Biosynthesis of UDP-4-keto-6-deoxyglucose and UDP-rhamnose in Pathogenic Fungi Magnaporthe grisea and Botryotinia fuckeliana*

Background: Rhamnose-containing glycans are involved in host-pathogen interactions. Results: UDP-rhamnose was identified in fungi; recombinant enzymes involved in its synthesis were characterized, and the genes involved are expressed in a tissue-specific manner. Conclusion: Fungi containing rhamnose likely utilize the UDP-rhamnose pathway. Significance: Understanding rhamnose pathways in fungi may provide new insight to fungus-host interaction. There is increasing evidence that in several fungi, rhamnose-containing glycans are involved in processes that affect host-pathogen interactions, including adhesion, recognition, virulence, and biofilm formation. Nevertheless, little is known about the pathways for the synthesis of these glycans. We show that rhamnose is present in glycans isolated from the rice pathogen Magnaporthe grisea and from the plant pathogen Botryotinia fuckeliana. We also provide evidence that these fungi produce UDP-rhamnose. This is in contrast to bacteria where dTDP-rhamnose is the activated form of this sugar. In bacteria, formation of dTDP-rhamnose requires three enzymes. Here, we demonstrate that in fungi only two genes are required for UDP-Rha synthesis. The first gene encodes a UDP-glucose-4,6-dehydratase that converts UDP-glucose to UDP-4-keto-6-deoxyglucose. The product was shown by time-resolved 1H NMR spectroscopy to exist in solution predominantly as a hydrated form along with minor amounts of a keto form. The second gene encodes a bifunctional UDP-4-keto-6-deoxyglucose-3,5-epimerase/-4-reductase that converts UDP-4-keto-6-deoxyglucose to UDP-rhamnose. Sugar composition analysis and gene expression studies at different stages of growth indicate that the synthesis of rhamnose-containing glycans is under tissue-specific regulation. Together, our results provide new insight into the formation of rhamnose-containing glycans during the fungal life cycle. The role of these glycans in the interactions between fungal pathogens and their hosts is discussed. Knowledge of the metabolic pathways involved in the formation of rhamnose-containing glycans may facilitate the development of drugs to combat fungal diseases in humans, as to the best of our knowledge mammals do not make these types of glycans.

Necrotrophic and biotrophic fungal pathogens are a major cause of global crop damage. Necrotrophic fungi, including Botryotinia fuckeliana (Botrytis cinerea), attach to the host and then secrete enzymes and toxins that kill host cells prior to pathogen invasion (1). In contrast, biotrophic pathogens, including Magnaporthe grisea (Magnaporthe oryzae), proliferate within living plant tissues and cause tissue damage and yield reduction. M. grisea, which causes rice blast disease (2), is a major threat to the food supply. It is estimated that M. grisea destroys enough rice to feed 60 million people in a given year (3). Likewise B. fuckeliana has an agricultural and economic impact on over 200 plant species (4).
There is increasing evidence that changes in the glycans present on the fungal cell surface play a role during the infection process (5,6). For example, the ability of a fungus to infect its host is reduced by biochemical treatments that impair glycan recognition or remove the glycan moiety (7). Here, we discuss the potential role of the surface residue rhamnose in fungi.
Rhamnose (Rha, 6-deoxy-L-mannose) has been reported to be present in a variety of glycoproteins, exopolysaccharides (EPS), 2 and some minor components of the fungal cell wall. For example, hyphae from the barley pathogen Rhynchosporium secalis contain rhamnomannans (8) that are tightly associated with the cell walls. Other types of rhamnomannans are solubilized by alkali. Rhamnose accounts for at least 30% of the sugars in the rhamnomannan of Cephalotheca purpurea and Cephalotheca reniformis (9) and less than 2% in Penicillium chrysogenum and Ophiostoma ulmi (10,11).
In addition, Cryphonectria parasitica, the causal agent of chestnut blight disease, secretes an EPS composed of a mannan backbone that is decorated with oligosaccharide side chains that are terminated with rhamnose (12). Interestingly, nonvirulent strains of this fungus do not secrete this EPS. The mucilage secreted by fungal spores of the corn pathogen, Colletotrichum graminicola consists of various glycoproteins containing Rha as well as galactose and mannose residues (13). These glycoproteins have been proposed to function as an antidesiccant thereby allowing Colletotrichum conidia to survive periods of drought and changes in temperature (6).
Although the biosynthesis of fungal core wall polysaccharides (e.g. glucan, chitin, and mannan) is relatively well understood, virtually nothing is known about the biosynthesis of Rhacontaining glycans in fungi, nor is the identity of the activated form of rhamnose used in the synthesis of these glycans known.

EXPERIMENTAL PROCEDURES
Expression and Purification of Recombinant UG4,6-Dh and U4k6dG-ER-BL21-derived E. coli cells containing the M. grisea and B. fuckeliana recombinant UG4,6-Dh and U4k6dG-ER in pET28bTev plasmids or a control empty vector (pET28b) were cultured for 16 h at 37°C in LB medium (20 ml) supplemented with kanamycin (50 g/ml) and chloramphenicol (35 g/ml). A portion (5 ml) of the cultured cells was transferred into fresh LB liquid medium (250 ml) supplemented with the same antibiotics and then grown at 37°C at 250 rpm until the culture had an A 600 of 0.8. Gene expression was then induced by the addition of isopropyl ␤-D-thiogalactoside (0.5 mM), and the cells were then grown for 20 h at 18°C at 250 rpm. Cells were harvested by centrifugation (6,000 ϫ g for 10 min at 4°C) and suspended in lysis buffer (10 ml). For UG4,6-Dh, 50 mM Tris-HCl, pH 7.5, containing 150 mM NaCl, 10% glycerol, and 1 mM EDTA was used. The same buffer, pH 9, supplemented with 2 mM DTT and 0.5 mM phenylmethylsulfonyl fluoride was used for U4k6dG-ER. Cells were lysed in an ice bath with a Misonix S-4000 sonicator (Misonix Inc, Farmingdale, NY) equipped with 1/8-inch microtip probe using 12 sonication cycles each (20-s pulse; 30 s off). The lysed cells were centrifuged (6,000 ϫ g for 10 min at 4°C); the supernatant was supplemented with 1 mM DTT and centrifuged again (20,000 ϫ g for 30 min at 4°C). The resulting supernatant (termed S20) was kept at Ϫ20°C. His-tagged proteins were purified on a fast-flow nickel-Sepharose (GE Healthcare, 2 ml of resin packed in a 1-cm inner diameter ϫ 15-cm polypropylene column). For UG4,6-Dh, the column was pre-equilibrated with loading buffer 50 mM sodium phosphate, pH 8 (pH 9 for U4k6dG-ER), containing 0.3 M NaCl and 1 mM DTT. The bound His-tagged protein was eluted with the same buffer containing increased concentrations of imidazole (10 -250 mM). The fraction containing the enzymatic activity, typically eluted with 250 mM imidazole, was dialyzed (6,000 -8,000 molecular weight cutoff, Spectrum Laboratories, Inc.) at 4°C three times against 50 mM Tris-HCl, pH 8 (pH 9 for U4k6dG-ER), containing 0.15 M NaCl, 10% glycerol, 1 mM DTT, 10 M NAD ϩ for 2 h. The dialysate was divided into 0.25-ml aliquots that were flash-frozen in liquid nitrogen and kept at Ϫ80°C.
Proteins extracted from E. coli cells expressing the empty vector were obtained using the same purification protocol and were used as controls in enzyme assays and SDS-PAGE analyses. Protein concentrations were determined using the Bradford reagent with bovine serum albumin (BSA) as standard. The native molecular weight of each recombinant protein was estimated by Superdex 75 size-exclusion chromatography as described (22) using dialysis buffer as eluant.
UG4,6-Dh and U4k6dG-ER Enzyme Assays-UG4,6-Dh reactions (50-l final volume) were carried out in 50 mM sodium phosphate, pH 7.6, containing 1 mM NAD ϩ , 2 mM UDP-Glc, and 13 g of purified recombinant UG4,6-Dh. U4k6dG-ER reactions (60-l final volume) were carried out in 50 mM sodium phosphate, pH 7.6, containing 3 mM NADPH, 1 mM NAD ϩ , 2 mM UDP-Glc, purified recombinant UG4,6-Dh, and purified recombinant U4k6dG-ER (17 g). Reactions were kept for up to 45 min at 30°C and then terminated (100°C water bath, 1 min). Chloroform (50 l) was added; the mixture was vortexed and centrifuged (12,000 rpm, 5 min at 22°C). The upper aqueous phase was collected. Reaction products were chromatographed on an anion-exchange column (Q15 resin in 1-mm inner diameter ϫ 250 mm, Amersham Biosciences; or Mono Q 5 ϫ 50 mm, Amersham Biosciences) eluted with a linear gradient of ammonium formate (5 mM to 0.6 M) (22) over 25 min, using an Agilent 1200 Series HPLC system equipped with an autosampler, diode array detector, and ChemStation software. Nucleotides were detected by their A 261 (maximum for UDP-sugar) and A 259 (maximum for NAD ϩ ). The amount of product formed was determined using a calibration curve of standard UDP-Glc. The products formed by the UG4,6-Dh reaction (eluted at 12.3 min from Q-column) and the U4k6dG-ER reaction (eluted at 13.2 min from the PA1 column) were collected, lyophilized, dissolved in water or 99.9% D 2 O, and analyzed by ESI-MS and by 1 H NMR spectroscopy.
Characterization of Recombinant Enzymes-UG4,6-Dh activities were determined using different buffers, different temperatures, and in the presence of selected cations or potential inhibitors. For pH studies, solutions of recombinant UG4,6-Dh (13 g) were mixed in each individual buffer (50 mM). Subsequently, NAD ϩ (1 mM) and UDP-Glc (2 mM) were added, and reactions were carried out for 30 min at 30°C. Inhibition assays were performed by first mixing the enzyme and buffer with different additives (e.g. nucleotides) on ice for 10 min. UDP-Glc and NAD ϩ (1 mM each) were then added. After 30 min at 30°C, the amount of UDP-sugar formed was determined by 1 H NMR spectroscopy.
Real Time 1 H NMR-based Enzyme Assay-Real time 1 H NMR spectroscopic monitoring of enzymatic reactions (150 l) was performed at 30°C in a mixture of D 2 O/H 2 O (9:1 v/v), containing 50 mM sodium phosphate at various pH values, and 1-1.5 mM UDP-Glc, 0.6 mM NAD ϩ , and 13 g of recombinant UG4,6-Dh. The combined UG4,6-Dh and U4k6dG-ER NMR assay used the same buffer supplemented with 1.5 mM NADPH. Real time 1 H NMR spectra were obtained using a Varian VNMRS spectrometer operating at 600 MHz and equipped with a 3-mm cold probe. Data acquisition was started between 2 and 5 min after the addition of enzyme to allow spectrometer acquisition conditions to be optimized. Sequential one-dimensional proton spectra with presaturation of the water resonance were acquired over the course of the enzymatic reaction, with spectra summed and averaged every 8 s. All chemical shifts (ppm) are referenced to 2,2-dimethyl-2-silapentane-5-sulfonate (␦0.00). 13 C-Enriched UDP-4-keto-6-deoxyglucose was prepared from UDP-[ 13 C]Glc using recombinant Sloppy (UDPsugar pyrophosphorylase (USP)) (21), UTP, and [ 13 C]Glc-1-P (OMICRON Biochemicals, Inc.). Subsequently, the 4,6-dehydratase with or without the U4k6dG-ER was added. The reaction mixtures, which contained UDP-[ 13 C]Glc, UDP-[ 13 C]4k6dG, and UDP-[ 13 C]Rha, were examined by 13 C HSQC and 13 C HMBC NMR experiments using standard Varian pulse programs.
Isolation and Glycosyl Residue Analyses of Fungal Polysaccharides-Freeze-dried spores of fungal strains, stored in Ϫ80°C on filter paper, were grown on Potato Dextrose Broth (B. fuckeliana) or oatmeal agar plates (M. grisea) for 10 -14 days at 22°C; conidia were washed off the agar with 10 ml of sterile water, and the suspension was filtered through eight layers of cheesecloth. The filtered suspension was allowed to settle, and the water above the spores was removed. The spores were resuspended in sterile water and diluted to 2 ϫ 10 5 spores per ml and used to inoculate 500 ml of Potato Dextrose Broth or CM-Suc (B. fuckeliana), Vogel-Suc, or CM-Suc (M. grisea) media. After 7 days of growth at 25°C with shaking (150 rpm) or without shaking, hyphae were collected on Miracloth, and the filtrate (cultured media) was saved. The hyphae ball was washed with water and freeze-dried, and the hyphal powder was weighed. The filtered culture media were concentrated to 100 ml by rotary evaporation and then 2 volumes of ethanol were added to precipitate EPS. The suspension was centrifuged (8,000 ϫ g for 15 min, 4°C), and the pellet was washed with aqueous 95% ethanol, dried, and resuspended in water.
Two extraction methods were used to isolate glycans from the hyphal powder. In the first procedure, the hyphal powder (1 g) was treated sequentially (twice each time) with 8 ml of EtOH (80%), 5 ml of chloroform/MeOH (1:1 v/v), and 4 ml of acetone. The sugar composition of the resulting residue (termed AIR) as well as of each derived organic extract (termed Et, CM, and Ac) was dried by a stream of air and determined. In the second procedure, the hyphal powder (5 g) was treated for 4 h at 22°C while stirring with 1 M NaOH (200 ml) three times. After centrifugation (10,000 ϫ g, 4°C, 10 min) the combined alkaline extracts were mixed with an equal volume of 96% ethanol. The pellet that formed was collected by centrifugation (10,000 ϫ g, 4°C, 15 min), resuspended in water (10 ml), and dialyzed (3,000 molecular weight cutoff) against two 10-liter changes of deionized water. The dialyzed solution was clarified by centrifugation (10,000 ϫ g, 4°C, 15 min) and the water-soluble supernatant (aWS) was collected and freeze-dried (yield ϳ250 mg). The residue remaining after centrifugation (termed aWi) was washed with water, dried, and weighed (ϳ150 mg). The glycosyl residue compositions of these extracts were then determined.
Glycosyl Residue Composition Analysis-Each extract (1 mg) was hydrolyzed for 2 h at 120°C with 2 M TFA (1 ml). The TFA was removed by evaporation under a stream of filtered air (40°C) followed by three cycles of wash with isopropyl alcohol (500 l) and solvent evaporation. The resulting dried neutral monosaccharides were converted into their corresponding alditols in 500 l of reducing solution (10 mg/ml sodium borodeuteride in 1 M NH 4 OH) for 1 h at 22°C. The reaction was quenched with glacial acetic acid (500 l), and the solution was concentrated to dryness. The residue was washed three times with MeOH/OHAc 9:1 v/v (1 ml) and then with MeOH (three times, 1 ml). The alditols were converted to their corresponding alditol-acetate derivatives (23) by treatment for 20 min at 120°C with acetic anhydride (150 l) and pyridine (100 l). The reagents were removed under a stream of air, and the residue was washed with hexane (three times, 1 ml), and then water (1 ml) was added. After sample sonication, the alditol acetates were extracted into dichloromethane (1 ml). The organic phase was concentrated to dryness, and the residue was dissolved in acetone (100 l). Samples were analyzed by gas-liquid chromatography (GLC) equipped with a flame ionization detector or with a mass selective detector (EI-MS). The sample (1 l) was injected in the split mode (1:50 split) using an Agilent 7890A autosampler to a Restek RTx-2330 (or SP-2330, Supelco) fused silica column (0.25 mm inner diameter ϫ 30 m, 0.2-m film thickness) with helium as gas carrier at a flow rate of 1.1 ml min Ϫ1 . The GC program was started with the oven temperature held at 80°C for 2 min, followed by an increase of 30°C min Ϫ1 to 170°C, then at 4°C min Ϫ1 to 235°C, and a hold at 235°C for 20 min. The column was then kept at 250°C for 7 min cooled to 80°C and kept at 80°C for 1 min prior to the next sample injection. The injection port and the flame ionization detector were kept at 250°C. GC-MS was performed with an HP 5973 MSD. Alditol-acetate derivatives of monosaccharide standards (50 g each of rhamnose, fucose, ribose, arabinose, xylose, mannose, glucose, and galactose) were prepared under the same condi-tions as samples. Myoinositol (20 g) was added as internal standard. For data analysis, sugar residue peaks were identified based on retention times of standard monosaccharides and their characteristic mass spectrum. The raw peak areas, obtained from the total ion chromatogram, were exported to Microsoft Excel and normalized to sample mass and the internal standard. The amount of individual residue in samples was calculated based on calibration curves of standards.
Nucleotide Sugar Extraction and Analyses-Fungal tissues (hyphae grown in 100 ml of CM media for 7 days at 25°C; spores from CM-agar) were collected on a filter, washed with water, and weighed. Cold 0.2 M sodium fluoride (1 ml) was added per 0.5 g wet weight cells. The suspension was vortexed for 1 min, then 6 volumes of cold chloroform/MeOH (1:1 v/v) was added, and the mixture was kept for 30 min at 4°C during which the sample was vortexed for 20 s every 3 min. The suspension was centrifuged (1,000 ϫ g for 4 min), and the aqueous layer was transferred to a new tube and centrifuged again to remove debris. Portions (30 l) were injected via HPLC at 0.25 ml min Ϫ1 to a Q15 column (1 mm inner diameter ϫ 250 mm) pre-equilibrated with 20 mM ammonium formate, and separa-tion was achieved over a 25-min gradient by increasing the concentration of ammonium formate (22). Column fractions (0.4 ml) were collected, and a portion (350 l) was lyophilized and then dissolved in D 2 O (150 l) for 1 H NMR analysis. The Expression of UDP-Rha Biosynthetic Genes-TRIzol reagent (Invitrogen) was used to isolate total RNA from a 14-day-old sporulated culture of M. grisea grown at 24°C on oatmeal agar (20), from 10-day-old mycelium grown on CM (20), and from 14-day-old spores of B. fuckeliana grown at 24°C on potato dextrose agar, and mycelium grown in PD broth. RNA (0.5 g) was reverse-transcribed with oligo(dT) as a primer using RTsuperscript III (Invitrogen). The cDNAs were used to amplify the coding sequences of corresponding UDP-Rha biosynthetic genes using Taq-DNA polymerase (Promega), and primers are listed in supplemental Table S1. For transcript control, primers to amplify Ces1 and Arg2 (20) were used. corresponding to UMP. Interestingly, the relative abundance of the UDP-Rha ion was higher than UDP-Glc in spores. To determine the genes involved in the synthesis of the rhamnoselinked glycans, we decided to first identify those that may be involved in the formation of the activated rhamnose, as these were not previously described to exist in fungi.

Identification
Identification, Cloning, and Biochemical Characterization of Fungal UG4,6-Dh-To identify fungal genes involved in the formation of activated rhamnose, we performed BLAST searches using amino acid sequences from bacterial and plant proteins known to be involved in this process. BLAST searches using bacterial rmlB (ADR74235 and BAS1138) and plant NDP-Rha synthase (At1g53500, URS) identified potential homologs in M. grisea and B. fuckeliana. The M. grisea homolog shares 37% amino acid sequence identity to a functional bacterial rmlB, dTDP-Glc 4,6-dehydratase (25,26), from Salmonella, Streptococcus, and 42% identity to the N-terminal portion of plant URS, Rhm (supplemental Fig. S3A) (18,19). Such sequence identity was not sufficient to establish the function of the gene or determine whether the encoded protein utilizes dTDP-Glc or UDP-Glc. Thus, we cloned and expressed these genes in E. coli and then determined the enzymatic properties of the recombinant proteins.
Distinct protein bands (ϳ52.5 and 52 kDa) were detected by SDS-PAGE of extracts from E. coli cells overexpressing recombinant UG4,6-Dh from B. fuckeliana and M. grisea, respectively (Fig. 3A, lanes 1 and 2), but were absent in E. coli cells express-ing the empty vector. The expressed UG4,6-Dh proteins were purified (Fig. 3A, lanes 4 and 5) and shown, using a HPLC-based assay, to convert UDP-Glc in the presence of NAD ϩ to a new UDP-sugar. The newly formed UDP-sugar eluted from the ionexchange column with a retention time of 12.3 min, although its substrate UDP-Glc eluted at 12.2 min (Fig. 3B, panel #2, marked by right arrow). Somewhat unexpectedly, the negative ion ESI mass spectrum of this material contained two ions, a major ion at m/z 565.04 and a less prominent ion at 547.06 (supplemental Fig. S4A). The ion at m/z 547.06 likely corresponds to UDP-4-keto-deoxyhexose (supplemental Fig. S4B). MS-MS analysis of the ion at m/z 565.04 gave an ion at m/z 323.77 (supplemental Fig. S4C), consistent with UMP, and suggested that this component was also a UDP-sugar. The masses of the two ions differed by 18 mass units suggesting a difference in hydration. However, the structure of the component corresponding to the ion at m/z 565.04 could not be determined from the MS data alone. Thus, the material eluting at 12.3 min was characterized by NMR spectroscopy (Fig. 4 and supplemental Table S2).
One-dimensional proton NMR analysis indicated that the enzymatic product UDP-4-keto-6-deoxyglucose exists in two forms (Fig. 4). The predominant component is the hydrated form and a minor component the keto form. The anomeric region of the 1 H spectrum contains two quadruplet signals with chemical shifts of ␦5.73 ppm for the keto and ␦5.54 ppm for the hydrated form (Fig. 4). The distinct chemical shifts (supplemental Table S2) of H1Љ and the coupling constant values of 3.7 Hz for J H1Љ, H2Љ and 7.0 Hz for J H1Љ,P are consistent with an ␣-linkage to the phosphate of UDP. The chemical shifts for the H6Љ (1.265 and 1.215 ppm for keto and hydrate, respectively), and H5Љ of the hydrate (4.09 ppm) are consistent with UDP-4keto-6-deoxyglucose (19). The 1 H NMR spectrum obtained at 45°C showed the H5Љ signal (4.74 ppm) of the keto form that is hidden by the water peak at lower temperatures. Correlations in a two-dimensional TOCSY spectrum confirmed the assignments of H2Љ, H3Љ, and H5Љ in both species. Like H5Љ, H3Љ adjacent to the 4-keto carbon is shifted downfield to 4.64 ppm. The J H2Љ,H3Љ large coupling constant of 10.2 Hz is consistent with retention of the ␣-gluco-configuration. As expected, no proton signal was observed for C4Љ; thus 13 C NMR was used to confirm the presence of a 4-deoxy carbon. An HSQC experiment with UDP-[ 13 C]4k6dG confirmed the results from the proton data (see supplemental Table S2), and an HMBC experiment gave a cross-peak at 95 ppm between the 6-deoxy protons (here split  (Fig. 3B, panel #2, marked by right-most arrow) corresponding to the product formed by UG4,6-Dh was collected and analyzed by 1 H NMR spectroscopy at 600 MHz. A portion of the one-dimensional NMR spectrum between 1.18 and 1.30, 2.5 and 4.5, and 4.9 and 8.1 ppm is shown. H denotes signals for uracil (Ura) protons, HЈ are ribose (Rib) protons, and HЉ are protons from 4-keto-6-deoxyglucose (U4k6dG_A hydrated and U4k6dG_K keto form). Note the diagnostic H-1Љ (anomeric) and H-6Љ signals for both major and minor products. * indicates HPLC column impurities.
by J C6,H6Љ ) and C4 of the hydrated form (Fig. 5B). The expected cross-peak between H6Љ and the 4-carbonyl of the keto form was not observed in this spectrum possibly due to its low concentration. The formation of two U4k6dG molecular ions (hydrated and keto) is also observed when real time NMRbased enzymatic reactions were carried out (see H1Љ anomeric region Fig. 6, A, and H6Љ protons region in B).
Our NMR data provide an explanation for the presence of the two molecular ions observed by ESI-MS. The major ion at m/z 565.04 originates from the hydrated species of U4k6dG and the minor ion at m/z 547.06 originates from the keto form of U4k6dG. Collectively, our data provides compelling evidence that the fungal gene encodes a protein with 4,6-dehydratase activity that converts UDP-Glc to UDP-4-keto-6-deoxyglucose (U4k6dG).
We then used time-resolved 1 H NMR to determine whether the UG4,6-Dh also converts dTDP-Glc to its 4-keto form (supplemental Fig. S5, A and B). However, this enzymatic reaction is much less efficient than the conversion of UDP-Glc to U4k6dG (compare Fig. 6A and supplemental Fig. S5A). Two dTDP-4k6dG molecular species were also observed in solution with chemical shifts for the H6Љ 1.216 and 1.258 ppm (supplemental Fig. S5B) for keto and hydrate, respectively. H5Љ of the hydrate (4.09 ppm) is consistent with dTDP-4-keto-6-deoxyglucose. Two forms of dTDP-4-keto-6-deoxyglucose have also been reported to exist in solution (14).
We identified several UG4,6-Dh homologs when we compared the M. grisea protein sequence with EST databases of transcribed fungal genes or with predicted ORFs from fungal genomes (see supplemental Fig. S3A). However, no comparable genes were identified in Saccharomyces cerevisiae, Schizosaccharomyces, and Cryptococcus neoformans indicating that not all fungi form UDP-4-keto-6-deoxyglucose. Phylogenic analysis (supplemental Fig. S3B) suggests that fungal UG4,6-Dh encoding genes are clustered in a distinct clade and are separated from other eukaryotic homologs. Moreover, fungi that affect plants and humans have UG4,6-Dh genes that cluster in separate clades.
Protein sequence alignment (supplemental Fig. S3A) showed conserved regions between the fungal UDP-Glc 4,6-Dh proteins, a plant UDP-Glc 4,6-Dh protein (22,23), and bacterial dTDP-Glc 4,6-Dh (rmlB) whose structure is known (16,27). The bacterial proteins belong to the short chain dehydrogenase/reductase family, and contain the tyrosine-dependent oxidoreductase sequence 230 YXXXK 234 (Tyr 167 in Salmonella). This sequence is believed to facilitate abstraction of the H4Љ hydride to NAD ϩ and, in concert with 194 TDE 196 , promote the oxidation of C-4Љ, the elimination of water, and the transfer of the hydride from NADH to C-6Љ (boldface in supplemental Fig.  S3A). Some of these amino acid sequences are also present in other UDP-sugar 4-epimerases and UDP-xylose synthase (28). The conserved 51 GXXGXX(G/A) is likely involved in the cofactor-binding site. In some of the enzymes, including UDP-xylose synthase, the NAD ϩ is tightly bound. By contrast, the fungal UG4,6-Dh loses the cofactor during chromatography and does not react with UDP-Glc unless NAD ϩ is added (data not shown).
The M. grisea and B. fuckeliana homologs were cloned, expressed, and their specific activities determined. A 36.8-and a 36.7-kDa protein was purified from E. coli strains overexpressing the recombinant protein from M. grisea and B. fuckeliana, respectively (Fig. 7A). No new products were detected when the M. grisea protein was reacted with U4k6dG alone (data not shown). However, when the reaction contained U4k6dG and NADPH, two new products were identified by HPLC (Fig. 7B, panel #2). The first peak at 11.5 min had a maximum UV absorbance at 259 nm and a retention time identical to NADP ϩ . The ESI mass spectrum of the second peak (12.2 min) contained a major ion at m/z 549.04 (supplemental Fig. S7A). Collision-induced fragmentation gave an ion at m/z 323.47 (supplemental Fig. S7B) indicative of UMP. Together, these data are consistent with the formation of a UDPdeoxyhexose.
NMR spectroscopic analysis of the enzymatic product ( Fig. 8  and supplemental Table S3) showed the presence of an anomeric proton with a chemical shift at ␦5.21 ppm and coupling FIGURE 5. HMBC analysis confirms 13 C-labeled UDP-4-keto-6-deoxyglucose in hydrate form. UDP-[ 13 C]Glc was reacted with UDP-Glc-4,6-dehydratase, and the product was analyzed by two-dimensional HMBC NMR. A, anomeric H-1Љ of the major product U4k6dG_A (hydrate) is coupled to 13 C-2Љ. B, H-6Љ of the hydrate product is coupled to 13 C-5Љ and 13 C-4Љ. No signal was observed for the minor product, U4k6dG_K, as concentration was too low. The corresponding one-dimensional 1 H spectrum is displayed above each magnified region. smaller than 2 for J H1Љ,H2Љ and 8.8 Hz for J H1Љ,P . This small coupling (Ͻ2) explains the apparent doublet anomeric signal rather than a quadruplet signal. Two-dimensional HMBC analysis of 13 C-labeled UDP-␤-L-rhamnose shows that the anomeric H-1Љ of UDP-Rha is coupled to 13 C-2Љ and that the H-6Љ of UDP-Rha is coupled to 13 C-5 (supplemental Fig. S8, A and B).
The chemical shifts and couplings of these signals are indicative of 13 C-labeled rhamnose (L-6-deoxymannose). To confirm a ␤-linkage in UDP-Rha, a 13 C HSQC NMR observation was carried over to determine the coupling between C1Љ and H1Љ.
Consistent with a ␤-linkage was the C-1 chemical shift of ␦95.51 and J C1Љ,H1Љ coupling of 161 Hz. The small chemical  Tables S2 and S3).
Real Time NMR Assays of UG4,6Dh and U4k6dG-ER Activity-The enzymatic progression of the dehydratase and the epimerase/reductase was examined to determine whether other products were formed. To this end selected regions (H-1Љ and H-6Љ) of the 1 H NMR spectra were analyzed (Fig. 6, A and B). The characteristic NMR signal having a quadruplet peak signal that corresponds to the anomeric proton of UDP-Glc (G) (labeled G H-1Љ at 5.58 ppm in Fig. 6A decreased, with the concomitant increase in a quadruplet signal corresponding to A H-1Љ of UDP-4-keto-6-deoxyglucose hydrated from U4k6dG_A). The anomeric proton of the keto form of U4k6dG_K H-1Љ (5.73 ppm) is also increased during the time course of the UG4,6-Dh reaction. No change was observed for the signals from NAD ϩ (N) protons (supplemental Fig. S9A). Clear differences in the chemical shifts for the proton attached to C-6 of glucose were observed during the formation of U4k6dG_A and U4k6dG_K from UDP-Glc (Fig. 6B, marked by  A H-6Љ and K H-6Љ, respectively).
The real time 1 H NMR spectra of the products formed by the bifunctional enzyme (see Fig. 9) revealed that the H-1Љ signals of U4k6dG (A and K) decreased with a concomitant increase in the Rha H-1Љ signal of UDP-Rha. The appearance of signals with chemical shifts corresponding to H-5Љ (3.43 ppm) and H-4Љ (3.35 ppm) and the diagnostic C-6 methyl group of the rhamnosyl moiety provide additional evidence for the formation of UDP-Rha. In addition, the real time 1 H NMR established that NADPH is converted to NADP ϩ during the C-4Љ reduction (supplemental Fig. S9).
Collectively, the real time NMR-based enzyme assay provides compelling evidence for the transformation of UDP-Glc to UDP-Rha via a UDP-4-keto-6-deoxyglucose intermediate. We next have tried to address if the 3,5-epimerization reaction occurs independently of the 4-reduction. The U4k6dG-ER was incubated with UDP-4-keto-6-deoxyglucose in the absence of NADPH, and the reaction was followed by NMR. However, no distinct chemical shift belonging to proton 2Љ or 3Љ was observed to indicate 3,5-epimerization. But when NADPH was added, UDP-Rha was formed as well as NADP ϩ . This suggests that the epimerization occurs, although the intermediate (UDP-4-ketorhamnose) is bound to the enzyme and not released.
Enzymatic Properties of Dh and ER-UG4,6-Dh had maximum activity between 25 and 30°C and was active across a wide pH range, with maximal activity at neutral pH (supplemental Fig. S10, A and B). Similar pH and temperature enzymatic activity profiles were observed with U4k6dG-ER (supplemental Fig.  S11, A and B). The UG4,6-Dh activity was eluted from size-exclusion column with mass of 140,000 suggesting the fungal enzyme is active predominantly as a dimer-trimer (supplemental Fig. S12). Our recombinant UG4,6-Dh preparation included for long term storage stability the addition of 10 M NAD ϩ . The fungal U4k6dG-ER, however, eluted from Superdex75 size-exclusion column at the void volume but was enzymatically inactive. Chromatography with different buffers failed to elute active U4k6dG-ER. Hence, the elution of U4k6dG-ER as a large mass may not reflect the size of active U4k6dG-ER in solution. It is possible that U4k6dG-ER aggregates on the gel, explaining, in part, no activity. UG4,6-Dh requires NAD ϩ for complete conversion of UDP-Glc to product. U4k6dG-ER requires NADPH for the reduction of U4k6dG to UDP-Rha. NADP ϩ did not substitute NAD ϩ for the dehydrogenation reaction nor did NADH substitute for NADPH for the reduction reaction. The latter result confirms earlier reports with plant NRS/ER (19) and viral U4k6dG-ER (29) and suggests a strict recognition of the U4k6dG-ER enzymes for the phosphoryl group O-CЈ2-linked to ribosyl group of the adenine moiety in NADPH. Substrate specificity of  6). B, HPLC analysis of the products formed by U4k6dG-ER reacting with UDP-4-keto-6-deoxyglucose (U4k6dG). U4k6dG-ER was reacted with U4k6dG for 30 min in the presence of NADPH (panel #2). The purified protein isolated from cells expressing control empty vector was incubated with U4k6dG, and NADPH for 45 min (panel #3) is a control. The distinct UDP-sugar peak marked by the longest arrow (panels 2 and 3 with the same retention time 12.1) was collected and analyzed by 1 H NMR spectroscopy and ESI-MS spectrometer.
UG4,6-Dh indicates that the fungal enzyme is unable to convert ADP-Glc, GDP-Glc, or UDP-GlcNAc to a product (data not shown). Although UG4,6-DH is able to convert dTDP-Glc to the corresponding dTDP-4-keto-6-deoxyglucose keto and hydrate forms, the catalytic efficiency (k cat /K m ) is 0.0027 M s Ϫ1 , ϳ400 times lower than that for UDP-Glc. Inhibition studies determine that UDP and to a lesser degree UTP are inhibitors of UG4,6-Dh. The UG4,6-Dh was not inhibited however by NADH. UTP had a similar adverse activity with U4k6dG-ER (supplemental Tables S5 and S6). Kinetic analyses of UG4,6-Dh and U4k6dG-ER are shown in supplemental Table S7, A and B.
Further synergistic reactions containing both enzymes, NAD ϩ and NADPH showed that UG4,6-DH is strongly inhibited, whereas U4k6dG-ER is minimally inhibited, possibly because of NAD ϩ . The inhibition of UG4,6-DH was likely due to UDP-Rha because in separate experiments 80% of UG4,6-DH activity was inhibited in the presence of 1 mM UDP-Rha (supplemental Table S5).
Transcripts and Glycosyl Analyses-The transcripts encoding UG4,6-Dh and U4k6dG-ER were detected in M. grisea spores but not in 10-day-old mycelium (Fig. 10A, compare  lanes 4 and 5 and 9 and 10). Glycan isolated from M. grisea (Fig.   10B) shows that spores consisted of Rha, Gal, and Man residues, but mycelium and the EPS fractions consisted of Man and Gal; and no Rha residue was found. Unlike M. grisea, the UDP-Rha biosynthetic genes of B. fuckeliana were expressed in both spores and mycelium (Fig. 10C, lanes 11-14). Rha-containing glycans were also detected in these tissues as well as in an extracellular glycan fraction, EPS (Fig. 10D).

DISCUSSION
Our study is the first to provide evidence for the existence of UDP-Rha in fungi and the identification of two enzymes required to form this activated sugar from UDP-Glc in M. grisea and B. fuckeliana.
We have shown that both M. grisea and B. fuckeliana synthesize rhamnose-containing glycans. Preliminary data 3 suggest that rhamnose-containing glycans are secreted 2 h after M. grisea spores germinate on artificial substrates. This glycan may protect the spore from desiccation and facilitate adherence to host, as has been suggested for the germination of Colletotrichum graminicola conidia (6,13). Interestingly, the rhamnose-3 J. Smith and M. Bar-Peled, unpublished data. containing glycans accumulate differentially in M. grisea. Similarly, the genes encoding enzymes involved in UDP-Rha are also expressed in a tissue-specific manner.
Unlike M. grisea, the UDP-Rha biosynthetic genes of B. fuckeliana are expressed in both spores and mycelium. Likewise, Rha-containing glycans were also detected in these tissues as well as in an extracellular fraction. The ratio of Rha-containing glycans varies among the fungal tissues, and further detailed structural analyses will be required. Nonetheless, the display of Rha-containing glycans correlates with the expression of the genes for the synthesis of UDP-Rha. This suggests that the formation of UDP-Rha and the synthesis of Rha-containing glycans in fungi could be regulated in a tissue-specific manner. What molecular mechanism is involved in such expression is currently under investigation.
Analysis of the literature describing rhamnose-containing glycans in other fungal species indicates that these glycans are also accumulated in a tissue-specific manner or in response to growth conditions. For example, the extracellular aspartic proteinase from the zygomycete fungus Phycomyces blakesleeanus is a glycoprotein that contains mannose and rhamnose (30). Secretion of this glycoprotein occurs only at later stages of growth and can be modified by the nutrient composition of the culture media. In contrast, a Rha-containing glycoprotein was reported to be secreted only when C. graminicola spores attach to the host surface (13). There is also evidence that different types of rhamnose-containing polysaccharides accumulate during development of Hericium erinaceus fruiting bodies (31). One of the Hericium polysaccharides contains Rha linked to a fucogalactan (31), in a second the Rha is 3-O-methylated, and in a third, Rha exists in a rhamnoglucogalactan (32). Rhamnose has also been detected in the mature fruiting body of the black Perigord truffle (Tuber melanosporum) but not in other stages of fungal growth (33). Taken together, synthesis of rhamnosecontaining glycans (glycoproteins or polysaccharides) in fungi appears to be regulated.
Some fungi that infect humans also have rhamnose-containing glycans. For example, Sporothrix schenckii, the etiological agent of subcutaneous mycosis disease, secretes a glycoprotein that is heavily glycosylated. The N-linked glycan portion of this glycoprotein is composed of mannose and rhamnose and is believed to be involved in the adherence of the fungus to dermal tissue (34). The opportunistic human pathogen Candida albicans, the causal agent of candidiasis, when grown under conditions that favor biofilm production secretes a large exopolysaccharide composed of rhamnose, mannose, glucose, and glucosamine (35). However, under normal growth conditions, the yeast and hyphal forms lack rhamnose. In the human pathogen Fonsecaea pedrosoi (36), the appearance of rhamnose-containing glycans is also under tissue-specific regulation as rhamnose is predominant in conidia, whereas galactose is predominant in mycelia.
Based on genomic DNA sequence data, fungal species that have Rha-containing glycans also carry genes involved in UDP-4-keto-6-deoxyglucose and UDP-rhamnose synthesis. For example, S. schenckii, Candida, C. graminicola, and T. melanosporum have genes with homology to the M. grisea and B. fuckeliana genes identified in this report (as shown in supplemental Figs. S3B and S6B). Interestingly, the UDP-rhamnose biosynthetic genes of plant fungi are closer by phylogeny analysis to the plant species, when compared with the animal fungal biosynthetic genes. Of note is the clustering of the two viral UGlc4,6-Dh proteins (29). The chlorella virus co-living with plant appears to cluster with UGlc4,6-Dh in plants, although the same protein from the giant DNA virus that infects members of the genus Acanthamoeba is clustered with Trypanosoma (supplemental Fig. S3B). It would be of interest in the The reaction was carried out at 30°C for 35 min. The selected region for the diagnostic anomeric protons of reactant and product is shown between 5.1 and 5.8 ppm. Note the appearance of a diagnostic quadruplet signal is a "doublet" due to very small coupling constant between phosphate and H1Љ.
future to look at many viral genomes and determine the role of UDP-Rha in these virus-host interactions.
As many as 200 fungi are known to impact human health (37). As far as we are aware, humans do not make Rha-containing glycans; thus, the metabolic pathways involved in the formation of rhamnose-containing glycans provide a potential target for drugs to control some fungal diseases in humans as well as other animals.
Additional research is required to structurally characterize the Rha-containing glycans synthesized by M. grisea and B. fuckeliana. Such information together with data obtained after reduction or elimination of these two UDP-Rha biosynthetic genes, will further illuminate the roles of rhamnose-containing glycans in fungi.