2.1. Cloning

pgaB was amplified by PCR from E. coli JM109 chromosomal DNA and cloned into a modified pET28 expression vector using BamHI and XhoI restriction sites. The primer sequences were CGGGATCCGCCCAGTCAAGAACATCATTTATACCG and CCGAGCCTCGAGTTACTGGAGGTTTTCGTCATAAAC. The PCR product was digested with BamHI and XhoI at 310 K for 2 h before purification using a QIAquick PCR Purification Kit (Qiagen). The purified PCR product was ligated into the cut vector using T4 DNA ligase (Wako) at room temperature for 1 h. The ligation mixture was used to transform E. coli DH5, and pET28b-pgaBN was prepared from cultures using QIAprep (Qiagen). This construct directs expression of residues Gln24-Gln330 of PgaB with a hexahistidine tag at the N-terminus that is cleavable with TEV protease.

2.2. Expression and purification

pET28b-pgaBN was transformed into E. coli BL21 (DE3) and cells were grown at 310 K with shaking in 3 l LB medium containing kanamycin (50 µg ml^-1). When the OD₆₀₀ of the culture reached 0.5-0.6, PGABN expression was induced by adding IPTG to a final concentration of 0.5 mM and growth was continued overnight at 288 K. The cells were collected by centrifugation at 3000g at 277 K for 30 min. The pellet was suspended in 50 mM Tris-HCl pH 8.0, 0.1 M NaCl and was then lysed by sonication on ice. The lysate was centrifuged at 38 000g and 277 K for 30 min. The supernatant solution was loaded onto a 10 ml nickel Sepharose column (GE Healthcare) equilibrated with 50 mM Tris-HCl pH 8.0, 0.1 M NaCl, 10 mM imidazole; after washing, it was eluted with 50 mM Tris-HCl pH 8.0, 250 mM imidazole. The major protein fractions were collected and digested with TEV protease overnight at 277 K during dialysis into 50 mM Tris-HCl pH 8.0, 0.1 M NaCl. The protease:PgaB ratio was 1:50. The protein was reloaded onto the washed nickel Sepharose column and eluted with 50 mM Tris-HCl pH 8.0, 0.1 M NaCl. The pooled fractions containing PGABN were dialyzed into 50 mM Tris-HCl pH 8.0, 0.1 M NaCl before concentration to 30 mg ml^-1 using Amicon centrifugal filter units (Millipore).

2.3. Crystallization and structure determination

Crystallization experiments were performed at 293 K using the hanging-drop vapour-diffusion method. Crystals grew in 16%(w/v) PEG 3350, 0.18 M sodium acetate, 0.1 M bis-Tris-HCl pH 6.5. Data were collected on beamline 17A of the Photon Factory, Tsukuba. The highest resolution data were collected from a crystal which had been soaked briefly in 1 mM mercury chloride but which appeared to be native in phasing trials. The data were used in the final refinement and revealed a bound Hg atom with low occupancy on refinement. A total of 250 images of 1° oscillation were collected for each data set. Data processing and scaling were carried out with HKL-2000 and SCALEPACK (Otwinowski & Minor, 1997). The space group was found to be P2₁, with two molecules in the asymmetric unit. Data statistics are given in Table 1. Multiple-wavelength data were collected to 2.5 Å resolution on the same beamline using a crystal soaked in 2 mM mercury chloride for 14 h. A single data set to 2.3 Å resolution was also collected using a crystal soaked in 2 mM K₂PtCl₄ for 24 h. The native data set and the Pt-soak data set were collected using incident radiation of 1.000 Å wavelength. Phases were calculated using PHENIX (Adams et al., 2010). Model building was carried out with ARP/wARP (Langer et al., 2008; Morris et al., 2003) and Coot (Emsley et al., 2010; Emsley & Cowtan, 2004). Refinement was carried out with REFMAC (Murshudov et al., 2011) and the CCP4 suite (Winn et al., 2011). Noncrystallographic symmetry restraints and TLS group refinement were not applied. The Ramachandran plot showed several residues in unusual positions, but the agreement between the two copies of the molecule was very good. Slight main-chain disorder at residues 66 and 100 led to several pairs of equivalent residues in chains A and B lying at different positions in the Ramachandran plot. Two Ramachandran outliers were the active-site residues His55 and Asp115. Isotropic temperature factors were refined with default restraints. Figures were prepared with PyMOL (DeLano, 2002). Water molecules were checked manually. A single Hg atom was modelled into the structure with an occupancy of 20%. Density around the Zn atom of chain B showed less than full occupancy of the acetate, and a partially ordered water molecule was also modelled at this site. The final model and structure factors have been deposited in the Protein Data Bank as entry 3vus .

Table 1 Data-collection and refinement statistics Values in parentheses are for the outer shell. Data set Native Hg, remote Hg, inflection Hg, peak Pt Space group P21 P21 P21 Wavelength (Å) 1.00000 0.99321 1.00938 0.99957 1.00000 Unit-cell parameters (Å, °) a = 39.6, b = 53.1, c = 144.2, = 95.2 a = 39.0, b = 51.9, c = 144.7, = 95.5 a = 39.3, b = 53.0, c = 144.4, = 95.5 Resolution range (Å) 50.0-1.65 (1.68-1.65) 50.0-2.50 (2.54-2.50) 50.0-2.30 (2.34-2.30) Reflections (measured/unique) 337723/63364 87449/19886 88848/19847 86805/19838 106033/25579 Completeness (%) 96.7/83.7 98.0/96.6 98.1/97.1 98.0/97.3 96.0/86.2 Rmerge# (%) 4.9/37.5 4.9/19.1 4.4/13.3 5.4/12.4 7.5/39.8 Multiplicity 4.9 4.4 4.5 4.4 4.2 <I/(I)> 49.0 44.3 47.9 48.2 34.9 Refinement statistics   Resolution range (Å) 25.0-1.65           R factor/free R factor 0.205/0.257           R.m.s.d. bond lengths (Å) 0.027           R.m.s.d. bond angles (°) 2.29           No. of water molecules 175           Average B factor (protein/water) (Å2) 33.4/35.2           Ramachandran plot, residues in (%)     Most favourable regions 87.6             Additional allowed regions 11.1             Generously allowed regions 0.9             Disallowed regions 0.4         #Rmerge = , where Ii(hkl) is the intensity of an observation, <I(hkl)> in the mean value for that reflection and the summations are over all equivalents.

3.1. Overall structure

Initial sequence analysis of the pgaB gene from E. coli suggested that the deacetylase activity resided in an N-terminal domain (PGABN) of roughly 300 residues in length. A threading search of known structures using the Wurst server (Torda et al., 2004) suggested a strong match to aldolase from Trypanosoma brucei (PDB entry 1f2j ), which has 14% sequence identity to the PgaB N-terminus (Chudzik et al., 2000). This TIM -barrel structure has no enzymatic activity in common with PgaB, but has eight structure repeats, whereas the CE4 enzymes have seven. PGABN shows sequence similarity to PdaA, an N-deacetylase from Bacillus subtilis, which is a CE4 member of known structure (Blair & van Aalten, 2004). On the basis of these results, a construct was made to express PgaB from Gln24 to Gln330 by PCR from E. coli genomic DNA, omitting the signal peptide and associated cysteine residue. This construct yielded 25 mg purified protein (with the N-terminal histidine tag removed) per litre of culture. Analytical ultracentrifugation showed the protein to exist as a monomer in solution (data not shown).

Crystals were grown that diffracted to almost 1.6 Å resolution and the structure (shown in Fig. 1a) was solved using two heavy-atom derivatives and multiple-wavelength anomalous measurements. Two copies of the molecule were found in the asymmetric unit. The ordered residues visible in the electron-density map began at Pro43 and ended at Val308 or Gln309, so that roughly 20 disordered residues at each end of the polypeptide were not modelled. EXAFS analysis of a native crystal clearly showed the presence of zinc, a metal not purposely added to the protein at any stage of preparation or purification. In each copy of the molecule a single metal ion was readily located and modelled (Fig. 1b). A search for related models in the PDB using DALI (Holm & Rosenström, 2010) yielded the best match as SlCE4 (PDB entry 2cc0 ), an acetylxylan esterase from Streptomyces lividans involved in plant cell-wall degradation which has a structure similar to PdaA (Taylor et al., 2006). Both enzymes are members of Pfam family PF01522. Overlaying the structures with SSM (Krissinel & Henrick, 2004) matched 128 residues of PdaA with PGABN with 18% sequence identity, giving an C r.m.s.d. of 2.3 Å. Similar results were obtained matching PGABN to SlCE4, giving an alignment of 138 residues (19.6% identity) and 2.4 Å r.m.s.d. Straightforward sequence searches against the PDB gave much poorer alignments, with the best fits being about 30% identical sequences of roughly 50 residues in length. Although most of these hits were to CE4 enzymes, they also included a subunit of the yeast ribosome and clearly were not all biologically relevant. Examination of the models quickly showed that PGABN is circularly permuted with respect to known CE4 members and that the last -strand of their TIM-like barrel corresponds to the first strand of PGABN. Simple sequence alignments of PGABN with other members of the CE4 family therefore suggested quite different matches to those found from comparisons of the actual models and did not match conserved active-site residues. A sequence alignment between SlCE4 and PGABN is shown in Fig. 2(a), demonstrating the shift of one strand of the central barrel from the N-terminus of PGABN to the C-terminus of other CE4 enzymes. The result of fitting PGABN and SlCE4 by SSM is shown in Fig. 2(b).

Figure 1 (a) A stereoview of the C trace of the PgaB N-terminal domain (PGABN), looking into the central barrel. The trace is coloured from blue (Pro43) to red (Val308), with -helices shown as coils and -strands as arrows. The figure was drawn using PyMOL (DeLano, 2002). Secondary structure was determined automatically. (b) The C trace of the PGABN monomer, coloured as in (a), is shown on the left, with the active-site zinc ion shown as a grey sphere. Residues coordinating the metal are shown as sticks, with O atoms coloured red and N atoms blue. The right-hand panel shows a surface representation of the monomer in the same orientation coloured by electrostatic potential. The zinc ion and acetate ligand were omitted from the calculation of the potential and are shown as ball-and-stick models over the protein surface. The strong negative charge of the active site is apparent. The remainder of the protein surface shows no distinct pattern of charge.

Figure 2 (a) A sequence alignment based on the crystal structures of PGABN and SlCE4 (PDB entry 2cc0 ). Conserved residues are shown in white on a red background. Red triangles show residues that coordinate the active-site zinc ion and blue triangles show nearby residues involved in catalysis. The residue numbering refers to PGABN. The first -strand of PGABN (residues 48-55) matches the last strand of SlCE4, the sequence of which has been shifted from the C-terminus to the N-terminus in this alignment. His55 of PGABN matches His155 of SlCE4. The figure was produced by ESPript (Gouet et al., 1999). (b) Overlay of the C traces of the PgaB N-domain and SlCE4 (PDB entry 2cc0 ), showing the helices and strands of both structures, but not coil. Alignment of the models was carried out with SSM (Krissinel & Henrick, 2004). Each structure is coloured from blue to red (N-terminus to C-terminus) as in (a), but circular permutation of the sequences leads to a marked difference in colouring, despite the close structural similarity. His55 of PGABN lies close to the active site, overlapping His155 of SlCE4, which is a conserved histidine in the CE4 family.

3.2. Active site

The active site of the CE4 family is found at the centre of the barrel-like -sheet and involves a metal ion in some, but not all, cases. PdaA notably has no metal in the active site on purification and does not bind zinc ions even when these are added; cadmium ions do bind, but only at very high concentrations (Blair & van Aalten, 2004). Other members of the CE4 family have been reported to require cobalt for activity or to prefer cobalt to zinc (Taylor et al., 2006). Nevertheless, two active-site residues (histidine and aspartic acid) were identified from a comparison of the PdaA and PgaB sequences; mutation of either residue in PgaB to alanine (D115A and H184A) blocked activity (Itoh et al., 2008). PGABN shares the common coordination pattern of a zinc ion bound by two histidines and an aspartic acid residue (Fig. 3). The initial crystallization screen and subsequent optimization showed that the inclusion of acetate in the buffer greatly improved the crystal quality, and an acetate ion can be found coordinated to the metal ion of each monomer in the asymmetric unit. However, in one of these sites the occupancy seems to be less than 1.0, and a water molecule with partial occupancy was also modelled close to the zinc ion. Well diffracting crystals could not be grown in the absence of acetate, suggesting that this ligand stabilized the protein. Attempts to remove acetate by soaking the crystals in acetate-free buffer limited the diffraction resolution.

Figure 3 A stereo figure of the 2mFo - DFc electron-density map for the refined model covering the active site. Electron density is shown at a level of 1. The zinc ion is shown as a grey sphere and water molecules are shown as red spheres. The acetate ion coordinating the zinc ion is clearly visible in the density map and is found to hydrogen-bond to the conserved His55. The conserved LTXDDG motif of the NodB domain includes Asp114 and Asp115.

Comparing PGABN with PdaA, it can be seen that many features of the active site are preserved. 128 C atoms from the core of the structures show roughly 17% sequence identity, and the zinc-binding residues His184 and His189 of PGABN closely overlap with His124 and His128 of PdaA. Loss of metal binding by PdaA is explained by the fact that it has no third coordinating residue; Asp115 of PGABN is replaced by an asparagine residue (Asn74) that points away from the active site (Fig. 4a). Other conserved residues of the active site of PdaA such as Phe98, Trp187 and Leu220 are not found in PGABN, the fold of which is quite different from PdaA in the region of the latter two residues. The active-site residue Asp73 of PdaA is preserved as Asp114 in PGABN, and in both enzymes this aspartic acid forms a salt bridge with an arginine side chain. This arginine residue (Arg163 in PdaA and Arg289 in PGABN) is found on a different -strand of the barrel; this does not reflect the altered order of the secondary-structure elements but is a consequence of a quite different spatial geometry. Another conserved histidine residue (His55) lies close to the active site at the end of the first -strand in PGABN, but its equivalent in PdaA (His222) lies at the end of the last -strand of the barrel. In PdaA this histidine forms a salt bridge (His222-Asp193), but the aspartate has no equivalent in PGABN and His55 is only stabilized by a water molecule. However, His55 does hydrogen-bond to the acetate placed in density near the zinc ion (Fig. 3) and must contact the substrate in the active site. The ion pairs in this region of PdaA have been proposed to play a role in stabilizing charges during catalysis (Blair & van Aalten, 2004), but the details of any such mechanism presumably differ in members of the CE4 family with or without active-site metal ions.

Figure 4 (a) A stereo overlay of PGABN (green) and PdaA (light blue) showing the active site. Labels in black indicate the residues in PGABN. O atoms are coloured red and N atoms blue. The models were fitted by overlapping 128 C atoms of the conserved core residues. The zinc ion of PGABN is shown as a grey sphere. PdaA (PDB entry 1w17 ) does not bind metal ions owing to the loss of the coordinating aspartate residue. Instead, Asn74 points away from the active site. Arg163 forms a salt bridge to Asp73, but the equivalent arginine in PGABN is Arg289, which sits on a different -strand to Arg163 of PdaA. (b) Stereo overlay of the active sites of PGABN (green) and SlCE4 (brown). Both structures have a bound zinc ion with very similar coordination geometry. Other residues around the active site show low sequence conservation between the two structures, and Tyr117 of PGABN has no counterpart in SlCE4.

Overlaying SlCE4 with PGABN gives a similar pattern to that with PdaA (Fig. 4b). Although SlCE4 is metal-dependent and PdaA has no metal centre, these two enzymes are not circularly permuted relative to one another, and the Asp73-Arg163 pair of PdaA is exactly mirrored by Asp12-Arg100 in SlCE4. His55 of PGABN is preserved as His155 in SlCE4. Thus, although the three enzymes share a common overall three-dimensional structure, each has an active site with characteristics not found in the others. PGABN and SlCE4 presumably share a common mechanism and the altered residues around the active site reflect the substrate preference.

3.3. Substrate binding

CE4 enzymes have a highly distorted barrel that lacks one of the repeats of regular TIM barrels, which creates a groove into which the extended polymer substrates of these enzymes can fit. PdaA has several positively charged residues lining this pocket (such as Lys34, Arg35 and Arg166) that are proposed to be involved in binding to the negatively charged peptidoglycan substrate. PGABN has no equivalent residues and acts upon an uncharged substrate; its enzyme activity is also presumably strongly affected by the insertion of two loops blocking one end of the substrate groove, apparently a feature unique to PGABN among the CE4 enzymes (Fig. 5). Residues Thr146-Phe164 face residues Tyr190-Pro202 on the opposite side of the groove. Neither loop appears to be stably fixed in position and residues 195-200 are not modelled in one copy of the molecule, but the presence of these loops may help to control access to the active site. A further loop (Arg207-Thr218) is more distant from the active site but may impede a long polymeric substrate from easy access to it. Overall, the structured domain of PGABN is rather longer than SlCE4 owing to these and other loops, which decorate its surface. Not all of these loops appear to be mobile; the surface loop from Ser276 to Pro288 appears to be fixed in place by hydrogen bonds formed between the side chain of Gln283 and the main-chain atoms of neighbouring residues.

Figure 5 Two views of the C trace of PGABN rotated 90° relative to one another about a horizontal axis. The zinc ion is shown as a grey sphere. Surface loops near the active site are shown in blue. These regions are presumably flexible in solution and have no counterpart in SlCE4 or PdaA. They may bind to PgaA in the periplasm. Residues 195-200 are not visible in the electron-density map for the copy of the molecule shown.

Repeated attempts to grow well diffracting crystals of PGABN without acetate failed. The reason why acetate, the product of the reaction, strongly promotes crystallization is unclear from the structure, but the enzyme appears to be product-inhibited. As with previous structures from this family, we were therefore obliged to attempt to fit substrate analogues into the binding site by docking (Blair et al., 2005). No clearly preferred binding sites were found which can be confidently predicted to bind single glucosamine residues, but the polymeric substrate may bind by weak interactions at several adjacent sites, just as a protease may require a minimum length of peptide in order to cut a single peptide bond. In vivo, it is also possible that the substrate is held in proximity to PgaB by interaction with other proteins in the synthetic machinery.