Structural Evidence for the Dopamine-First Mechanism of Norcoclaurine Synthase

Norcoclaurine synthase (NCS) is a Pictet-Spenglerase that catalyzes the first key step in plant benzylisoquinoline alkaloid metabolism, a compound family that includes bioactive natural products such as morphine. The enzyme has also shown great potential as a biocatalyst for the formation of chiral isoquinolines. Here we present new high-resolution X-ray crystallography data describing Thalictrum flavum NCS bound to a mechanism-inspired ligand. The structure supports two key features of the NCS “dopamine-first” mechanism: the binding of dopamine catechol to Lys-122 and the position of the carbonyl substrate binding site at the active site entrance. The catalytically vital residue Glu-110 occupies a previously unobserved ligand-bound conformation that may be catalytically significant. The potential roles of inhibitory binding and alternative amino acid conformations in the mechanism have also been revealed. This work significantly advances our understanding of the NCS mechanism and will aid future efforts to engineer the substrate scope and catalytic properties of this useful biocatalyst.

. X-ray data collection and refinement statistics.

Experimental Procedures
Protein purification and expression A construct containing a codon optimised, truncated Thalictrum flavum NCS 1 gene (ΔN33C196TfNCS), with an N-terminal hexahistidine tag and a TEV protease cleavage site was synthesised and cloned into pD451-SR (ATUM, CA, USA) 2 . The plasmid was transformed into BL21 (DE3) cells and a single colony inoculated 100 ml of Terrific broth media (TB) for 16 hours. One litre of TB was inoculated with 4% v/v of overnight culture and grown for 2 hours at 37 °C, then 1 hour at 25 °C. The protein was overexpressed by addition of 0.5 mM isopropylthiogalactoside, incubated for 3 hours at 25 °C and then harvested by centrifugation.
Cell pellets were suspended in binding buffer (50 mM Hepes, 100 mM NaCl, 20 mM Imidazole pH 7.5) and 10% v/v BugBuster 10X (Merck Millipore, Germany) was used to break the cells. After centrifugation at 25,000 g for 1 hour, the lysate was loaded onto 1 ml of Ni-Sepharose HP resin (GE Healthcare). The protein was eluted from the resin with elution buffer (50 mM Hepes, 100 mM NaCl, 500 mM imidazole, pH 7.5) after washing with binding buffer and washing buffer (50 mM Hepes, 100 mM NaCl, 50 mM Imidazole pH 7.5) for 5 column volumes respectively. The eluted fractions were pooled and 0.1 mg of TEV protease (containing a N-terminal His-tag) was added to the sample and dialysed in 4 litres of dialysis buffer (20 mM Tris, 50 mM NaCl, pH 7.5) for 16 hours at 4 °C.
The sample was loaded to a 1 ml of Ni-Sepharose HP resin to bind uncut NCS and TEV protease. Cut NCS was washed off the resin with wash buffer (20 mM Tris, 50 mM NaCl, 50 mM Imidazole, pH 7.5). Size exclusion chromatography was used to purify the NCS protein further using Superdex 75 16/600 column (GE, Healthcare). The eluents were pooled and concentrated using a 10 kDa cut off Vivaspin concentrator (Sartorius, Germany) to 12 mg/ml. The protein sample was either used directly to set up crystallization trials or stored at -80 °C.

Protein crystallisation and data processing
The truncated NCS apo protein crystals were grown by the sitting-drop method in 96-well crystallisation plates (Molecular Dimensions) in 10% w/v polyethylene glycol (PEG) 1000 and 10% w/v PEG 8000. Larger crystals were obtained by hanging-drop method. The protein was incubated with 10 mM of mimic compound 6 and crystallised in the same condition as the apo protein. The crystals were cryo-protected in crystallisation buffer containing 20% ethylene glycol. Diffraction data for the apo structure were collected at Soleil beamline Proxima 1 whereas the final mimic-bound dataset was collected at Diamond beamline I02. The diffraction images were processed using xia2 and XDS 3 software packages, scaled and merged using Aimless in the CCP4 program suite 4 . The initial phases of the apo NCS models were solved by molecular replacement with the program Phaser 5 using the previous apo NCS structure (PDB: 2VNE 6 ) as the search model. Model building was performed with COOT 7 and refinement was done with Refmac5 8 using TLS (one group per chain including the associated water molecules and ligand) and local noncrystallographic symmetry restraints. The positions of both aromatic rings of the mimic was clear in all three copies in the asymmetric unit of the mimic-bound structure from initial difference maps. There was a ring like density next to the dopamine ring despite there being no ring closure in the mimic. When the mimic is placed in the conformation proposed to be productive in the reaction mechanism refinement gave strong (>5 sigma) difference density where a 6 th atom could make up a second ring. Conversely if the nonproductive conformation where the dopamine is flipped and the rest of the molecule comes off the other side of the ring, is refined alone there is even stronger difference density where the C9 atom is in the first conformation. Neither of these positions correspond to water molecules in the apo structure probably ruling S4 out a mixed apo/ligand structure 9 . We propose that the structure is a mixture of these productive and unproductive ligand conformations.
Alternative ligands were tried: neither a five-membered ring oxidation product nor the product of the typical enzyme reaction gave plausible fits ruling out any structure with the R group coming from an atom adjacent to the dopamine. A tertiary amine fills the density but gives poorer R factors than the two-conformation fit, and such a compound is also chemically implausible in the conditions used. Placing a water in the difference density gives a lower Rfree than the two-conformation model and no difference density. However, the water is too close to the Nitrogen (1.6 Å) and the ring (1.8 Å) and is only on the very edge of the density. The final deposited model used the 'complete' occupancy refinement in Refmac5 such that the combined occupancy of the two ligands are constrained to 1.0 in each copy. This final occupancy is not particularly stable and depends on slight drift apart of the two ligands during refinement. The unproductive conformation often ends up with a lower occupancy and a higher B factor and can drift to very low occupancy and high B factor and move quite far out of the density resulting in a return of the difference density peak. Conversely more even occupancy results when the dopamine ring of the unproductive conformation moves away from the optimum individual fit to the ring allowing the ring linking atoms in minor conformation to be closer to the position of the major conformation. This results in less difference density and has been deposited. Other refinement packages did not give better results for the two ligand model in our hands.
Data collection and refinement statistics are summarized in Table S1. Figures and RMSD comparisons were performed using UCSF-Chimera (http://www.rbvi.ucsf.edu/chimera/) except the electron densities which were drawn with ccp4mg.

Computational docking
Subunit A of mimic-bound structure 5NON was used for docking experiments. Ligands and water molecules were removed. Ligands were MM2 energy minimised in ChemBio3D before docking with Chimera UCSF, using the AutoDock Vina plug-in 10 . The protein molecule was centred, and the docking box was position (-17.95, -7.43, 16.19) and size (18.14, 19.02, 28.24). The software was run with the settings: energy-range 3, exhaustiveness 8 and number of modes 10. Binding modes relevant to the dopamine-first mechanism were selected (see Table S2).

Enzyme assays
The time-courses of ΔN33C196TfNCS and Δ29TfNCS ( Figure S2) was conducted in triplicate. Each assay contained 10 mM dopamine, 10 mM hexanal, 10% v/v MeCN, 0.1 mg/mL purified enzyme, 5 mM sodium ascorbate and 50 mM Hepes pH 7.5. Samples were quenched with 100 mM HCl, diluted and analysed by HPLC. Enzyme activities (initial rates) for Δ29TfNCS and Δ29TfNCS-A79I ( Figure S3) were conducted in triplicate as previously reported 11 . Reactions contained dopamine (10 mM) and 4-HPAA or hexanal (2.5 mM) and were quenched after 30 seconds and analysed by HPLC. HPLC analyses were performed on a HPLC system consisting of an LC Packing FAMOS Autosampler, a P680 HPLC Pump, a TCC-100 Column oven and a UVD170U Ultraviolet detector (Dionex, Sunnyvale, CA, USA), and a C18 (150 x 4.6 mm) column (ACE, Aberdeen, UK). Samples were run with a gradient of H2O (0.1% trifluoroacetic acid) /MeCN from 9:1 to 3:7 over 6 min, at a flow rate of 1 mL.min -1 . The column temperature was 30 °C, and compounds were detection by monitoring A280. Retention times and concentrations were calculated based on chemically verified standards.

Synthesis of 4-{2-[(4-Methoxyphenethyl)amino]ethyl}benzene-1,2-diol
General. All chemicals were obtained from commercial suppliers and used as received unless otherwise stated. Thin layer chromatography (TLC) analysis was performed on Merck Kieselgel precoated aluminium-backed silica gel plates and compounds visualised by exposure to UV light, potassium permanganate or ninhydrin stains. Flash column chromatography was carried out using silica gel (particle size 40-60 µm). NMR: 1 H and 13 C NMR spectra were recorded at 298 K at the field indicated using Bruker Avance 300 and Brucker Avance 400 III spectrometers. Coupling constants (J) are measured in Hertz (Hz) and multiplicities for 1 H NMR couplings are shown as s (singlet), d (doublet), t (triplet) q (quartet) and m (multiplet). Chemical shifts (in ppm) are given relative to tetramethylsilane and referenced to residual protonated solvent. Mass spectrometry analyses were performed at the UCL Chemistry Mass Spectrometry Facility using a Finnigan MAT 900 XP and Waters LCT Premier XE ESI Q-TOF mass spectrometers. 3,4-Bis(benzoyloxy)dopamine 7 was synthesized as previously reported 12 .
Preparative HPLC conditions: Varian Prostar instrument with a UV-visible detector (monitoring at 280 nm) and a DiscoveryBIO wide Pore C18-10 Supelco column (25 Å~ 2.12 cm). A gradient of 5% to 90% of acetonitrile/water (0.1% TFA)) was used. See Figure S8 for NMR spectra. Tables   Table S1. X-ray data collection and refinement statistics.     H. Original density after one round of refinement of apo structure (including waters) direct with Refmac. The two data sets were isomorphous enough to obviate a molecular replacement step. 2Fo-Fc maps in blue at 1 sigma. Fo-Fc at +3 sigma (green) and -3 sigma (red). All maps clipped to the double mimic coordinates at 1.5 A (Fo-Fc) and 2 A difference maps. Drawn with CCP4mg.  Curly arrows represent electron movement, block arrows represent physical movement of residues/water.