Discovery of an RmlC/D fusion protein in the microalga Prymnesium parvum and its implications for NDP-β-l-rhamnose biosynthesis in microalgae

The 6-deoxy sugar l-rhamnose (l-Rha) is found widely in plant and microbial polysaccharides and natural products. The importance of this and related compounds in host–pathogen interactions often means that l-Rha plays an essential role in many organisms. l-Rha is most commonly biosynthesized as the activated sugar nucleotide uridine 5′-diphospho-β-l-rhamnose (UDP-β-l-Rha) or thymidine 5′-diphospho-β-l-rhamnose (TDP-β-l-Rha). Enzymes involved in the biosynthesis of these sugar nucleotides have been studied in some detail in bacteria and plants, but the activated form of l-Rha and the corresponding biosynthetic enzymes have yet to be explored in algae. Here, using sugar-nucleotide profiling in two representative algae, Euglena gracilis and the toxin-producing microalga Prymnesium parvum, we show that levels of UDP- and TDP-activated l-Rha differ significantly between these two algal species. Using bioinformatics and biochemical methods, we identified and characterized a fusion of the RmlC and RmlD proteins, two bacteria-like enzymes involved in TDP-β-l-Rha biosynthesis, from P. parvum. Using this new sequence and also others, we explored l-Rha biosynthesis among algae, finding that although most algae contain sequences orthologous to plant-like l-Rha biosynthesis machineries, instances of the RmlC-RmlD fusion protein identified here exist across the Haptophyta and Gymnodiniaceae families of microalgae. On the basis of these findings, we propose potential routes for the evolution of nucleoside diphosphate β-l-Rha (NDP-β-l-Rha) pathways among algae.


Figure S1-The target nucleotide-diphospho-sugars and their numbering.
From a synthetic point of view the common feature of the target nucleotide-diphospho-sugars 1 and 2 is the 1,2-cis configuration of the sugar ring that is difficult to make and once installed the resulting compounds are very acid labile with a high tendency to hydrolyse or to form cyclic 1,2-phosphodiesters. The key to the synthesis of compounds 1 and 2 is the preparation of the corresponding sugar-1-phosphate 6 (Scheme 1) with well-defined stereochemistry of the anomeric centre.
Currently available approaches to TDP-β-L-Rha (1) employ enzymatic transformation of TDPα-D-Glc and the product of this transformation has been isolated and characterised reasonably well by 1 H NMR [1][2][3].
Unfortunately, apart from mobility in paper chromatography, paper electrophoresis and nucleoside / total phosphate / acid-labile phosphate ratios the products were not further by chromatography, (see Offen et al [7]) and consequently, it would be fair to assume that the resulting nucleotide-diphospho-sugars prepared by Shibaev et al [4] have the correct anomeric configuration. However, in the absence of the analytical data to support this, uncertainty remains. A more recent synthesis of TDP-β-L-Rha (1) was published by Zhao and Thorson [8] starting from L-rhamnopyranosyl phosphate that in turn was prepared from 2,3,4tri-O-acetyl-α-L-rhamnopyranosyl bromide by a silver triflate promoted phosphorylation with dibenzyl phosphate in the presence of 2,4,6-collidine followed by a global deprotection. Given a certain ambiguity in the anomeric assignment of the sugar-1-phosphate generated by Zhao Here, we wish to report the first fully synthetic approach to TDP-β-L-Rha (1) that was also modified for the preparation of UDP-β-L-Rha (2).

Results and Discussion
Diphenyl (tri-O-acetyl-β-L-rhamnopyranosyl) phosphate [14] (5) (Scheme 1) was used as the starting point for our syntheses of both TDP-β-L-Rha (1) and UDP-β-L-Rha (2). Because of the above-mentioned issues with previous reports regarding the anomeric configuration of the synthetic TDP-L-rhamnose we decided to prepare and purify both anomers of 5 and confirm the anomeric configuration using extensive NMR studies.
L-Rhamnose was first subjected to a one-pot procedure comprising an acetylation, followed by bromination and hydrolysis of the corresponding tri-O-acetyl-L-rhamnosyl bromide to yield tri-O-acetyl-L-rhamnopyranose (4) which was obtained as an anomeric mixture containing 85% of the thermodynamically more stable α anomer (α/β anomers 5.7:1 by 1 H NMR).
To prepare the β phosphate 5β Sabesan and Neira used the fact that the β anomer of hemiacetal 4 is more reactive than the dominant α anomer and performed the subsequent phosphorylation at room temperature and limiting amount of the phosphorylating agent, diphenyl chlorophosphite [14]. In our hands these conditions resulted in an anomeric mixture of the diphenyl (tri-O-acetyl-α,β-L-rhamnopyranosyl) phosphates (5) that was indeed significantly enriched in the desired β anomer (α/β anomers 1:4 by 1 H NMR). Moreover, both anomers separated very well on silica gel TLC (5α Rf = 0.65 and 5β Rf = 0.58, see Experimental section for details) and were reasonably stable provided that the mobile phase contained a small amount of Et3N (1%) to account for limited acid stability of the phosphates.
Preparative silica gel column afforded α-anomer 5α (6%) and β-anomer 5β (49%) in pure forms. Analytical data ([α]D in CHCl3, 1 H and 13 C NMR in CDCl3) of both anomers were in good agreement with data published by Sabesan and Neira [14]. We have noticed, however, that both anomers of 5 undergo gradual decomposition on long exposure to chloroform or when solvent evaporation was attempted which we put down to traces of HCl present in this solvent degrading the acid labile phosphates. For this reason we acquired a second set of analytical data using toluene or toluene-d8, respectively. The anomeric phosphates 5 were stable in this solvent.
The phosphate 5β was globally deprotected using the same conditions as in Sabesan and Neira [14] to give the desired β-L-rhamnopyranosyl phosphate (6). The PtO2 catalysed hydrogenation to deprotect the diphenyl phosphate in 5β was fast and complete even at low hydrogen pressure (around 15 psi, originally used 55 psi). In summary, the presented synthesis of β-L-rhamnopyranosyl phosphate (6) offers a short and a much simpler alternative to methods published earlier [6, 13]. Although perhaps not the most obvious choice, the phenyl protecting groups in the phosphates 5α/5β greatly enhance the separation of the anomers on silica gel when compared to benzyl analogues and do not present any hurdles in terms of removal. Moreover, diphenyl chlorophosphate was commercially available whereas dibenzyl chlorophosphate was not at the time of manuscript preparation. The presented preparation of β-L-rhamnopyranosyl phosphate (6) is scalable to multi-gram quantities and the product in the bistriethylammonium form is stable to store over long periods of time.
The final step in the synthesis involved the standard pyrophosphate bond formation [18] between the pyridine soluble bistriethylammonium salt of β-L-rhamnopyranosyl phosphate (6) and the morpholidate activated [19] thymidine or uridine 5'-monophosphate, respectively, in dry pyridine. After 5 to 7 days stirring at 4 o C the yields (by HPLC) of the products were 55% for 1 and 70% for 2. An attempted purification of the desired nucleotide-diphospho-sugars 1 and 2 using strong anion exchange chromatography failed because of considerable degree of decomposition during the freeze-drying and vacuum assisted removal of the volatile ammonium bicarbonate buffer. Instead, a combination of Sephadex LH-20 gel permeation chromatography followed by C18 reverse phase chromatography was required to achieve a complete separation of the target compounds 1 and 2 from particularly the unreacted starting materials. After purification the resulting TDP-β-L-Rha (1) and UDP-β-L-Rha (2) were obtained in yields 16% for 1 and 21% 2 in high degree of purity as shown by NMR ( Figure S9, S11, S12) and HPLC (data not shown). 1 H NMR of TDP-β-L-Rha (1) produced enzymatically in this study ( Figure S10) is in good agreement with synthetically produce TDP-β-L-Rha (1) ( Figure   S9).
The 1H-tetrazole catalysed variant [20] of the pyrophosphate coupling has also been investigated. Although shorter reaction times were required to achieve similar conversions, the products were more difficult to purify from the reaction mixtures as they were more complex than the reaction mixtures obtained from the non-catalysed variant. High resolution accurate mass spectra were obtained using a Synapt G2 Q-Tof mass spectrometer using either positive or negative electrospray ionisation.

General Method A: Strong anion-exchange (SAX) HPLC on Poros HQ 50
The chromatography was performed on a Dionex Ultimate 3000 instrument equipped with UV/vis detector. An aqueous solution of a sample was applied on a Poros HQ 50 column (50×10 mm, column volume (CV) = 3.9 ml). The column was first equilibrated with 5 CV of 5 mM ammonium bicarbonate buffer, followed by linear gradient of ammonium bicarbonate from 5 mM to 250 mM in 15 CV, then hold for 5 CV, and finally back to 5 mM ammonium bicarbonate in 3 CV at a flow rate of 8 ml/min and an online UV detection to monitor A265. After multiple injections, the column was washed with 3 CV of 1 M ammonium bicarbonate followed by 3 CV of Milli-Q water.

General Method B: Gel filtration chromatography on Sephadex LH-20
The purification was performed on a Perkin Elmer Series 200 instrument equipped with UV/vis and RI detectors. A solution of a sample in MeOH was applied on a Sephadex LH-20 column (800×16 mm) and eluted isocratically with MeOH at a flow rate of 1 ml/min and detection with on-line RI detector and UV detector to monitor A265. Fractions containing the sugar nucleotide (the sugar nucleotides eluted typically around 130 ml) were pooled and the solvent was evaporated in vacuo.

General Method C: Reverse phase (RP) C18 purification
The purification was performed on a Dionex Ultimate 3000 instrument equipped with UV/vis detector. A solution of a sample in water was applied on a Phenomenex Luna 5 μm C18 (2) column (250x10 mm, CV = 19.6 ml) and eluted isocratically with 50 mM Et3NHOAc, pH 6.8 with 1.5% CH3CN in 8 CV at a flow rate of 5 ml/min and detection with on-line UV detector to monitor A265. Fractions containing the sugar nucleotide were pooled and freeze-dried.

Uridine 5'-diphospho-β-L-rhamnose bistriethylammonium salt (2)
In order to remove traces of water, β-L-rhamnopyranosyl phosphate bistriethylammonium salt Ethanol was evaporated at reduced pressure and ambient temperature and the aqueous residue was freeze dried. At this stage the sample can be stored at -80 o C for any length of time before the next step.
Lipophilic components were removed by partitioning the sample between water and butan-1ol [22]. In brief, the sample was dissolved in 9% aqueous butan-1-ol (3 x 2 ml) and transferred into a glass vial (10 ml volume). The solution was extracted with 90% butan-1-ol (3 x 2 ml) to remove the top layer containing lipids (but also chlorophyll, and insoluble polysaccharides such as paramylon forming middle layer). The bottom layer was collected and extracted again.
Centrifugation was used to speed up the separation of the layers (200 -800 x g, 4 o C, 5 min).
The clear aqueous layer was collected and freeze dried in a pear-shaped flask (foaming may appear under vacuum). Samples were stored at -80 o C before the next step.
Solid phase extraction (SPE) of sugar nucleotides was performed essentially as described by Rabina and co-workers [23]. In brief, a graphitised carbon column (EnviCarb, Supelco, 250 mg, 3 ml) was conditioned by washing with 80% aqueous acetonitrile containing 0.1 % trifluoroacetic acid (3 ml) followed by water (2 ml). The sample was dissolved in ammonium bicarbonate (5 mM, 500 µl) and applied on the SPE column. The column was washed with water (2 ml), followed by 25% aqueous acetonitrile (2 ml), and 50 mM triethylammonium acetate buffer (pH 7.0, 2 ml). Finally, the sugar nucleotides were eluted with 50 mM triethylammonium acetate buffer pH 7.0 containing 25% acetonitrile (1.5 ml). The sample was filtered using 0.45 µm disc filters (PTFE) and freeze dried. Samples were stored at -80 o C prior to LC-MS/MS analysis.
LC-MS/MS profiling of sugar nucleotides was performed on a Xevo TQ-S tandem quadrupole mass spectrometer (Waters) operated in multiple reaction monitoring (MRM) mode coupled to an Acquity UPLC. ESI-MS/MS analysis was performed in negative ion mode using a source with a capillary voltage of 1.5 kV, 500 °C desolvation temperature, 1000 l.hr -1 desolvation gas, 150 l.hr -1 cone gas, and 7 bar nebulizer pressure. MRM transitions for sugar nucleotide standards in negative ESI mode were generated using IntelliStart software (Table S1) (Table S1).
Limit of detection was determined to be 10 fmol on column using a serial dilution of UDP-α-D-Glc. Samples of extracted sugar nucleotides were reconstituted in buffer A (25 µl) and injected (5 µl, 20 % of total) using partial loop injection. Analysis of 3 biological replicates was performed. Where in doubt, co-injection of sample with appropriate standard sugar nucleotide was used for positive identification. Data processing was performed using MassLynx (Waters) software. Although between runs there were significant differences in absolute retention times of standards, relative retentions were reasonably reproducible (Table S1). To ensure maximum retention time (Rf) stability, after a batch of samples, the PGC column had to be regenerated and its performance was tested using UDP-α-D-Glc as a standard. The regeneration and column performance steps were performed using a standard HPLC system           Table S2 -List of organisms examined in this study with respective nucleic acid database identifiers and sequence identifiers found 0 for NDP-β-L-Rha biosynthesis. 1