activity plots versus time ( Fig. 2A). This behavior is in agreement with the rapid loss of [G]PLAP activity in the presence of EDTA and is indicative of the spontaneous dissociation of Zn 2ϩ from the active site. Replots of the slopes of these curves versus time were linear (Fig. 2B, r ϭ Ϫ0 Active Site Zn 2ϩ and Cooperativity-Fully metalated [G]PLAP incubated with increasing substrate concentrations up to 100 mM (1000-fold above the K m ) in 1 M DEA buffer, pH 9.8, shows no evidence of negative cooperativity. On the contrary, a mild inhibition is observed at substrate concentrations exceeding 10 mM (not shown). Similarly, measurements at pH 7.5, in 10 mM Tris-HCl buffer containing 20 M ZnCl 2 and 0.5 mM of MgCl 2 but no transphosphorylating alcohol still shows no evidence of negative cooperativity for pNPP up to 100 mM, at which concentration a clear-cut substrate inhibition of about 50% is observed (Fig. 3A). Experiments in the presence of 0.5 M guanidinium chloride, claimed to enhance AP activity (29), do not raise the AP activity measured at high pNPP concentrations, whereas in the presence of 1.0 M guanidinium chloride, a clear-cut drop in enzyme activity is observed (Fig. 3A). These experiments indicate that fully metalated APs do not display negative cooperativity.
Whereas Chelex-treated [G]PLAP displays negligible enzyme activity, Zn 2ϩ can restore the activity in a dose-dependent manner. Little or no activity is regained in the presence of 0.5 M of ZnCl 2 , partial reconstitution is observed with 2.5 M ZnCl 2 , and full reconstitution can be achieved in the presence of 20 M ZnCl 2 (Fig. 3B), whereas the subsequent addition of Because extrapolated V max values are identical for the partially reconstituted and fully reconstituted enzyme, this implies that on substrate saturation both of the enzyme's sites are occupied and that they function with identical catalytic rates, i.e. even though partial metalation increases k 1 , it does not affect the rate of phosphorylation k 2 and dephosphorylation k 3 (Scheme I). Together with the above findings, it follows that when fully metalated, AP dimers function noncooperatively.
Kinetics Properties of Heterodimers-To further study whether AP monomers acted independently, we made use of the differential L-Leu inhibition properties of PLAP and GCAP. Although the inhibition of wt GCAP (K i ϭ 0.54 Ϯ 0.01 mM) and wt PLAP (K i ϭ 9.2 Ϯ 1.2 mM) by L-Leu differ 17-fold, the inhibition of [S]PLAP (K i ϭ 19 Ϯ 1.5 mM) and [G]PLAP (K i ϭ 0.2 Ϯ 0.01 mM) differ 100-fold (18), and these mutants were therefore chosen for our experiments. To optimize the chromatographic separation of the AP heterodimers we also engineered two charge replacements, P209R and P479R in [G]PLAP ([RGR]PLAP). These substitutions were chosen, because they represent known allelic mutations in PLAP and GCAP and neither of them affect the catalytic properties of the enzyme (18,25). Fig. 4 illustrates the predicted inhibition curves of heterodimers constructed between PLAP and GCAP (Fig. 4A)  for the [S]PLAP-t-[RGR]PLAP heterodimer was, however, higher than the expected K m values predicted by the models described by Equations 1 (expected K m ϭ 0.2 mM) and 3 (expected K m ϭ 0.14 mM), i.e. is higher than the weighted average of the parent molecules regardless of the model chosen. These data suggested structural asymmetry to influence AP subunit communication but made evident that a simple Michaelis-Menten analysis did not suffice to distinguish between models 1 and 2.
According to Equations 5 and 6, the uncompetitive inhibition by L-Leu of a heterodimeric AP molecule is fundamentally different for an allosteric enzyme or a half-of-sites enzyme. To confirm that AP enzymes are classical allosteric but non- Structural Asymmetry in AP Heterodimers- Fig. 7 shows the well characterized electrophoretic migration on starch gel electrophoresis of the common F and S homodimeric allozymes of PLAP as well as the pattern for the heterozygous FS variant. The FS pattern displays a third band of activity corresponding to the heterodimeric F/S molecules, equidistant to the migration of the F and S homodimeric components. Resting JEG3 choriocarcinoma cells express relatively low levels of a PLAP, but when cultured in the presence of butyrate, these cells express GCAP (30,31). Starch gel electrophoresis of the desialylated AP extracted from JEG3 cells grown in the presence of increasing butyrate concentrations (Fig. 7) indicated that the lower migrating GCAP isozyme progressively increased in intensity with increasing butyrate concentration, whereas the upper PLAP isozyme band gradually diminished to disappear entirely at 2.0 mM sodium butyrate. In addition, the intermediate PLAP-GCAP heterodimeric band is induced maximally at 0.5 mM sodium butyrate. We have shown previously that PLAP and GCAP differ conformationally mainly as a result of the E429G substitution, which also causes the retardation of GCAP in starch gel compared with the migration of PLAP (25). The fact that heterodimers migrate at a position that is not the exact intermediate between that of the PLAP and GCAP dimers indicates that the overall conformation of the PLAP-GCAP heterodimer resembles more closely that of GCAP than that of PLAP and provides further evidence for the existence of structural cross-talk between the asymmetrical monomers in the PLAP-[G]PLAP heterodimers.
We have previously shown that the E429G substitution affected the resistance of the resulting mutants toward inactivation by heat (25 (Fig. 8). We reported that a surface loop, made up of amino acids 400 -430 substantially contributed to the heat-stability of PLAP (17), because modifications in this loop dramatically reduced the resistance of the resulting PLAP mutant (PLAP-t) to denaturation by heat (PLAP-t homodimers were inactivated by 90% after 20 min at 56°C; 17) without affecting the kinetic parameters of the PLAP-t mutant. To investigate in more detail whether this loop participates in the structural cross-talk between both AP subunits, we have also analyzed the heat stability of the [S]PLAP-t-[RGR]PLAP heterodimers. It is evident that the heat inactivation pattern for the [S]PLAP-t-[RGR]PLAP heterodimers resembles that of [G]PLAP (Fig. 8), confirming that these heterodimers behave more as [G]PLAP, irrespective of the presence of the t-loop substitution in the [S]PLAP subunit. Thus, the amino acids 400 -430 loop controls the active site stability in PLAP but is not involved in any stabilizing cross-talk between AP subunits, corroborating our findings that the E429G substitution in PLAP is associated with structural changes in this loop, facilitating the access for L-Leu (25) and EDTA.
Conclusions-Mammalian APs are allosteric enzymes in which both monomers act independently, at least when both AP subunits are completely metalated. It is, however, clear that for different AP isozymes, subtle amino acid substitutions in positions close to the active site may dramatically affect the affinity for Zn 2ϩ binding in the active site pocket. Therefore, in different tissues the mechanism of the actual AP catalysis will be determined by the local concentrations of available isozyme and zinc ions. It is also evident that heterodimers can form between structurally related mammalian APs. These heterodimers are not the weighted average of the parent homodimers; as a consequence of subunit interactions, AP enzymes are formed that are structurally less asymmetrical than expected and that have catalytic properties divergent from those of the parent homodimers. Hence, mammalian APs are noncooperative allosteric enzymes, but the stability and catalytic properties of each monomer are controlled by the conformation of the second AP subunit.