Molecular basis of sulfolactate synthesis by sulfolactaldehyde dehydrogenase from Rhizobium leguminosarum

Sulfolactate (SL) is a short-chain organosulfonate that is an important reservoir of sulfur in the biosphere. SL is produced by oxidation of sulfolactaldehyde (SLA), which in turn derives from sulfoglycolysis of the sulfosugar sulfoquinovose, or through oxidation of 2,3-dihydroxypropanesulfonate. Oxidation of SLA is catalyzed by SLA dehydrogenases belonging to the aldehyde dehydrogenase superfamily. We report that SLA dehydrogenase RlGabD from the sulfoglycolytic bacterium Rhizobium leguminsarum SRDI565 can use both NAD+ and NADP+ as cofactor to oxidize SLA, and indicatively operates through a rapid equilibrium ordered mechanism. We report the cryo-EM structure of RlGabD bound to NADH, revealing a tetrameric quaternary structure and supporting proposal of organosulfonate binding residues in the active site, and a catalytic mechanism. Sequence based homology searches identified SLA dehydrogenase homologs in a range of putative sulfoglycolytic gene clusters in bacteria predominantly from the phyla Actinobacteria, Firmicutes, and Proteobacteria. This work provides a structural and biochemical view of SLA dehydrogenases to complement our knowledge of SLA reductases, and provide detailed insights into a critical step in the organosulfur cycle.


Heterologous expression and protein purification
The sequence encoding RlGabD (WP_017967313.1) was amplified from Rhizobium leguminosarum SRDI-565 genomic DNA by PCR using the primers 5'-TTACTCATATGCCATCCAACTATGACAGC-3' and 5'-TCAGACTCGAGGTTCTGCGTCAACCCGGCATC-3'.The amplicon was cloned into the MCS of pET-29b(+) (Novagen) using the NdeI and XhoI sites, and sequence-verified by Sanger sequencing.This plasmid was transformed into chemically competent E. coli BL21(DE3) cells and starter cultures grown in LB-medium (5 mL) containing 100 µg mL -1 kanamycin for 18 h at 37 °C with shaking at 200 r.p.m. 1 L volume cultures were inoculated with the starter culture (5 mL) and incubated at 37 °C, with shaking at 200 r.p.m. until an A 600 of 0.6-0.8 was reached.
RlGabD expression was induced by addition of IPTG (0.5 mM) and shaking continued overnight at 18 °C at 200 r.p.m.The cells were harvested by centrifugation at 5000 g for 20 min and the pellet resuspended in 50 mM sodium phosphate buffer pH 7.4, containing 500 mM NaCl and 30 mM imidazole.Cells were disrupted by ultrasonication for 3 × 5 min, 30 s on; 30 s off cycles, and the suspension was centrifuged at 50,000 g for 30 min to yield a clear lysate.
The C-terminal His 6 -tagged protein was purified by immobilized-metal affinity chromatography (IMAC) using a Ni-NTA column, followed by size exclusion chromatography (SEC).For IMAC, the lysate was loaded onto a pre-equilibrated Ni-NTA column, followed by washing with a load buffer (50 mM Tris, 500 mM NaCl, 30 mM imidazole pH 7.5).The bound protein was eluted using a linear gradient with buffer containing 300 mM imidazole.RlGabD fractions were pooled, concentrated and loaded onto a HiLoad 16/600 Superdex 75 gel filtration column pre-equilibrated with 50 mM Tris, 300 mM NaCl pH 7.5 buffer.The protein was concentrated using a Vivaspin® 6 with a 30 kDa MW cut-off membrane, to a final concentration of 8 mg mL -1 for structural studies.

Generation of active-site variant constructs
The active-site variant constructs, RlGabD Glu261Ala and RlGabD Cys295Ala, were generated with the Q5 site-directed mutagenesis kit (New England Biolabs) using the primers 5'-cgTTGGGCGGCAATGCGC-3' and 5'-cCAGCGAGAGGCGCTTG-3' for Glu261Ala and 5'-CGGCCAGACCgcgGTTTGCGCCA-3' and 5'-GCATTGCGGAATTTGGAG-3' for Cys295Ala, respectively (lower-case letters indicate the mutated sequence).Mutagenesis was verified by DNA sequencing and the proteins were purified by the same method used for the wild-type protein.

SEC-MALLS analysis
Experiments were conducted on a system comprising a Wyatt HELEOS-II multi-angle light scattering detector and a Wyatt rEX refractive index detector linked to a Shimadzu HPLC system (SPD-20A UV detector, LC20-AD isocratic pump system, DGU-20A3 degasser and SIL-20A autosampler).Work was conducted at room temperature (20±2 °C).Solvent was 0.2 µm filtered before use and a further 0.1 µm filter was present in the flow path.The column was equilibrated with at least 2 column volumes of buffer (50 mM NaPi, 300 mM NaCl pH 7.4) before use and flow was continued at the working flow rate until baselines for UV, light scattering and refractive index detectors were stable.Sample injection volume was 100 µL RlGabD at 6 mg mL -1 in 50 mM Tris buffer, 300 mM NaCl pH 7.4 containing 2 mM NADH; Shimadzu LC Solutions software was used to control the HPLC and Astra V software for the HELEOS-II and rEX detectors.The Astra data collection was 1 min shorter than the LC solutions run to maintain synchronisation.Blank buffer injections were used as appropriate to check for carry-over between sample runs.Data were analysed using the Astra V software.
MWs were estimated using the Zimm fit method with degree 1.A value of 0.182 was used for protein refractive index increment (dn/dc).

Nanoscale Differential Scanning Fluorimetry (nanoDSF)
NanoDSF studies were performed on a Prometheus NT.48 (NanoTemper).Data recording and initial analysis was performed with PR.ThermControl software.All protein samples were at 2 mg.ml-1 in 50 mM Tris, 150 mM NaCl at pH 7.4, with a 15 μl capillary load per sample.
Temperature was ramped from 15 °C to 95 °C, at 1.0 °C/min with 10% excitation power.
Experiments were performed in duplicate.

Synthesis of SLA and determination of concentration
SLA was synthesized as a solution in water as reported, at a nominal concentration of 109 mM. 17The concentration of the SLA solution was measured by quantitative 1 H NMR spectroscopy using a calibration curve of methyl β-glucoside to determine the instrument sensitivity.The calibration curve was generated by measuring the absolute integration of H1 (δ 4.25 ppm in D 2 O) of methyl β-glucoside at 100, 50, and 10 mM.The absolute integration of H1 was 15738, 7753.06,1249.15,respectively (Fig. S1A).The SLA stock solution was diluted spectrum was 8647.6.The calculated concentration of SLA was 110 mM, which was used for all further calculations.

Measurement of consumption of SLA by RlGabD
The production of NADH from oxidation of SLA catalyzed by RlGabD was monitored using UV-Vis spectrophotometer at 340 nm, in triplicate.The reaction was carried out in 30 mM Tris buffer, 30 mM KCl pH 7 at 30 °C [NAD + ] = 1.5 mM, [DL-SLA] = 1 mM, and [RlGabD] = 15.8 nM.Phosphate buffer was not used as it has been shown to have a negative impact on NADH stability. 24The reaction appeared complete after 40 min (Fig. S1B).More RlGabD was added to a final [RlGabD] = 205 nM, and the reaction was monitored for 100 min with no further increase in absorbance.The extinction coefficient used for NADH was 6363 M -1 cm -1 .Error is standard error mean.

Michaelis-Menten kinetic analysis of RlGabD
Michaelis-Menten kinetic analysis were performed for SLA, GAP, NAD + and NADP + under pseudo first-order conditions in which the concentration of one substrate was varied while the other was held constant, and vice versa.The production of NADH/NADPH from oxidation of SLA catalyzed by RlGabD SLA dehydrogenase was monitored using a UV-Vis spectrophotometer at 340 nm.For the Michaelis-Menten kinetics of NAD + /NADP + , reactions were conducted in 0.1% BSA 30 mM Tris buffer pH 7.12 at 30 °C with 3 nM RlGabD, constant 0.25 mM D-SLA and varying concentrations of NAD + /NADP + (0.05-1 mM).For the Michaelis-Menten kinetics of D-SLA, constant 0.25 mM NAD + /NADP + and varying concentrations of D-SLA (0.025-0.25 mM) were used.For the Michaelis-Menten kinetics of racemic GAP, constant 0.25 mM NAD + /NADP + and varying concentrations of GAP (0.05-0.25 mM) were used.The apparent kinetic parameters, k cat , K M , and k cat /K M were calculated using the Prism 9 software package (GraphPad Scientific Software) (Table 1).No reaction was observed for >0.5 mM [D-SLA] and >0.25 mM [NADP + ].This appears to be a result of inhibition.
The activity of Cys295Ala and Glu261Ala variants were measured at 0.5 mM NAD + and 0.25 mM D-SLA, or 0.5 mM NADP + and 0.25 mM D-SLA) in 0.1% BSA 30 mM Tris buffer pH 7.12 at 30 °C.The concentration of each variant enzyme was 3000 nM.

Cryo-EM 3D-structure of RlGabD•NADH
A frozen aliquot of RlGabD was prepared at 4 mg/mL and complexed with 2 mM NADH for screening.R1.2/1.3 300 mesh UltrAuFoil gold grids (Quantifoil) were glow-discharged for 3 min at 20 mAmp/0.38 mBar.A total of 2.5 µL of this sample was applied to glow-discharged UltrAuFoil gold grids, which were subsequently blotted for 2 s with a blot force of 10, then were plunged into liquid ethane cooled by liquid nitrogen.Plunge-freezing was performed classified, without the use of symmetry constraints.The class showing well-defined structural features was then selected for 3D refinement with D2 symmetry imposed.This gave a reconstruction with a resolution of 3.37Å (FSC threshold of 0.143).Subsequently, CTF refinement 4 was performed for magnification anisotropy; optical aberrations (up to the fourth order); and per-particle defocus and per-micrograph astigmatism.Bayesian Polishing 5 was also used to optimise per-particle beam-induced motion tracks, followed by another round of autorefinement.CTF refinement, Bayesian Polishing and 3D refinement steps were repeated to yield a 2.52 Å map (FSC threshold of 0.143).Further details of the image processing and 3D reconstruction can be found in Table S1 and Fig. S6.

Model building, refinement, and validation
An initial model was built using the map_to_model function in Phenix. 6The initial model was using a Vitrobot Mark IV (Thermo Fisher Scientific) at 100% humidity and 22 °C.EER formatted movies of RlGabD•NADH complex were acquired on the Glacios microscope (Thermo Fisher Scientific), housed in the York Structural Biology Laboratory.The microscope was operated at an accelerating voltage of 200 kV with a Falcon 4 direct electron detector.The RlGabD•NADH dataset was acquired at a dose rate of 2.98 electrons per pixel per second, and a pixel size of 0.574 Å; target defocus values were -2 to -0.7 m.The autofocus function was run every 10 m, and the dataset was collected with a total dose of 50 electrons per Å 2 .Image processing and 3D reconstructionMovie frames of the RlGabD•NADH were motion corrected without binning, using a pixel size of 0.574 Å, and dose-weighted using the Motioncorr2 program.1 Contrast transfer function (CTF) corrections were performed using CTFFIND 4.1.2Most of the subsequent processing steps were carried out using RELION 3.3 Laplacian-of Gaussian (LoG) based automated particle picking was performed on the data.Particles were extracted and subjected to 2D classification.2D classes showing sharp structural features were chosen to build an initial 3D model.This initial model was used for template-based picking of particles (low pass-filtered at 20 Å).Picked particles were 2D classified to remove poor particles.Particles were 3D