|Home | About | Journals | Submit | Contact Us | Français|
The Campylobacter jejuni capsule is important for colonization and virulence in various infection models. In most strains, the capsule includes a modified heptose whose biological role and biosynthetic pathway are unknown. To decipher the biosynthesis pathway for the 6-deoxy-d-altro-heptose of strain 81-176, we previously showed that the 4,6-dehydratase WcbK and the reductase WcaG generated GDP-6-deoxy-d-manno-heptose, but the C3 epimerase necessary to form GDP-6-deoxy-d-altro-heptose was not identified. Herein, we characterized the putative C3/C5 epimerase Cjj1430 and C3/C5 epimerase/C4 reductase Cjj1427 from the capsular cluster. We demonstrate that GDP-6-deoxy-d-altro-heptose biosynthesis is more complex than anticipated and requires the sequential action of WcbK, Cjj1430, and Cjj1427. We show that Cjj1430 serves as C3 epimerase devoid of C5 epimerization activity and that Cjj1427 has no epimerization activity and only serves as a reductase to produce GDP-6-deoxy-d-altro-heptose. Cjj1430 and Cjj1427 are the only members of the C3/C5 epimerases and C3/C5 epimerase/C4 reductase families shown to have activity on a heptose substrate and to exhibit only one of their two to three potential activities, respectively. Furthermore, we show that although the reductase WcaG is not part of the main pathway, its presence and its product affect the outcome of the pathway in a complex regulatory loop involving Cjj1427. This work provides the grounds for the elucidation of similar pathways found in other C. jejuni strains and other pathogens. It provides new molecular tools for the synthesis of carbohydrate antigens useful for vaccination and for the screening of enzymatic inhibitors that may have antibacterial effects.
Campylobacter jejuni is the leading cause of bacterial gastroenteritis and food-borne disease in the Western world and accounts for 5% of food-related deaths (1, 2). In addition, C. jejuni infection is strongly associated with Guillain-Barré syndrome, a rapidly ascending paralysis, as well as inflammatory bowel disease and reactive arthritis (3, 4). Like numerous other bacterial pathogens, C. jejuni has developed antibiotic resistance (5–7). This phenomenon may be linked to the fact that C. jejuni is commensal in poultry and cattle and that sub-therapeutic antibiotics were used for years to increase productivity (8). This exposure of C. jejuni in the commensal host has often offered the bacterium an opportunity to develop antibiotic resistance before transfer to the human host via contamination of the meat during processing or via contamination of water tables from cattle runoffs or insufficiently processed manure (8). This has prompted the need to develop novel and more effective antimicrobials and vaccines (9), which in turn require a better understanding of the mechanisms that sustain the virulence of C. jejuni.
Several virulence factors contribute to the pathogenicity of C. jejuni, including the flagella that confer motility and are important for colonization (10, 11), the glycosylation of proteins that allows antigen protection/masking (12, 13), and surface glycolipids such as the lipooligosaccharide (14), and the capsule (15) that provide resistance to a wide array of innate immune defenses (16). The capsule is required for diarrheal disease and colonization in animal models of infection, as well as adherence, invasion, and serum resistance in human infections (17). The capsule of C. jejuni includes modified heptoses that can be C6-dehydrated or O-methylated on C6 hydroxyl and that can be present in a variety of ring configurations (18–22). For example, 6-O-methyl-l-gluco-heptose and 6-deoxy-d-altro-heptose have been identified in C. jejuni NCTC 11168 and 81-176, respectively (19, 22). Although it has been demonstrated that C. jejuni mutants deficient in capsule production have an altered immunoreactivity (23) and have a drastically diminished capacity to colonize 1-day-old chicks (24), the precise role of the heptose modification is unknown. The heptose modification may contribute to capsular function, as was previously observed in Yersinia pseudotuberculosis (25, 26). Answering this question requires elucidation of the heptose modification pathway, which our laboratory has been addressing via biochemical and genetic approaches. The biochemical characterization of the enzymes will determine unambiguously the role of each capsular gene in the modification pathway and ultimately will allow using the enzymes as novel targets for the development of therapeutic inhibitors. These enzymes could also be used to produce biosynthetic carbohydrates that could serve as epitopes for vaccination. Although the chemical synthesis of a modified heptose was recently described, it is quite difficult and cumbersome (27), and enzymatic synthesis represents an appealing alternative, being more efficient and malleable.
We reported recently the biochemical characterization of the C6 dehydratase WcbK and C4 reductase WcaG that had been identified in the capsular gene cluster of C. jejuni strain 81-176 by homology with the Y. pseudotuberculosis heptose-modifying enzymes DmhA and DmhB (28, 29). Our biochemical characterization of WcbK and WcaG showed that WcbK can dehydrate GDP-d-glycero-d-manno-heptose (herein referred to as GDP-manno-heptose) to produce compound P1, the anticipated precursor for GDP-6-deoxy-d-altro-heptose (18), and that its reaction product can be reduced by WcaG to yield product P3 (Fig. 1). As monitored by NMR spectroscopy, these activities did not give rise to the expected C3 epimerization step necessary to generate the d-altro epimer that is found in the capsule. This indicated that at least one more enzyme is involved in the required C3 epimerization step. Moreover, the kcat of WcbK on GDP-manno-heptose was low in comparison with the kcat of DmhA on the same substrate (29). This suggested that WcbK may have a higher kcat on the C3-epimerized (altro) heptose, implying an alternate enzymatic order for the heptose modification pathway that may not be led by WcbK. However, structural modeling data argued against this hypothesis. Therefore, it was necessary to identify the putative C3 epimerase of the complete biosynthesis pathway and to determine unambiguously the sequence of the reactions.
To date, no GDP-manno-heptose C3 epimerase has been characterized. Cjj1430 and Cjj1427 are the only enzymes with putative C3 epimerase activity present in the capsular gene cluster of C. jejuni strain 81-176 (18). Specifically, Cjj1430 is similar to the dTDP-6-deoxy-d-xylo-4-hexulose C3/C5 epimerases RmlC from Escherichia coli (26% identical, 57% similar) and RfbC from Salmonella enterica (30% identical, 59% similar), which are involved in dTDP-l-rhamnose synthesis (30–32). Cjj1427 is similar to the GDP-fucose synthases from E. coli (31% identical, 70% similar) and Helicobacter pylori (44% identical, 59% similar), which are GDP-4-keto-6-deoxy-α-d-mannose C3/C5 epimerases/C4 reductases (GMER)2 involved in the formation of GDP-l-fucose (33, 34). Although both Cjj1430 and Cjj1427 are predicted to have C3/C5 epimerase activities, Cjj1430 is a rather small protein of 181 amino acids, and Cjj1427 is a large protein of 352 amino acids, and they exhibit little similarity with one another (13% identity, 30% similarity). Therefore, their functions are not anticipated to be redundant.
We hypothesized that both Cjj1430 and Cjj1427 would be involved with the generation of the capsule-linked d-altro-heptose in C. jejuni strain 81-176. In this work, we cloned, overexpressed, and purified the yet uncharacterized Cjj1430 and Cjj1427 enzymes from C. jejuni strain 81-176, identified the order of the enzyme activities in the complete GDP-6-deoxy-d-altro-heptose modification pathway with respect to WcbK and WcaG, and determined the reactions catalyzed by each enzyme along the pathway using a combination of capillary electrophoresis, mass spectrometry, and NMR spectroscopy analyses.
The cjj1427 and cjj1430 genes from C. jejuni strain 81-176 were PCR-amplified from genomic DNA using primers CJPGFclP1 (AGGTACCATGGGCATGCAAAAAGATTCTAAAAATT) and CJPGFclP2 (GCTGGATCCCTATTGTCTTATATTTTGCT) for cjj1427 and primers CJPG1430P1 (AGGTACCATGGGCATGGCAAAGAATTTAATATAC) and CJPG1430P2 (GCTGGATCCTTATCCTTTATTTTTATTGCT) for cjj1430. The PCR was accomplished using PfuTurbo DNA polymerase (Stratagene) according to the manufacturer's instructions, with annealing at 50 °C for cjj1430 and 45 °C for cjj1427. The PCR products were digested with NcoI and BamHI and were cloned into the pET23 derivative (35) with an N-terminal hexahistidine tag. The recombinant plasmids were transformed into E. coli DH5α with ampicillin selection (100 μg/ml). The resulting plasmids pET-cjj1427 and pET-cjj1430 were purified using the GFX kit (GE Healthcare) and were verified by DNA sequencing (Robarts Institute Sequencing Facility, London, Canada). The cloning of WcbK and WcaG in the pET vector was reported previously (28).
Expression was performed in E. coli ER2566 for Cjj1427 and E. coli BL21(DE3)pLysS for WcbK, WcaG, and Cjj1430, using LB containing 100 μg/ml ampicillin and 34 μg/ml chloramphenicol as appropriate. Protein expression was induced by the addition of 0.1 mm isopropyl β-d-1-thiogalactopyranoside, and expression was carried out at 37 °C, except for Cjj1430 (25 °C). At the end of the induction period, the cells were harvested by centrifugation (8000 × g) and stored at −20 °C until needed.
The induced cell pellets were resuspended in binding buffer (0.1 m NaCl, 20 mm imidazole, 20 mm Tris-HCl adjusted at pH 8 for WcbK and pH 7 for all other enzymes). After mechanical disruption, the histidine-tagged proteins were purified by nickel chelation using a 1.6 ml of Poros MC 20 column (4.6 × 100 mm; Applied Biosystems) as described previously (28, 29). The protein fractions of interest were analyzed by SDS-PAGE and Coomassie staining and were stored in a final concentration of 25% glycerol at −20 °C.
CE was performed on a Beckman Gold instrument using a 57-cm bare silica capillary and separation with 200 mm borax buffer, pH 9, under 26 kV as described previously (28, 29). Substrate conversion was estimated by integration of the surface areas under the substrate and product peaks detected at 254 nm using the 32 Karat software.
The enzymatic preparation of GDP-d-glycero-d-manno-heptose from sedoheptulose 7-phosphate was performed as reported previously (28, 29). The reaction product that served as a substrate in this work was purified by anion exchange chromatography as described below.
Sugar nucleotides were purified by anion exchange chromatography using a High Q EconoPac 1-ml column (Bio-Rad) and a linear gradient of triethylammonium bicarbonate, pH 8.5 (50 mm to 1 m), at 1 ml/min with UV detection (28, 29). The fractions containing the product of interest were pooled and lyophilized twice in water. Quantification was performed using a Nanodrop spectrophotometer and using ϵGTP = 12000 mol−1 liter cm−1.
All reactions contained ~0.10 mm substrate, 0.12 mm NADPH/NADP+ mix (70:30%) in 200 mm Tris-HCl, pH 8.0, and were incubated at 37 °C. The amounts of enzymes added and incubation times are specified in the legends to the figures. The reaction products were analyzed by CE as described above. Large scale reactions were prepared by direct proportional increase of all components for anion exchange purification followed by MS or NMR spectroscopy analyses.
Enzymatic reactions were linearly scaled up to contain ~3 nmol of the product and were performed in volatile ammonium bicarbonate buffer instead of Tris-HCl. They were lyophilized in H2O twice before MS analysis. Alternatively, purified reaction products were analyzed by MS.
For deuterium incorporation experiments, in a first set of analyses, the WcbK reaction was performed in buffer prepared in D2O, and once full conversion to P1 was observed (as per CE analysis), Cjj1430 was added so that its reaction also occurred in D2O. The enzymes were removed by ultrafiltration (membrane cutoff of 3 kDa). In a second set of analyses, the WcbK reaction was performed in H2O; the enzyme was filtered out by ultrafiltration, and the product was lyophilized. The product was rehydrated in D2O, and Cjj1430 that had been washed in deuterated buffer by ultrafiltration was added to the rehydrated P1 product. After the Cjj1430 reaction had occurred in D2O, the enzyme was removed by ultrafiltration. All final samples were lyophilized to near dryness and rehydrated in H2O three times before MS analysis. All reaction products were analyzed by LC-MS/MS at the Rix Protein Identification Facility of the University of Western Ontario in the negative ion mode as reported before (28, 29).
MALDI MS analyses were performed at the MALDI mass spectrometry facility of the University of Western Ontario using a 4700 Proteomics Analyzer (Applied Biosystems, Foster City, CA) in the linear positive ion mode as described before (29).
Large scale reactions were performed using 0.75 μmol of GDP-manno-heptose, 0.87 μmol of NADH, 0.20 μg of WcbK, 0.26 μg of Cjj1430, 0.24 μg of Cjj1427, and 0.30 μg of WcaG (as appropriate) in 6.25 ml of 0.2 m triethylammonium bicarbonate, pH 8.5. The reactions were incubated for 1 h at 37 °C. The reactions were filtered through a 3-kDa cutoff ultrafiltration centrifugal device (Pall Filtron) to remove the enzymes, and the reaction products were purified by anion exchange chromatography as described above. The purity of the fractions was monitored by CE. The purified products (P5 and P6) were lyophilized repeatedly after resuspension in Milli Q water (twice) and in D2O (four times).
All 1H NMR data were collected with a Varian Inova 600 MHz NMR spectrometer at 25 °C. 1H NMR spectra were collected using a 2-s presaturation pulse for products P5 and P6, or WET sequence (36) for P3, centered on the residual HDO resonance. 1H and 13C assignments for the reaction products were determined from a two-dimensional 1H TOCSY experiment (37), using a 6-Hz spin lock for 256 complex increments and natural abundance 1H-13C HSQC experiment (38, 39). In addition, selective one-dimensional 1H TOCSY and NOESY experiments were used to confirm assignments and arrangement of the pyranose ring. All spectra were processed using unshifted Gaussian weighting in VnmrJ 2.1B and 1H and 13C chemical shifts referenced to 2,2-dimethyl-2-silapentane-5-sulfonate sodium salt at 0.00 ppm.
The overexpression, purification, and characterization of WcaG and WcbK were performed as reported previously (28). Similar results were obtained in this work (Fig. 2). Like WcaG and WcbK, Cjj1430 and Cjj1427 were also overexpressed in E. coli with an N-terminal histidine tag using the pET system. High yields of soluble proteins were obtained. The proteins were purified to near-homogeneity in a single step of nickel affinity chromatography and reacted with the anti-His antibody (Fig. 2). Although Cjj1427 migrated to the expected size of 41.7 kDa on SDS-polyacrylamide gels, Cjj1430 migrated close to 25 kDa, although it was expected at 22.1 kDa (Fig. 2). MALDI MS analysis indicated that both Cjj1427 and Cjj1430 proteins had the expected molecular weight, despite the anomalous migration of Cjj1430 on SDS-polyacrylamide gels.
We previously demonstrated that WcbK can dehydrate GDP-manno-heptose to generate GDP-6-deoxy-4-keto-heptose, referred to as P1 (Fig. 1) (28). However, the kcat of this reaction is slightly lower than that of DmhA, a similar GDP-manno-heptose C6 dehydratase in Y. pseudotuberculosis (28, 29). This raised the possibility that GDP-manno-heptose was not the preferred substrate of WcbK and that the epimerization step necessary to produce the altro ring configuration may actually occur before C6 dehydration by WcbK in vivo. Hence, the potential epimerization activities of Cjj1427 and Cjj1430 on GDP-manno-heptose were examined by CE, using UV absorbance to monitor the GDP portion of the GDP-linked heptose. No novel product was observed, indicating that Cjj1427 and Cjj1430 did not catalyze the GDP-manno-heptose substrate (data not shown). Likewise, we demonstrated earlier that WcaG has no activity on GDP-manno-heptose (28). Therefore, these data demonstrate that the GDP-6-deoxy-d-altro-heptose synthesis pathway is initiated by WcbK as highlighted in Fig. 1.
As we have shown previously, WcaG can use NAD(P)H to reduce the 4-keto group of the WcbK reaction product P1 to generate P3 (Figs. 1 and and3A,3A, trace d) (28). However, NMR spectroscopy has demonstrated that C3 of this WcaG product remains nonepimerized so that the sugar is GDP-6-deoxy-d-manno-heptose. It is not in the d-altro configuration that is present in the capsule of strain 81-176 (28). This suggested that P3 may subsequently be epimerized on C3 by additional enzymes such as Cjj1430 or Cjj1427. Therefore, the activities of Cjj1427 and Cjj1430 on P3 were assessed by CE. No new product peak was detected with the addition of Cjj1430 or Cjj1427, even in the presence of cofactor, indicating that P3 is not catalyzed by Cjj1430 or Cjj1427 (data not shown).
Because the data above indicated that neither Cjj1430 nor Cjj1427 used P3 as a substrate, and P3 was not in the proper configuration to lead to the altro-heptose derivative present in the capsule, it was likely that Cjj1430 and/or Cjj1427 could convert P1 into an altro form before its final reduction. Hence, the ability of Cjj1427 and Cjj1430 to catalyze P1 was assessed by CE analysis. When P1 was incubated with Cjj1427, no catalysis was observed, indicating that Cjj1427 did not use P1 as a substrate (Fig. 3A, trace b). In contrast, when P1 was incubated with Cjj1430, the formation of a novel reaction product P4 was observed, concurrently with a stoichiometric decrease in P1 (Fig. 3A, trace c). P4 was not present in control reactions where Cjj1430 was omitted or when Cjj1430 was present and P1 was absent. These data indicate that Cjj1430 used the reaction product of WcbK (P1) as a substrate (Fig. 1).
Regardless of the concentration of Cjj1430 added, and of the time allotted for catalysis, the maximum production of P4 represented 20–25% of the total GDP-d-glycero-d-manno-heptose used for the reaction, even when large amounts of P1 were still present (Fig. 3B). The reaction yield and the reaction kinetics were not improved when NAD(P)H and NAD(P)+ were added to the reaction. Over time, P1 is known to breakdown noncatalytically to release its GDP moiety (annotated as P2) (28). Interestingly, as P1 broke down into P2, there was a corresponding decrease of P4 levels (Fig. 3B). This suggested that the Cjj1430 reaction product P4 is also unstable and breaks down over time to contribute to the rise in P2 levels observed after long incubation times and/or that Cjj1430 catalyzed an equilibrium reaction between P1 and P4.
P4 was detected at m/z 616 by MS analysis. This m/z ratio and its MS/MS spectrum were identical to those observed previously for the 6-deoxy-4-keto intermediate P1 generated by WcbK (28) and which is used by Cjj1430 to form P4 (Fig. 4A). This is consistent with the following facts: (i) Cjj1430 was predicted to perform an epimerization reaction; (ii) Cjj1430 acted on a pre-formed 4-keto intermediate, and therefore no hydride extraction was needed to carry out the predicted epimerization reaction; and (iii) no cofactor was required for the activity of Cjj1430, and therefore, no reduction of the substrate was anticipated.
When the P1/P4 mixture from WcbK/Cjj1430 reactions was incubated with Cjj1427 in the presence of NAD(P)H, the formation of a novel reaction product, P5, was observed concurrently with the stoichiometric decrease in the P1 and P4 peaks (Fig. 3C, trace b). The P5 peak was not present in reactions when Cjj1427 was omitted or when Cjj1430 was omitted. The reaction was dependent on the presence of reduced cofactors NAD(P)H and led to the production of NAD(P)+. Far greater substrate conversion was obtained in the presence of NADPH rather than NADH indicating that the preferred cofactor of Cjj1427 is NADPH. Because P1 is not a substrate for Cjj1427, these data altogether indicate that Cjj1427 uses the Cjj1430 product P4 as a substrate in an oxido-reduction reaction that oxidizes the NAD(P)H cofactor provided (Fig. 1).
MS analysis identified the Cjj1427 reaction product P5 at 618 m/z, which by MS/MS yielded a fractionation pattern consistent with the predicted C4 reductase activity of Cjj1427 (Fig. 4B). It could not be excluded at this stage that P5 also underwent C3/C5 epimerizations compared with the original substrate P4.
Interestingly, when Cjj1427 was added to a P1/P4 mixture that contained active Cjj1430 in the presence of NADPH, not only did the Cjj1430 product P4 disappear but also the P1 peak decreased considerably (Fig. 3C, trace b), although P1 is not a substrate for Cjj1427 (Fig. 3A, trace b). These data confirm that P4 is used by Cjj1427 as demonstrated above, and they also indicate that Cjj1430 was catalyzing an equilibrium reaction between P1 and P4, as suspected from the data presented in Fig. 3B. The consumption of P4 by Cjj1427 appears to pull the Cjj1430 P1/P4 equilibrium toward production of more P4, therefore consuming more P1 than in the absence of Cjj1427 (see Fig. 1 for the full pathway). Considering that these assays were performed with more than twice as many moles of Cjj1430 than Cjj1427, these data also suggest that in this pathway the reaction catalyzed by Cjj1430 is rate-limiting.
In parallel reactions performed on the Cjj1430 reaction product P4 with WcaG instead of Cjj1427, P4 was still present, although a new product appeared upstream of P1 (Fig. 3C, trace c). This new product co-migrated precisely with P3, indicating that WcaG catalyzed conversion of P1 into P3 as observed previously and did not use P4. Importantly, under these parallel reaction conditions, WcaG catalyzed approximately two times less P1 than the Cjj1430/Cjj1427 combination did (Fig. 3C). This suggests that Cjj1430 and Cjj1427, but not WcaG, preferentially bind and catalyze P1 in the physiological 6-deoxy-d-altro-heptose synthesis pathway, whereas the WcaG-mediated conversion of P1 into P3 represents a side branch (Fig. 1). Finally, it was also noted that during the generation of P3 from P1 by WcaG, significant degradation of P1 occurred if reactions were incubated over 1 h at 37 °C. This is in contrast to Cjj1430 and Cjj1427 reactions with P1 in which minimal degradation of P1 was observed during the generation of P5 (data not shown). This further suggested that the generation of P3 is slow and not physiologically preferential in comparison with the rapid generation of P4 and P5.
As mentioned above, reactions comprising GDP-manno-heptose, WcbK, Cjj1430, and Cjj1427 lead to the formation of product P5 in the presence of reduced cofactor NAD(P)H. Concurrent addition of WcaG to such reactions led to the formation of a novel reaction product P6, accompanied by reduced formation of P5 (Fig. 3C, trace d). Full conversion of the GDP-manno-heptose substrate into P6 could be obtained under these conditions (data not shown). Because we showed above that WcaG cannot use the Cjj1430 reaction product P4 as a substrate, and only generates P3 out of the WcbK reaction product P1, these data could at first glance indicate that WcaG uses the P5 Cjj1427 reaction product as a substrate to generate P6. This interpretation is particularly appealing because in the absence of Cjj1427, the WcaG-mediated P1 to P3 conversion is very slow so that most P1 compound decomposes into P2. In contrast, in Cjj1427-containing reactions (also containing all other enzymes), efficient WcaG-mediated formation of the final product P6 was observed with hardly any formation of P2.
However, when the stable P5 Cjj1427 reaction product was purified by anion exchange chromatography and incubated with WcaG, no catalysis was observed, even in the presence of reduced cofactors and of large amounts of WcaG, and upon long incubation times (data not shown). This establishes that the P5 Cjj1427 reaction product is not a substrate for WcaG. This is consistent with the fact that P5 is already a reduced compound and WcaG is a reductase that could not possibly use P5 as a substrate.
As P6 was generated via the combined activity of the four candidate biosynthetic enzymes encoded by the genomic capsular cluster, it was anticipated that P6 would be the final product of the pathway, i.e. GDP-6-deoxy-d-altro-heptose. But the fact that P6 did not arise directly from WcaG-mediated catalysis of the Cjj1427 product P5 cast doubt on this assignment. Moreover, P6 co-migrated with P3 by CE analysis (Fig. 3C, traces c and d), although CE has enough resolution power that it can discriminate between epimers (40, 41). This suggested that P3 and P6 may be identical compounds and called for MS and NMR spectroscopy analyses. The P6 product was detected at m/z 618 by mass spectrometry (Fig. 4C), and its MS/MS pattern was identical to that of the reduced GDP-6-deoxy-manno-heptose isomer P3 obtained upon reduction of P1 by WcaG (28), indicating that P6 is also a reduced compound as expected. Like P3, P6 had an MS/MS pattern consistent with that expected for GDP-6-deoxy-d-altro-heptose. Further analysis by NMR spectroscopy clearly demonstrated that P3 and P6 were identical compounds (Table 1 and Fig. 5). Specifically, the chemical shifts for all 1H shifts for P6 were within 0.02 ppm compared with P3, including a characteristic downfield shift for H2 (4.02 ppm) and upfield shift for H4 (3.50 ppm). In addition, large coupling constants were noted for H3, H4 (3J3,4 = 9.5 Hz) and H4, H5 (3J4,5 = 9.5 Hz) indicating these three protons existed in trans arrangements around the pyranose ring (42). Comparison of the 1H spectra for P3 and P6 (Fig. 5) showed they were nearly identical.
Because the P6 product obtained upon activity of all four enzymes was not the expected final altro-heptose derivative, and involvement of two reductases (Cjj1427 and WcaG) in the pathway was not consistent with the anticipated reaction scheme, we investigated whether the P5 WcbK/Cjj1430/Cjj1427 product was the altro-heptose derivative by NMR spectroscopy.
The 1H and 13C chemical shift assignments for P5 (Table 1) were determined using two-dimensional TOCSY spectra, selective one-dimensional 1H TOCSY spectra, and 1H,13C HSQC data. An unusual feature of the data were the downfield-shifted 1H resonance for H5 (4.18 ppm) compared with this proton in either P3 or P6. Similar chemical shifts have been observed for deoxy-altro, ido, galacto, and talo-heptoses (42). Further comparisons to chemical shifts for H2 (C2), H3 (C3), and H4 (C4) were most similar to the altro form of the sugar. Analysis of 1H,1H coupling constants revealed small couplings between H1,H2 (3J1,2 = 2.3 Hz), H2,H3 (3J2,3 = 3.7 Hz), and H3,H4 (3J3,4 = 3.1 Hz) that was evident from the collapsed nature of these resonances in the 1H spectrum (Fig. 5). This shows that a trans arrangement does not exist between any of these proton pairs around the sugar ring. To confirm the 6-deoxy-d-altro-heptose configuration, selective one-dimensional 1H NOE experiments were completed. These experiments showed interactions between H1 with H2, H4 with H3 and H6, and H5 with H6 that are in agreement with the altro configuration and inconsistent with other arrangements. The absence of NOEs between H3 and H5 also indicated that these protons were not on the same side of the ring as found in the talo- and galacto-heptoses.
Altogether, these data clearly establish that the sequential activities of WcbK, Cjj1430, and Cjj1427 are sufficient to produce GDP-6-deoxy-d-altro-heptose from GDP-6-deoxy-manno-heptose. These data also confirm that the extra enzyme WcaG that is encoded in the close proximity of the wcbK, cjj1430c, and cjj1427 genes in the capsular cluster is not involved in the mainstream pathway.
The data presented above involve both Cjj1430 and Cjj1427 in GDP-6-deoxy-d-altro-heptose synthesis. Cjj1430 is similar to well characterized dTDP-6-deoxy-d-xylo-4-hexulose C3/C5 epimerases involved in dTDP-l-rhamnose synthesis (30–32), and Cjj1427 is similar to well characterized GDP-4-keto-6-deoxy-mannose C3/C5 epimerases/C4 reductases involved in GDP-l-fucose synthesis (33, 34). This raises the question of the need for two putative C3/C5 epimerase candidates (Cjj1430 and Cjj1427) to form GDP-6-deoxy-altro-heptose when a single C3 epimerization step is required to convert the manno sugar configuration of the substrate to the altro configuration of the final product. This indicates that the pathway may be more complex than anticipated and/or that the enzymes do not have the anticipated activities. Three different options need to be considered: (i) either Cjj1430 catalyzes both C3 and C5 epimerization steps and Cjj1427 only performs a C5 epimerization, so that the net result of the combined Cjj1430/Cjj1427 catalysis is a C3 epimerization, (ii) or Cjj1430 only catalyzes a C5 epimerization, and Cjj1427 performs both C3 and C5 epimerizations to yield the same net result. (iii) A third option is that Cjj1430 would perform a C3 epimerization and Cjj1427 would not have any epimerization activity. Note that in all three options, Cjj1427 is performing the C4 reduction step as shown by MS.
CE analyses indicated that each of Cjj1430 and Cjj1427 makes a single product (Fig. 3, A and C) but did not provide information as to the identity of the products observed. Likewise, standard MS analyses (Fig. 4) did not allow us to determine the epimerization status of the compounds analyzed, and the exact identity of the Cjj1430 product P4 could not be determined by NMR spectroscopy as this product was too unstable. To decipher whether Cjj1430 performs one or two epimerizations on the heptose substrate, reactions were performed in the presence of deuterated water, and the products obtained were analyzed by MS. In a first set of analyses, the WcbK reaction that supplies substrate to Cjj1430 was performed in deuterated buffer, and Cjj1430 was added once full conversion to P1 had occurred. In control reactions performed with WcbK only, the P1 reaction product obtained in deuterated buffer (referred to as P1D below) is detected by MS at m/z 617 (Fig. 6A), which is 1 mass unit bigger than P1 formed in H2O that is detected at m/z 616 (28). MS/MS showed that this extra mass unit was specifically located on the heptose moiety of the sugar nucleotide as peaks assigned to the GDP moiety were not affected by the presence of deuterium (data not shown). The deuterium incorporation upon reaction with WcbK reflects the well known fact that during the 4,6-dehydratase-catalyzed reaction, there is complete exchange of the C5 proton with solvent hydrogen, which, in deuterated buffer, results in a C5 H/D change (43–45). The characteristics of the P1D product formed under these conditions can be described as C5D, C3H, and m/z 617, to account for deuterium incorporation on C5 only.
Upon reaction of this P1D with Cjj1430 in the presence of deuterated buffer, the reaction product obtained P4DD was detected at m/z 618 (Fig. 6B), which is 2 mass units more than for the Cjj1430 reaction product P4 prepared in H2O (Fig. 4A) and 1 mass unit bigger than the P1D product. It was confirmed by MS/MS that the second deuterium was also on the heptose ring (data not shown). These data indicate that, in addition to the deuterium atom that was added at C5 by WcbK, Cjj1430 introduced a second deuterium atom in the reaction product. It has been established via conversion of reaction products to alditol acetate prior to MS analysis that C3/C5 epimerases introduce a deuterium atom at both C5 and C3 when reactions are performed in D2O (30, 46). Therefore, the Cjj1430-mediated introduction of a deuterium to P1D (where C5 H/D exchange had already taken place via WcbK) indicates C3 epimerization activity of Cjj1430 and results in P4DD (C5D, C3D, m/z 618).
Because Cjj1430 catalyzes an equilibrium between P1 and P4 in favor of P1, the reaction analyzed by MS contained mostly P1 (~80%). Thus, a significant peak corresponding to P1D (C5D, C3H, m/z 617) should a priori be expected. However, the Cjj1430-catalyzed back epimerization of P4 into P1 in D2O results in progressive conversion of the original P1D product (C5D, C3H, m/z 617) into P1DD product (C5D, C3D, m/z 618) via passage through the P4DD intermediate (C5D, C3D, m/z 618). Therefore, no original P1D product remained after reaction with Cjj1430 (Fig. 6B).
The data above allowed the determination that Cjj1430 has at least C3 epimerization activity. To determine whether it also has C5 epimerization activity, a second set of reactions was performed in which P1 was formed in H2O (P1 C5H, C3H, m/z 616), and the Cjj1430 reaction was performed in D2O. The final reaction also contained ~80% P1, 20% P4, as per CE analysis. By MS, the main product peak was observed at m/z 617 (Fig. 6C). The peak at m/z 617 includes a product in which a single deuterium has been incorporated. Because we showed in the prior experiment that Cjj1430 has C3 epimerization activity, this peak corresponds to the C3 epimer P4D (C5H, C3D). If Cjj1430 also had C5 activity, incorporation of two deuterium atoms would have been observed, yielding a peak at m/z 618. However, no signal was observed at 618 beyond the expected isotopic peak. Therefore, these data indicate that Cjj1430 only has C3 epimerization activity on the heptose substrate.
Further indication that Cjj1430 only has C3 epimerization activity came from the examination of the P1 peak. If Cjj1430 only has C3 epimerase activity, P4 only has a deuterium on C3 (C5H, C3D, m/z 617). Because of the P1/P4 equilibrium, the back epimerization of P4 into P1 by Cjj1430 would not exchange protons on C5 and would yield only P1D (C5H, C3D), which is detected at the same m/z 617 as P4D. This is clearly consistent with what is observed (Fig. 6C) as follows: a single main peak at m/z 617, which includes both P4D (C5H, C3D) and P1D (C5H, C3D). If Cjj1430 had both C3 and C5 epimerization activities, two deuteriums would be incorporated, which would generate P4DD (C5D, C3D), which would be detected at m/z 618. Moreover, the back epimerization of P4DD into P1 would convert all the original P1 (C5H, C3H at m/z 616) into P1DD (C5D, C3D), which would also be detected at m/z 618. Therefore, if Cj1430 performed both C3 and C5 epimerizations, the main peak observed by MS should be at m/z 618, but this is clearly not what is observed (Fig. 6C).
Taken together, the data above conclusively determine that Cjj1430 only has C3 epimerization activity and that its reaction product is therefore GDP-6-deoxy-4-keto-d-arabino-heptose, which can lead subsequently to the expected d-altro-heptose configuration by simple reduction at C4. Our CE data showed that Cjj1427 converts the Cjj1430 reaction product P4 into P5, which we determined by NMR spectroscopy to be in the altro configuration. Because a single C3 epimerization step was required to switch the manno configuration of the heptose substrate to the altro configuration of the product P5, and Cjj1430 is responsible for this C3 epimerization, the data imply that Cjj1427 does not have any epimerization activity at all but only carries a C4 reduction step to generate the final product P5.
The data presented above raised the question of the role of WcaG in GDP-6-deoxy-altro-heptose synthesis, if any. Although Cjj1427-dependent, the activity of WcaG did not use the Cjj1427 product P5 as a substrate but involved the P1 substrate to generate P3, as documented previously (28) and as supported by the data presented in Fig. 3, A and C. Moreover, the presence of WcaG in reactions containing active WcbK, Cjj1430, and Cjj1427 led to reduced formation of P5 from GDP-manno-heptose (Fig. 3C, traces b and d). This suggests a possible inhibition of Cjj1427 activity by WcaG and/or by the WcaG reaction product P3. The inhibition of Cjj1427 would prevent pulling the Cjj1430 P1/P4 equilibrium toward P4, so there would be more P1 available for catalysis by WcaG. This would result in enhanced formation of P3, as observed. An alternative possibility is that Cjj1427 would enhance WcaG activity, therefore resulting in fast P3 (= P6) formation and reduced P5 formation.
To test the hypothesis of inhibition of Cjj1427 by WcaG and/or P3, a stock of ~75% P1, 25% P4 was generated by reacting GDP-manno-heptose with WcbK and Cjj1430 (at equilibrium), and both enzymes were removed by ultrafiltration so that the composition of the P1/P4 stock would remain constant. The P1/P4 stock was then reacted with WcaG only, Cjj1427 only, or with a mix of Cjj1427 and WcaG. Upon reaction with WcaG only, the P1 peak decreased, and the P3 peak appeared as expected (Fig. 7A, trace b). The P4 peak was not affected. Upon incubation with Cjj1427 only, there was full conversion of P4 into P5 (Fig. 7A, trace c), but the P1 peak was not affected. The full disappearance of P4, although P1 was still present, also demonstrates that ultrafiltration had been efficient at removing Cjj1430, which would otherwise replenish P4 from P1. Finally, upon incubation with both WcaG and Cjj1427 at once, P4 was still present, P1 decreased, and P3 (= P6) appeared (Fig. 7A, trace d). The fact that P4 is still present when no replenishment is possible (due to the removal of Cjj1430) clearly demonstrates that Cjj1427 activity had been inhibited by the presence of WcaG and/or its reaction product and that Cjj1427 did not use P4 to generate any intermediate that could serve as a substrate for WcaG. Therefore, these data indicate that, although there is no requirement for WcaG to produce P5, WcaG and/or its product P3 exerts regulatory effects on P5 production via inhibition of Cjj1427 activity.
To dissect out which of WcaG or its reaction product P3 inhibited Cjj1427, the effect of addition of purified product P3 on Cjj1427 activity was monitored in reactions comprising GDP-manno-heptose, WcbK, Cjj1430, Cjj1427 and cofactor NADPH. Under the conditions tested, no noticeable effect of P3 addition was noted on the activities of WcbK and Cjj1430 (Fig. 7B, traces b–e). However, against expectations, the addition of P3 resulted in a significant increase in Cjj1427 activity, as indicated by the increase in peak P5 and total disappearance of the P1 peak (full conversion in trace g). The increase in Cjj1427 activity upon P3 addition was also quantified over a wide range of P3 concentrations (up to 0.15 mm) and under conditions where the total substrate was not limiting. P5 formation increased linearly with the amount of P3 added, reaching up to 3-fold under the conditions tested (data not shown). Therefore, altogether these data demonstrate that the addition of P3 in the absence of the WcaG results in enhanced formation of GDP-6-deoxy-d-altro-heptose (P5) via enhanced activity of Cjj1427. The effect observed was specific for P3 as no effect of addition of GDP-manno-heptose in lieu of P3 was observed (Fig. 7C). Note that P3 was not consumed in the process, which is consistent with the fact that P3 is not a substrate for Cjj1427. Overall, these data indicate antagonistic effects of WcaG and its reaction product P3 on Cjj1427 activity. WcaG inhibits Cjj1427 activity, whereas its reaction product P3 enhances Cjj1427 activity (Fig. 7D). Whether sufficient amounts of P3 would accumulate in vivo for the P3-mediated activation of Cjj1427 to be physiologically relevant is unlikely because full conversion of heptose into P3 (= P6) is readily obtained in the presence of WcaG.
In addition to inhibition of Cjj1427 by WcaG and activation of Cjj1427 by the P3 product, a closer examination of the data presented in Fig. 7A indicates that the interplay between WcaG and Cjj1427 also resulted in enhanced activity of WcaG. In reactions performed on the enzyme-free mixture of P1/P4 substrates, limited activity of WcaG was observed in the absence of Cjj1427 under the conditions used for this assay (Fig. 7A, trace b), but a significant level of WcaG activity was observed in the presence of Cjj1427 (trace d). The effect was not due to the Cjj1427 product itself, because minimal amounts of Cjj1427 product P5 were formed under these assay conditions. Such effects strongly suggest the existence of interactions between WcaG and Cjj1427.
Overall, these data demonstrate the intricacies of the GDP-6-deoxy-altro-heptose biosynthetic pathway and its regulatory aspects, with a complex regulatory loop involving Cjj1427, WcaG, and its reaction product P3 (Fig. 7D). Against expectations, WcaG activity does not lead to the GDP-6-deoxy-d-altro-heptose but prevents its formation and diverts all substrate conversion toward production of GDP-6-deoxy-manno-heptose (P3/P6).
To the best of our knowledge, this is the first report investigating the complete GDP-6-deoxy-d-altro-heptose biosynthetic pathway. The analysis of this pathway was extremely difficult due to the presence of the WcaG side branch with regulatory effects on the activity of Cjj1427 and due to the production of several compounds of identical molecular weight and similar MS/MS spectra. An extra level of complexity came from the fact that Cjj1430 operates in an equilibrium mode, whose direction is opposite to the production of altro-heptose. Such unfavorable equilibrium has been observed for hexose C3/C5 epimerases before (30, 47). Finally, several intermediates of the pathway are unstable keto products (P1 and P4) that readily break down to release their GDP moiety, further complicating the analyses. Only very detailed analyses performed with a variety of combinations of enzymes, comparison of data obtained with sequential versus concomitant additions of enzymes, processing of all samples extemporaneously (to avoid degradation of unstable intermediates), and analysis of multiple time course series with different enzyme concentrations allowed us to solve this pathway. Also, NMR spectroscopy analyses of the only two stable products of the pathway (P5 and P6) were necessary to allow full elucidation of the pathway and to determine unambiguously that, unexpectedly, P5 was the actual final product of GDP-6-deoxy-d-altro-heptose, whereas P6, which was obtained using the four candidate enzymes from the capsule cluster and was anticipated to be the altro-heptose derivative, only represented the by-product of the WcaG side branch. Generating enough pure material for these NMR spectroscopy analyses was a major undertaking, considering that generating the GDP-manno-heptose substrate from sedoheptulose phosphate requires four enzymatic steps (48), and generating the final P5 and P6 products from the substrate also required multiple enzymatic steps.
The pathway was anticipated to be fairly straightforward, simply requiring sequential C6 dehydration, C3 epimerization, and C4 reduction of GDP-manno-heptose. Our detailed analyses combining CE, MS, and NMR spectroscopy analyses demonstrated that the sequential activities of the C6 dehydratase WcbK (28), putative C3/C5 epimerase Cjj1430, and putative C3/C5 epimerase/C4 reductase Cjj1427 were required to generate GDP-6-deoxy-d-altro-heptose from GDP-d-manno-heptose. Based on reaction mechanisms of known C3/C5 epimerases, it is expected that Cjj1430 and Cjj1427 perform the necessary epimerization on the heptose while it is still in the 4-keto conformation (30, 33, 34, 49–51), which is consistent with their activity downstream of WcbK.
The requirement for two candidate C3/C5 epimerases (Cjj1430 and Cjj1427) was counter-intuitive because a single C3 epimerization step was necessary. Mass spectrometry analysis of reaction products generated in the presence of deuterium was necessary to resolve the issue. Such experiments have been used previously to characterize the C3/C5 epimerization activities of similar enzymes (30, 50, 51). Our experiments on deuterated reaction compounds clearly demonstrated that Cjj1430 only has C3 epimerization activity on the heptose substrate, which leads to the 4-keto-d-arabino sugar intermediate P4 (Fig. 1), and implied that Cjj1427 served as a reductase only. The fact that the P4 reduction step was exclusively carried out by Cjj1427 and that WcaG only reduced P1, which only differs from P1 by its C3 configuration, indicates that the configuration at C3 is a critical determinant of the specificity of these reductases.
It remains a bit of a puzzle that an enzyme such as Cjj1427, with potential for three activities (C3 and C5 epimerase activities and C4 reductase activity) only serves as a reductase, and a potentially bifunctional enzyme such as Cjj1430 only serves as a single epimerase. However, there seems to be an evolutionary component to the current state of the pathway in C. jejuni 81-176. In this strain, a C5 epimerization by Cjj1430 would need to be reverted by Cjj1427, therefore resulting in a fairly high energetic cost to switch the configuration back and forth. However, a permanent d to l configuration switch is in fact necessary for the formation of the l-gluco-heptose derivative present in C. jejuni strain NCTC 11168. This strain harbors a Cj1430 enzyme that is 81% identical and 98% similar to Cjj1430 from strain 81-176. Therefore, both enzymes can be anticipated to perform the same C3 epimerization activity in each strain. In contrast, although strain NCTC 11168 also contains a Cjj1427 homologue (Cj1428), it is only 57% identical and 92% similar to Cjj1427 from strain 81-176. Therefore, Cjj1427 and Cj1428 may differ in their C5 epimerization activities, whereby Cj1428 from NCTC 11168 may perform the necessary C5 epimerization step, although Cjj1427 from strain 81-176 cannot. Obviously, the biochemical activities of these enzymes will be worth investigating to determine whether this is the case.
A striking feature of this pathway is the direction of the equilibrium catalyzed by Cjj1430. The equilibrium lies in favor of P1, which is opposite to the direction of the mainstream pathway. Similar observations have been made for other C3/C5 epimerases, such as RmlC (30) and the GDP-mannose 3,5-epimerase (GME) (47). Only the reductase activity of Cjj1427 commits the pathway toward irreversible formation of the final product, effectively pulling the P1/P4 equilibrium toward P4. In the absence of Cjj1427 activity, the pathway would therefore generate copious amounts of P1, which would be readily transformed into P3 by WcaG.
The role of WcaG in this pathway is a puzzle. As schematically represented in Fig. 8, WcaG can reduce an intermediate (P1) from the GDP-6-deoxy-d-altro-heptose synthesis pathway; it affects the activity of Cjj1427 and therefore the outcome of the whole pathway, and it is affected in return by the presence of Cjj1427. Its activity is nevertheless not necessary for GDP-6-deoxy-d-altro-heptose synthesis in vitro. Because none of the enzymes of the pathway can use P3 as a substrate, P3 cannot merely represent a temporary storage of intermediates for the pathway. In addition, as also schematically represented in Fig. 8, we showed that P3 activates Cjj1427 in vitro. Although one could argue that these in vitro observations represent mere artifacts, we have observed by knock-out mutagenesis that WcaG is necessary for modified heptose synthesis in C. jejuni NCTC 11168.3 Considering that WcaGNCTC11168 is 97% identical to WcaG81-176, it can safely be assumed that WcaG81-176 and the WcaG81-176-dependent side branch that leads to P3 production do have in vivo relevance.
Although C3/C5 epimerases and C3/C5 epimerases/C4 reductases that function on deoxy-hexoses, have attracted much attention for their potential use as antibacterial targets (30, 33, 34, 50, 51), no such enzyme involved in heptose modification pathways has been characterized at the biochemical level to date. These enzymes also represent novel antibacterial targets because they are likely involved in the modification of the heptoses found in virulence-associated surface carbohydrates of a variety of pathogens, including several strains of C. jejuni (19, 21, 22), Campylobacter lari (52), as well as Burkholderia species (53). Also, BLAST searches reveal the presence of non-RmlC and non-GMER homologues (as defined as not identified by BLAST searches performed with the RmlC and GMER protein sequences) of Cjj1430 and Cjj1427 in a variety of bacteria for which the presence of modified heptose has not yet been demonstrated. For example Helicobacter bilis and winghamensis, several Serratia and several Prevotella species harbor non-RmlC homologues of Cjj1430 (e values 5e−58, 1e−53, 8e−24, and 1e−21 respectively), and Helicobacter hepaticus, Francisella philomiragia, Prevotella tannera, and Spirochaeta thermophila harbor non-GMER homologues (e values 1e−116, 7e−114, 4e−102, and 6e−112, respectively). Therefore, these homologues may also represent novel targets worthy of investigation in pathogenic Helicobacters, Prevotellae, Serratiae, and Francisellae. Inhibitors designed against the C. jejuni enzymes characterized in this study may also prove of therapeutic value against these other pathogens.
The data presented in this study will provide a comparison basis that will greatly facilitate the elucidation of similar heptose modification pathways, such as that of the 6-O-methyl-l-gluco-heptose that is present in C. jejuni NCTC 11168. As mentioned above, this pathway is also predicted to involve two candidate C3/C5 epimerases similar to Cjj1430 and Cjj1427 but that are bound to display slight differences in substrate/product specificities. Future structural and site-directed mutagenesis studies will allow defining the molecular basis for substrate specificity in this class of enzymes, which could lead to the rational design of inhibitors.
Overall, beyond the fundamental understanding of the intricacies of this pathway, the information generated in this study has the potential to lead to novel therapeutic or prophylactic agents useful to limit the production of essential virulence factors (capsule or lipopolysaccharide) in a variety of pathogens.
We thank Kathy Barber for help preparing some NMR samples and acquisition of NMR data and Dr. Anne Rintala-Dempsey for maintenance of the Biomolecular NMR Facility.
*This work was supported by Natural Sciences and Engineering Research Council of Canada Operating Grant RGPIN 240762-2010 (to C. C.) and Canadian Institutes of Health Research Grant MOP 93520 and the Canada Research Chairs Program (to G. S. S.).
3C. Creuzenet, A. Wong, and D. Lange, unpublished data.
2The abbreviations used are: