Engineering homooligomeric proteins to detect weak intersite allosteric communication: Aminotransferases, a case study


  • Edgar Deu,

    1. Department of Pathology, Stanford University School of Medicine, Stanford, California 94305-5324
    Search for more papers by this author
  • Jack F. Kirsch

    Corresponding author
    1. Department of Molecular and Cell Biology, University of California at Berkeley, Berkeley, California 94720-3206
    • Division of Biochemistry and Molecular Biology, Department of Molecular and Cell Biology, 572 Stanley Hall #3220, Berkeley, CA 94720-3220
    Search for more papers by this author


The existence of low levels of intersubunit communication in homooligomeric enzymes is often difficult to discover, as the identical active sites cannot be probed individually to dissect their interdependent contributions. The homodimeric paralogs, E. coli aspartate- (AATase) and tyrosine aminotransferase (TATase), have not been demonstrated to show allostery. To address this question, we engineered a hybrid aminotransferase containing two distinct catalytic pockets: an AATase and a TATase site. The TATase/AATase hybrid was constructed by grafting an engineered TATase active site into one of the catalytic pockets of E. coli AATase. Each active site conserves its specific catalytic and inhibitor binding properties, and the hybrid catalyzes simultaneously each aminotransferase reaction at the respective site. Importantly, association of a selective inhibitor into one of the catalytic pockets decreases the activity of the second active site by up to 25%, thus proving unequivocally the existence of allosteric communication between active sites. The procedure may be applicable to other homologous sets of enzymes.


Family Iα aminotransferases are homodimers with two pyridoxal 5′-phosphate- (PLP-) dependent active sites situated at the dimer interface. They catalyze the reversible transfer of an amino group from an amino acid to an α-ketoacid through a ping pong bi-bi mechanism.1 Residues essential for catalysis and substrate specificity are shared by both subunits. The active sites are identical and no communication between them has been demonstrated, except the strong negative cooperativity exhibited by the D222A E. coli aspartate aminotransferase (eAATase) variant.2 Is it possible to engineer an aminotransferase with the two active sites having a different activity?

AATases and tyrosine aminotransferases (TATases) are the best characterized members of this family. The former have very narrow substrate specificity and only accept L-aspartate (Asp) and L-glutamate (Glu) as the amino donors [Eq. (1)]. The latter have broader specificities and catalyze the reactions of both dicarboxylic [Eq. (1)] and aromatic amino acids [Eq. (2)].

equation image(1)
equation image(2)

where, KG: α-ketoglutarate, OAA: oxaloacetate, PP: phenylpyruvate, and HPP: p-hydroxyphenylpyruvate.

The residue determinants of substrate specificity in the E. coli enzymes (eAATase and eTATase) have been thoroughly investigated.3–6 Mutagenesis studies have shown that the specificity of these enzymes can be modulated through the modification of relatively few residues.7–11 For example, the HEX7 and SRHEPT8 mutants, which behave as TATases in terms of substrate specificity (ψ-TATase active sites), were obtained from rational redesign and directed evolution of eAATase, respectively. As their names indicate, they differ from eAATase by only six and seven point mutations per monomer. The HEX variant, which does not complement the E. coli TATase activity in vivo, was improved by directed evolution into an enzyme (HEX + A293D) that complements that function with only the single additional mutation.9 Moreover, A293D enhances the specificities of HEX9 and SRHEPT12; it makes them more TATase like. SRHEPT + A293D is a very efficient enzyme whose substrate specificity is within the range of those observed for natural TATases13 (Kathryn Muratore, PhD thesis, UC Berkeley), and it is very similar to that of eTATase.12

The accumulated knowledge on aminotransferase specificity was used here to design an enzyme in which one active site would behave as an AATase and the other as a TATase. The option of using natural AATase and TATase monomers as the constituents of this hybrid enzyme was discarded due to structural constraints: first, although most of the active site is nested within one monomer, some essential residues are provided by the opposite one. Second, the interface, which differs considerably between AATase and TATase, is a crucial governer of their global stability.14–16 Thus it is very unlikely that an AATase and a TATase monomer could form a stable heterodimer.

A priori, grafting an engineered ψ-TATase active site into the eAATase scaffold was deemed more likely to succeed in heterodimer formation. SRHEPT + A293D (OCT) is the best ψ-TATase constructed from eAATase, and was chosen for the construction of the ψ-TATase/AATase hybrid, OCT/WT. It differs from eAATase in only eight out of the > 800 residues, and will be heterodimeric because those mutations must be distributed between the two subunits (Scheme 1).

Scheme 1.

Substrate specificity and inhibitor selectivity of OCT/WT. The two subunits of the ψ-TATase/AATase hybrid are represented by the empty and dotted areas. The left and right inter-monomer cavities represent the SRHEPT + A293D and eAATase intersubunit active sites, respectively. Eight residues distributed between the two monomers (indicated by asterisks) differentiate the two catalytic pockets. The preferred affinities for amino acid substrates (upper structures) and inhibitors (lower compounds) at each site are indicated by arrows. A C-terminal Glu4-tag is present in the OCT5WT3 monomer to facilitate purification (Scheme 2).

The OCT/WT construct should allow a sensitive probe of intra sites allosterism. For example, if each binding pocket in the heterodimer retains the same specificity as it has in the respective homodimeric context, then each should preferentially associate with a specific competitive inhibitor: maleate (Mal) in the eAATase active site7 and hydrocinnamate (Hca) in the SRHEPT+A293D active site (Scheme 1).12 In that asymmetric scenario, long range communication between active sites should be easier to detect than in a homodimeric context, where it is not possible to perturb a single active site selectively.


ψ-TATase/AATase hybrid design

OCT/WT was designed to contain one SRHEPT + A293D and one native eAATase active site. The former differs from the latter by the eight mutations: A12T, P13T, N34S, T109S, G261A, S285G*, A293D*, and N297S* (asterisks indicate amino acids from the opposite subunit). To form the OCT/WT heterodimer one subunit must contain the first five mutations (OCT5/WT3) and the complementary monomer the remaining three (OCT3/WT5) (Scheme 1). Since these enzymes are expressed as homodimers, the subunits of OCT5/WT3 and OCT3/WT5 homodimers must be exchanged to form the OCT/WT heterodimer (Scheme 2).

Scheme 2.

Heterodimer formation and resolution. Purified OCT5/WT3 and OCT3/WT5 homodimers were incubated in 1M GdmHCl to enable rapid subunit exchange and formation of OCT/WT (left panel). The resulting hetero- and homodimeric species were separated by ion exchange chromatography. The Glu4-tag at the C-terminus of the OCT5/WT3 monomer adds 8 and 4 negative charges to OCT3/WT5 and OCT/WT, respectively. Chemical linkage of the cofactor to one of the homodimers with NaCNBH3 yields an inactive form of the active site (indicated by a cross) and leads to the formation of the half-reduced hybrids rOCT/WT (middle panel) and OCT/rWT (right panel) upon subunit exchange.

The tetra-glutamate tail (Glu4-tag) appended to the C-terminus of OCT5/WT3 facilitates separation of the heterodimer from the homodimers. The OCT/WT hybrid bears four extra negative charges with respect to OCT3/WT5 and four extra positive charges with respect to OCT5/WT3 (Scheme 2). The enzyme variants can thus be resolved by ion exchange chromatography.

Heterodimer formation and resolution

The subunits of OCT5/WT3 and OCT3/WT5 homodimers were exchanged as described above. The extent of hybrid formation was analyzed by native gel electrophoresis [Fig. 1(A,B)]. Three charge-differentiated bands are observed in the gel and correspond to OCT3/WT5, OCT/WT, and OCT5/WT3.

Figure 1.

Resolution of heterodimeric aminotransferases. A: Ion exchange chromatography elution profile of the OCT/WT construct. Protein concentration was monitored by A280. The three peaks correspond to the two homodimers and the OCT/WT heterodimer (middle). B: Native gel illustrating the charge difference between the two homodimers (OCT3/WT5 and OCT5/WT3) and the OCT/WT heterodimer. Three distinct species are observed following subunit exchange at 1M GdmHCl (OCT/WT Mix lane). The three lanes on the right correspond to the eluted peaks shown in panel A. C: Native gel electrophoresis of half-reduced hybrids. Bands corresponding to the protein mixture obtained after subunit exchange at 1M GdmHCl are labeled Mix. The middle bands correspond to the heterodimeric species. Resolution of the hybrid from the homodimers is evident in the bands corresponding to rOCT/WT and OCT/rWT.

Table I. Steady-State Kinetic and Inhibition Models for Aminotransferase Reactions
Inhibition modelSchemeaEquationbEq. no.
  • a

    Scheme legend: AA, amino acid; KA, keto acid product; I, inhibitor; Subindeces: S, substrate; P, product; C, competitive; AS, alternate substrate; AP, alternate product; and NC, non-competitive.

  • b

    Equation legend: ET is the total concentration of monomer, and β represents the factor by which kcat is decreased at saturating inhibitor concentrations.

  • c

    The alternate amino acid substrates of the AATase and TATase reactions are Phe and Asp, respectively.

No inhibitionThumbnail image of
equation image
CompetitiveThumbnail image of
equation image
Alternative substratecThumbnail image of
equation image
equation image
Partial non- competitiveThumbnail image of
equation image

The concentration of OCT5/WT3 is more depleted than that of OCT3/WT5 after subunit exchange, which indicates that the former is particularly unstable in the presence of guanidinium hydrochloride (GdmHCl) (a precipitate was observed at 1M GdmHCl). This also explains why the OCT/WT hybrid is not the most populated specie as expected from a binomial distribution. However, there is no reason to expect the 1:2:1 ratio between OCT3/WT5:OCT/WT:OCT5/WT3, because each contains several mutations in regions that are crucial for stability, that is, the dimer interface and the N-terminal tail.13–15

The diethylaminoethyl (DEAE) column elution profile in Figure 1(A) shows the successful resolution of OCT/WT from its homodimeric parents. Dimer purity was confirmed by native gel electrophoresis [Fig. 1(B)]. The half-reduced forms of the hybrid were purified following the same protocol (Scheme 2). The extent of half-reduced hybrid formation and resolution are shown in the native gel of Figure 1(C). Interestingly, more rOCT5/WT3 was formed compared to OCT5/WT3 (see lower protein bands in OCT/WT mix and rOCT/WT mix lanes). This is likely due to the stabilizing effect of the reduced cofactor. The stability of OCT5/WT3 seems to be the limiting factor in hybrid formation.

Steady-state kinetics of an ψ-TATase/AATase hybrid

The AATase and TATase activities of OCT/WT, rOCT/WT, and OCT/rWT were measured at varying substrate concentrations. Two data sets were collected and fitted to a ping pong bi-bi mechanism (Table I) [Eq. (3)] for each enzyme form and reaction. The averages of the steady-state parameters are reported in Table II (fits are shown in Supporting Information Figs. 1–3). The previously determined constants for eAATase17 and SRHEPT + A293D12 homodimers are included in Table II.

Table II. Steady-State Kinetic Parameters of Variant Aminotransferases
 Enzymekcat (s−1)Kmath image (mM)Kmath image (mM)kcat/Kmath image (mM s−1)kcat/Kmath image (mM s−1)equation image
  • N.S. No saturation. N.D. Not determined.

  • a


  • b


  • c

    These parameters were calculated from the sum of the two half-reduced form values, and are those expected in the absence of communication between active sites.

  • d


AATaseeAATasea159 ± 21.75 ± 0.040.48 ± 0.0191 ± 10332 ± 50.0013 ± 0.0002
rOCT/WT70 ± 103.7 ± 0.31.5 ± 0.518.8 ± 1.435 ± 50.0023 ± 0.0003
OCT/WT84 ± 53.2 ± 0.60.72 ± 0.0227 ± 5116 ± 30.5 ± 0.1
OCT/rWT13 ± 0.220 ± 32.1 ± 0.30.6 ± 0.16.3 ± 0.715 ± 3
SRHEPT+A293Db43 ± 319 ± 22.6 ± 0.32.2 ± 0.216 ± 215 ± 3
OCT/WT Expectedc83 ± 1019.4 ± 1.541 ± 60.45 ± 0.1
 Enzymekcat (s−1)Kmath image (mM)Kmath image (mM)kcat/Kmath image (mM s−1)kcat/Kmath image (mM s−1)
TATaseeAATasedN.S.N.S.N.S.0.119 ± 0.003N.D.
rOCT/WTN.S.N.S.N.S.0.043 ± 0.003N.D.
OCT/WT7.8 ± 10.64 ± 0.050.92 ± 0.0812.2 ± 0.89 ± 2
OCT/rWT5.2 ± 0.40.55 ± 0.030.92 ± 0.039.4 ± 15.7 ± 0.6
SRHEPT+A293Dd10.0 ± 0.60.30 ± 0.060.48 ± 0.0634 ± 521 ± 4
 OCT/WT expectedc5.2 ± 0.49.4 ± 1N.D.

AATases and TATases are generally classified according to their specificity factor, which is the ratio between kcat/Kmath image and kcat/Kmath image. While AATases exhibit very low TATase activity, TATases have broader specificity and catalyze both reactions with some preference for Phe over Asp as the amino donor. The specificity factor for OCT/WT (0.5 ± 0.1) indicates that this hybrid has no clear substrate preference.

Two explanations can be offered for the lack of specificity: (1) within the heterodimeric context both active sites broaden their substrate affinities, and/or (2) each active site behaves in the hybrid according to its homodimeric activity, that is, the eAATase active site as an AATase and the SRHEPT + A293D one as a ψ-TATase, thus presenting an averaged specificity. To study each of the active sites individually within an asymmetrical context (heterodimer), the aminotransferase activities of rOCT/WT and OCT/rWT were measured. The steady-state parameters are reported in Table II. In the OCT/WT hybrid each active site has the expected substrate specificity, the SRHEPT + A293D and eAATase active sites behave as a ψ-TATase and an AATase, respectively. They will be referred as the ψ-TATase and the AATase sites of the OCT/WT hybrid. However, some of the kinetic parameters are slightly different when compared with those measured for the corresponding homodimers. (Note that kcat and kcat/KM parameters reflect the activity of two active sites for fully active dimers but of a single one for the half-reduced enzymes).

The KM values for the AATase reaction of rOCT/WT and for the TATase reaction of OCT/rWT are 2- to 3-fold higher than expected, but the kcat values are within experimental error of those measured for the homodimers. Interestingly, the kcat value for the AATase reaction of OCT/rWT is 1.6-fold lower than that of SRHEPT + A293D but the KM values are identical. We believe that these differences can be explained by the presence of allosteric communication between active sites and the different substrate specificities of the ψ-TATase and the AATase sites (see Discussion section).

If no cooperativity exists between the dimeric active sites of a single enzyme, the kcat and kcat/KM parameters for a given reaction are expected to be equal to the sum of the individual active site parameters. At saturating concentrations of substrates, each active site functions at Vmax (= kcatPT), and the maximum reaction rate for the enzyme is equal to the sum of the kcat values multiplied by PT. At substrate concentrations ≪ KM, the velocity of the reaction = (kcat/KM) [S] PT. Therefore, the overall enzyme velocity will also be linearly dependent on substrate concentration by a factor equivalent to the sum of each active site kcat/KM. Lack of additivity in either of these parameters would indicate the presence of cooperativity between active sites.

The expected kcat and kcat/KM values for rOCT/WT and OCT/rWT in the absence of cooperativity are shown in Table II. Those are generally equivalent to the sum of the rOCT/WT and OCT/rWT parameters. In OCT/WT, kmath image/Kmath image and kmath image/Kmath image are 1.4- and 2.8-fold higher than expected but kmath image is the sum of the OCT/rWT and rOCT/WT values. Therefore, the higher kcat/KM values reflect a decrease in KM values with respect to the two half-reduced forms. In the case of the TATase reaction, a 1.5-fold increase in kcat was observed with no changes in the KM values with respect to OCT/rWT. These results indicate the possible existence of positive cooperativity. The allosteric communication between active sites was further investigated by selective inhibition.

Inhibition studies design

Mal and Hca are selective competitive inhibitors of eAATase7 and SRHEPT + A293D,12 respectively (Table III, last four rows). In the OCT/WT context, Mal is expected to bind preferentially to the AATase site and Hca to the ψ-TATase site (Scheme 1). Allosteric communication in the hybrid can be evaluated from a determination of the effect of an inhibitor at one site on the activity of the second. Each inhibitor a priori might affect the activity of both active sites simultaneously, one by direct competitive inhibition and the other allosterically. In order to decompose these two effects inhibition assays were performed on the half-reduced forms of OCT/WT.

Table III. Inhibition Constants for Aminotransferase Variants
EnzymeReactionInhibitorCompetitive Ki (mM)Non-competitive
Ki (mM)β
  • N.D. Not determined.

  • a

    The reported Ki value is the [Phe] where the velocity = (v0vfinal)/2.

  • b


  • c


rOCT/WTAATaseMal11 ± 2
Hca1.7 ± 0.70.86 ± 0.02
Phe0.9 ± 0.30.88 ± 0.01
OCT/rWTAATaseMal3.7 ± 20.88 ± 0.02
Hca1.02 ± 0.04
Phe0.56 ± 0.03
TATaseMal10 ± 30.81 ± 0.02
Hca1.04 ± 0.04
Asp19 ± 2
OCT/WTAATaseMal19 ± 2
Phe0.36 ± 0.16a
TATaseMal13 ± 40.73 ± 0.02
Hca0.7 ± 0.2
Asp4 ± 20.75 ± 0.03
eAATasebMal19 ± 1  
Hca0.6 ± 0.1  

Asp and Phe are alternative amino donor substrates of the ψ-TATase site (Table II) and therefore should decrease the TATase and AATase rates, respectively. Moreover, allosteric perturbation of the AATase site could be mediated through the binding of Phe at the ψ-TATase site.

Competitive inhibition

The inhibitory effect of 0–10 mM Hca on the aminotransferase activities of OCT/rWT and of 0–50 mM Mal on the rOCT/WT AATase activity were measured at 10 mM KG and 0.1–30 mM Asp or 0.03–10 mM Phe. The data are consistent with a competitive inhibition model (Table I), and were globally fitted to Eq. (4) (Supporting Information Figs. 4 and 5). Thus, Hca and Mal behave as competitive inhibitors of the ψ-TATase and AATase sites, respectively, in the half-reduced hybrids. Representative data are shown in the upper frames of Figure 2.

Figure 2.

Inhibition of OCT/rWT (left) and rOCT/WT (right). Representative data showing the decrease in aminotransferase activity due to added Mal (•), Hca (▴), Asp (♦), or Phe (▪). The amino acid substrate concentrations and the monitored activity (AATase or TATase) are given in each panel ([KG] = 10 mM). Each inhibition assay data set, in which both the inhibitor and the amino donor substrate concentrations were varied, was fitted to a competitive [Eq. (4)], alternate substrate [Eqs. (5) and (6)], or partial non-competitive [Eq. (7)] model (Table I). The lines were calculated from the numerical fits of each data set with the inhibition constants reported in Table III. Note that the upper middle graph has different X-axis scales for the Hca and Asp concentrations.

Asp and Phe are alternative substrates of OCT/rWT (Table II). As expected Asp and Phe compete as the amino donor substrate and inhibit the TATase and AATase reaction, respectively (Fig. 2, two upper left graphs and Supporting Information Figs. 4 and 5). The data were fitted to an alternative substrate model (Table I) [Eqs. (5) or (6)].

The competitive inhibition constants are reported in Table III. The Kmath image and Kmath image values are generally within 2-fold of the dissociation constants reported for eAATase and SRHEPT + A293D (last four rows in Table III) except for the Kmath image measured in the AATase reaction of OCT/rWT (3.7 ± 2 mM). The alternative substrate Kmath image and Kmath image values for OCT/rWT are within the experimental error of the KM values (Table II). These results demonstrate that the substrate and inhibitor binding preferences at each active site (Scheme 1) are maintained in the half-reduced forms of OCT/WT.

Non-competitive inhibition communicated between active sites

The above experiments show that Mal and Hca are competitive inhibitors of the AATase and ψ-TATase sites, respectively (Scheme 3, left panel). To test whether they allosterically perturb the second active site, the aminotransferase rates of OCT/rWT and rOCT/WT were measured in the presence of an inhibitor that preferentially binds at the catalytically inactive pocket, that is, Mal and Hca, respectively.

Scheme 3.

Inhibition models for the half-reduced forms of OCT/WT. Competitive inhibition (left panel) takes place when an inhibitor binds into the functional active site of a half-reduced enzyme. Phe and Asp act as competitive inhibitors of the AATase and TATase reactions of OCT/rWT, respectively, because they are alternate amino donor substrates of the ψ-TATase site. Partial non-competitive inhibition (right panel) is observed when a substrate or inhibitor binds into the inactive catalytic site (indicated with crosses) and allosterically induces a decrease in reactions catalyzed by the functional pocket. Note that Mal binds in the AATase site selectively, and Phe/Hca in the ψ-TATase one.

Small but consistently observed decreases in velocity (10–20%) were noted in these experiments (Fig. 2, lower graphs). The data were fitted to a partial non-competitive model (Table I) [Eq. (7)], and the resulting inhibition constants are shown in Table III (Supporting Information Figs. 4 and 5). The β parameter indicates the factor by which kcat is decreased at saturating inhibitor concentrations. Additionally, Phe was found to inhibit the activity of rOCT/WT in a partial non-competitive manner. The right panel of Scheme 3 illustrates this allosteric inhibition mechanism.

The Ki values obtained from the partial non-competitive and competitive models for each inhibitor are the same within experimental error and are consistent with the independently determined dissociation constant values for eAATase and SRHEPT + A293D. Additionally, Kmath image (Table II) is equal to the Kmath image values measured for rOCT/WT (Table III). Thus, the allosteric and competitive inhibition effects observed for the half-reduced forms of OCT/WT result from the same binding events: Mal and Hca associating with the AATase and ψ-TATase sites, respectively.

Inhibition of OCT/WT

As shown above, each inhibitor acts as a competitive inhibitor by binding in one active site and as a non-competitive one through allosteric inhibition of the second. Rigorous quantification of the inhibition of OCT/WT, in which both active sites are functional, is complex since both effects take place simultaneously and cannot be readily differentiated. It is nonetheless important to show that the observed allosteric inhibition is not an artifact due to the reduction of the PLP-Lys258 imine.

One of the inhibitory effects can be selectively favored under certain substrate concentration conditions. Only the ψ-TATase site of OCT/WT is able to turnover Phe (Table II). Thus, inhibition data for the TATase activity by Hca were fitted to a competitive model (Table I) [Eq. (4)] (Fig. 3, upper right and Supporting Information Fig. 6) and those resulting from association of Mal (measured at 1 mM KG and 10 mM Phe) to a partial non-competitive inhibition model [Eq. (7)] (Fig. 3, lower right).

Figure 3.

Inhibition of OCT/WT. Left and right plots show representative inhibitory effects observed in the AATase and TATase reactions, respectively, as a function of Mal (•), Hca (▪), Phe (▪), or Asp (♦) at the indicated substrate concentrations. The data were fitted to a competitive inhibitor [Eq. (4)] (upper graphs) or partial non-competitive [Eq. (7)] (lower right graph) model. Kmath image for the AATase reaction (lower left graph) was calculated as the concentration of Phe giving 50% inhibition change. The inhibition constants are reported in Table III.

At saturating concentrations of Phe and low KG concentration, association of Asp to the ψ-TATase site is negligible (Kmath imageKmath image). Under these conditions, partial non-competitive inhibition of the TATase reaction by 0-50 mM Asp is observed (Fig. 3, lower right).

The competitive effect dominates over the allosteric one for the Mal mediated inhibition of OCT/WT AATase activity for two reasons: first, this inhibitor binds only at the AATase site, which accounts for 80% of the AATase activity (Fig. 3, upper left and Supporting Information Fig. 6), and second, the β values (Table III) measured for OCT/rWT indicate that Mal inhibits <20% of the ψ-TATase site velocity via the allosteric mechanism, that is, only <4% of OCT/WT total AATase activity. Therefore, data for the inhibition of the hybrid AATase activity by Mal were fitted to a competitive inhibition model (Table I) [Eq. (4)] (Fig. 3, upper left). All of the inhibition constants thus obtained are consistent with those measured for the half-reduced forms (Table III).

Addition of 15 mM Phe results in a (10 ± 1)% decrease in AATase activity (measured at 20 mM KG and 50 mM Asp) (Fig. 3, lower left). This effect, due to the binding of Phe at the ψ-TATase site, has to be both competitive (Phe as an alternative substrate) and non-competitive simultaneously. A rigorous inhibition model should include the sum of the alternative substrate inhibitory effect in the ψ-TATase site and the partial non-competitive one in the AATase site [Eq. (6) + Eq. (7)]. However, the data set is too small to be fitted to such a complex model. An estimated Kmath image value was calculated as the concentration of Phe at which 50% of the inhibitory effect is observed (Table III). This value is consistent with Kmath image (Table II) and the other Kmath image values reported in Table III.


OCT/WT, an hybrid aminotransferase with dual specificity

The steady-state kinetic parameters of OCT/WT and of its two half-reduced forms prove that this heterodimeric enzyme contains an AATase and a ψ-TATase active site that function simultaneously. Each conserves the substrate specificity and catalytic efficiency of its respective homodimeric parent (eAATase and SRHEPT + A293D). The kmath image and kcat/KM values for OCT/WT are larger (Table II) than would be expected if both active sites function independently, that is, these parameters should be equal to the sum of those obtained for rOCT/WT and OCT/rWT. This result indicates weak positive cooperativity. The active sites of OCT/WT also maintain their inhibitor preferences: Mal and Hca competitively inhibit rOCT/WT and OCT/rWT, respectively, as is evidenced by the consistency between the competitive Kis measured here and the KD values reported for the eAATase and SRHEPT + A293D homodimers (Table III). Thus the ψ-TATase and AATase sites of OCT/WT maintain their expected catalytic properties and inhibitory selectivities.

Allosteric communication between active sites

Mal does not associate with the ψ-TATase site, nor does Hca bind to the AATase site. However, partial inhibition of OCT/rWT is observed with Mal and of rOCT/WT with Hca, and when the TATase activity of OCT/WT is monitored in the presence of Mal. This 12–25% decrease in activity is consistent with a partial non-competitive model. More importantly, the consistency in the measured Kis independent of the inhibition mechanisms proves that the competitive and allosteric effects originate from the same binding events, that is, Hca/Phe into the ψ-TATase site, and Mal into the AATase one.

Allostery is also observed in the absence of inhibitor. At saturating concentrations of Phe, the ψ-TATase site accepts almost exclusively the aromatic amino acid as the substrate. Under those conditions, inhibition of the TATase reaction by Asp can only be explained by the binding, and therefore turnover, of Asp at the AATase site. As mentioned above, the OCT/WT kmath image and kcat/KM parameters indicate the possibility of positive cooperativity between active sites because the value of these parameters exceeds that expected for a non-cooperative heterodimer (Table II). Therefore, allosteric communication not only takes place when each active site binds to a different substrate, but might also occur when both active sites catalyze the AATase reaction, or when the ψ-TATase site catalyzes the TATase reaction.

The last 40 years has witnessed several conflicting reports on whether other AATases are cooperative enzymes19–26; however the presence of allosteric communication between the active sites of eAATase has not been reported. This controversy has also arisen with other aminotransferases such as glutamate-1-semialdehyde aminotransferase for which crystal structures from Synechococcus27 and Bacillussubtilis28 show asymmetric and symmetric active sites, respectively. The present study demonstrates unequivocally communication between active sites of a class Iα aminotransferase enzyme during catalysis. However, the level of interdependence between the two active sites of OCT/WT is very weak compared to other homodimeric enzymes such as acyl-tRNA synthetases or thiamine diphosphate-dependent enzymes where the activity of both active sites is fully interdependent.29

The formation of heterodimers where one subunit has been inactivated or rendered more sensitive to inhibition has been previously used to study communication between active sites of homodimers.30–32 However, by modifying the substrate specificity of one of the active sites our method allows the investigator to measure allosteric effects between two active sites that are fully functional. The developed technology should be applicable to other homooligomeric enzymes for which it is possible to change the substrate specificity such as dehydrogenases33 or tRNA synthetases.34 For example, a single mutation in lactate dehydrogenase35 or five mutations in malate dehydrogenase36 are sufficient to invert the specificity of these two homooligomeric enzymes.

Putative model to explain the allosteric communication

Aminotransferases are homodimeric enzymes with two active sites located at the interface of the two subunits. Each is nested within the large domain of one monomer, but key catalytic residues are provided by the opposite subunit. The small domain, which is composed of Ct and Nt residues, rotates towards the active site upon substrate or inhibitor binding thus shifting the enzyme from an open into a closed conformation.37 Our partial non-competitive inhibition data shows that binding of an inhibitor into an active site, that is, locking that active site in the closed conformation, results in a slight decrease in the kcat of the second site. The decrease of kmath image for OCT/rWT with respect to SRHEPT + A293D can therefore be explained by the preferential binding of Asp to the reduced AATase site, which effectively locks that active site into the closed conformation. On the other hand, lack of activity in one active site seems to result in increased KM values at the second catalytic pocket. The lack of Phe turnover in the AATase site of OCT/rWT and OCT/WT might explain the increase in Kmath image values compared to SRHEPT + A293D, and the lack of Asp turnover at the ψ-TATase site of rOCT/WT might explain the higher Kmath image values with respect to WT-eAATase. Moreover, in OCT/WT the KM values for the AATase reaction are lower than those for rOCT/WT, which suggests that having the ψ-TATase site reacting with Asp increases the efficiency at the AATase site. However, association of an inhibitor into one active site does not decrease the KM values at the other since binding of Hca into the ψ-TATase site of rOCT/WT did not result in a decrease of Kmath image, and binding of Mal into the AAT site of OCT/rWT did not decrease Kmath image (data not shown). It remains unclear why the kmath image of OCT/WT is larger than that of OCT/rWT. Perhaps the low turnover rate of Phe at the AATase site is sufficient to make the ψ-TATase site more efficient. However, given the variability in the measurement of this parameter (Supporting Information Fig. 1 and Table II), it is possible that the value of kmath image might have been overestimated. Overall, our data suggest that locking one active site in the closed conformation decreases the catalytic rate of the second active site, and turnover of substrate at one active site seems to facilitate the binding of substrates at the second site.

It is hard to speculate on the structural mechanism that mediates the small allosteric effects observed in OCT/WT, especially because the active sites are distant from each other, and it is extremely difficult to predict long-range allosteric effects. However, we know that rotation of the small domain toward the active site of the same subunit shifts the enzyme from an open conformation to a closed one. The Nt-tail (residue 5–16), which is adjacent to the small domain, wraps around the opposite subunit and mediates important intersubunit interactions.14 Interestingly, some of the mutations in SRHEPT that are important to change the specificity of WT-eAATase into a ψ-TATase are located at the Nt-tail (A12T and P13T). It is therefore possible that small movements of the Nt-tail might be able to affect the activity of the opposite subunit. Also, the crystal structures of eAATase12, 37 and our stability studies14 show that the interface between the two large domains is formed by a large interconnected network of electrostatic interactions and hydrogen bonds. This might mediate the weak allosteric effects observed in this study.


The OCT/WT hybrid contains one AATase and one ψ-TATase active site. Each exhibits kinetic parameters that are very similar to those characterizing their homodimeric progenitors. The inhibitory effects observed with the use of selective inhibitors prove unequivocally that communication between family Iα aminotranferase catalytic pockets exists. Whether this effect is present in natural variants, or has been artificially introduced by the asymmetric context of OCT/WT is open to discussion. Nonetheless, this study illustrates a novel strategy to illuminate small allosteric effects that would normally escape detection. In terms of protein engineering, one of the two active sites of a dimer (eAATase) was transformed into a catalytic pocket with novel activity and the other into a regulatory site. The former catalyzes a new reaction (TATase activity) that can be regulated allosterically through the binding of a small molecule (Mal), which is not a natural inhibitor of this activity, at the second active site.


AA, amino acid; AATase, aspartate aminotransferase; DEAE, diethylaminoethyl; DTT, dithiothreitol; eAATase, Escherichia coli AATase; eTATase, Escherichia coli TATase; GdmHCl. guanidinium hydrochloride; Glu4-tag, 4-glutamate tag appended to the C-terminus of the enzyme; Hca, hydrocinnamate; HDH, D-2-hydroxyisocaproate dehydrogenase; HEX, V39L/K41Y/T47I/N69L/T109S/N297S eAATase hexamutant; His6-tag, 6-histidine tag appended to the C-terminus of the enzyme; HPP, p-hydroxyphenylpyruvate; KA, α-ketoacid; KG, α-ketoglutarate; Mal, maleate; MDH, malate dehydrogenase; Ni-NTA, nickel-nitrilotriacetate; OAA, oxaloacetate; OCT/WT, eAATase dimer containing the SRHEPT mutations plus A293D in one active site; OCT3/WT5, S285G/A293D/N297S eAATase triple mutant; OCT5/WT3, A12T/P13T/N34D/T109S/G261A eAATase-Glu4 pentamutant; OCT/rWT, OCT/WT with the eAATase catalytic site inactivated; PLP, pyridoxal 5′-phosphate; PP, phenylpyruvate; rOCT/WT, OCT/WT with the SRHEPT+A293D catalytic site inactivated; SRHEPT, A12T/P13T/N34D/T109S/G261A/S285G/N297S eAATase heptamutant; TAPS, N-[tris(hydroxymethyl)-methyl]-3-aminopropanesulfonic acid; Trizma, Tris(hydroxymethyl)aminomethane; ψ-TATase, an engineered AATase active site with 8 mutations that exhibits near WT TATase activity.

Materials and Methods

Cloning of OCT5/WT3 and OCT3/WT5

The former was constructed as follows: a Glu4-tag was inserted in the pKS + AAT plasmid upstream of the C-terminal 6-histidine tag (His6-tag) using Strategene's Quick-Change protocol. pKS + AAT + Glu4 and the pUC119 plasmid containing the open reading frame (ORF) of SRHEPT + A293D8 (pSRHEPT + A293D) were digested with BsmI and EcoRI. The restriction products were purified by gel electrophoresis, and the pSRHEPT + A293D fragment coding for residues 1–270 was ligated to the complementary fragment of pKS + AAT + Glu4 (Supporting Information Fig. 7) with T4 DNA ligase.

To obtain OCT3/WT5, pKS + AAT and pSRHEPT + A293D + A261G were cut at two sites with PstI. The latter plasmid was obtained by introducing the A261G mutation into pSRHEPT + A293D with Stratagene's Quick-Change protocol. The purified pSRHEPT + A293D + A261G fragment coding for residues 144–325 and the complementary one from pKS + AAT were ligated together (Supporting Information Fig. 7).

Purification and resolution of OCT/WT aminotransferase hybrids

The plasmids coding for OCT5/WT3 and OCT3/WT5 were expressed in MG204 electrocompetent cells (a gift from Ian Fotheringham, Nutrasweet corporation), and the homodimers purified by Ni-NTA affinity chromatography as previously described for WT.5 Resolution of the heterodimer from the homodimers was performed as previously described.2 30 μM each of OCT5/WT3 and OCT3/WT5 were incubated for 8 h in 1M GdmHCl, 20 mM potassium phosphate, 10 μM PLP, and 1 mM DTT, and dialyzed overnight against 0M denaturant at pH 7.5 and 4°C. The protein mixture was loaded into a DEAE-Sepharose Cl-6B column pre-equilibrated with 10 mM Trizma, 20 mM KCl, 1 mM DTT, and 10 μM PLP at pH 7.5. Proteins were eluted with a linear gradient of 20–350 mM KCl. Fractions containing protein were analyzed by native polyacrylomide gel electrophoresis. Heterodimers were stored at 4°C in 20 mM phosphate buffer, pH 7.2, 5% glycerol, 300 mM KCl, and 20 μM PLP. No subunit exchange was observed after two months under these storage conditions (data not shown). All activity and inhibition assays were performed within one month of heterodimer purification.

The half-reduced forms of the hybrid, rOCT/WT and OCT/rWT, in which one active site is catalytically inactive (SRHEPT + A293D and eAATase, respectively), were prepared in order to measure the activity of each catalytic pocket individually. The imine group linking the PLP cofactor to Lys258 was chemically reduced to an amine with NaCNBH3 in either OCT5/WT3 or OCT3/WT5 prior to subunit exchange and heterodimer resolution (Scheme 2). In the resulting hybrids (rOCT/WT and OCT/rWT, respectively) one of the active sites is catalytically dead but should conserve its binding properties. The reduction reaction was performed as previously reported.15

Steady-state kinetic assays

The AATase [Eq. (1)] and TATase [Eq. (2)] reactions were monitored with the maleate dehydrogenase (MDH)1, 3 and the D-2-hydroxyisocaproate dehydrogenase (HDH)38 coupled assays, respectively. Both coupling enzymes use NADH to reduce the α-ketoacid product of the aminotranferase reaction. Aminotransferase activity is thus proportional to the depletion rate of NADH, which was monitored at 340 nm with a Molecular Devices Spectra Max 190 plate-reader spectrophotometer. The assays were performed at pH 8 and 25°C in 200 mM TAPS-KOH, 140 mM KCl, 20 μM PLP, 25-75 nM enzyme, and 8 units/mL of coupling enzyme.12 The AATase rates were measured at 0.1–50 mM Asp and 0.15–25 mM KG, and the TATase activity at 0.03–10 mM Phe and 0.03–10 mM KG. The steady-state parameters were computed from a matrix of initial rates of 8 × 8 different substrate concentrations. The initial velocity values were globally fitted by non-linear regression to a ping pong bi-bi reaction mechanism (Table 1) [Eq. (3)] with the SAS statistical package (SAS institute, Cary, NC).

Inhibition assays

Aminotransferase velocities were measured as described above but in the presence of 0–60 mM Mal or 0–10 mM Hca. Additionally, the effects of 0–10 mM Phe on the AATase activity, and of 0–50 mM Asp on the TATase rates, were also investigated. In all assays, the concentration of KG was kept constant (usually at 10 mM), and those of amino acid substrate and inhibitor varied. The resulting matrices of initial rates, 4–8 inhibitor concentrations × 6–8 substrate concentrations, were fitted to the appropriate inhibition model shown in Table I (KM values were fixed to those obtained from the steady-state kinetic assays in the absence of inhibitor [Eq. (3)]). Each data set was fitted by non-linear regression using the NLIN program of the SAS statistical package (SAS institute, Cary, NC).


Authors thank Dr. Keith A. Koch for the cloning of pKS + AAT and Dr. Steven C. Rothman for the cloning of pSRHEPT+A293D.