Robust structural, kinetic and biophysical characterization of wild-type human ACOD1, selected mutants and their interaction with citraconate

Abstract

Aconitate decarboxylase 1, an enzyme member of the MmgE-PrpD family of proteins, has gained significant attention in the last decade as a therapeutic target for cancer and inflammatory diseases. Its product, itaconate, is a multifunctional metabolite shown to drive several disease states. Though extensively studied in cellulo and in vivo, this protein is biochemically and mechanistically under characterized and although a family of inhibitors has been described, no ligand-bound structures have yet been determined. In this work we present a thorough structural investigation that yielded the first ligand-bound structure of this protein family, which required the generation of artifact-free apo crystals. We also developed a novel, low-consumption, robust kinetic assay and investigated active site and allosteric mutants to further elucidate structural and dynamic activity relationships of this protein.

Graphical abstract

Highlights

• •An artifact-free apo structure of hACOD1, member of the MmgE-PrpD protein family.
• •First X-ray structure of an active-site ligand-bound protein of the MmgE-PrpD family.
• •A robust, low-consumption, continuous kinetic assay for hACOD1.
• •Solution scattering data for hACOD1 shows no large conformational changes for hACOD1.
• •Mutational analyses of hACOD1 indicate possible allosteric regulation.

1. Introduction

Human aconitate decarboxylase 1 (hACOD1) is a member of the MmgE-PrpD protein fold family and plays a pivotal immunoregulatory role in host-pathogen interactions (Nair et al., 2018; Shi et al., 2005; Daniels et al., 2019; Wu et al., 2022), pro- and anti-inflammatory responses (Dai et al., 2024; Lampropoulou et al., 2016; Cordes et al., 2016), and in the modulation of oxidative stress responses (Qian et al., 2024; Hall et al., 2018). hACOD1 has also been specifically shown to contribute to the maintenance of tumor microenvironments (Weiss et al., 2018; Wang et al., 2023). hACOD1 is encoded by the human immune regulatory gene 1 (hIRG1) and is responsible for catalyzing the decarboxylation of cis-aconitate to yield the terminal metabolite itaconate in the tricarboxylic acid (TCA) cycle (Fig. 1) (Michelucci et al., 2013). Itaconate was initially identified as a key metabolite that is highly upregulated in macrophages when triggered by inflammatory stimuli (Lee et al., 1995).

Fig. 1.

TCA cycle (Created in BioRender, 2026) and ACOD1 mechanism. A) overview of the TCA cycle showing metabolic intermediates, including the transformation of cis-aconitate to the terminal metabolite itaconate by ACOD1. CS: citrate synthase; ACON: aconitase; IDH: isocitrate dehydrogenase; a-KGDH: a-ketoglutarate dehydrogenase; SCS: succinyl-CoA synthase; SDH: succinyl dehydrogenase; FH: fumarase; MDH: malate dehydrogenase. B) Proposed reaction mechanism of cis-aconitate decarboxylation by ACOD1. The nature of any proton transfer steps between the protein and substrate is unknown, as well as whether the reaction follows a one- or two-base mechanism.

Itaconate has a direct regulatory effect on the TCA cycle, by inhibiting succinate dehydrogenase (SDH). Increased succinate concentration in cells impacts mitochondrial respiration and cytokine production in macrophages (Lampropoulou et al., 2016). These regulatory mechanisms lead to anti-inflammatory responses including halting of mitochondrial reactive oxygen species (ROS) production and downregulation of HIF-1a activity and of IL-1b expression. Itaconate reportedly also regulates antioxidant responses by modulation of KEAP1, stopping proteasomal degradation of the transcription factor NRF2 (Ahmed et al., 2017). The increase in NRF2 leads to ROS detoxification and protects cells against cytotoxic oxidative stress (O'Neill and Artyomov, 2019). Itaconate also directly interacts with TET2, an important player in innate immune homeostasis and tumor suppression (Li et al., 2023; Chen et al., 2022a). All these anti-inflammatory cascades, resulting from increased itaconate production, lead to a cellular environment more resilient to cytotoxic conditions and thus resistant to apoptotic mechanisms.

hACOD1 has been shown to have elevated expression levels in cancer cells and is a reported tumor growth factor (Weiss et al., 2018; Wang et al., 2023; Gu et al., 2023; Chen et al., 2023) making it a promising chemotherapeutic target for sensitizing cancer cells and eliciting ferroptosis (Zhao et al., 2023). Currently, citraconate and its derivatives are the only reported inhibitors of hACOD1 (Chen et al., 2025; Papathanassiu, 2020; Chen et al., 2022b). Citraconate is a structural mimetic itaconate (the product, Fig. 1). Though inhibition kinetics have been obtained and a proposed binding pose modelled in silico in the active site (Chen et al., 2019), no protein-bound structure has been determined so far.

In this work we fill this knowledge gap by determining the thus far elusive structure of hACOD1 bound to citraconate. We also introduce a novel, continuous, low-consumption enzymatic assay as an alternative to that previously described (Chen et al., 2022b). Together these tools support future inhibitor discovery campaigns, enabling structure-guided drug design and providing an orthogonal biochemical assay for mechanism of action characterization (Cornish-Bowden, 1986).

2. Results and discussion

2.1. An artifact-free structure of apo ACOD1

To date, only 11 structures for of the MmgE-PrpD family, which is comprised of ACOD1, 2-methylcitrate dehydratase and iminodisuccinate epimerase, have been deposited in the Protein Data Bank. Inspection of these 11 protein models showed all to be apo structures (at the active site), with one exception where a buffer component is modelled (tartrate, PDB 5MUX). All three enzymes catalyze the conversion of small, polydentate carboxylate molecules through acid-base catalysis mechanisms.

The only structures available for human and murine ACOD1 were solved in 2019 by Chen et al. (Chen et al., 2019). The crystals for these structures were obtained from crystallization cocktails containing tacsimate and citrate respectively, and the hACOD1 structure shows the presence of residual electron density in the putative active site and no well-defined water network, indicating the presence of undefined buffer components. Tacsimate is a mixture of polydentate carboxylates that are structurally similar to the substrate/product of hACOD1. It is thus, unsurprising, that these components bind to the protein, even if disordered. An artifact-free apo crystal without promiscuous binders is important for co-crystallization with inhibitors as well as high-throughput fragment screening campaigns.

After optimizing the expression and purification of hACOD1 to yield stable, active and pure protein suitable for structural and enzymatic studies (see Sup. Table 1 for final expression conditions), we set out to obtain reliable crystallization conditions for hACOD1 using high-throughput methods. Protein crystallization relies on empirical screening of crystallization conditions with chemicals known to drive crystal formation. We set 1.7 k crystallization experiments, with 576 different crystallization cocktails (sparse matrix screens) and varying protein-to-cocktail volume ratios. From this initial screening campaign, we obtained ∼30 crystallization hits, many of which contained unwanted polydentate carboxylates. From these, 11 conditions (Sup. Table 2) were chosen for further optimization and crystals diffracting routinely to 1.3 Å or better were obtained from two of the initial hits. Upon structure solution from data collected from crystals grown in sodium acetate, no evidence of bound acetate was found in the active site, showing only well-ordered water molecules. Thus, our high-throughput crystallization pipeline was successful in producing an artifact-free apo hACOD1 structure at high resolution (1.22 Å, Fig. 2, C). hACOD1 is a dimer, with each protomer composed of two well-folded domains, a lid domain composed of residues 273–410 and an alpha-helical domain, residues 1–267 and 413–461 (Fig. 2, A), matching the typical topology of the MmgE/PrpD superfamily (EMBL-EBI InterPro). The catalytic sites are located between the two domains, with catalytically relevant residues His103, His159 and Lys 272 in the alpha-helical domain and Tyr318 in the lid domain (Fig. 2, B). The small active site in the apo hACOD1 crystal structure is closed to the outside medium, through a hydrophobic barrier at the entrance by Try318 and Pro155 (Fig. 5, C). To verify that the crystallographic structure matches the native structure in solution, we collected Small-Angle X-ray Scattering (SAXS) data (Fig. 2, D). The SAXS data showed a single, well-folded species in solution and fitting to the calculated scattering curve from our experimental X-ray structure shows very good agreement. This correlation shows that this two-domain protein adopts a closed conformation in solution, rather than exploring large open/close conformational changes.

Fig. 2.

hACOD1 apo structure, active site cavity and hydrogen bonding network. A) overall architecture of the hACOD1 dimer, showing the lid domain in dark green and the alpha-helical domain in light green. B) and C) The active site of WT hACOD1, showing ordered waters (red spheres) and the hydrogen bonding network (black dashed lines), with the 2F_o-F_c electron density map shown as gray mesh, contoured at 1σ. D) Experimental SAXS profile of WT hACOD1 (green dots) fit to the calculated solution scattering profile derived from the X-ray crystal structure (black line), showing good agreement with low weighted fit residuals (black dots, Χ² = 0.221).

Fig. 5.

hACOD1-His103Ala and -Tyr318Ala mutants X-ray structures in yellow and teal (A and B respectively), showing the hydrogen bonding network within the active site (black dashes for 2.6–3.2 Å and gray for 3.2–3.5 Å). The His103Ala mutation opens the active site cavity and allows for the accommodation of a buffer glycerol molecule (shown as sticks).

Comparison of our apo hACOD1 structure with that previously reported by Chen et al. (PDB 6R6U), shows a virtually identical protein conformation, with the exception of the now well-resolved water network at the active site. This robust crystallization procedure was vital in also obtaining the first ligated structure for this protein family, described in the next section.

2.2. Structural comparison of apo-from and citraconate-bound ACOD1

The inhibitor-bound structure was obtained from co-crystallization of hACOD1 with citraconate. Crystals for the complex were grown in similar conditions to that of the apo protein (Sup. Table 3) and diffracted to high resolution (1.32 Å, Table 1). Fig. 2, B and Fig. 3, A/B show the active site architectures of apo- and citraconate-bound hACOD1 (see Fig. 3, C for overlay). The apo structure shows six highly ordered water molecules, making up a hydrogen-bonding network linking residues His103, His159, Lys272, Lys207 and Tyr308. Citraconate binding displaces 5 of the waters with the oxygen atoms taking the positions of the displaced waters and thus satisfying most of the hydrogen-bonding network. The citraconate-bound structure demonstrates that the ligand binds at the active site and corroborates the putative active site proposed by Chen et al. (Chen et al., 2019).

Table 1.

X-ray data and model statistics.

	apo hACOD1	citraconate-hACOD1	hACOD1-R273S	hACOD1-Y318A	hACOD1-H103A
PDB ID	12UX	12VD	12VW	12VT	12VQ
Diffraction data DOI	10.15785/SBGRID/1277	10.15785/SBGRID/1278	10.15785/SBGRID/1279	10.15785/SBGRID/1280	10.15785/SBGRID/1281
SASBDB code	SASD272	SASD282	SASD2A2	SASD2B2	SASD292
Diffraction Source	NSLSII 17-ID-1 (AMX)	NSLSII 17-ID-2 (FMX)	NSLSII 17-ID-1 (AMX)	NSLSII 17-ID-1 (AMX)	NSLSII 17-ID-1 (AMX)
Wavelength (Å)	0.920	0.979	0.919	0.920	0.920
Temperature (K)	100	100	100	100	100
Detector	EIGER 9 M	EIGER 16 M	EIGER 9 M	EIGER 9 M	EIGER 9 M
Crystal to detector distance (mm)	161.33	158.91	161.33	115.69	115.69
Total rotation range (°)	360°	360°	360°	360°	360°
Rotation per image (°)	0.2°	0.2°	0.2°	0.2°	0.2°
Exposure time per images	5 ms	5 ms	5 ms	5 ms	5 ms
Space group	P21212	P21212	P21212	P21212	P21212
a, b, c (Å)	101.8, 110.1, 75.8	102.2, 110.4, 76.2	101.87, 110.19, 75.73	101.71, 109.99, 76.01	101.85, 110.04, 75.87
α β, γ (°)	90.0, 90.0, 90.0	90.0, 90.0, 90.0	90.0, 90.0, 90.0	90.0, 90.0, 90.0	90.06, 90.06, 90.08
Resolution range (Å)	33.78–1.35 (1.49–1.32)	33.9–1.22 (1.45–1.22)	44.55–1.46 (1.54–1.46)	34.55–1.62 (1.817–1.62)	34.50–1.42 (1.54–1.42)
Ellipsoid diffraction limits (a*, b*, c*):	1.39, 1.43, 1.51	1.32, 1.44, 1.51	1.47, 1.48, 1.59	1.78, 1.70, 1.82	1.496, 1.489, 1.534
Total no. of reflections	2,000,190 (84956)	2,117,439 (93179)	1,128,351 (62100)	1,120,717 (57300)	1,805,157 (92007)
No. of unique reflections	146,042 (7306)	155,814 (7790)	129,053 (6453)	80,794 (4040)	129,811 (6491)
Completeness (%)	Spherical: 74.0 (13.0) Ellipsoidal: 94.1 (55.6)	Spherical: 61.4 (7.8) Ellipsoidal: 94.7 (54.9)	Spherical: 87.1 (28.4) Ellipsoidal: 96.2 (64.3)	Spherical: 74.0 (12.9) Ellipsoidal: 94.4 (46.3)	Spherical: 80.4 (17.6) Ellipsoidal: 95.4 (51.9)
Redundancy	13.7 (11.6)	13.6(12.0)	8.7 (9.6)	13.9 (14.2)	13.9 (14.2)
<I/σ(I) > merged data	9.8 (1.0)	10.3 (1.7)	8.4 (1.5)	8.4 (1.5)	10.1 (1.4)
CC1/2	0.998 (0.331)	0.996 (0.415)	0.997 (0.621)	0.995 (0.550)	0.998 (0.598)
Rmerge	0.204 (2.503)	0.187 (1.98)	0.136 (1.48)	0.260 (2.35)	0.176 (2.43)
Rmeas	0.220 (2.742)	0.202 (2.15)	0.154 (1.65)	0.28 (2.52)	0.189 (2.62)
Rpim	0.057 (0.789)	0.053 (0.595)	0.050 (0.537)	0.072 (0.664)	0.049 (0.696)
Wilson B factor (Å2)	11.4	9.9	17.5	19.2	16.1
Refinement
Resolution range (Å)	33.78–1.32 (1.48–1.32)	33.9–1.22 (1.45–1.22)	44.55–1.46 (1.54–1.46)	34.55–1.62 (1.82–1.62)	34.53–1.42 (1.55–1.42)
No. of reflections, working set	146,042	155,813	129,053	80,793	129,811
No. of reflections, test set	7369	8014	6449	4198	6630
Final Rcryst	15.8	15.1	0.154	0.16	0.15
Final Rfree	19.3	17.9	0.184	0.202	0.181
No. of non-H atoms	15,462	14,434	15,172	15,001	15,516
Protein/nucleic acid	14,417	14,434	14,179	14,141	14,433
Ions	2	2	1	4	4
Ligands	37	28	19	15	37
Waters	1006	1026	973	841	1042
R.m.s deviations
Bonds (Å)	0.01	0.01	0.0093	0.007	0.0092
Angels (°)	1.823	1.896	1.729	1.538	1.73
Average B factors (Å2)
Protein/nucleic acid	16.3	15	18.9	19.4	16.2
Ions	26.5	20.6	27.5	23	14.9
Ligands	46.9	19.9	31.5	35.8	29.1
Waters	29.7	28	30.3	27.4	28.3
Ramachandran plot (Refmac5)
Favored regions (%)	95.31	95.53	95.3	95.2	95.6
Outliers (%)	0.22	0.22	3.9	4	3.7
Unmodelled/incomplete residues (%)	0	0	0	0	0

Fig. 3.

Crystal structures of apo hACOD1 (green) and citraconate-bound hACOD1 (blue). A) Citraconate-bound hACOD1 showing the hydrogen-bonding network between the inhibitor and active site residues (dashed lines, with distances between non-H atoms shown) and the 2F_o-F_c electron density map contoured at 1σ. B) Overlay of apo and citraconate-bound structures, showing minimal movement of residues in the active site, except for Tyr318, which is partly rotated. The waters in the apo form sit in positions occupied by oxygens of the citraconate carboxylates. C) ITC titration curve of citraconate into hACOD1. D) SAXS scattering profile of citraconate-hACOD1 (blue dots) fitted to the calculated solution scattering profile (black line) derived from the hACOD1–citraconate X-ray crystal structure (with weighted residuals, Χ² = 0.231). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The root mean square deviation (RMSD) between the apo and citraconate-bound structures is very low (0.097 Å globally and 0.108 Å between same chains), indicating that the protein overall conformation does not differ between the apo and ligated forms. As with apo hACOD1, the citraconate-bound structure is a compact dimer in solution (Fig. 3, D). From all the residues comprising the active site and part of hydrogen-bonding network, only Tyr318 shows a conformational change between the two structures, and it is minimal at 1.7 Å (Fig. 3, B). hACOD1 is a cofactor-free decarboxylase, expected to follow an acid-base catalytic mechanism (Chun et al., 2020a), but the identity of the residue(s) involved in the proton transfer steps is still unknown. The active site is well-defined and previous work proposes that His103 should be involved due to its proximity (Chen et al., 2019). In our ligand-bound structure, His103 is indeed well positioned to transfer a proton at the C2 position of citraconate.

2.3. Development of a low-consumption, continuous enzymatic assay for hACOD1

To pursue robust kinetic characterization of hACOD1, we developed a new, continuous, low-consumption kinetic assay, using ¹H NMR for quantification. This assay has some advantages over the previous reported end-point HPLC assay (Chen et al., 2019). First, it allows for the direct visualization of reaction progress and quantification of product formation/substrate consumption by integration of assigned proton signals. During the course of the experiment, ¹H NMR also allows for identification of any side reactions, for example, spontaneous conversion of cis-aconitate to trans-aconitate or formation of side products, neither of which was observed (see Sup. Fig. 1 for cis-aconitate and itaconate 1H spectra for calibration curves and a comparison to an enzyme turnover experimental sample). Second, it is a continuous assay, where multiple time-points can be measured from the same sample, greatly diminishing sample consumption and reducing experimental error from sample preparation. Third, the continuous assay allows for the determination of initial vs steady state rates in one measurement, and robust calculation of initial velocities from multi-point fitting. By interleaving data collection of multiple samples and measuring multiple time-points from a single sample, full kinetics can be obtained in triplicate within 48 h of machine time and using only 15 μg of protein per substrate/inhibitor concentration.

In this assay, both the product and substrate signals can be concurrently measured as they have distinct chemical shifts. Decarboxylation of cis-aconitate by hACOD1 (Chun et al., 2020a) converts its tri-substituted alkene to a geminal alkene (Fig. 4, A). The signals for the protons on the alkene are distinct and quantifiable ¹H NMR (Fig. 4, B, Sup. Fig. 1). Quantification can be done either by the ratio of proton integrals for the product or direct quantification using an internal standard (see Sup. Fig. 1 for concentration dependent spectra). The initial rate for each kinetic assay was derived by fitting a linear regression to the linear regime of itaconate production, corresponding to the initial ∼10–20% turnover of cis-aconitate. Given the temperature-dependence of ¹H chemical shifts, we found that temperature equilibration of samples prior to adding protein to solution was necessary to extract accurate measurements. hACOD1 catalytic properties were modelled using Michaelis-Menten kinetics, giving a k_cat = 0.83 ± 0.03 s⁻¹ and K_M = 0.57 ± 0.06 mM at 25 °C. These values were comparable to those previously reported (Zhao et al., 2025).

Fig. 4.

Kinetics and inhibition of hACOD1 assays by ¹H NMR. A) structures of cis-aconitate, itaconate and citraconate showing the distinct chemical shifts associated with the alkene protons for each species. B) Overlay of ¹H NMR spectra of hACOD1 kinetic assay over a two-hour time course. C) hACOD1 enzymatic activity characterization by Michaelis-Menten kinetics. D) Double reciprocal plot of hACOD1 inhibition by citraconate showing clear direct competition with substrate. E) Dose-dependent inhibition of hACOD1 by citraconate for IC₅₀ calculation. F) Kinetic and inhibition constants for hACOD1.

We used this assay to determine the inhibition kinetics of hACOD1 by citraconate. While similar in structure, citraconate has distinct proton chemical shifts from cis-aconitate and itaconate, and thus the same data analysis methodology could be used (Fig. 4, A, Sup. Fig. 1). Similarly to the assays preformed on apo-protein, two-hour time-course experiments were used to measure reduction in hACOD1 enzymatic activity in the presence of inhibitor. The in vitro IC₅₀ measured in this study of 44.1 ± 9.0 μM is comparable to that previously determined biochemically and by cellular studies (Fig. 4, E/F) (Chen et al., 2022b). Evaluation of Michaelis-Menten kinetics at varying citraconate concentrations provided a Lineweaver-Burk plot, which yielded a K_i of 23.5 ± 4.8 μM. The y-axis intercept at 1/Vmax with changing citraconate concentration together with changing of x-axis intercept (not shown) shows clear competitive inhibition, as previously reported by Chen et al., showing similar results between two distinct assays (Chen et al., 2022b).

And finally, for further characterization of citraconate, we also measured the binding thermodynamics of the hACOD1-citraconate interaction using isothermal titration calorimetry (ITC, Fig. 3, C). Fitting of the ITC data to a single site mechanism yields a K_d of 35.5 ± 8.56 μM, within range of the calculated IC₅₀ for this inhibitor. The binding is strongly enthalpic (ΔH = 11 kcal•mol⁻¹) and shows a 1:1 interaction (N value of 1.07 ± 0.209) between protomer and ligand (2 ligand molecules per hACOD1 dimer).

2.4. Mutagenesis and molecular dynamics studies shed light on active site mechanisms

With a robust kinetic assay and thermodynamic measurements in hand together with high-resolution, artifact-free crystal structures of apo and inhibited hACOD1, we set out to do further mechanistic studies. From a sequence alignment of hACOD1 homologues (Sup. Fig. 3) and comparison to our crystal structures, we chose 4 mutants aimed at further elucidating structure-activity relationships.

The first mutant investigated was hACOD1-His103Ala, a mutation of a fully conserved residue in the active site. His103 mutants showed no appreciable activity in previous work (Chen et al., 2019). The point mutant yielded stable, well folded protein (solution SAXS showin in Sup. Fig. 2) and, as expected, the protein had no measurable catalytic activity. The crystal structure shows a well-defined active site (Fig. 5, A), matching the apo structure, but with changes to the ordered water network as the H-bond donor/acceptor pair of the histidine residue is removed. Furthermore, the structure also shows the presence of a bound glycerol molecule in the active site, promoted by the enlarged active site cavity from the reduced Ala103 amino-acid size. To elucidate whether His103 is only necessary for catalysis or also affects binding at the active site, we performed ITC binding experiments with citraconate and cis-aconitate, which showed no measurable binding. Co-crystallization in the presence of citraconate showed no evidence of binding from electron density (not shown). His103 is therefore necessary at the binding stage of the catalytic cycle.

The WT apo hACOD1 structure shows that the active site is small and shielded from bulk solvent by loops Asp153-His159 Ile313-Val319, specifically with Pro155 and Tyr318 blocking the active site entrance (Fig. 6, C). Tyr318 had been previously identified as a residue of interest for hACOD1 activity (Chen et al., 2019). We thus proposed that Pro155 and Tyr318 should be important for hACOD1 function, where Pro155 is not a conserved residue and Tyr318 is semi-conserved (see Sup. Fig. 3).

Fig. 6.

Dynamics in hACOD1. A) and B) MD simulation results for WT apo hACOD1, showing the tight distance distribution between Tyr318 and Pro155 Cα (A) and the general dynamics across the whole protein as a root-mean-square-fluctuation (RMSF, B). The RMSF shows similar dynamic patterns across the two protomers, with differences between the chains propagated from the initial crystallographic structure. A spike at residue ∼298 in chain A is an artifact carried from a conformational constraint caused by a crystal contact. This initial constraint is not resolved during the simulation first annealing step. C) surface and cavity representation of the X-ray structure, showing the active site shielded from bulk solvent by Tyr318. D) and E) surface and cavity representation of the closed and open conformations in chain B during the MD simulation. Swinging of Tyr318 causes the small, shielded active site cavity to open to open to the bulk solvent.

The hACOD1-Pro155Ala mutant was found to undergo slow aggregation after purification, preventing crystallization, and thus a structure could not be obtained. With freshly prepared protein, kinetic assays showed no catalytic activity. Therefore, the non-conserved Pro155 plays an important role on both protein stability and enzymatic function. By contrast, hACOD1-Tyr318Ala yielded stable, well folded protein (Sup. Fig. 2), but was not catalytically active and unable to bind citraconate or cis-aconitate (from ITC isotherms). Inspection of the hACOD1-Tyr318Ala structure (Fig. 5, B) shows a well-ordered active site, with fewer water molecules visible compared to the WT protein, indicating a loss of part of the hydrogen bonding network. The active site cavity is now exposed to the bulk solvent. The aspergillus ACOD1 homologue has a phenylalanine at this position (Sup. Fig. 3): a large, hydrophobic residue, with no hydrogen-bond forming sidechain. From these observations, stable binding of molecules to in the active site most likely requires the shielding of the cavity but not a hydrogen bond donor at this position, as initially expected from the hACOD1-citraconate structure where the hydroxyl group of Ty318-OH interacts with the ligand. For catalysis to occur, Tyr318 must change conformation to allow for access to the binding site cavity during substrate binding and product release.

To further elucidate the dynamics involved in this mechanism, we ran Molecular Dynamic (MD) simulations starting from our WT apo hACOD1 X-ray structure. Molecular dynamic simulations of the dimeric hACOD1 show clear conformational fluctuations around Tyr318 and Pro155 (in the alpha-helical and lid domains respectively, Fig. 6, B). Plotting the mean distance between the alpha‑carbons of these two residues samples during the simulation shows that their relative positions fluctuate minimally at around 9 Å (Fig. 6, A). SAXS data for the WT and Tyr318Ala mutant fit well to the solved crystal structures (Fig. 2, C and Sup. Fig. 2), corroborating that the global conformation of these proteins in solution is similar to that obtained by crystallography. To assess flexibility, we performed a dimensionless Kratky Analysis (Sup. Fig. 2), which showed low flexibility, similar across all the datasets, including the citraconate-bound WT hACOD1 sample. Together, these observations indicate that necessary conformational changes associated with accessibility of the active site occur on a local scale, rather than through a full domain open/closing mechanism. Frames from the MD trajectory indeed show different possible conformations for Tyr318, where the active site becomes accessible from swinging of the tyrosine side chain (Fig. 6, D and E). Though no experimental high-resolution structures of human ACOD1 show this open conformation, a recent structure of murine ACOD1 indeed captures the protein with an accessible active site, caused by the movement of these same residues (Chun et al., 2020b).

2.5. Molecular dynamics and mutagenesis studies shed light on allosteric mechanisms

Further inspection of frames from the MD trajectory, revealed significant local dynamics and conformational sampling. We decided to further investigate one such mobile residue, Arg273, a solvent exposed residue next to the conserved active-site Lys272. In murine ACOD1, residue 273 is a serine and the enzyme is known to be faster and more efficient (Chen et al., 2019). Previous work showed that an Arg273His mutation (a missense polymorphism found in hACOD1) leads to a decrease in K_M, with no significant change in k_cat (Chen et al., 2019).

hACOD1-Arg273Ser also shows changes in kinetic properties, with a slight decrease in K_M and k_cat compared to the WT protein (Fig. 7). ITC corroborates the K_M change, with citraconate binding more tightly at 20.7 μM for the Arg273Ser mutant compared to 35 μM for WT hACOD1 (Fig. 7C and Fig. 3 C). The ITC curve also shows saturation of the enzyme at an N value of 0.6, closer to a 1:2 binding per protein monomer than the 1:1 binding observed for WT hACOD1. As ACOD1 is a symmetric dimer, this lower N value hints at a cooperative mechanism between the active sites in the two protomers, but more work will be needed to corroborate this hypothesis. Crystal structure of hACOD1-Arg273Ser shows very clear changes in the environment around these residues. Arg273 (in the lid domain) binds to Asp417 (α-helical domain) through a single water-bridge (Fig. 7 A). In contrast, Ser273 loses this tight hydrogen bonding network (Fig. 7C). Though alone this mutation does not recapitulate the increased activity of murine ACOD1, these studies indicate that it should contribute. With the active site residue Lys272 adjacent to Arg273 and the changes in hydrogen bonding network between the two domains from the point mutation and the ITC data, we suggest that this region has a dynamic allosteric role in this enzyme's mechanism.

Fig. 7.

hACOD1-Arg273Ser mutant characterization. A) and B) show the local structure and water network surrounding residue 273, which is solvent exposed. The single-water bridge between Arg273 and Asp417 is lost upon mutation to serine. Hydrogen bonds are shown as dashes, with lengths of 2.6–3.2 Å in black and 3.2–3.5 Å in gray. C) ITC of hACOD1-Arg273Ser mutant binding to citraconate. D) Michaelis-Menten kinetic model of hACOD1-Arg273Ser.

3. Conclusion

Human aconitate decarboxylase is a target of interest in the oncometabolic drug discovery community. With many studies linking itaconate production to different cancer forms, the discovery of a clinically tractable inhibitor is of great importance. To pursue structure-guided drug discovery, detailed mechanistic and biochemical characterization is needed.

In this publication, we have expanded the biochemical, structural, and dynamic knowledge of hACOD1. We solved the first artifact-free apo structure of hACOD1, a necessary step towards obtaining ligand-bound structures. We also show the first hACOD1 structure with an orthosteric ligand (citraconate), which is also the first purposedly-bound structure of a member of the MmgE-PrpD family of proteins. We also developed a new, versatile, low-consumption, high-sensitivity, continuous kinetic assay.

The protein was found to have a similar structure in solution to that determined by X-ray crystallography. The dimensionless Kratky plot of the SAXS data shows no flexibility or dynamic changes between the apo and inhibited structures, indicating that any conformational changes associated with access to the active site occur on a local length-scale rather than through large conformational changes. Together with MD simulations and point-mutations, we demonstrated that Tyr318, though not fully conserved throughout the family, is important for both binding and catalysis and is involved in the shielding of the active site during catalysis. Local conformational changes of Tyr318 as visualized by MD are sufficient to open/close access to the small active site.

Other point mutants were also biochemically and structurally characterized. His103Ala, an active site mutation, abolishes both binding and catalysis. Arg273Ser, an allosteric mutation corresponding to the murine homologue, impacts stoichiometry of citraconate binding to the enzyme as well as the kinetic parameters. None of the point mutants showed large conformational or dynamic changes in solution. More biochemical and structural work will be needed to fully understand why homologous proteins have differing kinetics as well as to determine allosteric sites for drug discovery. Allosteric mechanisms allow for the design of molecules beyond active-site binders whose chemistry may be limited due to the highly specific chemical environment of this cofactor-free decarboxylase. The development of allosteric inhibitors may be optimal for hACOD1 due to the small pocket volume of the active site and the specific electrostatics necessary to bind the tri-carboxylic acid substrate.

4. Materials and methods

4.1. Protein expression

The expression plasmid pCAD29_hIRG1_4–461_pvp008 was a gift from Konrad Buessow (Addgene plasmid # 124843; http://n2t.net/addgene:124843; RRID:Addgene_124,843) (Chen et al., 2019). The plasmid codes for a truncated version of ACOD1, lacking short unstructured N- and C-terminal motifs. Escherichia coli BL21-CodonPlus (RIPL) cells (Novagen) were used for protein expression. Additionally, singe-point mutations H103A, P155A, R273S, Y318A of truncated ACOD1 (4–461) were prepared in pCOLA-Duet vector for mutational analysis of enzymatic catalysis. hACOD1 (4–461) was expressed and purified as reported previously with modifications. In brief, transformed cells were cultured in 5–10 mL of Luria-Bertani (LB) medium supplemented with 50 μg/mL kanamycin and 30 μg/mL chloramphenicol. The LB pre-culture was incubated at 37 °C with shaking at 200 rpm until the OD₆₀₀ reached 1.2–1.5 and 1 mL of this pre-culture was used to inoculate 50 mL of LB medium. After overnight growth at 37 °C, the small LB culture was transferred to 1 L of LB medium, incubated at 37 °C with shaking at 200 rpm up to an OD₆₀₀ of 0.75–0.80. At that point, the temperature and agitation were lowered to 22 °C and 130 rpm, respectively. After ∼1 h, protein expression was induced with isopropylthio-β-galactoside (IPTG), at a final concentration of 500 μM. Cells were grown at 22 °C for 18 h and harvested at 6000 x g for 10 min at 4 °C. Cell pellets were harvested and stored at −80 °C until further use.

4.2. Protein purification

Frozen cell pellets were resuspended in 50 mL of buffer A (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 10% v/v glycerol, 1 mM DTT) per 20 mg of cells. The resuspension was treated with a final concentration of 1 mg/mL lysozyme, 0.01 mg/mL RNAse, and 0.005 mg/mL DNAse and the cells disrupted by sonication (Fisherbrand Qsonica 505) at 60% power for 5 min (10 s pulse on and 20 s pulse off) on ice. The cellular lysate was clarified with 6–8 mL of 10% PEI followed by centrifugation at 16,500 xg for 1 h at 4 °C. After clarification, the supernatant was loaded onto an equilibrated 5 mL Strep-Tactin XT 4Flow column (IBA Lifesciences cat#: 2–5010-025) the column washed with 5 CV of buffer A (or until UV absorbance returned to baseline) and the protein eluted in buffer A supplemented with 500 mM Biotin. The elution fractions were checked using SDS-PAGE and those containing protein were pooled and dialyzed overnight in buffer A at 4 °C with 1.5% w/w TEV protease to cleave the Strep-tag. The protein was concentrated (3–4 mL) and further purified to homogeneity by size-exclusion chromatography (HiLoad 16/600 Superdex 200 prep grade, Cytiva) by isocratic elution with buffer B (10 mM HEPES, 150 mM NaCl, 10% v/v glycerol, 0.1 mM TCEP, pH 7.5). The purified protein was concentrated to 120 μM (6 mg/mL) and flash-frozen in small aliquots in liquid nitrogen for storage at −80 °C.

4.3. Protein quantification

Protein quantification was performed by UV–Vis spectroscopy, measuring the absorption at 280 nm. Accurate quantification was performed on an 8453 UV–Vis (Agilent) spectrophotometer and corrected using a three-point Morton-Stubbs correction (see Sup. Fig. 4 for details). The same sample was then measured on a NanoDrop™ UV–Vis Spectrophotometer (Thermo Scientific), yielding an approximate concentration, not baseline corrected. A ratio between the “true” and “approximate” concentrations was calculated and applied henceforth to all measurements routinely taken on the Nanodrop. This method showed that concentration using the Nanodrop was approximately 25% overestimated and correction yielded ITC fits for the hACOD1-citraconate binding much closer to an N of 1, validating this approach.

4.4. Protein crystallization screening and optimization

Initial crystallization conditions for hACOD1 were obtained from 6 sparse matrix crystallization screens (JSCG+ and Structure I + II HTS – Molecular Dimensions - and Wizard HTS, PACT HT, PEG/ION, and INDEX HT- Hampton Research). Screening was performed at room temperature (∼22 °C) using sitting drop vapor diffusion in MRC3 crystallization plates (SWISSCI) using 200 nL total volumes and varying the protein to cocktail ratio in 1:1, 2:1 and 1:2 v/v ratios for each condition and constant 40 μL reservoir volume. The drops were set using the Formulatrix NT8 drop setter, at a constant humidity of 80%. The plates were stored and imaged at 22 °C using a Fomulatrix R1000 imager, equipped with bright field and UV absorption capabilities. From the observed protein crystals, 11 hits were identified as devoid of polydentate carboxylate components.

Two optimized conditions yielded good diffracting crystals: 1) 100 mM Tris (pH 8.8), 0.2 mM CaAc, 35% PEG4000; and B) 200 mM NaF, 35% PEG4000. Optimization was performed by varying the concentrations and pH of the crystallization cocktail components using an automated formulating system (FORMULATOR, Formulatrix) in the same tray layout described above. Co-crystallization screening of citraconate with hACOD1 was performed by varying the citraconate concentration between 2.5 mM and 10 mM to reach 99.0% occupancy based on K_d = 35 μM measured by ITC. Fully grown crystals of both apo hACOD1 and citraconate-bound hACOD1 were obtained in 4–6 days. hACOD1 mutant proteins crystallized in the same conditions as the WT protein. Drops containing crystals were layered with LV cryoOil (Mitegen) as a cryoprotectant, looped and vitrified in liquid nitrogen.

4.5. Data collection and reduction and structure solution

Data for the apo and citraconate-bound structures were collected at beamlines 17-ID-2 (FMX) (Schneider et al., 2021) and 17-ID-1 (AMX) (Schneider et al., 2022) in fully automated mode. Parameters for data collection can be found in Table 1. Raw X-ray images for all datasets were deposited in SBGrid and corresponding DOIs can also be found in Table 1. The X-ray images were indexed, integrated, scaled and merged using the automated autoproc pipeline (Global Phasing) (Vonrhein et al., 2011) available at NSLSII. For datasets requiring further manual data reduction (apo- and citraconate-bound hIRG1), the output of the XDS (Kabsch, 2010) INTEGRATE step from autoproc was parsed through pointless (within Aimless (Evans and Murshudov, 2013), CCP4i2 (Potterton et al., 2018)) for space group determination and scaling. The unmerged pointless output mtz file was reduced in StarANISO (STARANISO, 2016) (Global Phasing) to yield an anisotropic resolution cut off. Scaled and merged files from either the autoproc pipeline or from manual StarAniso runs were imported into CCP4i2, solved by molecular replacement using MOLREP (Vagin and Teplyakov, 1997) with initial model PDB 6R6U (for the apo WT structure, with subsequent datasets solved using our apo structure) (Chen et al., 2019). Rounds of manual model building using Coot (Emsley et al., 2010) and refinement using Refmac5 (Murshudov et al., 2011) were performed in CCP4i2. Refinement was performed using automated TLS parameters. TLS groups were used for refinement instead of anisotropic b-factors as these caused clear overfitting, with a widening gap between Rfree and Rcryst and no significant decrease in Rfree.

4.6. Solution NMR spectroscopy data collection and processing

1D ¹H solution NMR spectra were collected at 298 K for cis-aconitic acid on a 14.1 T Bruker AvanceIII and 17.4 T Bruker AvanceIII spectrometer each equipped with a TCI four-channel inverse detection H/C/N/D cryoprobe. Samples were locked to 90:10 H₂O/D₂O. T ¹H 1D spectra were recorded at pH 7.5 to match protein and ligand stocks. All NMR data were processed using Bruker TopSpin and MestreMnova (MestreLab Research). ¹H chemical shifts were referenced to the water peak at 4.7 ppm. All solution spectra were analyzed using MestreMnova. ¹H solution chemical shifts were assigned based on the published cis-aconitate and itaconate chemical shifts (Biological Magnetic Resonance Data Bank entries bmse000705 and bmse000137, respectively).

4.7. Enzyme kinetic assay

Enzyme kinetics of hACOD1 were obtained by a continuous assay observing the conversion of cis-aconitic acid to itaconic acid by ¹H NMR. Initial velocity of enzyme progression curves was determined by measuring production of itaconate. Itaconate concentration was obtained by comparison of integrals against a standard curve of itaconate concentration.100 mM cis-aconitate and citraconate stock solutions were prepared using buffer B. 50–200 nM hACOD1 samples were prepared with a final cis-aconitate concentration ranging from 0.25 to 15.0 mM to a final volume of 400 μL. ¹H 1D spectra (32 K points) of final reaction mixtures were measured at 12-min time intervals over the course of two hours at low cis-aconitate concentrations (0.5–3.0 mM) and 60-min intervals for 8 h at high cis-aconitate concentrations (8.0–15.0 mM). Samples prepared for citraconate inhibitory assays were measured under constant protein (50 nM) and varying substrate concentrations (20–200 μM) with citraconate concentrations ranging from 0.2 to 1.0 mM. All time points at the different conditions described were measured in triplicate.

All samples were temperature equilibrated by water-insulated heating block at 25 °C. Data was fit by Prism 10 (GraphPad) using Michaelis-Menten non-linear regression to calculate catalytic rate constant (k_cat), maximum velocity (V_max) and Michaelis constant (K_M). Citraconate concentration resulting in 50% inhibition of hACOD1 activity (IC₅₀) was determined by non-linear regression of sigmoidal dose-response. Each point represents three independent experiments error bars are ±1 SEM. Mode of citraconate inhibition was determine by double reciprocal plot assaying inhibitor concentration (0–1000 μM citraconate) at varying substrate concentration (1–5 mM cis-aconitate).

4.8. Small-angle X-ray scattering data collection and analysis

In-line size-exclusion chromatography small-angle X-ray scattering (SEC-SAXS) experiments were performed at beamline 12-ID-B of the Advanced Photon Source (APS) at Argonne National Laboratory. WT apo and citraconate hACOD1 as well as single point mutants were analyzed at injection concentrations ranging from 5.0 to 6.05 mg/mL. Each sample was injected onto a Superdex 200 Increase 5/150 GL column connected to an ÄKTA micro FPLC system using a 100 μL sample loop. The samples were eluted at a flow rate of 0.3 mL/min and directed to the flow cell for simultaneous small- and wide-angle X-ray scattering (SAXS and WAXS) measurements. The running buffer consisted of 20 mM HEPES (pH 7.5), 150 mM NaCl, 0.5 mM TCEP, and 2% v/v glycerol. Data were collected at an X-ray energy of 13.3 keV (λ = 0.9322 Å). The combined SAXS/WAXS setup covered a momentum transfer range of 0.005 < q < 2.7 Å⁻¹, where q = (4π/λ)sinθ, 2θ is the scattering angle, and λ is the X-ray wavelength. A total of 1500–2000 frames were collected per sample with an exposure time of 0.2–0.3 s per frame.

Two-dimensional scattering images were corrected for detector geometry (solid angle per pixel) and reduced to one-dimensional scattering profiles using MATLAB-based beamline software. SEC-SAXS data were further processed using BioXTAS RAW. Frames corresponding to monodisperse regions of the elution peak were identified and averaged. Background scattering was estimated from frames collected before and after the elution peak, and baseline correction was applied where necessary. The radius of gyration (Rg) was determined by Guinier analysis, using data within the range qRg ≤ 1.3. Experimental SAXS profiles were compared with theoretical scattering curves calculated from atomic models derived from crystal structures using CRYSOL (Franke et al., 2017; Svergun et al., 1995). During fitting, parameters including the hydration layer, excluded volume, and contribution of implicit hydrogen atoms were taken into account.

All SAXS data was deposited to SASBDB and the corresponding dataset identifiers can be found in Table 1.

4.9. Isothermal titration calorimetry

Enzyme samples for isothermal titration calorimetry (ITC) were prepared to a final concentration of 35 μM in buffer B. Citraconate and cis-aconitate was prepared to a final concentration of 1 mM by 1:1000 dilution of 1 M stock in buffer B. The experiments were performed with an iTC200 Calorimeter (Malvern Panalytical/MicroCal, Netherlands/USA) at 25 °C. The experiment consisted of 18 injections (2.1 μL each) of citraconate or cis-aconitate into a hACOD1 in the cell (200 μL) at a stirring speed of 750 RPM. A sacrificial first injection of 0.5 μL of ligand was used to accommodate the interaction during pre-titration thermal equilibration at the tip of the titration syringe; this measurement point was excluded from the final set of data. An additional set of injections was run in a separate experiment with buffer B in the cell instead of the protein solution with identical run parameters as a blank experiment for subtraction during data processing. K_d, stoichiometry, enthalpy and entropy changes were determined from integrated binding isotherms using the “One Set of Sites” model in Origin 7.0 in the Malvern/MicroCal data analysis software.

4.10. MD simulations

Simulations were performed in GROMACS (Abraham et al., 2015). Each was started from apo-hACOD1 (residues 4–461) solvated in a rhombic dodecahedral water box; charges were neutralized with sodium and chloride ions and addition 150 mM NaCl was present to reproduce experimental conditions. The system was equilibrated for 1 ns in the NPT ensemble before starting production runs. Ten replicates of 200 ns with a 2 fs time step were run in parallel starting from the same equilibrated structure. Trajectories were analyzed with MDAnalysis (Michaud-Agrawal et al., 2011). The distance between the Cα of Tyr318 and Cα of Pro155 was measured for each protomer over every frame in the trajectories. The Cα root mean square fluctuations (RMSF) were determined by first calculating an average structure over all trajectories, aligning to this structure and calculation of the RMSF using the average structure as a reference.

CRediT authorship contribution statement

Brent Runge: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Hande Oktay: Formal analysis, Investigation. Ian J. Fucci: Investigation. Eric M. Merten: Investigation. Sergey G. Tarasov: Investigation. Lixin Fan: Investigation. Diana C.F. Monteiro: Conceptualization, Investigation, Methodology, Supervision, Writing – original draft, Writing – review & editing, Project administration, Funding acquisition, Formal analysis, Data curation.

Funding

This research was supported by the Intramural Research Program of the NIH. The Center for BioMolecular Structure (CBMS) is primarily supported by the National Institutes of Health, National Institute of General Medical Sciences (NIGMS) through a Center Core P30 Grant (P30GM133893), and by the DOE Office of Biological and Environmental Research (KP1605010). The Advanced photon source is a U.S. Department of Energy (DOE) Office of Science user facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. The National Synchrotron Light Source II, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Brookhaven National Laboratory under Contract No. DE-SC0012704.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material

Data availability

All protein models, X-ray diffraction and scattering data were deposited in the PDB, SBGrid repository and the SASBDB, respectively. Accession codes are given in Table 1.