A Catalytic Trisulfide in Human Sulfide Quinone Oxidoreductase Catalyzes Coenzyme A Persulfide Synthesis and Inhibits Butyrate Oxidation

Summary

Mitochondrial sulfide quinone oxidoreductase (SQR) catalyzes the oxidation of H₂S to glutathione persulfide with concomitant reduction of CoQ₁₀. We report herein that the promiscuous activity of human SQR supported the conversion of CoA to CoA-SSH (CoA-persulfide), a potent inhibitor of butyryl-CoA dehydrogenase, and revealed a molecular link between sulfide and butyrate metabolism, which are known to interact. Three different CoQ₁-bound crystal structures furnished insights into how diverse substrates access human SQR, and provided snapshots of the reaction coordinate. Unexpectedly, the active site cysteines in SQR are configured in a bridging trisulfide at the start and end of the catalytic cycle, and the presence of sulfane sulfur was confirmed biochemically. Importantly, our study leads to a mechanistic proposal for human SQR in which sulfide addition to the trisulfide cofactor eliminates ²⁰¹Cys-SSH, forming an intense charge-transfer complex with FAD, and ³⁷⁹Cys-SSH, which transfers sulfur to an external acceptor.

Graphical Abstract

eTOC Blurb

The catalytic promiscuity of human sulfide quinone oxidoreductase supports formation of CoA-persulfide, a known inhibitor of short-chain fatty acid oxidation in colonocytes. Landry et al describe four crystal structures of human SQR unraveling a different mechanism for sulfide oxidation, which is catalyzed by an active site cysteine trisulfide.

Introduction

Since its discovery as a gaseous signaling molecule (Abe and Kimura, 1996), an array of physiological effects have been associated with hydrogen sulfide (H₂S) in the cardiovascular, nervous and gastrointestinal systems (Filipovic et al., 2018; Kabil and Banerjee, 2010; Kimura, 2010). Cells maintain low steady-state levels of H₂S, a well-known respiratory toxin (Nicholls and Kim, 1982), and efficiently oxidize it via the mitochondrial sulfide oxidation pathway to generate thiosulfate and sulfate (Hildebrandt and Grieshaber, 2008; Libiad et al., 2014). Sulfide quinone oxidoreductase (SQR), a member of the flavin disulfide reductase superfamily (Argyrou and Blanchard, 2004), is anchored in the inner mitochondrial membrane and catalyzes the first and committing step in the sulfide oxidation pathway (Figure 1A). While 10–80 nM intracellular H₂S is typical in various tissues (Furne et al., 2008; Vitvitsky et al., 2012), ~0.2–2.4 mM H₂S has been reported in the colonic lumen (Deplancke et al., 2003; Macfarlane et al., 1992), which hosts ~100 trillion microbes (Turnbaugh and Gordon, 2009), including those that produce H₂S via sulfate reduction, or desulfuration of cysteine (Linden, 2014). Hence, in addition to producing the gas endogenously, colonic epithelial cells are routinely exposed to microbiota-derived H₂S. The hydrophobicity of H₂S promotes its unfacilitated membrane transport (Mathai et al., 2009), potentially increasing exposure of colonocytes to the gas. Accordingly, colonic epithelial cells are adapted to survive acute exposure to high H₂S levels (Levitt et al., 1999).

Figure 1. Intersection between sulfide and butyrate oxidation pathways and the proposed mechanisms for the SQR reaction.

(A) The canonical SQR reaction utilizes GSH as an acceptor to generate GSSH, with electrons driven into the CoQ₁₀ (Q) pool. ACADS via the electron transferring flavoprotein (ETF), also drives electrons into the Q pool via oxidation of butyrate. We hypothesized that SQR utilizes CoA as an alternate acceptor generating CoA-SSH, which inhibits ACADS and blocks electron flow from butyrate oxidation. PDO (or ETHE1), TST and SO represent persulfide dioxygenase, rhodanese and sulfite oxidase, respectively in the mitochondrial sulfide oxidation pathway. (B) The previously proposed SQR reaction mechanism in which the active site cysteines in the resting enzyme form a disulfide. Cys201 participates in a CT complex with FAD while Cys379 is the sulfane sulfur carrier. (C) A proposal for the SQR-catalyzed oxidative half reaction, informed by our crystal structures, invokes a trisulfide configuration of the active site cysteines. Sulfide addition leads to persulfides on both cysteines. ²⁰¹Cys-SS⁻ forms a CT complex with FAD while ³⁷⁹Cys-SSH serves as the sulfur donor.

Butyrate is the preferred fuel source for normal colonocytes and is generated via microbial fermentation of undigested fiber remnants that reach the large intestine (Roediger, 1980). Butyryl-CoA is metabolized via butyryl-CoA dehydrogenase (or ACADS) to acetyl-CoA, which enters the Krebs cycle or is converted to ketone bodies. An unusual feature of ACADS is that it is purified as a “green” protein due to the presence of an inhibitory charge-transfer (CT) complex between a tightly bound CoA persulfide (CoA-SSH) and the FAD cofactor (Williamson et al., 1982). The remarkable stability of this complex has raised the obvious but as yet unanswered question as to the biological source of CoA-SSH. Sulfide inhibition of ACADS-dependent butyrate oxidation in colonocytes is associated with ulcerative colitis (Babidge et al., 1998). Given the promiscuity of SQR with respect to its sulfane sulfur acceptor (Jackson et al., 2012; Libiad et al., 2014) and its key role in detoxifying colonic H₂S (Roediger et al., 1993), we hypothesized that SQR is a source of CoA-SSH whose formation allows prioritization of sulfide over butyrate oxidation.

SQR catalyzes two half reactions: (i) sulfur transfer from H₂S to an acceptor via an active site cysteine persulfide (Cys-SSH) intermediate, and (ii) electron transfer from H₂S to coenzyme Q₁₀ (CoQ₁₀) via an FADH₂ intermediate (Figure 1B) (Jackson et al., 2012; Landry et al., 2017; Mishanina et al., 2015). The first step in the proposed mechanism is addition of the sulfide anion to an active site disulfide between Cys201 and Cys379, generating a persulfide intermediate on Cys379 (³⁷⁹Cys-SSH) with concomitant release of the Cys201 thiolate (Figure 1B) (Jackson et al., 2012). SQR forms an intensely absorbing CT complex centered at ~675 nm (Jackson et al., 2012; Libiad et al., 2014), suggesting formation of an electronic species that is distinct from those seen in other members of the flavin disulfide reductase superfamily (Mishanina et al., 2015). While the identity of the physiological acceptor has been controversial (Jackson et al., 2012; Libiad et al., 2014), kinetic simulations predict that GSH is the physiological sulfur acceptor for SQR (Landry et al., 2017, 2018; Libiad et al., 2014; Mishanina et al., 2015). The resulting glutathione persulfide (GSSH) product is then used by persulfide dioxygenase (PDO or ETHE1), the next enzyme in the pathway (Kabil and Banerjee, 2012). In the final step, electron transfer to CoQ₁₀ regenerates FAD and connects SQR to the electron transfer chain at the level of complex III, making H₂S the only known inorganic substrate for ATP synthesis in humans (Goubern et al., 2007).

Prior to a recent report of the crystal structure of human SQR (Jackson et al., 2019), structural perspective on the reaction was provided by bacterial homologs (Brito et al., 2009; Cherney et al., 2010; Marcia et al., 2009) which are, however, mechanistically distinct. In bacterial SQRs, sulfane sulfur does not exit the active site at the end of each catalytic cycle; instead, a hydropolysulfide bridge extends between the active site cysteines and is released as a linear polysulfide chain or after cyclization to elemental octasulfur.

Herein, the known substrate promiscuity of SQR is expanded to include CoA as an alternate sulfur acceptor, forming CoA-SSH, which is an ACADS inhibitor. We also report four structures of human SQR in the absence and presence of CoQ₁, a soluble CoQ₁₀ analog, and with/without sulfite or sulfide soaking, at resolutions ranging from 2.0–2.8 Å. Notably, soaking SQR-CoQ₁ crystals with sodium sulfide provided evidence for in crystallo catalysis and captured CoQ₁ exiting a long hydrophobic channel leading from FAD to the putative membrane-facing surface of SQR. Importantly, the structures together with biochemical data, led us to postulate a different mechanism for human SQR by assigning the ²⁰¹Cys-SS⁻ (versus the Cys201 thiolate) as the species involved in CT complex formation with FAD, and ³⁷⁹Cys-SSH as the sulfane sulfur donor to an external acceptor (Figure 1C).

Results

CoA is an alternative sulfur acceptor for SQR

To test our hypothesis that SQR is a potential source of CoA-SSH, we characterized the kinetics of sulfur transfer from sulfide to CoA. Since the absorption spectrum of CoA interfered with our ability to monitor CoQ₁ reduction at 278 nm in the steady-state SQR assay, we developed an alternative coupled assay using PDO, which oxidizes CoA-SSH in an O₂-dependent reaction (Kabil and Banerjee, 2012). Michaelis-Menten analysis of SQR activity at varying CoA concentrations yielded a specific activity of 88 ± 4 μmol O₂ min⁻¹ mg⁻¹, K_M(CoA) of 2.3 ± 0.4 mM, and k_cat of 69 ± 3 s⁻¹ at 25 °C (Figure 2A). The k_cat/K_M for CoA (3 × 10⁴ M⁻¹s⁻¹) was ~2-fold higher than for GSH (1.6 × 10⁴ M⁻¹s⁻¹) (Landry et al., 2017).

Figure 2. Pre-steady state kinetic analysis of ndSQR-mediated sulfur transfer to CoA.

(A) Dependence of ndSQR activity on CoA concentration assessed in the coupled assay described under Star Methods. The data represent the mean ± S.D. of three independent experiments, each performed in duplicate. (B) ndSQR (40 μM) in 100 mM potassium phosphate buffer, pH 7.4, was rapidly mixed 1:1 (v/v) with Na₂S (80 M) for ~35 msec to form the sulfide-induced CT complex at 675 nm (dashed black line), followed by a second rapid 1:1 (v/v) mixing with CoA (3 mM). Spectral changes associated with CT complex decay in ndSQR and concomitant FAD reduction (red line) were monitored over a period of 3 sec. (C) Comparison of the decay kinetics of the sulfide-induced CT complex in ndSQR with or without mixing with CoA (3 mM) monitored at 675 nm over 3 sec. (D) Dependence of the k_obs for sulfide-induced CT complex decay on CoA concentration. The data represent the mean ± S.D. of two independent experiments.

The sulfide-induced CT intermediate on SQR (λ_max = 675 nm) is stable for several seconds, but decayed upon rapid mixing with CoA, and led to the concomitant reduction of FAD (Figure 2B,C). A k_on of 1.05 × 10³ M⁻¹s⁻¹ at 4 °C was obtained from the dependence of the k_obs for CT complex decay on the concentration of CoA (Figure 2D). CoA, like other thiols (Landry et al., 2018), was also capable of nucleophilic addition to the resting enzyme, forming a CT complex (Figure S1A). Compared to solubilized SQR, the intensity of the CT complex with nanodisc-embedded SQR (ndSQR), formed under the same conditions, was ~80% lower (Figure S1B), consistent with a role for the membrane environment in limiting this unwanted side reaction, as seen previously with GSH (Landry et al., 2018).

SQR-derived CoA-SSH forms a CT complex with ACADS

Incubation of oxidized ACADS with chemically prepared CoA-SSH led to the appearance of a CT band centered at 710 nm and a shift in the FAD absorption peak from 445 to 425 nm (Figure 3A), which were consistent with the formation of a CoA-SSH to FAD CT complex (Williamson et al., 1982). Next, the ability of SQR-derived CoA-SSH to form a CT complex on human ACADS was examined. Addition of a catalytic amount of ndSQR to the reaction mixture containing sulfide, CoA and CoQ₁ reproduced the spectral changes in ACADS, consistent with the formation of a CoA-SSH to FAD CT complex (Figure 3B).

Figure 3. CT complex formation in ACADS by ndSQR-derived CoA-SSH.

(A) ACADS (10 μM) in 100 mM potassium phosphate buffer, pH 7.4 at 25 °C (black line) was mixed with CoA-SSH (20 μM) and incubated for 1 min. Formation of the CoA-SSH-induced CT complex in ACADS was indicated by the 710 nm absorbance band (red line). (B) ACADS (10 μM) in the same buffer as in (A), was mixed with Na₂S (150 μM), CoA (3 mM), and CoQ₁ (60 μM) and incubated for 1 min (black line), followed by addition of ndSQR (100 nM) and incubation for 1 min. Formation of the CoA-SSH-induced CT complex in ACADS (red line) was observed. The data are representative of three independent experiments.

Sulfide inhibits butyrate oxidation in colon epithelial cells

The effect of sulfide on butyrate oxidation was monitored by the release of ¹⁴CO₂ from ¹⁴C-butyrate in the malignant human colon epithelial cell line HT-29. ¹⁴CO₂ release was inhibited 20 or 37% upon bolus exposure to 0.3 or 0.5 mM Na₂S, respectively (Figure S1C). While our data and those reported previously with human colonocytes exposed to 1.5 mM H₂S (Babidge et al., 1998) could be consistent with sulfide inhibition of butyrate metabolism via CoA-SSH, it is important to note that both the sulfide and butyrate oxidation pathways are dependent on Complex IV, which is also inhibited by sulfide (Nicholls and Kim, 1982).

Crystal structures of human SQR

Table 1 summarizes the crystallographic data collection and refinement statistics for the following four structures: SQR (SQR, 2.80 Å), SQR co-crystallized with CoQ₁ (SQR-CoQ₁, 2.03 Å), SQR-CoQ₁ soaked with sulfite (SQR-CoQ₁ + sulfite, 2.21 Å), and SQR-CoQ₁ soaked with sulfide (SQR-CoQ₁ + sulfide, 2.55 Å). The initial phases were obtained from selenomethionine-substituted SQR at 3.27 Å resolution using the single-wavelength anomalous dispersion (SAD) method. The SQR model built from the SAD map was then employed for molecular replacement to calculate electron density maps for other high-resolution datasets. Two copies of monomeric SQR were present in the asymmetric unit forming a crystallographic dimer (Figure 4A). The overall structure is similar to the Acidithiobacillus ferrooxidans SQR (Cherney et al., 2010) and the flavoprotein subunit of the periplasmic Allochromatium vinosum flavocytochrome sulfide dehydrogenase (Chen et al., 1994) which uses cytochrome c instead of quinone as an electron acceptor (Figure S2). Each monomer of human SQR comprises the tandem Rossmann fold repeat signature of the flavin disulfide reductase superfamily, along with two amphipathic C terminal helices that anchor SQR to the membrane (Figure 4A), as also seen in A. ferrooxidans SQR (Cherney et al., 2010), and described in detail for human SQR (Jackson et al., 2019). The four SQR structures overlay well, with Cα rmsd ranging from 0.19–0.32 Å (Figure S3A), revealing the absence of major conformational changes. Our SQR structure overlays with a Cα rmsd of 0.29 Å on the recently described SQR structure (PDB ID: 6MP5) (Jackson et al., 2019). The membrane anchoring helices are at opposite ends in the crystallographic dimer, indicating that the observed dimer of detergent-solubilized SQR is not biologically relevant.

Table 1.

Crystallographic Data collection and refinement statistics

	SeMet-apo-SQR	SQR	SQR-CoQ1	SQR-CoQ1 + sulfite	SQR-CoQ1 + sulfide
Space group	P212121	P212121	P212121	P212121	P212121
Unit cell parameters (Å)	a=77.58	a=76.729	a=78.516	a=76.11	a=76.295
	b=111.42	b=110.424	b= 111.512	b=108.75	b=111.768
	c=133.92	c=132.124	c=133.648	c=131.06	c=137.032
	α=β=γ=90°	α=β=γ=90°	α=β=γ=90°	α=β=γ=90°	α=β=γ=90°
Wavelength (Å)	0.9786	1.12713	1.12713	1.12713	1.12713
Data collection statistics
Resolution range (Å)	66.92–3.27 (3.33–3.27)*	50.00–2.80 (2.85–2.80)	51.46–2.03 (2.08–2.03)	43.69–2.11 (2.16–2.11)	50.00–2.55 (2.60–2.55)
Number of unique reflections	18411 (831)	26934 (1198)	73848 (4449)	61813 (4480)	38478 (1642)
Completeness (%)	99.72 (92.03)	96.5 (89.1)	96.9 (78.9)	98.0 (97.3)	99.2 (85.7)
Rmerge	0.508 (0.864)	0.134 (0.747)	0.184 (1.664)	0.140 (1.300)	0.202 (1.139)
Rpim	0.127 (0.306)	0.077 (0.485)	0.079 (0.798)	0.065 (0.588)	0.095 (0.522)
CC1/2	0.315 (0.537)	1.002 (0.568)	0.875 (0.302)	0.982 (0.637)	0.984 (0.673)
CC1/2(anom)	0.091 (0.323)
Redundancy	24.31 (9.87)	3.6 (3.1)	6.6 (5.8)	5.6 (5.8)	5.3 (5.4)
Mean I/σ	9.09 (2.24)	11.7 (1.6)	6.0 (1.3)	6.4 (1.1)	4.5 (1.3)
Refinement statistics
Resolution range (Å)		33.18–2.81	41.38–2.03	41.92–2.21	50.93–2.56
Rwork/Rfree (%)		21.74/27.32	22.01/25.53	19.02/22.60	22.41/27.19
RMSD bonds (Å)		0.005	0.007	0.012	0.005
RMSD angles (deg)		0.882	0.951	1.243	0.825
Average B factor (Å2)		48.8	24.8	30.6	34.1
Number of water molecules		30	191	101	41
Ramachandran		95	94	97	97
favored (%)
allowed (%)		5	6	3	3
not allowed (%)		0	0	0	0
PDB ID		6O15	6O1B	6O1C	6O16

^*Values in parentheses are for highest-resolution shell

Figure 4. Overall structure and active site architecture of SQR.

(A) Overall structure of SQR-CoQ₁ in which the monomers in the crystallographic dimer are shown in pale green and pink, respectively. The C terminal membrane anchoring helices are highlighted in cyan and salmon pink. FAD and CoQ₁ are shown in stick display. The electron density maps (2F_o-F_c) contoured at 1.0 σ of Cys379, Cys201, FAD and CoQ₁ are shown as mesh in SQR-CoQ₁ (B), SQR-CoQ₁ + sulfite (C) and SQR-CoQ₁ + sulfide (D). Note that the cysteine backbone carbons are shown in green in B-D and that CoQ₁ has moved away from FAD in the sulfide-soaked crystal structure in (D). (E-G) Close up of the active site redox centers in the indicated structures in which the distances between sulfur atoms and between sulfur atoms and the flavin C4a are labeled. The diffraction data precision indicator values are 0.37 Å (E), 0.16 Å (F), and 0.28 Å (G), respectively.

Configuration of redox active cofactors

The cache of redox cofactors in SQR comprises a noncovalent FAD and a pair of cysteine residues, Cys201 and Cys379 (Figure 4B–D, Figure S3B). FAD is bound in an extended conformation via multiple electrostatic and hydrogen-bonding interactions with main chain and side chain groups (Figure S3C). The FAD walls off the active site into a re-face and matrix-facing sulfur substrate channel, and a si-face and membrane-facing CoQ channel. An additional solvent-exposed channel connects FAD to the protein surface. Sulfur atoms in the active site were assigned using composite or polder omit maps calculated in the absence of FAD, CoQ₁ and sulfur atoms, to limit biases introduced during model building. In the apo-SQR structure, the extra electron density between the cysteine thiols is assigned as a sulfur atom connecting Cys201 and Cys379 in a trisulfide bridge (Figure S3B). This configuration, also seen in the recent human SQR structure, was designated as the catalytically inactive “thiocystine” state of the enzyme (Jackson et al., 2019). The S-S distances in the trisulfide are 2.1 Å and 2.0 Å between the bridging sulfur and the sulfurs on Cys201 and Cys379, respectively (Figure 4E). The additional sulfur in human SQR as isolated was confirmed biochemically by cold cyanolysis, which detected 1.0 ± 0.2 mol sulfane sulfur per mol monomer.

In the SQR-CoQ₁ structure, electron density assigned to CoQ₁ is present in a hydrophobic pocket that opens to the putative membrane side (Figure 4B, Figure S4A,B), identified by the location of the membrane anchoring C-terminal helices (Jackson et al., 2019). Assumptions about equivalent thermal parameters or occupancy were not made, although the higher B-factors for CoQ₁ likely reflect incomplete ligand occupancy and/or greater mobility. Unexpectedly, the sulfur bridge between Cys201 and Cys379 is broken in this structure. The strong additional density near Cys201 is assigned as the major ²⁰¹Cys-SSH species; the weak electron density at Cys379 is assigned as the minor ³⁷⁹Cys-SSH species (Figure 4B, Figure S4C). The outer sulfur in ²⁰¹Cys-S-S⁻ is 3.7 Å away from ³⁷⁹Cys-SH, i.e. too far for an S-S bond (Figure 4F). Instead, it is rotated toward the FAD and is at a distance of 3.4 Å from C4a, forming the basis of our assignment that the major species in the SQR-CoQ₁ structure is the ²⁰¹Cys-S-S⁻ persulfide forming a CT complex with FAD. In the minor ³⁷⁹Cys-S-S⁻ species, the distance between the outer sulfur and ²⁰¹Cys-SH is 4.2 Å, also too far for a bridging ²⁰¹Cys-S-S-S-³⁷⁹Cys trisulfide.

In principle, soaking sulfite into SQR-CoQ₁ crystals, which have a sulfane sulfur at either Cys201 or Cys379, could have led to a single sulfur transfer reaction to form thiosulfate (Mishanina et al., 2015). The active site cysteine configuration in the SQR-CoQ₁ + sulfite structure is similar to that in the SQR-CoQ₁ structure (Figure 4B,C), except that the additional electron density in the vicinity of Cys379 is almost completely absent in the polder omit map (Figure S4C). Density assigned to CoQ₁ is found in the active site in chain B, and in the active site or at the putative membrane-entry site in chain A (Figure 4C).

Soaking SQR-CoQ₁ crystals with sulfide could, in principle, support in crystallo catalysis (Jackson et al., 2012). Unexpectedly, sulfide addition to the SQR-CoQ₁ crystals restored the resting trisulfide configuration (Figure 4D,G). The S-S distances in the trisulfide are 2.1 Å and 2.0 Å between bridging sulfur and the sulfurs on Cys201 and Cys379, respectively similar to those in the apo-SQR structure (Figure 4E). The distance between the flavin C4a and the Sγ of Cys201 versus C4a and the bridging sulfur is 3.5 and 4.3 Å, respectively. Furthermore, the CoQ binding site proximal to FAD is vacant (Figure 4D). In solution, when SQR was allowed to undergo 1.7 turnovers in the presence of sulfide and CoQ₁, the sulfane sulfur atom was retained on the enzyme and detected by cold cyanolysis (1.5 ± 0.2 mol sulfane sulfur per mol monomer), consistent with the cysteine trisulfide being regenerated following turnover. We note that the slight increase in the sulfane sulfur detected at the end, compared to the beginning of turnover, might have resulted from spurious persulfidation of a surface-exposed Cys127 by HSSH, the reaction product.

Structural perspective on GSH and CoA binding

The structures of detergent-solubilized SQR reported herein, correspond to the promiscuous form of the enzyme, which allows diverse nucleophiles to add into the oxidized cysteine cofactor and/or serve as sulfur acceptors. On the other hand, embedding SQR in nanodiscs restricts some of these adventitious reactions (Landry et al., 2018). The matrix-facing side of solubilized SQR provides some structural perspective on its catalytic promiscuity, as discussed for the apo-SQR structure (Jackson et al., 2019). A large channel leads from the surface to the relatively exposed ³⁷⁹Cys-SH or ³⁷⁹Cys-SSH in the active site (Figure S5). The electropositive potential energy surface could be important for guiding deprotonated substrates (sulfide, GSH, or CoA) during the sulfide oxidation and sulfur transfer reactions (Figure 5A). The top docking poses for GSH and CoA bound to SQR-CoQ₁ reveal that they are lodged in the concavity at the surface of the channel (Figure 5B,C), with GSH being in a similar orientation as reported (Jackson et al., 2019). In these modeled structures, the thiols of GSH and CoA reach into the active site with their sulfur atom pointing toward, but distant from the Cys379 thiol (~6.3 Å for CoA and ~8.2 Å for GSH). These thiols would be closer to the outer sulfur in the putative ³⁷⁹Cys-SSH intermediate generated during the catalytic cycle.

Figure 5. The sulfur substrate entry site is on the matrix side.

(A) Electrostatic potential energy surface map showing the sulfur substrate entry/exit site in SQR-CoQ (dashed box and arrow, left panel). In the right panel, the structure is shown in the same orientation as in the left panel, with the FAD and CoQ₁ displayed as yellow and dark gray spheres, respectively. (B) Docking model of SQR-CoQ₁ with CoA (left) and close-up view of the active site (right). (C) Docking model of SQR-CoQ₁ with GSH (left) and close-up view on the active site (right).

CoQ₁ movement through a hydrophobic tunnel

The largely electropositive potential energy surface on the putative membrane-facing side of SQR frames a central hydrophobic patch that narrows into a deep tunnel (Figure 6A). CoQ₁, with its C1 rather than C10 isoprenoid tail found in the natural CoQ₁₀ substrate, resides at the end of the tunnel. CoQ₁ is held at the edge of the isoalloxazine ring via interactions with active site residues on the si face of FAD (Figure 6B), similar to the interactions noted with a docked decylubiquinone moiety (Jackson et al., 2019). The C10-isoprenoid tail of the physiological substrate for SQR presumably occupies the length of the tunnel in the membrane environment. In the SQR-CoQ₁ + sulfite structure, density assigned as CoQ₁ is observed both at the entrance and the end of the hydrophobic tunnel, while it is only seen at the entrance in the SQR-CoQ₁ + sulfide structure (Figure 4C–D, 6C).

Figure 6. CoQ entry from the mitochondrial membrane side.

(A) Electrostatic potential energy surface map showing the CoQ entry/exit site denoted by the dashed line in the SQR-CoQ₁ structure (left panel). In the right panel, the structure is shown in the same orientation as the left panel, with FAD and CoQ₁ displayed as yellow and grey spheres, respectively. (B) A close-up showing the interactions between CoQ₁ and active site residues. (C) Location of CoQ₁ near FAD at the end of the hydrophobic tunnel (in mesh representation) as seen in the SQR-CoQ₁ structure (green) versus at the entrance to the tunnel (cyan) in the SQR-CoQ₁ + sulfide structure.

The apparent movement of CoQ₁ away from the active site and to the tunnel entrance in the sulfide-treated SQR-CoQ₁ crystals suggests that in crystallo catalysis occurred, as discussed below. Additional density assigned as CoQ₁ is seen at a third location in chain A in the SQR-CoQ₁ and SQR-CoQ₁ + sulfite structures (not shown). This site is likely to be nonspecific since it is on the surface of the matrix-facing side of SQR, and would not be contact with the CoQ₁₀ pool in the mitochondrial membrane.

Discussion

SQR exists at the membrane-water interface, and its development as a potential therapeutic target for regulating H₂S has been impeded by the lack of structural perspective on its catalytic mechanism, and complicated by its substantial substrate promiscuity. We report for the first time that CoA can serve as a sulfur acceptor for human SQR, generating CoA-SSH. The specificity constant (k_cat/K_M) for sulfide oxidation is ~2-fold higher with CoA than with GSH. Intracellular CoA levels are estimated to be ~15–175 μM in various tissues (Shurubor et al., 2017), while GSH is present at 1–10 mM (Kosower and Kosower, 1978), predicting that CoA-SSH production will be minor relative to GSSH. However, it is not known whether colonocytes, which preferentially rely on β-oxidation for their energy metabolism, have higher CoA concentrations. Our study establishes the feasibility of SQR-derived CoA-SSH to bind to and inhibit ACADS, thus identifying a biological source of CoA-SSH associated with “green” ACADS preparations that were first reported 65 years ago (Green et al., 1954; Shaw and Engel, 1984; Steyn-Parve and Beinert, 1958).

Both butyrate and sulfide oxidation rely on the electron transfer chain for provision of electron acceptors (Figure 1A). A possible rationale for sulfide modulation of ACADS in colonocytes via SQR-derived CoA-SSH is prioritization of sulfide oxidation under conditions of acute H₂S exposure from gut microbial metabolism. Under these conditions, the high sulfide oxidation activity of SQR would rapidly decrease ambient H₂S levels if competition for the oxidized CoQ pool is reduced by inhibition of ACADS (Figure 1A). We posit that CoA-SSH-induced ACADS inhibition helps shield the electron transfer chain from the effects of H₂S poisoning, and allow cells to return to butyrate-oxidation fueled metabolism when H₂S levels decrease. Impaired butyrate utilization is correlated with ulcerative colitis (Roediger, 1980a), and sulfide exposure reportedly leads to increased butyryl-CoA and decreased crotonyl-CoA in human colonocytes (Babidge et al., 1998). While our observation that HT-29 cells exhibit sulfide-sensitive impairment of butyrate oxidation (Figure S1C) is consistent with the postulated inhibition of ACADS by CoA-SSH (Shaw and Engel, 1987), it is important to note that Complex IV poisoning by H₂S also contributes to inhibition of β-oxidation. Exploring the role of SQR in this process in vivo is an intriguing next step that merits investigation.

While our manuscript was in preparation, the structure of human apo-SQR at 2.59 Å resolution was reported (Jackson et al., 2019). The observed “thiocystine,” which is analogous to the active site trisulfide described in this study, was interpreted as the noncatalytic state of SQR and decylubiquinone was modeled into the CoQ₁₀ binding pocket. The structure was used as a framework to propose a chemically unusual mechanism for converting the “noncatalytic” trisulfide into the active disulfide form (Reaction 1). Jackson et al postulated that sulfide (or an alternative nucleophile) attacks the trisulfide, generating ²⁰¹Cys-SS⁻ and ³⁷⁹Cys-SSH; the outer sulfur atom of ²⁰¹Cys-SS⁻ then attacks Cβ of Cys379 eliminating hydrodisulfide and forming the ³⁷⁹Cys-S-S-Cys²⁰¹ disulfide (Jackson et al., 2019). Notably, the sulfide sulfur atom replaces the original Cys379 Sγ in the proposed mechanism to build the active site disulfide.

$${}^{379}\text{Cys}-\mathbf{\text{S}}-\mathbf{\text{S}}-\text{S}-{\text{Cys}}^{201}+{\text{HS}}^{-}{\to }^{379}\text{Cys}-SS\text{H}{+}^{-}\text{SS}-{\text{Cys}}^{201}{\to }^{379}\text{Cys}-\text{S}-\text{S}-{\text{Cys}}^{201}+\text{H}S{S}^{-}$$ [Rxn 1]

We propose an alternative mechanism (Figure 1C), which is supported by analysis of the sulfane sulfur associated with SQR at the start and end of catalytic turnover. Furthermore, the movement of CoQ₁ out of the active site and reformation of the resting trisulfide in the SQR + CoQ₁ + sulfide structure, provide evidence for in crystallo sulfide-dependent turnover and support the postulated mechanism.

We propose that SQR, as isolated with a cysteine trisulfide cofactor (Figure S6, [1]), represents the active form of the enzyme. Surprisingly, co-crystallization with CoQ₁ leads to loss of continuous electron density between the cysteines and to S-S bond distances that indicates the absence of a trisulfide. Instead, the extra electron density was modeled as a mixture of a major ²⁰¹Cys-SS⁻ [2] and a minor ³⁷⁹Cys-SS⁻ species [3]. The ruptured trisulfide suggests that the S-S bond was either reduced, or that nucleophilic addition led to elimination of ²⁰¹Cys-SS⁻ (major) or ³⁷⁹Cys-SS⁻ (minor), followed by hydrolysis of the resulting Cys-S-Nuc species. The source of the nucleophile, if any, is unknown, and could be an impurity introduced with CoQ₁ (2.5 mM) or the carrier, dimethylsulfoxide (2.5% v/v) used during crystallization. We tentatively attribute the ²⁰¹Cys-SS⁻ species as the product of a nucleophilic addition/hydrolysis, as a nucleophile would more readily access the exposed Cys379 to cleave the trisulfide. Alternatively, intermittent light exposure during crystallization conditions, which included carboxylates that can participate as electron donors, could have led to CoQ₁ photoreduction (Gorner, 2004) and electron transfer to FAD. This could account for the formation of the ³⁷⁹Cys-SS⁻ species by reduction of the proximal ²⁰¹Cys-S-SR bond in the trisulfide leading to [3]. Notably, electron density assigned as CoQ₁ proximal to FAD was observed in the SQR-CoQ₁ crystals, as also seen in the homologous A. ferrooxidans SQR structure (Zhang et al., 2016).

Sulfite can act as a sulfur acceptor, forming thiosulfate. The structure of SQR-CoQ₁ + sulfite revealed loss of the extra electron density on Cys379 seen in the SQR-CoQ₁ structure, but retention of a strong electron density peak in the polder omit map on the Sγ of Cys201 (Figure S4C). Implications of the observed sulfite-induced changes are that: (i) the CT complex with FAD involves ²⁰¹Cys-SS⁻ rather than the postulated ²⁰¹Cys-S⁻, and (ii) sulfur transfer to the acceptor occurs from ³⁷⁹Cys-SSH. The distance between either sulfur in ²⁰¹Cys-SS⁻ and C4a on flavin is 3.5 Å in the SQR-CoQ₁ + sulfite structure. Electropositive stabilization of the persulfide-CT intermediate could be provided by the helix dipole of α-helix 11 with its N-terminus oriented towards the FAD ring. The α-amino group of Lys207, oriented towards the outer sulfur in ²⁰¹Cys-SS⁻ at a distance of ~3.65 Å, could also help stabilize the persulfide intermediate. Strikingly, the CT complex in “green” human ACADS also arises from the interaction between a persulfide and FAD, in which the outer sulfur of CoA-SS⁻ is 3.8 Å away from the flavin C4a (Figure S3D).

In the presence of sulfite, we presume that sulfane sulfur transfer occurred from the accessible ³⁷⁹Cys-SSH position leading to the presence of [2] in the crystal structure (Figure S6). Sulfide can act as both a sulfur donor and an acceptor in the SQR reaction (k_cat =74 s⁻¹ with sulfide, versus 113 s⁻¹ with sulfide + GSH) (Libiad et al., 2014). Addition of sulfide to the SQR-CoQ₁ crystals resulted in regeneration of the resting trisulfide and movement of CoQ₁ out of the active site.

Based on our model, neither species observed in the SQR-CoQ₁ crystals is catalytically active. The trisulfide could be rebuilt from [2] and [3] in crystallo in the presence of HSSH, a common contaminant in sulfide (Greiner et al., 2013), during sulfide soaking of SQR-CoQ₁ crystals. Once reconstituted, the active enzyme would support sulfide oxidation and reduction of CoQ₁, which was captured exiting the active site in the SQR-CoQ₁ + sulfide structure (Figure 4D). Sulfane sulfur, detected by cold cyanolysis, remained associated with SQR following catalytic turnover in solution, consistent with the proposed regeneration of the resting trisulfide seen in crystallo. A CT complex and a C4a adduct between a persulfide rather than a thiolate on the proximal cysteine has been proposed for A. ferrooxidans SQR (Cherney et al., 2010). Similarly, a catalytic mechanism featuring a CT complex between a cysteine persulfide and FAD and a trisulfide resting form, has been postulated for the Aquifex aeolicus SQR (Marcia et al., 2009), and now visualized in our structures. The mechanism by which a redox active trisulfide cofactor is initially built in human SQR is currently under investigation in our laboratory.

As is clear from the above discussion, the redox activity of sulfur compounds introduces a layer of complexity in working with them, but fortuitously, led to the in crystallo formation and trapping of reactive species in human SQR. Structures of bacterial SQRs with linear and branched polysulfides extending between the analogous active site cysteine residues have been reported (Cherney et al., 2010; Marcia et al., 2009). The distance between the cysteine thiols in bacterial SQRs is significantly greater (~8.8 Å in A. ferrooxidans and 8.3 Å in A. aeolicus SQR) than in human SQR (~4.6 Å), consistent with the difference in length of the bridging sulfane sulfur intermediates that are formed in their respective catalytic cycles.

The large cavity extending from the matrix face to the active site exposes Cys379 and explains the vulnerability of solubilized SQR to errant attack by nucleophiles (sulfite, GSH, homocysteine, cysteine and methanethiol). Attack on the Sγ of the exposed Cys379 in the resting trisulfide would release ²⁰¹Cys-SS⁻ and form the CT complex (Figure 1C). However, these adventitious reactions lead to a dead end, and the active enzyme was only slowly recovered by reversal of the reaction and release of the nucleophile (Landry et al., 2018). Promiscuity is equally a problem in the subsequent sulfur transfer step, i.e. from ³⁷⁹Cys-SSH to an acceptor. Small molecule acceptors like sulfite and methanethiol exhibit higher catalytic efficiencies than GSH or CoA, which could reflect their more facile access to the active site (Jackson et al., 2012; Landry et al., 2018; Libiad et al., 2014). Docking models reveal that either GSH or CoA can be accommodated at the mouth of the cavity with their thiol moieties oriented to interact with the ³⁷⁹Cys-SSH intermediate (Figure 5B,C). The significantly greater intracellular concentration of GSH than sulfite or methanethiol, except perhaps under pathological conditions (Shih et al., 1977; Yaegaki and Sanada, 1992), is predicted to favor GSSH formation. In contrast to human SQR, the A. ferrooxidans SQR has a sheltered active site in which sulfide is polymerized rather than being transferred to an exogenous acceptor (Cherney et al., 2010). Since the propensity for adventitious addition of GSH (Landry et al., 2018) or CoA (Figure S1A, B) to the resting form of ndSQR is ~100-fold lower than the solubilized enzyme, we speculate that the active site in ndSQR might be less exposed than observed in the crystal structures of the solubilized enzyme. Membrane anchoring and/or dimerization might reduce the size of the substrate entry channel, thereby restricting access of larger substrates in the resting enzyme.

In summary, we have identified CoA as an additional sulfur acceptor for human SQR, and demonstrated that its generation in situ leads to formation of the inhibitory CoA-SS⁻ to FAD CT complex in human ACADS. Our results implicate SQR in the coordinate regulation of sulfide and butyrate oxidation, potentially enabling the prioritization of sulfide clearance in colonic epithelial cells exposed to acute H₂S stress. In addition, our structures have provided unexpected insights into the catalytic mechanism of SQR and implicated a cysteine trisulfide as the active form of the redox cofactor. Another feature of the postulated mechanism, also supported by the crystal structure, is that the CT complex forms between FAD and ²⁰¹Cys-SS⁻, rather than with the thiolate ²⁰¹Cys-S⁻. A potential catalytic advantage of this redox motif is that the nucleophilic addition of sulfide to a trisulfide is chemically more facile than addition to a disulfide, explaining in part the ~2×10⁷ rate enhancement for sulfide addition to SQR versus sulfide addition to a disulfide in solution (0.6 M⁻¹ s⁻¹ at pH 7.4, 25 °C) (Cuevasanta et al., 2015).

STAR METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ruma Banerjee (rbanerje@umich.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

HT-29 colorectal adenocarcinoma cells (ATCC® HTB-38™, adult Caucasian female origin) were passaged at 37 °C in 5% CO₂ and 95% air in RPMI 1640 medium lacking glucose, supplemented with fetal bovine serum (10%) and penicillin/streptomycin (0.5 mg mL⁻¹) to induce differentiation (Zweibaum et al., 1985) and promote butyrate utilization (Leschelle et al., 2000). The cells were initially grown from a frozen stock that was authenticated by short-tandem repeat profiling performed by ATCC before this study.

METHOD DETAILS

Preparation of solubilized human SQR, ndSQR, and ACADS

Polyhistidine-tagged human SQR lacking the N-terminal mitochondrial leader sequence (amino acids 1–40) (Jackson et al., 2012) was expressed and purified as the selenomethionine-labeled enzyme as described previously (Libiad et al., 2014) with the following modifications. Briefly, a 1 L overnight culture of E. coli in Luria Bertani medium co-expressing plasmids for human SQR and cold-adapted chaperonins was pelleted and resuspended in 12 L of complete minimal medium supplemented with amino acids (80 μg mL⁻¹ each, lacking methionine), glycerol (0.4% v/v), thiamine (5 μg mL⁻¹), and the appropriate antibiotics. The initial A₆₀₀ of the culture was ~0.4. The culture was incubated at 37 °C, 230 rpm for 2 h, followed by addition of selenomethionine (40 μg mL⁻¹) and incubation for 1 h until A₆₀₀ was ~0.9. The cells were then incubated at 17 °C, 170 rpm for 1 h before induction with isopropyl β-D-1-thioglactopyranoside (300 μM) and incubation was continued overnight under the same conditions. The purification procedure for SeMet-SQR was identical to that described previously (Libiad et al., 2014), except that 1,2-diheptanoyl-sn-glycero-3-phosphocholine was substituted with n-dodecyl-β-D-maltoside (DDM, 0.05% w/v) as the solubilizing detergent.

For biochemical studies, SQR was purified as described (Libiad et al., 2014). Solubilized SQR was incorporated into nanodiscs formed with the MSP1E3D1 scaffold protein (Denisov et al., 2007) containing 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine as described to generate ndSQR (Landry et al., 2017). The concentration of solubilized SQR and ndSQR was estimated from the absorbance of the bound FAD cofactor using an extinction coefficient of 11,300 M⁻¹cm⁻¹ at 450 nm. Human ACADS (lacking the mitochondrial leader sequence and extending from residues 25–412) was expressed using the ACADS-pNIC-CTHF vector (plasmid #38870, from Addgene, deposited by Nicola Burgess-Brown). Briefly, a 200 mL overnight culture of E. coli in Luria Bertani medium co-expressing plasmids for human ACADS and the GroEL/GroES chaperonins was diluted 1:50 (v/v) into 6 L of Terrific Broth supplemented with the appropriate antibiotics, and grown at 37 °C with shaking at 230 rpm. Upon reaching an A₆₀₀ of ~0.8, the cells were incubated for 1.5 h at 18 °C before induction with isopropyl β-D-1-thioglactopyranoside (300 μM) and incubation was continued overnight under the same conditions. The cells were harvested at 4 °C, and all subsequent purification steps were performed at the same temperature. The cells were re-suspended in 50 mM Tris-acetate, pH 7.6, containing 150 mM NaCl and protease inhibitor cocktail (Roche), lysed by sonication, and centrifuged at 34,000 rpm for 45 min. The supernatant was loaded onto a 10 mL Ni-NTA column pre-equilibrated with 50 mM Tris-HCl buffer, pH 8.0, containing 300 mM NaCl. The column was then washed with 10 column volumes of the same buffer containing 20 mM imidazole, and ACADS was eluted with 20 mL of 250 mM imidazole in the same buffer. The eluted fractions were pooled, concentrated, and applied to a 25 mL Superdex 200 column (GE Healthcare) equilibrated with 50 mM Tris-HCl buffer, pH 8.0, containing 300 mM NaCl. The concentration of the eluted ACADS was estimated from the bound FAD cofactor using an extinction coefficient of 12,500 M⁻¹cm⁻¹ at 450 nm (Engel and Massey, 1971). CoA-SSH used for generating the CT complex in ACADS was prepared by reacting Na₂S with CoA disulfide (CoAS-SCoA), as described previously (Kabil and Banerjee, 2012)

Stopped-flow spectroscopy

Stopped-flow spectroscopy experiments were conducted in 100 mM potassium phosphate buffer pH 7.4, at 4 °C on a SF-DX2 double mixing stopped flow system from Hi-Tech Scientific, equipped with a 300–700 nm range photodiode array detector. The spectra and k_obs for CoA-dependent decay of the sulfide-induced CT complex in ndSQR were obtained in similar double mixing experiments as described previously (Landry et al., 2017, 2018). For these, ndSQR was first rapidly mixed 1:1 (v/v) with Na₂S and aged for ~35 ms to generate the CT complex, followed by a second 1:1 (v/v) rapid mixing with different concentrations of CoA. The kinetic traces at 675 nm monitoring the CT complex decay were fitted to a single-exponential equation using the KinetAsyst software (Cousins, 1997). The spectra and k_obs for CoA-induced CT complex formation in solubilized SQR and ndSQR were obtained from similar single mixing experiments as described (Landry et al., 2017, 2018). For these, solubilized SQR or ndSQR were rapidly mixed 1:1 (v/v) with different concentrations of CoA, and the kinetic traces at 675 nm monitoring the CT complex formation were fitted to a single-exponential equation using the KinetAsyst software. All enzyme and substrate concentrations for stopped-flow experiments reported in the respective figure legends are before any 1:1 (v/v) mixing steps.

Coupled enzyme activity assay for ndSQR with CoA as an acceptor

Recombinant human PDO (Tiranti et al., 2009) was purified as described previously (Kabil and Banerjee, 2012). Oxygen consumption rates in the coupled enzyme assay for ndSQR with PDO were determined as described (Landry et al., 2018). All reactions were conducted using a Clark oxygen electrode housed in a temperature-controlled, 1.5 mL Gilson-type chamber with constant magnetic stirring. PDO (2 μM) in 100 mM potassium phosphate, pH 7.4, at 25 °C was mixed with 0.06 mg mL⁻¹ BSA, CoA (0–15 mM), and CoQ₁ (150 μM). The reactions were initiated by injection of Na₂S (150 μM), followed immediately by injection of ndSQR (1 nM). Oxygen consumption was monitored with a Kipp and Zonen BD single channel chart recorder, with oxygen consumption expressed as μmol O₂ consumed min⁻¹ mg⁻¹ enzyme. To ensure complete coupling of the ndSQR and PDO reactions, the assay described above was first conducted using GSH (0–40 mM) as the acceptor, and the oxygen consumption rates were checked for consistency with CoQ₁ reduction rates obtained previously from ndSQR steady-state kinetic analyses (Landry et al., 2017).

Butyrate oxidation assay

Butyrate oxidation in HT-29 cells was monitored by conversion of [1-¹⁴C] butyrate to ¹⁴CO₂ as described previously (Roediger, 1982), with the following modifications. Briefly, cells were divided into 2 × 10⁶ cells per well in 6-well plates. The plated cells were incubated for 24 h, after which the media was replaced with serum-free RPMI 1640 medium lacking glucose, supplemented with 1 mM butyrate (containing 10 nCi [1-¹⁴C] butyrate) and L-carnitine (100 μM). Cells were then incubated for 3 h at 37 °C at atmospheric CO₂, with each well covered by a filter paper saturated with NaOH (1 M) to capture the released ¹⁴CO₂ from butyrate as ¹⁴C-bicarbonate. The reactions were terminated by addition of 125 mM citric acid, pH 6.0, and incubated for an additional 3 h at 25 °C to capture the remaining ¹⁴CO₂. The filter papers were then collected and suspended in EcoLite(+) liquid scintillation cocktail for counting. Counts were corrected for background levels of butyrate volatilization by using filter papers from plates with the same culture medium but lacking cells.

Crystallization of SQR

All SQR crystals were grown at 20 °C by the hanging drop vapor diffusion method using solubilized human SQR (17.4 mg mL⁻¹) in 50 mM Tris-HCl pH 8.0, containing sodium chloride (300 mM) and n-dodecyl β-D-maltoside (0.05% w/v). SeMet SQR crystals were grown by mixing protein (14 mg mL⁻¹) at a 1:1 (v/v) ratio with reservoir solution containing 100 mM N-(2-acetamido) iminodiacetic acid (ADA) pH 6.5, lithium sulfate (300 mM) and PEG 400 (30% w/v). Wedge-shaped crystals grew within a couple of weeks. The crystals were harvested in the aforementioned reservoir solution, which was supplemented with an additional 20% (w/v) PEG 400 as a cryoprotectant. Crystals were then flash-frozen in liquid N₂ and used for data collection. Apo-SQR crystals were grown by mixing purified SQR with an equal volume of reservoir solution containing 100 mM ADA pH 7.0, PEG 400 (30% w/v) and lithium sulfate (400 mM). The crystals were harvested in the aforementioned reservoir solution supplemented with glycerol (25% v/v) as a cryoprotectant, and flash-frozen in liquid nitrogen. SQR-CoQ₁ crystals were grown by combining SQR (17.4 mg mL⁻¹) and CoQ₁ (5 mM, in DMSO), followed by 1:1 (v/v) mixing with reservoir solution containing 200 mM ammonium tartrate dibasic, pH 6.6, and PEG 3350 (20% w/v) to yield a final CoQ₁ concentration of 2.5 mM, with initial crystals used as seed stock. Crystals from SQR-CoQ₁ co-crystallization wells were used for sulfite soaking, which was performed by incubation with Na₂SO₃ (2 mM). Similarly, sulfide-soaking of SQR-CoQ₁ crystals was performed by addition of Na₂S (2 mM). Crystals were then cryoprotected in their respective reservoir solutions supplemented with glycerol (35% v/v) and flash-frozen in liquid nitrogen.

X-ray data collection, structure determination, and data deposition

Diffraction data for SeMet SQR crystals was collected at the LS-CAT beamline 21-ID-G (Advanced Photon Source, Argonne National Laboratory) at 0.9786 Å wavelength. The space group and the unit cell size of the crystal were P2₁2₁2₁ and a=77.58Å, b=111.42Å, c=133.92Å, α=90°, β=90°, γ=90°, respectively. Three single-wavelength anomalous dispersion (SAD) data sets (3.27 Å resolution) from SeMet SQR crystals were merged, indexed, integrated and scaled by DIALS (Winter et al., 2018) via Xia2 (Winter et al., 2013). A total of 22 selenium sites from two copies of SeMet SQR in the asymmetric unit were found using SHELXC/D/E/ (Sheldrick, 2010). The low-resolution initial model was built using the buccaneer program in the CCP4 package (Cowtan, 2006). The model was further built manually using Coot (Emsley and Cowtan, 2004), and refinement calculations were carried out using Phenix.Refine (Afonine et al., 2010). Ligand restraints and libraries were applied during the refinements. A single asymmetric unit contains two copies of SeMet SQR. The final refined SeMet SQR model was then used as an input model for molecular replacement toward higher resolution datasets obtained from SQR (2.80 Å resolution), SQR-CoQ₁ (2.03 Å resolution), SQR-CoQ₁ + sulfite (2.21 Å resolution), and SQR-CoQ₁ + sulfide (2.55 Å resolution) crystals. Crystallographic data statistics are summarized in Table 1. Diffraction data for SQR crystals were collected at the LS-CAT beamline 21-ID-D (Advanced Photon Source, Argonne National Laboratory). The diffraction images were processed using HKL2000 (Otwinowski and Minor, 1997) for SQR, and DIALS for SQR-CoQ₁, SQR-CoQ₁ + sulfite, and SQR-CoQ₁ + sulfide. The molecular replacement solutions for SQR, SQR-CoQ₁, SQR-CoQ₁ + sulfite, and SQR-CoQ₁ + sulfide were determined using the SeMet SQR model as a search model. The final structures were completed using alternate cycles of manual fitting in Coot (Emsley and Cowtan, 2004) and refinement in REFMAC5 (Murshudov et al., 1997). The stereochemical quality of the final models was assessed using MolProbity (Chen et al., 2010). The omit maps (2F_o-F_c) were calculated by iteratively omitting 10% of the total region in the absence of sulfur, CoQ₁ and FAD using the composite omit map program in the PHENIX package (Afonine et al., 2012). Polder omit maps were calculated for the FAD cofactor, active site cysteines, and CoQ₁ ligand were calculated by omitting the surrounding electron density derived from bulk solvent, using the polder maps program in the PHENIX package. The diffraction data precision indicator values were estimated using SFCHECK in CCP4 (Vaguine et al., 1999). Structures presented in figures were generated using PyMOL 2.0.6.

Detection of sulfane sulfur in SQR

Solubilized SQR (60 M) in 50 mM Tris-HCl, pH 8.0, containing NaCl (300 mM) and DHPC (0.03%) was assayed for sulfane sulfur according to the cold cyanolysis method as previously described (Wood, 1987). SQR was combined with KCN (50 mM) and ammonium hydroxide (80 mM) in a volume of 1 mL and incubated for 45 min at 25 °C, followed by addition of formaldehyde (0.76%) and 200 L of Goldstein’s reagent (50 mM ferric nitrate in 25% nitric acid). Precipitated protein was removed by centrifugation at 15,000 × g, and samples were measured at 460 nm for formation of the ferric thiocyanate complex. Sulfane sulfur concentration was calculated using a standard curve generated with thiocyanate (0–500 M). To correct for absorbance from FAD released from acidification of SQR, a sample containing an equimolar concentration of FAD in buffer was used as a control. To detect sulfane sulfur in SQR after catalytic turnover, the same enzyme was mixed with Na₂S (200 μM) and CoQ₁ (300 μM) and incubated for 5 min at 25 °C, followed by buffer exchange at 4 °C by passage through a PD-10 desalting column before cyanolysis. The data for mol sulfane sulfur per mol monomer are presented as the mean ± SD for three independent preparations of SQR.

Docked models for GSH and CoA binding to human SQR

The AutoDock Vina program (The Scripps Research Institute, CA, USA) (Trott and Olson, 2010) was used for the docking simulations. GSH and CoA structures were obtained from the PDB. The grid maps for docking studies were centered on the putative GSH and CoA binding site on the matrix facing side of SQR and comprised 22 × 22 × 22 and 20 × 20 × 20 points with 1.0 Å spacing, respectively. The docked models were selected among the nine top candidates based on the orientation of the thiol group in GSH and CoA.

QUANTIFICATION AND STATISTICAL ANALYSIS

Kinetic data are presented as mean ± SD, with the number of technical repeats indicated in the corresponding figure legends. A Student’s paired t-test was used to estimate the significance of sulfide treatment on butyrate oxidation in HT-29 cells, where p<0.05 was considered statistically significant, with the number of technical repeats indicated in the corresponding figure legend.

DATA AND CODE AVAILABILITY STATEMENT

The atomic coordinate and structure factors for SQR (6OI5), SQR-CoQ₁ (6OIB), SQR-CoQ₁ + sulfite (6OIC), and SQR-CoQ₁ + sulfide (6OI6), were deposited in the RCSB Protein Data Bank.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Virus Strains
E. coli BL21 (DE3)	Invitrogen	Cat#C600003
Chemicals, Peptides, and Recombinant Proteins
Selenomethionine	Molecular Dimensions	MD12–503B
Human SQR	Libiad et al 2015	N/A
1,2-diheptanoyl-sn-glycero-3-phosphocholine	Avanti Polar Lipids	850306P
n-dodecyl-β-D-maltoside	Sigma-Aldrich	D4641
MSP1E3D1 nanodisc scaffold protein	Landry et al 2017	N/A
Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine	Anatrace Lipids	P516
Human ACADS	This study	N/A
Coenzyme Q1	Sigma-Aldrich	C7956
Sodium sulfite	Sigma-Aldrich	S0505
Sodium sulfide nonahydrate	Sigma-Aldrich	431648
Human PDO	Kabil et al 2012	N/A
Coenzyme A persulfide	Kabil et al 2012	N/A
Butyric acid, sodium salt	Alfa-Aesar	A11079
[1-14C] butyric acid, sodium salt	ViT rax	VC 171
Deposited Data
Crystal structure of selenomethionine-labeled SQR	This study	PDB ID: 6OI5
Crystal structure of SQR-CoQ1	This study	PDB ID: 6OIB
Crystal structure of SQR CoQ1 + sulfite	This study	PDB ID: 6OIC
Crystal structure of SQR CoQ1 + sulfide	This study	PDB ID: 6OI6
Crystal structure of native SQR	Jackson et al 2019	PDB ID: 6MP5
Crystal structure of A. ferrooxidans SQR	Zhang et al 2016	PDB ID: 3T31
Crystal structure of A. vinosum FCSD	Chen et al 1994	PDB ID: 1FCD
Experimental Models: Cell Lines
HT-29 human colorectal adenocarcinoma	ATCC	HTB-38
Recombinant DNA
Expression vector pET23a+-MATopzSQOR	Jackson et al 2012	N/A
Expression vector pCPN10/60	Aligent Technologies	230192
Expression vector pNIC-CTHF-ACADS	Burgess-Brown, unpublished	Addgene 38870
Expression vector pET28b+-MSP1E3D1	Denisov et al 2007	Addgene 20066
Expression vector pMW172-ETHE1 (PDO)	Tiranti et al 2009	N/A
Software and Algorithms
KinetAsyst	Cousins 1997	https://www.hi-techsci.com/instruments/kinetasyst-stopped-flow-systems/
DIALS	Winter et al 2018	https://dials.github.io/
Xia2	Winter et al 2013	https://xia2.github.io/
SHELXC/D/E	Sheldrick 2010	http://shelx.uni-goettingen.de/
CCP4	Cowtan 2006	https://www.ccp4.ac.uk/
Coot	Emsley and Cowtan 2004	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Phenix.Refine	Afonine et al 2010	https://www.phenix-online.org/
HKL2000	Otwinowski and Minor 1997	http://www.hkl-xray.com/
REFMAC5	Murshudiov et al 1997	http://www.ccp4.ac.uk/html/refmac5.html
MolProbity	Chen et al 2010	http://molprobity.biochem.duke.edu/
PHENIX	Afonine et al 2012	https://www.phenix-online.org/
SFCHECK	Vaguine et al 1999	http://www.ccp4.ac.uk/html/sfcheck.html
PyMOL	Schrodinger	https://pymol.org

Highlights.

• Human SQR utilizes CoA as an alternate sulfur acceptor generating CoA-persulfide

• SQR-derived CoA-persulfide inhibits ACADS and butyrate oxidation

• Four crystal structures of human SQR provide snapshots of its reaction coordinate

• Cysteine trisulfide is the active cofactor and reforms at the end of turnover

Significance.

Mitochondrial membrane-anchored SQR catalyzes the committing step in H₂S clearance in a reaction comprising sulfur transfer from H₂S to a thiophilic acceptor and electron transfer from H₂S via FAD to CoQ₁₀, which links to the electron transfer chain. We have identified CoA as an alternate sulfur acceptor for SQR, and demonstrated that the CoA-SSH product forms a stable CT complex with FAD in human ACADS, which is known to be inhibitory. The ability of SQR to generate CoA-SSH has important implications for regulating colon bioenergetics, since colonocytes preferentially utilize microbiota-derived butyrate as a biofuel. Our study suggests the hypothesis that CoA-SSH could help prioritize sulfide over butyrate oxidation to shield cells from respiratory poisoning during acute gut exposure to H₂S. Our crystallographic analyses have provided rich structural insights and lead us to postulate a different reaction mechanism. We have demonstrated that the active form of the enzyme has a redox-active cysteine trisulfide motif before and after in crystallo catalysis. The spectrally intense CT complex on SQR forms between ²⁷⁹Cys-SS⁻ and FAD, rather than between ²⁷⁹Cys-S⁻ and FAD, and was captured in sulfite-soaked SQR-CoQ₁ crystals. By implication, the subsequent flavin C4a adduct involves a covalent interaction with ²⁷⁹Cys-SS⁻. The structural basis for the substrate promiscuity exhibited by solubilized SQR is amply explained by the structures, which reveal a large entrance exposing the active site trisulfide to solvent and providing access to the thiolate sulfur acceptor moieties of GSH and CoA alike.