Conformational interdomain flexibility in a bacterial α-isopropylmalate synthase is necessary for leucine biosynthesis

Abstract

α-Isopropylmalate synthase (IPMS) catalyzes the first step in leucine (Leu) biosynthesis and is allosterically regulated by the pathway end product, Leu. IPMS is a dimeric enzyme with each chain consisting of catalytic, accessory, and regulatory domains, with the accessory and regulatory domains of each chain sitting adjacent to the catalytic domain of the other chain. The IPMS crystal structure shows significant asymmetry because of different relative domain conformations in each chain. Owing to the challenges posed by the dynamic and asymmetric structures of IPMS enzymes, the molecular details of their catalytic and allosteric mechanisms are not fully understood. In this study, we have investigated the allosteric feedback mechanism of the IPMS enzyme from the bacterium that causes meningitis, Neisseria meningitidis (NmeIPMS). By combining molecular dynamics simulations with small-angle X-ray scattering, mutagenesis, and heterodimer generation, we demonstrate that Leu-bound NmeIPMS is in a rigid conformational state stabilized by asymmetric interdomain polar interactions. Furthermore, we found removing these polar interactions by mutagenesis impaired the allosteric response without compromising Leu binding. Our results suggest that the allosteric inhibition of NmeIPMS is achieved by restricting the flexibility of the accessory and regulatory domains, demonstrating that significant conformational flexibility is required for catalysis.

Enzyme catalysis is a structurally dynamic process consisting of multiple steps including substrate binding, transition state stabilization, and product release (1, 2, 3, 4). The activity of an enzyme can be regulated by allostery, which effectively tunes catalysis by ligand-binding remote to the active site. Allostery occurs via altering the accessible dynamic conformational landscape of the enzyme, which is often associated with either conformational changes or subtle dynamic changes (4, 5, 6, 7, 8, 9, 10). Allostery plays a crucial role in the regulation of many biosynthetic pathways (11, 12, 13, 14, 15). In these cases, some allosterically controlled enzymes adopt multidomain assemblies so that the allosteric signal must be transmitted between catalytic and regulatory domains (15, 16, 17, 18, 19). Because of the versatile and complicated structural and biochemical characteristics of multidomain enzymes, the understanding of the molecular underpinnings of crossdomain allosteric communication is limited.

α-Isopropylmalate synthase (IPMS) catalyzes the reaction between acetyl-coenzyme A (AcCoA) and α-ketoisovalerate (α-KIV) to produce CoA and α-isopropylmalate, in the first committed step in the biosynthesis of the amino acid leucine (Leu). The catalytic activity of IPMS is allosterically inhibited by the pathway end product Leu (20, 21). Crystal structures of IPMS from Mycobacterium tuberculosis (MtuIPMS) reveal a homodimeric quaternary structure. Each protomer contains an N-terminal TIM-barrel catalytic domain and a C-terminal regulatory domain that hosts the Leu-binding site. The catalytic and regulatory domains are connected by an accessory domain that is divided into two parts, subdomains I and II (Fig. 1A) (22, 23). The homodimer adopts a domain-swapped arrangement, with the accessory and regulatory domains from one monomer interacting with the catalytic domain of the other monomer (Fig. 1A) (23). The crystal structures of MtuIPMS show marked asymmetry because of different positions adopted by the accessory and regulatory domains relative to the catalytic domain. Superimposition of the two chains at the catalytic domain showed a distance of 62.3 Å between the center of mass of the two regulatory domains (d_reg) (Fig. 1B). One of the active sites is capped by the accessory domain of the other chain and can be described as in a “closed” conformation (Fig. 1C), whereas the other active site is in the “open” conformation with the accessory subdomain II and the regulatory domain positioned at the side of the catalytic domain (Fig. 1C).

Figure 1.

The crystal structure ofMtuIPMS shows an asymmetric homodimer.A, asymmetric domain–swapped dimer of MtuIPMS with Leu bound (Protein Data Bank ID: 3FIG). Chain A is shown in white, chain B is colored to show the constituting domains, including catalytic domain (yellow), subdomain I (blue), and subdomain II (pink) of the accessory domain and the regulatory domain (green). Leu bound in the allosteric sites is displayed as orange spheres. The active site metal ion (Zn²⁺) is shown as a cyan sphere. B, superimposition of the two chains at the catalytic domain reveals large conformational differences. Distance between the center of mass of the two regulatory domain is 62.3 Å. Location of the active site is indicated by the Zn²⁺ ion shown as a cyan sphere. C, the “open” active site with the accessory and regulator domains positioned at the side of the catalytic barrel and the “closed” active site with the accessory and regulatory domains capping the top of the catalytic barrel. The protein chain containing the “closed” or “open” active site is shown in surface representations, and the other protein chain is displayed as cartoons. Chain color scheme is maintained the same as that used in (A). D, superimposition between the α-KIV bound MtuIPMS (Protein Data Bank ID: 1SR9, blue) and Leu-bound MtuIPMS (PDB ID: 3FIG, red). α-KIV, α-ketoisovalerate; MtuIPMS, IPMS from Mycobacterium tuberculosis.

Previous studies have revealed that the accessory domain is essential to the catalytic activities of IPMS enzymes. Catalytic activity was significantly impaired in truncated forms of MtuIPMS, IPMS from Neisseria meningitidis (NmeIPMS) (16, 22), and for the IPMS isoenzymes from Leptospira biflexa (LbiIPMS1 and LbiIPMS2) that lack the regulatory domain and part of the accessory domain (24). The accessory domain was also shown to contribute to the allosteric regulation of IPMS. Amino acid substitution of residues from subdomain I (His379 or Tyr410) and II (Asp444) showed disrupted allosteric response to Leu in MtuIPMS (25, 26). Despite the lack of notable conformational changes between α-KIV-bound or Leu-bound MtuIPMS crystal structures (Fig. 1D), solution-phase amide hydrogen/deuterium exchange experiments revealed a decreased deuterium incorporation percentage of residues 453 to 457 in subdomain II associated with Leu binding, which indicated a shift in conformational equilibria of this region (27).

The dynamic and asymmetric structures of IPMS enzymes pose a challenge for the study of the molecular detail in their catalytic and allosteric mechanisms. Consequently, the role of the accessory domain in propagating the allosteric signals and the contribution the structural asymmetry makes to catalysis and allostery in IPMS is unclear. In this study, we have conducted molecular dynamics (MD) simulations to explore the changes in domain dynamics and identify interdomain polar contacts important for the allosteric mechanism in NmeIPMS. Variant enzymes that contain amino acid substitutions in either one or both chains of the homodimer were generated and characterized structurally and biochemically to verify the role of the accessory domain and asymmetry in the allosteric mechanism of NmeIPMS. Our findings reveal that allosteric inhibition of NmeIPMS by Leu is effected by locking the enzyme in a rigid conformational state stabilized by asymmetric interdomain polar interactions.

Results

Leu binding is associated with changing conformational dynamics of the accessory domains of NmeIPMS

To investigate the changes in the conformational dynamics of NmeIPMS associated with allostery, we conducted MD simulations for ligand-free, KIV-bound, and Leu-bound NmeIPMS. Three replicates of MD simulations were conducted for each system, and the equilibrated periods from the MD simulations were analyzed.

All MD simulations started from the same asymmetric conformation as observed in the crystal structure of MtuIPMS (Protein Data Bank [PDB] ID: 1SR9). The active site of chain A is in the open conformation, and the active site of chain B is capped by the accessory domain of chain A and is in a closed conformation. While all three systems (ligand-free, KIV-bound, and Leu-bound NmeIPMS) remained in an asymmetric dimer conformation during the MD simulations, subtle differences were observed in the positions of the accessory and regulatory domains between the ligand-bound and ligand-free systems; the KIV-bound NmeIPMS sampled slightly less asymmetric conformations (d_reg = 80 ± 5.0 Å) than Leu-bound NmeIPMS (d_reg = 94 ± 6.0 Å) (Table S1). MD average structures of KIV-bound and ligand-free NmeIPMS, when aligned at the catalytic domains, showed that the accessory domain of chain A was placed further away from the top of the active site of chain B in KIV-bound NmeIPMS, than that in ligand-free NmeIPMS, giving rise to a more open chain B active site in the KIV-bound system (Fig. 2A). Interestingly, comparison of MD average structures of Leu-bound and ligand-free NmeIPMS displayed the opposite trend (Fig. 2B). The active site of chain B in Leu-bound NmeIPMS was found in a more closed conformation than the ligand-free system, as the accessory and regulatory domains of chain A moved further in to cap the top of the active site of chain B.

Figure 2.

Changes in conformations and dynamics of the accessory domains of NmeIPMS associated with Leu binding revealed by MD simulations. Overall conformational changes in terms of subdomain positions (A) between ligand-free (yellow) and KIV-bound (blue) NmeIPMS and (B) between ligand-free (yellow) and Leu-bound (red) NmeIPMS. Each domain (catalytic barrel, subdomain 1 and linker, subdomain 2, and regulatory domain) is displayed as a sphere with its position defined by the mass-weighted center of mass of the corresponding domain, and the domains were joined by cylinders. The direction of movement from ligand-free to KIV-bound or Leu-bound conformations is indicated by black dashed arrows. C, fluctuations of the accessory domains were measured as RMSF values of the center of mass of the individual domains during each MD simulation. Individual RMSF values from the three replicate MD simulations are plotted, and the mean of the RMSF values is plotted as a horizontal line. The error bars are calculated as standard deviations. Trajectories were aligned at the N-terminal catalytic domains. KIV, ketoisovalerate; MD, molecular dynamics; NmeIPMS, IPMS from Neisseria meningitides; RMSF, root mean square fluctuation.

The presence of Leu appeared to have also affected the domain dynamics in NmeIPMS (Fig. 2B). Positional fluctuations of the accessory and regulatory domains during MD simulations showed comparable levels of flexibilities between the ligand-free and KIV-bound NmeIPMS. However, Leu-bound NmeIPMS showed to some extent a reduction in fluctuations of subdomain II and the regulatory domain in both chains, indicating the Leu-bound conformations of NmeIPMS were more rigid.

In parallel, we examined the solution phase structures of NmeIPMS in the absence and presence of different ligands (α-KIV or Leu) using size-exclusion chromatography coupled with small-angle X-ray scattering (SEC–SAXS) experiments. SAXS data indicated that there were no significant changes in the average size and overall shape of NmeIPMS in the presence of either α-KIV or Leu, as demonstrated by the similar SAXS profiles and parameters obtained for ligand-free, KIV-bound, and Leu-bound NmeIPMS (Fig. 3A and Table 1). The SAXS data were compared with the theoretical scatterings of the average structures from MD simulations and showed reasonable consistencies between the observed and calculated structural profiles (Fig. S1). Ensemble fitting of the SAXS data using asymmetric MD average conformations and hypothetical symmetric conformations suggested that asymmetric conformations accounted for the majority of NmeIPMS structural populations in all the three experimental conditions (Fig. S2). The predicted SAXS scatterings from MD average structures of ligand-free, KIV-bound, and Leu-bound NmeIPMS were highly similar, which is consistent with the insignificant variations in the experimental scattering profiles. However, flexibility analysis using Porod–Debye transformations showed a structural rigidification of NmeIPMS in the presence of Leu (Fig. 3B) (28), matching the observed decreasing flexibilities of the accessory domains in MD simulations (Fig. 2C).

Figure 3.

Analysis of the SAXS data for ligand-free NmeIPMS and NmeIPMS in the presence of α-KIV (blue) or Leu (red).A, SAXS profiles (log I(q) versus q) and their corresponding Guinier plots (ln I(q) versus q²). B, Porod–Debye plot (q⁴ ∗ I(q) versus q) limited to the range of the SAXS data for which the Guinier linearity is observed. Each SAXS experiment in the presence of different ligands was repeated three times and presented consistent results. α-KIV, α-ketoisovalerate; Leu, leucine; NmeIPMS, IPMS from Neisseria meningitides; SAXS, small-angle X-ray scattering.

Table 1.

SAXS parameters of ligand-free NmeIPMS and NmeIPMS in the presence of different ligand combinations

SAXS parameter	Ligands present
	Ligand-free	α-KIV	Leu
Guinier analysis
Rg (Å)	34.0 ± 0.4	34.7 ± 0.4	35.9 ± 0.5
I(0) (cm−1)	0.03 ± 0.00	0.03 ± 0.00	0.03 ± 0.00
Correlation coefficient, R2	0.99	0.99	0.99
Pair wise distribution analysis
Rg (Å)	34.1 ± 0.1	34.7 ± 0.3	35.8 ± 0.4
I(0) (cm−1)	0.03 ± 0.00	0.03 ± 0.00	0.03 ± 0.00
Dmax (Å)	110.4	113.6	114.3
VP (nm3)	164	153	175
MMporod
MM (kDa)	96	90	103
Number of subunits	2	2	2

Abbreviation: MM_porod, molecular mass estimated from Porod volume.

The uncertainties represent standard deviations estimated by the error propagation from the experimental data.

Leu binding is associated with changing polar interactions involving the accessory domains of NmeIPMS

The network of polar interactions established between different domains and chains of ligand-free, KIV-bound, and Leu-bound NmeIPMS during MD simulations was identified (Tables S2–S4), and the residues involved were mapped onto the corresponding MD average conformations (Fig. 4). In general, a larger number of polar interactions were established in the Leu-bound NmeIPMS that could stabilize its asymmetric conformations, in agreement with the reduced flexibility observed from both the MD simulations and SAXS experiments. A total of 16 polar interactions were established or became more stable in the inhibited Leu-bound NmeIPMS compared with the active KIV-bound NmeIPMS (Table 2). The position of chain A accessory domain over the closed chain B active site in the Leu-bound NmeIPMS was stabilized by clusters of polar interactions (e.g., Tyr313_A–Asp144_B, Ser299_A–Glu236_B, His297_A–Glu236_B, and Arg77_B–Asp304_A). The relative position between accessory and regulatory domains of chain A in Leu-bound NmeIPMS was stabilized by the interaction between Arg310 and Glu503. On the other side of the dimer, the established polar interactions in the Leu-bound NmeIPMS formed a connected interaction pathway from the regulatory domain of chain B to the catalytic domain of chain A via chain B accessory domain, including Arg470_B–Glu353_B, Arg476_B–Glu385_B, Tyr313_A–Asp144_B, Arg35_A–Asp345_B, and Arg32_A–Asp375_B (Fig. 4).

Figure 4.

Polar interactions established with average occupancy above 30% during MD simulations for ligand-free, KIV-bound, and Leu-bound NmeIPMS. Residues involved in the interactions are displayed as spheres and surface representations in light orange. Leu molecules are displayed as green spheres in the Leu-bound structure. The two pairs of residues selected for amino acid substitution experiments are shown in red. KIV, ketoisovalerate; Leu, leucine; MD, molecular dynamics; NmeIPMS, IPMS from Neisseria meningitides.

Table 2.

Hydrogen bonds and salt bridges with occupancy changes more than 30% between Leu-bound and KIV-bound NmeIPMS

H-bond donor	H-bond acceptor	Occupancy (KIV-bound) (%)	Occupancy (Leu-bound) (%)	Occupancy differences (%)	Domains
ARG362-Side (A)	GLU314-Side (A)	0	56	56	Subdomain II–subdomain I
TRP323-Side (A)	GLY291-Main (A)	0	42	42	Subdomain I–catalytic
ARG310-Side (A)	GLU503-Side (A)	0	41	41	Subdomain I–regulatory
ARG470-Side (B)	GLU353-Side (B)	0.19	78	78	Regulatory–subdomain II
ARG327-Side (B)	GLU298-Side (B)	38	84	45	Linker–subdomain I
ARG476-Side (B)	GLU385-Side (B)	0	44	44	Regulatory–subdomain II
ARG35-Side (A)	ASP345-Side (B)	0.72	59	59	Catalytic–subdomain II
THR461-Side (B)	ASP433-Side (A)	1.1	52	51	Regulatory–regulatory
ARG32-Side (A)	ASP375-Side (B)	0.93	46	45	Catalytic–subdomain II
LYS332-Side (A)	GLU18-Side (B)	0	43	43	Linker–catalytic
TYR313-Side (A)	ASP144-Side (B)	0	41	41	Subdomain I–catalytic
TYR180-Side (B)	ASP255-Side (A)	2.6	43	41	Catalytic–catalytic
SER299-Side (A)	GLU236-Side (B)	37	73	36	Subdomain I–catalytic
HSD297-Side (A)	GLU236-Side (B)	20	55	34	Subdomain I–catalytic
LYS332-Side (A)	ASP56-Side (B)	9.2	42	33	Linker–catalytic
ARG77-Side (B)	ASP304-Side (A)	5.5	37	31	Catalytic–subdomain I
ARG476-Side (A)	GLU385-Side (A)	37	0	−37	Regulatory–subdomain II
LYS501-Side (B)	GLU385-Side (B)	50	13	−37	Regulatory–subdomain II

Positive occupancy differences indicate the interactions were formed in Leu-bound NmeIPMS, and negative differences indicate the interactions were broken in Leu-bound NmeIPMS.

Generation of heterodimeric NmeIPMS

To identify the role of the accessory domains in the allosteric mechanism of NmeIPMS, we selected to probe two polar interactions (Arg470_B–Glu353_B and Arg32_A–Asp375_B) in the pathway from the regulatory domain of chain B to the catalytic domain of chain A (Fig. 4). To account for the asymmetry of the interactions formed in the NmeIPMS homodimer, we first developed a protein assembly approach that enabled the generation of heterodimers of NmeIPMS to better pinpoint the effect of proposed substitutions on allosteric inhibition of NmeIPMS.

An expression vector pRSFDuet-1 that contained two independent expression cassettes was utilized, allowing simultaneous expressions of two polypeptide chains mediated by T7 RNA polymerase/promoter. Using restriction digestion and ligation reactions, two genes encoding NmeIPMS ORF that were N-terminally tagged with either a His tag or Strep tag were sequentially cloned into the pRSFDuet-1 vector and transformed into BL21 cells (Fig. 5). The monomeric units with different tags were expressed by IPTG induction and assembled as dimers either homogeneously or heterogeneously in vivo, forming His-tagged homodimer (NmeIPMS_His), Strep-tagged homodimer (NmeIPMS_Strep), and the heterodimer carrying both His and Strep tags (NmeIPMS_His/Strep) (Fig. 5). The dimeric species were separated by passing through two affinity columns, HisTrap and StrepTrap, which were specific to His tag and Strep tag (Fig. 5). Both NmeIPMS_His and NmeIPMS_His/Strep dimers were retained in the HisTrap, whereas NmeIPMS_strep flowed through. The mixture eluted from the HisTrap was further separated by StrepTrap, which retained the heterodimer NmeIPMS_His/Strep. Mass spectrometry revealed the presence of both Strep- and His-tagged monomers. Both tags were then removed from the purified dimeric species by tobacco etch virus (TEV) protease cleavage and SEC.

Figure 5.

Schematic illustration of the homodimeric and heterodimeric assemblies of His-tagged and Strep-tagged NmeIPMS and a two-step purification method using HisTrap and StrepTrap in sequence for separating three dimeric species of NmeIPMS.NmeIPMS, IPMS from Neisseria meningitides.

Two polar interactions involving accessory subdomain II contribute to the communication of allosteric signals in NmeIPMS

The role of the polar interactions Arg470_B–Glu353_B and Arg32_A–Asp375_B in the allosteric mechanism of NmeIPMS were investigated by substituting Arg470 and Arg32 with Gln. Four variant enzymes were generated using the duet protein assembly method, including two heterodimers (NmeIPMS_R470Q^hetero and NmeIPMS_R32Q^hetero) that contained the corresponding substitutions in only one chain and two homodimers (NmeIPMS_R470Q and NmeIPMS_R32Q) that contained the substitutions in both chains. Kinetic properties of the variant enzymes and their inhibition responses to Leu were characterized and compared with the WT enzyme.

The kinetic assays showed that R32Q and R470Q substitutions had a small effect on the catalytic activity of NmeIPMS, except in NmeIPMS_R32Q, where the turnover number (k_cat) was decreased by half (Table 3).

Table 3.

Kinetic characterizations of WT and variant NmeIPMS enzymes and Leu inhibition effect on WT and variant NmeIPMS enzymes

SAXS parameter	KmAcCoA (μM)	KmKIV (μM)	kcat (s−1)	kcat/KmAcCoA (μM−1. s−1)	kcat/KmKIV (μM−1. s−1)	KiApp,AcCoA (μM)	KiApp,KIV (μM)
NmeIPMS	39 ± 5	37 ± 5	12.8 ± 0.3	0.33	0.35	43 ± 4	48 ± 3
NmeIPMSR32Qhetero	35 ± 5	48 ± 6	12 ± 0.5	0.34	0.25	70 ± 4	84 ± 6
NmeIPMSR32Q	40 ± 5	47 ± 3	5.4 ± 0.2	0.14	0.11	7624 ± 2223	10,913 ± 4453
NmeIPMSR470Qhetero	38 ± 4	37 ± 3	15.4 ± 0.3	0.41	0.42	61 ± 3	70 ± 4
NmeIPMSR470Q	48 ± 3	53 ± 6	14.2 ± 0.6	0.30	0.27	5291 ± 1067	4147 ± 1399

To determine apparent K_i values, three Leu concentrations (20, 50, and 100 μM) were applied separately in the kinetic assays where either AcCoA or α-KIV was fixed at 300 μM.

In contrast, R470Q and R32Q substitutions had striking impacts on the allosteric inhibition responses to Leu in the homodimeric variants (NmeIPMS_R470Q and NmeIPMS_R32Q). The apparent inhibitory constants (K_i) of Leu with respect to either AcCoA or α-KIV showed significant increases in NmeIPMS_R470Q and NmeIPMS_R32Q compared with the WT enzyme (Table 3). Interestingly, the heterodimeric variants NmeIPMS_R470Q^hetero and NmeIPMS_R32Q^hetero showed only small increases in K_i values (Table 3), suggesting that maintaining the connected polar interaction pathway on one side of the dimer was sufficient to stabilize the Leu-bound conformation and achieve allosteric inhibition. This observation supported the predicted asymmetry of the polar interaction networks exposed by MD simulations.

The effect of the R32Q or R470Q substitutions on Leu binding was inspected using the NMR technique water-ligand observed via gradient spectroscopy (waterLOGSY) (Fig. S3) (29). Disassociation constants (K_D) calculated from the waterLOGSY data demonstrated that all the four variants maintained similar affinities for Leu to that in the WT NmeIPMS (Table 4 and Fig. S4), suggesting that the loss of allosteric inhibition in the homodimeric variants NmeIPMS_R470Q and NmeIPMS_R32Q was not because of loss of the allosteric binding but was the result of disrupted allosteric signal.

Table 4.

K_D constants showing the affinities of NmeIPMS and its variants for Leu obtained from waterLOGSY assays

NmeIPMS	KD (μM)
NmeIPMS	376 ± 48
NmeIPMSR32Q	352 ± 37
NmeIPMSR32Qhetero	324 ± 65
NmeIPMSR470Q	386 ± 72
NmeIPMSR470Qhetero	370 ± 55

Discussion

The gateway enzyme of the Leu biosynthesis pathway, IPMS, is a dynamic multidomain enzyme with the constituent domains joined by flexible linkers. IPMS adopts a domain-swapped homodimer quaternary structure with asymmetry illustrated by the markedly different relative domain conformations of the two chains in the crystal structures of MtuIPMS (23). Owing to challenges posed to structural studies because of the dynamic and asymmetric nature of IPMS enzymes, the roles of structural dynamics and asymmetry in the allosteric mechanisms are yet to be fully understood.

Previous hydrogen/deuterium exchange experiments on MtuIPMS revealed a shift in conformational equilibria of subdomain II in the presence of the allosteric inhibitor Leu (27). Here, we investigated the allosteric mechanism of NmeIPMS using MD simulations and SAXS experiments. We found that, similar to what was observed for MtuIPMS, allosteric inhibition of NmeIPMS was associated with positional changes and reduced flexibilities of the accessory domains.

MD simulations revealed a network of asymmetric interdomain polar interactions differentially stabilizing the Leu-bound conformation of NmeIPMS. We chose to probe the contributions of two interactions, Arg470_B–Glu353_B and Arg32_A–Asp375_B, by substituting Arg470 or Arg32 with Gln. Using a dual expression protein assembly approach, we generated two homodimeric (NmeIPMS_R470Q and NmeIPMS_R32Q) and two heterodimeric (NmeIPMS_R470Q^hetero and NmeIPMS_R32Q^hetero) variants. We observed that the complete loss of these interdomain interactions was detrimental to the communication of the allosteric signal in the homodimeric variants, as the allosteric response was significantly impaired despite Leu binding not being compromised. Previous studies also identified Tyr410 in MtuIPMS to be essential for allostery (25, 27). Here, the equivalent residue in NmeIPMS, Tyr313, was also found to establish asymmetric interdomain polar interactions in the presence of Leu (Table 2). In general, it appears that the conformational dynamic changes in the accessory domains associated with allostery in NmeIPMS were stabilized by these interdomain polar interactions, and allosteric inhibition was achieved by locking NmeIPMS in the rigid Leu-bound conformational state.

In contrast to the homodimeric variants, which completely lost the ability to establish the interdomain interactions, the heterodimeric variants retained the ability to form the interdomain interactions on one side of the dimer. Interestingly, we observed that the allosteric response to Leu was maintained in these heterodimeric variants, with only small increases in K_i values. These observations suggested that establishing the interdomain polar interactions on one side of the dimer was sufficient to lock the enzyme in the inhibited rigid conformations and underscored the importance of asymmetry in the allosteric mechanism of NmeIPMS.

IPMS is closely related to the enzyme citramalate synthase (CMS) (30), which is the first enzyme in the threonine-independent pathway of isoleucine biosynthesis and is feedback inhibited by the pathway end product Ile (31). Crystal structures of separate catalytic and regulatory domains of CMS exhibit high structural similarities to those of MtuIPMS (31, 32); however, the full structure of CMS is yet to be obtained. Based on sequence similarity, CMS likely adopts similar domain organizations and asymmetric conformations as IPMS, with the dynamic accessory domain between the catalytic TIM barrel and the regulatory domain that binds Ile. Therefore, we predict that CMS may also adopt a very similar allosteric mechanism to IPMS. Another closely related enzyme to IPMS is homocitrate synthase (HCS), which catalyzes the first step in lysine biosynthesis (33, 34). HCS is dimeric and consists of only catalytic TIM barrel domains and flexible accessory domains but lacks the regulatory domains. HCS is feedback regulated by the pathway end product lysine by competitive inhibition at the active site (33, 35, 36, 37). Interestingly, the dimeric structures of HCS appear to be symmetric, which suggests that appending regulatory domain dimers to the accessory domains likely leads to the asymmetry that is integral to the allosteric mechanisms in IPMS and CMS.

In summary, here we showed that the interdomain flexibility between the accessory and the catalytic domains is key to IPMS catalysis. We probed interactions that communicate allosteric signal to the active site of IPMS using mutagenesis and demonstrated that the contacts formed in the asymmetric conformation are manifestly important for allostery. Allostery in IPMS exploits the vulnerability of the enzyme to asymmetric conformations, and inhibition is achieved by putting constraints on the accessible conformations via stabilizing interdomain interactions.

Experimental procedures

Construction of plasmids

The expression plasmid carrying the gene encoding NmeIPMS with a cleavable N-terminal His tag, pFH01, was already available from previous work (22). To construct duet NmeIPMS expression plasmid, a two-step gene cloning was conducted using pFH01 as PCR template to insert NmeIPMS gene into multiple cloning sites (MCSs) 1 and 2 of pRSFDuet1 (Merck) sequentially. In the first step, NmeIPMS gene was amplified via two rounds of PCR using two pairs of primers, Duet_I_FW1/Duet_I_RV and Duet_I_FW2/Duet_I_RV (Table 5), in sequence to introduce an N-terminal TEV protease cleavage site (in bold) and BamHI and HindIII restriction sites (underlined). After gel purification, the final PCR product was doubly digested by BamHI and HindIII and then ligated into the MSC 1 and transformed into Escherichia coli DH5α-competent cells. The recombinant pRSFDuet1-MSC1-NmeIPMS plasmid was isolated from the transformant screened by LB agar medium containing 100 μg/ml kanamycin and verified by gene sequencing. It is worth noting that pRSFDuet1 contains a gene sequence encoding 6× His at the upstream of MSC 1, thus His-tag sequence was not included in the Duet_I_FW2 primer. In the second step, NmeIPMS gene was amplified and incorporated with a sequence encoding an N-terminal Strep tag (in italic), a TEV cleavage site (in bold) between the Strep tag and NmeIPMS gene, and restriction sites, BglII and XhoI (underlined). To achieve this, two rounds of PCR were carried out sequentially using two primer pairs, Duet_II_FW1/Duet_II_RV and Duet_II_FW2/Duet_II_RV (Table 5). Following gel purification, BglII/XhoI digestion, ligation and transformation as described in step 1, the PCR product was inserted into MCS 2 of pRSFDuet1-MSC1-NmeIPMS. The obtained recombinant construct (pRSFDuet1-NmeIPMS) carries two NmeIPMS ORFs, respectively, at two MCS sites. As well, the ORFs at MCS 2 of pRSFDuet1-NmeIPMS were sequence verified.

Table 5.

Primers for construction of expression vectors

Abbreviations: GGATCC, BamH I restriction site; AAGCTT, Hind III restriction site; AGATCT, Bgl II restriction site; CTCGAG, Xho I restriction site.

The designated mutations are highlighted in red.

Single mutations, R32Q or R470Q, were introduced into the His-tagged NmeIPMS ORF at MSC 1 of pRSFDuet1-NmeIPMS using overlap extension-PCR–based site-directed mutagenesis and a pair of mutagenic primers containing the designated mutation (in red). Three rounds of PCR were included in the mutagenesis procedure, two of which were carried out using pFH01 as the template, and respectively, using two primer pairs, Duet_I_FW2/reverse mutagenic primer (Duet_R32Q_RV for R32Q and Duet_R470Q_RV for R470Q) and forward mutagenic primer (Duet_R32Q_FW for R32Q and Duet_R470Q_FW for R470Q)/Duet_I_RV. The resulting two PCR products having an overlapping segment were purified and mixed as the template for the final PCR using primers, Duet_I_FW2/Duet_I_RV. This PCR generated a full-length gene sequence encoding NmeIPMS with the designated mutation, which is then purified and cloned into the MCS 1 of pRSFDuet1-NmeIPMS to substitute for the WT NmeIPMS ORF. The plasmid carrying the mutation variant R32Q or R470Q at MCS 1 (pRSFDuet1-R32Q/WT or pRSFDuet1-R470Q/WT) was sequence verified.

Protein expression and purification

Plasmids pFH01, pRSFDuet1-NmeIPMS, pRSFDuet1-R32Q/WT, and pRSFDuet1-R470Q/WT were transformed into E. coli BL21 (DE3) competent cells for protein expression. Cultures were grown at 37 °C in LB medium supplemented with 100 μg/ml ampicillin for the cell strain containing pFH01 or 100 μg/ml kanamycin for the cells transformed by duet-expression plasmids. Expression of target protein was induced by the addition of 0.5 mM IPTG into the culture at midexponential phase, which is followed by further overnight growth at 23 °C.

Cells were harvested by centrifugation (12,000g for 15 min at 4 °C) and resuspended in lysis buffer (20 mM sodium phosphate buffer [pH 8.0], 500 mM NaCl, 20 mM imidazole, and 2 mM DTT). Cells were lysed by sonication and clarified by centrifugation (40,000g for 40 min at 4 °C).

Standard purification of NmeIPMS

The clarified cell lysate containing WT His-tagged NmeIPMS encoded by pFH01 was filtered and loaded to a HisTrap HP column (GE Healthcare), pre-equilibrated with binding buffer (20 mM sodium phosphate buffer [pH 8.0], 500 mM NaCl, and 20 mM imidazole), and then eluted with elution buffer (20 mM sodium phosphate buffer [pH 8.0], 500 mM NaCl, and 500 mM imidazole). The eluted fractions were pooled together and desalted, which was followed by overnight treatment of TEV protease at 4 °C to cleave the His tag. After TEV cleavage, the protein was reapplied to the HisTrap column to remove the TEV protease, the His tag, and any noncleaved protein. The untagged proteins were further purified by SEC (HiPrep 16/60 Sephacryl S-200 HR column; GE Healthcare) using a buffer consisting of 20 mM Bis–Tris propane buffer (pH 8.5), 200 mM NaCl, and 10 mM MgCl₂.

Separation of dimeric species expressed by duet expression plasmids

The crude cell extract containing the expression products of pRSFDuet1-NmeIPMS or pRSFDuet1-R32Q/WT or pRSFDuet1-R470Q/WT was subjected to a HisTrap to retain the His-tagged dimeric species following the standard procedure as described previously. The elution fraction from HisTrap was transferred to Strep-Tactin binding buffer (20 mM sodium phosphate buffer [pH 7.4], 280 mM NaCl, and 6 mM KCl) and subsequently loaded to a StrepTrap column HP (GE Healthcare) pre-equilibrated with Strep-Tactin binding buffer. The homodimeric His-tagged NmeIPMS or variants flew through StrepTrap column, whereas the heterodimeric species carrying a His tag and a Strep tag was retained in the column, which was then eluted by the Strep-Tactin binding buffer supplemented with 2.5 mM desthiobiotin. The separated homodimeric and heterodimeric species were individually treated by TEV protease and subjected to SEC in sequence following the standard purification procedure for removal of tags and further purification.

Kinetic assay

The rate of NmeIPMS catalyzed reaction was measured by a dithiodipyridine (DTP)-coupled assay, which uses DTP to monitor the formation of CoA product at 324 nm (ε = 1.98 × 10⁴ l × M⁻¹ cm⁻¹). All kinetic assays were carried out in 1 cm path length quartz cuvettes at 25 °C and measured by a Varian Cary 100 UV–visible spectrophotometer. The typical reaction condition contained 100 μM DTP, 20 mM KCl, 20 mM MgCl₂, and 3.0 × 10⁻² μM enzyme in 50 mM HEPES buffer (pH 7.5), and all reactions were initiated by the addition of α-KIV. To determine apparent kinetic parameters of WT NmeIPMS and variant enzymes, the α-KIV concentration was fixed at 300 μM, and the concentration of AcCoA was fixed at 300 μM, while carrying out assays with varied concentrations of the other substrates.

Leu inhibition assay

The inhibitory effect of Leu on activities of WT NmeIPMS and variant enzymes was characterized under the standard kinetic reaction condition. To determine the apparent K_i value for Leu with respect to either substrate, kinetic assays were conducted at three Leu concentrations (20, 50, and 100 μM) separately where the α-KIV concentration was fixed at 300 μM or the AcCoA concentration was fixed at 300 μM, whereas varied the concentration of the other substrate. The recorded data with respect to either α-KIV or AcCoA were analyzed by Prism and fits to noncompetitive inhibition model well.

SAXS

SAXS measurements were conducted at the Australian Synchrotron SAXS/WAXS beamline equipped with a Pilatus detector (1 M, 170 × 170 mm, effective pixel size, 172 × 172 mm) (38, 39, 40). The applied wavelength of X-ray was 1.0332 Å, and a sample detector at a distance of 1.6 m from the sample, which recorded scattered X-ray intensities covering the q range of 0.0015 to 3.0 Å⁻¹. The q is the magnitude of the scattering vector, which is defined as q = (4π/λ) sinθ, where λ is the wavelength and 2θ is scattering angle. The WT NmeIPMS was eluted from a SEC column (Superdex increase 200 5/150; GE Healthcare) using the ligand-free SEC–SAXS buffer (20 mM Bis–Tris propane buffer [pH 8.5], 200 mM NaCl, 10 mM MgCl₂, and 3% [v/v] glycerol) or the buffers, respectively, containing 500 μM α-KIV and Leu. The elution traces in all applied conditions showed nicely consistent single peaks (Fig. S5). The eluted sample passed into a 1.0 mm thin-walled glass capillary where the sample was detected using X-rays at 25 °C at 1 s intervals. The reduction and buffer subtraction of raw scattering data was performed by Scatterbrain developed at the Australian Synchrotron. Processed data were plotted as I(q) versus q and analyzed using Primus (Biological Small Angle Scattering Group) (41). Each SAXS dataset containing ∼650 data frames over the range 0.005 ≤ q ≤ 0.35 Å⁻¹ where a single symmetric scattering peak over ∼80 frames was observed. To verify conformational homogeneity, three groups of frames, respectively, covering the maximum scattering intensity and either side of the scattering maximum were summed and compared. The same scattering profile was observed across each scattering peak. Guinier plots (lnI(q) versus q²) of scattering data showing good linearity in the q range of q ≥ 1.3/R_g demonstrated good monodispersity of samples and evaluated the R_g of the protein particle. The Porod volume (V_p) and the maximum distance (D_max) of the protein particle were evaluated, respectively, from the Porod plot and the pair-distance distribution (P(r)) function obtained via Fourier transformation. MM_porod (molecular mass estimated from Porod volume) was obtained by dividing the Porod volumes by 1.7.

WaterLOGSY

WaterLOGSY can be used to detect ligand–protein binding via ¹H NMR observation of small-molecule ligand (29, 42). In solution, proton (¹H) magnetization of small-molecule ligand can be transferred directly from excited bulk water molecules or indirectly from excited proton at protein surface through protein–ligand complex. Direct and indirect magnetization transferring, respectively, yields positive and negative nOe signals, which is taken advantageous of by waterLOGSY to quantify the ligand bound to protein for further analysis of protein–ligand affinity.

WaterLOGSY spectra were recorded at a ¹H frequency of 600 MHz using a JEOL 600 MHz Nuclear Magnetic Resonance Spectrometer type JNM-ECZ600R with a 5 mm FG/RO Digital Auto Tune Probe, at 25 °C. A 5 mm diameter NMR tube with a sample volume of 600 μl was used in all experiments. Samples consisted of 10 mM Hepes (pH 8.0), 3 mM MgCl₂, 3 mM KCl, 10% (v/v) D₂O, and Leu at varied concentrations (0, 50, 150, 250, 500, and 750 μM). WaterLOGSY experiments were conducted using the pulse sequence without a water flipback pulse as described by Dalvit et al. (29). The parameter used was a 30 ms pulse length with an initial water selective 180° REBURP pulse. W5 composite inversion pulses were used for a double-pulsed field gradient spin echo WATERGATE water signal suppression. The mixing time was 1 s and a 2 s relaxation delay before each scan was used. A T1p Filter was used to suppress the protein signals.

WaterLOGSY spectrum for each sample was recorded in the absence of enzyme first, which was followed by a second measurement with the addition of enzyme (WT NmeIPMS, or NmeIPMS_R32Q or NmeIPMS_R32Q^hetero, or NmeIPMS_R470Q or NmeIPMS_R470Q^hetero) with a final concentration of 10 μM. Subtraction of the two waterLOGSY signal intensities recorded, respectively, in the absence and presence of enzyme at varied Leu concentration resulted in the Leu-binding isotherm. K_D value was obtained by fitting the binding isotherm to the equation as follows (43, 44, 45):

$$I=\frac{I_{\max }+\left[L_T\right]}{K_D\times \left[L_T\right]}$$

where I denoted changes in corrected waterLOGSY intensity resulted by titrations. I_max indicates the maximum intensity change. [L_T] is the concentration of titrated ligand.

Homology model and MD simulations

The full structural model of KIV-bound NmeIPMS was first generated by homology modeling using Prime (Schrödinger) (46, 47). The crystal structure of KIV-bound MtuIPMS (PDB ID: 1SR9) was used as the template, which shares 31% homology to the amino acid sequence of NmeIPMS. The metal ion in the active site of NmeIPMS was modeled as Mg²⁺ (whereas in MtuIPMS crystal structure, Zn²⁺ ions were bound). The α-KIVs in the active site of MtuIPMS were kept in the homology model, and metal coordination to α-KIVs observed in MtuIPMS was retained. Ligand-free NmeIPMS model was then obtained by deleting the α-KIVs in the active sites but keeping the Mg²⁺ ion. Leu-bound NmeIPMS was generated by merging the Leu molecules bound in the regulatory domain of MtuIPMS (PDB ID: 3FIG) onto corresponding Leu-binding sites in the ligand-free NmeIPMS homology model. The homology models adopted the same asymmetric conformation as in the crystal structure of MtuIPMS and align to the crystal structure of MtuIPMS (PDB ID: 1SR9) with an RMSD value of 1.2 Å (aligning backbone atoms). The catalytic domains in the homology models compared well to the crystal structure of the isolated catalytic domains of NmeIPMS (PDB ID: 3RMJ), with an RMSD value of 2.2 Å (aligning backbone atoms of residues 1–297).

MD simulations were conducted using NAMD 2.12 (Theoretical and Computational Biophysics Group at the University of Illinois), and trajectories were visualized and analyzed in VMD. The NmeIPMS systems were solvated in a water box with explicit TIP3P water molecules, and the charges were neutralized by adding Na⁺ and Cl⁻ ions. Initial force field topology and parameters for α-KIV were obtained from CGenFF server (https://cgenff.paramchem.org) (48, 49, 50, 51) and further refined by Force Field Toolkit (52). MD simulations were conducted with CHARMM36 force field (53) at a constant temperature of 310 K and pressure of 1 atm. For MD simulations of ligand-free and KIV-bound NmeIPMS, the systems were first minimized for 5000 steps followed by dynamic production runs at 2 fs time steps. For Leu-bound NmeIPMS, the simulation system was first relaxed with two phases of short MD simulations. The first phase consisted of a series of 20 simulations, each started with 5000 steps of minimization followed by 0.1 ns of dynamics simulation. The second phase of relaxation consisted of 80 simulations, each started with 5000 steps of minimization followed by 1 ns of dynamics simulation. The production run for Leu-bound NmeIPMS was started after the relaxation runs. Three replicates of MD simulations were conducted for ligand-free (310.9, 357.4, and 239.9 ns), KIV-bound (360.7, 298.8, and 239.1 ns), and Leu-bound (332.4, 282.3, and 191.4 ns for production runs) NmeIPMS. The RMSD values of protein backbone atoms during the MD simulations were plotted, and the equilibrated periods were identified for analyses (Fig. S6). A total of 648.2, 563.6, and 606.1 ns of equilibrated trajectories were obtained for ligand-free, KIV-bound, and Leu-bound NmeIPMS systems, respectively. It should be noted that the lattice contacts in the crystal structure of MtuIPMS that may have stabilized the packing of the asymmetric conformation were neither in the NmeIPMS homology models nor in the MD simulations; therefore, the observed asymmetric conformations from MD simulations were not likely to be biased by those lattice contacts.

Data availability

The data for the experimental work described herein are available on request by email to emily.parker@vuw.ac.nz.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Edited by Wolfgang Peti