The Crystal Structure of AbsH3: a Putative FAD-dependent Reductase in the Abyssomicin Biosynthesis Pathway.

Abstract

Natural products and natural product-derived compounds have been widely used for pharmaceuticals for many years, and the search for new natural products that may have interesting activity is on going. Abyssomicins are natural product molecules that have antibiotic activity via inhibition of the folate synthesis pathway in microbiota. These compounds also appear to undergo a required [4+2] cycloaddition in their biosynthetic pathway. Here we report the structure of an FAD-dependent reductase, AbsH3, from the biosynthetic gene cluster of novel abyssomicins found in Streptomyces sp. LC-6-2.

Introduction:

Abyssomicins are novel microbial natural products produced by Verrucosispora and Streptomyces species (1-15). Abyssomicin C is a polycyclic polyketide-type antibiotic that was initially detected in a screen for inhibition of p-aminobenzoate (pABA) synthesis, which is a key cofactor in the tetrahydrofolate (THF) biosynthesis pathway. Unique to microorganisms, the THF pathway is an attractive drug target for antibiotics (16). Abyssomicin C is the first natural product inhibitor of this pathway, and it inhibits the aminodeoxychorismate synthase enzyme (ADCS), which converts chorismate and glutamine into 4-amino-4-deoxychorismate and glutamate. Other inhibitors of this pathway include sulfonamides and trimethoprim, which inhibit much further along the pathway at dihydropteroate synthase (sulfonamides) or dihydrofolate reductase (trimethoprim) (17-19). Abyssomicin C inhibits ADCS via covalent modification of the ADCS (20). In this reaction, abyssomicin C functions as a Michael acceptor, with the Cys263 sulfur acting as the nucleophile. After this reaction abyssomicin C is thought to rearrange into an abyssomicin D derivative, thus irreversibly inhibiting ADCS. Other interesting activities have been ascribed to abyssomicin compounds separate from their ability to inhibit the folate pathway in gram-positive bacteria, such as antitumor properties and activation of latent HIV replication (10, 21).

As exemplified by abyssomicin, versipelostatin and pyrroindomycin biosynthetic studies (22-24), novel Diels-Alderase-catalyzed stereoselective [4 + 2] cycloaddition reactions are central to forming the cyclic abyssomicin core architecture. Recent structural and mechanistic studies of AbyA5 (25), an esterase implicated in the formation of the key Diels-Alderase substrate, and biochemical characterization of AbmV (26), a P450 oxidase central to bridged ether formation, further extend understanding of abyssomicin core construction. However, biochemical and/or structural studies of the tailoring enzymes involved in abyssomicin core scaffold maturation remain sparse. Here we report the crystal structure of AbsH3 (GenBank Accession number: ARE67860, Uniprot: A0A1V0QH64), a putative FAD-dependent reductase encoded by the abyssomicin biosynthetic gene cluster found in the coal mine fire isolate Streptomyces sp. LC-6-2 (GenBank accession number KY432814) (14). AbsH3 is homologous to a number of mono-oxygenases, including TetX2 (PDB ID: 3P9U, 3V3N) and PhzS (PDB ID: 2RGJ) (30% and 33% identity, respectively) (27, 28). Unique Streptomyces sp. LC-6-2 abyssomicin metabolic features to which enzyme-catalyzed oxidation may contribute include the production of novel “enantiomeric”-like abyssomicin metabolites (abyssomicins M–X), unprecedented core cyclization patterns (e.g., the abyssomicin W 8/6/6/6 tetracyclic core) and linear spirotetronates (e.g., abyssomicin X).

Materials and Methods:

Cloning, Expression, and Purification.

Genomic DNA of Streptomyces sp. LC-6-2 was extracted using GenElute Bacterial Genomic DNA Kit (Sigma-Aldrich) and stored in Tris-EDTA buffer. The gene absH3 was amplified by PCR using Phusion Hot Start II DNA Polymerase (Thermo Scientific) using genomic DNA as template. The DNA fragments of the expected sizes were separated by agarose electrophoresis and purified by gel extraction kit (Zymo Research) and annealed with T4 DNA ligase into pET30 vector (Novagen) as previously described (29). Cloned genes were verified by DNA sequencing. The primers used for cloning were as follows: absH3-pET30-fw 5’-GGTATTGAGGGTCGCATGAACACGACCGAC-3’/ absH3-pET30-rv 5’-AGAGGAGAGTTAGAGCCTCAGCTCGCCGCTGT-3’.

E. coli BL21 DE3 was transformed with pET30-absH3 and seed culture was started from a single colony in LB medium supplemented with 50 μg/mL kanamycin. A sample of 2 mL seed culture was transferred to 1 L of LB medium supplemented with 50 μg/mL kanamycin. Cultures were grown at 37 °C until the cell density reached an OD600 of 0.4–0.6, when protein expression was induced with 0.1 mM IPTG. After growing at 18°C for 12 hours, bacterial cells were harvested at 4,000 rpm and lysed by sonication. The resulting cell debris was removed by centrifugation at 21,000 rpm. Supernatant was loaded on HisPur Ni-NTA resin (ThermoFisher) for affinity purification. The resin was first washed with wash buffer (25 mM HEPES, 30 mM imidazole, 200 mM NaCl, 5 mM β-mercapto ethanol (BME), pH 7.5). Proteins were eluted from the resin using elution buffer (25 mM HEPES, 300 mM imidazole, 200 mM NaCl, 5 mM BME, pH 7.5). Fractions containing desired protein (visualized on an SDS-PAGE gel) were concentrated. The imidazole-containing buffer in the concentrated protein was removed using a desalting column (GE PD-10) and exchanged with storage buffer (25 mM HEPES, 200 mM NaCl, pH 7.5). The purity of protein was determined by SDS-PAGE. Purified proteins were promptly flash frozen in sealed microcentrifuge tubes using liquid N2.

Crystallization and Structure Determination.

A Mosquito crystallization robot was used for initial screening of AbsH3 with the Hampton® crystal screens HR2-086, HR2-130, and HR2-134. Screens were conducted via the sitting drop method using 0.2 μL of protein solution and 0.1 μL of screening solution or, alternatively, 0.1 μL of protein and 0.2 μL of reservoir. Starting protein solution concentration was 10mg/mL. Initial crystals of AbsH3 observed in HR2-134 well D8: 100 mM HEPES pH 7.5 and 25% w/v PEG 3350. The final crystal condition after optimization was 22% PEG3350, 20 mM magnesium chloride, 100 mM HEPES, pH 7.4, with drops composed of of 0.2 μL protein and 0.2 μL reservoir solution.

X-Ray diffraction data were collected at the LRL-CAT beam line at the Advanced Photon Source at Argonne National Laboratory at a wavelength of 0.9793 A. Diffraction patterns were processed via De-ice in order to remove ice rings prior to indexing (30). De-Ice is a program that removes ice rings via modeling the ice and subtracting the average intensity working out from the beam stop radially on the raw diffraction image. By subtracting the radial average of intensity in the ice ring, ice ring diffraction can be at least partially accounted for without losing Bragg peaks at the resolution of the ice rings. Indexing and scaling was done in XDS (31). Molecular replacement was used to solve the structure, using PDB ID 3v3n (32) split via SCEDS (33) into multiple search models in Phaser (34). Real-space refinement was carried out manually in Coot (35) with final refinement via Phenix.refine (36). Analysis of the structure was aided using a 3D semi-immersive collaborative graphics system (37).

Docking Experiments.

The X-ray crystal structure was used to explore how each abyssomicin analogue may interact with the protein in any possible binding modes. For this purpose, each ligand molecule was docked into the protein by using the Autdock Vina program (38). The molecular docking was carried out by using the default optimization parameters of the Autdock Vina program for the pose sampling, with a large grid map (74 Å × 80 Å × 64 Å) to cover the entire protein. The docking score was calculated by using the Lamarckian algorithm (39). Finally, the docked protein-ligand binding structures were energy-minimized by using the Amber12 program (40).

Results and Discussion:

Diffraction data was collected at LRL-CAT beamline at the Advanced Photon Source. The highest resolution (1.99 Å) data set exhibited ice rings and attempts to recollect data with cryo-protection (20% ethylene glycol, 10-20% glycerol, saturated sucrose solution) instead of looping from reservoir solution resulted in lessened diffracting power. Therefore, as a workaround to preserve resolution, data were preprocessed with De-Ice (30).. Since the structure of AbsH3 was solved, a new global background model has been developed in DIALS for better estimation of Bragg peak intensities within ice ring data (41). Using this pipeline yields similar statistical results to De-Iced data and requires less intervention on the part of the crystallographer.

The structure was solved with molecular replacement in Phaser, utilizing PDB ID 3V3N for the search model split into two “domains” with the SCEDS program. The SCEDS program separated the TetX2 structure 3v3n into two domains which involved multiple cut sites. Fragment A contained residues 14-54,70-79,122-189,287-317, and 330-366. Fragment B contained residues 55-69, 80-121, 190-246, 249-286,318-329, and 367-382. The crystal structure of AbsH3 was solved to 1.99 Å resolution in space group P2₁2₁2₁. The final refined structure had R-work and R-free scores of 0.183 and 0.236 respectively. Table 1 provides diffraction and modeling statistics. Data collection and merging statistics are for data after Deice processing.

Table 1.

Data collection and refinement statistics.

PDB ID	6N04
Wavelength	0.9793
Resolution range	35.62 - 1.998 (2.07 - 1.998)
Space group	P 21 21 21
Unit cell	53.316 109.491 143.637 90 90 90
Total reflections	379890 (37404)
Unique reflections	53291 (5578)
Multiplicity	7.1 (6.7)
Completeness (%)	91.96 (97.22)
Mean I/sigma(I)	12.06 (1.97)
Wilson B-factor	29.53
R-merge	0.1237 (1.015)
R-meas	0.1336 (1.101)
R-pim	0.04993 (0.4195)
CC1/2	0.997 (0.71)
CC*	0.999 (0.911)
Reflections used in refinement	53221 (5560)
Reflections used for R-free	2023 (211)
R-work	0.1828 (0.2713)
R-free	0.2365 (0.2986)
CC(work)	0.964 (0.835)
CC(free)	0.945 (0.731)
Number of non-hydrogen atoms	6573
macromolecules	5559
ligands	108
solvent	906
Protein residues	731
RMS(bonds)	0.005
RMS(angles)	0.77
Ramachandran favored (%)	98.19
Ramachandran allowed (%)	1.81
Ramachandran outliers (%)	0.00
Rotamer outliers (%)	1.46
Clashscore	3.64
Average B-factor	34.84
macromolecules	33.65
ligands	25.70
solvent	43.21

Even though AbsH3 behaves as a monomer in solution, there are two molecules in the asymmetric unit, which have an interesting interaction. The two molecules are aligned in an anti-parallel fashion where the lid domain β-sheets hydrogen bond to each other at the interface of the two monomers, maintaining the secondary structure motif. This feature is accomplished by aligning the lid domains of both monomers, as shown in Figure 1. In addition, in this packing arrangement an AbsH3 monomers obstructs their neighbor’s active site with a C-terminal tail that maintains helical secondary structure. This capping of the active site appears to have stabilized the apo structure and locked the lid domain into place. This hints at possible self-inactivation at very high protein concentration, but it is unclear at this time if this phenomenon is in any way biologically relevant.

Figure 1.

Crystal Structure of AbsH3 (PDB ID: 6N04) A. Comparison of frames before and after De-Ice processing. On left, frame as collected at LRL-CAT, compared to after processing, right. Inset shows entire frame from which zoomed in imaged is pulled. B. Crystal packing of AbsH3. The crystal packing in the structure shows the C-terminal tail of the purple monomer plunged into the active site of the other aqua AbsH3 monomer. C. FAD density and polar contacts in the core domain of AbsH3. 2Fo-Fc map of FAD at 2.0σ level showing high quality density of this co-factor. Polar contacts with sidechains are also shown. D. C-terminal tail occludes the active site of adjacent monomer in the crystal and may be a self-inactivation function at high AbsH3 concentration.

Similar oxidoreductases to AbsH3 are reliant on FAD as a prosthetic group to supply electrons for their chemical processes, and AbsH3 is no exception. In our structure, FAD is clearly bound, and, in fact, has lower average B-factors than the overall structure. It sits down in the core domain, with the electron-donating ring system extending into the bottom of the active site. This arrangement is similar to homologous reductases including TetX2, which hold their respective FAD molecules in similar fashions (27). PhzS, even though it has a similar overall fold and utilizes a FAD cofactor for oxidoreductase activity, appears to have different substrate binding morphology due to different FAD prosthetic group orientation in the active site, as well as a more open lid domain in AbsH3 that should be accessible from both sides.

In order to better understand possible substrate binding modes and help determine a likely substrate, docking studies with novel abyssomicin compounds discovered by Wang et al. (14) from the Streptomyces strain that AbsH3 was cloned from were performed. These docking studies showed a number of potentially important residues for binding in the active site. See Supplemental Table 1 and Figure 2 for docking energy table and structures, respectively, for representative docked structures, see Supplemental Figure 1. In site A, the putative active site, Arg89, Tyr162, Gln164, and Arg206 potentially use electrostatic interactions or hydrogen bonds to guide the abyssomicin analogues toward the quinone of the FAD cofactor. The rest of the binding pocket stabilizes the large multi-cyclic ring system via hydrophobic interactions. Leu114, Leu241, Met239, and Leu247 appear to be especially important for stabilizing the compounds in the hydrophobic pocket. Tyr162, Met187, and Arg206 seem to gate the compounds into this pocket and lock them in orientations that allow exposed oxygen moieties to be reduced by the FAD cofactor. An additional binding site, site B, in the FAD binding groove allows for the possibility of diffusion across the surface of the AbsH3 molecule into the active site. Interestingly, compounds 8-12 show tighter binding in this second site rather than the putative active site. Future work unraveling this Abyssomycin biosynthetic pathway will yield new insights into how this fascinating natural product is produced.

Figure 2.

Binding sites from docking studies and abyssomicin analogs used in the docking calculations. At left, FAD model showing the two sites where abyssomicin analogues bound in the docking study. Site A is the putative active site, while Site B is in the FAD binding groove. The presence of this second binding site may allow substrates to diffuse across the enzyme surface to the active site, even when not recruited directly to Site A. At right, chemical structures of the abyssomicin analogues used in the docking computations. Interestingly, compounds 8-12 preferred Site B to Site A, which is the active site.