A distinct MaoC-like enoyl-CoA hydratase architecture mediates cholesterol catabolism in Mycobacterium tuberculosis.

The Mycobacterium tuberculosis (Mtb) igr operon plays an essential role in Mtb cholesterol metabolism, which is critical for pathogenesis during the latent stage of Mtb infection. Here we report the first structure of a heterotetrameric MaoC-like enoyl-CoA hydratase, ChsH1-ChsH2, which is encoded by two adjacent genes from the igr operon. We demonstrate that ChsH1-ChsH2 catalyzes the hydration of a steroid enoyl-CoA, 3-oxo-4,17-pregnadiene-20-carboxyl-CoA, in the modified β-oxidation pathway for cholesterol side chain degradation. The ligand-bound and apoenzyme structures of ChsH1-ChsH2(N) reveal an unusual, modified hot-dog fold with a severely truncated central α-helix that creates an expanded binding site to accommodate the bulkier steroid ring system. The structures show quaternary structure shifts that accommodate the four rings of the steroid substrate and offer an explanation for why the unusual heterotetrameric assembly is utilized for hydration of this steroid. The unique αβ heterodimer architecture utilized by ChsH1-ChsH2 to bind its distinctive substrate highlights an opportunity for the development of new antimycobacterial drugs that target a pathway specific to Mtb.

The resurgence of tuberculosis disease (TB) in the 1990s highlighted that tuberculosis is no longer a disease of the past. HIV/AIDS infection, multidrug resistant TB (MDR-TB), and extensively drug resistant TB (XDR-TB) have made the problem more complex.[1] A quarter of HIV infected deaths are caused by TB, and MDR-TB is present all over the world.[2] The drugs currently used to treat tuberculosis require long treatment times and have limited efficacy and significant toxicities. The first line drugs, including isoniazid, rifampicin, and pyrazinamide, require more than half a year’s treatment. Existing drugs only target a small fraction of essential growth pathways, and it is not difficult for Mycobacterium tuberculosis (Mtb) to overcome their inhibition by activating compensatory pathways. In 2010, 8.8 million people were infected with TB and 1.4 million died from TB.[2] New drugs based on novel pathways or targets that can shorten the therapy and eliminate drug-resistant strains are urgently needed.

In order to develop new strategies against TB, a better understanding of the virulence and pathogenesis of Mtb’s infection lifecycle is needed. Macrophages are responsible for defending against invasion by foreign pathogens. However, Mtb can reside in macrophages and replicate within them. Mtb successfully assimilates nutrition from the macrophage and hides from attack by the immune system. As an intracellular pathogen, the first challenge Mtb faces is how to obtain nutrients. Mtb preferentially metabolizes host-derived lipids during infection.[3,4] Among a total of 4000 Mtb genes identified through genome sequencing, at least 250 genes are potentially involved in lipid metabolism,[5] which is 5 times more lipid-metabolism genes than Escherichia coli. Many of the lipid-metabolism genes are induced by intracellular growth and infection.[5] Knowing the functions of genes and their corresponding products involved in lipid-utilization pathways in Mtb is critical for understanding Mtb’s infection biology and may provide new targets to develop anti-TB drugs.

In the process of defining genes required for survival of Mtb in macrophages by transposon library screening, several experiments highlighted the importance of the intracellular growth operon (igr operon), comprised of Rv3540c to Rv3545c, for survival in macrophages.[6,7] Deletion of the operon (H37Rv:Δigr) results in attenuated growth in both macrophages and mice.[6] The genes of the igr operon were predicted to encode lipid β-oxidation enzymes.[8] However, the operon is not required for growth on even or odd chain fatty acids.[6] During the course of Mtb infection, cholesterol has been demonstrated to serve as a nutrient and is significant for Mtb persistence. Many steroid-utilizing genes have been shown to be critical both in vitro and in vivo.(9) The location of the igr operon in the 82-gene cholesterol catabolism cluster suggested a potential function in cholesterol metabolism,[10] and the operon was found to be required for growth of Mtb in vitro in the presence of cholesterol.[11] Metabolite profiling of the H37Rv:Δigr strain using isotopically labeled cholesterol resulted in accumulation of a metabolite. The structure of the metabolite led to the hypothesis that the igr operon is required for β-oxidation of C20–C22 in the cholesterol side chain[12] (Figure 1a,b).

Figure 1

Biochemical function of igr operon and gene constructs to express ChsH1-ChsH2, ChsH1-ChsH2N. (a) The igr operon in Mtb; chsH1 (Rv3541c) and chsH2 (Rv3542c) are colored in cyan and pink, respectively. Other genes are shown in gray. (b) The biochemical function of the gene products encoded by the igr operon. (1) 3-oxo-4-pregnene-20-carboxyl-CoA (3-OPC-CoA); (2) 3-oxo-4,17-pregnadiene-20-carboxyl-CoA (3-OPDC-CoA); (3) 17-hydroxy-3-oxo-4-pregnene-20-carboxyl-CoA (17-HOPC-CoA); (4) androstenedione. (c) Constructs for expression of ChsH1-ChsH2 and ChsH1-ChsH2N. An N-terminal 6× His tag is encoded at the 5′ end of the first gene of each construct. The ltp2 gene following ChsH1 is not shown for clarity. The number of residues encoded by each gene is labeled. The factor Xa cleavage sequence is introduced after the Arg187 codon in the chsH2 gene. (d) SDS-PAGE analysis of system I isolated by IMAC and of system II isolated by IMAC and then subjected to factor Xa cleavage. Protein identities were confirmed by tryptic digest and MALDI-TOF MS fingerprinting.

The six genes in the igr operon are annotated as a lipid transfer protein (Ltp2/Rv3540c), MaoC-like hydratases (Rv3541c, Rv3542c), acyl-CoA dehydrogenases (Rv3543c, Rv3544c), and a cytochrome P450 (cyp125/Rv3545c).[11] Enzymatic and structural studies demonstrated that Cyp125 oxidizes cholest-4-en-3-one or cholesterol at C-26 to cholest-4-en-3-one-26-oic acid.[13] However, in the H37Rv strain employed in the igr studies, loss of this activity is compensated by Cyp142.[14] Biophysical characterization of the acyl-CoA dehydrogenases Rv3543c-Rv3544c, named ChsE1-ChsE2, showed that they form an obligate α2β2 heterotetramer with two active sites and two FAD cofactor binding sites.[12,15] ChsE1-ChsE2 catalyzes the C17–C20 dehydrogenation of steroid acyl-CoA thioester 3-oxo-4-pregnene-20-carboxyl-CoA.[15] Earlier pull down experiments demonstrated that Rv3541c and Rv3542c form a protein complex.[12] All of the above information established the importance of the igr operon for cholesterol metabolism by Mtb and highlighted the possibility that a second heteromeric structural motif may be encoded in the operon.

Based on a protein sequence BLAST search against the nonredundant protein sequence database (blastp), the igr operon encodes two potential MaoC-like enoyl-CoA hydratases, Rv3541c and Rv3542c. However, their role in cholesterol side chain degradation was unclear since known fatty acid β-oxidation cycles in bacteria, as well as eukaryotes, utilize a crotonase family member to catalyze hydration of fatty enoyl-CoA metabolites. MaoC-like enoyl-CoA hydratases are structurally distinct from crotonases, and in bacteria, they are employed in polyhydroxyalkanoate energy storage pathways instead of for energy metabolism.[16] In eukaryotes, MaoC-like enoyl-CoA hydratases are employed in peroxisomal β-oxidation of di- and trihydroxy-5β-cholestanic acid to chenodeoxycholic acid and cholic acid, respectively. However, these hydratases are part of a multifunctional enzyme (MFE) that includes a fused 3-hydroxy-acyl-CoA dehydrogenase domain.[17,18] Both the bacterial and peroxisomal MaoC-like enoyl-CoA hydratases characterized to date are all homodimers,[16,19] although the peroxisomal enoyl-CoA domains are larger than their bacterial homologues.

Sequence similarities indicated that both Rv3541c and Rv3542c have MaoC-like hydratase domains, and in addition Rv3542c includes a DUF35/DUF35_N domain. Here, we demonstrate that Rv3541c and Rv3542c form a heterodimer, two of which further dimerize to form an α2β2 heterotetramer. The unique architecture of this heterotetramer is in stark contrast to previously characterized structural homologues, for example, peroxisomal and bacterial enzymes, which form homodimers. We also demonstrate that Rv3541c-Rv3542c catalyzes the hydration of a steroid substrate, 3-oxo-4,17-pregnadiene-20-carboxyl-CoA (3-OPDC-CoA). We now refer to Rv3541c-Rv3542c as ChsH1-ChsH2 to emphasize its function in cholesterol side chain enoyl-CoA hydration. Our data suggest that Mtb metabolizes sterol side chains through a unique assembly of β-oxidation enzymes. The structures described here highlight opportunities for the design of inhibitors specifically targeting cholesterol catabolism in actinobacteria, particularly Mycobacterium tuberculosis.

Results and Discussion

Enzymes that degrade the core structure of steroids like cholesterol are not present in humans. On the other hand, Mtb has the ability to metabolize completely steroids of mammalian origin and utilize these catabolic intermediates for both energy and other processes related to virulence and pathogenesis. Despite the fact that many of these steroid-utilizing genes have been shown to be critical both in vitro and in vivo, no enzyme crystal structure to date has been elucidated of a steroid utilizing β-oxidation enzyme.[9] Given the unclear structural and functional homologies of ChsH1 and ChsH2, as well as their association, we undertook investigation of the structure and activity of the ChsH1-ChsH2 complex in order to understand its role in Mtb cholesterol metabolism.

ChsH1-ChsH2 Forms an Obligate α2β2 Heterotetramer and Preferentially Catalyzes Hydration of Steroid Enoyl-CoA

Initially, ChsH1 and ChsH2 were expressed individually as N-terminally His6-tagged proteins in E. coli. However, both ChsH1 and ChsH2 precipitated directly after elution from an ion affinity chromatography (IMAC) column indicating that ChsH1 and ChsH2 are not stable under these expression conditions.

Construct pigr3(12) (Table 1) was expressed heterologously to provide ChsH1-ChsH2 that was isolated by IMAC and further purified using gel filtration chromatography. SDS-PAGE gel analysis confirmed the coisolation of ChsH1 and ChsH2 (Figure 1c,d). Analytical gel filtration chromatography demonstrated that ChsH1-ChsH2 forms a stable oligomeric complex in solution (Figure 2a) with an approximate 1:1 stoichiometry. Further analysis by analytical ultracentrifugation (AUC) equilibrium experiments established that it forms a complex with a molecular weight of 99 kDa (Figure 2b). Thus, given the relative stoichiometry and the individual molecular weights of ChsH1 (14 kDa) and ChsH2 (36 kDa), the ChsH1-ChsH2 complex is an α2β2 heterotetramer. To confirm the stoichiometry of the complex, the assembly was analyzed by LC/UV/MS (Figure 2d). Integration of the ChsH1 and ChsH2 peaks and calculation of their relative molar concentration provided a relative molar stoichiometry of one ChsH1 to one ChsH2, consistent with formation of an α2β2 heterotetramer.

Table 1

Expression Constructs Used in This Work

construct name	genes	restriction sites	purified enzyme	source/ref
pET28b				Novagen
pChsH1	Rv3541c	NdeI/XhoI	N-His6-ChsH1	this worka
pChsH2	Rv3542c	NdeI/NotI	N-His6-ChsH2	this worka
pigr3	Rv3542c, Rv3541c, Rv3540c	NdeI/HindII	N-His6-ChsH2-ChsH1	(12)
pChsH1ChsH2N	Rv3542c, Rv3541c, Rv3540c	NdeI/HindII	N-His6-ChsH2-ChsH1	this worka
pigr3H1D29A	Rv3542c, Rv3541c, Rv3540c	NdeI/HindII	N-His6-ChsH2-ChsH1D29A	this worka
pigr3H1H4A	Rv3542c, Rv3541c, Rv3540c	NdeI/HindII	N-His6-ChsH2-ChsH1H34A	this worka
pCasI	Ro05822	NdeI/XhoI	N-His6-CasI	this workb

Genes were cloned from H37Rv genomic DNA and ligated into pET28b with the indicated restriction sites to include an N-terminal His6 fusion tag.

The Rhodococcus jostii RHA1 casI (Ro05822) gene was synthesized and ligated into pET15b with the indicated restriction sites to include an N-terminal His6 fusion tag.

Figure 2

ChsH1-ChsH2 and ChsH1-ChsH2N are heterotetramers in solution and prefer to catalyze hydration of a steroid enoyl-CoA substrate. (a) ChsH1-ChsH2 and ChsH1-ChsH2N were analyzed by analytical size exclusion chromatography on a Superdex 75 column. (b) ChsH1-ChsH2 and (c) ChsH1-ChsH2N were centrifuged at speeds of 10000, 20000, and 30000 rpm at 20 °C in an analytical ultracentrifuge. A representative fit for each sample is shown. The solid line shows the fit of the data to the ideal species model, and the residuals of the fit are graphed below the fit. The best global fit for each protein provided molecular weights of 98.6 and 65.2 kDa for ChsH1-ChsH2 and ChsH1-ChsH2N, respectively. (d, e) Reverse phase LC/UV/MS chromatograms of ChsH1-ChsH2 and ChsH1-ChsH2N. Peaks A, B, and C were identified as ChsH2, ChsH1, and ChsH2N, respectively, by deconvolution of multiple charged states in the corresponding ESI+ MS spectra. The absorbance of peaks were integrated, and relative concentrations were determined from the calculated extinction coefficients of ChsH2, ChsH1, and ChsH2N [ε280(ChsH2) = 59993 M–1 cm–1, ε280(ChsH1) = 16960 M–1 cm–1, and ε280(ChsH2N) = 33523 M–1cm–1]. (f) Specific activities of ChsH1-ChsH2 and ChsH1-ChsH2N. ChsH1-ChsH2 and ChsH1-ChsH2N were assayed with octenoyl-CoA, decenoyl-CoA, and 3-oxo-4,17-pregnadiene-20-carboxyl-CoA (3-OPDC-CoA).

We assayed the ChsH1-ChsH2 enzyme complex with three different substrates: octenoyl-CoA, decenoyl-CoA, and 3-oxo-4,17-pregnadiene-20-carboxyl-CoA (3-OPDC-CoA). The hydration process was monitored spectrophotometrically, and the formation of the hydrated products was confirmed by matrix-assisted laser desorption ionization (MALDI)-time of flight (TOF) mass spectrometry. Although ChsH1-ChsH2 catalyzed the hydration of all three substrates, the highest turnover activity was observed with 3-OPDC-CoA (Figure 2f). Therefore, ChsH1-ChsH2 has the ability to bind bulky substrates and preferentially catalyzes the hydration of steroid enoyl-CoA compared with aliphatic enoyl-CoAs, thus establishing a role in cholesterol metabolism for this enzyme.

A Truncated Form of ChsH1-ChsH2N Is Sufficient for Catalytic Activity

In the process of crystallizing full length ChsH1-ChsH2, diffraction quality crystals were obtained. The crystals diffracted to 1.52 Å resolution (Table 2), and the structure was solved using single wavelength anomalous dispersion. The crystals contained four αβ heterodimers (two α2β2 heterotetramers) in the asymmetric unit (Figure 3a, Supplementary Figure 1 and Note, Supporting Information). The electron density was of excellent quality and allowed a model of full length ChsH1 and the N-terminal domain of ChsH2 (180 amino acids out of 311 amino acids) to be built. The absence of further density for the C-terminal domain of ChsH2 prompted us to analyze the protein content of our crystals. Further analysis by SDS-PAGE and MALDI-TOF mass spectrometry confirmed that around a 130 amino acid C-terminal fragment of ChsH2 was lost through proteolysis prior to crystallization. We henceforth refer to this fragment as ChsH2C.

Table 2

Data Collection and Refinement Statistics

	ChsH1-ChsH2N	ChsH1-ChsH2N/3-OPC-CoA
Data Collection
space group	P21	P6122
cell dimensions
a, b, c (Å)	51.8, 134.1, 90.4	51.9, 51.9, 436.2
α, β, γ (deg)	90, 91.6, 90	90, 90, 120
resolution (Å)	37.6–1.52 (1.50–1.54)a	72.70–1.76 (1.768–1.762)a
Rmerge	0.033 (0.434)	0.068 (0.605)
Rmeas	0.038 (0.585)	0.073 (0.642)
Rpim	0.026 (0.408)	0.024 (0.208)
CC1/2	0.999 (0.886)	0.999 (0.852)
I/σI	19.2 (2.2)	18.8 (8.8)
completeness (%)	99.2 (95.7)	100 (100)
multiplicity	13.4 (11.2)	8.0 (8.8)
Wilson B-factor	17.04	21.19
Refinement
resolution (Å)	37.6–1.52	72.70–1.76
no. reflns	180 565	40 369
Rwork/Rfree	0.1768/0.2046	0.1895/0.2322
CC*	1.0 (0.976)	1.0 (0.963)
no. atoms
total	10,103	2,607
protein	8,920	2,286
ligand		76
water	1163	263
B-factors
protein	24.0	34.8
ligand		41.0
water	33.7	40.6
rms deviations
bond lengths (Å)	0.009	0.016
bond angles (deg)	1.15	1.57
PDB ID	4W78	4W7B

Values in parentheses are for the highest-resolution shell.

Figure 3

Overall architecture and topology of ChsH1-ChsH2N. (a) Two α2β2 heterotetramers of ChsH1-ChsH2N in one asymmetric unit (ASU). Surface representation of the structure corresponds to chain color. (b, c) ChsH1-ChsH2N adopts a heterotetrameric assembly with two ChsH1 chains and two ChsH2N chains. One ChsH1 chain and one ChsH2N chain fold into a heterodimer and a pair of heterodimers assembles into a tetramer. The identities of the chains are labeled. The structure in panel c is obtained by a 90° rotation of the structure in panel b around the Y-axis. (d) Schematic representation of one heterodimer from the tetramer. Each chain possesses a conserved “hot-dog” fold. α-Helices and β-strands from ChsH1 and ChsH2N are shown as ribbons and arrows, respectively; C-termini and N-termini are labeled. 310-Helices are labeled η. (e) Topology representation of one heterodimer. The color-coding is the same as in panel b. Secondary structure was assigned using DSSP.

In order to reliably obtain the protein complex ChsH1-ChsH2N and to reproduce the crystallization, an expression plasmid, pChsH1ChsH2 (Table 1), was constructed that introduced a factor Xa cleavage site followed by four glycine residues after Arg187 of ChsH2 (Figure 1c). The ChsH1-ChsH2 protein complex was obtained upon cistronic expression using the construct pChsH1ChsH2. The full-length protein was purified using IMAC and gel filtration chromatography and then cleaved with factor Xa to provide ChsH1-ChsH2N as confirmed by SDS-PAGE gel and analytical gel filtration (Figure 1d). The molecular weight of ChsH1-ChsH2N was determined to be 65 kDa by AUC (Figure 2c) consistent with the predicted molecular weights of ChsH1 and His6 tagged ChsH2N (14 kDa and 23 kDa, respectively). The relative molar stoichiometry of the subunits remained 1:1 based on LC/UV/MS analysis (Figure 2e). Therefore, ChsH1-ChsH2N still forms an α2β2 heterotetramer. The C-terminal fragment of ChsH2, referred to as ChsH2C, precipitated during factor Xa cleavage.

In order to test whether ChsH1-ChsH2N is still catalytically active, we analyzed its ability to catalyze hydration of the substrates described above. Importantly, ChsH1-ChsH2N retained catalytic activity, indicating that ChsH2C is not required for catalysis of hydration (Figure 2f). Furthermore, we were able to obtain crystals of the new protein complex ChsH1-ChsH2N. These crystals diffracted to 1.54 Å and yielded a structure identical to that originally obtained.

The Structure of ChsH1-ChsH2N Reveals a Unique MaoC-like Enoyl-CoA Hydratase Architecture

Consistent with annotations based on BLAST alignments, the crystal structure shown here establishes that ChsH1-ChsH2N belongs to the MaoC-like enoyl-CoA hydratase family. This family contains homodimeric MaoC-like enoyl-CoA hydratases that are comprised of a β-sheet that wraps around a central α-helix, which is known as a hot-dog fold. A second structural motif in MaoC-like hydratases is a three helix segment, referred to as an active site housing segment, and this motif is usually fused with the hot-dog fold domain in the family.[20]

Surprisingly, rather than the prototypical homodimeric structure, ChsH1-ChsH2N forms a dimer of a unique αβ heterodimeric assembly comprising one standard and one nonstandard MaoC-like hydratase. In the ChsH1-ChsH2N tetramer, one ChsH1 chain and one ChsH2N chain form an intimate heterodimer. Two of these heterodimers combine to form the heterotetramer (Figure 3b,c,d). As far as we know, ChsH1-ChsH2 is the first example of an α2β2 heteromeric MaoC-like enoyl-CoA hydratase.

The interaction between the two heterodimers involves helices α1′ from ChsH1 and α2 from ChsH2N. The helices from one heterodimer form a compact interface with the other heterodimer through salt bridges and hydrogen bonding interactions (Supplementary Figure 2a,b,c, Supporting Information). Arg37 from ChsH1 in one heterodimer forms a double salt bridge with Asp38 from the second ChsH1. Because of the 2-fold rotational symmetry in the dimer–dimer interface, four salt bridges are formed. Six hydrogen bonds are contributed from helices α2 of the ChsH2N chains (Supplementary Figure 2e,f, Supporting Information).

Importantly, this heterodimer–heterodimer interface is clearly divergent with respect to other MaoC-like hydratase homologues. A DALI[21] search revealed homologues from several species, ranging from bacteria (Aeromonas caviae; AcRH) and yeast (Candida tropicalis; CtRH) to mammals (Homo sapiens; HuRH). Interestingly, the ChsH1-ChsH2N heterotetramer is comparable to a CtRH homodimer and a HuRH homodimer. The two monomers in CtRH or HuRH dimerize through a similar four-helix bundle (Supplementary Figure 2d, Supporting Information). However, in those cases, the helices from one monomer interact with the helices from the other monomer orthogonally[22] in a very different structural arrangement from ChsH1-ChsH2 (Supplementary Figure 2b,d, Supporting Information).

The ChsH1-ChsH2N Heterodimer Is Structurally Distinct from Its MaoC-like Hydratase Homologues

In the ChsH1-ChsH2N heterodimer, five β-strands (β1′−β3′−β4′−β5′−β2′) from ChsH1 and four β-strands (β1−β4−β3−β2) from ChsH2N form an antiparallel nine-stranded β-sheet. Helix α3′ from ChsH1 and α4 from ChsH2N are above the antiparallel β-sheet in the viewpoint shown in Figure 3d. Helices α2−η1−α3 from ChsH2N and α1′−η1′−α2′ from ChsH1 form a third layer on top of α3′ and α4 and are almost 2-fold symmetric (Figure 3d,e). This architecture and the extensive interface between the monomers imply that ChsH1 and ChsH2N form an obligate heterodimer.

The overall structure of the ChsH1-ChsH2N heterodimer from Mtb is disparate from its homologues across different species. The ChsH1-ChsH2N heterodimer has two hot-dog folds in total. Interestingly, a similar architecture is present in other structural homologues, but the two hot-dog folds are present as a single monomer (CtRH or HuRH) or a homodimer (AcRH) (Supplementary Table 1, Supporting Information). The fold of ChsH1 is most similar to the C-terminal domain of the CtRH or HuRH monomer and corresponds to the structure of a single chain of AcRH (Supplementary Table 1, Supporting Information). Surprisingly, ChsH2N, although similar to the N-terminal domains of CtRH and HuRH monomers, deviates from the standard hot-dog fold of AcRH. In addition, both the structure and sequence have low similarity in the active site housing segment (Figure 5c, Supplementary Figure 3, Supporting Information).

Figure 5

Structure comparison and secondary structure sequence alignment between ChsH1 and ChsH2N. (a, b) ChsH1 possesses a standard hot-dog fold, and ChsH2N adopts a nonstandard hot-dog fold. The cartoon representations are colored by secondary structure. The secondary structures are labeled as in Figure 3d,e. N-termini and C-termini are labeled. In ChsH1, the five-stranded β-sheet (β1′-β3′-β4′-β5′-β2′) wraps around α3′, forming a hot-dog fold, and the proposed active site housing segment is comprised of α1′, η1′, and α2′. In ChsH2N, a four-stranded β-sheet (β1-β4-β3-β2) wraps around α4, assembling into a nonstandard hot-dog fold; α2, η1, and α3 fold to form a potential active site housing segment. The potential active site residues are colored blue, and the simulated annealing omit maps of those residues are calculated to reduce the effects of model bias. Electron density is shown as mesh (σ = 3). (c) Cartoon representation of superimposed ChsH1 (yellow) and ChsH2N (magenta). Structurally similar parts are rendered transparent, and the variations are marked. The structure on the right is obtained by rotating the structure on the left 90° about the X-axis. Potential active site residues are shown as sticks. (d) Secondary structural sequence alignment between ChsH1 and ChsH2N. Identical residues are highlighted in black. The secondary structure elements from ChsH1 (yellow) and from ChsH2N (magenta) correspond to panels a and b. Residues that constitute active site housing segments are highlighted in the green box. The central α-helices from ChsH1 and ChsH2N are highlighted in the red box.

The ChsH1-ChsH2N Heterotetramer Forms a Complex with 3-Oxo-4-pregnene-20-carboxyl CoA

In order to investigate the structural basis of preferential hydration of a steroid enoyl-CoA instead of aliphatic enoyl-CoA’s, we cocrystallized ChsH1-ChsH2N with a substrate analog, 3-oxo-4-pregnene-20-carboxyl CoA (3-OPC-CoA) (Figure 4). The crystals diffracted to a resolution of 1.76 Å (Table 2), and molecular replacement was used to solve the ChsH1-ChsH2N/3-OPC-CoA structure using the apoprotein structure as the template.

Figure 4

ChsH1-ChsH2N complexed with 3-oxo-4-pregnene-20-carboxyl-CoA (3-OPC-CoA). (a) Overall atomic picture of ChsH1-ChsH2N/3-OPC-CoA. The holo-heterotetramer is in the same orientation as Figure 3b. A single heterotetramer is in one asymmetric unit (ASU). ChsH1 chains are colored blue, and ChsH2N chains are colored magenta. Ligand 3-OPC-CoA is rendered as sticks and colored by atom. Molecular surfaces are rendered transparent. (b) Superimposed heterodimer apo-ChsH1-ChsH2N structure and ChsH1-ChsH2N/3-OPC-CoA structure. Superimposed ChsH1 chains are colored in blue, and their surface is shown. The ChsH2N chains from Apo-ChsH1-ChsH2N and ChsH1-ChsH2N/3-OPC-CoA are shown in pink and magenta, respectively. The variations occur in the segments of the ChsH2N chains that are labeled. (c) The difference in the opening of the binding pocket before and after 3-OPC-CoA binding. (d) A simulated annealing Fo – Fc omit map of 3-OPC-CoA contoured at 2.5σ was calculated to reduce the effects of model bias. A chemical structure of 3-OPC-CoA is shown and colored by atom. (e) Ligand-binding interactions in the ChsH1-ChsH2N heterodimer. The hydrogen bonds are shown as black dashes. Residues that interact with 3-OPC-CoA are labeled. 3-OPC-CoA fits between β1 from ChsH2N and β2′from ChsH1.

The atomic picture of the ChsH1-ChsH2N/3-OPC-CoA complex established that one binding site exists per ChsH1-ChsH2N heterodimer. In each heterodimer, the binding pocket is located at the ChsH1-ChsH2N interface and intrudes into ChsH2N (Figure 4a). Binding of 3-OPC-CoA results in a large quaternary structural change compared with the apo-ChsH1-ChsH2N complex (Figure 4b,c). The major changes occur in the ChsH2N chain. Movements of α1, α5, loop I, and loop II (6–7 Å) and the smaller movements of β1−β4−β3−β2 generate enough space to accommodate the 3-OPC-CoA ligand in the tunnel (Figure 4b,c). We posit that the heterodimer interface serves to enable the large structural change required for binding the polycyclic steroid system, which could not occur in a monomeric binding site like that seen in CtRH or HuRH.

The ligand is in a boomerang conformation, which is similar to the orientation of 3-hydroxy-octanoyl-CoA in CtRH.[19] The “boomerang” sits between the ChsH1 chain and the ChsH2N chain with its elbow, cysteamine, positioned across β1 and β2′ (Figure 4e). The whole boomerang bends toward ChsH2N with the coenzyme A moiety exposed to solvent and the four-ring steroid system buried inside ChsH2N (Figure 4a,c; Supplementary Figure 4, Supporting Information).

Binding interactions with ChsH2N anchor the coenzyme A moiety to the protein. The amine group (N1″) is hydrogen bonded to Asn106 in β1, N2″ is hydrogen bonded to Arg163 in β5, and O3″ is stabilized by Ala137 from loop II. The elbow of the boomerang is stabilized through a hydrogen bond between N4″ nitrogen and Gly81 from β2′ in ChsH1. The steroid part of the “boomerang” has fewer specific interactions with the enzyme compared with the coenzyme A moiety, consistent with the steroid’s hydrophobicity. The thioester O5″ is stabilized through a hydrogen bonding interaction with Ala104-ChsH2N. No polar interactions were found that stabilize the O3–C3 ketone (Figure 4d, 4e).

ChsH1 possesses a canonical MaoC-like hydratase hot-dog fold providing the active site, but ChsH2N possesses a modified hot-dog fold that confers the ability to bind bulky substrates

ChsH1 includes a complete hot-dog fold and an active site housing segment (Figure 5a). Five antiparallel β strands (β1′-β3′-β4′-β5′-β2′) act as a bun, wrapping around the sausage-like central helix α3′ to form a standard hot-dog fold (Figure 5a), characterized by the four-turns in the central α-helix.[19] The region from α1′, η1′ to α2′ comprises the active site housing segment. The strictly conserved active site residues, Asp29 and His34 (Supplementary Figure 5a), located in the active site housing segment (α1′−η1′−α2′) in ChsH1 (Figure 5a), are almost in identical positions with respect to the active sites in AcRH, CtRH and HuRH (Figure 6b). From the ChsH1-ChsH2N:3-OPC–COA structure, the C17–C20 bond, which is the site of hydration, sits in apposition to the active site residues Asp29/His34 from ChsH1 (Figure 4d, 4e).

Figure 6

Comparisons of MaoC-like enoyl-CoA hydratases across species. (a, b) Three-dimensional structural conservation between AcRH monomer (gray), C-terminal domain of CtRH (C-CtRH) (residue numbers 163–280; purple), C-terminal domain of HuRH (C-HuRH) (residue numbers 174–298; yellow), and ChsH1 chain (cyan). The structural elements are labeled following Figures 3d and 5a. The conserved active sites are enlarged in the box. (c) Three-dimensional structural comparison between AcRH monomer (gray), N-terminal domain of CtRH (N-CtRH) (residue numbers 1–135; purple), N-terminal domain of HuRH (NHuRH) (residue numbers 1–146; yellow), and ChsH2N chain (pink). Active site residues from the other chain of AcRH, C-terminal domains of CtRH and HuRH, or ChsH1 are shown as sticks. The central α-helices are colored red. Molecular surfaces are rendered transparent. The length of the central α-helices correspond to the 3-D structures above them. For simplicity, 310 helices and α-helices are both shown as cylinders. Protein sequences shown underneath are the amino acids comprising the central helices. The sizes of the central helices dictate the substrate preferences. The preferred substrates correspond to the structure above them. AcRH (PDB code 1iq6); CtRH (PDB code 1pn2); HuRH (PDB code 1s9c).

To establish that Asp29 and His34 correspond to the active site of the ChsH1-ChsH2 complex, Asp29 and His34 were mutated to alanine in independent constructs (Table 1). The purified protein complexes were analyzed by SDS-PAGE and by analytical size exclusion chromatography. They both had the same elution profile as wild-type ChsH1-ChsH2. Therefore, ChsH1D29A-ChsH2 and ChsH1H34A-ChsH2 still form α2β2 heterotetramers. The activities of these two mutant protein complexes were tested with octenoyl-CoA, decenoyl-CoA, and 3-OPDC-CoA. ChsH1D29A-ChsH2 retained 2%–4% of the wild-type activity. Moreover mutation of His34 to alanine abolished the catalytic activity within detection limits (less than 0.0062% activity) (Supplementary Figure 5b, Supporting Information).

In striking contrast to ChsH1, the structure of ChsH2N varies substantially with respect to the standard MaoC-like hydratase structure, for example, ChsH1, AcRH, or the C-terminal domains of CtRH or HuRH (Figures 5b,c,d and 6). Only four antiparallel β-strands wrap around the central helix α4. Furthermore, helix α4 contains two turns, and one turn is a 310 helix (Figure 5b,d). The short central helix (α4) renders the hot-dog fold incomplete (Figure 5c). This short, modified helix results in an open binding pocket formed by the four-stranded β-sheet. Helix α1 and α5 are two new flexible elements (Figures 3d and 5c) that are not present in other known MaoC-like hydratase structures. Even though the active site housing segment (α2– η1−α3) is structurally conserved, the active site histidine is replaced with a tyrosine (Figure 5b,d).

The distinctively short central α-helix (α4) is a source of another major difference (Figures 5c and 6c) and has the most important functional implication. The short helix α4 and the following long loop generate a bigger space in the hot-dog fold compared with other MaoC-like hydratases (Figure 6c), and the presence of this space is critical to allow the ChsH1-ChsH2N heterodimer to bind the steroid substrate. Consistently, comparison of this structural feature in the structures of ChsH1-ChsH2N and its homologues (Figure 6c) reveals that as the central α-helix of the hot-dog fold becomes shorter, the binding cavity becomes wider and sufficiently spacious to accommodate increasingly larger and bulkier substrates. It appears that upon shifting from linear alkyl chain and branched alkyl chain substrates to the rigid rings of a steroid system, the protein architecture transitions from small monomer to extended monomer, and finally to a heterodimer (Figure 6c).

For example, in the structure of N-CtRH complexed with its physiologic product, (3R)-hydroxy-octanoyl-CoA, the ligand is accommodated in the interface between the nonstandard N-terminal hot-dog fold and the complete C-terminal hot-dog fold.[19] The pantothenate and adenosine triphosphate of the ligand are exposed to solvent, and the ten-carbon acyl chain points toward the shorter central α-helix in the N-terminal nonstandard hot-dog fold.[19] Interestingly, the ChsH1-ChsH2N dimer has a similar, yet distinct, substrate binding mode; ChsH1 provides the active site and ChsH2N provides the binding pocket (Figure 4e). The available space in the suggested binding tunnel of the hot-dog fold is restricted by the rigid central helix, so the length of the helix influences substrate preferences (Figure 6c; Supplementary Table 1, Supporting Information). The existence of a nonstandard hot-dog fold with shorter central helices in the N-terminal domains of CtRH, HuRH, and ChsH2N chain suggests that one active site has evolved into a binding site to accommodate bulky substrates concomitant with the sacrifice of one set of active site catalytic residues (Figure 6c; Supplementary Table 1, Supporting Information).

The standard central helix of 16 amino acids in the complete hot-dog fold confers specificity for four to six carbon enoyl-CoAs.[23] In CtRH, the bent central helix of 13 amino acids in the N-terminal hot-dog fold enables accommodation of long-chain enoyl-CoAs (C10–C22) in the binding site.[19] The shorter central α-helix of 11 amino acids and bigger pocket in the N-terminal hot-dog fold of HuRH extends the accepted enoyl-CoA ester chain length up to C26 and methyl-branched enoyl-CoA’s. The central helix α4 of eight amino acids and adjacent flexible loop in ChsH2N generates an even more flexible active site with a larger pocket. The enlarged pocket in ChsH1-ChsH2N makes it possible to accommodate steroid CoA thioesters as substrates, consistent with its function in cholesterol degradation in Mtb(12) (Figure 6c; Supplementary Table 1, Supporting Information).

Phylogenetic Relationships of ChsH1-ChsH2

Our characterization of the unusual ChsH1-ChsH2 architecture prompted us to assess whether the heteromeric assembly might also exist in other organisms, and if so, whether those organisms were known to metabolize cholesterol. The presence of a hot-dog motif is insufficient to identify MaoC-like hydratase candidates because it is present in other enzyme families.[20] However, sequences in the active site housing segment are highly conserved and, in combination with the hot-dog fold, clearly define MaoC-like hydratases (Supplementary Figure 5a, Supporting Information). By comparing the MaoC-like hydratase sequences from human, fungi, and bacteria, we identified a generalized motif [VIL]-[AVI]-[SA]-X-[AY]-[ILRA]-[AL]-[ST]-[RGEN]-D-[FYWR]-[NQEF]-[PDN]-[VLG]-H-[ILH]-[PRK]-[ANDE]-X-A, which is located in the active site housing segments (Figure 6a; Supplementary Figure 5a, Supporting Information).

We used this motif to identify additional family members through individual BLAST queries with the protein sequences of ChsH1 and ChsH2 against the nonredundant protein sequence database. Then we examined their genomic environment to assess the prevalence of the heterotetrameric MaoC-like hydratase motif.[24] We found five different genomic contexts for ChsH1 and ChsH2 homologues (Figure 7a). Importantly, inclusion of genomic proximity in the search for homologues allowed us to identify three different motifs for encoding structural homologues of ChsH1 and ChsH2. All of these homologues are found in bacterial families known to catabolize steroids.[25]

Figure 7

Phylogenetic relationships of ChsH1-ChsH2. (a) Homologues of ChsH1-ChsH2 were identified in five different genomic contexts by BLAST. The arrow lengths are scaled to gene lengths. SHD (cyan), single hot-dog fold; DHD (cyan), double hot-dog; SHD+DUF35 (magenta), single hot-dog fold fused with DUF35/DUF35_N domain; ACAD (gray), acyl-CoA dehydrogenase. (b) Phylogenetic tree for five different genomic contexts of chsH1-chsH2 gene homologues. (1) Domain Opisthoknot; (2) Domain Bacteria; (3) Phylum Actinobacteria; (4) Phylum Proteobacteria; (5) Suborder Corynebacterineae; (6) Class Gammaproteobacteria. Representative organisms were selected for each category. The Roman numerals in parentheses after each organism correspond to the gene organization in panel a.

The two most common classes of homologues are single hot-dog fold and double hot-dog fold MaoC-like hydratases encoded without any neighboring MaoC-like enoyl-CoA hydratase genes. Typically, they form homodimers like AcRH or CtRH as described above. These are evolutionarily most distant from chsH1 and chsH2 and represent the previously identified MaoC-like hydratases. The third category contains two homologous genes in the same adjacent context as chsH1, a single hot-dog fold, and chsH2, a single hot-dog fold fused to a DUF35/DUF35_N domain, which is hypothesized to bind and deliver acyl-CoA moieties in acyl-CoA-utilization processes[26,27] (Supplementary Note). This arrangement is predominantly in Corynebacteria, which includes Mycobacteria (Figure 7b).

In less closely related Actinobacteria, the chsH1-chsH2 motif exists as a single fused gene. The sequence alignments suggest that these proteins will form a homodimer that is structurally similar to the ChsH1-ChsH2 heterotetramer, that is, a Rosetta fusion protein.[28,29] Whether these homologues catalyze the hydration of a steroid enoyl-CoA or a structurally simpler steroid metabolite remains to be determined. The fifth category is found in Proteobacteria and is comprised of operons that encode a chsH2 homologue separated from a chsH1 homologue by an acyl-CoA dehydrogenase (ACAD/FadE) homologue. Interestingly, in Agrobacterium tumefaciens strain C58, these three proteins are encoded as a single Rosetta fusion protein suggesting that the enoyl-CoA hydratase uses the acyl-CoA dehydrogenase product as a substrate, as is the case for the igr-encoded enzymes (Figure 1a,b). However, the organization of the igr operon is different; the acyl-CoA dehydrogenase genes (chsE1 and chsE2) are adjacent to chsH2 (Figure 1a). In addition, we have found no evidence for a tight association between the ChsH1-ChsH2 enzyme and the ChsE1-ChsE2 enzyme in our work.[15]

Within Mtb, 14 proteins are annotated as possessing hot-dogs folds and belong to the MaoC-like-hydratase family.[30−32] Analysis of their sequences suggests that they either assemble as single hot-dog folds or double hot-dog folds and form homodimers like AcRH and CtRH, respectively. We found no other Mtb homologues like ChsH1 and ChsH2, which would be predicted to form a dimer of heterodimers. This unique utilization of a heterotetrameric MaoC-like hydratase is in contrast to the presence of six cholesterol-regulated heterotetrameric acyl-CoA dehydrogenases in Mtb.[25] We conclude that in Mtb, this heterotetrameric enoyl-CoA hydratase is only employed in the final step of cholesterol side chain β-oxidation.

Conclusion

The enzymes encoded by the igr operon diverge structurally from the β-oxidation enzymes typically found in bacteria or eukaryotes. Oxidation of 3-OPC-CoA is catalyzed by a heterotetrameric acyl-CoA dehydrogenase (ChsE1-ChsE2) in place of the typical homotetrameric or homodimeric ACAD assembly. Hydration of the product 3-OPDC-CoA is catalyzed by an α2β2 heterotetrameric MaoC-like enoyl-CoA hydratase instead of the usual mitochondrial or bacterial crotonase family member or homodimeric peroxisomal multifunctional enzyme-2 (MFE-2) family member. The heterotetrameric architectures of these two Mtb enzymes appear to have evolved in order to accommodate the large steroid-CoA substrates. The structures of ChsH1-ChsH2 elucidated in this work provide insight into binding site cavities that differ substantially from its homologues and will allow discrimination between host and pathogen enzymes by inhibitors in the future.

Methods

Materials, Strains, Media, and General Methods

Ferricenium hexafluorophosphate was purchased from Sigma-Aldrich. Coenzyme A was purchased from MP Biomedicals. Isopropyl β-d-1-thiogalactopyranoside was from Denville Scientific. Tryptone, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), and tris(hydroxymethyl)aminomethane (TRIS) were purchased from Fisher Scientific. Kanamycin is from IBI Scientific. Yeast extract was purchased from Research Products International Co. iProof DNA polymerase was from Bio-Rad. Restriction endonucleases, T4 DNA ligase, T4 polynucleotide kinase, factor Xa, and protein ladder were from New England Biolabs. HisTrap FF columns and Superdex 200 HiLoad 16/60 and 10/300 GL columns were from GE Healthcare Biosciences Corp. Oligonucleotides were from IDT Inc. Total genomic DNA of M. tuberculosis H37Rv was obtained from the TB Research Materials Facility at Colorado State University (NIAD NO1-AI40091). MALDI mass spectra were acquired on a Bruker Autoflex II TOF/TOF. Big Dye DNA sequencing (Applied Biosystems; performed by the Stony Brook University Sequencing Facility) was used to verify the coding sequence of the expression plasmids. BL21(DE3) E. coli was obtained from BioRad. The 2× YT media is composed of 16 g of tryptone, 10 g of yeast extract, and 5 g of NaCl per liter. Buffer A consists of 20 mM Tris-HCl buffer, pH 8.0, supplemented with 300 mM NaCl and 10 mM imidazole. Buffer B consists of 20 mM Tris-HCl buffer, pH 8.0, supplemented with 300 mM NaCl and 500 mM imidazole. Buffer C consists of 50 mM Tris-HCl buffer, pH 8.0, supplemented with 200 mM NaCl. The casI (Ro05822) gene sequence was synthesized by GenScript USA Inc. Primary crystallization screens were from Hampton research.

Expression Plasmid Construction

ChsH1 (Rv3541c) and ChsH2 (Rv3542c) were amplified from M. tuberculosis H37Rv total genomic DNA by PCR using forward and reverse primers. The PCR product was digested with the appropriate restriction endonuclease and ligated into similarly digested pET28b (Table 1). DNA sequencing of the plasmids confirmed that the sequence was correct and that no mutations were introduced during PCR. Active site mutants were prepared using Quick Change site-directed mutagenesis.[33] The mutations were confirmed by DNA sequencing.

To produce the C-terminal truncation of ChsH2, a factor Xa cleavage site, followed by four glycine residues, was introduced at Arg187 of ChsH2. Insertion was conducted by PCR with primers 5′-ATC GAC GGG CGT GGT GGT GGT GGT CGT CCC TCG TCG TCG CGG GAC-3′ and 5′-CAT CAT AGC GTC GGG ATC CAA ATC GTC AGG-3′ to give construct pChsH1ChsH2.

Protein Expression and Purification

To express ChsH1 or ChsH2 individually, construct pChsH1 or pChsH2 was transformed into BL21(DE3) E. coli. Single colonies were selected on LB plates containing 30 μg/mL kanamycin and cultured in 2× YT media at 37 °C. Expression was induced at OD600 ≈ 0.6–0.8 by the addition of 50 μM to 1 mM IPTG, and cells were grown 20 h at 16–25 °C. Purified proteins were analyzed by reducing SDS-PAGE. CasI was expressed as described for ChsH1 or ChsH2 using construct pCasI and 100 μg/mL ampicillin in the media.

ChsH1-ChsH2 Complex Was Obtained Using Construct pigr3(12)

Expression was induced at OD600 ≈ 0.6–0.8 by the addition of 1 mM IPTG, and cells were grown 20 h at 25 °C. Cells were lysed by French press or sonication in buffer A, and cellular debris was removed by centrifugation at 125 000g for 1 h. ChsH1-ChsH2 was purified by IMAC, with HisTrap FF column and buffers A and B. Protein was further purified by size exclusion chromatography on a Superdex 200 HiLoad 16/60 column equilibrated with buffer C. Active site mutants were expressed and purified following the method for pigr3 using constructs pigr3H1 and pigr3H1.

ChsH1-ChsH2N was expressed as described for pigr3 and isolated by IMAC. ChsH2 was cleaved by factorXa at 23 °C in buffer C supplemented with 2 mM CaCl2. Factor Xa protease reaction was monitored by SDS-PAGE, and cleaved ChsH1-ChsH2N was purified by size exclusion chromatography on a Superdex 200 HiLoad 16/60 column equilibrated with buffer C.

Solution-State Biophysical Analysis of ChsH1-ChsH2 and ChsH1-ChsH2N

ChsH1-ChsH2 (10 mg mL–1) and ChsH1-ChsH2N (7 mg mL–1) were analyzed by analytical gel-filtration on a Superdex 200 (10/300 GL) column (GE Healthcare). The column was equilibrated in buffer C. Samples were eluted isocratically in buffer C, monitoring at 220 and 280 nm. Several standard proteins were analyzed under the same conditions to generate standard curves to estimate molecular weights of analyzed proteins.

Molecular weights were determined using analytical ultracentrifugation sedimentation equilibrium (Beckman Optima XL-A). ChsH1-ChsH2 (5.7, 2.8, and 1.4 μM) and ChsH1-ChsH2N (7.3, 3.7, and 1.8 μM) were centrifuged at speeds of 10000, 20000, and 30000 rpm at 20 °C. Scans were acquired after 18 and 20 h of centrifugation at each speed monitoring at 280 nm. The protein partial-specific volume of 0.7359 for ChsH1-ChsH2 and 0.7379 for ChsH1-ChsH2N and a solvent density 1.0079 for buffer C were calculated using SEDNTERP. Data were fit globally to the ideal, single species model using Heteroanalysis to determine the molecular weight.

Protein complex stoichiometries of ChsH1-ChsH2 and ChsH1-ChsH2N were confirmed by LC/UV/MS. Samples were separated on a XBridge BEH 300 C4 3.5 μm column (2.1 mm × 100 mm) at 40 °C with a linear gradient from 95% A to 95% B over 15 min, where A is 5% isopropanol/0.1% trifluoroacetic acid and B is 99.9% isopropanol/0.1% trifluoroacetic acid. MS spectra were collected in ESI positive ion mode with a cone voltage of 40 V, a capillary voltage of 4.5 kV, and source temperature of 150 °C. MS spectra were deconvoluted using ESIprot 1.0,[34] and peaks in the UV 280 nm chromatograms were integrated using R. The integrated peak areas of each protein were divided by the corresponding molar extinction coefficient for the protein to yield the molar concentrations. Protein stoichiometries were determined from the ratio of the molar concentrations.

Synthesis of Enoyl-CoAs

trans-2-Decenoyl-CoA and trans-2-octenoyl-CoA were synthesized using the mixed anhydride method.[35] Briefly, the mixed anhydride was prepared by mixing the corresponding acid (0.2 mmol) with ethyl chloroformate (0.4 mmol) in 4 mL of dry THF in the presence of TEA (56 μL). After approximately 25 min, the acid was completely converted into the corresponding anhydride according to TLC. The mixed anhydride was filtered though glass wool in a disposable Pasteur pipet into newly prepared lithium CoA solution (55 mg) dissolved in 5 mL of H2O and THF (3:2 v/v) (pH, 8). The reaction was stirred at rt for 2 days. The pH of the reaction mixture was adjusted to 3, and the unreacted acid was removed by extraction with ether. The enoyl-CoAs were purified by HPLC using 20 mM ammonium bicarbonate, pH 8, with a linear gradient from 0 to 90% methanol in 20 mM ammonium bicarbonate.

3-Oxo-4-pregnadiene-20-carboxylic acid

3-Oxo-4-pregnadiene-20-carboxylic acid was synthesized following a previously reported method starting from stigmasta-4,22-diene-3-one.[12,36]

3-Oxo-4-pregnene-20-carboxyl CoA, 3-OPC-CoA

3-Oxo-4-pregnene-20-carboxyl CoA was synthesized following the previously reported CoA synthetase assay. 3-Oxo-4-pregnene-20-carboxylic acid (3-OPC) was dissolved to 50 mM in 94% ethanol containing 60 mM NaOH. The thioesterification was performed for 4 h at 22 °C in a 1 mL reaction volume containing 5 mM 3-OPC, 10 μM CasI (Ro05822), 100 mM HEPES (pH 7.4), 5 mM MgCl2, 2.5 mM ATP, and 1 mM CoASH.[37,38] The reaction was quenched by adding 1 mL of MeOH to precipitate the enzyme. The supernatant was isolated by centrifugation and filtration. 3-Oxo-4-pregnadiene-20-carboxyl CoA was purified by HPLC using 100 mM ammonium acetate, pH 4.5, with a linear gradient from 0 to 90% MeOH in 100 mM ammonium acetate to yield 3-OPC-CoA.

3-Oxo-pregna-4,17-diene-20-carboxyl-CoA, 3-OPDC-CoA

3-Oxo-4-pregnene-20-carboxyl CoA was redissolved to 2.8 mM in H2O. The oxidation was performed for 3 h at 20 °C in 1 mL of 0.5 mM 3-OPC-CoA, 3 μM ChsE1-ChsE2 (FadE28-FadE29), 100 mM N-tris[hydroxymethyl]methyl-3-aminopropanesulfonic acid (TAPS) (pH 8.5), and 1 mM ferricenium hexafluorophosphate.[15] The product was purified by HPLC in 100 mM ammonium acetate, pH 4.5, with a linear gradient of 90% MeOH. MeOH was removed under reduced pressure, and H2O was removed by lyophilization to yield 3-OPDC-CoA.

Hydratase Assay

ChsH1-ChsH2, ChsH1-ChsH2N, ChsH1D29A-ChsH2, and ChsH1H34A-ChsH2 were assayed for hydratase activity with substrates octenoyl-CoA, decenoyl-CoA, and 3-oxo-pregna-4,17-diene-20-carboxyl CoA in 100 mM HEPES, pH 7.4, buffer and 100 nM (50 nM for 3-oxo-pregna-4,17-diene-20-carboxyl CoA) enzyme at 25 °C. Reactions were monitored at 263 nm, and initial velocities were determined. Product formation was quantified using ε263nm = 6700 M–1 cm–1, which corresponds to the α,β-unsaturation of the enoyl-CoA substrates. Hydration product formation was confirmed by analysis of the assay mixture by MALDI-TOF mass spectrometry.

Crystallization

The ChsH1-ChsH2N apoenzyme crystals were obtained by hanging drop vapor diffusion at rt. Briefly, 1 μL of 15 mg mL–1 protein was mixed 1:1 with a reservoir solution of 20%–24% PEG 3350, 20 mM CaCl2, 20 mM CdCl2, and 20 mM CoCl2 at pH 7 or pH 6.5 and equilibrated against 500 μL of the reservoir solution. The crystals were then harvested and transferred to a cryoprotectant solution containing the mother liquor and 17% glycerol. For co-crystallization, ChsH1-ChsH2N was mixed with 3-OPC-CoA. Crystals were obtained at 4 °C by hanging drop vapor diffusion, in mother liquor containing 1 mg mL–1 3-OPC-CoA, 20 mM CaCl2, 20 mM CdCl2, 20 mM CoCl2, and 200 mM NaCl at pH 7.0 and 25% PEG 3350. The crystals were harvested and transferred to a cryoprotectant solution containing the mother liquor supplemented with 1 mg mL–1 3-OPC-CoA and 17% ethylene glycol. All crystals were flash cooled in liquid N2 before data collection.

X-ray Data Collection and Structure Determination

Diffraction data for ChsH1-ChsH2N were collected on beamline X25 of the National Synchrotron Light Source at Brookhaven National Laboratory (Upton, NY) at a wavelength of 1.7 Å, and for ChsH1-ChsH2N/3-OPC-CoA, data were collected on beamline 23ID-D of the Advanced Photon Source at Argonne National Laboratory (Argonne, IL) at a wavelength of 1.075 Å. All data sets were processed using XDS[39] and Aimless[40] as implemented in autoPROC.[41] Four cadmium sites in ChsH1-ChsH2N were located with SHELX,[42] phases were calculated to 1.54 Å via single-wavelength anomalous dispersion[43] using SHARP,[44] and automated modeling building with ARP/wARP[45] produced a starting model consisting of 85% of ChsH1 and 90% of ChsH2N. For the ChsH1-ChsH2N/3-OPC-CoA structure, phases were calculated via molecular replacement with Phaser, using ChsH1-ChsH2N as a search model. Six cadmium sites were identified in the ChsH1-ChsH2N/3OPC-CoA structure; four of them are located in the tetramer–tetramer packing interface, and two of them coordinate with Asp29/His34 in ChsH2N, blocking the active site (Supplementary Figure 5b and Note, Supporting Information). For both structures, manual model building was carried out in Coot[46] followed by refinement with Refmac[47] and Phenix.[48] Final model quality was assessed using MolProbity.[49] Data collection and refinement statistics are shown in Table 2.