Isozyme-specific modules on human aldolase A molecule. Isozyme group-specific sequences 1 and 4 are required for showing characteristics as aldolase A.

Vertebrate aldolase molecules bear at least four stretches of isozyme group-specific sequences (referred to as IGS). The IGSs of the type A isozyme are known to endow the aldolase molecules with some characteristics typical of A. In order to locate the type A regions, 4 chimeric enzymes were constructed between human aldolases A and B and 5 mutant enzymes with single or double mutations in the IGS-1 region. Among engineered proteins, the chimeric enzymes bearing the type A IGS-1 to -4 (BABA34-108:306-363) and the IGS-1 and -4 (BABA34-55:306-363) exhibited similarities to isozyme A in many respects. On the other hand, neither chimeric enzyme bearing the type A IGS-1 to -3 (BAB34-108) nor that bearing the IGS-1 alone (BAB34-55) exhibited properties as isozyme A. Four mutant aldolases A (carrying single mutation in the IGS-1 region) maintained the original activity as A. Similarly, the BA306 chimera with the type B-->A substitution at positions 41 and 45 (BA306 N41K:R45S) failed to exhibit the A-like properties although the activities toward Fru-1,6-P2 and Fru-1-P significantly increased. Conclusively, the type A IGS-1, together with the IGS-4, act as indispensable modules in determining the characteristic properties of human aldolase A.

nerve tissues and therefore shows an intermediate activity toward Fru-1,6-Pz/Fru-l-P with the ratio of approximately 10 (2).
Recently, x-ray crystallographic studies indicated that muscle-type aldolases of rabbit (3) and humans (4,5) have a pseudo-8-fold P/a-barrel structure. The core structures of Drosophila aldolase have also been shown to be essentially identical with those of the vertebrate aldolases (6,7). Comparison of amino acid sequences of vertebrate aldolase isozymes reveals that the enzyme molecules are composed of several short and long stretches of amino acid sequences belonging to three different categories: ( a ) the sequences commonly conserved among three isozymic groups (referred to as CCS), ( b ) isozyme group-specific sequences (referred to as IGS), and (c) divergent sequences (8). The CCSs are the major composition of the P/a repeating structures and their connecting turns that form the %fold P/a-barrel (8). The IGSs are located at three sites in the amino-terminal regions (IGS-1, -2, and -3) and in the carboxyl termini (IGS-4) of the enzymes (8). These molecular features of aldolases made the determination of a possible role of the IGSs in connection with an isozyme group-specific function of the enzyme very tempting. Thus, chimeric enzymes were systematically constructed between human aldolases A and B, and their characteristics were analyzed (8). Previous studies have shown that for aldolase A, the carboxyl-terminal region bearing the  and the amino-terminal region spanning amino acid residues 34-108 (IGS-1-3) serve as the determinants that exhibit characteristics of isozyme A (8). The significance of Tyr-363 and the proximate carboxyl-terminal region in determining the characteristics of isozyme A have been elucidated by several different studies using the chimeric enzyme construction (8), the enzymatic modifications (9, lo), and the site-directed mutagenesis (11,12). The importance of the carboxyl termini on isozyme-specific catalysis can be drawn from the fact that Drosophila aldolase has three isozymic forms with distinct carboxyl-terminal sequences corresponding to the IGS-4 of vertebrate aldolase (13)(14)(15). ' In order to precisely locate the regions or residues necessary for the determination of isozyme specificity, various types of chimeric enzymes were constructed between human aldolases A and B and also the mutant aldolases (with single or double mutations in the IGS-1 region). The enzymatic properties of these constructs were later analyzed. This communication postulates that the type A IGS-1, together with the IGS-4, act as indispensable modules in determining the characteristic properties of aldolase A. * R. Zhang, T. Kai, Y . Sugimoto, Y. Takasaki, K. Kaga, and K. Hori, unpublished data.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-Escherichia coli K12 strain JM83 from laboratory stock was used for the cloning and expression experiments. pINIII (16), an E. coli expression vector that contains lipoprotein promoter (Ipp), lac UV5 promoter operator (lac), and lactose repressor gene (lacI), were as previously described (17). pHAAL116-3, a cDNA clone of human aldolase A, and pHABL120-3, a cDNA clone of human aldolase B, both of which were isolated in the laboratory as previously described (17), were used to construct the expression plasmids.
Enzymes and Other Materiak-Restriction enzymes, T4 DNA ligase, T4 polynucleotide kinase, and other enzymes were purchased from Takara Biochemicals, Nippon Gene and Toyobo Co., Ltd. Mutagenesis was carried out by using the Amersham oligonucleotidedirected in uitro mutagenesis kit. 32P and the DNA sequencing kit were obtained from ICN Radiochemicals and United States Biochemical Corp., respectively.
Construction and Expression of Plasmids for Chimeric Enzymes-cDNAs for chimeric enzymes were constructed by connecting cDNA fragments for isozymes A and B (11) and that for chimeric enzyme BA34 (Fig. 1I) (8). pHA-A47, an expression plasmid of human aldolase A, pHA-B141, an expression plasmid of human aldolase B, and pH-BA34, an expression plasmid of chimeric enzyme BA34, were digested with restriction endonucleases at various sites common to the cDNAs. First, pHA-BAB34-108, which encodes a chimeric protein bearing 74 amino acid residues of isozyme A (positions 34 to 108 from the amino terminus) in between the amino-terminal and the carboxyl-terminal fragments of isozyme B, was constructed as follows; a 324-base pair-long EcoRI-AccI fragment of pHA-BA34 which covered the sequence of a chimeric enzyme BA34 cDNA encoding the amino-terminal 108 residues was ligated to an 851-base pair-long AccI-Hind111 fragment of an isozyme B cDNA from pHA-B141. The 1,175-base pair-longEcoRI-Hind111 fragment thus generated was then inserted into an E. coli pUC13 cloning vector. The positive colony for I. this chimeric enzyme was screened directly by dideoxy DNA sequencing (18) and cloned again into an E. coli expression vector to construct expression plasmid pHA-BAB34-108 ( Fig. 111, a). pHA-BAB34-55 was constructed by procedures similar to that used for the construction of pHA-BAB34-108 ( Fig.111, b ) . In this case, prior to constructing the expression plasmid, a CfrlOI site (marked with an open star) was created in the isozyme B cDNA by changing a triplet for Arg-55 from CGC to CGG by a site-directed mutagenesis. Then, cDNAs for isozyme B and BA34 were cleaved at the CfrlOI site by a restriction endonuclease and ligated as shown in Fig. 1. pHA-BABA34-108306-363 and pHA-BABA34-55306-363 were constructed by connecting EcoRI-HinfI fragments of pHA-BAB plasmids with a HinfI-Hind111 fragment of pHA-A47 (Fig. lII, c and d). These constructs were sequenced by the dideoxynucleotide chain termination method (18) to ascertain whether they carried the entire nucleotide sequences for the respective aldolase cDNAs. The respective numbers of the chimeric aldolases represent the amino acid positions from the chimeric boundary. Chimeric enzymes encoded by the expression plasmids pHA-BAB34-108, pHA-BAB34-55, pHA-BABA34-108306-363, and pHA-BABA34-55:306-363 are referred to as BAB34-108, BAB34-55, BABA34-108306-363, and BABA34-55306-363, respectively. Other chimeric enzymes employed in this study were obtained as previously described (8).
Purification of Chimeric Enzymes Generated in E. coli Transfected with the Expression Plasmids-E. coli JM83 carrying the plasmids was grown overnight in brain heart infusion (Difco) containing 50 pg/ml ampicillin, and the cells were harvested as previously described (17). Normal aldolases A and B and chimeric enzymes expressed in E. coli were purified as previously described (8) by a modification of the method of Penhoet et al. (19) for vertebrate aldolase. All purification procedures were carried out at 4 "C in the presence of 1 mM phenylmethylsulfonyl fluoride unless otherwise indicated.
Determination of Enzyme Actiuity-Aldolase activity was determined by two methods: activity staining (zymogram) and spectropho-

FIG. 1. Construction of E. coli expression plasmids harboring cDNA encoding human aldolase A and B chimeric proteins between isozymes A and B.
Upper, restriction maps of human aldolase A, B, and BA34 cDNAs. I, construction diagrams of the expression plasmids for two BAB chimeric enzymes; II, construction diagrams of the expression plasmid for the BABA chimeric enzymes. The coding regions of isozyme A and B expression plasmids and the corresponding amino acid sequences are represented by stippled and closed boxes, respectively. The noncoding regions of the expression plasmids are represented by open boxes. Arrows in expression plasmids represent E. coli lipoprotein (Ipp) and lactose (lac) promoters derived from the pINIII vector as shown in a previous paper (17). Thin lines in expression plasmids represent the sequence of the expression vector. Ac, AccI; Bg, BglII; Cf, CfrlOI; E, EcoRI; H, HindIII; Hf, HinfI. pHA-A47, an E. coli expression plasmid for isozyme A, pHA-B141, an expression plasmid for isozyme B, and pHA-BA34, an expression plasmid for chimeric isozyme BA34, were constructed as described in a previous paper (17). tometric methods. In the staining method, 10 rl each of enzyme preparations was applied on cellulose polyacetate strips, subjected to electrophoresis for 40 min at 250 V at 4 "C, and stained for aldolase activity in the presence of 5 mM EDTA (to inhibit E. coli class 11 aldolase activity) (20). The spectrophotometric assay was performed at 30 "C according to Rajkumar et al. (21).
Determination of Molecular Sizes-The molecular sizes of the wild type and the engineered aldolases were determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (22). Those of the native tetrameric forms of aldolases were determined as previously described (23).
Oligonucleotide-directed, Site-specific Mutagenesis-Mutagenesis was carried out by using the Amersham oligonucleotide-directed in uitro mutagenesis kit (25-27). The cDNA inserts of human aldolase A (pHA-A47) and BA306 (pHA-BA306) in E. coli expression plasmids were cloned into the M13mp9 phage to synthesize the template for mutagenesis. The mutagenic oligonucleotides employed in this study and the corresponding amino acid substitutions are listed in Table I. Screening of positive plaques for mutants was performed directly by dideoxy DNA sequencing (18). The mutant enzymes of aldolase A bearing a single replacement, Lys-41 + Arg, Lys-41 -+ Asn, Arg-42 -+ Lys, Ser-45 -+ Arg, are referred to as A-K41R, A-K41N, A-R42K, and A-S45R, respectively. The mutant enzyme of BA306 carrying double mutations, Asn-41-Lys and Arg-45 + Ser, is referred to as BA306N41K:R45S.

Comparison of Primary Structures of Vertebrate Aldolases
There were at least 7 highly conserved sequences (CCS) and 4 isozyme group-specific sequences (IGS) in the primary structure of the vertebrate aldolases from various sources ( Fig. 2 and Ref. 8). The CCSs appeared to be the modules for the basal framework of @/a-barrel structure and are required t o exhibit the activity common to the three isozymic forms of vertebrate aldolase (8). In particular, CCS-5 is the longest stretch bearing Lys-229 as the active site of the enzyme and Glu-187/189 as the presumptive C-1-P binding residues (3, 5), each of which is highly conserved through aldolases A, B, and C. By contrast, the IGSs were conserved within a single isozyme group or two. As shown in Fig. 2, IGSs are the modules where amino acid substitutions occurred in an isozyme group-specific manner and basically overlap the addi-tional a-helices which could be inserted into or duplicated locally by a regular repetition of P/a-structures to give three a/a helices, A1/AP, B1/B2, and H1/H2(3) of aldolase molecule when evolved.

Analyses with Chimeric Molecules of Human Aldolases
A and B

Construction of Chimeric Enzymes between Human Aldolases
A and B To ascertain whether these IGSs really determine isozymespecific functions, E. coli expression plasmids were constructed for four chimeric enzymes between isozymes A and B (Fig. 1). In this study, local sequences of isozyme A were introduced but restricted to IGS-1, IGS-1 through -3, and IGS-4 with four different combinations into the respective regions of isozyme B. The two BAB chimeric enzymes, BAB34-55 and BAB34-108, carried the amino acid sequences with either the type A IGS-1 (amino acid residues 34-55) or IGS-1-3 (amino acid residues 34-108) in between the aminoterminal and the carboxyl-terminal fragments of isozyme B. The two other chimeric enzymes, BABA34-55:306-363 and BABA34-108306-363, were the derivatives of BAB34-55 and BAB34-108, respectively, in which the type A IGS-4 (the amino acid residues 306-363) was substituted for that of the type B isozyme (Fig. 1).

Expression of Chimeric Enzymes
These constructions were transfected into E. coli JM83 as described under "Experimental Procedures." Cell extracts of E. coli transfected with the expression plasmids were assayed for aldolase activity using the activity staining method (20).
BAB34-108 did not show activity at all. BAB34-55, BABA34-108:306-363, and BA306 as controls gave detectable activity spots, while BABA34-55:306-363 was as strong as aldolase A (Fig. 3). The generated enzymes were purified by homogeneity as previously described (11). These chimeric proteins moved toward the anode with electrophoretic mobilities predicted from their isoelectric points. I l e A l a L y s Arg Leu G l n Ser I l e G l y Thr * 1

Characteristics of Chimeric Enzymes
Electrophorctic Mohilities of Chimeric Enzynes-In a previous study, it. was shown that human aldolases A and R moved with different mobilities in a SDS-PAGE as if the latt,er was apparent.ly much smaller than the former although they have almost the same molecular sizes (8). When the purified enzymes were suhjected to a SDS-PAGE, the BAR chimeric enzymes resembled aldolase R in their electrophoretic mobilities, while the RARA chimeras rather resembled aldolase A in their mohilities (data not shown).
Reactivity of These Enzymes with Monoclonal Antihodivs. mAhlA2 and mAh4C2"To ascertain if these chimeric enzymes were constructed with a structure as expected, an immunoblotting test was run by using monoclonal antihodies for human aldolase A, mAblA2 and -4C2. mAblA2 reacts with the carboxyl-terminal region of the enzyme spanninE amino acid residues 306-368. while mAh4C2 recognizes an epitope at the amino-terminal region of the enzyme present within amino acid residues 34-108 (24). As shown in Fig. 4, the two RARA chimeric enzymes as well as aldolase A reacted with mAhlA2, whereas the two HAH chimeric enzvmes w-hich lacked the type A carboxyl-terminal sequences and the human aldolase H did not react with the antibody. Similarly, chimeric enzymes carrying the t-ype A sequences from 34 to 108 (RAR34-108, HARA34-108:306-:lfi3) interacted with mAb4C2 as expected, whereas chimera with the thpe A sequences from 34 to  and HARA.34-:i;i::l06-.36:l) escaped from an interaction with the antitmdv. These results indicate t.hat the chimeric enzvmes employed in this study were constructed as expected and, furthermore, showed that the epitope recognizable by mAh4C2 should he on the amino acid sequences from 56 to 108 of the t.ype A isozyme.

General Properties of Chimeric I,'nzymrps
In this study, the same criteria as those used in a previous work to discriminate the t-ype A and H chimeric enzymes from  Table 11). This chimera was extremely thermolabile (data not shown). As can be seen in Fig. 2, the IGS regions appear to he entirely different in a primary sequence and thus possibly in secondary structure among three isozymic groups, the types A, R, and C. X-ray crystallographic analyses on rabbit ( 3 ) and human (4,5) aldolase A suggest that for the t.ype A isozyme, the carboxyl-terminal tail situates in proximity to the &barrel and is in close contact with the C-6phosphate and the C-1-phosphate binding residues located in the amino-terminal regions. This arrangement of these groups leads to a stable conformation of the type A enzyme. However, for the type H isozyme, the regions spaning IGS-2 and -3 rather than IGS-4 were found to be responsible for its catalytic activity and stability (8). Therefore, insertion of the t-ype A IGS-1-3 into the isozyme €3 to generate RAH34-108 was likely to induce a structural change in the molecule that altered its st.al)ilitv and subsequently resulted in a loss of catalytic activity.
HAHA3.I-lOR:306-369-This chimeric enzyme showed a similarity to aldolase A (Table 11); a high activity toward Fru-1 ,6-Pz and a low activity toward Fru-1-P, and a high K , for Fru-1-P. Its h,,,, toward Fru-1,6-P2 was nearly .?-fold that of aldolase H, resulting in the activity ratio for Fru-l,6-P2/Fru-I-P of about 7. T h e K , for Fru-1 -I' remained as l o w a s thnt for aldolase H. These results indicate that the type A l(;S-l-3 together with the IGS-3 are capahle of elevating the Fru-1,6-P2 cleavage activity with a concomitant reduction o f t h e activity toward Fru-1-1' to a level o f t h e isozyme (' I I , 2 ) . In spite of being more stable than HA%-108, this enzyme was also shown to lose the activity t o less than I O f . ; that ofcontrol when incubated for 30 min at 45 "C. This result not only supports the previous data showing that the total IGSs o f t h r t-ype A isozyme are required to exhihit ch;Ir;lcteristics as A, but also suggests that the type A I(;S c o u l d not suhstitrrte lor the corresponding sequences o f the type 13 isozyme in terms of isozyme specific function for stability of thcb molecule. HAH3.I-55-The HAH chimera never exhi1)itetl enzymatic properties similar to the type A isozyme because o f the absence of the type A carboxyl-terminal sequences ( I ( S 3 ) that are indispensable for aldolase A activity. A l t hough the specific activities of the chimera toward Fru-I,f;-I'? and Fru-1-1' were a t most q50°;, of the aldolase H ( Table 11). it rather resembled in many respects: the Fru-1.6-I',/E'ru-l-I'activity ratio, the K , for both substrates, and physicochemical properties. This chimera was, however, considerably stable (data not shown).
RAHA3.1-,55:30fi-:~h'9-The HAHA chimera exhibited a striking contrast to HAR34-5.5 and HAH3.1-108 in many respects: the activity toward Fru-1,6-P2 was as high as 4-fold that of the isozyme H while that toward Fru-1 -I' was down t o the level of the isozyme A, resulting in the Fru-1,6-P2/Fru-l-P activity ratio of 19, nearly one-third that of the isozyme A. Furthermore, the K , values for Fru-1,6-P2 and Fru-1-1' were also the same level as those of the isozyme A. Conclusively. this chimeric enzyme exhibits characteristics close t o those of aldolase A.
These results strongly suggest that the type A I(iS-1 and -4 are mainly responsible for exhibiting characteristics as the isozyme A and also indicate that ICs-2 and -3 are not likely to play a role in the functions directly related t o the activity toward Fru-l,6-P2 and Fru-1-1'. In a previous paper (8), it was shown that an internal region spanning positions 108-212 would he responsihle for regulating the catalytic activities toward Fru-1-P; low for aldolase A and high for aldolase B (8). This was presumably because the region carried amino acid residues implicated to be the C-1-phosphate hinding sites ( 3 , 5); moreover, HA108 chimera exhibited a low catalytic activity toward Fru-1-P like aldolase A, whereas RA212 showed I\ high Fru-1-P activity like aldolase H ( X ) . However, the present results show definitely that the type A IGS-1, in conjunction with IGS-4, modulate the catalytic activity of aldolase A toward Fn1-1.6-1'~ at a high level and that toward Fru-1-P at a low level. This indicates that the ICs-1 would be superior to the internal sequences covering positions 108 to 212 in determining the enzyme activity t nward Fru-1-P. It is also likely that the t+ype H IGS-2 and -:i would be responsible for the conformational stability o f aldolase I3 since HAR:14-.5.5 was more stable than HAH:b-lOX. General properties of these chimeric proteins are summarized in Table 11.

I)ouhlr, Mutation in IC;S-I
The results obtained with four chinwrir enzymes constructed in this study strongly suggest that I(iS-1 play n crucial role in exhibiting the properties ofthe type A isozyme. As can be seen in Fig. ? specificity toward Fru-1,6-P2 and Fru-1-P. The type A IGS-1 bears Lys-41 and Arg-42, both of which have been implicated t o be the C-6-phosphate binding sites from x-ray crystallographic studies ( 5 ) , and Ser-45, a residue also implicated to be of importance in determining the specificity toward Fru-1,6-P2 and Fru-1-P ( 5 ) . Thus, four mutant enzymes of isozyme A were constructed, each of which had single mutations at the respective sites, to examine whether the residues in the parent enzyme were really indispensable for the activity (Table 111). The A-K41R and the A-K41N were derivatives of human aldolase A in which Lys-41 was replaced by Arg, a positively charged residue, in the former and by the type B residue, Asn, in the latter, respectively. The A-R42K is a derivative of the isozyme A carrying the Arg-42 + Lys substitution. Similarly, the A-S45R is a derivative of isozyme A with the Ser-45 + Arg substitution. All derivatives of isozyme A were active and maintained almost the same activities toward Fru-1,6-Pz and Fru-1-P as that of the parent isozyme A (Table 111). Thus, it appears that amino acid residues a t positions 41 and 42 are replaceable with another residue having a similar property. The Ser-45 of aldolase A is likely to be dispensable since the mutant with the Ser-45 + Arg substitution was still active (Table  111). BA306 N41K:R45S is a derivative of the chimeric enzyme BA306 carrying the double mutations at positions 41 and 45 as the Asn-41-Lys and the Arg-45 + Ser substitutions. The effects of the double mutations on the catalytic activities toward Frul,6-Pz and Fru-1-P were prominent, and the kcat values for these substrates were likely to be 1.6-fold as compared to those of BA306 (Table HI), indicating that the type A residues, Lys-41 and Ser-45, play together a role of endowing the enzyme with significantly high catalytic activities toward Fru-1,6-P2 and Fru-1-P, but did not fulfill the role played by the type A IGS-1. Therefore, the type A IGS per se could be required as a module for exhibiting the activity like isozyme A, although there is another possibility that amino acid residues other than Lys-41 and Ser-45 play a role in the process.
Although the data presented here lead to the conclusion that IGS-1, together with IGS-4, would be particularly important in conferring isozyme-specific catalysis of the particular isozyme, more evidences are needed to show which residues or what structure is really responsible for the activity. At present, the functions of IGS-2 and -3 remain unknown. Since the amino acid sequences covering the IGS-2 and -3 are situated at the surface opposite to the region where the subunit contact occurs (3)(4)(5), the IGS-2 and -3 regions might serve as the isozyme-specific modules for conformational stabilization, the tissue-specific interaction with other glycolytic enzymes, or cytoskeletal proteins in cellular compartments (37,38).
The results obtained with several chimeric enzymes and the mutant enzymes with single or double mutations clearly indicate that the isozyme group-specific sequences in human aldolases A and B would play a positive role in elucidating isozymic functions such as a high or low catalytic activity toward Fru-1,6-P2 and Fru-1-P, the K,,, for these substrates and so on. Therefore, it may be speculated that the three isozymic forms of vertebrate aldolase might be generated from a n ancestor through multiple mutations accumulated mainly in the IGS regions which can be mapped to the A1/A2, B1/B2, and HI/H2 sites, the ala-helices located at the flanking region of the regular 8-fold p/a-barrel ~t r u c t u r e .~