CP (capsular polysaccharide) is an important virulence factor during infections by the bacterium Staphylococcus aureus. The enzyme CapF is an attractive therapeutic candidate belonging to the biosynthetic route of CP of pathogenic strains of S. aureus. In the present study, we report two independent crystal structures of CapF in an open form of the apoenzyme. CapF is a homodimer displaying a characteristic dumb-bell-shaped architecture composed of two domains. The N-terminal domain (residues 1–252) adopts a Rossmann fold belonging to the short-chain dehydrogenase/reductase family of proteins. The C-terminal domain (residues 252–369) displays a standard cupin fold with a Zn2+ ion bound deep in the binding pocket of the β-barrel. Functional and thermodynamic analyses indicated that each domain catalyses separate enzymatic reactions. The cupin domain is necessary for the C3-epimerization of UDP-4-hexulose. Meanwhile, the N-terminal domain catalyses the NADPH-dependent reduction of the intermediate species generated by the cupin domain. Analysis by ITC (isothermal titration calorimetry) revealed a fascinating thermodynamic switch governing the attachment and release of the coenzyme NADPH during each catalytic cycle. These observations suggested that the binding of coenzyme to CapF facilitates a disorder-to-order transition in the catalytic loop of the reductase (N-terminal) domain. We anticipate that the present study will improve the general understanding of the synthesis of CP in S. aureus and will aid in the design of new therapeutic agents against this pathogenic bacterium.
- bifunctional enzyme
- capsular polysaccharide
- crystal structure
- cupin fold
- short-chain dehydrogenase/reductase
- Staphylococcus aureus
The proliferation of antibiotic resistance challenges our ability to effectively combat disease [1,2]. This situation is particularly problematic in connection with the opportunistic pathogen Staphylococcus aureus. S. aureus is a commensal bacterium residing in ~20% of the general population. Although generally harmless, S. aureus may cause life-threatening infections through sporadic breaches of the local defences [3–5]. In fact, S. aureus is a leading cause of nocosomial infections in hospitals worldwide . Resistance to antibiotics of last resort such as methicillin and vancomycin further increases the risks associated with this bacterium [1,7,8]. It is thus desirable to boost the quality and quantity of our therapeutic arsenal against this pathogen.
CP (capsular polysaccharide) and its biosynthetic machinery are potentially useful therapeutic targets to combat S. aureus. CP is a polymeric carbohydrate attached to the outer surface of S. aureus that enhances pathogenesis by rendering the bacterium resistant to phagocytosis [9–14]. More than 70% of clinical isolates of S. aureus belong to either the CP5 or CP8 serotypes. These strains synthesize a polysaccharide containing repeating units of N-acetyl-L-fucosamine, N-acetyl-D-fucosamine and N-acetyl-D-mannosamine uronic acid.
The biosynthetic pathway of nucleotide-activated UDP-L-FucNAc (uridine diphosphate N-acetyl-L-fucosamine) is well conserved among several pathogenic bacteria producing CP, such as S. aureus, Pseudomonas aeruginosa, Staphylococcus pneumoniae and Bacteroides fragilis . Mechanistic studies have shown that UDP-L-FucNAc is produced from the precursor UDP-D-GlcNAc (uridine diphosphate N-acetyl-D-glucosamine) through the sequential action of three enzymes CapE, CapF and CapG in S. aureus [15,16]. These three enzymes catalyse a total of five chemical transformations (Figure 1).
The first two enzymes CapE and CapF are proposed to belong to the SDR (short-chain dehydrogenase/reductase) superfamily of proteins. This idea is supported by sequence homology, and by prediction of a conserved catalytic triad and a classical nucleotide-binding motif GXXGXXG [15,17]. However, no high-resolution structural data for any of these enzymes has been reported so far. Detailed knowledge of their three-dimensional structures is needed to clarify their mechanism and function. Moreover, obtaining high-resolution structural information is a necessary step towards designing antibiotic therapeutics based on the inactivation of these enzymes.
In the present study, we report the new crystal structure of enzyme CapF obtained in two different space groups at resolutions of 2.45 Å (1 Å=0.1 nm) and 2.7 Å. The structure revealed that CapF is a homodimer arranged in a unique architecture not observed previously. CapF comprises a larger N-terminal region that adopts a typical SDR fold, and a smaller C-terminal region displaying a classical cupin motif with a bound Zn2+ metal ion. Mutational analysis demonstrated that each domain catalyses one enzymatic reaction: the cupin motif is responsible for an epimerization step, whereas the SDR domain catalyses the subsequent NADPH-dependent reduction.
Expression and purification of untagged full-length CapF
Full-length SeMet (selenomethionine)-labelled and wild-type CapF were produced and purified as described previously . Briefly, Escherichia coli Rosetta2 (DE3) cells expressing SeMet-labelled or wild-type protein were harvested, lysed by sonication and purified in a two-step procedure. First, we employed anion-exchange chromatography in a HiTrap Q XL column (GE Healthcare). Secondly, fractions containing CapF were subjected to gel-filtration chromatography in a HiLoad 26/60 Superdex 200 column (GE Healthcare) equilibrated with 50 mM Tris/HCl, pH 8.0, and 500 mM NaCl. Fractions containing CapF were concentrated with a 10 kDa Centriprep filtration unit (Millipore). Protein concentration was determined spectrophotometrically at 280 nm using the calculated molar extinction coefficient of CapF . Protein was stored at −20°C in 40% (v/v) glycerol.
Expression and purification of CapF variants
Muteins and deletion constructs of CapF with a His6 tag at the N-terminal end were expressed in E. coli Rosetta2 (DE3) cells. A scheme with all of the constructs prepared in the present study is shown in Supplementary Figure S1 (at http://www.BiochemJ.org/bj/443/bj4430671add.htm). Cells were harvested by centrifugation at 7000 g for 10 min at 4°C, and washed with lysis buffer containing 50 mM Tris/HCl, pH 8.0, and 500 mM NaCl. Cells were disrupted on ice for 15 min using a sonicator (Tommy) set at 150 W. Cell lysate was centrifuged at 40000 g for 30 min at 4°C, and insoluble debris was discarded. Supernatant was loaded on to a HisTrap column (GE Healthcare) equilibrated with a buffer containing 50 mM Tris/HCl, pH 8.0, and 500 mM NaCl. Protein bound to the affinity column was initially washed with 20 ml of lysis buffer and was subsequently eluted with the same buffer supplemented with 0.5 M imidazole. Protein fractions containing CapF were further purified on a HiLoad 26/60 Superdex 200 gel-filtration chromatography column (GE Healthcare) equilibrated with a solution containing 50 mM Tris/HCl, pH 8.0, and 500 mM NaCl. Fractions of CapF were concentrated and stored as above. The UV–visible spectrum showed no evidence of NADPH bound to CapF.
Expression and purification of CapE
Full-length CapE with a His6 tag at the N-terminal end was expressed in E. coli Rosetta2 (DE3) cells, and purified as described for CapF above. CapE fractions were concentrated and stored at −20°C in a solution containing 40% (v/v) glycerol. Protein concentration was determined spectrophotometrically at 280 nm using the calculated molar extinction coefficient of the protein (ϵ=19200 M−1·cm−1).
Crystals of SeMet-labelled CapF suitable for X-ray diffraction analysis were obtained by the hanging drop method. We mixed 1 μl of fresh protein solution at 15 mg·ml−1 in 10 mM Tris/HCl, pH 8.0, and 1 μl of crystallization solution composed of 100 mM Mes, pH 7.2, 20% (v/v) glycerol, 100 mM (NH4)2SO4, 300 mM NaCl and 3.9 M sodium formate. Diamond-shaped crystals grew to an approximate size of 0.3×0.2×0.2 mm3 within 8 weeks. Alternatively, crystals of SeMet-labelled CapF were obtained in a solution containing 17.5% (w/v) PEG [poly(ethylene glycol)] 3350 and 0.2 M sodium malonate.
Data collection and phasing
Suitable crystals of CapF were identified, harvested and mounted under a stream of cold nitrogen (100 K) at beamline BL17A of the Photon Factory (Tsukuba, Japan). Three datasets each at a different wavelength were collected on the basis of the absorption edge of selenium: 0.9792 Å (peak, maximum f”); 0.9796 Å (edge, minimum f'); and 1.0000 Å (remote). Diffraction data were indexed and processed with HKL2000 software . Phase determination and construction of an initial model was performed using the automated module of the SOLVE/RESOLVE program [21–23].
We refined data of SeMet-labelled crystals of CapF belonging to trigonal (P3221) and orthorhombic (C2221) space groups. Refinement of crystals of SeMet-labelled protein has been used as a surrogate for native protein in numerous structures (e.g. [24–26]). Data were processed using MOSFLM , and were merged and scaled using the SCALA program of the CCP4 suite . Atomic co-ordinates of CapF obtained using SOLVE/RESOLVE were used as an input model to determine the structure of CapF in the trigonal space group by molecular replacement with the program PHASER . Data were further refined using REFMAC5  with TLS (Translation/Libration/Screw) parameterization , and COOT . During all of the refinement procedures SeMet replaced methionine residues. The refined structure of the trigonal crystal was used as the input model to determine the structure of CapF in the orthorhombic space group with PHASER . Refinement was carried out as described for the trigonal crystal. Because the data of the orthorhombic crystal was collected at the peak of the anomalous scattering of selenium, coefficients f' and f” of selenium were incorporated in the refinement with REFMAC using separate Bijvoet pairs F+ and F− . Nonetheless, the contribution of anomalous scattering was minimal as demonstrated by the small statistical differences with respect to a mock refinement using averaged amplitudes [less than 0.6% for each Rfactor and Rfree, and RMSD (root mean square deviation) between equivalent atomic co-ordinates below 0.1 Å]. Model quality was assessed with PROCHECK . The small difference between Rfree and Rfactor in the structure of the trigonal crystal (Rfree−Rfactor=2.2%) is probably explained by its very high solvent content (Supplementary Table S1 at http://www.BiochemJ.org/bj/443/bj4430671add.htm). Co-ordinate and structure factor files corresponding to refined models have been deposited in the PDB under accession codes 3ST7 and 3VHR.
Identification of Zn2+
The identity of the heavy metal ion bound in the cupin pocket was determined by X-ray anomalous diffraction analysis [34–37]. This procedure required two datasets collected at a wavelength before (1.270 Å) and after (1.300 Å) the absorption edge of Zn2+. Each dataset was indexed using MOSFLM and scaled using SCALA to a resolution of 3.5 Å. We used PHASER to obtain an initial model of CapF, followed by refinement using REFMAC5 (until Rfree<24%). Anomalous difference electron density maps (Bijvoet-difference Fourier map) were calculated using programs SFALL, CAD, FFT and MAPMASK of the CCP4 suite .
The CapF substrate was generated in situ by employing the enzyme CapE (which precedes CapF in the biosynthetic pathway of UDP-L-FucNAc). A typical enzymatic assay contained 1.7 mM substrate UDP-D-GlcNAc (Wako), 0.7 mM coenzyme NADPH or NADP+ (Wako), and enzymes CapE (His6-tagged) and CapF (untagged wild-type or His6-tagged muteins; Supplementary Figure S1) both at a concentration of 2 μM. The buffer used was 20 mM Tris/HCl, pH 8.0, and 10 mM MgCl2 in a total volume of 100 μl. Reaction mixtures were incubated at 37°C for 2 h. Enzymatic reactions were stopped by addition of 100 μl of ice-cold phenol/chloroform/isoamyl alcohol in a 25:24:1 molar ratio. Supernatant containing sugars were mixed with 100 μl of chloroform and analysed by HPLC using a CarboPac PA1 anion-exchange column (Dionex) as described previously . Identification of monosaccharides was partly on the basis of previous reports [15,16]: substrate UDP-D-GlcNAc, 13.4 min; intermediate 2 UDP-2-acetamido-2,6-dideoxy-α-D-xylo-4-hexulose, 13.9 min; intermediate 3 UDP-2-acetamido-2,6-dideoxy-β-L-arabino-4-hexulose, 20.2 min; product UDP-2-acetamido-2,6-dideoxy-β-L-talose, 12.9 min; and coenzyme NADPH or NADP+, 17.9 min.
ITC (isothermal titration calorimetry)
Binding constants and thermodynamic parameters of the interaction between CapF and nucleotides NADPH or NAD+ were determined with an ITC200 instrument (GE Healthcare) at 25°C. Samples were equilibrated with a buffer composed of 20 mM Hepes and 150 mM NaCl at pH 8.0. Aliquots of a solution containing nucleotide NADPH at a concentration of 500 μM were injected stepwise into a cell filled with a solution of CapF at a concentration of 50 μM. The concentrations of NADP+ and protein in the titration with NADP+ were increased to 3.80 mM and 96 μM respectively to observe the binding isotherm under low-affinity conditions . Heat generated from the dilution of NADPH and protein was negligible, but that of NADP+ had to be subtracted prior to data analysis. Titration curves were fitted to a one-site binding isotherm .
DSC (differential scanning calorimetry)
The thermal stability of CapF was determined in a VP-capillary calorimeter (GE Healthcare). Protein samples at a concentration of 50 μM were equilibrated in a solution containing 50 mM Hepes, pH 8.0, and 200 mM NaCl. Thermograms were recorded between 283 K and 373 K at a rate of 1 K·min−1. The buffer baseline was subtracted and data were normalized by protein concentration using Origin software.
RESULTS AND DISCUSSION
CapF adopts a unique two-domain architecture
The crystal structure of CapF in trigonal space group P3221 was solved by the multi-wavelength anomalous diffraction method and refined to a resolution of 2.45 Å (Table 1). The asymmetric unit consisted of a single polypeptide chain (residues 1–369). CapF adopted a characteristic dumb-bell-shaped architecture comprising two domains. The N-terminal domain included residues Met1 to Pro251, whereas the C-terminal domain was composed of residues Leu252 to Val369 (Figure 2A).
At three-dimensional search in the DALI server  found high homology between the N-terminal domain of CapF and the SDR superfamily of proteins (Supplementary Table S2 at http://www.BiochemJ.org/bj/443/bj4430671add.htm) [43,44]. SDR enzymes are characterized by a nucleotide-binding Rossmann motif [45,46]. An analogous analysis showed that C-terminal domain adopted a cupin fold [47,48]. The cupin domain of CapF displayed a heavy metal atom bound at the bottom of the β-barrel. We unambiguously identified the metal as Zn2+ by X-ray anomalous scattering analysis (Supplementary Figure S2 at http://www.BiochemJ.org/bj/443/bj4430671add.htm). Although Rossmann and cupin domains are ubiquitous in nature with more than 45000 members of each family in the Pfam database , the structure of CapF described in the present study constitutes, to the best of our knowledge, the first documented example of a protein combining both motifs within the same polypeptide chain.
Crystallographic symmetry analysis revealed that CapF forms a homodimer stabilized by a very large interaction surface (Figure 2B). Analysis of the contact interface in the PISA server  indicated that CapF buries 6800 Å2 of surface area upon dimer formation (3400 Å2 from each monomer). This extensive surface is composed mostly of residues of the C-terminal domain (2124 Å2, 63% of the total). The N-terminal domain contributed with a smaller interaction surface (1276 Å2, 37%). Protein dimerization was verified by gel-filtration chromatography, confirming that CapF is also a homodimer in solution (Supplementary Figure S3 at http://www.BiochemJ.org/bj/443/bj4430671add.htm).
Active site of the reductase domain is disordered in a second crystal structure
We determined a second crystal structure of CapF in the orthorhombic space group C2221 at a resolution of 2.7 Å (Table 1). This independent structure superimposed well on the trigonal structure, achieving an average RMSD value of 0.84 Å over 1371 mainchain atoms (Figure 3). The contribution of each domain to the dimer interface was nearly identical with that in the trigonal crystal (interaction surface=3336 Å2; cupin, 63%; SDR, 37%). A loop of the SDR domain containing residues Gly55 to Phe65 could not be traced in the electron density maps of the orthorhombic crystal. The probable reason was that these residues clashed with a neighbouring protein chain of an adjacent unit cell of the crystal. These same residues displayed the highest temperature factors in the trigonal structure, indicating a propensity to become disordered (Figure 4). Other SDR proteins such as RmlB  or the dehydrogenase domain of ArnA  undergo order–disorder transition of selected loops (not belonging to the catalytic site) in response to cofactor and/or substrate binding.
Strikingly, the loop containing the predicted catalytic residues Tyr103 and Lys107 of the reductase domain was also disordered in the orthorhombic structure (Supplementary Figure S4 at http://www.BiochemJ.org/bj/443/bj4430671add.htm). Specifically, the residues between Ser95 and Gln109 could not be traced in the difference electron density (Fo−Fc) maps. We did not find evidence of clashes between this loop and other regions of CapF in the crystal structure that could explain the dynamic disorder. The high mobility of this loop clearly contrasted with the orderly electron density features seen in the same loop of CapF crystallized in the trigonal space group (Supplementary Figure S4).
This interesting observation suggested that the active site of CapF undergoes a large structural remodelling during its catalytic cycle. To the best of our knowledge a similar order–disorder transition in the active site of this family of proteins has not been described previously. These fluctuations could be related to the unique primary sequence of CapF in this region compared with similar SDR proteins. In fact, the catalytic loop of CapF comprises between 16 and 31 fewer residues than that of homologous enzymes (Supplementary Figure S4). The relative smaller size of this region of CapF could facilitate an enhanced dynamic behaviour in response to external stimuli such as substrate and/or coenzyme binding.
Cofactor is needed only in the second reaction catalysed by CapF
Armed with the valuable structural information above, we embarked on a detailed structure–function analysis of CapF. Enzymatic assays were first employed to assess the cofactor requirements of CapF. We note that the enzyme CapE was employed to generate the substrate of CapF in situ (see Figure 1). HPLC chromatograms are shown in Figure 5, and integrated data are displayed in Table 2.
Control experiments showed that the substrate UDP-D-GlcNAc was not modified when incubated with NADPH alone (Figure 5A). However, catalytic amounts of enzyme CapE with or without coenzymes generated a new major peak corresponding to intermediate 2 (only the reaction with NADPH is shown; Figure 5B). We also observed a minor peak at higher retention times that was tentatively assigned to intermediate 3. Appearance of this minor peak was probably caused by enolization of intermediate 2, similar to what occurs with the homologous protein FlaA1 . However, we cannot completely exclude that this peak corresponded to a reaction by-product.
The presence of enzymes CapE and CapF in the reaction mixture (with or without oxidized coenzyme NADP+) resulted in a major peak at high retention times, which we interpreted as epimerization of intermediate 2 at the C3 position (Figures 5C and 5D). Finally, when the substrate was incubated with the three active components CapE, CapF and NADPH, it generated the expected product UDP-2-acetamido-2,6-dideoxy-β-L-talose (Figure 5E). The yield was quantitative within experimental error (mean±S.D.=44.6±3.7%) when taking into consideration that NADPH was the limiting reactant in our enzymatic assay (the molar concentration of NADPH was 41% that of the substrate UDP-D-GlcNAc). The appearance of product was coupled to consumption of intermediate 3, but not of intermediate 2.
Binding of NADPH is enthalpy-driven
We have shown above that the reduction step required one equivalent of NADPH. However, because neither NADPH nor NADP+ were observed in the crystal structure of CapF, we determined their binding properties by ITC (Figure 6 and Table 3).
Titration of CapF with the reduced coenzyme NADPH gave rise to a typical sigmoid binding isotherm. Thermodynamic parameters indicated that binding of NADPH to CapF was driven by a very favourable enthalpy change (ΔH=−16.7±0.2 kcal·mol−1; 1 kcal=4.184 kJ) and displayed moderate affinity (Kd=1.6±0.2 μM). Binding of NADPH to CapF was strongly opposed by the entropy term (−TΔS=8.8 kcal·mol−1), which accounted for approximately half of the absolute value of the enthalpy change. One monomer of CapF binds one molecule of NADPH (n=0.91±0.01).
The binding isotherm of NADP+ was characterized by markedly lower affinity for the apoenzyme. The dissociation constant determined for NADP+ (Kd=61±6.1 μM) was ~40-fold higher (lower affinity) than that for NADPH. The enthalpy change was significantly reduced in absolute terms (ΔH=−9.8±0.8 kcal·mol−1). Binding of NADP+ to CapF was also opposed by an unfavourable entropy term (−TΔS=4.0 kcal·mol−1). Overall, the affinities of both NADPH and NADP+ were smaller than those reported for similar enzymes [54,55], which explains why NADPH or NADP+ were not found in the crystal structures of CapF.
Importantly, the affinity gap between NADPH and NADP+ resulted in −2.1 kcal·mol−1 of free energy that can be utilized by the enzyme in each catalytic cycle. We propose that the order–disorder transition observed in the active-site loop of the SDR domain (Figure 4 and Supplementary Figure S4) could be coupled to the thermodynamic switch described just above. We expect that new crystal structures of CapF with either NADPH or NADP+ will reveal the identity of the residues interacting with the coenzyme.
Identification of catalytic domains
Two different constructs of His6–CapF, the first one comprising residues of the reductase domain (residues 1–243; CapFREDUCTASE) and the second one composed of residues of the cupin domain (residues 252–369; CapFCUPIN), were cloned, expressed and purified to homogeneity. Each construct was assayed in the presence of substrate, CapE and NADPH (Table 4). However, none of these two constructs were active when acting separately, or when added together into the reaction mixture. The structural integrity of each domain, evaluated by CD, ruled out that enzymatic inactivation was caused by large-scale unfolding (Supplementary Figure S5 at http://www.BiochemJ.org/bj/443/bj4430671add.htm) . These experiments demonstrated that the intact topology of the native dimer was needed to preserve catalytic prowess in CapF.
To further investigate the enzymatic mechanism of CapF, we prepared two different muteins of full-length CapF containing point mutations in key residues: (i) at the catalytic site of the SDR domain (mutein S94A/Y103A); and (ii) belonging to the co-ordination sphere of Zn2+ (H290L). These constructs maintained a native-like dimeric conformation in solution, as indicated by size-exclusion chromatography (results not shown). Their enzymatic activity was measured as described above (Table 4 and Figure 7).
We noticed that none of these two muteins generated a final product. However, as shown below, their mechanism differed greatly. The HPLC elution profile of mutein H290L resembled that of assays without CapF, indicating that this mutein was completely inactive (compare Figures 7A and 7C). However, the activity of mutein S94A/Y103A resembled that of native CapF in the absence of coenzyme NADPH, i.e. it catalysed the reaction to intermediate 3, but not the reduction step because the SDR domain was catalytically dead (compare Figures 7D, 5C and 5D).
Importantly, CapFREDUCTASE rescued the full enzymatic activity of mutein S94A/Y103A (Figure 7E). When these two constructs are together in the reaction mixture, the intermediate 3 species generated by mutein S94A/Y103A (at its intact cupin domain) was properly processed by variant CapFREDUCTASE. We concluded that the catalytic potential of SDR remained intact even when the cupin domain was absent from the primary sequence. In contrast, the lack of activity of the CapFCUPIN construct suggested that this catalytic unit is very sensitive to changes in the architecture of the overall enzyme.
The integrity of the Zn2+-binding pocket of the cupin domain influenced the overall thermostability of the enzyme (Figures 8A and 8B). Removal of Zn2+ after incubation with 1.0 mM EDTA diminished the overall stability of the protein by 15°C (Figure 8C). The stability of CapF was recovered upon reconstitution with Zn2+. Similarly, mutein H337L and mutein H288A/H209A (both constructs modified the direct co-ordination sphere of Zn2+) suffered from diminished thermal stability (Supplementary Figure S6 at http://www.BiochemJ.org/bj/443/bj4430671add.htm). We also note that muteins F297Y and T364Y of the second co-ordination sphere of Zn2+ (Figure 8A) exerted a minor effect on the activity of CapF (Table 4), in contrast with the full inactivation observed in H290L.
Comparison with RmlC and RmlD
The present study has revealed clear parallelisms between CapF and the pair of enzymes RmlC/RmlD. These enzymes catalyse similar sequential reactions to those of CapF, but in the biosynthetic pathway of L-rhamnose (a diastereoisomer of L-fucose). Specifically, RmlC catalyses a double C5,C3-epimerization reaction [57–61], whereas RmlD is responsible for the subsequent NADPH-dependent reduction of the dTDP-4-hexulose species generated by RmlC [62,63].
Although the overall three-dimensional structure of RmlC and the cupin domain of CapF are very similar (RMSD=1.7 Å over 95 residues) and the substrate of RmlC fits nicely in the pocket of CapF, their constellations of catalytic residues differed substantially from each other (Figure 9A and Supplementary Figure S7 at http://www.BiochemJ.org/bj/443/bj4430671add.htm). RmlC does not display a Zn2+ ion (or metals) in the pocket of the β-barrel motif. Likewise, Zn2+ (or metals) are not required for enzymatic activity. All of these differences precluded unambiguous identification of the catalytic residues of the cupin domain based exclusively on structural comparisons.
RmlD and the N-terminal domain of CapF displayed homologous three-dimensional structures (RMSD=2.4 Å over 189 residues) and both enzymes required the coenzyme NADPH for catalysis (Figure 9B). NADPH and the substrate analogue of RmlD (dTDP-L-rhamnose) fitted well on to the binding pocket of CapF upon superposition of both structures. We found severe bumps (distances below 2.2 Å) at only four residues of CapF in the binding configuration shown in Figure 9(B): the substrate analogue bumped into Asn132 and Asn162, whereas NADPH bumped into Arg33 and Pro59. This rough model thus could recapitulate the approximate orientation of coenzyme and intermediate 3 when bound to CapF.
The biggest difference between the aligned residues of CapF and RmlD occurred in Loop55–72 of CapF, which corresponded to Loop63–84 of RmlD (Figure 9B). Loop55–72 of CapF is largely shifted towards the solvent compared with the equivalent region of RmlD (Figure 9B). In RmlD this loop stabilizes the binding of NADPH through several hydrogen bonds (distance≤3.0 Å) with Ala63 and Thr65 (not shown). In contrast, Loop55–72 of CapF bumps with the adenosine moiety of NADPH in this configuration. Intriguingly, we have shown above that a large portion of Loop55–72 of CapF was the most dynamic region within the crystal structure of CapF (Figures 3 and 4).
On the basis of this comparison, we suggest that Loop55–72 of CapF closes towards the core of the enzyme upon binding of the coenzyme and/or substrate. All of this evidence suggested that the structure of apoCapF described in the present study corresponded to an ‘open’ form in the SDR domain. A similar mechanism to that proposed in the present study has been documented in another homologue SDR protein known as RmlB (Supplementary Table S2). Loop82–102 of RmlB swings between an open and a closed form upon binding of the substrate [46,51,64]. We expect to see analogous closure movements in forthcoming crystal structures of CapF.
We solved the three-dimensional structure of CapF in two different crystal forms. CapF forms a distinctive homodimer with a unique topology built on two functional domains. Comparative analysis demonstrated that the N-terminal domain belongs to the SDR family of proteins, whereas the C-terminal domain is homologous to the cupin motif and displays a bound Zn2+ ion. Each domain possesses a functional role in the reactions catalysed by CapF. The cupin domain is necessary for the C3-epimerization reaction and does not require coenzyme. Meanwhile, the N-terminal domain catalyses NADPH-dependent reduction of the UDP-4-hexulose intermediate. Because the crystal structures corresponded to an open form of the SDR domain, we proposed that CapF experiences a large reconfiguration of its active site during each catalytic cycle. In fact, binding and release of coenzyme from the SDR domain depended on the thermodynamic properties of the interaction. Looking ahead, we will need to address why CapF of S. aureus evolved this unique enzyme containing both an epimerase and a reductase domain within a single polypeptide chain.
Takamitsu Miyafusa designed the research, performed experiments, analysed data and edited the manuscript prior to submission. Jose Caaveiro designed the research, performed experiments, analysed data and wrote the paper. Yoshikazu Tanaka designed research, performed experiments and analysed data. Kouhei Tsumoto designed the overall study, analysed data, provided guidance and edited the paper prior to submission.
This work was supported in part by a Grant-in-Aid for General Research from the Japan Society for the Promotion of Science [grant number 21360398 (to K.T.)]. T.M. was supported by a predoctoral fellowship from the Japan Society for the Promotion of Science [grant number 09J07066].
We thank members of Photon Factory (Tsukuba, Japan) for their assistance during X-ray data collection. We thank Professor M. Kuroda (Center for Pathogen Genomics, National Institute of Infectious Diseases, Tsukuba University) and Professor T. Ohta (Department of Microbiology, Tsukuba University) for providing plasmids containing full-length CapF and CapE. We also appreciate useful suggestions from Professor M. Yao and Professor I. Tanaka (both at the Division of Biological Science, Hokkaido University).
The atomic co-ordinates and the structure factors of CapF crystallized in space group P3221 and space group C2221 will appear in the PDB under accession codes 3ST7 and 3VHR respectively.
Abbreviations: CP, capsular polysaccharide; ITC, isothermal titration calorimetry; RMSD, root mean square deviation; SDR, short-chain dehydrogenase/reductase; SeMet, selenomethionine; UDP-L-FucNAc, uridine diphosphate N-acetyl-L-fucosamine; UDP-D-GlcNAc, uridine diphosphate N-acetyl-D-glucosamine
- © The Authors Journal compilation © 2012 Biochemical Society