The role of the allosteric B site in the fumarase reaction

Abstract

The role of a malate binding site in a concavity external to the more deeply situated active site has been a major mystery of the fumarase reaction. The malate, within 12 Å of the active site, was bound by hydrogen bonds to two main-chain amides and to two basic residues, H129 and R126. Mutation of the His of this so-called B site of Escherichia coli fumarase had little effect on the overall initial rate kinetics of the enzyme, which has obscured an understanding of the critical role of the site. Contrary to the WT enzyme, which is rate-limited in the recycling of free enzyme isoforms that follows product release, the enzyme with both basic residues modified is rate-limited in the product release step itself. A loss of complexity in the mutated, but still functional, step is indicated by a greatly reduced sensitivity of its rate to changes in temperature. Unlike the inhibition by glycerol shown with normal enzyme and attributed to a viscogenic effect on the recycling rate, the product-release step of the B-site mutants is accelerated by glycerol, suggestive of a structural effect on the 12-Å space between the A and B sites. It is proposed that the “extra” malate represents a stage in the transfer of substrate and product between the solvent and the “buried” active site of the enzyme.

The role of the so-called B site of Escherichia coli fumarase isozyme FumC has been a matter of speculation since an unexpected molecule of malate (M) was detected in crystals containing citrate at the active or A site (1). The A and B sites are only 12 Å apart. It has been suggested that an allosteric site, possibly the B site, could be responsible for the frequently observed property of activation of fumarase by high concentrations of substrates (2–4). However, the importance of substrate activation, reported to occur in many if not all fumarases of this class, would seem very problematic, given that the reaction catalyzed by fumarase is readily reversible, with a very high turnover rate, and is not found at the beginning of a multienzyme sequence. In addition, activation of fumarase by a variety of common anions (4, 5) would obscure activation by substrate in a physiological setting. Therefore, another role for the B site, perhaps related to the operation of the enzyme, has seemed a more likely possibility.

This article will conclude that the B site should be considered a functioning part of the active site, serving as a stage in the transfer of substrate and product between the bulk solvent and the active site. In the “fumarase fold” of this homotetrameric enzyme, different segments of three identical 50-kDa subunits contribute in forming any one of the four A/B sites, each of which is found within an ≈20-Å-deep concavity of the enzyme's surface, each with a volume of 2,429 ų (6). Of the 15 charged residues that make up the lining of each A/B- containing cavity, 10 are negatively charged, and of these, 7 are completely conserved among the known fumarases. Of the five basic residues only one is conserved. Thus, the “lining” of the cavity appears to be negatively charged. The positively charged B site, defined by the allosteric M, is formed by a stretch from R126, K127, V128, H129, P130, and N131 to D132 that comes from the same subunit that contributes S98, T100, and N141 to the A site. In part, the B site is built from a single turn of π-helix placed between the A site and the bulk solvent. The B site contacts the allosteric M by two hydrogen bonds between its C1 carboxyl oxygens and the main-chain –NH of residues N131 and D132, both of which are totally conserved, and by ionic bonds between R126 and H129 and the two oxygens of the C4 carboxyl of the M. The H129 equivalent position is retained in all of the fumarase members. Lys, rather than Arg, is the residue preferred at or next to the R126 equivalent position. N135 provides the linkage between the A and B sites through a series of hydrogen bonds as illustrated in Fig. 1.

Fig. 1.

A and B sites within FumC. The concavity opening to the A and B sites is illustrated in ball-and-stick fashion. The A site, identified by the location of both the citrate and water 26 (W-26) molecules, lies at the back of the opening ≈20 Å from the enzyme surface. The B site is located ≈12 Å from the A site and identified by the position of the bound M. Dashed lines in the drawing denote hydrogen bonds. Each residue has been labeled to identify the subunit from which it is derived. Atoms are colored as follows: carbon (gray), nitrogen (dark gray), and oxygen (black). The illustration was made with the chimera package from the Computer Graphics Laboratory, University of California, San Francisco [supported by National Institutes of Health Grant P41 RR-01081 (www.cgl.ucsf.edu/chimera)].

By using steady-state kinetic methods with yeast fumarase (4, 7), the stages in the catalytic cycle for the dehydration of M have been identified (Scheme 1). A M-specific form of the enzyme in its acidic-protonated form, , combines with M, which is then dehydrated in a two-step process in which a basic residue first abstracts the C3 proton to form an -carbanion intermediate, which then uses the acidic proton of the enzyme to abstract the C2-OH to form H₂O and E_H·F. The F (fumarate) may dissociate immediately from the product complex, giving an F-specific isoform, E_Hf, or through parallel recycling steps, may dissociate with the formation of the nonspecific isoform, (E_f bypass, the path taken by E·F to E_mf·F when M is activating). In both cases the acidic proton is returned to the enzyme from the solvent at a rate that is 2- to 3-fold slower in D₂O. This step is in series with a glycerol-sensitive conformational change. Both the D₂O and glycerol effects are V_max effects, indicating that recycling is rate-limiting for the overall catalytic cycle. That the C3 proton of M is lost primarily in the step is indicated by two observations: (i) tritium of [3-³H]M undergoes almost complete transfer to F if the latter is present during the dehydration, and (ii) the extent of this transfer is competitive with the added M (8), showing that the tritium came from a nonspecific intermediate. Specific intermediates are recognized and estimated by substrate analogs that are noncompetitive inhibitors. Thus, the K_ii (the concentration of a noncompetitive inhibitor required to decrease V_max by 2-fold) of S-2,3-dicarboxyaziridine (DCAZ) is inversely proportional to E_f in the M → F direction (7) and the K_ii of meso-tartrate is a measure of E_m in the F → M direction (4).

Scheme 1.

Little is known of the kinetic properties of E. coli FumC. Of special interest is the observation made in the present studies that, unlike the enzymes from yeast and heart, which are activated by both M and F (3), FumC shows activation by M, but not by F. Thus, if the B site is responsible for this activation, it is not acting as a general control switch. Differences in “control” properties in the two directions can be accommodated if, as in yeast fumarase, M and F were to activate by interaction with different enzyme isoforms, E_m, E_mf, and E_f. The three isoforms differ in their conformations and patterns of protonation, reflecting the changes required in the chemical phase of the cycle (7). The different isoforms are recognized by substrate analogs and their pattern as kinetic inhibitors. Thus, the K_ii of DCAZ, specific for E_f, provides a measure of the amount of E_f present in the steady state when it is used as an inhibitor of the M → F reaction (7), and the K_ii of meso-tartrate in the F → M reaction is a measure of E_m (4).

Materials and Methods

Site-directed mutagenesis was carried out according to the protocol for the QuikChange Kit (Stratagene), using primers with pEFC001, which harbors a fumC gene sequence with an in-frame carboxyl-terminal five-His tag. Each mutant fumC gene was subjected to automated DNA sequencing to confirm the position of the replacement. Each plasmid was transformed by using a standard CaCl₂ method into JM105 for protein expression. Protein purification was by metal-chelate chromatography as described (1).

The kinetic experiments make use of the strong absorbance of F in the UV region (225–260 nm). With the moderate equilibrium constant of M/F = 4.4, and convenient K_m values, the kinetic properties can be readily analyzed in both directions. Data are reported in the conventional 1/v vs. 1/[S] manner. The reciprocal of the ordinate values indicates the absorbance change per minute obtained as an initial rate. All incubations were done at 25°C unless otherwise noted.

Results and Discussion

To investigate the role of the B site, the initial rate kinetics of the WT and the mutant fumarases were compared. H129N has already been reported to have a rate similar to that of WT (9), and this observation led to the conclusion that the B site was not part of the active site. When compared with WT in both directions over an extended substrate range, V_max and V_max/K_m values for the individual mutants (Table 1) were similar within an order of magnitude, suggesting, at first sight, that the B site is not required for normal function. Indeed, K127D behaves like WT in all parameters that have been investigated.

Table 1. Kinetic properties of FumC and B-site mutants.

	M → F		F → M
Enzyme	Vmax, s-1	Vmax/Km, M-1·s-1	Vmax, s-1	Vmax/Km, M-1·s-1
WT	129	4.3 × 105	60	6 × 105
K127D	180	—	68	7.5 × 105
R126A	30	3.3 × 104	22	5 × 104
H129N	41	6.0 × 105	60	6 × 105
R126A/H129N	15	7.5 × 103	9.6	1.7 × 104

All rates were determined in a temperature-controlled spectrophotometer in 20 mM Tris acetate, pH 7.5.

However, marked qualitative differences are seen between the other mutants and the WT, if their rates are compared in D₂O and H₂O at saturation levels of either substrate (Table 2). The inhibition by D₂O, arises in the recycling phase because of the slow conversion (7). The greatest loss of D₂O effect is seen with R126A/H129N, the double mutant (DM). R126A and H129N are about equal and intermediate in the loss of the D₂O effect, and K127D is like WT. Thus, the extent to which recycling, an A-site function, loses its rate-limiting control strongly depends on the status of the B site. Loss of each basic residue, used equally in binding one of the carboxyls of the allosteric M, defining the B site in the native enzyme, has contributed to some prior step becoming rate-limiting. The fact that the same phenomenon is seen in both directions, resulting from an effect on a non-A site, suggests that the new rate-limiting step reflects a role of the B site in the product release processes. If the mutations lower the efficiency of the product release steps, they themselves could eventually become rate-determining for V_max. A shift to rate-limiting product release as the rate-determining step is confirmed by determining the K_ii values of meso-tartrate and DCAZ of the DM (Table 3). Because the rate constant for the recycling of E_m in the F → M reaction should not be different in a B-site mutant the 160-fold increase in K_ii of meso-tartrate with the DM should be attributed to an ≈160-fold decrease in the rate of the off step producing M. Knowing from Table 1 that V_max = (k_off · k_c)/(k_off + k_c) is 6-fold decreased, one calculates the ratio of the off rate to the recycling rate, k_off/k_c to be ≈30/1 before mutation and 9/1 after the double mutation. Similar effects are calculated for the M → F direction with DCAZ as the inhibitor.

Table 2. D₂O effects on WT and mutants of the B Site.

	Vmax in H2O/Vmax in D2O
Enzyme	M → F	F → M
WT	2.33	2.57
K127D	2.10	3.00
R126A	1.40	1.60
H129N	1.49	1.81
DM	1.06	1.24

Comparisons were made (at 25°C in 10 mM Tris chloride/50 mM sodium acetate, pH 7.7) in H₂O and D₂O (95%) at 20 mM M or 4 mM F.

Table 3.

Changes to E_m and E_f in the steady state of B-site mutants

	F → M			M → F
Enzyme	meso-Tartrate, mM	Kii	Em(rel)	DCAZ, μM	Kii	Ef(rel)
WT	1.0	0.6	1	80	19	1
H129N	1.0	5.0	0.12	—	—	—
DM	100	100	0.006	140	1,900	0.01

K_ii values for meso-tartrate were measured in the F → M direction, pH 7.5; F varied from 0.2 to 4 mM. K_ii values for DCAZ were measured in the M → F direction, pH 7.5; M varied from 0.3 to 3 mM.

Next, we characterized the product-off steps of the DM in each direction of reaction. The kinetic structure of an enzyme reaction sequence may be examined by determining the effect of change in temperature on the intercepts and slopes of the usual double-reciprocal plot. The use of temperature change (ΔT) has the advantage of not having to otherwise change the conditions of the reaction as with the use of D₂O, glycerol, or other reagents that may have unexpected side effects. On the other hand, all steps are expected to increase with temperature, but only the rate-limiting steps should influence the rate of the overall reaction. Intercepts normally increase ≈2-fold/10°C, because of the enthalpy contribution to the activation energy of the rate-limiting step at substrate saturation. Values of V_max/K_m that are not sensitive to ΔT imply that the recycling of ligand-free isoforms E_m or E_f, formed after the release of reaction product is rate-limiting. [This rule follows from Cleland's explanation of Ping-Pong kinetics (10) that sequences that follow the first irreversible step should not affect the slope parameter, only the intercept.] Converging linear slopes should be found when V_max/K_m of the enzyme is determined by steps up to and including product release, but do not distinguish between the reaction itself and the off step as rate-limiting. There may be an unusually low effect of ΔT on V_max if the rate-limiting step is a process with a very low activation energy such as simple diffusion or rupture of particular hydrogen bonds that are replaced from solvent.

Fig. 2 shows ΔT effects on V_max and V_max/K_m for K127D (same as WT) and DM in the F → M reaction. Rate-limiting recycling, when the reaction sequence and the M-off steps are rapid relative to recycling, is shown by the parallel slopes with WT. With the DM, Fig. 2B shows an unusually low ΔT, ≈10% increase in V_max in going from 20°C to 35°C, compared with a 190% increase with WT. (The ≈10% value is the average of six rates each at 20°C and 35°C, at >10 times the K_m of F, using Mops buffer for its temperature invariance of pH.) The low activation energy of V_max makes it unlikely that release of product occurs directly from the immediate product complex E_m·M in the DM, or that in the M → F direction, M-on occurs at E_m. Without both R126 and H129, the hydrogen bonds observed (1) between C1 carboxyl of the allosteric M, the main chain of the B site (Fig. 1) may be sufficient to constitute a transient intermediate with only entropic factors in the transition state to dissociation.

Fig. 2.

ΔT of F → M with WT and DM. (A) K127D (like WT) is rate-limited in recycling (in 20 mM Tris acetate, pH 7.5), ^Δ20°V_max = 1.9 and ^Δ20°Slope = 1.0. The units for v are ΔA/min. (B) DM is about the same at 20°C and 35°C.

As shown in Fig. 3A, WT FumC is also rate-limited in the recycling sequence in the M → F direction (i.e., reaction and F-off are not rate-limiting). As indicated in Table 2, this is not the case for DM. With the DM, the ΔT effect on V_max of 180%/20°C indicates that the new rate-limiting step is not of the low activation energy type seen in the F → M direction. A significant ΔT effect remains on V_max/K_m, ≈160%/20°C, requiring that there are steps with normal energies of activation that precede or include the release of F. As shown in Scheme 1, there are two ways to reach free F from E_H·F. Disruption of the B site in the DM may cause a low ΔT path to be invisible as in Fig. 2B. In the second path, , F may be more tightly bound in the complex from which it is slowly liberated from the DM. This difference can be explained by the fact that the production of F from M → F does not enter the B site.

Fig. 3.

ΔT of WT and DM in M → F. (A) WT, 20 mM Tris acetate, pH 7.56, ^Δ15°V_max = 1.75 and ^Δ15°Slope = 1.0. (B) DM, 10 mM Tris acetate + 50 mM sodium acetate, pH 7.8, ^Δ20°V_max = 1.8 and ^Δ20°Slope = 1.6.

A second consequence of modifying the B site is seen in the effect of glycerol on the rate-limiting step. Glycerol at 25% inhibits V_max severalfold without affecting V_max/K_m of the WT enzyme (Fig. 4A), consistent with rate-limiting recycling. On the other hand, glycerol activates the DM enzyme severalfold, depending on substrate concentration (Fig. 4 B and C) and pH. The effect on slope is consistent with activation by glycerol of the product-off step of the mutant enzyme. Glycerol has little or no effect on V_max with M or F as substrate. High concentrations of substrate could occupy the B site in competition with the generated bound product, specificity allowing. Product would then remain bound during recycling until an isoform that did not have a conventional B-site requirement was reached, perhaps E_mf. The basis for the activating effect of high glycerol remains to be determined (11).

Fig. 4.

Effect of glycerol on WT and DM rates. (A) WT in 10 mM Tris chloride, pH 7.2, and F from 0.025 to 0.15 mM. (B) DM in 10 mM 2-amino-2-methyl-1-propanol, pH 9.0, and F from 0.2 to 4 mM. (C) DM in 20 mM Tris acetate + 50 mM sodium acetate, pH 7.5, and M from 5 to 40 mM.

Mutation of R126 and H129 of the B site fundamentally changes the rate structure of the fumarase reaction. The nature of the change is to greatly decrease the product-release step, making it the rate-limiting step of the overall reaction. The properties of the modified B site suggest that the B site is an extension of the active site. Its functions are to promote the reaction product out of the 20-Å concavity to the enzyme surface and to conduct the substrate into the active site from the solvent. The lining of the cavity is highly negatively charged, and it would prevent interactions with the substrate and leaving product, both of which would be focused toward the very positive B site. Based on the original x-ray data of fumarase with citrate at the A site (1), the B site engages M by hydrogen bonds to main-chain amides of N131 and D132 and electrostatically to basic residues, R126 and H129, perhaps acting as a third rail. Loss of the third rail in the DM decreases the efficiency of transport, which now has a very low enthalpy of activation for the release step when M is the product, probably a consequence of the two hydrogen bonds that must be broken. F is also transported by a glycerol-activated carrier, with activation that is also lost at high substrate. A significant thermal effect on its rate of dissociation indicates a tighter binding of F to the B site of the DM enzyme. R126 and H129 contribute to binding of F in the native enzyme as shown by the changes in the D₂O effect in the mutants when assayed in M → F (Table 2).

Many questions remain about the structure and operation of the apparatus for the transfer of reactants between the active site of fumarase and the bulk solvent. Clearly, the apparatus works differently in the two reaction directions, suggesting that the behavior of the structure may be influenced by aspects of the A site, which was assumed to be the basis for specificity in the binding of M and F. The structure reported of allosteric M with citrate on the A site (1) corresponds best to conditions when M is the product. Leaving only two hydrogen bonds to bind M in the DM seems consistent with the very low temperature effect on the F → M reaction. The position of the allosteric M, only about half-way toward solvent, divides the product transit route into two compartments. How is the ≈160-fold decrease in product off rate in the DM distributed between the two compartments? Which one is influenced by glycerol? A unique B site for F as product having different on and off routes would provide the easiest explanation for the absence of inhibition at high substrate concentrations, a problem that seemed resolved when specificity could be ascribed to specific A-site isoforms. Why are the sequences of the A/B pocket so highly conserved, and perhaps, finally, why did this kind of system survive during evolution, and how extensive is it in nature?