Human α-amino-β-carboxymuconate-ε-semialdehyde decarboxylase (ACMSD): A structural and mechanistic unveiling

Abstract

Human α-amino-β-carboxymuconate-ε-semialdehyde decarboxylase determines the fate of tryptophan metabolites in the kynurenine pathway by controlling the quinolinate levels for de novo NAD biosynthesis. The unstable nature of its substrate has made gaining insight into its reaction mechanism difficult. Our electron paramagnetic resonance (EPR) spectroscopic study on the catalytically active, Cu-substituted enzyme suggests that the native substrate does not directly ligate to the metal ion. Substrate binding did not result in a change of either the hyperfine structure or the super-hyperfine structure of the EPR spectrum. We also determined the crystal structure of the enzyme in its native state (at 1.99 Å resolution), a substrate analogue-bound form (2.50 Å resolution), and a selected active site mutant form with one of the putative substrate binding residues altered (2.32 Å resolution). These structures illustrate that each asymmetric unit contains 3 pairs of dimers. Consistent with the EPR findings, the ligand-bound complex structure shows that the substrate analogue does not directly coordinate to the metal ion but is bound to the active site by two ariginine residues through non-covalent interactions.

INTRODUCTION

α-Amino-β-carboxymuconate-ε-semialdehyde (ACMS) is a metabolic intermediate found in two catabolic pathways: the kynurenine pathway (Scheme 1), which is responsible for tryptophan catabolism, and the 2-nitrobenzoic acid biodegradation pathway.¹ ACMS is unstable and non-enzymatically dehydrates to quinolinic acid (QUIN) with a t_1/2 of 33 min at 20 °C, pH 7.0.² QUIN is the universal precursor for the de novo biosynthesis of NAD, providing the pyridine ring.^3,4 In humans, QUIN levels must be tightly regulated (< 100 nM) because it is also an agonist of N-methyl-D-aspartate receptors, and its overproduction can cause overexcitement of neurons and cell death in the central nervous system.^5–9 Elevated QUIN concentrations in bodily fluids have been observed in an exceptionally wide range of neuropsychiatric diseases including anxiety, depression, epilepsy, and neurodegenerative conditions such as Alzheimer’s and Huntington’s diseases.^5,10–15 A 300-fold elevation of QUIN has also been detected in the brain of patients with HIV infection.¹⁶ Moreover, over the past two decades, an increasingly accumulating volume of evidence unanimously suggest that abnormal activity of the kynurenine pathway participates in the initial phases of neuropathological processes,^13,17 the pathogenesis of AIDS-dementia complex,^18,19 and Alzheimer’s disease.^6,10,12,20 When QUIN was injected into healthy rats, the characteristic neuropathological features of Huntington’s disease started to develop.⁵ Thus, the connection between QUIN and related diseases flows both ways, i.e. in symptomatic patients QUIN levels are elevated, and in healthy mammals QUIN injection induces symptoms of disease. Notably, ACMS decarboxylase (ACMSD) directs the vast majority of metabolites to further catabolic steps controlled by enzymes and, by doing so, avoids the overproduction of QUIN (Scheme 1).^1,3 Thus, the decarboxylase ACMSD is of significant biomedical importance.

Scheme 1.

ACMSD competes with a non-enzymatic spontaneous cyclic reaction (marked with dotted arrow) and directs the vast majority of ACMS to enzyme-mediated catabolic pathway.

Prior to this work, the structure of human ACMSD (hACMSD) in its catalytically active form was unavailable. The elucidation of this structural information is crucial for the design of small molecule regulators and for the interpretation of biochemical and spectroscopic studies. We are also interested in the structural and mechanistic divergence between hACMSD from the tryptophan kynurenine pathway and Pseudomonas fluorescens (pfACMSD) from the 2-nitrobenzoic acid metabolic pathway. However, hACMSD has proven to be a considerable challenge to express in prokaryotic systems. Although expression of hACMSD as a weakly active enzyme has been achieved in mammalian cell lines²¹ and the eukaryote, Pichia pastoris,²² prior attempts to express hACMSD in bacterial systems were all unsuccessful.^21–23

In the present work, we successfully expressed hACMSD in E. coli with the assistance of chaperone co-expression and obtained catalytically active form of the enzyme. By using substrate analogue and metal-substituted hACMSD, we employed a combined structural and spectroscopic approach to tackle the question of how the unstable substrate interacts with the enzyme’s active site.

MATERIALS AND METHODS

Expression of Human ACMSD

To construct an N-terminally His₁₀-tagged hACMSD expression plasmid, the hACMSD cDNA,²¹ a generous gift from Dr. Fukuoka, was amplified by the polymerase chain reaction with the forward primer 5′-GGAATTCCATATGAAAATTGACATCCATAG-3′ and the reverse primer 5′-CCGCTCGAGTCATTCAAATTGTTTTCTCTC-3′ (built-in NdeI and XhoI sites are underlined). The PCR product was purified from a 0.8% agarose gel, digested with NdeI and XhoI, and ligated in the equivalent sites of pET16b (Novagen). After sequencing, the positive clone was used for transformation into E. coli BL21 (DE3) containing the GroEL/ES overexpression plasmid pGro7, pColdII-pGro7, pColdII-pG-Tf2, or pET16b-pG-Tf2, respectively (Takara Corp., Japan).

Cell Culture

Single colonies of E. coli containing the appropriate pET-hACMSD plasmid were used to inoculate 50 mL of M9 growth medium (1×M9 medium supplement with 1 mM MgSO₄, 0.1 mM CaCl₂, 2 μg/mL vitamin B₁, 0.04% casamino acids, 0.4% D-glucose, and a divalent metal ion described below). To obtain hACMSD with desired metal at the active site, a divalent metal ion of either CoCl₂, CuSO₄, Fe(NH₄)₂(SO₄)₂, or ZnCl₂ was added to a final concentration of 0.05 to 0.5 mM prior to induce expression. The cell culture medium also contained 100 μg/mL ampicillin, 40 μg/mL chloramphenicol and 500 μg/mL L-arabinose, the latter of which induces the expression of the protein chaperone GroES-GroEL. The cells were grown at 37 °C with shaking until the OD₆₀₀ reached 0.4, at which time the temperature was lowered to 25 °C. hACMSD expression was induced by 0.1 mM isopropyl β-D-thiogalactopyranoside at an OD₆₀₀ of 0.6. The cultures were allowed to further grow at 25 °C for 16 h before being harvested by centrifugation at 8000 × g for 15 min at 4 °C. The wet cells were washed with 50 mM potassium phosphate buffer, pH 8.0, twice before storage at −80 °C.

Protein Isolation

Human ACMSD was purified using a modified procedure described for the bacterial analogue.² Frozen cells were re-suspended in 50 mM potassium phosphate, pH 8.0, containing 300 mM NaCl and 0.1 mM protease inhibitor phenylmethylsulfonyl fluoride. The cell slurry was passed through an M-110P Microfluidics cell disruptor and the debris was removed by centrifugation at 27,000 × g for 30 min at 4 °C. The supernatant was loaded onto a HiPrep immobilized metal ion affinity chromatography (Ni-NTA) 26/20 column (80 mL resin) for isolation of the native enzyme or FF 16/10 columns charged with either Co²⁺, Cu²⁺ or Zn²⁺ depending on the corresponding metal-specific hACMSD preparations in order to minimize possible metal cross-contamination in purified protein. The proteins were eluted in a two-step gradient with increased concentration of imidazole in the buffer. The major fractions with ACMSD decarboxylase activity were pooled, concentrated, and further purified on a Superdex 75 column (26/60) using an ÄKTA FPLC protein purification system with a 25 mM HEPES buffer, pH 7.0, containing 5% glycerol. Protein concentrations were determined using Coomassie Plus protein assay reagent (Pierce). The expression level and enzyme purity were determined by SDS-PAGE on 12% polyacrylamide gels.

Metal Analysis

The metal content of ACMSD was assayed in triplicate by inductively coupled plasma optical emission spectrometry (ICP-OES) using a Spectro Genesis spectrometer (Spectro Analytical Instruments GmbH & Co. KG, Germany) as previously described ²⁴. Metal ions loosely or nonspecifically bound to the enzyme were removed by an overnight incubation of 0.5 mM EDTA with the protein solution at 4 °C followed by desalting on a 5-mL HiTrap column with 10 mM Tris-HCl (pH 7.4) and buffer wash with the ICP buffer and ultrafiltration.

EPR Spectroscopy

The interaction between ACMS and the enzyme-bound metal center was measured by EPR spectroscopy using Cu-containing ACMSD in the absence or presence of 3 equivalents of ACMS in the assay buffer. The EPR samples were obtained by mixing the protein solution with the substrate solution at 25 °C using a freeze-quenching apparatus, Update Instruments, System 1000. The reaction mixture was shot directly to an EPR tube and frozen in liquid nitrogen. The entire mixing and freezing process was completed in less than 5 seconds. In a parallel EPR experiment, the response of the addition of 3 equivalents of ACMS to copper sulfate was determined with the same buffer system. The same X-band EPR instrumentation is used as those previously described.^1,25

Steady State Kinetics

The enzyme activity of the recombinant hACMSD was determined according to the procedures described in detail in earlier reports.^1,25,26 Briefly, ACMS was generated from 3-hydroxyanthranilate by an enzymatic method.¹ The ACMSD enzyme assay mixture contained 0 – 120 μM ACMS and an appropriate amount of ACMSD protein in 25 mM HEPES buffer, pH 7.0 with 5% glycerol. Specific activities were calculated from the initial velocities of ACMS decay monitored by the loss of absorption at 360 nm using an absorbance coefficient constant of 47,500 M⁻¹cm⁻¹ when the substrate concentration was equal to or lower than 20 μM, or at 320 nm using an absorbance coefficient constant of 9,600 M⁻¹cm⁻¹ when the substrate concentration was between 20 – 120 μM. In the enzyme concentration dependence assay, 800 μM Zn-hACMSD was diluted to 200 nM by the reaction buffer. The initial rate of the reaction was monitored from 1 to 240 min frequently in triplet after dilution: a saturated substrate concentration of 15 μM was used.

The inhibition pattern of pyridine-2,6-dicarboxylic acid (PDC) acting on hACMSD was obtained by determining the k_cat and K_m in the presence of varying concentrations of PDC (0, 20, 40, and 80 μM) in both buffer and substrate solution. Apparent K_m values for each inhibitor concentration were plotted as a function of inhibitor concentration and inhibition is expressed as K_i values in micromolar (K_i = y-intercept/slope).

X-ray Protein Crystallography

Catalytically active form of Zn-hACMSD was used to screen crystallization conditions in Art Robbins 96-well Intelli-Plates using an ARI Gryphon crystallization robot. The initial hit was obtained from the Index Screening Kit (Hampton Research). After optimization, crystals were obtained from drops assembled with 1.5 μL of 28 mg/mL hACMSD mixed with 1.5 μL of a reservoir solution containing 0.2 M lithium sulfate monohydrate, 0.1 M Tris-HCl pH 6.5 - 7.5, and 20 - 25% polyethylene glycol 3,350, by hanging drop diffusion in VDX plates (Hampton Research). PDC-hACMSD co-crystals were obtained by pre-incubating hACMSD with 10 equivalents of PDC for 30 min prior to crystallization. Crystals suitable for X-ray diffraction were obtained ca. 10 days after crystal growth at 18 °C. Crystallization mother liquor containing 20% glycerol was used as a cryoprotectant. All crystals obtained under these conditions belong to the P2₁2₁2₁ space group and data were processed with HKL-2000.²⁷ The three dimensional structures of hACMSD were solved by molecular replacement using our previously published Pseudomonas fluorescens (pfACMSD) structure as a search model (PDB code: 2HBV). Molecular replacement and refinement were carried out using PHENIX.²⁸ Electron density analysis and model building were carried out in Coot.²⁹ The unit cell images were generated with PyMol³⁰ with the aid of the SuperSym plugin available at http://supersym.sourceforge.net.

RESULTS

Overexpression of the Human Enzyme in E. coli

hACMSD expressed from the constructed vectors, pColdII-pGro7, pColdII-pG-Tf2, or pET16b-pG-Tf2 was only found in inclusion bodies. Fortunately, the constructed pET16b-pGro7 plasmid expressed hACMSD as a soluble and enzymatically active protein. The enzyme expressed from this system was purified to near homogeneity as described under “Experimental Procedures”. Five to ten mg of pure enzyme was obtained from each liter of cell culture when gown in the LB medium; however the yield was significantly reduced when a minimal medium was used. Purified hACMSD is stable at 4 °C for weeks and can endure freeze/thaw cycles with liquid nitrogen multiple times without a significant loss of activity.

Metal Content and Catalytic Activity

Metal analysis of hACMSD purified from cells grown in the LB medium contained several metals. Iron, zinc and copper were present and the total metal content was substoichiometric (Table 1). Metal analysis of hACMSD purified from the cells cultured in the M9 minimal media supplemented with 100 μM of varying metal ions is presented in Table 1. Activity assays demonstrated that hACMSD expressed in the presence of Zn²⁺ exhibited activity comparable to what was found for the most active forms of pfACMSD bound with a Zn²⁺ or Co²⁺ ion.¹⁹ Co²⁺-containing hACMSD, on the other hand, exhibited very little specific enzymatic activity relative to Co-pfACMSD. It should be noted that the contribution of Zn ion to the activity of Co-, Cu-, and Fe-hACMSD was subtracted based on the metal analysis data (Table 1) with regard to the Zn-hACMSD activity.

Table 1.

Percentage of metal occupation and corresponding activity of hACMSD from different medium cultures.

	LB	M9	M9 + Co2+	M9 + Cu2+	M9 + Fe2+	M9 + Zn2+
Co	0	0	96.8	0.4	0.8	0
Cu	2.2	2.9	0.7	47.9	3.6	9.7
Fe	28.7	9.8	1.0	10.0	55.1	7.5
Zn	13.8	9.8	0.8	0.6	7.7	64.4
Specific activity (nmol/min/mg)	910	800	234* 190	123* 90	823* 400	3540

The metal ions of hACMSD were incorporated into the enzyme during protein assembly. Metal content was measured as described in Experimental Procedures. The concentration of the metal ions supplemented to the M9 medium was 0.1 mM in each of the above experiments. The values were the average numbers of three independent experiments with standard deviation smaller than 5%.

^*indicating data prior to subtracting the contribution of Zn ion.

Unlike pfACMSD,¹ the metal ions in hACMSD could not be removed by the metal chelators EDTA, 1,10-phenanthroline or 8-hydroxy-quinoline-5-sulfonic acid, and the enzyme activity was not affected by the addition of these metal chelators. Moreover, the addition of one to ten equivalents of Co²⁺, Cu²⁺, Fe²⁺, or Zn²⁺ to purified hACMSD did not increase enzyme activity, indicating that the purified human enzyme cannot be reconstituted in vitro.

Since in vitro metal reconstitution was not successful for the human enzyme, we attempted to add zinc ions to the M9 medium at various concentrations prior to induction of the cells. The enzyme isolated in this set of experiments contained different amount of naturally incorporated zinc ion during protein synthesis. Fig. 1 shows that the activity of hACMSD is proportional to the zinc content of the enzyme. These results suggest that hACMSD is a zinc-dependent metalloenzyme. Zn-hACMSD yielded steady-state kinetic parameters similar to those of Zn²⁺-containing pfACMSD.³¹ A k_cat value of 4.8 s⁻¹ was measured for hACMSD with 64.4% zinc ion occupancy. This value is anticipated to be 7.5 s⁻¹ with a fully-loaded zinc ion. The K_m (5.8 μM) and the k_cat (7.5 s⁻¹) values of hACMSD are comparable to the K_m (9.6 μM) and the k_cat (6.5 s⁻¹) reported for Zn-pfACMSD with ca. 80% zinc occupancy.³¹

Figure 1.

The specific catalytic activity of hACMSD as a function of Zn content in the protein. The variation of Zn content was achieved by growing cells in M9 minimum medium supplemented with 0 – 500 μM concentrations of ZnCl₂.

When hACMSD was diluted from 800 μM to 200 nM, it lost 90% of activity after 4 h (Fig. 2). Since proteins prefer to form higher oligomer state at higher concentration, the hACMSD dimer will start to dissociate once its concentration drops. Losing catalytic activity after dilution implies that hACMSD is unlikely to be active in the monomeric state. Similar observations have been reported for the bacterial ACMSD.²⁶

Figure 2.

The catalytic activity of hACMSD decays after dilution, indicating that hACMSD has better activity at higher oligomerization state. The relative activity of concentrated and diluted samples was measured under the same conditions as described in the text. The experimental data was fitted with first exponential decay (red line).

Probing the Nature of the Enzyme-Substrate Interactions

Although ACMS is an unstable molecule, we find that it tends to form stable complexes with various metal ions. Because Zn²⁺ is EPR silent while Cu²⁺ is EPR active with well refined hyperfine signals sensitive to X-band EPR, copper substituted ACMSD is used in this section of experiment. Fig. 3(A) shows that ACMS binds to a cupric ion in solution. The EPR signal intensity of CuSO₄ solution is weak due to the coupling of copper ions. Addition of unstable ACMS to CuSO₄ led to the formation of stable Cu²⁺-ACMS complex which gives rise to well-resolved type II copper EPR spectrum.³² Since this technique is highly sensitive to the changes of the electronic/chemical structure of a Cu²⁺ ion, especially when the superhyperfine structure is resolved, we analyzed ACMS perturbation to the enzyme-bound copper center by EPR spectroscopy.

Figure 3.

Low-temperature X-band EPR spectroscopic study confirms that the substrate binds to the free metal ion but does not directly coordinate to the catalytic metal ion in ACMSD. (A) Free CuSO₄ and CuSO₄-ACMS (1:3) at 200 μM concentration, (B) as-isolated Cu-hACMSD (200 μM) and Cu-hACMSD-ACMS (1:3), and (C) a magnified view of superhyperfine structure of Cu-hACMSD in the absence/presence of ACMS (600 μM). The spectrometer conditions: temperature, 20 K; modulation amplitude, 0.3 mT; microwave power, 0.5 mW; time constant, 40.96 ms; and sweep time 1.79 mT/s for the field from 225 to 375 mT.

To probe whether the substrate ACMS is ligated directly to the metal, we obtained Cu-substituted hACMSD. Since the metal ion cannot be extracted and reconstituted, we circumvented this problem by isolating the human enzyme from cell culture of a metal-depleted minimal medium supplemented with cupric ion. The Cu-substituted protein is catalytically active although it is a relatively poor catalyst. Its k_cat value was 0.1 ± 0.02 s⁻¹ and K_m value was 3.3 ± 0.2 μM.

Both hyperfine splitting due to the nuclear spin of copper (I = 3/2) and nitrogen (I = 1)-induced superhyperfine interactions are present in the EPR spectrum of Cu-hACMSD [Figure 2(B)]. The EPR parameters obtained from a simulation of the spectrum of wild-type Cu-hACMSD have an A_// value of 18.7 mT and a g_// value of 2.218, both typical for a type II Cu²⁺ center. In the perpendicular region, a multiple-line superhyperfine structure with a splitting of about 1.5 mT is well resolved in the spectrum [Fig. 3(C)]. The superhyperfine structure originates from the spin-spin interaction between the nuclear spin of the nitrogen atoms (nuclear spin I = 1) and the electron spin of the Cu²⁺ ion. When 3 molar equivalents of ACMS were added to Cu-hACMSD, frozen in less than 5 s, and subsequently measured by EPR at 20 K, no spectral change was observed. In sharp contrast to the free copper ion in solution, the presence of ACMS is unable to induce any disturbance in the hyperfine and superhypefine structures of the Cu center in hACMSD. Thus, there is no direct ligation between ACMS and the Cu center in the enzyme.

Crystal Structure

The structure of hACMSD was obtained from native, catalytically competent form of the enzyme and was refined to 1.99 Å resolution (Table 2, PDB entry: 4OFC). Each unit cell contains four asymmetric units in which six protomers are found as a trimer of dimer, as shown in Fig. 4. The dimer interface contains a surface area of about 2,400–2,460 Ų and is primarily formed by three helical regions which are composed of residues 191 to 201, 230 to 244, and 271 to 281 of each subunit. The central area of the dimer interface is strictly hydrophobic. Analysis of the dimer interface reveals a total of 24–33 direct hydrogen bonding interactions and 18–21 salt bridges for each dimer.

Table 2.

X-ray crystallography data collection and refinement statistics

Data collection	hACMSD	PDC-hACMSD	R47A hACMSD
Space group	P212121	P212121	P212121
Unit cell lengths (Å)	a=88.704, b=101.082, c=232.820	a=88.368, b=101.655, c=233.117	a=89.142, b=101.686, c=232.613
Unit cell angles (°)	α=γ=β=90	α=γ=β=90	α=β=γ=90
Wavelength (Å)	1.0	1.0	0.8
Temperature (K)	100	100	100
Resolution (Å)a	45.00–1.96 (2.03–1.99)	45.00–2.50 (2.50–2.54)	50.00–2.33 (2.37–2.33)
Completeness (%)a	87.6 (81.3)	75.4 (65.7)	98.4 (82.8)
Rmerge (%)a, b	15.3 (78.2)	9.9 (29.3)	13.2 (48.6)
I/σI a	20.6 (2.3)	10.3 (1.5)	17.9 (2.0)
Multiplicity a	10.1 (8.0)	9.7 (8.5)	25.3 (13.7)
Refinement
Resolution (Å)	1.96	2.49	2.32
No. reflections; working/test	130135/6497	55192/2809	85220/4492
Rwork (%)c	19.9	23.3	20.8
Rfree (%)d	24.1	29.8	27.6
No. of protein atoms	15984	15810	15774
No. of ligand atoms	6	78	6
No. of solvent sites	994	286	821
Average B factor (Å2)
Protein	29.7	33.3	30.0
Zn(II) or PDC	26.3	39.4	23.6
Solvent	31.3	28.4	32.2
Ramachandran statisticse
Preferred (%)	96.15	91.82	97.7
Allowed (%)	2.60	5.86	2.3
Root mean square deviation
Bond lengths (Å)	0.008	0.029	0.015
Bond angles (°)	1.155	1.262	1.785
PDB entry	4OFC	4IH3	4IGN

^aValues in parentheses are for the highest resolution shell.

^bR_merge = Σ_i |I_hkl,i − 〈I_hkl〉|/Σ_hkl Σ_i I_hkl,i, where I_hkl,i is the observed intensity and 〈I_hkl〉 is the average intensity of multiple measurements.

^cR_work = Σ||F_o|−|F_c||/Σ|F_o|, where |F_o| is the observed structure factor amplitude, and |F_c| is the calculated structure factor amplitude.

^dR_free is the R factor based on 5% of the data excluded from refinement.

^eBased on values attained from refinement validation options in COOT.

Figure 4.

The crystal structure of hACMSD is solved as a homodimer with a zinc ion bound at the catalytic center. (A) Each unit cell contains 4 asymmetric units of 3 pairs of dimers. (B) The catalytic center contains two arginine residues, Arg47 and Arg235*. Star indicates Arg235 is from a neighboring subunit.

Due to the instability of ACMS, the substrate-bound ACMSD structure is not yet available. However, pyridine-2,6-dicarboxylic acid is a heterocyclic, stable ACMS analogue that competitively inhibits the bacteria ACMSD.²⁶ We found that it is also an effective competitive inhibitor of hACMSD [Fig. 5(A)] with a K_i value of 15.2 ± 0.5 μM. We successfully co-crystallized this compound with hACMSD (Table 2, PDB entry: 4IH3) and the overall structure is nearly identical to the ligand-free enzyme. PDC is bound at the active site and forms ionic interactions with Arg47 and Arg235*, where the asterisk indicates this residue is from a neighboring subunit [Fig. 5(B)]. The PDC-bound structure shows that the inhibitor is fairly close to the zinc ion with the C4 ca. 2.6 Å from the zinc ion. However, the closest oxygen is 5.6 Å away from the metal ion. Thus, the substrate analogue does not directly coordinate to the metal ion.

Figure 5.

The crystal structure of hACMSD in complex with a substrate analogue sheds light on the enzyme-substrate interaction mode. (A) Competitive inhibition of hACMSD by 0, 20, 40, and 80 μM of pyridine-2,6-dicarboxylic acid (PDC) as the black, red, green, and blue traces, respectively. (B) The co-crystalized structure shows that the competitive inhibitor PDC is bound by Arg47 (right Arg in the figure) and Arg235* (left) near the zinc ion (pink ball).

To further prove that the unstable substrate binds to the enzyme by the active site arginines, we mutated Arg47 and Arg235* to alanine. Both R47A and R235A variants showed no detectable catalytic activity, proving that they are important residues for catalysis. The result of this analysis is consistent with our structural findings shown in Fig. 4(B). While R235A eludes crystallization, the crystal structure of the R47A mutant was solved at 2.32 Å resolution (Table 2). Similar to the wild-type enzyme structure, six protomers in three dimers are present in each asymmetric unit. The same interface region is observed between the protomers as observed in the wild-type enzyme. The RMSD value is only 0.254 Å from 14,034 out of 14,040 atoms for the structures of R47A and wild type.

DISCUSSION

Our work on human ACMSD solved several long-standing mysteries. The first pertains to the cofactor identity. Our previous work on the analogous bacterial system indicates that pfACMSD is a metal-dependent enzyme.^1,25,31 Thus, one would expect hACMSD is a metalloprotein. However, all prior studies on hACMSD suggest that the catalytic activity of human enzyme is independent of metal ions.^21,22 This is likely due to the observation that neither a metal chelator, nor external metal affects the activity of purified hACMSD. The first crystal structure of hACMSD was later determined with a zinc ion bound at the enzyme active site, but how the zinc ion is associated with catalytic activity was not discussed,²³ leaving a questionable cofactor-free conclusion for hACMSD in literature. Here, by showing protein expressed in a minimal medium supplemented with increasing concentrations of zinc ions, giving higher decarboxylase activity together with the data present in Table 1, we found that the zinc enzyme presents the highest catalytic activity. The catalytic efficiency of Zn-hACMSD is significantly higher than that previously reported for the same enzyme expressed in Pichia pastoris, which has a similar K_m but a lower k_cat value of 1.0 s^−1.22 Thus, it becomes evident that the hACMSD is a zinc-dependent enzyme.

Since metal chelators are demonstrated to be able to extract the metal ion from pfACMSD but not hACMSD, hACMSD apparently binds its metal more tightly. The crystal structures solved provide a molecular understanding for this nonconformity. An aspartate and three histidine residues in hACMSD coordinate the zinc ion. Notably, His174 is a metal ligand in all six protomers of each asymmetric unit of the structure, with a bond distance of 2.2–2.3 Å. In contrast, the equivalent histidine in pfACMSD, His177, presents significant flexibility in the structure we have solved (PDB entry: 2HBV). In subunit A, His177 coordinate to the zinc ion with distance of 2.2 Å. In subunit B, His177 is no longer a metal ligand and is 4.2 Å away from the zinc ion. Thus, His177 is a weaker ligand in pfACMSD. The observed metal affinity deviation therefore becomes understood.

It is challenging to predict the cofactor dependency of a protein base on the primary structure if the cofactor is a single atomic metal. Although characteristic sequences can be used to identify some metal binding motifs, including zinc finger, calcium EF hand, and iron-sulfur cluster, drawing conclusions based solely on sequence can be misleading. For example, LigI is a member of the amidohydrolase superfamily, whose members were all previously demonstrated to be divalent metal dependent including ACMSD. Based on sequence alignment, LigI has all four conserved residues, three histidines and one aspartate, for metal coordination. However, later mechanistic and structural studies revealed that LigI is actually metal independent and the four metal ligand residues are with new catalytic roles.³³ Sometimes, even with the assistance of biochemical study, prediction of metal cofactor can still be misjudged. The enzyme uronate isomerase catalyzes the isomerization reaction of D-glucuronate and D-fructuronate was initially thought to be metal independent because metal chelators do not affect the enzyme activity.⁴ Later on, an improved study suggested that uronate isomerase is actually a mononuclear zinc-containing enzyme.³⁴ Hence, a combined knowledge of sequence, biochemical assay, and structure is needed to correctly identify a metal cofactor.

Since its initial identification from rat liver, ACMSD has been known for 58 years³⁵ and this enzyme has arisen significant biomedical interest. Following our success in obtaining the bacterial enzyme structure in 2006,³¹ a structure of the enzyme in its inactive form, complexed with 1,3-dihydroxyacetone-phosphate (DHAP), was obtained through molecular replacement in 2009 by Garavaglia et al. (PDB entry 2WM1).²³ It should be noted that the DHAP-bound structure is monomeric with only one subunit in each asymmetry unit, whereas the three hACMSD structures present in this study, including a substrate-free and ligand-bound structure, and all of the pfACMSD structures determined thus far,^24,26,31 are homodimers.

A quaternary structure variety is apparently present in the structural studies of hACMSD. We think that the diversity brings up the following questions: (1) is the quaternary structure of ACMSD linked to the catalytic activity, and (2) is a ligand-induced structural change a potential mechanism of catalytic activity regulation? While the latter question will be a subject for future study, the results presented in this work and the comparison with our previous studies on the bacterial enzyme shed insightful clues, as discussed below.

We have recently shown that pfACMSD is a mixture of monomer and dimer in solution, and that the former is catalytically inactive while the latter is functionally competent.²⁶ The decarboxylase activity is dependent on the presence of two arginine residues (Arg51 and Arg239*, star indicates that Arg239 is from the neighboring subunit), which play a key role in substrate binding ²⁶. Likewise, when hACMSD is in the dimeric form, the conserved arginine residues (Arg47 and Arg235*) are both present in the active site in a manner resembling what has been described for pfACMSD (Fig. 6). When hACMSD is in the monomeric form bound with the DHAP, Arg235 is 25 Å from the catalytic metal and cannot be a residue involved in substrate binding.²³ Thus, these conserved arginine residues are effective probes for determining the functional assembly of ACMSD.

Figure 6.

A side-by-side comparison of the active site architecture of (A) pfACMSD and (B) hACMSD shows similar substrate binding environment. Two arginine residues, including one from the neighboring subunit (with its carbon atoms highlighted in yellow), must be present in the active site for the decarboxylase activity.

The enzyme 5-carboxyl-uracil decarboxylase (IDCase) is a member of the ACMSD subfamily under the amidohydrolase superfamily and has very high sequence similarity to ACMSD.^36,37 This enzyme is the most closely related neighbor of ACMSD in the reconstructed evolution tree. Recently, the overall structures of IDCase from both Cordyceps militaris and Metarhizium anisopliae mostly resemble the structure of ACMSD.³⁸ Interestingly, in both of the substrate 5-carboxyl-uracil and inhibitor 5-nitro-uracil bound binary complex structures, the catalytic metal ion does not directly interact with the substrate and inhibitor. The two conserved arginine residues, Arg68 (equivalent to Arg47 of hACMSD) and Arg262* (equivalent to Arg235* of hACMSD) from the neighboring subunit play major roles in substrate/inhibitor binding. Specifically, Arg262* binds to the leaving carboxyl group in the substrate bound structure (PDB entry: 4LAM). This observation further suggests that Arg47 and Arg235* are responsible for substrate binding not only in ACMSD but also in proteins from the same subfamily. As a result, a dimeric quaternary structure is most likely required for enzyme activity through the ACMSD subfamily.

Since mammals have no mechanism to store free amino acids not used for protein synthesis, tryptophan is primarily directed towards catabolism. During starvation, when tryptophan levels are low, the priority of catabolism is to preserve the kynurenine pathway intermediates for NAD biosynthesis and thus, the activity of ACMSD must be lessened. Using a glycolic intermediate, DHAP, to force ACMSD to change its structure to the catalytically inactive monomeric form is a possibly unprecedented metabolic interrelation between the tryptophan kynurenine catabolic pathway and glycolysis.

CONCLUSION

The biochemical, spectroscopic, and structural results led us to conclude that hACMSD is a zinc-dependent dimeric enzyme that harnesses its unstable, negatively charged substrate by its two positively charged residues rather than by the metal ion during substrate positioning at the early stage of the catalytic cycle. The quaternary structure variety, previously demonstrated in the bacterial analogous enzyme²⁶ and now implicated in the human enzyme, is directly linked to the enzyme catalytic activity. Thus, the structural arrangement of the catalytic center offers a potential regulatory mechanism at a critical junction of the downstream tryptophan kynurenine pathway.

Abbreviations

QUIN

quinolinate or quinolinic acid

ACMS

α-amino-β-carboxymuconate-ε-semialdehyde

ACMSD

ACMS decarboxylase

DHAP

1,3-dihydroxyacetone-phosphate

IDCase

5-carboxyl-uracil decarboxylase

PDC

pyridine-2,6-dicarboxylic acid