Characterization of Homoisocitrate Dehydrogenase Involved in Lysine Biosynthesis of an Extremely Thermophilic Bacterium,Thermus thermophilus HB27, and Evolutionary Implication of β-Decarboxylating Dehydrogenase*

Although the presence of an enzyme that catalyzes β-decarboxylating dehydrogenation of homoisocitrate to synthesize 2-oxoadipate has been postulated in the lysine biosynthesis pathway through α-aminoadipate (AAA), the enzyme has not yet been analyzed at all, because no gene encoding the enzyme has been identified until recently. A gene encoding a protein with a significant amino acid sequence identity to both isocitrate dehydrogenase and 3-isopropylmalate dehydrogenase was cloned from Thermus thermophilus HB27. The gene product produced in recombinantEscherichia coli cells demonstrated homoisocitrate dehydrogenase (HICDH) activity. A knockout mutant of the gene showed an AAA-auxotrophic phenotype, indicating that the gene product is involved in lysine biosynthesis through AAA. We therefore named this genehicdh. HICDH, the gene product, did not catalyze the conversion of 3-isopropylmalate to 2-oxoisocaproate, a leucine biosynthetic reaction, but it did recognize isocitrate, a related compound in the tricarboxylic acid cycle, as well as homoisocitrate as a substrate. It is of interest that HICDH catalyzes the reaction with isocitrate about 20 times more efficiently than the reaction with the putative native substrate, homoisocitrate. The broad specificity and possible dual function suggest that this enzyme represents a key link in the evolution of the pathways utilizing citrate derivatives. Site-directed mutagenesis study reveals that replacement of Arg85 with Val in HICDH causes complete loss of activity with isocitrate but significant activity with 3-isopropylmalate and retains activity with homoisocitrate. These results indicate that Arg85 is a key residue for both substrate specificity and evolution of β-decarboxylating dehydrogenases.

Conversion of homoisocitrate to 2-oxoadipate is supposed to occur in lysine biosynthesis through ␣-aminoadipate (AAA), a pathway that has been believed to be present only in fungi and yeast ( Fig. 1) (10 -13). The enzyme containing ␤-decarboxylating dehydrogenation activity involved in this process is homoisocitrate dehydrogenase (HICDH; EC 1.1.1.115). The reactions catalyzed by HICDH proceed in a manner similar to those by IPMDH and ICDH. In addition, IPMDH and ICDH show distinct identity in amino acid sequence to each other. Until recently, no genes have been identified and characterized as hicdh even in Saccharomyces cerevisiae, although its whole genome sequence has been determined (14). Recently, Chen and Jeong (3) reported that the gene named YIL094C of S. cerevisiae and the gene with the accession number T38612 of Schizosaccharomyces pombe could be likely candidates encoding HICDHs based on the amino acid sequence alignment for this protein family and the three-dimensional structures of IPMDH and ICDH. They further showed that the T38612 product actually had HICDH activity but did not characterize it further. This new member of ␤-decarboxylating dehydrogenases is expected to lead us not only to elucidation of the substrate recognition but also to understanding evolution of the protein family as well as cellular metabolism including amino acid biosynthesis. However, unlike with ICDH and IP-MDH, there is little information on properties of HICDH.
Thermus thermophilus HB27, an extremely thermophilic bacterium, optimally grows at 70°C. We recently found that this bacterium synthesizes lysine through AAA as an intermediate (15) like fungi and yeast, although all bacteria and plants are believed to synthesize lysine through diaminopimelate. Sequence analysis of the cloned genes that were shown to be involved in the lysine biosynthesis suggested that the first half of the biosynthesis (from 2-oxoglutarate to AAA) proceeded in the same manner as that of fungi and yeast. The reactions were also suggested to be similar to those of leucine biosynthesis and a part of the tricarboxylic acid cycle (16). The second half of the biosynthesis from AAA to lysine is, on the other hand, similar to the conversion of glutamate to ornithine in arginine biosynthesis (17). We have succeeded in cloning of most of the genes involved in lysine AAA biosynthesis in T. thermophilus HB27 (15)(16)(17)(18). However, the gene encoding HICDH that converts homoisocitrate to 2-oxoadipate was not contained in the cloned DNA fragments. In this paper, we describe cloning of the gene encoding HICDH from T. thermophilus HB27 and, for the first time, detailed analysis of properties for HICDH including sitedirected mutagenesis study. The evolutionary relationship among leucine biosynthesis, the tricarboxylic acid cycle, and newly identified lysine biosynthesis are also discussed.
Molecular Cloning and Sequencing of hicdh Gene-Oligonucleotides used in this study are listed in Table I. Based on amino acid sequence alignment among IPMDHs, ICDHs, and putative HICDHs from various sources, oligonucleotides (HICD-N1 and HICD-C1) were designed and used as degenerate primers for PCR for obtaining a DNA fragment corresponding to the hicdh gene of T. thermophilus HB27. The following thermal cycle was used: step 1, 95°C, 5 min; step 2, 95°C, 1 min; step 3, 68°C, 1 min; step 4, 75°C, 1 min; step 5, 75°C, 7 min. Steps 2-4 were repeated 25 cycles. An amplified 635-bp fragment was blunted, phosphorylated, and ligated into the EcoRV site of pBluescript II KS (ϩ) (Stratagene). The resulting plasmid was digested with EcoRI and Hin-dIII, and the smaller fragment containing the amplified DNA fragment was used as the probe for the following Southern hybridization. Southern hybridization was carried out with a Random Primer Fluorescein Labeling kit (PerkinElmer Life Sciences) against chromosomal DNA of T. thermophilus HB27 digested with several restriction enzymes. An approximately 2.6-kb SacI fragment showing positive hybridization with the probe was cloned into pUC18 and introduced into E. coli DH5␣. Colony hybridization was carried out using the same probe for Southern hybridization. A plasmid contained in a hybridization-positive colony was named pHICD-Sac2600. The nucleotide sequence was determined by the method of Sanger et al. (22).
Knockout of hicdh in T. thermophilus HB27 and Auxotrophic Complementation Test-The plasmid for knockout of hicdh was constructed as follows. Two independent PCRs using DFHICN1/DFHICN2 for amplifying 590 bp of a 5Ј-portion of the hicdh gene and DFHICC1/ DFHICC2 for amplifying 619 bp of a 3Ј-portion of the hicdh gene as primers were performed with pHICD-Sac2600 as the template. PCR conditions were as follows: step 1, 94°C for 2 min; step 2, 94°C for 15 s; step 3, 68°C for 30 s; step 4, 68°C for 1 min; step 5, 68°C for 7 min. Steps 2-4 were repeated 30 times using a thermostable DNA polymerase, KOD-plus (TOYOBO). After the amplified 5Ј-and 3Ј-fragments were digested with EcoRI/XbaI and XbaI/HindIII, respectively, both the resulting fragments were ligated into pUC18 previously digested with EcoRI/HindIII. The resulting plasmid was named p18DFHIC. The next PCR for amplifying the heat-stable kanamycin nucleotidyltransferase gene (23) was performed using KmR-Nt/KmR-Ct as the primers and pUC39 -442KmR (24) as the template. PCR conditions were as follows: step 1, 94°C for 2 min; step 2, 94°C for 15 s; step 3, 55°C for 30 s; step 4, 68°C for 2 min; step 5, 68°C for 7 min. Steps 2-4 were repeated 30 times using KOD-plus. An amplified fragment was digested with XbaI and cloned into p18DFHIC previously digested with the same enzyme. The resulting plasmid named p18DFHICKmR was used for knockout of the hicdh gene of T. thermophilus HB27 by the method described previously (17). Colonies growing on TM plate (20) supplemented with 50 g/ml of kanamycin were picked, and disruption of the genes was confirmed by Southern hybridization. For the auxotrophic complementation test, the knockout mutant, named MJ101, was inoculated on the minimal medium (MM plate) (21) and the MM plate supplemented with 0.1 mM AAA or 0.1 mM lysine.
Phylogenetic Analysis-A phylogenetic tree was constructed by using the neighbor joining method. Sequence data for the analysis were obtained from the GenBank TM and PIR data bases. The amino acid sequences were aligned by using ClustalW (25) on the DDBJ. Using this aligned data, a phylogenetic tree was constructed by a computer program, Mega 2 (26).
Expression of hicdh, icd, and leuB from T. thermophilus HB27 in E. coli and Preparation of the Crude Extract-NdeI and EcoRI recognition sites were introduced around the start codon and the termination codon of hicdh from T. thermophilus HB27, respectively, by PCR using synthetic oligonucleotides, 19HICN and HICDH-Ct. PCR conditions were as follows: step 1, 94°C for 2 min; step 2, 94°C for 1 min; step 3, 65°C for 1 min; step 4, 72°C for 2 min; step 5, 72°C for 7 min. Steps 2-4 were repeated 30 times using Ex-taq (Takara Shuzo). The amplified fragment was digested with HindIII and EcoRI and cloned into pUC18. After the nucleotide sequence was verified, an NdeI-and EcoRI-digested fragment was cloned into pET26b (ϩ) (Novagen, Darmstadt, Germany). The resulting plasmid, pET-tHICDH101, was used for expression of hicdh. To construct the plasmids for overexpression of leuB and icd genes from T. thermophilus HB27, PCRs with the chromosomal DNA were performed by using primers 19leuBN1/19leuBC2 and ICN/ ICC, respectively. PCRs were performed under the same conditions used for amplifying the hicdh gene. Amplified fragments were digested with HindIII and EcoRI, subcloned into pUC119 for leuB and pUC18 for icd, to yield pUC119-sleuB and pUC18-ICD, respectively. For overproduction of HICDH, E. coli BL21-Codon-Plus (DE3)-RIL harboring pET-tHICDH101 was cultured in 2ϫ YT medium containing 50 g/ml kanamycin and 30 g/ml chloramphenicol. When the E. coli cells were grown to give an A 600 of 0.5, isopropyl-␤-D-thiogalactopyranoside (final concentration 1 mM) was added. The culture was continued for an additional 12 h after the induction. For overexpression of leuB and icd, E. coli JM109 harboring pUC119-sleuB or pUC18-ICD was cultured in the same way for the hicdh overexpression.
E. coli cells that overexpressed hicdh, icd, or leuB from T. thermophilus HB27 were suspended in 5 ml of buffer IV (20 mM potassium phosphate, pH 7.5, 0.5 mM EDTA) and disrupted by sonication. The supernatant prepared by centrifugation at 40,000 ϫ g for 20 min was heated at 70°C for 20 min, and denatured proteins from E. coli cells were removed by centrifugation.
Purification of Recombinant HICDH-E. coli cells of 10 g (wet weight) harboring pET-tHICDH101 were suspended in 20 ml of buffer IV and disrupted by sonication. The supernatant prepared by centrifugation at 40,000 ϫ g for 20 min was heated at 85°C for 20 min, and denatured proteins from E. coli cells were removed by centrifugation. Supernatant fractions were applied onto an anion exchange column, DE-52 (Whatman Japan, Tokyo, Japan), pre-equilibrated with buffer IV, washed with buffer IV containing 0.1 M NaCl, and eluted with buffer IV containing 0.2 M NaCl. Ammonium sulfate was added to the eluted fractions at a final concentration of 45%, and the resultant precipitate was collected by centrifugation at 40,000 ϫ g for 30 min, dissolved in buffer IV, and applied onto a Hi-load 26/60 Superdex200 prep grade column (Amersham Biosciences) pre-equilibrated with buffer IV containing 0.2 M NaCl. ICDH from T. thermophilus HB27 was also purified by the method described previously (27).
Purity of the recombinant enzyme was verified by SDS-12% PAGE. Proteins were determined by the method of Bradford (28) by using a Bio-Rad protein assay kit (Nippon Bio-Rad, Tokyo, Japan). Sedimentation equilibrium analysis of the purified HICDH was carried out with a Beckman Optima XL-A analytical ultracentrifuge fitted with a Beckman An-60Ti analytical rotor. The subunit organization of the enzyme was analyzed according to the procedure of Van Holde and Baldwin (29).
Heat Stability of HICDH and ICDH of T. thermophilus HB27-Heat stability of HICDH and ICDH was analyzed as follows. For calculating the melting temperature (T m ) of the enzymes, purified enzyme dissolved in buffer IV (0.5 mg/ml) was incubated at various temperatures for 20 min and immediately chilled in ice water. Then aggregated proteins were removed by centrifugation (23,500 ϫ g for 5 min), and the remaining activity in the supernatant was measured. For determining the half-life (t1 ⁄2 ) of the enzymes, the enzyme dissolved in buffer IV (0.5 mg/ml) was incubated at 90°C, and an aliquot was taken at an appropriate interval followed by immediate cooling in ice water. After centrifugation to remove denatured proteins, the remaining activity of the supernatant was measured.
Enzyme Assays-Ten microliters of the enzyme solution (0.1 mg/ml for isocitrate and 0.2 mg/ml for homoisocitrate) was added to the reaction mixture (50 mM HEPES, pH 8.0, 200 mM KCl, 5 mM MgCl 2 , 1 mM NAD ϩ , and 25-5000 M homoisocitrate) that was preincubated at 60°C for 5 min. For determination of the kinetic constants for isocitrate, 4 -400 M isocitrate was added to the reaction mixture in place of homoisocitrate. The reaction was monitored at 60°C by following the increase in absorbance at 340 nm. Specific activity was determined by using a 400 M concentration of each substrate, and 1 unit of the enzymes was defined as the amount of enzyme that produced 1 mol of NADH or NADPH/min at 60°C in the enzymatic reaction.
Kinetic parameters were calculated by using an initial velocity program, HYPER, of Cleland (30).

Cloning and Sequence Analysis of the hicdh Gene-Putative candidates for HICDHs have been proposed in two yeasts (3).
In addition, the whole genome sequence has been determined (31) for a radiation-resistant bacterium, Deinococcus radiodurans, that is closely related to T. thermophilus in taxonomy and suggested to possess the lysine biosynthetic system through AAA found in T. thermophilus (32). D. radiodurans possesses a homolog of yeast HICDH genes in addition to putative leuB and icd. Based on the amino acid sequences of the putative HICDHs from yeast and the Deinococcus strain and Blast analysis against the incomplete T. thermophilus HB27 genome sequence (available on the World Wide Web at www.g2l.bio.uni-goettingen.de/blast/blast_search.html), we designed a pair of primers (HICD-N1 and HICD-C1) to amplify a portion of hicdh of T. thermophilus HB27 by PCR. The resulting amplified DNA fragment of about 640 bp carried a sequence that was similar to those of the hicdh/leuB/icd family but obviously different from those of leuB and icd from T. thermophilus (33)(34)(35). We then tried to clone the DNA fragment of T. thermophilus HB27 that hybridized against the amplified DNA fragment.
When the chromosomal DNA was digested with several restriction enzymes and subjected to Southern hybridization using the amplified DNA fragment as the probe, SacI digestion gave a positive signal of ϳ2.6 kb. The DNA fragment was recovered and inserted into pUC18. A clone positive in the colony hybridization assay was picked, and the plasmid contained in the transformant was named pHICD-Sac2600. In this SacI fragment, three open reading frames were found. One of the open reading frames encoded a protein showing 44 and 45% identity in amino acid sequence to IPMDH and ICDH from T. thermophilus HB8, respectively (33)(34)(35). The other two open reading frames did not show amino acid sequence similarity to other proteins whose functions were identified.
Based on the crystal structures of IPMDH and ICDH, amino acid residues responsible for the recognition of a common portion, the malate moiety, of the substrates and amino acid residues determining the substrate specificity can be easily predicted (36 -38). As expected, amino acid residues involved in the recognition of the malate moiety of the substrate are all conserved even in the putative Thermus HICDH (Fig. 2). Furthermore, by comparison of the amino acid residues at positions 255 and 266, the enzyme is suggested to utilize NAD ϩ as the coenzyme. On the other hand, the amino acid residues responsible for the recognition of the ␥-moieties in IPMDH and ICDH are different in HICDH. These observations suggest that this homolog does not serve as IPMDH and ICDH, but HICDH. Phylogenetic analysis of this protein family also supports this hypothesis because the homolog forms one group with putative HICDHs that separated from groups of IPMDHs and ICDHs (Fig. 3). Knockout of hicdh in T. thermophilus HB27-We performed a knockout experiment of the homolog to examine whether or not this gene is related to lysine biosynthesis in T. thermophi-lus HB27. A knockout mutant, MJ101, did not grow on minimal medium (Fig. 4). However, growth of the mutant was restored by the addition of AAA or lysine, indicating that the homolog is related to lysine biosynthesis of the microorganism and has a role in a reaction step before AAA synthesis as expected. We hereafter call this homolog hicdh.
Specific Activities of HICDH, ICDH, and IPMDH from T. thermophilus HB27-We next constructed plasmids for expression in E. coli of hicdh, icd, and leuB from T. thermophilus HB27 and determined the specific activities of the gene products after removal of most of the proteins in the E. coli crude extract by heat treatment and successive centrifugation. Thermus HICDH gave a specific activity of 564 units/mg with homoisocitrate as the substrate in an NAD ϩ -dependent manner. Surprisingly, the enzyme was able to catalyze the reaction using isocitrate as a substrate with much higher specific activity (8,579 units/mg). On the other hand, HICDH did not recognize 3-isopropylmalate as a substrate. A similar analysis was also performed with crude extract containing ICDH or IPMDH. ICDH catalyzed the reaction with both isocitrate (14,835 units/ mg) and homoisocitrate (764 units/mg), whereas, as expected, IPMDH recognized 3-isopropylmalate (2,418 units/mg) as a substrate but did not utilize homoisocitrate and isocitrate as substrates at all. Therefore, we analyzed only HICDH and ICDH in detail in the following experiments.
Expression and Purification of HICDH-16 mg of HICDH was prepared through three purification steps from 1 liter of culture with over 95% purity on SDS-PAGE (data not shown). We carried out sedimentation equilibrium analysis to evaluate the quaternary structure of HICDH. By using a partial specific volume of 0.745 that was calculated from the amino acid composition, data were fitted best with the dimer-tetramer equilibrium: equilibrium constants for monomer-dimer and monomer-tetramer association of 1.19 ϫ 10 47 M Ϫ1 and 3.80 ϫ 10 95 M Ϫ3 , respectively. According to the equilibrium constants, we estimate that HICDH is present as a mixture of homodimer and homotetramer with a ratio of about 25 and 75%, respectively, when the enzyme concentration is in the range of 0.25-1.0 mg/ml. We therefore treated HICDH as a homotetramer in the kinetic analysis described below. The heat stability of HICDH and ICDH was analyzed by two different assay methods. When melting temperatures (T m ) were examined by incubating the purified enzyme at various temperatures for 20 min, FIG. 2. Amino acid sequence alignment of HICDHs, ICDHs, and IPMDHs from various sources. Closed circles, amino acid residues that interact with the malate moiety of isocitrate and 3-isopropylmalate; open circles, amino acid residues that interact with the adenineribose portion of NAD ϩ in T. thermophilus HB8 IPMDH (37). Amino acid residues necessary to recognize the unique ␥-moiety of 3-isopropylmalate or isocitrate are shown in white on black. Amino acid residues proposed to recognize ␥-moiety of homoisocitrate are boxed. Amino acid residues conserved in all the enzymes are marked with asterisks. TtICDH, ICDH from T. thermophilus HB27 (this study); EcICDH, ICDH from E. coli (P08200); ScHICDH, putative HICDH from S. cerevisiae (P40495); SpHICDH, HICDH from S. pombe (T38621); TtHICDH, HICDH from T. thermophilus HB27 (this study); TtIPMDH, IPMDH from T. thermophilus HB27 (this study); TfIPMDH, IPMDH from Thiobacillus ferrooxidans (JX0286). T m of 93.6 and 77.2°C were obtained for HICDH and ICDH, respectively. Next, we determined the half-life (t1 ⁄2 ) of each enzyme at 90°C. The half-life of HICDH was determined to be 16.7 h, whereas that of ICDH was only 14 s. Thus, HICDH is much more stable than ICDH.
Kinetic Properties of HICDH Activity-To characterize HICDH in detail, kinetic analysis was carried out with purified HICDH. As shown above, because HICDH recognized both homoisocitrate and isocitrate as the substrates, we used both compounds for the analysis. The result shows that the k cat value (171 s Ϫ1 ) of HICDH is almost the same for the reaction with either homoisocitrate or isocitrate (Table II). However, the K m value of HICDH for homoisocitrate was much larger than that for isocitrate. Consequently, the catalytic efficiency (k cat / K m ) of the reaction for homoisocitrate is about 20 times smaller than that for isocitrate. When a similar analysis was carried out for ICDH, ICDH also recognized homoisocitrate as a substrate, although the catalytic efficiency for homoisocitrate was about 20 times lower than that for isocitrate. It was of interest that the k cat /K m value of HICDH for homoisocitrate was 2 times smaller than the corresponding value of ICDH for homoisocitrate. Although the reaction with homoisocitrate was less efficient than the reaction with isocitrate for both HICDH and ICDH, the reason for the reduced efficiency differs between the two enzymes. HICDH has a larger K m value for homoisocitrate, whereas ICDH has a lower k cat value for the reaction with homoisocitrate.
Site-directed Mutagenesis to Alter the Substrate Specificity of HICDH-As shown above, Thermus HICDH recognizes both homoisocitrate and isocitrate as substrates, in contrast to HIC-DHs from yeast, S. pombe and S. cerevisiae, which utilize homoisocitrate but not isocitrate as the substrate (3). 2 Structural studies for IPMDH and ICDH revealed that only a few amino acid residues in a loop between ␤-strand 3 and ␣-helix 4 determine the substrate specificity of the enzymes of the same enzyme family (36 -38). Since HICDH shows significant amino acid sequence identity to both enzymes, we hypothesized that a limited number of residues also determine the substrate specificity even in HICDH. In order to elucidate the residues determining the difference in the substrate specificity between yeast and Thermus HICDHs, we constructed six altered HICDHs from T. thermophilus HB27 in each of which 1-7 amino acid replacements were introduced in the loop region. An altered enzyme, HICDH/7Sc, containing the same sequence as that of S. cerevisiae HICDH in the loop displayed activity for homoisocitrate but not for isocitrate (Fig. 5). On the other hand, two other altered enzymes, HICDH/4Sc and HICDH/5Sc, containing YSSP sequence like HICDH/7Sc and the yeast enzymes, recognized isocitrate as well as homoisocitrate as substrates, although the activity was significantly decreased in the altered enzymes. These results suggested that Arg 85 and/or Tyr 86 were required for recognizing isocitrate. Enzyme assays for HICDH/R85V, HICDH/2Sc, and HICDH/3Sc, all of which possessed Val at position 85, revealed that the altered enzymes completely lost activity for isocitrate but acquired activity with 3-isopropylmalate as a substrate. DISCUSSION A gene coding for a new member of ␤-decarboxylating dehydrogenase, HICDH, has been cloned from an extreme thermophile, Thermus thermophilus HB27, and its involvement in lysine biosynthesis has been demonstrated. Although HICDH is indispensable for lysine AAA biosynthesis and its primary function in vivo must be the conversion of homoisocitrate to 2-oxoadipate, kinetic analysis showed that the enzyme can also utilize isocitrate, an intermediate of the tricarboxylic acid cycle, 20 times more efficiently compared with using homoisocitrate as a substrate (Table II, Fig. 3). This may indicate that the kinetic constant of an enzyme determined in vitro does not always reflect its physiological role, or it may be possible that HICDH will substitute for ICDH in some cases. On the other hand, the catalytic efficiency in terms of k cat /K m of ICDH for the ICDH reaction is in a range similar to that of HICDH for the same reaction. Therefore, this may suggest that ICDH can function in lysine AAA biosynthesis as well.
HICDH and ICDH share similarities in primary sequence and reaction mechanism and are therefore thought to have diverged from a common ancestral enzyme (3). A common ancestor is thought to have broad substrate specificity (1), so an HICDH from T. thermophilus HB27 that can catalyze reactions contained in both lysine AAA biosynthesis and the tricarboxylic acid cycle may still have features of the ancestral-type enzyme. It should be noted that we recently found that homocitrate synthase, which catalyzes synthesis of homocitrate from 2-oxoglutarate and acetyl-CoA, the first reaction in lysine biosynthesis of T. thermophilus HB27, was able to catalyze the synthesis of citrate from oxaloacetate and acetyl-CoA, the corresponding reaction by citrate synthase in the tricarboxylic acid cycle, although homocitrate synthase has no structural similarity to citrate synthase (39). Taken together with our previous results on LysJ and LysK (17,18), both of which are able to catalyze the reactions in lysine and arginine biosynthesis, dual functions and/or broad specificity may be common features of lysine biosynthetic enzymes of T. thermophilus HB27.
Amino acid residues responsible for binding the malate moiety of the substrates are completely conserved in ␤-decarboxylating dehydrogenases, including HICDH (Fig. 2). Therefore, the corresponding residues are presumably involved in substrate binding even in HICDH. However, amino acid residues in the ␤-strand 3-␣-helix 4 loop that recognize the ␥-moiety (Fig. 1) are not conserved between ICDH and IPMDH. According to the three-dimensional structure of IPMDH of T. ferrooxidans, three residues (Glu 88 , Leu 91 , and Leu 92 in T. ferrooxidans IPMDH numbering (38)) are specific determinants for recognition of 3-isopropylmalate (Fig. 6). Leu 91 and Leu 92 make hydrophobic contacts with the ␥-isopropyl group of 3-isopropylmalate, and Glu 88 interacts with the nicotinamide ring of NAD ϩ for stabilizing the Michaelis complex. Glu 88 also serves to prevent IPMDH from binding the ␥-carboxylate group of isocitrate via charge repulsion. On the other hand, in ICDH from E. coli, two residues (Ser 113 and Asn 115 in the numbering of E. coli ICDH) are indicated by x-ray structural analysis to be responsible for isocitrate recognition (36). Ser 113 forms a hydrogen bond to the ␥-carboxylate group of isocitrate, and Asn 115 interacts both with the ␥-carboxylate group of isocitrate and with NADP ϩ . Due to the polar environment of the ␥-moietybinding site of the ICDH, the hydrophobic isopropyl moiety of 3-isopropylmalate is excluded from the site. In HICDH, regions probably recognizing the ␥-moiety are also occupied by amino acid residues different from those in ICDH and IPMDH (Figs. 1 and 2). Chen and Jeong (3) aligned amino acid sequences of several enzymes in this protein family and suggested Tyr 106 and Ile 110 (numbering according to HICDH from S. cerevisiae) as the residues that were necessary for recognition of the ␥-moiety of homoisocitrate. They speculated that the phenolic hydroxyl group of Tyr 106 formed a hydrogen bond to ␦-carboxylate of homoisocitrate analogous to Ser 113 in E. coli ICDH. In HICDH from T. thermophilus HB27, however, the corresponding position is occupied by phenylalanine. This indicates that the mechanism for recognition of homoisocitrate has to be reconsidered in HICDH.
Thermus HICDH recognizes both homoisocitrate and isocitrate as substrates (Table II) in contrast to HICDHs from S. pombe and S. cerevisiae, where the yeast enzymes catalyze the reaction with only homoisocitrate and neither isocitrate nor 3-isopropylmalate (3). 2 To elucidate the difference in the mechanism for substrate recognition between Thermus and yeast HICDHs, we introduced amino acid replacements into the ␤-strand 3-␣-helix 4 loop region of Thermus HICDH. An altered enzyme, HICDH/7Sc, containing the same sequence as that of S. cerevisiae HICDH in the loop displayed activity for homoisocitrate but not for isocitrate, indicating that the loop actually does contain the determinants for substrate specificity in HICDH. Among the altered HICDHs, all of the enzymes with Val at position 85 completely lost the ability to utilize isocitrate as a substrate. These results indicate that Arg 85 is required for recognition of isocitrate as a substrate in Thermus HICDH.  However, this result looks confusing, because the corresponding position (116 in E. coli ICDH) is also occupied by Val in ICDH. Val 116 in E. coli ICDH does not directly interact with substrate nor NADP ϩ , but replacement of the residue by Leu or Ser converted the ICDH to preference toward 3-isopropylmalate (9). This suggests that substrate recognition of the loop residues differs between these enzymes.
The phylogenetic tree of HICDH, ICDH, and IPMDH from various sources indicates that HICDH from T. thermophilus HB27 belongs to one group along with the HICDHs from two yeasts and DR1674 from D. radiodurans that is suggested to synthesize lysine through AAA (Fig. 3). According to the hicdh knockout experiment, the gene was shown to be involved in lysine biosynthesis. Nevertheless, unlike the case for yeast HICDHs, Thermus HICDH has higher activity for isocitrate. Our present study indicates that only a few amino acid replacements are enough to convert Thermus HICDH to an enzyme that is specific to homoisocitrate. Since HICDH/7Sc displayed extremely high thermostability comparable with that of the wild-type enzyme, which may ensure that the microorganism grows at extremely elevated temperatures, we may assume that it would be possible for the organism to acquire mutations that could generate a homoisocitrate-specific enzyme. We at present do not know what causes the enzyme to have dual functions or quite higher activity for isocitrate. Physiology and metabolism will have to be investigated in detail for T. thermophilus HB27 to understand this.
We previously reported that an anaerobic hyperthermophilic archaeaon, Pyrococcus horikoshii, has a gene cluster that is very similar to that of T. thermophilus HB27, with each component of the archaeon tightly linked to the corresponding counterpart in T. thermophilus HB27 in phylogenetic analysis, suggesting that the archaeaon probably synthesizes lysine through AAA as in the case of T. thermophilus HB27 (16). However, PH1722, an HICDH homolog, in the gene cluster of P. horikoshii does not belong to this group in the phylogenetic tree but rather to the IPMDH group (Fig. 3). Similar gene clusters were also found in two related thermoacidophilic archaea, Sulfolobus tokodaii strain 7 and Sulfolobus sulfataricus, whose genome sequences have recently been determined (40,41). Recently, expression of the gene cluster was shown to be regulated by LysM, an Lrp-like transcriptional regulator, depending on the presence or absence of lysine (42), although the functions of the cluster have not yet been analyzed. Based on these observations, it is speculated that the two Sulfolobus species also synthesize lysine through AAA. However, it is of interest that these microorganisms carry no hicdh-homologous genes other than icd and leuB in their genomes. IPMDH (LeuB) was isolated from S. tokodaii strain 7 and analyzed in detail (2). The Sulfolobus IPMDH cannot utilize homoisocitrate or isocitrate as substrates (2). 2 This may suggest that the HICDH reaction of lysine AAA biosynthesis in Sulfolobus could be catalyzed by a putative ICDH. In order to prove this hypothesis, substrate specificity of the putative ICDH has to be analyzed. In any case, PH1722 in Pyrococcus and the putative ICDHs in the Sulfolobus species could be keys to elucidate evolution of ␤-decarboxylating dehydrogenase. It should be also noted that the HICDH group is monophyletic, with a deep divergent point between IPMDH and ICDH groups. This suggests that HICDH has not evolved from IPMDH or ICDH but rather directly from an ancestral enzyme of ␤-decarboxylating dehydrogenase.
Amino acid sequence identity between HICDH and ICDH is 45%, and the basic mechanisms for reactions and substrate recognition to bind a malate moiety seem similar between the two enzymes. However, although these enzymes also showed similar catalytic efficiency (k cat /K m ) for the reaction with homoisocitrate or isocitrate, profiles in the kinetic parameters were different; HICDH has larger k cat and K m values, and ICDH has smaller k cat and K m values. In addition, HICDH is a much more heat-stable enzyme than ICDH. Determination of structures of both enzymes may lead us to understand not only the detailed catalytic mechanism but also molecular evolution along with the heat stability of these enzymes.