A four-gene operon in Bacillus cereus produces two rare spore-decorating sugars

Bacterial glycan structures on cell surfaces are critical for cell-cell recognition and adhesion and in host-pathogen interactions. Accordingly, unraveling the sugar composition of bacterial cell surfaces can shed light on bacterial growth and pathogenesis. Here, we found that two rare sugars with a 3-C-methyl-6-deoxyhexose structure were linked to spore glycans in Bacillus cereus ATCC 14579 and ATCC 10876. Moreover, we identified a four-gene operon in B. cereus ATCC 14579 that encodes proteins with the following sequential enzyme activities as determined by mass spectrometry and one- and two-dimensional NMR methods: CTP:glucose-1-phosphate cytidylyltransferase, CDP-Glc 4,6-dehydratase, NADH-dependent SAM:C-methyltransferase, and NADPH-dependent CDP-3-C-methyl-6-deoxyhexose 4-reductase. The last enzyme predominantly yielded CDP-3-C-methyl-6-deoxygulose (CDP-cereose) and likely generated a 4-epimer CDP-3-C-methyl-6-deoxyallose (CDP-cillose). Some members of the B. cereus sensu lato group produce CDP-3-C-methyl-6-deoxy sugars for the formation of cereose-containing glycans on spores, whereas others such as Bacillus anthracis do not. Gene knockouts of the Bacillus C-methyltransferase and the 4-reductase confirmed their involvement in the formation of cereose-containing glycan on B. cereus spores. We also found that cereose represented 0.2–1% spore dry weight. Moreover, mutants lacking cereose germinated faster than the wild type, yet the mutants exhibited no changes in sporulation or spore resistance to heat. The findings reported here may provide new insights into the roles of the uncommon 3-C-methyl-6-deoxy sugars in cell-surface recognition and host-pathogen interactions of the genus Bacillus.

Bacterial cells produce myriad glycan structures that are secreted or attached to cell surfaces. Often, being at the outermost surface of the cells, glycans have significant impact on cell-cell recognition, cell-to-cell adhesion, and host-pathogen interactions (1)(2)(3)(4). Glycans also play a profound role in shaping the cell (5). In recent years it has also been recognized that spore-producing bacteria like members of the Bacillus cereus sensu lato group are decorated with glycan structures (6,7). Although very few of these glycans have been studied, they are found as glycoproteins of the spore exosporium, the outermost surface layer surrounding the spore coat. Among the exosporium glycoproteins that have been characterized are ExsH and the collagen-like protein BclA (8,9). The BclA protein plays a central role in Bacillus anthracis pathogenesis by promoting the interaction of spores with the host phagocytic cell. This is proposed to facilitate transport of the spores to sites of spore germination and bacterial outgrowth (10,11). These two glycoproteins characterized in the B. cereus ATCC 14579 exosporium have different sugar composition and glycosylation patterns (12). The reason for the presence of several surface glycoproteins differently glycosylated on B. cereus ATCC 14579 exosporium remains unknown, but various glycosylation patterns suggest specific recognition of different cell receptors.
Genome sequence analyses have revealed the presence of at least 13 different collagen-like protein-encoding genes, and proteomic studies indicate that some of them are in the exosporium (13). Whether these are also glycoproteins with glycans that differ in their sugar composition and sequence remains unknown. This is a critical feature for study because spore glycoepitopes may explain host recognition and strain specificity. Thus, using a systematic screen to identify sugar-containing polymers in spores, we have identified by GS-MS analyses new unknown sugar-like residues in spore glycans.
Here we have provided the evidence of two rare sugars, 3-Cmethyl-6-deoxyhexoses, displayed on Bacillus spores. We show that the activated forms of the sugars are CDP-3-C-methyl-6-deoxyhexose isomers, as illustrated in Fig. 1. The formation of the C-methyl-sugar-containing glycans on spore surface occurs when cells are induced to grow in a sporulation medium. different growth environments influence glycan composition. Furthermore, some sugar residues are labile to common harsh hydrolytic conditions (high temperature and high concentration of acid).
In this study, we examined sugar composition under different hydrolytic conditions using B. cereus ATCC 14579 cells. In addition, B. cereus was grown in different media because its glycans change as the medium is altered (14). Using this approach, we detected two unknown sugar peaks (labeled B and C in Fig. 2A) by GC-MS analysis. The EI-MS 2 spectra of peaks B and C ( Fig. 2A) showed identical fragment ion patterns, suggesting that they are isomers. Calculations of the potential structure of the alditol-acetate derivatives implied that peaks B and C ( Fig. 2A) are sugar residues, predicted to be 3-C-methyl-6-deoxyhexoses. The GC chromatograms and MS fragmentation patterns of such structures have not been reported in Bacillus. The peaks at m/z 245 and 231 ( Fig. 2A) suggest a cleav-age between C3 and C4; the peaks at m/z 189, 129, and 87 as well as m/z 203 and 143 are likely secondary ion fragments due to the loss of m/z 42 (ketene) or 60 (acetate) from m/z 231 and 245, respectively. Interestingly, hydrolyses of samples at a very low molarity of trifluoroacetic acid (TFA), HCl, and acetic acid gave different sugar composition profiles. However, no peaks for the B and C sugar residues ( Fig. 2A) were observed when samples were hydrolyzed with 2 or 4 M TFA, suggesting that they are labile.

In-microbe analyses of genes of the C3CM operon in Escherichia coli show synthesis of CDP-3-C-methyl-6-deoxyallose
Of the two experimental approaches used to determine the function of genes within the C3CM operon, the first one is named "in-microbe." The in-microbe experimental setup consisted of a combination of up to four recombinant genes of the C3CM operon that were co-transformed into E. coli. After isopropyl-1-thio-␤-D-galactopyranoside (IPTG) gene induction, the NDP-sugars were directly extracted from the cell and separated by HILIC column. Chromatographic peaks were further purified and analyzed by NMR. E. coli induced to express the first gene, BC3358, gave a distinct peak (  (Fig. 3Av). Subsequently, the E. coli (Bc3358)-produced NDP-sugars were purified by HPLC and analyzed by NMR spectroscopy. The NMR data confirmed that the m/z 564 peak was CDP-Glc (supplemental Fig. S1). The chemical shift assignments for CDP-Glc are summarized in Table 1. The large J H2Љ,H3Љ and J H3Љ,H4Љ coupling constants of 9.9 and 9.8 Hz, respectively, are consistent with a sugar in gluco-configuration. The chemical shift of 5.59 ppm of the anomeric proton and J H1Љ,H2Љ coupling constants of 3.4 Hz are consistent with an ␣-linkage. This suggests that enzyme Bc3358 is a CTP:glucose-1-phosphate cytidylyltransferase.
E. coli induced to express both BC3358 and BC3359 gave a broad peak eluting from an HILIC column between 16.5 and 18 min. This peak was characterized as CDP-4-keto-6-deoxyglucose (CDP-4k-6d-Glc) ( Fig. 3Aii) with an ion [M-H] Ϫ at m/z 546 that gave by MS/MS an ion fragment of m/z 322 (Fig. 3Bii). This suggests that BC3359 encodes a CDP-glucose 4,6-dehydratase. Similarly, E. coli induced to express three Bacillus genes (BC3358, BC3359, and BC3360) produced yet another broad peak with a retention time (14 -15.5   2. Two uncommon 3-C-methyl-6-deoxyhexose sugar residues were detected upon mild hydrolysis of spore glycans from Bacillus. A, GC-MS of alditol-acetate derivatives of sugars released by mild acid hydrolyzes from spore glycans of B. cereus ATCC 14579 grown in Msgg medium. The two novel sugar peaks eluted at 22.7 and at 23.3 min and are labeled B and C, respectively. *, 6-deoxyhexose peak. The EI-MS fragmentation pattern of peaks B and C with their major m/z ion fragments is shown in the right spectrum; and the boxed insert shows the predicted primary and secondary MS fragments (labeled in bold) of the CDP-3-C-methyl-6-deoxyhexose alditol-acetate derivative. B, GC-MS analyses of cereose and cillose standards show that peak B is cereose and peak C is cillose. Standards were obtained from mild hydrolysis of corresponding CDP-3-C-methyl sugars purified from an in-microbe-based assay (see Fig. 4 for CDP-cillose) and after an enzymatic assay (see Fig. 8 for CDP-cereose). Alditol-acetatesugar derivative peaks of ribose and arabinose are also shown. C, structures of cereose and cillose. three combined Bacillus genes yielded a C-methylated CDP-4keto-6-deoxyglucose. The MS/MS of the product (m/z 560) gave an ion fragment at m/z 237 (Fig. 3Biii), consistent with a methyl modification of the 4-keto sugar-1-P [M-H 2 O] Ϫ . Hence, we tentatively suggest that BC3360 encodes a CDP-4keto-6-deoxyglucose C3-methyltransferase. Lastly, when all four genes were co-expressed in E. coli, a new peak was observed ( Fig. 3Aiv) with an ion [M-H] Ϫ at m/z 562, which gave MS/MS ion fragments of m/z 322 and 239 (Fig. 3Biv). This increase of 2 atomic mass units indicates that the 4-keto moiety of the substrate is reduced. Such a peak was not detected in the E. coli control. Hence, the last gene (BC3361) when expressed in E. coli suggested encoding a CDP-3-C-methyl-4-keto-6-deoxyhexose 4-reductase.

One-and two-dimensional NMR experiments reveal the chemical structure of CDP-3-C-methyl-6-deoxyallose
Column purification of the final four-gene product (CDP-3-C-methyl-6-deoxyhexose) produced by the in-microbe experiment (Fig. 3Aiv, see peak point indicated by arrow) was analyzed by NMR spectroscopy (Fig. 4A). The one-dimensional NMR spectrum (Fig. 4B) shows protons H5 and H6 belonging to the Cyt base ring and anomeric signals belonging to ribose H1Ј (5.98 ppm) and 3-C-methyl-6-deoxyallose H1Љ (5.52 ppm), as well as ring protons (3-4.5 ppm) and the diagnostic H5Љ proton of the 6-deoxysugar. In addition, two distinct signals in the methyl region were observed. The first peak (H7Љ, 1.29 ppm) is a singlet and was later assigned to the methyl group attached to C3Љ; the second peak is a doublet and was later assigned to the methyl group at C6Љ. The singlet peak suggested that no proton is connected to the C3Љ on the sugar ring, whereas the doublet peak (i.e. split) at H6Љ is the result of coupling to H5Љ. If the C3Љ-methyl was O-linked to C3Љ (like in methanol), we would expect a H3Љ signal and a doublet signal of H7Љ due to the split of H3Љ. ii, E. coli expressed the combined BC3358ϩBC3359 genes, producing CDP-4-keto-6-deoxyglucose; iii, E. coli expressed the combined BC3358ϩBC3359ϩBC3360 genes, producing a 3-Cmethylated CDP-4-keto-6-deoxyhexose; iv, E. coli expressed the combined four genes, producing a new UV peak, later assigned as CDP-cillose; v, E. coli expressed an empty vector control, producing no such peak.  A two-dimensional COSY experiment (Fig. 5A) showed the connectivity between H1Љ and H2Љ, H4Љ and H5Љ, and H5Љ and H6Љ on the cillose ring. The absence of a proton connected to the C3Љ position and the absence of connectivity of methyl H7Љ to any adjacent protons suggest that this methyl group is connected via a carbon at the 3Љ position. In addition, our HMBC experiment (Fig. 5B) also indicates the connectivity of methyl H6Љ to C4Љ and C5Љ and methyl H7Љ to C2Љ, C3Љ, and C4Љ, suggesting a sugar ring with 3-C-methyl-6-deoxyallose configuration.
Complete determination of the stereoisomerism of this nucleotide sugar came by measuring the coupling constants in one-dimensional NMR experiments. The large coupling constant J 4,5 ϭ 10.1 Hz between H4Љ and H5Љ (Table 1) strongly supports an allose-configuration and not a gulose-configuration. The distinct chemical shift of the anomeric proton H1Љ and the J coupling constant value of 3.79 Hz for J H1Љ,H2Љ are consistent with ␣-linkage to the phosphate moiety of CDP. Additionally, a one-dimensional ROESY experiment was performed (Fig. 5C). When H4Љ was irradiated, dipolar couplings (NOE effect) between H4Љ and H2Љ, H4 and H6Љ, and H4Љ and H7Љ were observed, confirming an allose-configuration. Taken together, this nucleotide sugar is characterized as CDP-3-C-methyl-6-deoxyallose, which we named CDP-cillose.
In addition to NMR analysis, we also utilized a CDP-cillose sample as a standard for GC-MS analysis. When the CDP-cillose standard was hydrolyzed, derivatized to its alditol acetate, and analyzed by GC-MS, the retention time, EI-MS ion, and fragments were identical to sugar peak C in Fig. 2B.  Note that there is no proton connected through H2Љ and H4Љ indicating no proton linked to C3Љ. B, selected spectral region of HMBC experiment showing protons on the methyl region connected to adjacent carbons that are 2-3 bonds away. Note that the methyl group (H7Љ) is connected to C3Љ indicating that cillose is C3-methylated. C, one-dimensional ROESY experiment with H4Љ irradiated. Dipolar couplings (NOE effect) are observed between H4Љ and H2Љ also indicating an allose-configuration.

In vitro assays of C3CM operon reveal that the combined enzymatic product is CDP-3-C-methyl-6-deoxygulose
Although the first method provided evidence that the four Bacillus genes expressed in E. coli produced CDP-cillose, the second approach based on in vitro enzymatic assays suggests that the last step gives a different product with altered sugar configuration, as described below. In the second method, each individual recombinant protein was expressed and purified by affinity column (supplemental Fig. S2). Our biochemical data confirm that recombinant His 6 -Bc3358 is indeed a CTP:Glc-1-P cytidylyltransferase as determined by LC-MS/MS (Fig. 6A), whereas the control empty vector gave no product. During the initial characterization of the recombinant enzyme, we determined that the activity required magnesium, Glc-1-P, and CTP, as no activity was observed without the metal. The enzyme is specific to CTP as a nucleotide because ATP, dTDP, and GTP are not substrates in the conditions described under "Experimental procedures"; however, a less than 1% conversion was observed with UTP. Gal-1-P is not a substrate. MS analyses of the enzymatic peak product yielded an ion at m/z 564 and an MS/MS ion fragment of m/z 322, consistent with [CDP-Glc-H] Ϫ and [CMP-H] Ϫ , respectively. As CDP-Glc is unavailable commercially, we purified a large amount of it, using a HILIC column, for further characterization of the subsequent enzymes.
The second enzyme is a specific 4,6-dehydratase. The recombinant enzyme, His 6 -Bc3359, readily converted CDP-Glc to a CDP-4-keto-6-deoxyglucose, whereas a control reaction yielded no product. Fig. 6B shows LC-MS/MS analyses of the activity of Bc3359, where the enzymatic products (labeled CDP-4k-6d-Glc, referring to CDP-4-keto-6-deoxy-Glc) eluted from the HILIC column as a broad peak between 10   A, purified His 6 -Bc3358 was reacted with CTP and Glc-1-P to yield a new peak of CDP-Glc eluting from HILIC column at 14 min. The dashed line indicates the elution of substrates. B, purified His 6 -Bc3359 converts CDP-Glc to a new peak, CDP-4-keto-6-deoxy-Glc, eluting as a broad peak between 10 and 12 min. C, His 6 -Bc3360 methylates and converts CDP-4-keto-6-deoxy-Glc to CDP-3-C-methyl-4-keto-6-deoxyhexose. D, when reaction shown in panel B was terminated and co-incubated with purified His 6 -Bc3360 and His 6 -Bc3361, a new peak eluting at 11 min was observed; this peak was later identified as CDP-3-C-methyl-6-deoxygulose (CDP-cereose). peared. Product K is CDP-4-keto-6-deoxy-Glc and H is the hydrated form of K. The H-4Љ proton signal of the substrate (labeled C, H-4Љ) decreased over time, reflecting the disappearance of H-4Љ when forming the 4-keto structure product. Two signals also appeared at the 6-deoxy regions around 1.20 and 1.30 ppm. These signals correspond to the C6-methyl protons (H-6Љ) of products K and H sugar moieties, respectively. The collective MS/MS and NMR data provide evidence that Bc3359 encodes a CDP-Glc 4 -6-dehydratase. Further assays were conducted to determine substrate specificity, but dTDP-Glc and UDP-GlcNAc were not substrates; however, less than 1% of the UDP-Glc was converted to UDP-4-keto-6-deoxy-Glc as evident by MS/MS. The next enzyme activity was found as C3-MetT. Preliminary assays established that the recombinant Bacillus Bc3359 C3-MetT requires magnesium and NADH for activity and is SAM-dependent. The purified recombinant C3-MetT (His 6 -Bc3360) was reacted with CDP-4-keto-6-deoxyglucose (the reaction product of Bc3359). A new broad peak with a retention time between 9.5 and 10.5 min (Fig. 6C) was detected, whereas a control reaction gave no product. The mass spectrum showed an ion [M-H] Ϫ at m/z 560, proposing that the product was also a CDP-4-keto-sugar, but the increase of 14 atomic mass units when compared with the substrate suggested that the product gained a methyl group. Hence, the product of Bc3360 is CDP-3-C-methyl-4-keto-6-deoxyhexose. Interestingly, the activity of the C3-MetT was relatively low and only a ϳ10% conversion was obtained. However, adding the last enzyme in the pathway (4-reductase, Bc3361) drove the reaction further to produce a new product. Fig. 6D shows the enzymatic assay of the combined C3-MetT and 4-reductase. Both purified recombinant C3-MetT (His 6 -Bc3360) and 4-reductase (His 6 -Bc3361) readily reacted with CDP-4-keto-6-deoxyhexose to yield a final product, whereas negative control reactions gave no product. A peak with a retention time at 11 min ( Fig. 6D) was detected having an ion [M-H] Ϫ at m/z 562. MS/MS analysis of this ion gave fragments at m/z 322 and 239, consistent with CMP and sugar-1-P derivatives for CDP-3-C-methyl-6-deoxyhexsose.
When substituting Mg 2ϩ for Zn 2ϩ or Mn 2ϩ in the Bc3360 assay, no C3-MetT activity was detected. On omitting NADH, no C3-MetT activity and only low activity was observed when NADPH was substituted for NADH or when pyridoxal phosphate (PLP) was used. Lastly, the C3-MetT activity requires SAM as the methyl donor with the highest activity at 22°C and with little activity at 37°C. Further support for the in vitro C3-MetT-specific activity came from our in-microbe experiment. Engineered E. coli cells co-expressing the C3-MetT gene with a gene encoding an enzyme that forms dTDP-4-keto-6deoxy-Glc, UDP-4-keto-6-deoxy-Glc, or UDP-4-keto-6-deoxy-N-acetylhexosamine were not able to methylate these derivatives. The last enzyme in the pathway is a 4-reductase. It has the highest activity with NADPH rather NADH. In the inmicrobe experiments, the 4-reductase activity also appeared specific with no activities toward dTDP-4-keto-6-deoxy-Glc, UDP-4-keto-6-deoxy-Glc, or UDP-4-keto-6-deoxy-Nacetylhexosamine.

One-and two-dimensional NMR reveals the chemical structure of CDP-3-C-methyl-6-deoxygulose
The column-purified product (see product peak in Fig. 6D) of the in vitro combined enzymatic assays of His 6 -Bc3360 and His 6 -Bc3361 was analyzed by NMR spectroscopy. The one-dimensional NMR spectra (Fig. 8) showed anomeric signals belonging to the H1Љ peak (5.54 ppm), ring protons H2Љ and H4Љ (3.25-3.75 ppm), and the diagnostic H5Љ proton of the 6-deoxysugar (4.56 ppm). Similar to CDP-cillose, two distinct methyl signals were observed: the singlet peak H7Љ (1.31 ppm), which represents the methyl group attached to C3Љ; and the doublet peak H6Љ (1.19 ppm), assigned to the methyl group at C6Љ. The chemical shifts acquired are different from those of CDP-cillose, indicating that the enzymatic product from the in vitro assay of the C3CM operon is different from the E. coli-based in-microbe reaction. The chemical shift differences between CDP-cillose made in E. coli compared with the in vitro enzymatic product led us to acquire more NMR data to determine the structural nature of this material.
Further two-dimensional COSY and ROESY experiments were performed to determine the stereoisomerism of H4Љ (Fig.  9). The COSY experiment (Fig. 9A) showed the connectivity between H1Љ and H2Љ and H5Љ and H6Љ on the ring. No proton connected the C3Љ position, as expected; however no coupling between H4Љ and H5Љ was observed, indicating that the coupling constant between H4Љ and H5Љ is very small (J 4,5 Ͻ1). An additional ROESY experiment was conducted to confirm that H4Ј is in an axial position (Fig. 9B). Dipolar couplings were observed between H4Ј and H6Ј and H4Ј and H7Ј, but no NOE effect was observed between H4Ј and H2Ј (data not shown), suggesting that H4Љ is in an axial position, consistent with a gulose-configuration. The NOE effect observed between H2Љ and H7Љ (Fig. 9B) also indicates that the methyl on C3Љ is equatorial. Taken together, the product of the in vitro enzymatic assay from the C3CM operon is CDP-3-C-methyl-6-deoxygu- lose. We decided to name it CDP-cereose because while this work was in progress, a study on spore glycoprotein BclA in B. cereus ATCC 14579 (18) identified cereose as attached as a terminal sugar that is ␤-1,4-linked to 3-O-methyl-rhamnose on the oligosaccharide of BclA. The complete chemical shifts of CDP-cereose are listed in Table 1. We also utilize a purified CDP-cereose sample as a standard for the GC-MS analysis, and the retention time, the EI-MS ion, and fragments are identical to sugar peak B in Fig. 2B.
To solve the discrepancy between the E. coli in-microbe analysis and the in vitro enzymatic activities, we decided to examine whether the 4-reductase is also a 4-epimerase. However, CDPcillose was not converted by Bc3361 to CDP-cereose when assays utilized NADPH or NADH as determined by NMR.

Bacillus produces predominantly CDP-cereose in vivo during sporulation
We next addressed which CDP-3-C-methyl-6-deoxyhexose sugar(s) Bacillus produces in vivo and when. For that purpose we analyzed NDP-sugars as well as the expression of genes belonging to the C3CM operon. We first examined whether the genes are constitutively expressed or are induced by specific environmental cues. To that end we found that rich BHI (brain-heart infusion) medium suppressed transcription of the four genes (Fig. 10A), whereas under a defined nutrient medium (Msgg) these genes were transcribed as determined by RT-PCR. It took, however, about 48 h for the genes of the C3CM operon to be transcribed upon shifting from BHI to Msgg medium.
Because Msgg medium gave rise to both biofilm and spores, we tested whether DSM (Difco sporulation medium), which induces spore formation, also gave rise to an accumulation of CDP-3-C-methyl-6-deoxyhexose. To that end, NDP-sugars were extracted from Bacillus grown in either Msgg or DSM. A peak with a retention time at 11 min (Fig. 10B) was detected with m/z 562, an ion consistent with CDP-3-C-methyl-6-deoxyhexose. In both DSM and Msgg the m/z 562 was identified on the first day, with the highest level at day 2 and reduced amounts by day 4. The HILIC column was unable to separate the CDP-cillose from its 4-epimer, CDP-cereose. To determine the identity of peak 562, the NDP-sugars were isolated, mildly hydrolyzed, derivatized to alditol acetates, and analyzed by GC-MS. The elution time (22.7 min) and MS fragmentation data indicate that the major CDP-sugar is CDP-cereose (data not shown). Based on the retention time (23.3 min), the amount  of cillose derived from CDP-cillose was too low to be designated conclusively. Hence, the combined LC-MS and GC-MS data suggest that Bacillus makes predominately CDP-cereose in vivo, but we cannot exclude the possibility that CDP-cillose is made, albeit in substantially smaller amounts.

C3CM operon is required to form cereose-containing glycan
We first addressed whether the C3CM operon is involved in cereose-glycan formation. For this purpose, we deleted the gene encoding Bc3360, the C3-MetT in B. cereus ATCC 14579 (supplemental Fig. S3). The data of spore-derived glycans in ⌬BC3360 not surprisingly showed the lack of the cereose sugar residue (Fig. 11). These data suggested that genetically C3-MetT of the C3CM operon is involved in cereose-containing spore glycans. We then examined genetically the function of Bc3361 and whether it is a specific 4-reductase. To that end, we generated a BC3361 gene deletion strain, ⌬BC3361. The spore glycans were analyzed by GC-MS, and the data showed a significant reduction in the amount of cereose. This strongly suggests that Bc3361 is indeed involved in the formation of cereosecontaining glycans in spores. Further investigation and protein crystallography will be required to fully understand the contribution of Bc3361 to the formation of cillose-containing glycans.
We next addressed whether all Bacillus sp have the capability of forming cereose-containing glycans. We examined 17 differ-ent Bacillus strains (Table 2) of the Bacillus cereus sensu lato group: B. cereus, Bacillus thuringiensis, and B. anthracis. Genomic sequences revealed that four of the 17 strains do not harbor DNA sequences for the C3CM operon; and these strains showed no cereose residue by GC-MS. But all strains having the C3CM operon produced cereose. Thus, it is evident that the C3CM operon is indeed involved in the formation of cereosecontaining glycans.

C3CM operon may contribute to spore germination
We next addressed whether mutants of the C3CM operon have an effect on spore properties, sporulation, or spore germination. Upon the initial analysis of ⌬BC3360 and ⌬BC3361, we did not detect any difference in the sporulation rate after 8 days of growth on DSM-agar plates compared with the wild type. Similarly, the lack of 3C-methyl sugars had no impact on spore stability, and mutant spores preincubated at 80, 85, and 90°C had similar heat stability properties as the wild type (data not shown). However, we observed a 10 -20% increase in the germination rate of ⌬BC3360 and ⌬BC3361 when compared with the wild type (Fig. 12). Taken together, cereose glycosylation may contribute to spore germination but seems not to impact the stability of spores under heat.

Timing of cereose-containing glycan expression
In Msgg or DSM, over 90% of the B. cereus ATCC 14579 cells sporulated by day 3-4. Analyses of glycans isolated from 1-day cultures provided no evidence of cereose-containing glycans. Cereose-containing glycans started to appear in day 2-3 in cultures, with higher amounts accumulating by day 6. To determine whether cereose-containing glycans are stable, we analyzed spores isolated from cultures grown for up to 14 days in Msgg. No degradation of the cereose-sugar signal was observed between days 6 and 14, suggesting that cereose-containing glycans were not degraded. Similar amounts of cereose were observed in spores produced by DSM. In general, the amount of cillose-containing glycans, based on GC-MS analyses, was not sufficient to be quantified for the timing analysis when compared with cereose-containing glycans.  The amount of cereose in spores was calculated to be 0.2-1% of the dry mass. This significant amount may play an important role in spore biology.

Cereose sugar is not exclusive to spore glycoprotein BclA
Following the identification of CDP-cereose in this report, we further examined whether cereose is found only in BclA and enriched in the spores by analyzing the enriched exosporium fraction in the BclA mutant as described earlier (18,19). Cereose indeed appeared in the exosporium preparation, but a significant amount of cereose was also detected in the spores lacking exosporium (Fig. 13A), suggesting that cereose may not be unique to BclA glycoprotein. In addition, upon examination of the BclA as well as the ExsJ mutants from B. cereus ATCC 10876 (20), we found cereose present in the spores indicating that cereose may be attached to other glycans as well (Fig. 13B).
Consistent with the above observation, we treated the spores with urea and SDS, as the combined reagents are known to release glycoproteins such as BclA (21) and BclB (22). Analysis of B. cereus ATCC 14579 spores after urea/SDS treatment showed two clear GC-MS signals for cereose and cillose at a ratio of 36:1, respectively (Fig. 13C). This suggests that a large portion of the cereose-and cillose-containing glycans still decorates the spores. The explanation for this will require further intensive research; it could suggest that some of the cereose/ cillose glycan structures on the spores are either cross-linked to other macromolecules, urea-resistant glycoproteins, or SDSresistant glycolipids.

Discussion
In this study, we have identified and biochemically characterized a new metabolic pathway involved in the formation of uncommon sugar residues in Bacillus. The characterized pathway includes four B. cereus ATCC 14579 recombinant proteins (Bc3358 -Bc3361). Fig. 1 summarizes our proposed biochemical route for the formation and conversion of CDP-glucose to CDP-3-C-methyl-6-deoxygulose (named CDP-cereose) and potentially its 4-epimer CDP-cillose. The first Bc3358 protein is a magnesium-and CTP-dependent enzyme that attaches the cytidyl group from CTP to Glc-1-P to from CDP-glucose. The second enzyme, Bc3359, is a specific 4,6-dehydratase that converts CDP-Glc to a CDP-4-keto-6-deoxy-Glc. This intermediate exists in two forms, the hydrated and keto forms. Existing in two forms appears to be a common feature observed with other 4-keto nucleotide-sugar derivatives including UDP-4-keto-6deoxyglucose (23), UDP-4-keto-6-deoxy-AltNAc (24), UDP-4keto-6-deoxy-GlcNAc (25), and UDP-4-keto-xylose (26). The third enzyme is a distinct C3-methyltransferase (C3-MetT, Bc3360) that requires magnesium to catalyze the transfer of the methyl group from SAM and link it to C3Љ of the sugar moiety of the CDP-sugar to form CDP-3-C-methyl-4-keto-6-deoxyhexose. Like some other characterized C3-methyltransferases (27,28), Bc3360 is NADH-dependent. There is no apparent formation of NAD ϩ , and the enzyme does not appear to function as a reductase in the absence of SAM. It is also able to use NADPH as co-factor with 50% less conversion than NADH as a co-factor. Other co-factors, like pyridoxal phosphate (PLP), which may be suitable for other C3-methyltransferase (29), showed very little conversion (less than 1%). Finally, in vitro Bc3361 reduces in an NADPH-dependent manner the 4-keto  moiety of the methyl sugar and converts it to CDP-cereose and NADP. The analog, NADH, could not substitute for NADPH in the 4-reductase reaction.
The order of genes within the C3CM operon, the nucleotide space, and the overlap among the genes are highly similar to other Bacillus spp. For example, BC3360 is overlapped by one nucleotide with BC3361 and 1-and 12-nucleotide gaps are found between the coding regions of BC3358 and BC3359 and BC3359 and BC3360, respectively. Such gene overlaps and reading-frame offsets appeared to be conserved in other Bacillus spp including Bacillus weihenstephanensis FSL R5-860, B. cereus ATCC 10876, B. thuringiensis serovar israelensis ATCC 35646, B. thuringiensis serovar kurstaki str. HD-1, Bacillus toyonensis BCT-7112, and B. cereus m1293. It is therefore possible that the expression of the C3CM genes is under the same regulation mechanism. Not all strains of the B. cereus group harbor the C3CM operon; for example the B. anthracis strains including 34F2, Ames, and Vollum do not have the C3CM genes in their genome. On the other hand, some B. cereus strains that are also human pathogens (supplemental Table S1), like B. cereus G4264 and G9241 or food and dairy spoilage strains like B. wiedmannii, do have C3CM operon. This suggests that cereose decoration on spores may function in assisting spore survival under certain environmental conditions or in host infection via specific mechanisms.
The enzyme activity of recombinant proteins showed that the last protein, a 4-reductase, gave a gulose-configured sugar residue, cereose. However, in E. coli the 4-reductase produces predominantly the 4-epimer with an allose-configuration, cillose. The discrepancy was not easily discerned, specifically because cillose and cereose are both present in Bacillus. One possibility is that the predominant activity of the 4-reductase is to generate CDP-cereose, and little activity is expended to form CDP-cillose. The second possibility is that an additional 4-epimerase in Bacillus or E. coli epimerizes CDP-cereose to CDPcillose; or alternatively, the 4-reductase activity could be regulated post-translationally by other factors, resulting in this enzyme having a different activity compared with its activity in vitro. The specific mechanism behind this variance is unknown. Further protein crystallography will be required to elucidate the enzymology of Bc3361.

Strains and culture conditions
The strains used in this study are listed in Table 2. Each strain was stored in 16% glycerol at Ϫ80°C. The media (agar or liquid) used were Luria Bertani (LB) (10 g/liter tryptone, 5 g/liter yeast extract, and 10 g/liter NaCl), BHI, DSM, and Msgg medium (30). E. coli DH10B cells were used for cloning and Rosetta2(DE3)pLysS (Novagen) cells were used to express protein.
B. cereus cells were cultured in either liquid or agar plates with Msgg or DSM plates and incubated at 30°C for 5 days to achieve a sporulation rate over 90% as confirmed by microscopy and cfu counts after an 80°C heat stability treatment. The spores were purified as described (31), subsequently washed repeatedly with sterile double-distilled water (ddw), and finally centrifuged (8000 ϫ g, 10 min).

Cloning the C3CM operon genes
A single colony of B. cereus cells grown on BHI-agar was suspended in 50 l of sterile ddw, heated for 5 min at 95°C, and then centrifuged (10,000 ϫ g, 30 s). A portion of the supernatant (5 l) was used as the DNA source to amplify each specific gene by PCR. Each primer was designed to include a 15-nucleotide extension with exact sequence homology to the cloning site of the target plasmid at the 5Ј-end. This facilitated cloning between the NcoI and Bpu1101I sites of pET28_Tev1.15 (32) or the SacI site of the pCDF-ParaB vector. pCDF-ParaB is a modified pCDF duet where expression is controlled by arabinose via the regulator gene araC. The pET28_Tev1.15 vector was (32) opened with primer sets ZL169 and KB_T7t. The individual genes (BC3358, BC3359, BC3360, and BC3361) were PCR-amplified with primer sets KB2 S and KB2 AS; KB5 S and KB5 AS; KB3 S and KB3 AS; and KB4 S and KB4 AS, respectively (see supplemental Table S2), using high fidelity Pyrococcus DNA polymerase (0.4 unit of Phusion Hot Start II, Thermo Scientific), including buffer, dNTPs (0.4 l at 10 mM), B. cereus DNA template (5 l), and PCR primer sets (1 l each at 10 M) in a 20-l reaction volume. The PCR conditions were: a 98°C denaturation cycle for 30 s followed by 25 cycles (8 s denaturation at 98°C, 25 s annealing at 54°C, and 20 s elongation at 72°C) and 4°C. After PCR, portions (2 l each) of the amplified vector and insert were mixed, treated for 15 min at 37°C with 0.5 unit if FastDigest DpnI (Thermo Scientific), and then transformed into DH10B (Life Technologies)-competent cells. Positively transformed clones were selected on LB-agar containing kanamycin (50 g/ml for pET vector) or spectinomycin (100 g/ml for pCDF vector). Clones were verified by PCR and DNA sequencing. The resulting recombinant plasmids are: pET28b: His 6 -Tev-Bc3358#3 yielding N-terminal His 6 -Tev-tagged Glc1P-CytT; pET28b:His 6 -Tev-Bc3359#10 yielding N-terminal His 6 -Tev-tagged CDP-Glc 4,6-DH; pET28b:His 6 -Tev-Bc3360#12 yielding N-terminal His 6 -Tev-tagged C3-MetT; and pET28b:His 6 -Tev-Bc3361#2 yielding N-terminal His 6 -Tev-tagged 4-reductase. Plasmids were isolated and purified using a PureLink Quick Plasmid miniprep kit (Invitrogen) and transformed into Rosetta2(DE3)pLysS-competent cells for recombinant protein expression. Once protein activities were confirmed, the DNA sequences of the plasmids harboring specific genes were deposited in GenBank TM with the following accession numbers: KY445942 for Bc3358, KY445943 for Bc3359, KY432406 for Bc3360, and KY432407 for Bc3361.

His 6 -tagged protein expression and purification
Rosetta2(DE3)pLysS strains harboring pET expression plasmids were grown at 37°C and 250 rpm in 250 ml of LB medium supplemented with chloramphenicol (35 g/ml) and kanamycin (50 g/ml). Gene expression was induced when cellular A 600 reached 0.6 by adding 0.5 mM IPTG. After induction, cells were grown for 18 h at 18°C and 250 rpm and then harvested by centrifugation (6000 ϫ g) for 10 min at 4°C. The cell pellets were washed with ddw and then suspended in 10 ml of lysis buffer (50 mM Tris-HCl, pH 7.4, 10% (v/v) glycerol, 1 mM EDTA, 2 mM DTT, and 0.5 mM PMSF). Cells were lysed by sonication (26), and after centrifugation (6000 ϫ g for 15 min at 4°C), the supernatant was supplemented with 1 mM DTT and 0.5 mM PMSF and again centrifuged (20,000 ϫ g for 30 min at 4°C). The supernatant was applied to a nickel-Sepharose fastflow column (GE Healthcare Life Sciences; 2 ml of resin packed in a polypropylene column, inner diameter 1 ϫ 15 cm). Each column was pre-equilibrated with buffer A (50 mM Tris-HCl, pH 8, 4% (v/v) glycerol, and 100 mM NaCl). The column was washed with 30 ml of buffer A containing 20 mM imidazole and then with 10 ml of buffer A containing 40 mM imidazole. Histagged proteins were eluted with 5 ml of buffer A containing 250 mM imidazole. The eluates containing these proteins were divided into small aliquots, flash-frozen in liquid nitrogen, and kept at Ϫ80°C. Proteins were separated by SDS-12.5% PAGE and visualized by staining with Coomassie Blue.

Nucleotide-sugar analyses derived from "in-microbe"-based assay
Nucleotide sugars produced by the C3CM operon were extracted as described previously (33). Briefly, Rosetta2(DE3) pLysS strains harboring expression plasmids were grown at 37°C and 250 rpm in 5 ml of LB medium supplemented with chloramphenicol (35 g/ml) and kanamycin (50 g/ml) (pET vector) or with chloramphenicol (25 g/ml) and kanamycin (35 g/ml) with additional spectinomycin (35 g/ml) (pET plus pCDF vector co-expression). Gene expression was induced after 2 h by adding 0.5 mM IPTG. After induction, cells were grown for 3-4 h at 30°C and 250 rpm and then harvested by centrifugation (10,000 ϫ g, for 1 min at 2°C). The cell pellets were washed with 4 volumes of 10 mM sodium phosphate, pH 7.5, and 150 mM NaCl (PBS) and then suspended in 1 volume of ddw. Ten volumes of cold chloroform/methanol (1:1, v/v) was added, and the sample was mixed for 15 min on ice. Samples were centrifuged (12,000 rpm, 3 min, 22°C), and the upper aqueous phase enriched in nucleotide sugar was collected. Portions of this aqueous phase were chromatographed on an HILIC column with HPLC-UV or LC-ESI-MS/MS (liquid chromatography-electrospray ionization-tandem mass spectrometry), using a Shimadzu ESI-MS/MS IT-TOF mass spectrometry system that included a Nexera UFPLC LC-30AD pump, a Sil-30AC autosampler, and a column heater at 37°C. HPLC peaks of nucleotide sugars were detected by UV A 271 nm (maximum for CDP-sugars), collected, and lyophilized prior to analysis by NMR. The purified product of His 6 Bc3358 (CDP-Glc) was also used as substrate for the downstream enzymatic reactions. The purified product of entire C3CM operon (CDP-cillose or CDPcereose) was suspended in D 2 O (99.9%) for NMR analysis.
For analyses by HPLC-UV or LC-MS/MS, an Accucore 150 Amide HILIC column (150 ϫ 4.6 mm, 2.6-m particle size, Thermo Scientific) was used for chromatography with a solvent system of 40 mM ammonium acetate, pH 4.3 (solvent A), and acetonitrile (solvent B). The column was equilibrated at 0.4 ml/min with 25% A and 75% B prior to sample injection (20 l). Following injection, the HPLC conditions were: 0 -1 min, 0.4 ml/min with 25% A/75%B and then a gradient to 50% A and 50% B over 24 min. The flow rate was then increased to 0.6 ml/min with a gradient to 25% A and 75% B over 5 min. The column was then washed for 5 min with 25% A/75% B prior to the next injection.

In vitro enzyme reactions
The activity of recombinant His 6 Bc3358 (referred to herein as Bc3358 or Glc1P-CytT) was examined by HPLC-UV and LC-ESI-MS/MS. The total reaction volume of 50 l consisted of 50 mM Tris-HCl, pH 7.4, 3.3 mM MgCl 2 , 1.5 mM CTP, 1 mM Glc-1-P, and up to 15 l of purified Bc3358. Reactions proceeded for 1 h at 30°C. An aliquot (20 l) was mixed with acetonitrile (40 l) and 0.5 M ammonium acetate, pH 5.3 (2 l), and a portion (30 l) was chromatographed and detected by LC-ESI-MS/MS operating in the negative ion mode. Enzyme products were identified based on their retention time, the mass of their parent ion, and their mass spectral fragmentation pattern.
The activity of recombinant His 6 Bc3359 (referred to herein as Bc3359 or 4,6-DH) was examined similarly as described above. To a total reaction volume of 50 l was added 50 mM Tris-HCl, pH 7.4, CDP-Glc, and up to 15 l of purified Bc3359. Reactions proceeded for 2 h at 30°C followed by enzyme inactivation and extraction with 50 l of chloroform (32). The activity of recombinant His 6 Bc3360 (referred to herein as Bc3360 or C3-MetT) was examined using the reaction product of Bc3359 supplemented with 1 mM NADH, 3.3 mM MgCl 2 , 1 mM SAM, and up to 15 l of purified Bc3360 in a total volume of 25 l. Reactions proceeded for 2 h at 22°C. For the co-incubation assays, which included both Bc3360 and Bc3361, the enzymatic reaction product of Bc3359 was supplemented with 1 mM NADH, 1 mM NADPH, 3.3 mM MgCl 2 , 1 mM SAM, 4 l of purified Bc3360, and 4 l of purified Bc3361 in a total volume of 25 l. Reactions proceeded for 2 h at 30°C. Aliquots of reactions were prepared as described previously and analyzed by HILIC LC-ESI-MS/MS.

NMR spectroscopy used to characterize the structure of products
Time-resolved NMR for the formation of CDP-4-keto-6-deoxy-Glc by recombinant Bc3359 was carried out in a final volume of 180 l, essentially as described (24). The HPLC peaks of CDP-Glc, CDP-cillose, and CDP-cereose collected as described previously were lyophilized, dissolved in D 2 O (99.9%), supplemented with 1 l of 10 mM DSS (4,4-dimethyl-4-silapentane-1sulfonic acid), and analyzed by NMR spectroscopy (Agilent DD2 600 MHz NMR spectrometer equipped with a cryogenic 3-mm probe). Proton and carbon chemical shifts were referenced to an internal DSS peak set at 0.00 ppm for both proton and carbon spectra. The one-dimensional proton NMR spectrum was recorded using a water-presaturated pulse sequence and obtained with a spectral width of 6 kHz, a 90°pulse field angle (7.5 s), a 2.7-s acquisition time, and a 2-s relaxation delay. Proton chemical shifts were assigned by a COSY (correlation spectroscopy) experiment. Carbon chemical shifts as well as the connectivity between carbons and protons were determined by a heteronuclear multiple bond correlation (HMBC) experiment. The configuration of each hydroxyl group on CDP-cillose or CDP-cereose was determined by onedimensional proton and ROESY (rotating-frame Overhauser spectroscopy) experiment. Data processing and plotting were performed using MestreNova software.

RNA isolation and RT-PCR
A 1:100 dilution of an overnight BHI culture of B. cereus ATCC 14579 was used to inoculate 50 ml of liquid medium (BHI or Msgg), and the cultures were grown for up to 72 h at 30°C with shaking (200 rpm). Cells were rapidly pelleted by centrifugation (10,000 ϫ g for 1 min at 4°C), resuspended in 0.8 ml of lysozyme solution (14), and incubated at room temperature for 10 min. To each sample 80 l of 10ϫ EB (0.3 M NaOAc, pH 5.2, 50 mM EDTA, 5% sarkosyl, and 1.42 M ␤-mercaptoethanol) was added. After 3 min at 65°C, 1 volume of 70°C preheated acidic phenol (phenol:AcE buffer: 50 mM NaOAc, pH 5.1, and 5 mM EDTA (1:1 v/v)) was added, and the mixture was incubated at 65°C for 7 min. The sample was centrifuged (10,000 ϫ g for 5 min at 4°C), and the top aqueous layer was collected and mixed with an equal volume of chloroform. Nucleic acid partitioned to the upper phase was collected, mixed with 1 volume of 30 mM NaOAc, pH 5.2, and 3 volumes of cold ethanol, and nucleic acids were precipitated at Ϫ20°C. To digest the remnant genomic DNA, an aliquot of 5 g of crude RNA was treated with DNase I. The RNA was ethanolprecipitated, and a portion was used for transcript analyses by reverse transcriptase (RT) and PCR reactions. The RT reaction (20 l, final) consisted of 250 ng of RNA, 1 l of 10 M genespecific reverse primers, buffer, 0.2 M dNTPs, and 1 unit of reverse transcriptase (SuperScript III, Invitrogen). Negative control RT reactions were done without added reverse transcriptase. Transcripts of genes in the C3CM operon, negative controls, and a positive control (SigA) were amplified, each at final 25 l, by PCR reaction that included 2 l of RT reaction, buffer, dNTPs, 1 unit of TaqDNA polymerase (Promega), and gene-specific sense and antisense primers (0.4 M) (primer sets TM001 and TM002 for BC3358, TM003 and TM004 for BC3359, TM005 and TM006 for BC3360, TM007 and TM008 for BC3361, and ZL347 and ZL348 for SigA) (see supplemental  Table S2). Following PCR, a portion (8 l) of each RT-PCR reaction was loaded on a 1% agarose-TAE gel casted with 10 g/ml ethidium bromide, separated by gel electrophoresis, and UV-imaged using Gel Imager.

GC-MS analyses for identification of cereose and cillose
As cereose and cillose are labile sugars, mild hydrolysis was used. Purified spores collected either from liquid culture or scraped from agar plates were washed and suspended. Around 1 ϫ 10 9 spores were supplemented with 10 g of inositol and hydrolyzed with 0.2 M TFA for 2 h at 70°C, and the released monosaccharides were reduced to their alditols (35) and acetylated. The resulting alditol-acetate derivatives were analyzed using a GC/EI-MS system (Agilent 7890a/5975c) equipped with an autosampler injector (Agilent 7693). A 1-l sample was injected into the GC column (Equity-1 or DB-5, 30 m ϫ 0.25 mm, 0.25-m film thickness) using split mode (1:50) with an injector inlet setting of 250°C (helium at 3 ml/min), and chromatography was performed as described previously (14). In some cases the GC column used was Rtx-2330 by RESTEK. MS data were collected after a solvent delay of 5 min, and the ion abundance in the range of 50 to 550 m/z was recorded. The spectra were analyzed using MSD ChemStation. To determine the elution time of authentic cereose and cillose and the EI-mass fragments formed, CDP-cereose and CDP-cillose were produced and purified over an HILIC column, TFA-hydrolyzed, converted to alditol acetate, separated by GC using a DB-5 column, and analyzed by EI-MS using the above GC conditions.

Mutant generation in C3CM operon
Gene knock-out was achieved by double crossover recombination as described previously (14). Two DNA fragments flanking gene BC3360 and two DNA fragments flanking gene BC3361 were PCR-amplified by specific primer sets ZL479 and ZL480, ZL481 and ZL482, ZL489 and ZL490, and ZL491 and ZL492, respectively (see supplemental Table S2). The flanking regions were cloned individually by primer sets ZL007 and ZL008 and ZL009 and ZL123, respectively, into the pZL-KO shuttle plasmid (14), a derivative of pBCB13 that harbors the thermosensitive ORI (origin of replication) from pMAD (37,38). The cloning strategy for chromosomal integration at a nonpermissive temperature and subsequently a double crossover recombination event was conducted as described previously (14) to achieve the resulting B. cereus mutant strains ⌬BC3360 and ⌬BC3361. For complementation of mutants, PCR was used to amplify the BC3360 and BC3361 genomic DNA region from wild-type B. cereus ATCC 14579 with primer sets ZL483 and ZL484 and ZL485 and ZL486, respectively. Each DNA fragment was cloned into the pDZ vector, a derivative E. coli/Bacillus shuttle plasmid of pDG148-stu (34), which was generated by primer sets ZL213 and ZL214 as described previously (14).

Germination analysis of mutants
Spores from B. cereus wild-type and mutant strains ⌬BC3360 and ⌬BC3361 were prepared as stated previously. Prior to germination, the purified spores were heated at 70°C for 30 min and washed again with ddw. Equal amounts of spores (calculated to an optical density, A 600 ϭ 0.8) were resuspended in germination solution (50 mM inosine in 10 mM Tris-HCl, pH 8, and 10 mM NaCl), and the degree of germination was monitored at A 600 over a time course of 32 min (36).

Exosporium extraction and urea-SDS treatment
Purified spores from B. cereus wild type were washed twice in water, and the exosporium extraction was conducted as described previously (19). Briefly, the spores were subjected to four successive passages through a French press (20,000 p.s.i.), and the exosporium-free spores were recovered after centrifugation (3000 ϫ g, 30 min, 4°C). Exosporium in the supernatant was pelleted by Ultracentrifuge at 120,000 ϫ g for 30 min at 4°C, treated by mild acid hydrolysis, derivatized to alditol acetates, and analyzed by GC-MS.
Spores were also suspended in urea-SDS buffer (50 mM Tris-HCl, pH 7, 2% SDS, and 8 M urea) and boiled for 15 min as described (20,21). The sample was then centrifuged at 10,000 ϫ g for 10 min at 4°C, and the supernatant was dialyzed against ddw and lyophilized followed by mild acid hydrolysis, alditolacetate derivatization, and GC-MS analyses.