Crystal structure of homoisocitrate dehydrogenase from Schizosaccharomyces pombe

Abstract

Homoisocitrate dehydrogenase (HICDH) catalyzes the conversion of homoisocitrate to 2-oxoadipate, the third enzymatic step in the α-aminoadipate pathway by which lysine is synthesized in fungi and certain archaebacteria. This enzyme represents a potential target for anti-fungal drug design. Here, we describe the first crystal structures of a fungal HICDH, including structures of an apoenzyme and a binary complex with a glycine tri-peptide. The structures illustrate the homology of HICDH with other β-hydroxyacid oxidative decarboxylases and reveal key differences with the active site of Thermus thermophilus HICDH that provide insights into the differences in substrate specificity of these enzymes.

Introduction

Lysine biosynthesis in fungi, euglena and certain archaebacteria occurs through the α-aminoadipate pathway.¹ Enzymes in the first steps of this pathway have been proposed as potential targets for the development of antifungal therapies, as they are absent in animals but are conserved in several pathogenic fungi species, including Candida, Cryptococcus and Aspergillus.² One potential antifungal target in the α-aminoadipate pathway is the third enzyme in the pathway, homoisocitrate dehydrogenase (HICDH), which catalyzes the divalent metal-dependent conversion of homoisocitrate to 2-oxoadipate (2-OA) using nicotinamide adenine dinucleotide (NAD⁺) as a cofactor¹.

HICDH belongs to a family of β-hydroxyacid oxidative decarboxylases that includes malate dehydrogenase, tartrate dehydrogenase, 6-phosphogluconate dehydrogenase, isocitrate dehydrogenase (ICDH) and 3-isopropylmalte dehydrogenase (IPMDH). ICDH and IPMDH are well-characterized enzymes that catalyze the decarboxylation of isocitrate to yield 2-oxoglutarate (2-OG) in the citric acid cycle and the conversion of 3-isopropylmalate to 2-oxoisovalerate in the leucine biosynthetic pathway, respectively.³ Recent structural and biochemical studies of HICDH reveal that this enzyme shares sequence, structural, and mechanistic homology with ICDH and IPMDH.^3-7 To date, the only published structures of HICDH are from the archaebacteria Thermus thermophilus (TtHICDH).^6,8 Fungal HICDHs diverge from TtHICDH in several aspects, including their thermal stability, oligomerization state and substrate specificity, thus warranting further characterization. To gain insights into these differences, we determined crystal structures of a fungal Schizosaccharomyces pombe HICDH (SpHICDH) as an apoenzyme and as a binary complex with the additive tripeptide glycyl-glycyl-glycine (GGG) to 1.55 Å and 1.85 Å resolution, respectively. Finally, a comparison of the SpHICDH and TtHICDH structures reveal differences in their active sites that help explain the variations in their respective substrate specificities.

Materials and Methods

Cloning, expression and purification

Full length HICDH encoded by the Lys12⁺ gene in S. pombe was amplified from the genomic clone SPAC31G5.04 (Sanger Institute) and was subcloned into a modified pET15b N-terminal 6xHIS-tag expression vector (Novagen) that contains a tobacco etch virus protease cleavage site. The enzyme was overexpressed in the Rosetta 2 strain (EMD Biosciences) by induction with 0.1 mM isopropyl β-D-1-thiogalactopyranoside at 22 °C for 16 h. Cells were resuspended in 50 mM sodium phosphate, pH 7.0, 500 mM NaCl, 5 mM β-mercaptoethanol and were lysed with lysozyme followed by sonication. Soluble enzyme was purified on a Talon (Clontech) Co(II) column and the 6xHIS tag was removed by incubating with tobacco etch virus protease during dialysis overnight against 50 mM sodium phosphate, pH 7.0, 150 mM NaCl and 5 mM β-mercaptoethanol. The His-tagged protease was removed by passing the sample over a second Talon Co(II) column. SpHICDH was further purified by gel filtration via isocratic elution with 50 mM sodium phosphate, pH 7.0, 150 mM NaCl and 1 mM tris(2-carboxyethyl)phosphine on a Superdex 200 column (GE healthcare). The protein was concentrated to 100-120 mg/ml (A₂₈₀, ε =17,420) and stored at -80 °C.

Crystallization, data collection and structure determination

Crystals of SpHICDH were initially obtained by vapor diffusion in a condition with 0.2 M lithium acetate and 20% (w/v) PEG 3350. The Additive Screen HT (Hampton Research) was used to determine that the addition of 30 mM GGG improved crystal quality. Optimization was performed at 20 °C in hanging drops composed of 1.5 μL of 25 mg/mL SpHICDH, 1.5 mM 2-OA, 60 mM GGG as an additive and 1.5 μL crystallization solution (0.22 M lithium acetate and 18% (w/v) PEG 3350). SpHICDH was also crystallized by hanging drop vapor diffusion using 1 μL of 15-24 mg/mL SpHICDH, 2 mM MgCl₂, 6.6 mM 2-OA and 2 mM NAD⁺ in 1 μL 0.1-0.25 M lithium acetate, 12-16% PEG 3350. Crystals were cryo-protected in their crystallization solution supplemented with 20% glycerol and subsequently frozen in liquid nitrogen. Diffraction data were collected at LS-CAT, beamline 21ID-G at the Advanced Photon Source (Argonne National Labs), and data sets were integrated and scaled using HKL2000⁹ (Table I). A model of SpHICDH was generated in Chainsaw¹⁰ from the sequence alignment of SpHICDH and TtHICDH and the coordinates of the TtHICDH structure (PDB 1X0L),⁶ which was subsequently used as a search model for molecular replacement in MOLREP.¹¹ The model of SpHICDH bound to GGG was built in Coot¹² and refined in REFMAC¹³ using TLS refinement.¹⁴ This structure was used as a model for molecular replacement for SpHICDH apoenzyme crystallized in the presence of MgCl₂, 2-OA and NAD⁺, which was built and refined as described for the SpHICDH-GGG complex. The final models were validated using Molprobity¹⁵ (Table I). Root mean square deviations (RMSDs) were calculated for Cα atoms using SSM Superpose in Coot.¹² A simulated annealing omit map of GGG was generated in CNS¹⁶ and structural figures were created using MacPyMOL (The PyMOL Molecular Graphics System, Schrödinger, LLC.)

Table I. Data Collection and Refinement Statistics.

	SpHICDH (apoenzyme)	SpHICDH (GGG complex)
Data Collection
Beamline	21-ID-G	21-ID-G
Space group	P21	P21
Cell dimensions
a, b, c, (Å)	52.85, 94.45, 76.08	53.24, 92.86, 75.47
β (°)	106.6	106.7
Wavelength (Å)	0.97856	0.97856
Resolution (Å)a	25.0-1.55 (1.61-1.55)	20.0-1.85 (1.89-1.85)
Rmerge (%)b	5.5 (45.9)	7.2 (41.5)
I/σI	30.2 (2.9)	22.1 (2.7)
Total Reflections	713,906	380,206
Unique Reflections	103,345	60,681
Redundancy	6.9 (4.3)	6.3 (3.9)
Completeness (%)	99.3 (94.2)	99.5 (94.7)
Refinement
Resolution range (Å)	24.86-1.55	19.70-1.85
No. of reflections	97,599	56,698
No. of atoms	6084	5761
Protein atoms	5524	5371
Ligand/Glycerol	-/60	26/24
Water atoms	500	340
Rwork/Rfreec	18.8/21.5	19.3/23.0
Average B-factors (Å2)
Overall	22.1	26.5
Protein	22.2	26.7
Ligand/Glycerol	--/24.6	43.8/32.6
Water	20.9	21.5
RMSDs
Bond length (Å)	0.012	0.014
Bond angles (°)	1.409	1.413
MolProbity scored	1.32	1.49
Percentilee	96th	95th
Ramachandran favored (%)	97.3	97.2
Ramachandran allowed (%)	2.7	2.8
Ramachandran outliers (%)	0	0

^aValues in parentheses are for the highest-resolution shell.

^bR_merge = Σ_hkl Σ_j |I_j - I |/Σ_hklΣ_j I_j where I is the mean intensity of reflection hkl.

^cR_work = Σ_hkl ||F_o|- |F_c ||/Σ_hkl |F_o|; R_free = 5.1% of the total reflections.

^d0.42574*log(1+clashscore) + 0.32996*log(1+max(0,pctRotOut-1)) + 0.24979*log(1+max(0,100-pctRamaFavored-2)) + 0.5.

^eFor structures with a resolutions ± 0.25 Å of the resolution of each SpHICDH structure. 100^th percentile is the best; 0^th percentile is the worst.

Results and Discussion

Overall structure of SpHICDH

The crystal structures of SpHICDH apoenzyme and complex with GGG were determined by molecular replacement to 1.55 Å and 1.85 Å, respectively (Table I). The quaternary structure of SpHICDH consists of domain-swapped monomers [Fig. 1(A)] with density observed for residues 5-360 and 4-360 of each monomer in the apoenzyme and residues 5-362 of monomer A and 5-53 and 57-359 of monomer B in the GGG-bound complex. Each SpHICDH monomer is composed of a large domain (residues 1-101 and 266-362), a small domain (residues 102-129 and 156-265), both of mixed α/β topology, and a clasp domain¹⁷ forming an anti-parallel β-sheet at the dimer interface (residues 130-156) [Fig. 1(A)]. The secondary structure of a monomer of SpHICDH adopts a modified Rossmann fold¹⁷ with ten β-strands (β1-β5 and β8-β12) creating a mixed parallel and anti-parallel β-sheet [Fig. 1(B)]. The remaining two β-strands (β6-β7) compose the clasp domain, which interacts with the adjacent monomer. Additional interactions at the dimer interface are predominantly mediated through the formation of a four-helix bundle between α7 and α8 from both monomers. This fold is homologous to other β-hydroxyacid oxidative decarboxylases including ICDHs (RMSDs 1.77 Å and 2.68 Å for PDB entries 1AI2 and 2QFV, respectively),^18,19 IPMDHs (RMSDs 1.92 Å and 1.84 Å for PDB codes 1IPD and 1A05, respectively)^20,21 and TtHICDH (RMSD 1.23 Å for PDB entry 1X0L)⁶.

Figure 1. Crystal structure of SpHICDH.

(A) Ribbon diagram of a homodimer of SpHICDH with the large domain, small domain and clasp domain of monomer A depicted in dark blue, light blue and cyan, and of monomer B in dark orange, orange, and light orange, respectively. (B) Cartoon representation of the secondary structure of a monomer of SpHICDH with secondary structure elements labeled. (C) Ribbon diagram illustrating the superimposition of a monomer of SpHICDH (colored as described for monomer A in (A)) with those of EcICDH (PDB ID 1AI2, light orange), TfIPMDH (PDB: entry 1A05, orange) and TtHICDH (PDB code 1X0L, gray). Rearrangement of the 3₁₀-helix between the SpHICDH and TtHICDH apoenzyme structures is circled.

A structural comparison of SpHICDH with Escherichia coli ICDH (EcICDH) in a complex with isocitrate and NADPH¹⁸, Thiobacillus ferrooxidans IPMDH (TfIPMDH) bound to isoporpylmalate²⁰ and TtHICDH⁶ illustrates similarities and differences among these enzymes [Fig. 1(C)]. Each of these enzymes adopts a closed conformation. The closed and open forms of these enzymes refers to the width of the NADH binding pocket, which is located in a cleft between the large and small domains below the active site [Fig 1(A)]. For example, the EcICHD apoenzyme has been crystallized in the open and closed states, whereas substrate-bound EcICHD adopts a closed form, which is a result of a ∼16° rotation of the large domain relative to the small domain that narrows the NADH binding cleft^17,22. The major difference between SpHICDH and other β-hydroxyacid oxidative decarboxylases is the conformation of the clasp domain and the loop region following β4. The 3₁₀-helix preceding the clasp also differs in its orientation between SpHICDH and TtHICDH apoenzyme (discussed below).

Active site of SpHICDH

The active site of SpHICDH is located in a cleft between the small and large domains of one monomer, and also contains residues from the small domain of the neighboring monomer. [Fig. 1(A)]. Although 2-OA, NAD⁺ and Mg(II) were included in the crystallization condition, the ligands and metal ion were not observed in the electron density maps. Despite their absence, the apoenzyme's active site adopts a conformation homologous to the closed state EcICDH bound to isocitrate and NAD⁺,¹⁸ Saccharomyces cerevisiae mitochondrial NADP-dependent ICDH (ScICDH) in a complex with isocitrate¹⁹, and TfIPMDH bound to isopropylmalate.²⁰ Similarly, the TtHICDH apoenzyme also displays a closed state⁶ [Fig. 1(C)], suggesting that HICDH may not undergo an extensive active site conformational change as observed for IPMDH^23,24 and ICDH enzymes.^17,19,22 Alternatively, the crystallization conditions for the HICDH enzymes may simply force the enzyme into a closed state. Additional structures of the HICDH apoenzyme and complexes with substrates are needed to resolve this question.

The residues in the active site responsible for substrate binding and catalysis have been identified through homology with ICDH and IPMDH and by biochemical characterization of S. cerevisiae HICDH (ScHICDH).³ A lysine-tyrosine pair composed of Lys-196′ and Tyr-133 in SpHICDH (the prime (′) denotes a residue in the adjacent monomer) have been implicated in the acid-base catalyzed reaction that converts homoisocitrate to 2-OA, based on kinetic studies of ScHICDH [Fig. 2(A)].⁵ Additionally, three conserved aspartate residues (Asp232′, Asp256 and Asp260) have been proposed to coordinate the divalent-metal ion necessary for catalysis.⁶ Because SpHICDH adopts a closed conformation in the absence of substrates, its structure can be superimposed with that of an EcICDH-isocitrate complex²⁵ to gain insight into HICDH substrate recognition. Homoisocitrate is predicted to bind in a similar orientation as isocitrate, with its C1 carboxylate group and 2-hydroxyl groups coordinated to the active site metal ion. Three structurally conserved arginine residues (Arg97, Arg107 and Arg126) further stabilize the substrate through interactions with its C1 and C6 carboxylate groups [Fig. 2(A)]. To visualize the binding position of the second substrate, NAD⁺, a product complex of HICDH bound to 2-OA, NADH and Mg(II) was modeled by aligning the active site residues to that of a product complex of mitochondrial ScICDH bound to NADPH and 2-OG¹⁹ [Fig 2(B)]. The modeled NADH binds in the active site cleft, adjacent to the product 2-OA, and is flanked by α9 and the loop between β3 and α4 of one monomer and α7 of the adjacent monomer [Fig. 1(B) and Fig. 2(B)]. The modeled NADH fits well into the active site cleft of SpHICDH; however, there is a clash between the ribose hydroxyl groups of the nicotinamide mononucleotide of NADH and Ser81 located in the β3-α4 loop of the protein, suggesting that a conformational change of this loop is required for coenzyme binding. In summary, the structural alignments illustrate that the active site residues of SpHICDH responsible for substrate binding and catalysis are conserved with the residues in other β-hydroxyacid oxidative decarboxylases.

Figure 2. Active site of SpHICDH.

(A) Stick diagram of the active site of apo-SpHICDH with isocitrate and Mg(II) modeled from a superimposition with isocitrate dehydrogenase (PDB ID 5ICD). Isocitrate is modeled in stick configuration with yellow carbon atoms and the Mg(II) ion is shown as a green sphere. (B) A model of the product complex of SpHICDH. A surface representation of SpHICDH monomer A (orange) and monomer B (blue) illustrates the NAD⁺ binding cleft. The NADH and 2-OA products were modeled by aligning apo-SpHICDH with a product complex of ScICDH bound to NADPH and 2-OG (PDB code 2QFX). The Mg(II) ion is modeled as a green sphere, and 2-OA and NADH are illustrated with violet and green carbon atoms, respectively. (C) Stick diagram of the active site of SpHICDH bound to a GGG peptide, (green carbons). Electron density of the F_o-F_c simulated annealing omit map (contoured to 2.0 σ̤around GGG peptide is shown in gray. Water molecules and hydrogen bonds are shown as red spheres and green dashes, respectively. (D) Overlay of the active site of apo-SpHICDH (blue), GGG-bound SpHICDH (orange) and TtHICDH (gray). Residues in SpHICDH and TtHICDH are labeled in blue and gray, respectively.

Unexpectedly, we observed that the tri-peptide additive GGG binds in the active site of SpHICDH. Two conserved arginine residues (Arg97 and Arg126), which are proposed to interact with homoisocitrate, form contacts with GGG through a network of direct and water mediated hydrogen bonds [Fig. 2(C)]. The tri-peptide is further stabilized by hydrogen bonding to the backbone atoms of residues Ser85, Val87 and Gly89, which are located in the loop between β3 and α4. Additional hydrogen bonds occur between the α-amine group of GGG and Glu285 and Pro286. Together, these interactions orient the GGG tri-peptide into a similar location as the modeled NADH and 2-OA products [Fig. S1]. Specifically, the backbone carbonyl of the second glycine is nearly superimposable with the modeled nitrogen in the ring of the nicotinamide moiety of NADH. Additionally, the C-terminal carboxylate of GGG is positioned between the amide nitrogen in the nicotinamide group of NADH and the C1 carboxylate of 2-OA, respectively.

Comparison of SpHICDH to TtHICDH

A comparison of the SpHICDH and TtHICDH structures reveal some key differences between the fungal and archaeal enzymes. First, the two enzymes exhibit distinct oligomerization states. SpHICDH eluted as a homodimer in size exclusion chromatography (data not shown) and crystallized in the same oligomerization state (Table I). In contrast, TtHICDH functions as a homotetramer, with the clasp domain of two dimers forming hydrophobic inter-subunit interactions.^6,7 This tetramerization state is presumably unique to archaeal HICHDs and may enhance the thermostability of the enzyme, as a mutation that disrupts tetramerization exhibits a lower thermal inactivation temperature.⁶ Moreover, the active site conformation differs between SpHICDH and TtHICDH apoenzyme despite their homologous structures [Fig. 1(C) and Fig. 2(D)]. In TtHICDH, the active site adopts a non-productive conformation in which Tyr125 is flipped out of the active site cleft [Fig. 2(D)]. This observation lead Miayzaki et al. to speculate that either substrate binding is coupled to a conformational change in the active site or the TtHICDH structure represents a non-physiologically relevant structure, possibly induced by the low pH of the crystallization condition.⁶ Conversely, in SpHICDH, the corresponding tyrosine (Tyr133) is orientated in a catalytically competent conformation due to a shift of the 3₁₀ helix towards β6 and β7 of the clasp domain. In a recent study, the structure of TtHICDH in complex with a designed inhibitor is reported.⁸ Similar to SpHICDH, the catalytic tyrosine (Tyr125) is flipped inward and interacts with the inhibitor, indicating that it is important for catalysis, and possibly substrate binding. These finding suggest that the SpHICDH active site adopts a conformation that is conducive to substrate binding, whereas the non-productive configuration observed in the TtHICDH apoenzyme structure may be a consequence of the crystallization condition. Further structural studies of HICDH substrate and product complexes are needed to ascertain the extent of the conformational changes that occur in the active site during ligand binding.

In addition to variations in their oligomerization state and active site conformation, the substrate specificities of the fungal and archaeal HICDHs also differ. TtHICDH utilizes both isocitrate and homoisocitrate as substrates, whereas yeast HICDHs display greater specificity for homoisocitrate versus isocitrate.^4,7,26 Mutagenesis studies in TtHICDH determined that substrate specificity is governed by a key active site residue, Arg85 in TtHICDH or Val94 in SpHICDH, which occupy the same position in their active sites (Fig. 2[D]).⁷ This observation is particularly interesting in that the structurally equivalent residue in ICDHs is also a valine. ICDH specificity towards its preferred substrate is instead governed by serine and asparagine residues (Ser113 and Asn115 in EcHICDH), which form interactions with the C5 carboxylate of isocitrate [Fig. S2].^25,27 The serine residue is conserved in HICDHs, but the asparagine is an isoleucine in yeast and archaeal HICDHs. Based on an SpHICDH-isocitrate model [Fig. 2(A)], Val94 may form van der Waals interactions to isocitrate, whereas in TtHICDH the corresponding Arg85 could potentially hydrogen bond to C5 carboxylate of isocitrate, equivalent to the interaction of the asparagine residue in ICDHs. Thus, this model offers a potential explanation for the ability of TtHICDH to efficiently oxidize both citrate and homocitrate. In summary, these observations suggest potential routes to exploit structural variations between the active sites of fungal and other β-hydroxyacid oxidative decarboxylases in order to design HICDH-specific inhibitors.