Structural analysis of Atopobium parvulum SufS cysteine desulfurase linked to Crohn's disease

Abstract

Crohn’s disease (CD) is characterized by the chronic inflammation of the gastrointestinal tract. A dysbiotic microbiome and a defective immune system are linked to CD, where hydrogen sulfide (H₂S) microbial producers positively correlate with the severity of the disease. Atopobium parvulum is a key H₂S producer from the microbiome of CD patients. In this study, the biochemical characterization of two Atopobium parvulum cysteine desulfurases, ApSufS and ApCsdB, shows that the enzymes are allosterically regulated. Structural analyses reveal that ApSufS forms a dimer with conserved characteristics observed in type II cysteine desulfurases. Four residues surrounding the active site are essential to catalyse cysteine desulfurylation, and a segment of short-chain residues grant access for substrate binding. A better understanding of ApSufS will help future avenues for CD treatment.

Abbreviations

ApSufS, SufS from Atopobium parvulum

BsSufS, SufS from Bacillus subtilis

CD, Crohn’s disease

DTT, dithiothreitol

GI, gastrointestinal tract

H₂S, hydrogen sulfide

IBD, inflammatory bowel disease

IPTG, isopropyl β-d-1-thiogalactopyranoside

ISC, iron-sulfur cluster system

k_cat, product turnover

K_m, Michaelis–Menten constant

LB media, Luria–Bertani media

NIF, nitrogen fixation system

OD, optical density

PBS, phosphate-buffered saline

PLP, pyridoxal 5′ phosphate

SUF, sulfur formation system

Inflammatory bowel disease (IBD) is a global disease, where newly industrialized countries are seeing an accelerated incidence [[1]] and burden remains high in western countries [[2]]. One subtype of IBD is Crohn’s disease (CD), a chronic condition characterized by the excessive inflammation of the gastrointestinal (GI) tract [[3]]. Unfortunately, the treatments available to CD patients only attempt to reduce symptoms and mitigate its advancement to more severe stages. The molecular events underlying the appearance and progression of CD are still debated; however, evidence suggests both hereditary and environmental factors can promote chronic intestinal inflammation through microbial dysbiosis [[4-7]]. Recently, network correlation analysis has identified Atopobium parvulum as a key hub of hydrogen sulfide (H₂S) producers with a link to gut inflammation and CD pathogenesis [[8]]. The relative abundance of A. parvulum positively correlated with CD severity in paediatric patients. A. parvulum induced severe pancolitis in specific pathogen-free Il10^−/− mice, and bismuth (III)-subsalicylate, a known H₂S scavenger, alleviated Atopobium-induced colitis suggesting that H₂S production plays an important role in triggering colitis or in maintaining inflammation [[8]].

In humans, there are three sources of H₂S in the GI tract. First, H₂S can transit between tissues to eventually penetrate the gut. Second, H₂S can be produced locally by sulfate-reducing bacteria through anaerobic respiration of sulfate (SO₄). Lastly, H₂S can be released as a by-product of cysteine desulfurases expressed by bacteria belonging to the GI microbiome [[9, 10]]. Cysteine desulfurases are pyridoxal 5′-phosphate (PLP)-dependent homodimeric enzymes that facilitate sulfur mobilization by cleavage of the C-S bond in L-cysteine [[11, 12]]. This cleavage is the initial step of thio-cofactor biogenesis required for metabolic activities such as cellular redox sensing, electron-transfer reactions and maintaining physiological levels of free sulfur and iron [[12, 13]]. This cysteine desulfurase activity is typically coupled with a mechanism that allows for the transfer of the sulfide from the cysteine desulfurase to various acceptor molecules. With the sulfide transfer, the remaining L-alanine is released to regenerate the free cysteine desulfurase [[14]].

Currently, there are three main machineries of cysteine desulfurases that all play a role in regulation of iron–sulfur (Fe-S) biogenesis: ISC (iron–sulfur cluster), SUF (sulfur formation) and NIF (nitrogen fixation) systems [[15-17]]. Cysteine desulfurases can be separated into two groups based on their structures and catalytic activities. Common to both groups, the PLP cofactor binds to a conserved lysine residue to form an internal aldimine bond in the resting state of the cysteine desulfurase [[11]]. Additionally, a conserved nucleophilic cysteine residue facilitates the attachment of the sulfhydryl group from the substrate cysteine to the enzyme, forming a cysteine persulfide intermediate (R-S-SH) [[11]]. Group I cysteine desulfurases harbour a flexible extended loop that allows for improved active site dynamics to facilitate catalysis [[12]]. However, the extended loop anchoring this catalytic cysteine is significantly different between group I and II cysteine desulfurases [[18]]. Conversely, group II cysteine desulfurases display relatively decreased activity that is proposed to be a consequence of their shortened and structurally rigid catalytic loop [[19, 20]]. Furthermore, group II cysteine desulfurases can interact with different acceptors that enhance their enzymatic activity [[21]].

Recent evidence shows the impact of cysteine desulfurases on the growth and pathogenesis of diverse bacterial species. For example, the cysteine desulfurase IscS is mainly responsible for H₂S generation in Escherichia coli under anaerobic conditions, maintaining E. coli growth and bioenergetics [[22]]. Under oxidative stress or iron-starved conditions, the SUF operon is activated and contributes substantially to the release of H₂S through the cysteine desulfurase SufS [[23-25]]. The A. parvulum genome does not contain an IscS homolog, but does express SufS and CsdB, two putative group II cysteine desulfurases. Due to the positive correlation of the H₂S producer A. parvulum with CD severity, we sought to characterize ApSufS and ApCsdB. Biochemical characterization reveals that ApSufS and ApCsdB are bona fide cysteine desulfurases. Kinetic characterization demonstrates that ApSufS is a more efficient cysteine desulfurase compared to ApCsdB, and that at least four key residues are critical for H₂S formation in ApSufS. The crystal structure of ApSufS shows conserved features of group II cysteine desulfurases, including the extended loop harbouring the catalytic cysteine. With the global rise of IBD cases, a better understanding of the ApSufS structure will allow us to explore its inhibition as a more specific and compelling avenue of IBD treatment.

Materials and methods

ApSufS and ApCsdB purification

The cDNA of full-length A. parvulum SufS and CsdB were cloned in the pET-11a (Genscript, Piscataway, NJ, USA) vector using Nde1 and BamH1 restriction sites with codons optimized for expression in E. coli. Mutants were generated using the QuickChange II XL Site-Directed Mutagenesis kit from Agilent and mutations were confirmed by sequencing. The plasmids encoding wild-type or mutant ApSufS as well as ApCsdB were transformed in Rosetta cells and grown in LB media supplemented with ampicillin and chloramphenicol. Transformed cells were grown at 37 °C to ~ OD⁶⁰⁰ of 0.5 before induction with 0.1 mm isopropyl β-d-1-thiogalactopyranoside (IPTG) and then left at 18 °C for 16 h. The cells were harvested and suspended in 1X PBS pH 7.4 supplemented with 5 mm β-mercaptoethanol and then frozen down at −80 °C for storage. On the day of purification, cells were lysed by sonication and the resuspension was clarified by centrifugation at 30 883 g, 4 °C and filtration with a 0.45 µm filter. The soluble fraction showed the characteristic yellow coloration expected due to the binding of cofactor PLP to ApSufS and ApCsdB. The supernatant of each cell lysate was applied to TALON^® Metal Affinity pre-equilibrated in 1X PBS pH 7.4 and 5 mmβ-mercaptoethanol. After washing extensively with 1X PBS, proteins were eluted using 1X PBS supplemented with 500 mm imidazole. The eluted proteins were dialysed in 50 mm HEPES pH 7.9, 150 mm NaCl and 5 mm β-mercaptoethanol for 16-18h. After concentration on an Amicon®Ultra-15 Centrifugal Filter Unit, the proteins were further purified by size-exclusion chromatography pre-equilibrated in 10 mm Hepes pH 7.9, 150 mm NaCl and 5 mm β-mercaptoethanol. Following concentration, the proteins were flash-frozen in 5% glycerol and stored at −80 °C until further use for crystallography or enzymatic assays.

Crystallization of ApSufS

ApSufS crystals (wild-type and mutants) were grown by sitting drop vapour diffusion using various conditions (Table 1). Crystals were harvested, then soaked in the mother liquor and cryo-preserved in liquid nitrogen. A full data set was obtained using a Micromax-007HF equipped with image plate detectors (Raxis IV⁺⁺). Data were reduced and scaled using Structure Studio. The structure of ApSufS was solved by molecular replacement using Bacillus subtilis SufS (PDB 5J8Q) as a search model. Using Phaser, one molecule was placed in the asymmetric unit for the wild-type ApSufS protein. Following several rounds of refinement and model building using Phenix and Coot respectively, the final model contains 417 amino acids and 302 water molecules. ApSufS mutants crystallized in various space groups with varying number of molecules in the asymmetric unit (Table 1). These structures were solved using wild-type ApSufS as search model and the model was improved as for the wild-type protein.

Table 1. Crystallographic data and refinement statistics for ApSufS. R-factor: R_working = Σ || F_o | – | F_c || / Σ | F_o | ; R_free = Σ_T || F_o | – | F_c || / Σ_T | F_o |, where T is a test data set of 5% of the total reflections randomly chosen and set aside before refinement.

	WT	C375S	A34Y	K235R
Crystallography conditions	10% glycerol 20% PEG300 10% PEG8000	22% PEG3350 50 mm NaMalate	24% PEG3350 50 mm NaMalate	10% PEG8000 12% ethylene glycol 50 mm Hepes pH 7.1
Crystal parameters
Space group	P 6122	P222	P212121	C121
Unit cell
a (Å)	92.01	71.12	76.31	91.74
b	92.01	110.61	86.78	159.26
c	178.24	122.34	133.43	178.06
α, β, γ (°)	90, 90, 120	90, 90, 90	90, 90, 90	90.00, 90.16, 90.00
Data collections statistics
Resolution range (Å)	29.7–1.8 (1.86–1.80)	29.8–2.2 (2.28–2.20)	36.3–2.1 (2.18–2.10)	38.9–3.0 (3.10–2.99)
Total reflections	734 687	992 863	52 261	204 383
Unique reflections	41 844 (3898)	48 408 (2393)	52 321 (5123)	49 720 (4738)
Rmeas (%)	0.062 (0.313)	0.064 (0.345)	0.075 (0.337)	0.084 (0.258)
I/sigma (I)	23.9 (5.1)	17.1 (3.9)	26.1 (8.1)	12.3 (5.1)
Completeness (%)	99.4 (94.6)	97.2 (98.0)	99.6 (99.1)	96.4 (93.3)
Refinement statistics
Resolution range (Å)	29.7–1.8 (1.86–1.80)	29.8–2.2 (2.28–2.20)	36.3–2.1 (2.18–2.10)	38.9–3.0 (3.10–2.99)
Reflections (Fo > 2σ)	41 839 (3898)	47 836 (4388)	52 251 (5121)	49 640 (4783)
Final model
Number of ApSufS in the A.U.	1	2	2	6
Protein atoms	3176	6318	6128	17 597
Ligand(s)	1	2	2	8
Water	302	352	172	0
R-Factors
R working	0.18 (0.24)	0.22 (0.29)	0.19 (0.19)	0.24 (0.27)
R free	0.22 (0.26)	0.26 (0.33)	0.24 (0.23)	0.29 (0.37)
Molprobity scores	1.04	1.14	1.22	2.24
R.M.S.
Bond length (Å)	0.018	0.004	0.008	0.006
Bond angles (°)	1.36	0.94	0.93	1.21
Average B—factors (Å^2)	31.41	30.54	24.41	57.64
Protein	30.51	30.45	24.51	57.65
Ligands	25.24	23.00	17.25	54.65
Water	41.16	32.79	22.10	–
PDB Code	7TLM	7TLQ	7TLR	7TLP

In vitro cysteine desulfurase assay

Reactions were performed in 384-well plates and initiated by the addition of purified wild-type ApCsdB (2.25 µm), wild-type ApSufS (1.125 µm) or mutant ApSufS (2.25 µm) into a reaction mixture composed of 0–15 mm of L-cysteine, 25 mm Hepes pH 7.9, 2.5 mm ZnCl₂, and 5 mm DTT. Reactions were sealed and quenched at different time points using a diamine solution (0.10 mm N,N-dimethyl-p-phenylenediamine sulfate, 0.40 mm FeCl₃). Following a 30 min incubation in the dark, absorbance was measured at 670 nm. Reaction velocities were calculated using the linear range of product produced vs. time. Time points 0–8 min were used for ApSufS, and 0–120 min were used for ApCsdB to calculate initial velocities for each cysteine concentration. Prism 8 was used to generate reaction velocities vs. substrate concentration plot and calculate kinetic parameters. Calibration was performed using Na₂S as standard.

Results and Discussion

Sequence analysis reveals four cysteine desulfurases in Atopobium parvulum

Two predominant sources can produce H₂S in the GI microbiome: sulfate-reducing bacteria or cysteine desulfurases. Given that a recent study reported no link between sulfate-reducing bacteria and CD patients [[8]], we decided to focus on cysteine desulfurases in the key H₂S producer A. parvulum. Using the B. subtilis genome as a reference, we found that A. parvulum contains two members from each group of cysteine desulfurases (Fig. 1A). This contrasts with the B. subtilis genome which encodes four group I cysteine desulfurases (YrvO, NifS, NifZ and YcbU) and one group II cysteine desulfurase (SufS). The A. parvulum genome contains a YrvO homolog that clusters with MnmA and is likely involved in 2-thiouridine modifications of tRNA [[26]]. The second group I cysteine desulfurase shares high homology with BsNifS and will be referred to as ApNifS. Intriguingly, the canonical NifS from B. subtilis and A. vinelandii is often clustered with its acceptor NifU. However, the ApNifS gene is not found in proximity of a sulfur acceptor, suggesting that the ApNifS may function differently from BsNifS and AvNifS. Akin to B. subtilis, a putative group II cysteine desulfurase gene linked to the Fe-S cluster biosynthetic pathway is found in the SufCBDSU operon, including the group II cysteine desulfurase SufS and its enhancer SufU (Fig. 1C). Curiously, in specific A. parvulum strains (such as DSM 20469), an additional group II cysteine desulfurase can be found in a short operon without known proximal sulfidic acceptor genes. Although it shows poor conservation with other group II cysteine desulfurases in the β-hairpin region, it contains a conserved motif in the extended loop. Thus, we have classified this gene as a group II cysteine desulfurase referred to as ApCsdB (Fig. 1B).

Fig. 1

Sequence analysis of cysteine desulfurases in Atopobium parvulum. (A) Phylogenetic analysis of cysteine desulfurases in Atopobium parvulum (Ap), Bacillus subtilis (Bs), Mycoplasma pneumoniae (Mp) and Escherichia coli (Ec). Enzymes are classified as Group I (orange) or Group II (green). (B) Multiple sequence alignment of cysteine desulfurase homologs in the regions of the β-hairpin and the extended loop region. Conserved sequence features are indicated in orange or green box for groups I and II, respectively. The conserved lysine residue (K235 in ApSufS) which forms the Schiff-base, the cysteine residue (C375) which forms the persulfide bond, and the arginine residue (R390) are indicated with blue, red and purple stars, respectively. (C) Schematic representation of the SUF operons.

ApSufS and ApCsdB are bona fide cysteine desulfurases

Given that the group II cysteine desulfurase SufS is known to enhance Mycoplasma pneumonia pathogenesis through the generation of H₂S [[27]], we sought to investigate whether A. parvulum group II cysteine desulfurases (ApSufS and ApCsdB) are bona fide H₂S producers. A methylene blue assay was selected and optimized to characterize the kinetics of ApSufS and ApCsdB [[28, 29]]. This assay uses diamine to quantify the amount of gaseous H₂S. Cysteine desulfurases catalyse the initial step in thio-cofactor synthesis by extracting the sulfur group from its substrate L-cysteine. Upon formation of the persulfide bond, conventionally, a sulfur acceptor releases the sulfide and L-alanine product. To analyse the enzyme kinetics of ApSufS and ApCsdB, this assay required optimization of reducing conditions for ApSufS by usage of DTT. This optimization allows for the conversion of the intermediate persulfide bound ApSufS to free ApSufS enzyme [[30]]. Curve fitting analysis reveals that both ApSufS and ApCsdB display a sigmoid with Hill coefficients of 2.99 ± 0.26 and 4.25 ± 0.24 for ApSufS and ApCsdB, respectively (Fig. 2), indicating that both enzymes are allosterically regulated. Both enzymes display similar K_m values (Table 2), but the k_cat of ApSufS is approximately 17-fold higher than ApCsdB. The observed K_m values for ApSufS and ApCsdB are considerably larger than the intracellular cell concentration of cysteine [[31]]. This can be partly explained by the absence of the required accessory protein for cysteine desulfurases to efficiently mobilize sulfur [[24, 32]]. As such, our data reinforce the inefficiency of group II cysteine desulfurases without its accessory proteins [[20]]. This is further supported by previous kinetic analyses performed in absence of sulfur acceptors yielding K_m values within the millimolar range, comparable to our K_m values [[29, 33, 34]]. Interestingly, both enzymes are inhibited with high concentration of cysteine. Overall, the assay serves our purpose to compare the activity between ApSufS and ApCsdB and investigate the impact of mutations on ApSufS activity.

Fig. 2

Michaelis–Menten plot of H₂S production versus substrate concentration. The k_cat and K_m values were derived from a sigmoidal fit of the kinetic data using Prism. H₂S production was quantified based off methylene blue production at 670 nm and standard curve was prepared using Na₂S (0–150 µm). (A) ApSufS assays were performed using enzyme concentration of 1.125 µm and cysteine concentrations of 0-15 mm. Initial velocities were calculated using time points from 0 to 8 min. (B) ApCsdB assays were performed using enzyme concentration of 2.25 µm and cysteine concentrations of 1–15 mm. Initial velocities were calculated using 3 time points from 0–120 min.

Table 2. Kinetic parameters of wild-type and mutant ApSufS. Kinetic data were obtained by performing a methylene blue assay that measures H₂S production. Standard deviations were calculated over three replicates for each cysteine concentration and time point. K_m, k_cat and k_cat/K_m were calculated using Prism 8. No activity: n.a.

ApSufS	Km (mm)	kcat (min^−1)	kcat/Km (mm^−1·min^−1)
Wild-type	4.0 ± 0.1	6.2 ± 0.2	1.6
E377V	3.8 ± 0.1	9.2 ± 0.2	2.4
C375S	2.7 ± 0.2	0.23 ± 0.02	8.5 × 10^−2
H132A	n.a	n.a	–
K235R	n.a	n.a	–
A34Y	n.a	n.a	–

Crystal structure of ApSufS

As ApSufS is a more potent H₂S producer relative to ApCsdB, we were interested in its structural and biochemical characterization. We solved its crystal structure at a resolution of 1.80 Å by molecular replacement, using the Bacillus subtilis SufS (PDB 5J8Q) as a search model (Table 1). Only one protomer could be modelled in the asymmetric unit and contains 416 residues. Owing to the lack of electron density, residues 57–60, 88 and 421–429 were not modelled. The ApSufS protomer is composed of 18 α-helices and 12 ß-strands (Fig. 3A). Commonly seen with other group II cysteine desulfurases, ApSufS can be split into two distinct domains (Fig. 3A). The large N-terminal domain (residues 1–306) forms a tightly packed structure around the active site, consisting of several α helical inserts between parallel ß-sheets. There is a seven-stranded ß-sheet where ß strands A–F are in parallel orientation, and the final β-strand G is antiparallel (Fig. 3A). The opening of the active site pocket contains a stretch of conserved short-chain amino acids (D31-T35), most likely to minimize steric hindrance around the active site. The PLP resides in the active site where it is covalently linked to the side chain of K235, resulting in the formation of an aldimine bond (Fig. 3B). Residues 266–275 assume a characteristic feature of group II cysteine desulfurases in the form of a ß-hairpin structure (Fig. 3A). The small C-terminal domain (residues 307–429) contains a four-stranded antiparallel ß-sheet, where two of these strands flank an extended loop containing a short α-helix (α17). This extended loop harbours the conserved cysteine residue C375 (Fig. 3B). Interestingly, ApSufS is found in its persulfurated form, suggesting that the protein carried the initial step of its enzymatic reaction during overexpression. As ApSufS exists as a dimer, PyMOL was used to model the dimer structure (Fig. 3C).

Fig. 3

Crystal structure of ApSufS. (A) One protomer of ApSufS contains 18 α-helices and 12 ß-strands. The large N-terminal domain (teal) includes the ß-hairpin secondary feature and a segment of short-chain amino acids. The small C-terminal domain (light green) contains the extended loop. (B) The active site consists of residues from both the large N-terminal domain and small C-terminal domain. K235 is covalently linked to the PLP cofactor to form an internal aldimine. Residues are highlighted as stick with carbon, oxygen and nitrogen atoms rendered in grey, red and blue, respectively. (C) Surface representation of ApSufS dimer.

ApSufS shares structural similarities to other group II cysteine desulfurases

We next investigated whether ApSufS displays structural similarities with other group II cysteine desulfurases. Alignment of BsSufS with ApSufS shows that both enzymes share high structural similarities with a root-mean-square deviation of 0.696 Ų. The large domain is conserved between the two structures (Fig. 4A). Notably, the position of the residues involved in PLP binding is highly similar (Fig. 4B). Ultimately, this results in the position of PLP to remain nearly identical in both structures (Fig. 4B). Within the large domain of ApSufS, an extra α-helix at the N terminus extends towards α-helix 14 of the small C-terminal domain (Fig. 4A). BsSufS has an extended α-helix within the small domain, whereas this region is disordered in ApSufS (Fig. 4A). This C-terminal region seems to be specific to A. parvulum as it is not observed in other SufS structures. However, the overall architecture of ApSufS bears a close resemblance to other group II cysteine desulfurases as it assumes a fold similar to type I aminotransferase class V structure [[35]]. We examined the ß-hairpin structure of ApSufS and BsSufS as it is a characteristic feature of group II cysteine desulfurases. The BsSufS ß-hairpin contains bulkier residues than those in the corresponding region of ApSufS (Fig. 4C). The ß-hairpin loop has been suggested to fix the extended loop that contains the catalytic cysteine [[11]]. Specifically, the ß-hairpin loop from the other monomer has been implicated in stabilizing the extended loop that contains the catalytic cysteine residue [[19]]. With the presence of these bulkier residues in BsSufS, the resulting steric hindrance and decreased flexibility can impede the ß-hairpin interaction and stabilization of the extended loop. Thus, the increased flexibility of the ß–hairpin in ApSufS could be a factor that affects reaction kinetics in the presence of a sulfur acceptor.

Fig. 4

ApSufS structure shares similarities with the type 2 family of cysteine desulfurylases. (A) Ribbon diagram showing the alignment of ApSufS (grey) and BsSufS monomers (purple, PDB: 5J8Q). Highlighted boxes in orange show major structural differences between the two enzymes. (B) Zoomed view of the active site with residues involved in the stabilization of the PLP. (C) Residues from the ApSufS ß-hairpin structure with notable structural differences when compared to BsSufS residues are indicated.

H132 and K235 are essential residues for ApSufS desulfurylase activity

The presence of a histidine in position 132 is critical for EcSufS activity and is highly conserved among group II cysteine desulfurases [[36]]. Thus, we were interested in determining whether H132 is important ApSufS enzymatic activity. In contrast to the wild-type ApSufS, substitution of H132 by an alanine diminishes the activity below our level of detection (Table 2). The K235 residue is critical for forming the Schiff-base with the PLP in other cysteine desulfurases. Accordingly, wild-type ApSufS K235 holds the PLP by forming an aldimine bond using its ε-amine (Fig. 5). Based on the structure, we hypothesized that K235 is important for ApSufS desulfurase activity. Accordingly, the replacement of K235 with an arginine fails to show desulfurase activity compared to the wild-type (Table 2). We then sought to investigate whether the formation of the Schiff-base was critical for PLP binding. The crystal structure of ApSufS K235R shows that six protomers can be modelled in the asymmetric unit (Table 1). Surprisingly, unambiguous electronic density for the PLP is observed in the active site of ApSufS K235R (Fig. 5). Compared to the WT structure, PLP forms pi-stacking interactions with H132 and hydrogen bonds with the side chains of D209, Q212, H234, along with the main chain of T103. Unlike WT ApSufS, the side chain of R235 forms hydrogen bonds with the phosphate group of PLP, but fails to form a Schiff-base with its methanoyl group. We also identified additional electronic density close to the methanoyl group of PLP which the ε-amino group of K235 attacks to form the Schiff-base. Unfortunately, we could not accurately determine the PLP derivative at this site, which could be modelled as water or the PLP-alanine intermediate (Fig. 5). Interestingly, the side chain of R390 points towards PLP and likely stabilizes the PLP intermediate or the water molecule by a hydrogen bond. This raises the question of whether an arginine in that position can partially replace the function of a lysine and allow the formation of specific reaction intermediates. Though this substitution has not been examined in previous cysteine desulfurases, replacing the lysine residue corresponding to K235 in tryptophanase enables external aldimine formation, but significantly hinders the conformational changes required for the formation of the quinonoid intermediate [[37]]. Akin to Mycoplasma pneumonia, ApSufS could non-enzymatically hydrolyse the external aldimine to produce H₂S and PLP-alanine, which further breaks into ammonium and pyruvate. In this case, the K235R mutant would generate PLP-alanine without forming the quinonoid intermediate via a non-enzymatic hydrolysation. As we did not detect H₂S in our enzymatic assay, it is possible that such non-enzymatic reaction occurred during the crystallization process to slowly generate PLP-alanine. To confirm whether the K235R mutant can produce PLP-alanine through this non-enzymatical reaction, future enzymatic studies and a high-resolution structure will be required.

Fig. 5

The Schiff base forming residue in ApSufS is not essential for PLP binding. Panels indicating the residues surrounding the PLP binding pocket of wild-type ApSufS (left) or the mutant structure in which the PLP (centre) or the PLP-ALA (right) was refined in the structure. PLP and residues are shown as stick. Water molecules are and hydrogen bonds are indicated red spheres and yellow dash line, respectively. The electronic density surrounding of R/K235 residues is shown as mesh. The 2Fo-Fc map is contoured at 1σ.

A conserved short segment is required for substrate access to the active site in cysteine desulfurases

During our analysis of the wild-type ApSufS, we identified a conserved segment of short-chain amino acids located in the N-terminal domain of the enzyme. To determine whether this region is required to maintain access to the active site, we mutated A34 to a tyrosine residue and performed enzymatic assays. In contrast to wild-type ApSufS, A34Y is completely inactive suggesting that a bulky residue in that location restricts the access of cysteine to the active site. To address this question, we solved its crystal structure at a resolution of 2.1 Å (Table 1). Alignment of wild-type and A34Y ApSufS shows that both structures are similar (r.m.s.d. ~ 0.31 Å). Interestingly, alignment of the A34Y ApSufS with a substrate-bound SufS (PDB: 6O11) reveals that the tyrosine side chain blocks access to the cysteine binding site and likely prevents the formation of the aldimine in both protomers (Fig. 6A). Contacts with R370 stabilizes the orientation of Y34 via pi-stacking interactions, likely locking its confirmation in a position to block the active site (Fig. 6B) and explaining the loss of activity of the mutant. These findings also highlight the importance of preserving the short-chain amino acids in this region of SufS to minimize steric hindrances around the active site.

Fig. 6

A segment of short-chain residues is needed to maintain access to SufS catalytic site. (A) Alignment of wild-type ApSufS (grey) with the A34Y mutant (light blue) structure. (B) Y34 forms pi-stacking interactions with R370 hindering its desulfurase activity. (C) Surface representation of wild-type ApSufS (light grey) and A34Y active sites (light blue). The C6P (cysteine bound PLP, light green) was modelled in the active site using the substrate bound structure of EcSufS (PDB: 6O11).

The nucleophile C375 enhances ApSufS desulfurase activity

To confirm that C375 is the nucleophilic cysteine residue facilitating the attachment of the sulfhydryl group and forming the persulfide intermediate, we biochemically characterized ApSufS C375S. As shown in Table 2, the ApSufS C375S mutation results in a 1.5-fold decrease in K_m and a 20-fold loss in k_cat/K_m, an observation that is predominantly driven by a significant decrease in the product turnover. To understand the residual activity of ApSufS C375S, we solved its structure at a resolution of 2.2 Å (Table 1). As indicated with a RMSD of 0.17 Å, ApSufS C375S is highly similar to the wild-type protein (Fig. 7A). However, close inspection of the alignment reveals one major change between C375S and the wild-type SufS in the active site. While S375 occupies a similar orientation compared to C375 from the ApSufS wild-type (Fig. 7B), the H373 imidazole ring in the C375S mutant structure is facing away from the PLP (Fig. 7B). Interestingly, the PLP remains identically positioned between the ApSufS C375S mutant and wild-type (Fig. 7B).

Fig. 7

Structural analysis of C375S mutant. (A) Alignment of wild-type ApSufS (grey) with C375S mutant (pink). (B) zoomed view on residues 375 in which residues are indicated as in Fig. 4.

Akin to our findings, the structure of the C375S mutant of SufS from A. thaliana mimics the resting state of the SufS enzyme [[38]]. The residual activity of C375S can be explained by the low r.m.s.d of the residues involved in the desulfurase activity between wild-type and C375S ApSufS. Yet the replacement of the cysteine’s thiol by a weaker nucleophile prevents the ability of the enzyme to attack the L-cysteine-ketamine to form the persulfated form of ApSufS. However, the H₂S product may be a result of slow side reaction. Similar results were observed in the SufS of the Synechocystis species, where the corresponding C372A variant displayed some activity. In that model, a slow side reaction uses an unbound L-cysteine molecule to attack the L-cysteine-ketimine adduct, forming the cysteine persulfide which is further reduced to cysteine to generate H₂S in the presence of DTT [[20]]. Finally, the reorientation of H373 may grant better access to the active site and allow the attack by the unbound L-cysteine molecule, explaining the residual activity of the C375S mutant.

Concluding remarks

The ApSufS structure shares several features of other group II cysteine desulfurases and assumes a fold similar to the aminotransferase class V structure. We have identified four residues that serve distinct and important roles in ApSufS desulfurase activity. ApSufS H132 is required for ApSufS catalysis, while C375 is critical for ApSufS turnover numbers. These findings are reminiscent to a recent study showing that C375 is critical for optimal positioning of the cysteine-aldimine intermediate for subsequent Cα deprotonation, while H132 protonates the Ala-enamine intermediate in EcSufS [[36]]. The residue K235 forms an internal aldimine with the PLP, and despite its substitution to an arginine residue, the mutant enzyme can maintain the PLP in the active site in an orientation similar to the wild-type. Finally, a segment of short-chain amino acids reduces steric hindrance around the active site of ApSufS. Our observation that a tyrosine in position 34 hinders the access of cysteine to the active site pocket confirms the importance of maintaining short-chain amino acids in the active site of ApSufS. H₂S is a potential virulence factor in CD patients, where cysteine desulfurases from A. parvulum are major contributors of H₂S in the GI tract [[8]]. With a better understanding of the molecular underpinnings related to the cysteine desulfurase ApSufS, we will be able to specifically target this enzyme’s active site to develop novel therapeutics for CD patients.

Acknowledgements

This work was supported by Canadian Institutes of Health Research (CIHR) grants (PJT-148533) and internal funds through the Shanghai Institute of Materia Medica–University of Ottawa Joint Research Centre on Systems and Personalized Pharmacology. Dr Figeys is supported by Genome Canada, CIHR grants and Natural Science Engineering Research Council (NSERC) grants. Dr Stintzi is supported by the Government of Canada through Genome Canada and the Ontario Genomics Institute (OGI-149), the Ontario Ministry of Economic Development and Innovation (project number 13440), the W. Garfield Weston Foundation and Canadian Institutes of Health Research (ECD-144627). Dr Giguere is funded by an NSERC grant.

Conflict of interest

A Stintzi and D Figeys have co-founded MedBiome, a company that focuses on the development of precision microbiome therapeutics. The remaining authors declare that they have no competing interests.

Authors contributions

GK purified SufS wild-type and mutant proteins, completed the cysteine desulfurase assays and obtained kinetic data, performed data analysis of kinetic and structural data and writing. VT obtained ApSufS crystals and JFC solved its structure. VT and YY optimized the cysteine desulfurase assay. AS and AB identified the Atopobium species as an upregulated H₂S microbe in CD patients and provided guidance on the analysis of A. parvulum genome. JB obtained the initial datasets while PG offered guidance on the optimization of the high-throughput assay. DF performed mass spectrometry to optimize the protein constructs for crystallization purposes.

Open Research

Data accessibility

The atomic coordinates and structure factors of ApSufs wild-type and mutants have been deposited to the PDB with the following accession codes: wild-type–7TLM, C375S mutant–7TLQ, A34Y mutant–7TLR, K235R mutant–7TLP.