Isolation and purification of celery and rose RG-II
Celery petioles were purchased from a local supermarket. Suspension-cultured ‘Paul's Scarlet’ rose (Rosa sp.) cells were a kind gift from Dr. Stephen C. Fry (The University of Edinburgh)36. Rosa cells were grown at 23°C under low light in a basal salt liquid medium36, containing glucose (2% w/v, 70% 12C glucose and 30% 13C glucose) on an orbital shaker. Cell walls were prepared from celery petioles and Rosa cells as their alcohol insoluble residues (AIR) as described.20 Starch was removed by treating suspensions of the AIR in 50 mM sodium acetate, pH 5.2, for 24 h at 45°C with the glucoamylase Spirizyme® (60 µL/g; Novozymes A/S, Denmark) and the α-amylase Liquozyme® (300 µL/g; Novozymes A/S, Denmark). The de-starched AIR was collected by filtration through nylon mesh (100 µm pore size) and washed with deionized water. The AIR (5–25 g) was then suspended in 50 mM NaOAc, pH 5.2 (1g/100 mL) and treated for 24 h at 30°C with endo-polygalacturonanase M2 (EPG; 5 U/g) from Aspergillus aculeatus (Megazyme, Ireland). The suspension was shaken at 250 rpm. The EPG-treated AIR was collected by filtration through nylon mesh (100 µm pore size) and then retreated with EPG. The combined EPG-solubilized fractions were concentrated by rotary evaporation at 37°C, dialyzed (Spectrum™ Spectra/Por™, 3500 Dalton MWCO) against deionized water, and freeze dried.
Solutions of the EPG-soluble material from celery AIR (500–700 mg) in 50 mM NaOAc, pH 5.2 (~ 10 mL), were separated using a Sephadex-75 column (1000 x 3.4 cm) eluted at 1 mL/min with 50 mM NaOAc, pH 5.2 20. Fractions (10 mL) were collected, and a portion (100 µL) assayed colorimetrically for uronic acid.44 Pooled fractions were dialyzed and freeze dried. The fraction containing RG-II was then chromatographed on a column (20 x 2.0 cm) of fast-flow DEAE-Sepharose (Cytiva, USA) to remove a contaminating galactose-rich polymer as described 20. Solutions of the EPG-soluble material from rose AIR (5–10 mg) in 50 mM ammonium formate, pH 5, were separated using a Superdex-75 Increase column (30 x 1 cm, Cytiva, USA) eluted at 0.5 mL/min with 50 mM ammonium formate, pH 5, with refractive index detection (Avci et al., 2018).45 Fractions containing the RG-II dimer were collected manually and repeatedly freeze dried to remove the ammonium formate.
Preparation of the RG-II monomer
Solutions of the celery and rose RG-II (~ 95% dimer) in 0.1 M HCl were kept for 1h at room temperature to generate RG-II monomer.20 The solution was immediately dialyzed (3500 Dalton MWCO) against deionized water and freeze-dried. The RG-II monomer was chemically de-O-acetylated by treatment for 16 h at 4°C with 0.1 M NaOH. The solution was neutralized with acetic acid, dialyzed (3500 Dalton MWCO) and freeze dried.
Generation of the 13C labeled RG-II FAD fragment
13 C-enriched RG-II isolated from isotopically labelled Rosa cells was treated for 16 hours at 40°C with 0.1M TFA. This treatment releases mostly the B side chain from the monomer.17 The fragments generated were separated using size-exclusion chromatography with a Superdex-75 Increase column (Cytiva, USA) eluted at 0.5 mL/min with 50 mM ammonium formate, pH 5. The B-chain-depeleted RG-II eluted between 27 min and 32.5 min and was collected and then repeatedly freeze dried to remove the ammonium formate. A solution of this fraction (32 mg) in 50 mM Tris-HCl, pH 8, containing 1mM CaCl2 was treated for 24 h at 20°C with a pectate lyase from Aspergillus sp. (2 µl/mg, Megazyme, 7500 U), with shaking at 150 rpm. The lyase-treated material was then fractionated using the Superdex-75 column. The fragment comprising a portion of the galacturonan backbone with chains A, F and D attached to it (FAD fragment) eluted between 30–35 min.
Enzymatic removal of arabinose from chain D
A solution of celery RG-II monomer (8 mg) in water (400 µL) was mixed with 100 mM NaOAc, pH 5.2 (200 µL), and the Dha hydrolase/β-L-arabinofuranosidase from Bacteroides thetaiotaomicron added (10 µL, 1 mg/mL; NzyTech, Portugal). The mixture was kept overnight at 37°C. The solution was diluted to 2 mL with water, dialyzed (3500 Dalton MWCO) against deionized water and freeze dried.
NMR spectroscopy
NMR spectra were recorded with a Varian NMR spectrometer (Agilent Technologies) operating at 600 MHz using a 5 mm cold probe or with a Bruker NEO (Bruker) spectrometer operating at 900 MHz. Solutions of the RG-II monomers and the FAD fragment in deuterium oxide (0.2 mL, 99.9%; Cambridge Isotope Laboratories, Tewksbury, MA, USA) were transferred to a 3 mm NMR tube. 1H and 13C NMR spectra were obtained using standard Varian and Bruker pulse programs. NOESY experiments were performed using 0.2 sec mixing time. Chemical shifts were measured relative to internal DMSO (δ 1H 2.721, δ 13C 39.39). Data were processed using MestReNova software (Mestrelab Research S.L., Santiago de Compostela, Spain).
In silico model construction and simulation set-up
The experimental NMR observations were used to guide the setup of molecular simulations to refine structural models and to test which models best depicts RG-II in its natural state. The eight RG-II models (Fig. 1, SI Figure S1) were built, refined according to the NMR results, simulated, and analyzed to determine which ones best fit the experimental results.
A molecular-mechanics model of the RG-II molecule requires a full set of parameters for bonds, angles, dihedrals, charges, and nonbonded interactions. Although many of the parameters required to model RG-II already existed in the CHARMM carbohydrate force field,39, 40, 46 the presence of unusual saccharide residues (Kdo,Dha, Aceric Acid, Apiose) along with the many new linkages between them and other known saccharide residues required the generation of new CHARMM monosaccharide residues and patches for linkages. These methods are described in SI Section S2 and SI Table S4. We built each of the eight Models (Fig. 1, SI Figure S1) with a backbone of ten GalA residues that was equilibrated to adopt a relaxed configuration. Coordinates for each sidechain were obtained one residue at a time by attaching the residue to the backbone, then rotating the added residue around its glycosidic linkage into an energetically feasible configuration followed by minimization to form a relaxed structure.
The resulting initial structure for each model was subjected to molecular dynamic simulations in a fully solvated environment, with ions to maintain charge neutrality, at 600 K in the NVT ensemble. The final stage of equilibration involved running four stepwise restrained simulations of each replica of each model for totally 4 ns to slowly pull atoms together that are known to be close from NMR experimental data (Table 1). This is a form of NMR structure refinement and was undertaken to start the unbiased simulations as close to structures giving correct NMR distances as possible for any given model and replica. The MD simulations on the equilibrated structures continue for 64 ns at 600 K in the NVT ensemble.
The trajectories at 600 K were then selected every 1 ns to produce 64 distinct configurations which are used as replicas in the subsequent replica-exchange molecular dynamics (REMD) simulations for enhanced configurational sampling.47 The temperature range in REMD for this study is from 300 to 600 K with 64 replicas to obtain an average acceptance probability of ~ 16% between neighboring replicas at an interval of 500 steps. Two different cases are conducted in the REMD simulations: i) the simulations were conducted for 10 ns while maintaining all the NOE restraints, then the NOE restraints were removed, and the simulations were continued for 20 ns. ii) starting with the same initial conformations of the 64 replicas, the simulations were conducted for 20 ns while removing all the NOE restraints. Convergence analysis was performed by checking the 1st 10 ns of the trajectory against the overall 20 ns trajectory.
All simulations were conducted using the CHARMM program48 with the CHARMM carbohydrate force field, all36_carb, including the additionally parametrized residues and linkages specific to RG-II included in the Supplemental Information.37, 38 The DOMDEC fast parallel CHARMM method49 was used in the molecular dynamics. The simulations were performed using a 2 fs time step and SHAKE50 to keep the length of bonds to hydrogen fixed. Periodic solvated systems used a non-bonded cut-off of 12 Å and the particle mesh Ewald method51 for long-range electrostatics. Simulation results were recorded as time series of RG-II structural coordinates, saved at every 2 ps for 50 ns of simulations under three cases per model, or a total of 3.2 µs of simulation time per model. The sampled RG-II structures at 300 K were used for data analysis, while all other structures sampled at higher temperatures are discarded. To calculate the root mean square fluctuations of the RG-II atoms, structural alignments of trajectories were performed using the ten RG-II backbone residues as a reference.
All the force-field parameters and topology files required to simulate RG-II in CHARMM are provided as part of SI data (rg2_charmm_ff.zip).