Two RmlC homologs catalyze dTDP-4-keto-6-deoxy-d-glucose epimerization in Pseudomonas putida KT2440

l-Rhamnose is an important monosaccharide both as nutrient source and as building block in prokaryotic glycoproteins and glycolipids. Generation of those composite molecules requires activated precursors being provided e. g. in form of nucleotide sugars such as dTDP-β-l-rhamnose (dTDP-l-Rha). dTDP-l-Rha is synthesized in a conserved 4-step reaction which is canonically catalyzed by the enzymes RmlABCD. An intact pathway is especially important for the fitness of pseudomonads, as dTDP-l-Rha is essential for the activation of the polyproline specific translation elongation factor EF-P in these bacteria. Within the scope of this study, we investigated the dTDP-l-Rha-biosynthesis route of Pseudomonas putida KT2440 with a focus on the last two steps. Bioinformatic analysis in combination with a screening approach revealed that epimerization of dTDP-4-keto-6-deoxy-d-glucose to dTDP-4-keto-6-deoxy-l-mannose is catalyzed by the two paralogous proteins PP_1782 (RmlC1) and PP_0265 (RmlC2), whereas the reduction to the final product is solely mediated by PP_1784 (RmlD). Thus, we also exclude the distinct RmlD homolog PP_0500 and the genetically linked nucleoside diphosphate-sugar epimerase PP_0501 to be involved in dTDP-l-Rha formation, other than suggested by certain databases. Together our analysis contributes to the molecular understanding how this important nucleotide-sugar is synthesized in pseudomonads.

www.nature.com/scientificreports/ They are a paradigm of metabolically versatile microorganisms being able to recycle organic wastes and are key players in the maintenance of environmental quality 24 .
Following an unbiased approach and utilizing a restriction based genomic library, we identified the two paralogous proteins PP_1782 (now termed RmlC1) and PP_0265 (now termed RmlC2) as dTDP-4-dehydrorhamnose 3,5-epimerases while the last step namely the reduction to dTDP-l-Rha seems to be solely catalyzed by PP_1784 (RmlD). By contrast, two further candidate genes that were identified by database mining and homology analyses-PP_0500 and PP_0501-are not involved in dTDP-l-Rha biosynthesis. Taken together, our findings contribute to the molecular understanding how dTDP-l-Rha is synthesized in Pseudomonas putida KT2440.

Results
A screening system that allows for the discovery of dTDP-l-Rha synthesis genes. To identify genes involved in dTDP-l-Rha biosynthesis, we took advantage of cross functionality of pseudomonal EF-P in Escherichia coli and the fact that activation of the translation factor strictly depends on the nucleotide sugar as donor substrate. This cannot necessarily be expected, as the E. coli endogenous EF-P significantly differs from its pseudomonal counterpart 25 : although both proteins alleviate ribosome stalling at polyproline stretches 6,26 , their modes of activation are phylogenetically unrelated 6,27 . While E. coli EF-P (EF-P Eco ) strictly depends on (R)-β-lysylation 22,[28][29][30] and hydroxylation 31 of a conserved lysine, Pseudomonas EF-P (EF-P Ppu ) is rhamnosylated at an arginine by the glycosyltransferase EarP (EarP Ppu ) at the structurally equivalent position 6,21 . Despite these apparent distinct post-translational modifications, a combination of efp Ppu and earP Ppu from P. putida KT2440 can compensate for a lack of efp in E. coli (Δefp) as long as the endogenous dTDP-l-Rha pathway remains intact ( Fig. 2A,C) 6 . Interestingly, loss of any synthesis gene-here exemplified with a ΔrmlC strain-does not simply phenocopy Δefp but even results in more severe growth defects, as can be concluded from the corresponding doubling times (Fig. 2B): E. coli Δefp cross complemented with efp/earP Ppu grows twice as fast as the same strain additionally lacking rmlC (Δefp ΔrmlC). These growth defects are also reflected by the size of the colonies (Fig. 2C). The differences in growth rates provide us with a selection regime to identify dTDP-4-dehydrorhamnose-3,5-epimerase genes from a P. putida genomic library.
The library was constructed by partial restriction digestion of the P. putida genome with the dam and CpG methylation insensitive enzyme StuI (NEB) (Fig. 3). The average fragment size was set to 5 kb to ensure that at least one gene was completely covered (average gene size: 1.132 kbp). These were cloned into SmaI linearized pBAD33, which allows for high-level expression by induction of the P BAD promoter with l-arabinose 32 . Bottom left: Rhamnosylation of translation elongation factor EF-P in about 10% of all bacteria 6 . Bottom middle: biosynthesis of Streptomycin inter alia originating from dTDP-l-Rha 10 . Bottom right: Glycosylated flagella with a linking l-Rha moiety in certain Pseudomonads 11 . (B) dTDP-β-l-rhamnose biosynthesis pathway. Glucose-1-phosphate thymidylyltransferase, the first enzyme of the pathway, transfers a thymidylmonophosphate nucleotide to glucose-1-phosphate, which is further oxidated by dTDP-d-glucose 4,6-dehydratase at the C4 hydroxyl group of the saccharide. The double epimerization reaction at positions C3 and C5 is catalyzed by the dTDP-4-keto-6-deoxy-d-glucose 3,5-epimerase. Finally, the reduction of the C4 keto group by the dTDP-4keto-6-deoxy-l-mannose reductase leads to dTDP-l-Rha.  33,34 containing 0.2% l-arabinose. Considering duplication times (Fig. 2B) and genome coverage, we expect rmlC copies to accumulate already to a single-digit percentage of the total population within latest two days (= ~ 16 generations with mutant growth phenotype and ~ 32 for wild-type phenotype), even under unfavorable circumstances. Indeed, when plating the second overnight culture on LB agar we obtained colonies of two different sizes. Consequently, we isolated plasmids from 16 large clones and sequencing identified 12 times PP_1782 and four times PP_0265 as the insert. PP_0265 (from now on rmlC2/RmlC2) resides next to genes encoding a putative two component signal-transduction system (Fig. 4A). PP_1782 (from now on rmlC1/RmlC1) on the  Screening strategy for the identification dTDP-l-Rha biosynthesis genes in P. putida. Chromosomal DNA of P. putida (green) was fragmented by restriction digestion. Fragments with an average size of 5 kb were then ligated into the arabinose inducible vector (pBAD33). The resulting library was transformed into an E. coli Δefp P cadBA ::lacZ reporter strain that concomitantly lacks either rmlC or rmlD (ΔrmlC/ ΔrmlD) and additionally encodes earP Ppu and efp Ppu (orange) of P. putida in trans. Genes cross-complementing ΔrmlC and ΔrmlD recover the impaired growth phenotype and can be selected by size (white arrows).  (Fig. 4C). This protein, on the contrary, shares similarities with RmlD both at the sequence level (29% identity) as well as structurally.
To test the putative role of PP_0500 and PP_0501 in dTDP-l-Rha biosynthesis we made again benefit of EarP mediated activation of P. putida EF-P and its functionality in E. coli. Hence, we cloned the two genes into pBAD33 simultaneously adding a His 6 -tag coding sequence for immunodetection in order to ensure proper protein production (Fig. 5). rmlC1, rmlC2 and rmlD were also included in the study. The resulting plasmids pBAD33-rmlC1, pBAD33-rmlC2, pBAD33-PP_0501 as well as pBAD33-rmlD and pBAD33-PP_0500 were introduced into E. coli Δefp ΔrmlC + efp/earP Ppu and Δefp ΔrmlD + efp/earP Ppu , respectively. Of note, these are reporter strains in which EF-P functionality is coupled to LacZ expression (Fig. 5A). Whereas β-galactosidase activity is low in cell with an incomplete dTDP-l-Rha biosynthesis pathway, introduction of either rmlC1, rmlC2 (Fig. 5B) or rmlD (Fig. 5C) into the respective mutant strains led to a significant increase. By contrast, neither PP_0500 nor PP_0501 were able to rescue the Δefp Eco mutant phenotype.

Discussion
In the scope of this study, we have investigated the dTDP-l-Rha pathway of P. putida KT2440 with a focus on the epimerization of dTDP-4-keto-6-deoxy-d-glucose. Combining an unbiased approach and utilizing a genomic library, we identified two paralogous proteins RmlC1 and RmlC2. Duplication of rmlC is not restricted to P. putida KT2440 but certain other pseudomonads such as P. monteilii, P. fulva, P. plecoglossicida or P. asiatica harbor also two gene copies. In fact, functional redundancy in the dTDP-l-Rha biosynthesis pathway is nothing unusual. As an example, the two enzymes RffH and RffG of E. coli are paralogous to RmlA and RmlB, respectively 13 . Such duplications may be useful, e.g., to compensate for bottleneck reactions in the dTDP-l-Rha biosynthesis 44 . Such bottlenecks can occur at different stages as the pathway is not only utilized to ultimately generate dTDP-l-Rha. Specifically, dTDP-4-keto-6-deoxy-d-glucose is also a precursor of dTDP-3-acetamido-α-d-fucose 45 and TDP-d-viosamine 46 which are found as part of the glycan pattern in P. syringae 47 . Similarly, the two paralogs RmlC1 and RmlC2 in P. putida KT2440 might serve as starting point of similar but so far unknown reactions. Moreover, gene duplications open the gate for regulated expression in turn allowing the precise adjustment of the desired ratio of distinct NDP-sugars depending on parts of the dTDP-l-Rha biosynthesis pathway. It would also allow for the accumulation of educts or products of the preceding reactions such as dTDP-glucose and Glc-1P. Notably, whereas rmlC1 is part of an operon in which presumably the full dTDP-l-Rha pathway is encoded, the rmlC2 resides in the vicinity of two genes encoding a two-component system (TCS) of thus far unknown www.nature.com/scientificreports/ function. Based on the predicated domain composition, this specific TCS presumably transduces external signals into gene transcription. One might therefore speculate on regulated expression of rmlC2 according to the environmental conditions. While our genomic library revealed two RmlC paralogs in P. putida database mining indicated a further enzyme with similar activity PP_0501. However, our in vivo rhamnosylation assay disproved the initial hypothesis. Notably, the UDP-N-acetylglucosamine C4-epimerase PelX from P. protegens Pf-5 is structurally the closest homolog (identity 67%) 48 . PelX is involved in the biosynthesis of the GalNAc-rich bacterial polysaccharidepolysaccharide Pel, that is essential for pellicle biofilm formation 48,49 . One can hence hypothesize, that PP_0501 and the adjacent putative reductase PP_0500 might be involved in that pathway, instead.

Material and methods
Bacterial strains and growth condition. All strains and plasmids used in this study are listed and described in Table 1. E. coli cells were grown in Miller modified Lysogeny Broth (LB) 33,34 at 37 °C aerobically under agitation, if not indicated otherwise. LB agar plates contained 1.5% agar. Mean diameters were measured from 20 colonies from 2 different LB agar plates from of the respective strain after incubation at 37 °C for 16 h. Growth measurements were conducted in 96 well plates. Therefore, 200 µl LB was inoculated with o/n cultures at an OD 600 0.001. OD 600 was monitored in 10-min intervals for 12 h in a Tecan Spark with 240 rpm at 37 °C. The medium was supplemented with antibiotics at the following concentrations: 50 µg/ml kanamycin sulfate and 30 µg/ml chloramphenicol. Plasmids carrying the P BAD promoter 32 were induced with l-arabinose at a final concentration of 0.2% (w/v).

Molecular biology methods.
Oligonucleotides used in this study are listed and described in the Supplementary Table S1. Plasmid DNA was isolated using the Hi Yield Plasmid Mini Kit from Süd Laborbedarf according to manufacturer's instructions. DNA fragments were purified from agarose gels using the Hi Yield Gel/PCR DNA fragment extraction kit from Süd Laborbedarf. All restriction enzymes, DNA modifying enzymes and the Q5 high fidelity DNA polymerase for PCR amplification were purchased from New England BioLabs and used according to manufacturer's instructions.
Genomic library. The genomic DNA (gDNA) was isolated from 50 ml o/n culture of P putida KT2440 according to the protocol described in reference 54 . Further purification was achieved using Phase Lock Gel (QuantaBio) with Phenol-Chloroform. After the centrifugation, isopropanol precipitation was repeated. The pellet was resuspended in water, the final amount was 60 µg DNA.
Plasmid DNA was purified as described in "Molecular biology methods" from 12 ml E. coli DH5α cells. The plasmid DNA was diluted in water, the final amount was 10 µg DNA.
The library was constructed using SmaI (pBAD33 vector) and StuI (gDNA) for digestion resulting in an average size of 5 kb per insert (Bionexus, Inc.). After ligation, the plasmids were transformed into E. coli DH10 B (Lucigen). Quality control was done by restriction digest of library clones with BamHI. All restriction enzymes were produced by New England Biolabs, Frankfurt. The library was reisolated from E. coli DH10B as described in "Molecular biology methods" and transferred into corresponding reporter strains.