Subdomain II of α-Isopropylmalate Synthase Is Essential for Activity

Background: Isopropylmalate synthases (IPMSs) with and without a regulatory domain were found. Results: IPMS subdomain II is essential for activities and likely involved in acetyl-CoA binding-mediated conformation transition. Conclusion: The N-terminal domain and the two subdomains comprise the complete and independently functional catalytic module of IPMS. Significance: The IPMS catalytic module was defined and characterized, which inferred a probable feedback inhibition mechanism. The committed step of leucine biosynthesis, converting acetyl-CoA and α-ketoisovalerate into α-isopropylmalate, is catalyzed by α-isopropylmalate synthase (IPMS), an allosteric enzyme subjected to feedback inhibition by the end product l-leucine. We characterized the short form IPMS from Leptospira biflexa (LbIPMS2), which exhibits a catalytic activity comparable with that of the long form IPMS (LbIPMS1) and has a similar N-terminal domain followed by subdomain I and subdomain II but lacks the whole C-terminal regulatory domain. We found that partial deletion of the regulatory domain of LbIPMS1 resulted in a loss of about 50% of the catalytic activity; however, when the regulatory domain was deleted up to Arg-385, producing a protein that is almost equivalent to the intact LbIPMS2, about 90% of the activity was maintained. Moreover, in LbIPMS2 or LbIPMS1, further deletion of several residues from the C terminus of subdomain II significantly impaired or completely abolished the catalytic activity, respectively. These results define a complete and independently functional catalytic module of IPMS consisting of both the N-terminal domain and the two subdomains. Structural comparison of LbIPMS2 and the Mycobacterium tuberculosis IPMS revealed two different conformations of subdomain II that likely represent two substrate-binding states related to cooperative catalysis. The biochemical and structural analyses together with the previously published hydrogen-deuterium exchange data led us to propose a conformation transition mechanism for feedback inhibition mediated by subdomains I and II that might associated with alteration of the binding affinity toward acetyl-CoA.

The crystal structure of IPMS from Mycobacterium tuberculosis (MtIPMS) (9) revealed an asymmetrical homodimer architecture. Each monomer is composed of an N-terminal domain and a C-terminal regulatory domain connected by two subdomains, designated as I and II. The N-terminal domain assumes a (␤/␣) 8 TIM barrel with one Asp and two His residues located at the active site to bind the substrate ␣-Kiv and a Zn 2ϩ ion. Based on the structural results, a possible four-step catalytic mechanism was proposed with enol formation and stabilization mediated by interaction with the positively charged side chain of Arg-80 as one of the critical steps (9,10). Later, the crystal structure of the N-terminal domain of (R)-citramalate synthase (EC 2.3.1.182; CMS) from Leptospira interrograns (LiCMSN), an analogous allosteric enzyme for isoleucine biosynthesis, in complex with acetyl-CoA identified the acetyl-CoA binding site, and the structural and biochemical data together revealed the molecular basis of the substrate specificity (11). In addition, the structure of the C-terminal regulatory domain of the same enzyme (LiCMSC) complexed with isoleucine clearly demonstrated the ligand-binding site and its selectivity for feedback inhibition (12).
One of the important questions raised during these studies is whether the regulatory domain is required for the catalytic activity (13,14). Several recent studies seem to support the idea that the activity of IPMS is dependent on its C-terminal regulatory domain (13)(14)(15)(16). However, as early as 2003, when the genome of L. interrogans serotype lai was initially sequenced, three copies of leuA-like genes were annotated (7). One of them, LA_0469 was proved to encode an active IPMS with a significantly shorter peptide than those encoded by the other two genes (LA_2350, i.e. cimA and LA_2202, i.e. leuA1); it apparently lacks the regulatory domain and thus was designated as leuA2. This finding implied that the catalytic activity of the IPMS-like enzymes might be independent of the regulatory domain and thus directly led us to characterize the enzymatic features of both the long and short forms of IPMSs from Leptospira biflexa. We determined the crystal structure of the short form IPMS. The following molecular and biochemical comparative analyses identified the minimal key structural elements responsible for the full activity of IMPS without the regulatory domain. Further structural comparison with MtIPMS revealed conformational difference between the long and the short forms of the enzymes, which may represent different stages of catalysis. Combining the previously published hydrogen-deuterium exchange data with our mutational analysis data, a potential mechanism involving acetyl-CoA binding mediated by the two subdomains and associated with acetyl-CoA binding is proposed for feedback inhibition.
These gene fragments were digested by NheI/NotI or NdeI/ HindIII and inserted into the pET-28b expression plasmid, resulting in N-terminal hexahistidine (His 6 )-tagged pET-28b-LbIPMS1/2. The plasmid was transformed into Escherichia coli BL21(DE3) strain (Novagen), and the transformed bacterial cells were cultured at 37°C in LB medium containing 50 g/ml kanamycin.
Protein expression was induced by adding isopropyl ␤-Dthiogalactoside into the medium to a final concentration of 1 mM. The cells were harvested by centrifugation at 5000 ϫ g for 10 min at 4°C, resuspended in a lysis buffer (50 mM Tris-HCl (pH 8.0), 300 mM KCl, and 1 mM PMSF), and then disrupted using a French press. The recombinant protein was purified with affinity chromatography using a Ni 2ϩ -nitrilotriacetate Superflow column (Qiagen) pre-equilibrated with buffer A (50 mM Tris-HCl (pH 8.0) and 300 mM KCl) and then washed with buffer B (buffer A supplemented with 10 mM imidazole for IPMS2 or 40 mM imidazole for IPMS1) to remove nonspecifically bound proteins. The target protein was eluted with buffer C (buffer A supplemented with 100 mM imidazole for IPMS2 or 300 mM imidazole for IPMS1), and the eluted fractions were dialyzed against buffer D (20 mM Tris-HCl (pH 8.4) and 50 mM KCl). After purification, the target protein was of sufficient purity (above 95%) and was then concentrated to approximately 10 mg/ml in buffer D by ultracentrifugation for further structural and biochemical studies.
Selenomethionine-substituted LbIPMS2 suitable for structure determination was prepared following the method described previously (17). Purification of the selenomethionine LbIPMS2 protein was performed using the same methods as for the native protein. Expression and purification of the LbIPMS2 truncations were the same as for the full-length protein as described above.
Crystallization and Diffraction Data Collection-Purified LbIPMS2 protein was used for the crystallization experiments, which were performed at 4°C using the hanging drop vapor diffusion method. Crystals were grown in a drop containing equal volumes (1 l) of the protein solution (5 mg/ml) and the reservoir solution (0.1 M Tris-HCl (pH 8.5), 0.8 mM sodium formate, and 30% PEG 2000 monomethyl ether). The crystals of LbIPMS2 were later found to contain the substrate ␣-Kiv; therefore this structure represents the protein complexed with the substrate ␣-Kiv (LbIPMS2-Kiv). The LbIPMS2-IPM complex crystals were prepared by co-crystallization with substrates ␣-Kiv, Zn 2ϩ , and acetyl-CoA. Attempts to crystallize the LbIPMS2 in the presence of acetyl-CoA have been unsuccessful so far.
For diffraction data collection, the LbIPMS2 crystals were first cryoprotected using Paratone oil (Hampton Research) and then flash cooled in liquid nitrogen. Selenium single wavelength anomalous dispersion diffraction data of LbIPMS2 were collected to a resolution of 2.9 Å from a flash cooled crystal at 100 K at the Photon Factory (Japan), beamline BL5A. The native diffraction data were collected to a resolution of 2.0 Å at beamline NW12. LbIPMS2-IPM was collected to 2.2 Å at beamline BL17U of Shanghai Synchrotron Radiation Facility (China). All the diffraction data were processed, integrated, and scaled together using the HKL2000 suite (18). The crystals of native LbIPMS2 belong to space group P3 1 21, containing one LbIPMS2 molecule in the asymmetric unit with a solvent content of 60%. The statistics of the diffraction data are summarized in Table 2.
Structure Determination and Refinement-The structure of the LbIPMS2-Kiv complex was solved using the single wavelength anomalous dispersion method implemented in the program SOLVE (19). The single wavelength anomalous dispersion phases were improved by statistical density modification, including solvent flattening and histogram matching, using the program RESOLVE (20), increasing the overall figure of merit from 0.39 to 0.77 at 2.9-Å resolution. The resultant electron density map was of high quality, and RESOLVE automatically built 60% of the polyalanine model. The full-structure model was built manually using the program Coot (21). Structure refinement was carried out against the 2.0-Å native data using the program PHENIX with standard protocols (22). The structure of the LbIPMS2-IPM complex was solved by molecular replacement using PHENIX with the structure of the LbIPMS2-Kiv complex used as the starting model. In the structures, there is strong electron density at the active site that matched the bound substrate or product very well. In addition, there was a residual electron density near the bound substrate that could be fitted with a divalent metal ion. The statistics of the structure refinement and the quality of the structure models are summarized in Table 2.
Enzymatic Activity Assay-The enzymatic activity of both wild-type and mutant LbIPMS1/2 were assayed by monitoring the production of reduced CoA (HS-CoA) over time as described previously (11,23). Specifically, a typical reaction mixture consists of varied concentrations of ␣-Kiv (or other ␣-keto acids in the determination of substrate specificity) and acetyl-CoA, 2 mM MnCl 2 , 50 mM KCl, 0.1 M Tris-HCl (pH 8.5), and 25 nM LbIPMS1/2 in a total volume of 50 l. To determine the effect of ␣-Kiv, the concentration of acetyl-CoA was fixed at 4 mM (about 4 times K m of acetyl-CoA for LbIPMS2), and the concentration of ␣-Kiv was varied from 50 M to 5 mM. To determine the effect of acetyl-CoA, the concentration of ␣-Kiv was fixed at 4 mM (about 8 times K m of ␣-Kiv for LbIPMS2), and the concentration of acetyl-CoA was varied from 200 M to 6 mM. The reaction was performed in a water bath at 45°C for LbIPMS2 and 37°C for LbIPMS1 for 15 min and then stopped by cooling on ice for 5 min.
For measurement of the HS-CoA produced, 35 l of 1 M Tris-HCl (pH 8.0), 25 l of 10 mM 5,5-dithiobis(2-nitrobenzoic acid) dissolved in 0.1 M Tris-HCl (pH 8.0), and 390 l of distilled H 2 O were added to the reactant to a total volume of 500 l. The yellow 5-mercapto-2-nitrobenzoic acid generated was quantified at 412 nm using a Beckman Du800 spectrometer and blanked against an identical incubation sample without the substrate (either ␣-Kiv or other ␣-keto acids). The concentration of HS-CoA was calculated from a linear standard curve generated with known concentrations (0 -400 M) of 2-mercaptoethanol. The production of HS-CoA was found to be linear over the time period of the assay, and the product formation was a linear function of the amount of the enzyme added. All experiments were repeated at least twice under the same conditions. The allosteric models were shown by Scatchard plot, and the Hill numbers were determined by plotting log[v/(1 Ϫ v)] versus [S]. All kinetic parameters (V max , K m , and k cat ) were calculated by non-linear regression analysis using Prism 4.0 for Windows (GraphPad).
The optimal pH for LbIPMS1/2-catalyzed reactions was determined in the pH range of 6.5-10.1 in about 0.4-pH unit increments under conditions otherwise the same as that of standard, but the Tris-HCl buffer (0.1 M) in reaction assays was replaced by MES (pH 6.5), HEPES (pH 6.5-8.1), Tris-HCl (pH 8.1-8.9), or glycine (pH 8.9 -10.1) buffer (0.1 M in all cases). Because an excess amount of divalent (2 mM Mn 2ϩ ) and monovalent ions (50 mM K ϩ ) was included in the standard enzymatic assay systems, a further ionic strength change caused by pH adjustment may be avoided. The optimal reaction temperature for LbIPMS1 and LbIPMS2 was determined in the reaction temperature range of 0, 20, 37, 40, 45, 50, 60, and 65°C. The thermostability assay was performed by treating LbIPMS1/2 at different temperatures of 0, 20, 30, 37, 40, 45, 50, 55, 60, and 65°C for 30 min before cooling on ice. The above treated enzyme was then used in the catalytic reaction.
In the experiments to measure the effects of different metal ions on the enzymatic activity of LbIPMS1/2, we removed the bound metal ions co-purified with the enzyme by using a reported protocol (11). Then the protein was used for the activity assay in the presence of different metal ions.
Complementation Assay-A complementation assay was carried out in E. coli CV512 strain, which lacks functional IPMS. The strain is able to grow on M9 medium only when transformed with functional IPMS. The full length and truncations of LbIPMS1/2 cDNA were amplified and linked with the plasmid pTRC99a. The resulting new plasmid was transformed to E. coli CV512, and the empty vector pTRc99a was used as a negative control. The strain was cultured under 37°C on an M9 medium plate supplemented with 1 mM isopropyl ␤-D-thiogalactoside.

LbIPMS2 Exhibits Catalytic Activities Comparable with That of LbIPMS1 with a Unique Characteristic of Unresponsiveness
to Monovalent Cation Activation-Similar to L. interrogans, the genome of L. biflexa also bears three copies of leuA-like genes. LEPBI_I1291, which shares high amino acid sequence similarity to cimA of L. interrogans (LA_2350), is likely to encode a CMS. LEPBI_I1845 and LEPBI_I1108 are most likely to encode IPMSs based on their high sequence similarities to their counterparts in L. interrogans, LA_2202 (leuA1) and LA_0469 (leuA2), respectively (24). The 500-residue protein encoded by LEPBI_I1845 (leuA1) is designated as LbIPMS1, and the 394residue protein encoded by LEPBI_I1108 (leuA2) is designated as LbIPMS2. Sequence analyses show that LbIPMS1 has a structure arrangement similar to that of MtIPMS and is correspondingly composed of an N-terminal domain, a C-terminal regulatory domain, and two connecting subdomains I and II, whereas LbIPMS2 contains only the N-terminal domain and the connecting subdomains, apparently lacking the C-terminal regulatory domain (Fig. 1).
The two leuA genes from L. biflexa were individually expressed in E. coli, and the recombinant proteins were affinity chromatographically purified to Ͼ90% as detected by SDS-PAGE. The enzymatic activities indicate that, in addition to the natural substrate ␣-Kiv, both LbIPMS1 and LbIPMS2 can catalyze the condensation reaction of transferring the acyl group from acetyl-CoA to several ␣-keto acids, and LbIPMS1 shows a broader substrate spectrum than LbIPMS2. It is noteworthy that the K m values of LbIPMS2 for both acetyl-CoA and ␣-Kiv are 2-fold higher than those of LbIPMS1, leading to an approximately 2-fold decrease of the catalytic efficiency ( Table 1).
Activation of IPMS by monovalent cations is well documented for the enzymes from several organisms, including the activation of MtIPMS by K ϩ (23,(25)(26)(27)(28). Additional divalent metal ions cannot activate the enzymes from yeast, Salmonella typhimurium, or Alcaligenes eutrophus (23,(27)(28)(29) but are able to further improve the activities of both MtIPMS (14,26) and LiCMS (11). In this study, we demonstrated that the activities of both LbIPMS1 and LbIPMS2 can be potentiated by several divalent metals with Mn 2ϩ as the most effective activator followed by Mg 2ϩ , Co 2ϩ , and Ca 2ϩ , whereas Zn 2ϩ may slightly inhibit the activities of both enzymes ( Fig. 2A). The activation of LbIPMS1 and LbIPMS2 by Mn 2ϩ was concentration-dependent with the maximum at 0.5 mM Mn 2ϩ (Fig. 2B). However, it was unexpected that although the activity of LbIPMS1 was coactivated by monovalent metal ions with K ϩ /NH 4 ϩ being the . This finding suggested a speculative association with the short form structure of LbIPMS2 among all of the IPMSs studied so far. However, we failed to identify any of its other biochemical properties that might account for or be caused by this unique character. Nevertheless, to keep the assay conditions consistent, 50 mM K ϩ and 2 mM Mn 2ϩ were added in the reaction systems for both enzymes throughout the study ("Experimental Procedures"). The optimum pH values for both LbIPMS1 and LbIPMS2, measured as V max values, were 8.5. This result is highly similar to that previously characterized for IPMSs from M. tuberculosis (10), S. typhimurium (23), A. eutrophus (29), and plant, i.e. the two isozymes from Arabidopsis (30). The optimum reaction temperature for LbIPMS2 (50°C) was much higher than that for LbIPMS1 (37°C) (Fig. 2E).The stability of LbIPMS2 was remarkably decreased when the temperature exceeded 45°C and was higher than that of LbIPMS1at 37°C. Therefore, the enzymatic assay was carried out at 45°C with LbIPMS2 and 37°C with LbIPMS1.
Crystal Structure of LbIPMS2 Reveals the Active Center for Binding the Substrate and the Product-We solved the crystal structure of LbIPMS2 by the single wavelength anomalous dispersion method ("Experimental Procedures"). Each asymmetric unit contains one LbIPMS2 molecule bound with the substrate ␣-Kiv and a Zn 2ϩ at the active center, representing the substrate complex (LbIPMS2-Kiv). As we did not add ␣-Kiv during protein purification or crystallization, the substrate should be acquired from the host E. coli. Acetyl-CoA and Zn 2ϩ were added in other co-crystallization trials, and the final electron density map showed a good match for the reaction product ␣-IPM, representing the product complex (LbIPMS2-IPM). The statistics of the final structure models are summarized in Table 2. As the two structures do not show significant conformational differences, we used the structure of LbIPMS2-Kiv as the representative for description.
The structure of LbIPMS2-Kiv contains residues 3-301 and 310 -387 with residues 302-309 disordered. Similar to MtIPMS, the overall structure of LbIPMS2 can be divided into three parts: the N-terminal domain, subdomain I, and sub-domain II (Fig. 3A). The N-terminal domain (residues 3-289) adopts a (␤/␣) 8 TIM barrel fold with ␣-Kiv and Zn 2ϩ bound at the active center located in the middle of the barrel. Subdomain I (residues 290 -333) forms an ␣-helix (␣10), a short 3 10 helix, and two loops packing against ␣10. Subdomain II (residues 334 -394) consists of three ␣-helices, ␣11, ␣12, and ␣13, which form a compact helical bundle mediated by hydrophobic residues Ile-343/Leu-346/Leu-347 of ␣11, Ile-359/Leu-362/Phe-363/Ala-366 of ␣12, Leu-383/Val-384/Leu-386 of ␣13, and Phe-334/Ile-352/Val-354/Ile-378 of the connected loops. Two types of interactions (type A and type B) between the symmetry-related molecules were identified. The type A interactions were formed between the N-terminal domains and buried 4709 Å 2 (30.2%) of the total surface area. Similar types of interactions have also been identified in the structures of MtIPMS and LiCMSN, suggesting that the dimerization of LbIPMS2 is biologically relevant (Fig. 3B), and this was supported by size exclusion chromatography of the active enzyme in solution (data not shown). The type B interactions were formed between two subdomains II via mainly hydrogen bonds and buried 1785 Å 2 (11.3%) of solvent-accessible surface area (Fig. 3C).
At the active center of LbIPMS2, the substrate ␣-Kiv is stabilized by a metal ion. In the structures of LiCMSN and MtIPMS, a Zn 2ϩ ion was identified to bind at the same site (9,11). As the bound metal ion in LbIPMS2 is coordinated by six ligands (His-209, 2.3 Å; His-207, 2.2 Å; Asp-15, 2.1 Å; water 14, 2.3 Å; and two carbonyl oxygens of ␣-Kiv, both 2.3 Å) with an octahedral geometry in a similar way as the Zn 2ϩ ion in the structures of LiCMSN and MtIPMS, we speculate that it is very likely a Zn 2ϩ ion as well even though we did not determine it chemically. The substrate ␣-Kiv also makes hydrogen-bonding interactions with two residues, Thr-178 and Arg-14 (2.3 and 2.9 Å, respectively) (Fig. 4A). All of these residues are highly conserved among IPMSs from different species (Fig. 1), suggesting a common binding mode of the substrate. When the product is formed as the result of the catalytic reaction, as shown by the structure of the LbIPMS2-IPM complex, the conformation of LbIPMS2 does not show much difference from the LbIPMS2-Kiv complex (root mean square deviations are 0.6 Å). In the structure of LbIPMS2-IPM, a metal ion was also found to bind at the active center with similar coordination geometry, and most of the residues constituting the active center assume similar positions and conformations as observed in the structure of LbIPMS2-Kiv. However, the side-chain conformation of Gln-18 in LbIPMS2-IPM rotates about 180°compared with that in the structure of LbIPMS2-Kiv and forms two hydrogen bonds with the carbonyl oxygen of ␣-IPM; one is direct, whereas the other is via a water molecule (water 123). In addition, water 14 also forms a hydrogen bond (2.4 Å) with the carbonyl oxygen of ␣-IPM (Fig. 4B). In general, the structures of the enzyme bound with either the substrate or the product are highly similar to each other, albeit we cannot exclude the possibility of potential conformational changes during the reaction process, especially for acetyl-CoA binding. Intact Subdomain II Is Required for the Catalytic Activity of Both Short Form and Long Form LbIPMSs-Because LbIPMS2, which lacks the regulatory domain, has enzymatic activity com-parable with that of LbIPMS1, we hypothesized that the regulatory domain might not be required for the activity of LbIPMS. To evaluate this hypothesis, we truncated the regulatory domain of LbIPMS1 at residue Arg-385 (LbIPMS1-R385) (hereafter, the truncation mutant is named as "protein-truncated residue" for short) to construct an LbIPMS2-mimetic protein based on the structure-based sequence alignment (Fig. 1). The enzymatic assay showed that LbIPMS1-R385 retained 90% of the activity of the full-length LbIPMS1 (Fig.  5A). The activation of LbIPMS1-R385 by di-or monovalent metals also showed patterns similar to that of the full-length LbIPMS1 (Fig. 2). These results confirm our hypothesis that the presence of the regulatory domain is not required for the enzymatic activity and further suggest that activation of LbIPMS1 by monovalent metals is independent of the regulatory domain. The crystal structure of LbIPMS2 shows a novel organization such that subdomain II is positioned away from the N-terminal domain of the same molecule and close to the active center of the adjacent molecule in the biologically relevant dimer (Fig.  3B). To clarify the functional role of subdomain II in the enzymatic activity, we further truncated LbIPMS2 from its C terminus in different lengths and measured the enzymatic activities of the resultant mutants (Fig. 5A). The results show that either removal or disruption of subdomain II could lead to complete loss or a dramatic decrease of the enzymatic activity, and the minimum length required for the activity is LbIPMS2-S389, which retains about 64% activity of the full-length LbIPMS2. Further deletion of more residues either almost disrupted the activity (i.e. LbIPMS2-S387; 0.3% activity remained) or completely abolished the activity (i.e. LbIPMS2-R376). These data are consistent with the in vivo complementation assay results. The leuA Ϫ mutant containing LbIPMS2-S389 grew normally; in contrast, that containing the LbIPMS2-S387 grew very slowly, and that containing LbIPMS2-R376 did not grow at all (Fig. 5A). In the structure of LbIPMS2, Ser-389 is located at the C terminus of ␣13, which is on the distal site of the active center of the adjacent monomer and packs against ␣11 and ␣12 through hydrophobic interactions (Figs. 1 and 3C). LbIPMS2-S387 (i.e. removing the C-terminal residues of ␣13) or LbIPMS2-R376 (i.e. complete deletion of ␣13) would expose the hydrophobic core of subdomain II to the solvent and therefore destabilize the overall structure of subdomain II (Fig. 3D).
Consistently, these two LbIPMS2 truncation mutants purified from the heterogeneous expression system seemed unstable (Fig. 5B). To verify these observations, we generated several equivalent truncation mutants of LbIPMS1 based on the sequence alignment with LbIPMS2. Enzymatic activity assays and genetic complementation phenotypes of these mutants showed similar results as the LbIPMS2 mutants, strongly suggesting that an intact subdomain II is required for the enzymatic activity of LbIPMS2 and LbIPMS1 (Figs. 1 and 5A).
Subdomain II Is Likely Involved in the Conformation Transition during Catalysis That Might Be Associated with the Binding to Acetyl-CoA-In the dimeric MtIPMS, the only known structure for the long form IPMSs, the two monomers exhibit significant conformational differences. In particular, subdomain II of monomer A is located on the top of the active site of monomer B, whereas subdomain II of monomer B is far away from the active center of monomer A (9, 31). In contrast, in the short form dimeric LbIPMS2, two symmetry-related monomers form a homodimer (Fig. 6), and the conformation of both subdomains II are similar to that of monomer A but substantially different from that of monomer B in MtIPMS (Fig. 6). Because of the flexibility of the linker between subdomains I and II, it is reasonable to hypothesize that subdomain II may adopt different conformations, and one of them should represent the state during the catalytic reaction.
Full-length LbIPMSs may be involved in two kinds of allosteric conformational changes during catalysis. One is the homotropic positive cooperation between the two monomers with either of the two substrates as an effector. A particularly complex example of this kind of allostery is found in the IPMS from A. eutrophus H16. With its Hill coefficient for either of the two substrates alternating between values higher and lower than 1.0, the cooperativity of the enzyme seems to change between positive and none along with variation of the concentration of the substrates (32). Another form is heterotropic inhibition caused by the binding of the ligand, for instance L-leucine for IPMSs of many different origins (1,23,33). The homotropic positive cooperative reaction kinetics features of LbIPMS2 and its homolog LbIPMS1-R385 (Fig. 7) implicate that the two monomers of the functional dimer may adopt different conformations corresponding to different catalytic states with or without the substrates bound. The Hill numbers of these enzymes indicated that the positive cooperativity against acetyl-CoA was stronger (or more complex; for n H Ͼ 2) than that against ␣-Kiv (n H ϭ 1.8) (Fig. 7). In fact, similar positive cooperativity was found in the long form LbIPMS1 as well (derived from data in Fig. 8), but the Hill number for either of the two substrates (1.6 -1.8) was lower than that of the short form enzymes.
The questions of whether and how subdomain II is related to this allosteric effect was further studied. Among the series of truncation mutants of LbIPMS2, LbIPMS2-Q353, which contains only a small portion (helix ␣11) of subdomain II (Fig. 1), is special because it could not catalyze the catalytic reaction for ␣-Kiv but retained the ability to catalyze the condensation reaction of ketobutyrate with very low activity (more than a 97% decrease of the k cat ). However, its K m toward ketobutyrate was   nearly unaffected (Table 3), consistent with the previous report that deletion of the regulatory domain of MtIPMS did not affect the ␣-Kiv binding (16). Conversely, the K m of this truncated enzyme for the cofactor acetyl-CoA decreased by 2-fold (Table  3). This result implies that, as an essential part of the intact catalytic unit, subdomain II might be functioning in the recruitment and binding of acetyl-CoA.
We also investigated the feedback inhibition for the fulllength LbIPMS1. Similar to that observed for MtIPMS (1), a V-type inhibition was shown in the case of LbIPMS1 against the substrate ␣-Kiv with its K m value almost unchanged (Fig. 8C). However, when the inhibitory effect against acetyl-CoA was measured for LbIPMS1 in the presence of 1 mM L-leucine, the k cat was about 68.8% of that of the apoenzyme, indicating a V-type inhibition. Meanwhile, the K m was about twice of that of the apoenzyme, inferring a K-type inhibition mechanism (Fig.  8B). Therefore, the inhibitory effect against substrate acetyl-CoA is complex, but a mechanism involving acetyl-CoA binding in the process of feedback inhibition might be inferred to the first order of approximation (Fig. 8). (The activity of truncation mutants of LbIPMS1/2 were experimentally tested, and the relative activity was calculated based on the specific activity of the wild-type LbIPMS2 (2.8 ϫ 10 3 nmol/mg⅐min), which was 100. a, not detected (N); b, undetectable (ND) (i.e. the specific activity is less than 0.1% of that of LbIPMS2); c, reported undetectable; d, for MtIPMS-F457, which retains 12% of the activity of full-length MtIPMS (100%) (14).The clones expressing truncations of LbIPMS1 and LbIPMS2 were transformed into a leuA-null mutant to observe their capability of complementing the auxotrophic growth defect of the host. ϩϩ, growth similarly to the full length; ϩ, growth slower than the full length; Ϫ, cannot grow. B, the truncation mutants were expressed heterologously and purified, and the purity and stability of the resulting proteins were checked by SDS-PAGE. M, molecular mass markers.

DISCUSSION
In this study, we characterized structurally and biochemically the natural short form LbIPMS2 from L. biflexa that lacks the regulatory domain but has catalytic activity comparable with that of the long form LbIPMS1 from the same species. Based on these analyses, we further tested a series of truncation mutants from the C terminus of subdomain II of LbIPMS2 to pinpoint the minimal structural components essential for catalysis (Fig. 5A). The results indicate that either removal or disruption of subdomain II may lead to the complete loss or a dramatic decrease of the enzyme activities, whereas deletions that left the subdomain II unaltered had little effect upon the activities. Therefore, we conclude that the catalytic activity of IPMS is independent of its regulatory domain but requires an intact subdomain II. In other words, the previously defined "N-terminal catalytic domain" in MtIPMS and LiCMSN (9,11,12) is actually incomplete, and by integrating the N-terminal domain hosting the catalytic center with the complete subdomain I and subdomain II, a catalytic module, which is functionally and structurally independent of the regulatory domain, is defined.
This conclusion appears to conflict with the results of previous mutation analyses conducted in IPMS/CMSs regarding the role of the regulatory domain in maintaining the full enzymatic activity. In the case of Saccharomyces cerevisiae IPMS (ScIPMS), an "R" region (514 -552) deletion leads to complete loss of the ScIPMS activity. Because this R region is likely located in the middle of the regulatory domain corresponding to the structure of MtIPMS (Fig. 1), this result echoes our data, at least to a certain extent, that partial truncation of the regulatory domain would cause significant, albeit not necessarily complete, loss of the LiCMS activity (Fig. 5A). In the other two typical studies (15,16), truncation of the C-terminal regulatory domain in both IPMSs from Neisseria meningitidis and M. tuberculosis is shown to result in the complete loss of catalytic function with limited structural changes to the active sites of either of the truncated proteins observed (16). However, structure-based sequence alignment matching the C-terminal residues of those truncations, i.e. the NmIPMS-E365 and MtIPMS-V425, to LbIPMS2 residues, i.e. the Ser-368 and Thr-331, respectively, indicated that not only removal of the regulatory domain but also disruption of subdomain II can occur (Figs. 1 and 5A). Therefore, the catalytic module was disrupted in these studies.
By searching the protein structure database, we found that the homocitrate synthase from Schizosaccharomyces pombe (SpHCS) is another short form Claisen enzyme that catalyzes the condensing reaction of ␣-ketoglutarate and acetyl-CoA. Composed of an N-terminal domain followed by subdomain I and subdomain II, SpHCS is not only structurally (Protein Data Bank code 3IVS) (34) similar to LbIPMS2, but its subdomain II has been shown to be essential for catalytic activity (35). Therefore, together with the structure/function characteristics of SpHCS, the basic catalytic module defined in this study, using LbIPMS2 as a representative, should be regarded as a general model for the IPMS-like Claisen enzymes no matter whether an extra regulatory domain is present or not.
The essential function of an intact subdomain II with regard to the catalytic activity of IPMS is somewhat unexpected given the distance from it to the catalytic center. In our previous work, the structure of LiCMSN complexed with its substrates (pyruvate and acetyl-CoA) indicates that acetyl-CoA is bound in the deep surface groove of the TIM barrel near the bound pyruvate and is stabilized through Phe-83, Arg-16, Gln-20, and Glu-146 (11), all located on one side of the active center, which does not seem sufficient for stable binding of acetyl-CoA. In this study, we identified that the K m of acetyl-CoA was increased about 2-fold in the truncated active enzyme LbIPMS2-Q353 (Table 3), inferring that subdomain II might be another element involved in stable binding with acetyl-CoA. In addition, the moderate distance of subdomain II to the active center in the symmetric dimer allows us to speculate that the binding of acetyl-CoA may induce the reorientation of subdomain II, which might further adjust the active site residues to the proper position required for catalysis to occur. This speculation is also supported by the homotropic allosteric kinetics property of the IPMSs (23, 32), including LbIPMS2 (Fig. 7), showing an obvious cooperative effect of substrate binding. We also identified significant difference in the subdomain II orientation between the structure of LbIPMS2 and the structure of MtIPMS in complex with ␣-Kiv, respectively (Fig. 6). Because the MtIPMS structure with ␣-Kiv bound to both monomers was determined via soaking the substrates into the crystal where the preexistent crystal packing might prevent any substrate binding-induced conformational change, the structure may represent the preactive state. Meanwhile, our LbIPMS2-Kiv structure may represent the active state because the substrate was co-purified and prebound to the enzyme during crystallization. Therefore, the difference in the orientation of subdomain II between the two structures may reflect the conformational change between the active and preactive states in the widely accepted allosteric model in which it is generally understood that allosteric enzymes in the preactive state can be converted to the active state by substrate binding (3, 36 -38). Comparison of the crystal structure of MtIPMS with or without L-leucine binding revealed no evident structural changes at the active site (9). However, a mechanism of feedback inhibition/allosteric regulation involving interdomain communication via a change in conformational equilibrium rather than changes in a static structure was proposed previously (39,40) based on the mutation analysis of the critical residue Tyr-410 of MtIPMS subdomain I ( Fig. 1; corresponding to Tyr-316 in LbIPMS2) laid over the active site of the adjacent monomer (39). In that study, the Y410F mutant lost feedback inhibition to L-leucine; meanwhile, its K m value toward acetyl-CoA decreased about 10-fold, and the k cat values were reduced by 35-fold relative to that of the wild-type enzyme. Those results suggest that the active site of the Y410F mutant is more accommodatable for acetyl-CoA binding, whereas the mutation breaks the pathway for transmitting the signal of L-leucine binding to the catalytic center (39). A similar phenomenon can be found in the mutation analysis of Asp-444 residue located in subdomain II of MtIPMS (corresponding to Arg-350 of LbI-PMS2) that also demonstrated higher affinity for acetyl-CoA and less sensitivity to L-leucine feedback inhibition than the wild-type enzyme (31). Considering that the signal of L-leucine binding should pass subdomain II prior to subdomain I from   the regulatory domain, we further hypothesize that subdomain II may have a similar function as subdomain I.
In a hydrogen-deuterium exchange assay, three peptides of L-leucine-bound MtIPMS showed decreased deuterium incorporation compared with the apoenzyme, indicating that conformational changes may happen in these peptides when L-leucine binds; i.e. feedback inhibition occurs (40). Structure-based alignment (Fig. 1) mapped two peptides (residues 453-457 and residues 488 -495) in subdomain II of L-leucine-bound MtIPMS to the corresponding residues in subdomain II of LbIPMS2, i.e. residues 359 -363 and residues 391-394, respectively. The former peptide, located at the second ␣-helix of subdomain II adjacent to Gln-353 (this truncation only affects the binding affinity of acetyl-CoA) ( Table 3), can be inferred to participate in the binding of acetyl-CoA, although the latter peptide, located at the C terminus of subdomain II adjacent to Ser-389 (Fig. 5A), may not be essential for catalysis per se. However, we cannot exclude its possible role in the transition of the allosteric signal. The third peptide in the TIM barrel domain (residues 78 -87) can be mapped to residues 14 -23 in LiCMS (residues 12-21 in LbIPMS2) in which Arg-16 and Gln-20 are known to stabilize the binding of acetyl-CoA (11), indicating that the conformational changes induced by L-leucine binding may interfere with binding of acetyl-CoA to these residues. Thus, the inhibition occurring in the active center is most likely achieved through the transition of the signal induced by L-leucine binding from the regulatory domain to the catalytic module via a certain conformational change of the intermediate domains (subdomain I and subdomain II), which may directly or indirectly be associated with the binding of acetyl-CoA.
This proposed bifunctional role of subdomain II in both acetyl-CoA binding and allosteric transmission between the regulatory domain and the catalytic module is further supported by the significant difference in the orientation of LbIPMS2 subdomain II from that of MtIPMS (Fig. 6). This structural difference is likely attributable to the active versus preactive states of the enzyme corresponding to the homotropic allostery and might be equivalent to the active versus inhibited states of the long form enzyme corresponding to the heterotropic allostery. Therefore, an alternative scenario might account for the fact that there is no change observed in the structure with or without the presence of L-leucine.
Recently, it was shown that V-type inhibition in MtIPMS is due to perturbation of the hydrolysis step by binding of L-leucine, which has no effect on the K m for acetyl-CoA (41). Conversely, the isoleucine feedback inhibition effect on LiCMS is K-type for either pyruvate or acetyl-CoA (11), and acetyl-CoA binding to LiCMS does show low level of cooperativity. 5 In this study, our data about LbIPMS1 present a mixed mechanism of feedback inhibition in between that of the other two enzymes with its K m and k cat both affected (Fig. 8). In addition, binding of either of the two substrates demonstrates cooperativity; albeit both are weaker than that of the short form LbIPMSs (Fig. 7). All of these data present a complex spectrum of mechanisms of action in IPMSs and their related enzymes such as LiCMS. Nevertheless, because our data match well with most of the previous biochemical measurements and the truncated LbIPMS1-R385 performs with almost identical kinetics as that of LbIPMS2 (Fig. 7), we propose that the function of subdomain II might be involved not only in acetyl-CoA binding in the short form IPMS but also in the conformation transition process during feedback inhibition of the long form IPMS along with affecting the binding of acetyl-CoA to a certain extent. This proposed process may be similar to or mimicked by the Y410F mutant of MtIPMS in the absence of L-leucine in which residues 440 -447 and 457-462 in the subdomain II portion of the linker domain of the Y410F enzyme exhibited an increase in deuterium incorporation from exchange (40). However, the structure of long form IPMS in complex with the substrate acetyl-CoA is awaited to disclose the mechanism in detail.