Snapshots of the Reaction Coordinate of a Thermophilic 2′-Deoxyribonucleoside/ribonucleoside Transferase

Nucleosides are ubiquitous to life and are required for the synthesis of DNA, RNA, and other molecules crucial for cell survival. Despite the notoriously difficult organic synthesis of nucleosides, 2′-deoxynucleoside analogues can interfere with natural DNA replication and repair and are successfully employed as anticancer, antiviral, and antimicrobial compounds. Nucleoside 2′-deoxyribosyltransferase (dNDT) enzymes catalyze transglycosylation via a covalent 2′-deoxyribosylated enzyme intermediate with retention of configuration, having applications in the biocatalytic synthesis of 2′-deoxynucleoside analogues in a single step. Here, we characterize the structure and function of a thermophilic dNDT, the protein from Chroococcidiopsis thermalis (CtNDT). We combined enzyme kinetics with structural and biophysical studies to dissect mechanistic features in the reaction coordinate, leading to product formation. Bell-shaped pH-rate profiles demonstrate activity in a broad pH range of 5.5–9.5, with two very distinct pKa values. A pronounced viscosity effect on the turnover rate indicates a diffusional step, likely product (nucleobase1) release, to be rate-limiting. Temperature studies revealed an extremely curved profile, suggesting a large negative activation heat capacity. We trapped a 2′-fluoro-2′-deoxyarabinosyl-enzyme intermediate by mass spectrometry and determined high-resolution structures of the protein in its unliganded, substrate-bound, ribosylated, 2′-difluoro-2′-deoxyribosylated, and in complex with probable transition-state analogues. We reveal key features underlying (2′-deoxy)ribonucleoside selection, as CtNDT can also use ribonucleosides as substrates, albeit with a lower efficiency. Ribonucleosides are the building blocks of RNA and other key intracellular metabolites participating in energy and metabolism, expanding the scope of use of CtNDT in biocatalysis.


Protein sequence
The coding sequence for protein expression (derived from the sequence available at Uniprot, accession K9TVX3) was as below, where the region in bold is removed after cleavage with TEV protease:

Size exclusion chromatography
100 ml of CtNDT (1 mg/mL) was injected onto a Superdex TM 75 Increase 10/300 GL to estimate the protein quaternary structure by relative mass.Buffer was 50 mM MES, 250 mM NaCl, pH 6.5 with a flow rate of 0.2 mL/min.The elution peak from the SEC column was around 10.03 mL, corresponding to a mass of ~ 62 kDa calculated with a calibration curve (Figure .S2).For the calibration curve, a Gel Filtration Calibration Kit LMW (Biorad) was used, including aprotinin (bovine lung), ribonuclease A (bovine pancreas), Carbonic anhydrase (bovine erythrocytes), Ovalbumin (hen egg), Conalbumin (chicken egg white) and blue dextran 2000.The MS was operated in ESI+ and scanned from 500 -2500m/z with lock mass of LeuEnk.The protein spectrum elution at 3 minutes was combined and the raw data processes to mass using MaxEnt algorithm at 0.1 resolution using peak width of half height of 0.4Da.

Standard enzyme kinetics of CtNDT with (2´deoxy)nucleoside substrates
A standard assay contained CtNDT, a 2´deoxynucleoside and a base (50 µl reaction volume) in a mixed buffer solution (final 30 mM CHES, MES and HEPES, pH 8.5) incubated at 45°C for 5, 10 and 15 minutes.At these times, an aliquot from the reaction mixture was removed and quenched with 200 µl of 10 M Urea, centrifuged at 24,000g for 10 minutes.100 µl of quenched mixture were taken and placed into 96-well round bottom microplate (Agilent Technologies) and 10 µl of each sample were injected into the HPLC column (using a Shimadzu Prominence HPLC and the method described below

HPLC conditions for reaction monitoring
A 2.5 mm, 50 mm X 4.6 mm HSS T3 column (Waters TM ) was used with buffer A (10 mM trimethylammonium acetate, pH 7) and buffer B (ACN + 0.1% TFA).A gradient elution was used to separate the compounds: from 0-10 mins, 99% to 85% buffer A and 1% to 15% buffer B, from 10-15 mins, 85% to 0% buffer A and 15% to 100% buffer B. The oven temperature was set at 40°C and the absorbance was set at 260 nm.The column was equilibrated with buffer A: buffer B (1:99) for at least 15 mins before each injection.Retention times for reference standard compounds were obtained by running each standard (Figure S3, top).

Differential Scanning Fluorimetry for protein melting temperature determination
Enzyme was prepared the day before the assays and the concentration of the sample was calculated as 1000/ !"#$%&' )#'#)%" mM.The concentration of sypro orange (5000X stock in DMSO) was prepared as 10X in the protein purification buffer.Enzyme (5 ml), the test buffer (20 ml) and the sypro orange working dye solution (25 ml) were mixed in 0.2 ml non-skirted 96-well PCR plates.Data were acquired using a Stratagene Mx3005p instrument, under a temperature gradient from 25°C -95°C in 1°C increments for a total of 1.5h.For each condition, triplicate experiments were carried out.Data were analysed by fitting to a Boltzmann equation in Prism 9.0 to obtain melting temperatures in different conditions.

Viscosity effect on kcat
To obtain information about rate limiting steps in the temperature range in which deviation from linear Arrhenius behaviour was observed, we determined kcat at increasing concentrations of glycerol.Reactions were carried out with 20nM wild type CtNDT and 0, 15 and 24% glycerol at 35 degrees or 0, 18 and 27% glycerol at 55 degrees and saturating concentration of substrates (1mM 2´-dGuo and 10 mM Ade).Formation of 2´-dAdo was quantified.Test assays with 2mM 2´-dGuo showed both 1mM and 2mM 2´-dGuo were saturating at the maximum % glycerol used (24% at 35 degrees and 27% at 55 degrees).Data are shown on Figure S3.

Under initial rate, [B1] and [N2]
~0 so all terms with those also disappear:

this is the kcat/KM for Nucleoside1
Under initial rate, [B1] and [N2]~0 so all terms with those also disappear.
Under initial rate, [B1] and [N2]~0 so all terms with those also disappear.[B2] is small so this term also disappears: )}, => this is the kcat/KM for nucleobase2

Rate limiting steps:
The reaction catalysed by CtNDT is fully reversible, and both substrates and products are (or can be) identical, as depicted below where forward rate constants lead to nucleoside bond breaking and ribosylated protein while reverse rate constants lead to nucleoside formation.Therefore, our pre-steady-state data makes clear that 2¢-deoxynucleoside release is not rate limiting, as binding experiments showed association and dissociation rate constants for 2¢-deoxyadenosine 100 times faster than kcat.Because identical kcat values were determined for 2¢-deoxyadenosine and 2¢-deoxyguanosine when Gua or Ade were employed as nucleobases, respectively (Table S2 Steady-state kinetics of CtNDT), steady-state turnover is likely to be limited by the same step when these different 2¢-deoxynucleosides are used as substrates, which argues against different values for kon and koff for different purine nucleosides.Therefore, one can consider the CtNDT reaction coordinate (and more generally for dNDTs) is a "mirror image", meaning the binding of the first substrate is akin kinetically to the release of the last product as they are both nucleosides.Scheme S1 exemplifies this point: Scheme S1: First and second half reactions for CtNDT Therefore, given 1) the maximum viscosity effect of 1.0 observed which points towards a diffusional step limiting the reaction; 2) the fast rate constants for k1 and k-1 100x faster than kcat; and 3) viscosity experiment performed with 10 mM nucleobase, more than 10x over KM, resulting in a fast observed rate of nucleobase association; the rate imiting step for CtNDT is likely nucleobase1 release (k3 in the scheme above or k6 in Scheme 1 shown in the main paper).Solvent viscosity effect on kcat using glycerol as a microviscogen: A) replot of relative rate in function of relative viscosity, using 20 nM CtNDT, 1 mM 2¢dGuo and 10 mM adenine.Experiments were performed in duplicate and data show as mean and standard error of the mean.A hypothetical curve with slope = 1 is shown (purple dotted line) as a reference for a reaction fully limited by diffusional steps.B) control experiment comparing reaction rates with 1mM and 2mM 2¢dGuo, demonstrating no change in slopes within experimental error, and therefore that the rate observed on (A) corresponds to kcat.C) control experiment with PEG 8K (no effect), which would be acting as a macroviscogen.D) Residual plots for fits on Figure 2a and E) Figure 2b.

General scheme for reactions monitored:
Data were fitted using Kintek Global Explorer.Details for each experiment are as below: • 2´deoxyadenosine binding to CtNDTE88A.
This experiment was used to determine the rate constants for 2´deoxyadenosine binding, k1 and k-1.

Figure
Figure S2Intact mass spectrum of CtNDT and its mutants.

Figure S5b :
Figure S5b: Initial velocity assays with wild type CtNDT, while varying 2´-dGuo with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5c :
Figure S5c: Initial velocity assays with wild type CtNDT, while varying 2´-dIno with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5d :
Figure S5d: Initial velocity assays with wild type CtNDT, while varying 2´-dCyd with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5e :
Figure S5e: Initial velocity assays with wild type CtNDT, while varying 2´-dUrd with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were

Figure S5f :
Figure S5f: Initial velocity assays with wild type CtNDT, while varying 2´-dThd with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5h :Figure S5g :
Figure S5h: Initial velocity assays with wild type CtNDT, while varying Gua with fixed 2´-dAdo.Reactions were performed in duplicate according to the methods reported herein.Left panel

Figure S5i :
Figure S5i: Initial velocity assays with wild type CtNDT, while varying Hyp with fixed 2´-dAdo.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5j :
Figure S5j: Initial velocity assays with wild type CtNDT, while varying Cyt with fixed 2´-dAdo.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5k :
Figure S5k: Initial velocity assays with CtNDTD62N mutant, while varying 2´-dGuo with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5l :
Figure S5l: Initial velocity assays with CtNDTE88Q mutant, while varying 2´-dGuo with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5m :
Figure S5m: Initial velocity assays with CtNDTM120C mutant, while varying 2´-dGuo with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5n :
Figure S5n: Initial velocity assays with wild type CtNDT, while varying Ado with fixed Hyp.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

Figure S5o :
Figure S5o: Initial velocity assays with wild type CtNDT, while varying Guo with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated

Figure S5p :
Figure S5p: Initial velocity assays with wild type CtNDT, while varying Ino with fixed Ade.Reactions were performed in duplicate according to the methods reported herein.Left panel depicts raw HPLC data with peaks for compounds detected labelled.Areas were integrated and converted into concentration product formed per unit of time for the Michaelis Menten plot shown on the right.

FigureFigure S10 :Figure
Figure S6: 2´-deoxyadenosine binding to CtNDTE88A.a) Raw data for binding experiment (dots), line is a fit to a single exponential equation, only used as a starting point for Kintek Global Explorer analysis; b) replot of analytically fitted rate constants in function of ligand concentration.c) Fitspace analysis of the fitted model.

Figure S13 :Figure S14 :
Figure S13: Immucillin-H binding to CtNDT and CtNDTE88Q.Raw data collected with 0.75 µM enzyme and 50 µM and 100 µM ImmH.Inset shows a replot of the analytical fitting to a single exponential equation of each transient.Data for CtNDTE88Q was noisy (with large Fitspace confidence boundaries) and therefore not further interpreted mechanistically.

Figure S18 :
Figure S18: Intact protein mass spectra of CtNDT and its mutants in the absence and presence of Clofarabine.In this experiment, the same protein batch as in Fig. S4 -CtNDT WT, expected MW 17861.5 Da was used.a) mass spectrum of wild type CtNDT which underwent same incubation time as other conditions under comparison without the addition of ligands.b) Protein mass spectrum of wild type CtNDT incubated with 1mM clofarabine.c) Protein mass spectrum of CtNDTE88A incubated with 1mM clofarabine.

Time (mins) %A (98% water with 0.1% Formic acid and 2% acetonitrile) %B (100% acetonitrile with 0.1% Formic acid) 0.200 98
20 µM enzyme alone or after overnight incubation with 1mM Clofarabine by end-over-end rotation were used.Protein samples were analysed at the University of St Andrews mass spectrometry and proteomics facility.For analysis, 20ul of sample at 1:20 dilution were injected onto a Waters MassPrep micro column 2mmx5mm on a Waters Xevo LC-MS system optimised for protein analysis.A short gradient elution was used to desalt and then elute the protein as follows: Data fitting on Kintek Global Explorer yielded (model for 1 step binding):