Spatial Clustering of Isozyme-specific Residues Reveals Unlikely Determinants of Isozyme Specificity in Fructose-1,6-bisphosphate Aldolase*

Vertebrate fructose-1,6-bisphosphate aldolase exists as three isozymes (A, B, and C) that demonstrate kinetic properties that are consistent with their physiological role and tissue-specific expression. The isozymes demonstrate specific substrate cleavage efficiencies along with differences in the ability to interact with other proteins; however, it is unknown how these differences are conferred. An alignment of 21 known vertebrate aldolase sequences was used to identify all of the amino acids that are specific to each isozyme, or isozyme-spe-cific residues (ISRs). The location of ISRs on the tertiary and quaternary structures of aldolase reveals that ISRs are found largely on the surface (24 out of 27) and are all outside of hydrogen bonding distance to any active site residue. Moreover, ISRs cluster into two patches on the surface of aldolase with one of these patches, the terminal surface patch, overlapping with the actin-binding site of aldolase A and overlapping an area of higher than average temperature factors derived from the x-ray crystal structures of the isozymes. The other patch, the distal surface patch, comprises an area with a different electrostatic surface potential when comparing isozymes. Despite their location distal to the active site, swapping ISRs between aldolase A and B by multiple site mutagenesis on recombinant

Vertebrate aldolases exist as three or more isozymes with different tissue distributions: aldolase A (expressed predominantly in muscle), aldolase B (expressed predominantly in liver), and aldolase C (expressed predominantly in brain) (1). The sequences of the mammalian isozymes are highly conserved, exhibiting 81% sequence identity between aldolases A and C (4). Aldolase B is slightly more divergent with ϳ70% sequence identity to both aldolases A and C (5). Consistent with this sequence similarity, isozymes A and C exhibit comparable kinetic properties. Aldolases A and C have evolved to perform the glycolytic reaction, Fru 1,6-P 2 cleavage, more efficiently than aldolase B as demonstrated by a 20 -30-fold higher k cat (6,7). Conversely, aldolase B cleaves Fru 1-P three times more efficiently than aldolase A or C (6). In addition, aldolases B and C demonstrate a 10-fold lower K m for glyceraldehyde 3-phosphate and dihydroxyacetone phosphate, implying that they have evolved to perform the gluconeogenic reaction, Fru 1,6-P 2 synthesis, more efficiently than aldolase A (8). Although these kinetic properties are consistent with their physiological roles and tissue-specific expression (9), the structural features that confer these isozyme-specific differences are unknown.
The three-dimensional structures of aldolase A (10) and B (11) have been determined and reveal homotetramers of very similar ␣/␤-barrels with the active site of each monomer located at the center of the ␣/␤-barrel. The 18 -20 residues at the C terminus of the protein are likely to be flexible as this portion is either missing from many structures (10,12) or found in a variety of conformations (11,13). The C-terminal region has been implicated in determining the isozyme-specific differences of aldolase (14,15). Mutations in this region diminish the kinetic distinctions among isozymes. For example, the swapping of two regions (amino acids 34 -55 and the C-terminal region 306 -363) of human aldolase A onto human aldolase B causes changes in the kinetic properties of aldolase B. Like aldolase A, the chimera had a 5-fold increase in the k cat and a 4-fold increase in K m for Fru 1,6-P 2 cleavage, as well as a 3-fold decrease in cleavage of Fru 1-P (15). Similarly, the naturally occurring mutation, A337V, which is located in the C-terminal portion of human aldolase B, is found in patients with hereditary fructose intolerance and results in a 43% loss in the ability to cleave Fru 1-P while maintaining wild-type cleavage rates of Fru 1,6-P 2 (16). In contrast, switching the C-terminal 6 residues between isozymes A and B produces no clear effects on steady-state kinetics (14).
Given that aldolase isozymes have evolved distinct kinetic properties, residues unique to each isozyme must be conserved. Such isozyme-specific residues (ISRs) for aldolases B and C have been identified previously (17,18). The ISRs were identified based on the criteria that the residue is conserved within orthologs and is distinct among paralogs. In this report, we combined this evolutionary information with the three-dimensional structures of aldolase A, B, and a model of C to identify structural features that may be involved in conferring isozymespecific kinetic preferences. To this end, an alignment of all six vertebrate aldolase A sequences was performed to identify aldolase A ISRs, along with a more extensive alignment of aldolase B and C sequences. All three sets of ISRs were mapped onto the conserved three-dimensional structure of the enzyme, and two patches on the surface of aldolase, where ISRs from all three isozymes cluster, were identified. The terminal surface patch (TSP) was associated with an area of relatively high flexibility as demonstrated by high x-ray crystallographic temperature factors. In addition, the F-actin-binding domain overlapped with this patch; thus, isozyme-specific surface moieties were correlated with macromolecular interactions.
ISR clusters on the surface of aldolase may confer kinetic specificity by influencing the active site directly via the Cterminal region or indirectly through conformational changes that alter the structure of the active site from a distance. To test this hypothesis, PCR-assisted multiple site-directed mutagenesis was used to incorporate all 12 aldolase B ISRs into the appropriate position of aldolase A while at the same time replacing the 11 aldolase A ISRs with the analogous residue of aldolase B. The resulting protein was expressed in Escherichia coli and purified. Kinetic constants were determined, and a direct correlation between kinetic constants and ISR type was found. This type of analysis of ISRs can easily be applied to other protein families to reveal functional residue clusters not apparent from primary sequence alignments.

EXPERIMENTAL PROCEDURES
Materials-Restriction endonucleases, T4 DNA ligase, exonuclease III, and DNA polymerase I were from New England Biolabs. Calf intestine alkaline phosphatase and glycerol-3-phosphate dehydrogenase/triose phosphate isomerase were from Roche Applied Science. Deoxynucleoside triphosphates, CM-Sepharose ® CL-6B, and Sephadex ® G-150 were from Amersham Biosciences. Oligonucleotides used for site-directed mutagenesis and sequencing were either synthesized on DNA synthesizers using phosphoramidite chemistry and the manufacturer's protocols or purchased from The Midland Certified Reagent Company, Inc. Ultrapure sucrose was from Invitrogen. Fru 1,6-P 2 , triethanolamine, and other chemicals were from Sigma.
Identification of Isozyme-specific Residues-The amino acid sequences of the following aldolase A enzymes were aligned: human (19), mouse (20), rabbit (21), rat (22), chicken, 2 and Xenopus (23). The alignment was generated using Clustal V (24). Amino acids that are unique to each isozyme and distinct from the analogous amino acid in paralogs were identified and defined as ISRs. This comparison included the addition of mouse B (25) and C (26), salmon B (27), and pufferfish C 3 to the previously identified ISRs (17,18). The presence of non-conservative (radical) and conservative differences between isozyme groups were evaluated using the three-dimensional structure of aldolase A (Protein Data Bank (PDB) accession code 1ADO) (13) and the graphics program O (28) to assess whether the ISR played a crucial role in stabilizing tertiary or quaternary structure. These determinations were based on the degree of structural alteration caused by each substitution, location of the residue in the three-dimensional structure of the aldolase tetramer, polarity, charge, size, and aromaticity (29). Conservative changes included residues that were located in a structural motif such as an ␣-helix or ␤-sheet and those residues that had side chains accessible to the solvent. In these cases, replacement with a residue of similar charge and size was unlikely to alter the overall structure of the protein.
Characterization of Mutant Aldolases-Activity of Fru 1,6-P 2 and Fru 1-P cleavage was measured as described previously (31) by the decrease in A 340 in an assay coupled to NADH oxidation by glycerol-3phosphate dehydrogenase (35). Sucrose density gradient sedimentation velocity was determined for the H156E mutant using a modified procedure of Martin and Ames (36) as described previously (37). Briefly, 0.1 mg of sample was loaded on 5-20% sucrose gradients at 20°C and sedimented for 14.5 h at 38,000 rpm in a SW60 rotor ( 2 t ϭ 8.31 ϫ 10 11 ). Circular dichroism spectra were collected between 185 and 260 nm on an Aviv 62DS spectrometer equipped with a temperature controller at 20°C as described previously (37). Irreversible thermal inactivation was performed as described previously (37) using 10-min incubations at temperatures between 30 and 55°C in 5°C increments, and the activity was measured at 30°C.
Construction of Aldolase C Homology Model-The model of aldolase C was derived from the high resolution structure of aldolase A (PDB accession code 1ADO) (13) using the Swiss-Model protein modeling server (38). Comparisons of amino acid and nucleotide sequences have shown that aldolase C and aldolase A are more closely related to each other than either is to aldolase B (4). The protein modeling server performs superposition of related tertiary structures, generation of a multiple sequence alignment with the sequence to be modeled, generation of a structural framework of the new sequence, rebuilding of missing loops, completion and correction of backbone and side-chain geometry, and refinement of the structure by energy minimization and molecular dynamics.
Three-dimensional Mapping of ISRs-The locations of ISRs were mapped to the tertiary and quaternary structure of aldolase A using the program O (28). The coordinates of aldolase A and B were taken from high resolution crystal structures (PDB accession codes 1ADO and 1QO5, respectively) (11,13), and those of aldolase C were generated from the homology model (see above). The electrostatic potential surfaces of aldolase A, B, and C were calculated and displayed using the program GRASP (39).

RESULTS AND DISCUSSION
Identification of Aldolase A ISRs-The amino acid sequences of 21 vertebrate aldolase isozymes were analyzed to identify individual residues specific to each isozyme. Multiple sequence alignment revealed a total of 11 aldolase A ISRs (Fig. 1). The alignment of these aldolase A ISRs was compared in order to further categorize these isozyme-specific differences into either conservative or non-conservative (radical) amino acid substitutions in an effort to identify which substitutions are more likely to confer functional changes to the isozyme. From this analysis, it was determined that among the aldolase A ISRs, Ile-39, His-156, Tyr-327, and Ser-353 were conservative differences and that Ser-45, Pro-71, Arg-91, Gly-102, Lys-311, Glu-354, and Ser-359 were non-conservative differences with respect to the other isozymes (Fig. 1). Due to the lack of a definitive structure, the ISRs in the C-terminal region, Ser-353, Glu-354, and Ser-359, were categorized without consideration of the three-dimensional location.
Aldolase A ISRs were compared with the aldolase B (17) and aldolase C ISRs (18). The inclusion of mouse and salmon aldolase B sequences in the aldolase B alignment revealed that the previously identified residues Ser-68 and Cys-268 are no longer conserved (Pro-68 and Thr-268 in salmon) and were thus excluded from the aldolase B ISRs. Similarly, the inclusion of the pufferfish aldolase C sequence revealed that the previously identified aldolase C ISR, Leu-57, was not conserved (Arg-57 in pufferfish), and therefore this residue was not included among the aldolase C ISRs. As a result, a total of 27 ISRs were identified: 11 aldolase A ISRs, 12 aldolase B ISRs, and 4 aldolase C ISRs (Fig. 1). Notably, the aldolase C ISRs were found within the last 49 residues of the protein sequence, whereas aldolase A-and B-specific residues were found throughout the primary sequence.
Locations of ISRs in the Three-dimensional Structure-The conservation of ISRs among orthologs implies that they play a role in conferring the specific kinetic properties of each isozyme. Thus, the expectation was that they would map largely at or near the active site. More specifically, the expectation was that they might reside within hydrogen bonding distance (2.6 -4.0 Å) of either Lys-229 or Asp-33, which are residues known to be involved in catalysis (12). The locations of all 27 ISRs in three-dimensional space were examined by mapping the residues onto the high resolution crystal structure of rabbit aldolase A (PDB accession code 1ADO) (13), which includes several conformations for the C terminus (residues 345-363). After analyzing the C termini of the four different monomer structures, the structure with the C terminus closest to the ␣/␤-barrel portion of the protein was chosen. Surprisingly, none of the ISRs were within hydrogen bonding distance of the active site as defined by the ⑀-amino group of the Schiff base Gly 3 Ile AB_All4 His 3 Gln AB_All7, AB_All8 forming Lys-229 and the carboxyl group of Asp-33 (Fig. 2a).
The closest residue to the active site was 8.2 Å from the catalytic Asp-33.
To determine whether the ISRs were involved in the subunitsubunit interactions that might stabilize the tetramer, the locations of ISRs were examined with regard to the tetramer structure. Only one ISR, His-156 of aldolase A, was located at the subunit interfaces and was near the 2-fold symmetry axis of the dimers. Its location implies that His-156 may play a role in stabilization of the tetramer. The role of this residue was examined by site-directed mutagenesis. His-156 was mutated to a glutamic acid in an attempt to reverse the point charge of this residue. The purified aldolase A H156E mutant enzyme demonstrated similar kinetics (Table II) and stability to the wildtype enzyme. The kinetic constants for recombinant wild-type aldolase A and B were obtained in the same fashion as for the mutant enzymes, and the kinetic constants agree with the published data (Table II). Comparison of the H156E far ultraviolet circular dichroism spectrum with that of wild-type aldolase A did not reveal any differences in secondary structure. Gel filtration analysis demonstrated a quaternary structure similar to wild type. In addition, thermal stability, as assessed by activity, was similar to that of wild-type enzyme (data not shown). Thus, there was no evidence supporting the stabilization of quaternary structure by ISR His-156.
Remarkably, the majority of ISRs (24 out of 27) were located on the surface of the protein and generally cluster in two patches, regardless of the particular isozyme (Fig. 2b). The other 3 ISRs, 58, 182, and 327, were buried in the ␣/␤-barrel portion of the protein and were not solvent-accessible. The first ISR surface patch, termed the distal surface patch (DSP), was related in location to the active site by a 180 o rotation of the monomer. The DSP was composed of the residues 60, 71, 91, 102, 320, 324, and 332, which were from different portions of the primary sequence. The second ISR patch included residues from the C-terminal region (residues 345-353) and those residues on the surface of the protein proximal to the C-terminal region. This patch was called the TSP and was composed of the residues 38, 39, 40, 45, 311, 314, 350, 353, 354, 358, 359, and 361. The ISRs 350, 353, 354, 358, 359, and 361 were assigned to the TSP based on the primary sequence since a definitive structure has not been determined for this region. The remaining surface residues, 113, 156, 262, and 296, were not located in either patch. Excluding the C-terminal region (residues 345-363), 9 of the ISRs identified here (38 -40, 45, 58, 60, 71, 91, and 102) were located in an area of higher root-mean-square deviation as compared with the average root-mean-square deviation of main chain atoms between the aldolase A and B structures (data not shown). The location of these residues in areas of structural divergence between the isozymes provided support for their role in conferring isozyme specificity.
ISRs Impact on Enzymatic Function-Although the majority of ISRs cluster on the surface of aldolase, one model is that these residues influence kinetic properties at a distance from the active site. Long distance effects have been documented in other enzymes, such as trypsin (40) and lactate dehydrogenase (41). The complete set of ISRs from one aldolase isozyme incorporated into the primary sequence of another isozyme would test this model and verify the importance of these ISRs. At the same time, the ISRs of the latter isozyme need to be removed. Specifically in this study, a mutant enzyme was engineered by swapping the ISRs from aldolase B into aldolase A, whereas at the same time exchanging the aldolase A ISRs with the analogous residue from aldolase B (Table I). Due to the fact that no structure is available for aldolase C and the kinetic differences are greater between aldolases A and B, the ISRs of only these two isozymes were swapped. To accomplish this, PCR-assisted multiple site-directed mutagenesis was used to engineer 32 nucleotide substitutions in an expression plasmid for aldolase A (Table I). It should be noted that even with the number of substitutions, the resulting enzyme, AB_All, named in reference to the swapping of all ISRs, was 93% identical to aldolase A and only 75% identical to aldolase B. After expression and purification of AB_All, the steady-state kinetics toward the substrates Fru 1,6-P 2 and Fru 1-P were determined using an enzyme-coupled assay (35). Comparing these data with data obtained for isozymes A and B revealed that both K m and k cat values toward Fru 1,6-P 2 decreased significantly from those of aldolase A and are nearly identical to that of aldolase B (Table  II). Similarly, the K m of AB_All toward Fru 1-P decreased, whereas the k cat increased appropriately to mimic the kinetic parameters of aldolase B (Table II). These results verify that the set of ISRs identified here is necessary and sufficient for determining isozyme-specific kinetic properties and suggest that some of these residues impact their effects at a distance from the active site. In addition, the fact that the k cat toward the substrate Fru 1-P is increased relative to the parent protein demonstrates that we have not simply generated a mutant with impaired function.
There are two ways in which these apparent kinetic effects could occur from a distance. First, these residues may act as structural modulators, modifying the structure of the active site from a distance through conformational effects as in trypsin (40) and lactate dehydrogenase (41). Second, the residues may act through direct contact with the active site via the mobile C-terminal region. For example, in the case of the TSP, a conformational change could position a portion of the TSP proximal to the active site. It has been hypothesized that the C-terminal region of aldolase is involved in catalysis through a possible conformational change coordinated with substrate cleavage (10,12,14). One piece of evidence for this involvement comes from previous studies involving a recombinant chimera of human aldolase B (15). The section of primary sequence exchanged between aldolase A and B included 17 ISRs identified in this study, encompassing all the TSP residues.
The prediction of functionally important residues in proteins has been investigated previously using more generalized meth-ods (42)(43)(44)(45). Recently, three-dimensional cluster analysis was used to identify functionally relevant residue clusters of aldolase (46). Patches of conserved residues, which could not be identified previously by sequence alignment and phylogenetic analysis, included the highly conserved active site and subunit interfaces along with a patch near the C-terminal region and another that overlapped with the actin-binding region. The C-terminal and actin-binding patches they identified overlap with the TSP discussed here. Mutations in the C-terminal portion implicate this region in conferring isozyme specificity (14 -16, 46, 47). However, an interpretation of the mechanism by which these residues confer specificity cannot be substantiated without the definitive location and functional assessment within this apparently flexible region.
Potential Role of ISRs in Protein-Protein Interactions-The binding region on aldolase A by F-actin has been identified by peptide mapping with a synthetic peptide corresponding to the aldolase sequence 32-52 (48) and by site-directed mutagenesis of binding residues Arg-42, Lys-107, and Arg-148 (49). This actin-binding site overlaps with the TSP, which includes ISRs 38, 39, 40, and 45. The residue Arg-42 is required for actin binding (49) and is located in the center of the TSP. This residue is conserved in all aldolase isozymes, which is consistent with the finding that all three isozymes bind to the cytoskeleton (50). Although all isozymes have the ability to bind to the cytoskeleton, the interaction is preferable for cytoskeleton derived from the same tissue from which the isozyme is expressed. Aldolase A binds most tightly to cytoskeleton isolated from muscle tissue, and aldolase B binds most tightly to  cytoskeleton isolated from liver tissue. Aldolase C, however, shows only slight preference for liver-and brain-isolated cytoskeleton (50). These results, together with the fact that the actin-binding site overlaps with the TSP, suggest that, although Arg-42 is required for binding actin, the surrounding ISRs confer preference for tissue-specific actin. The other ISRs that do not overlap with the F-actin-binding site are also located on the surface. This suggests that they may participate in tissue-specific protein-protein interactions. Orthologous conservation of these surface patches reveals potential sites of protein-protein interactions and implies that isozyme specificity is mediated through surface interactions. In addition to F-actin (49), the aldolase A binding partners include vacuolar H ϩ -ATPase (51) and GLUT4 (52). The interaction between aldolase A and F-actin or GLUT4 can be disrupted by substrates of aldolase, thus implicating structural connectivity of binding residues with residues of the active site (52). Aldolase B has been shown to interact with vacuolar H ϩ -ATPase (51) and the hepatocyte matrix (53). Aldolase C has demonstrated binding to inositol 1,4,5-trisphosphate (54) and S-100 A1 (55) protein, thus implicating aldolase C in calcium signaling. Such protein-protein interaction with proteins not involved in glucose metabolism suggests alternate cellular functions for aldolase, which may be isozyme-specific.
The method of mapping evolutionarily conserved residues on the tertiary structure reveals interesting structural attributes that might correlate with protein-protein interactions. A correlation between elevated temperature factors derived from x-ray crystal structures and sites of protein-protein interactions is apparent from studies of proteases and their inhibitors (56). Several proteases have demonstrated a preference for cleavage at relatively flexible areas of a protein substrate as defined by crystallographic temperature factors. In this study, support for the role of ISRs in protein-protein interactions is suggested by the relatively high temperature factors of these residues as compared with the other residues in the crystal structure. The average temperature factor for all residues in the aldolase A structure (13) is 21.6 Å 2 , whereas the average temperature factor for the ISRs is 37.7 Å 2 . Similarly, all residues in the aldolase B structure (11) have an average temperature factor of 40.0 Å 2 , whereas the ISRs average 50.9 Å 2 . The elevated temperature factor among ISRs indicates that they are located in an area of relative flexibility. Moreover, this flexibility cannot be attributed solely to the C-terminal region as the majority of ISRs not located in this region also demonstrate elevated temperature factors. Mapping of temperature factors on the threedimensional structure reveals that the TSP and the actinbinding site overlap with an area of flexibility in both aldolase A (Fig. 3a) and B (Fig. 3b) 3. The terminal ISR patch overlaps with an area of elevated temperature factor. In a, the surface model of the aldolase A monomer (PDB accession code 1ADO) (13) is shown, and those residues with an average temperature factor greater than 50 Å 2 are colored magenta. In b, the surface model of the aldolase B monomer (PDB accession code 1QO5) (11) is shown, and all residues with an average temperature factor greater than 60 Å 2 are colored magenta. Outlines of the ISR patches, as described in the legend for Fig. 2, are superimposed. These surface renderings were made with the program GRASP (39). , and aldolase C (c) are generated from 1ADO (13), 1QO5 (11), and an aldolase C homology model, respectively. The model of aldolase C was derived from the high resolution structure of aldolase A (PDB accession code 1ADO) (13) using the Swiss-Model protein modeling server (38). Outlines of the ISR patches, as described in the legend for Fig. 2, are superimposed. Note that the TSP of aldolase B is different from that of the aldolase A structure. Figures were created with GRASP (39) and colored based on surface potential in units of k with the scale (Ϫ5.8, 0, 7.3) in GRASP notation where negative values are red and positive values are blue. aldolase A and B structures, respectively.
If each isozyme interacts with specific proteins in tissues where they are expressed via ISR interactions, they might show different electrostatic surface potentials at these putative binding sites. To test this idea, three-dimensional atomic structures of aldolase A and B and a homology model of aldolase C were used to calculate the electrostatic surface potentials in the area of the mapped ISR patches (Fig. 4). The electrostatic surface of each isozyme in the DSP was indeed distinct between the three isozymes, whereas they were similar in the TSP. Other areas of the surface were also similar among the isozymes. Despite its higher overall sequence identity, the electrostatic surface of the DSP of aldolase C was the most negatively charged, in contrast to those of aldolase A or B (Fig. 4).
In this study, the information generated from the primary sequence alignment of aldolase isozymes is combined with the information generated from the three-dimensional structures of these isozymes to attribute functional roles for these conserved amino acids. This analysis revealed two surface patches containing the majority of the ISRs. These experiments confirm the computational prediction of isozyme-specific patches distinct from the active site (46). Swapping the ISRs between two isozymes via site-directed mutagenesis revealed that the isozyme specificity of aldolase is conferred either through conformational changes influencing the active site from a distance and/or causing the C-terminal region to interact directly with the active site. To explore these possible mechanisms, subsets of ISRs need to be swapped and subjected to the same analysis as described here to attribute a role for each conserved residue cluster. It is not likely that the entire set of ISRs is exclusively involved in influencing the kinetic differences between isozymes. Specific kinetic roles for subsets of these ISRs, if any, as well as possible roles in protein-protein interactions need to be investigated in subsequent studies.