Three-dimensional structure of RBcel1, a metagenome-derived psychrotolerant family GH5 endoglucanase

The crystal structure of the psychrotolerant endoclucanase RBcel1, an enzyme implicated in bacterial cellulose synthesis, has been determined to 1.4 Å resolution.

Abstract

RBcel1 is an endoglucanase belonging to glycoside hydrolase family 5 subfamily 5 (GH5_5) that was recently identified from a soil metagenome library from the Antarctic. Unlike its closest structural homologue (Cel5A from Thermoascus aurantiacus), this enzyme was reported to be able to catalyze transglycosylation reactions and has putatively been implicated in the bacterial cellulose-synthesis process. Here, the structure of RBcel1 at 1.4 Å resolution, solved by molecular replacement, is reported. The structure and putative substrate-binding site are described and compared with those of other GH5_5 subfamily members.

1. Introduction  

RBcel1 was isolated from the metagenomic library PP1 by using functional and genotypic screening. This library was constructed from Antarctic soil samples in order to isolate novel cold-adapted enzymes, i.e. proteases, laccases, esterases/lipases and glycoside hydrolases (Berlemont et al., 2011 ▶). A detailed biochemical analysis revealed that RBcel1, which was first identified as an endoglucanase, is also able to catalyze the polymerization of cellulosic material and therefore behaves as a so-called ‘transglycosylase’ (Berlemont et al., 2009 ▶). Furthermore, based on genomic sequence homology to Pseudomonas stutzeri, Berlemont and coworkers proposed that RBcel1 is physiologically involved in bacterial cellulose production. Indeed, several bacteria (including the Pseudomonas genus) are known to be able to synthesize cellulose as an extracellular matrix for mechanical and chemical protection (Bielecki et al., 2005 ▶). To date, at least four proteins have been proposed to be required for cellulose biosynthesis: a cellulose synthase, a cyclic di-GMP-binding protein, an enzyme involved in cellulose crystallization and a cellulase (Römling, 2002 ▶). However, the involvement of the cellulolytic enzyme in cellulose biosynthesis remains unclear. Several hypotheses have been proposed, including the release of a lipid-linked polysaccharidic precursor (Römling, 2002 ▶), adjustment of the cellulose fibril length (Robledo et al., 2012 ▶) and control of the alignment of individual β-­1,4-glucan strands to form cellulose microfibrils (Mazur & Zimmer, 2011 ▶). In this paper, we report the crystal structure of the psychrotolerant metagenome-derived RBcel1. Such knowledge could help to advance studies of the biological role of bacterial GH5 glycosidases involved in cellulose synthesis.

2. Materials and methods  

2.1. RBcel1 production and purification  

RBcel1 production and purification were performed as described previously by Berlemont et al. (2009 ▶), except that production was carried out at 301 K. The protein concentration was determined from the absorbance at 280 nm using a Specord 50 spectrophotometer (Analytik Jena; Δ∊ = 80 455 M ⁻¹ cm⁻¹; Wilkins et al., 1999 ▶). The purity of the enzyme was checked by densitometry analysis on 15% SDS–PAGE gels.

2.2. Crystallization  

Initial crystallization screening of RBcel1 stored in 20 mM sodium phosphate pH 6.4 buffer was performed by hanging-drop vapour diffusion at 292 K using Qiagen EasyXtal Tool plates. Drops consisting of 2 µl protein solution and 2 µl mother liquor were equilibrated against a reservoir containing 500 µl precipitant solution. Hampton Research Crystal Screen and Crystal Screen 2 and the Sigma PEG Grid Screening Kit (Sigma–Aldrich) were used in the assays. The most promising condition (0.1 M Tris–HCl pH 8.0, 30% PEG 600) was optimized by varying the PEG and Tris concentrations at different pH values.

The best diffracting crystals were obtained with a protein concentration of about 10 mg ml⁻¹, a precipitant concentration of 17%(w/v) PEG 600 and 0.1 M Tris–HCl buffer pH 6.0. Single crystals appeared after 1 d and grew to maximum dimensions within two weeks.

2.3. Data collection and processing  

Diffraction data were collected on the FIP-BM30a beamline at ESRF, Grenoble, France using a Quantum 315r CCD detector (ADSC, USA) from a crystal in a cold nitrogen stream at ∼100 K (Oxford Cryosystems Ltd). For cryoprotection, additional PEG 600 was added to the crystal, bringing the final concentration to 25%(w/v). 270 images were recorded at a wavelength of 0.9797 Å with an exposure time of 20 s, an oscillation range of 1° per image and a crystal-to-detector distance of 140.9 mm. Diffraction data were processed using the XDS program package (Kabsch, 2010 ▶). The data-collection and refinement statistics are presented in Table 1 ▶.

Table 1. Data-collection and processing statistics.

Values in parentheses are for the outer shell.

Data collection
Diffraction source	FIP-BM30A
Wavelength ()	0.9797
Temperature (K)	100
Detector	ADSC Quantum 315r
Crystal-to-detector distance (mm)	140.90
Rotation range per image ()	1
Exposure time per image (s)	20
No. of images	270
Processing software	XDS/XSCALE [xdsme]
Space group	P212121
Unit-cell parameters ()	a = 52.2, b = 63.09, c = 98.92
Completeness (%)	99.0 (94.5)
Cutoff	3.0
Multiplicity	9.5 (8.0)
I/(I)†	20.5 (3.3)
R r.i.m. ‡ (%)	6.4 (68.4)
Overall B factor from Wilson plot (2)	11.9
Refinement statistics
Resolution range ()	30.01.38 (1.401.38)
Total No. of reflections	642628 (74389)
No. of unique reflections	67049 (9218)
R work/R free	0.154/0.176
No. of non-H atoms
Protein	2656
Tris	24
Water	557
Total	3237
R.m.s. deviations
Bonds ()	0.01
Angles ()	1.35
Average B factors (2)
Protein	14.54
TRS401	16.63
TRS402	22.95
TRS403	14.81
Water	30.43
Overall	17.27
Ramachandran plot
Favoured regions (%)	98.24
Additionally allowed (%)	1.76
Outliers (%)	0

^† I/(I) is the average intensity divided by its standard deviation.

^‡ R _r.i.m. = , where I_i(hkl) is the ith intensity measurement, N(hkl) is the number of observations of reflection hkl, including symmetry-related reflections, and I(hkl) is its average intensity.

2.4. Structure solution and refinement  

Structural analyses were performed using the PHENIX software package (v.1.6.4-486; Adams et al., 2010 ▶). The initial structure of RBcel1 was determined by molecular replacement with the AutoMR Wizard using the coordinates of Cel5A from Thermoascus aurantiacus (PDB entry 1gzj; Lo Leggio & Larsen, 2002 ▶) as the search model. The initial model was then rebuilt using the AutoBuild Wizard followed by numerous reiterative cycles of manual building in the Crystallographic Object-Oriented Toolkit (Coot; Emsley et al., 2010 ▶) and restrained maximum-likelihood refinement with phenix.refine. The final model included 321 amino-acid residues, three Tris molecules identified from clear difference electron density (F _o − F _c) and 557 water molecules in the asymmetric unit. The stereochemical quality of the model was assessed using MolProbity (Chen et al., 2010 ▶).

2.5. Molecular-dynamics simulations  

The RBcel1–cellotetraose and glycosyl-enzyme complexes were soaked in a neutralized water box containing 0.9% NaCl. The box extended 15 Å around all atoms. A distance cutoff of 10 Å was used for the van der Waals interactions, while the electrostatic interactions were treated by the Particle Mesh Ewald (PME) method, i.e. within periodic boundary conditions. The geometry of the whole system was optimized using the YAMBER3 force field (Krieger et al., 2004 ▶). A molecular-dynamics simulation at 298 K was then run for 10 ns. In order to let the system reach equilibrium, snapshots after 7.5 ns were analyzed and compared. Several starting conformations were considered, as well as different protonation states for some His, Asp and Glu residues in the catalytic site. The simulations were performed with the program YASARA (http://www.yasara.org). The cello­tetraose molecule used in this study was from the crystallographic structure of cellulase E1 from Acidothermus cellolyticus in complex with cellotetraose (PDB entry 1ece; Sakon et al., 1996 ▶).

3. Results and discussion  

3.1. Overall structure  

According to the amino-acid sequence-based classification, RBcel1 was recently assigned to subfamily 5 of the family 5 glycoside hydrolases (GH5_5; GenBank accession No. ACO55737; Aspeborg et al., 2012 ▶). The GH5 family is classified within the GH-A clan (also described as the 4/7 superfamily), which gathers enzymes with related structures that cleave glycosidic bonds with net retention of anomeric configuration through a double-displacement mechanism involving the formation of a covalent glycosyl-enzyme intermediate.

Fig. 1 ▶ displays the overall structure of RBcel1. As expected for a member of the 4/7 superfamily, the structure possesses a typical TIM-barrel fold consisting of a core composed of an eight-stranded parallel β-sheet surrounded by eight α-helices. Compared with the canonical fold, the structure has only three extra elements of secondary structure: a two-stranded antiparallel β-sheet and two one-turn helices (Figs. 1 ▶ a and 1 ▶ b). A disulfide bond between Cys270 and Cys312 was found connecting the L_α7β8 and L_CT loops. There are three cis-peptides in the structure between Leu19 and Pro20, Trp276 and Ala277, and Gly316 and Pro317. The nonprolyl cis-peptide bond involving Trp276 occurs near the end of β-strand 8 and is conserved in the GH5 family and some other members of the 4/7 superfamily (Domínguez et al., 1996 ▶; Juers et al., 1999 ▶).

Figure 1.

Overall structure and active site of RBcel1. (a) Amino-acid sequence of RBcel1. Secondary structures are shown above the sequence. (b) Ribbon stereo diagram showing the eightfold β/α-barrel. (c) Stereoview of the molecular surface of RBcel1 showing the hypothetical −2, −1, +1 and +2 sugar-binding subsite positions. β-Strands, α-helices and loops are shown in orange, blue and grey, respectively. The noncanonical elements of secondary structure are shown in green. The disulfide bond (diS) is shown in pink. Side chains of the active-site residues are represented as stick models and are coloured according to the amino-acid sequence in (a).

At the C-terminal face of the β-barrel, an open crescent-shaped cleft (30 Å long, 8 Å deep and between 6 and 13 Å wide) runs across the surface of the protein from β-strand 1 to β-strand 6 (Figs. 1 ▶ b and 1 ▶ c). A similar cleft is observed in all of the other GH5 structures and is described as accommodating the active site. Two glutamate residues (Glu135 and Glu245) are located at the bottom of the cleft at the C-­terminal ends of β4 and β7 on a single β-bulge and the so-called ‘crankshaft-like structure’ (a double β-bulge; Harris et al., 1996 ▶; Pickersgill et al., 1998 ▶), respectively. These glutamate residues are strictly conserved in GH5 family members, in which they have been shown to be catalytic residues (Davies, 1998 ▶). Besides other structural constraints, the β-bulges are needed to orient the side chains of the catalytic residues with their carboxylate moieties positioned some 5 Å apart (4.5 Å in RBcel1). Such a distance is consistent with the formation of the glycosyl-enzyme intermediate and the expected ‘retaining hydrolytic mechanism’ (Vocadlo & Davies, 2008 ▶).

3.2. Structural comparison with subfamily members  

A structural similarity search was performed with the atomic coordinates of RBcel1 using the DALI server v.3 (Holm & Rosenström, 2010 ▶). The DALI server identified five ‘strong matches’ with Z-score > 28 (the cutoff was set to n/10 − 4, where n is the number of residues in the query structure; Holm et al., 2008 ▶). The two highest DALI Z-scores of 40.4 and 34.2 were associated with the endo­glucanases Ta_Cel5A from T. aurantiacus (PDB entry 1gzj) and Hj_Cel5A from Hypocrea jecorina (PDB entry 3qr3; Lee et al., 2011 ▶). These endoglucanases are the only two GH5_5 subfamily members with known structures. The other most similar structures found by DALI were endoglucanases belonging to either GH5_25 or GH5_37, suggesting that these subfamilies are the closest structural homologues in the GH5 family.

The overall folds of RBcel1, Ta_Cel5A and Hj_Cel5A are very similar (Supplementary Fig. S41). In particular, they all share equivalent noncanonical elements of secondary structure (α_a and α_b α-­helices and β_a and β_b β-strands; Fig. 1 ▶ b), suggesting that these elements are characteristic of the GH5_5 subfamily. However, there are several significant disparities between these structures. In the immediate vicinity of the active site, the main structural difference involves the L_β1α1 loop (Supplementary Figs. S1 and S4a). This loop is substantially longer in Hj_Cel5A and projects towards the active site, while at the equivalent position in RBcel1 and Ta_Cel5A a small cavity is connected to the main crescent-shaped cleft (orange arrows in Supplementary Figs. S2a and S2b). The occluding loop in Hj_Cel5A redefines the active site, which seems even more tailored to accommodate a cellulose strand (orange arrows in Supplementary Figs. S2c and S4a). This suggests that the cavity excrescence is not essential for endoglucanase activity.

3.3. Substrate-binding cleft of RBcel1  

As mentioned above, the active site of RBcel1 is positioned in the C-terminal region of the β-barrel. As in other members of subfamily 5, it is essentially made up of β-strands (at the bottom) surrounded by loops. The cleft can be subdivided into three parts. Positioned close to the active reaction centre, the central part is formed by the C-­terminal ends of the β₆ and β₇ β-strands and the N-­terminal halves of the L_β4α4, L_β5αb and L_β6α6 loops (Fig. 1 ▶ b). The typical mechanistically important amino acids of GH5 enzymes (Asn134, Glu135, His199, Tyr201 and Glu245 in RBcel1) are found within this part of the cleft (Fig. 2 ▶). All of these residues are essential for the catalytic hydrolysis. Glu135 is proposed to act as a proton donor (Glu A/B) and Glu245 as a nucleophile (Glu nuc). Asn134 stabilizes the transition states (Correia et al., 2011 ▶) by making a critical hydrogen bond to the sugar 2-hydroxy group at subsite −1 (the nomenclature is as described by Davies et al., 1997 ▶). The equivalent of His199 has been proposed to correctly position the Glu nuc side chain (Ducros et al., 1995 ▶) and to be directly involved in the catalytic mechanism by acting as an electron donor or acceptor between the two glutamates (Zheng et al., 2012 ▶). To date, the preponderant role of the Tyr201 equivalent remains unclear. Different studies suggest that this residue may orient and activate the Glu nuc (Sakon et al., 1996 ▶; Correia et al., 2011 ▶) or stabilize the positively charged transition state (Sidhu et al., 1999 ▶), or may be the proton donor that mediates the formation of the glycosyl-enzyme intermediate (Kim & Ishikawa, 2011 ▶). A molecular-dynamics simulation performed on RBcel1 showed that Tyr201 interacts with the O atom of the ester linkage forming the glycosyl-enzyme intermediate. This residue seems to be well placed for involvement in the hydrolysis of the intermediate. The catalytic centre is flanked on both sides by wide and narrow clefts (corresponding to the nonreducing and reducing ends of the substrate-binding cleft; Fig. 1 ▶ c). Negatively numbered subsites (Davies et al., 1997 ▶) are found in the wider cavity, which is principally made up of the C-terminal end of the β₈ β-strands and the N-terminal halves of the L_β1α1, L_β3αa and L_β8α8 loops (Fig. 1 ▶ b). On the other side, the positive subsites are built up by both the L_β5αb and the L_β6α6 loops.

Figure 2.

Close-up stereoview of the substrate-binding cleft. The side chains of residues lining the substrate-binding cleft are shown in yellow (substrate-binding residues), cyan (conserved residues) and red (catalytic residues). The disulfide bond is shown in pink. The Tris inhibitor is depicted in dark green.

Similarities between the active sites of RBcel1 and other GH5 members permit the inference of subsites involved in the substrate binding. Hence, residues His90, Tyr92 and Asn134 are well positioned to interact with the sugar at subsite −1 (on the nonreducing end side of the scissile bond; Figs. 1 ▶ c and 2 ▶). Glu13 can interact with the glucose at subsite −1 (via a hydrogen-bonded water molecule) or at subsite −2. Phe14, which is implicated in aromatic–aromatic interactions with Trp276, Trp282 and Tyr285, can stack against substrate in subsite −2. However, molecular-dynamics simulations in the presence of cellotetraose revealed that Phe14 is too far away and Trp282 seems to be incorrectly positioned to properly orient the corresponding glucose residue. Tyr201 and Trp276 lie between subsites −1 and −2, where they are likely to be able to make contact with the corresponding sugar units. These two amino acids are also in contact in the glycosyl-enzyme intermediate. Indeed, as mentioned above, Tyr201 makes a direct interaction with the intermediate, whereas Trp276 orients and prevents it from tipping over in subsite −1. Subsites −3 and −4 are far less similar in RBcel1 and other GH5 members. At the narrower end of the cleft, analysis of regions contributing to binding subsites +1 and +2 shows that Trp171 forms the basis of both subsites, while Arg176 brings a positive charge into the vicinity of subsite +2. Some additional interactions may occur in the subsites between the substrate and the main-chain backbone of the L_β6α6 loop (in particular residues 204–208).

Interestingly, the main differences in RBcel1 in comparison with the members of GH5_5 (Ta_Cel5A and Hj_Cel5A) are observed in subsites +1 and +2, suggesting that the aglycone affinity may have significantly changed in RBcel1 (Supplementary Fig. S4). Indeed, the Trp171 indole ring is flipped by about 180° and as a result is parallel to the principal direction of the narrow cleft. Besides, to our knowledge, Arg176 has no equivalent similarly located in other GH5 members.

3.4. The Tris molecule bound to the active site  

One Tris molecule (TRS401) derived from the crystallization buffer occupies subsite −1 of RBcel1, in which it forms hydrogen bonds to several residues of the catalytic centre (Figs. 2 ▶ and 3 ▶), as frequently reported for ‘retaining’ glycoside hydrolases (Aghajari et al., 1998 ▶; Pell et al., 2004 ▶; Linden et al., 2003 ▶). The positioning of this Tris molecule is very similar to that observed in Trichoderma reesei GH5 β-mannanase (PDB entry 1qns; Sabini et al., 2000 ▶). The positively charged N atom of the Tris molecule binds to the catalytic nucleophilic and proton-donor glutamates, and the hydroxyl moieties at the C2 and C3 positions of the Tris molecule interact with Glu nuc and Asn134 and with Glu A/B and Tyr92, respectively. In addition to these hydrogen bonds, hydrophobic interactions made with the His90 and Trp276 residues contribute to the binding of the Tris molecule.

Figure 3.

Schematic representation of the active-site residues interacting with the Tris molecule. (a) LIGPLOT schematic showing residues interacting with the Tris inhibitor. Hydrogen-bond lengths are shown in green. Hydrophobic interactions with specific atoms are shown in red as depicted in the key. This figure was rendered with the program LigPlot+ (Laskowski & Swindells, 2011 ▶). (b) Close-up view of the Tris molecule.

It is well known that Tris competitively inhibits various glycosidases by structural mimicry of a reaction intermediate (Aghajari et al., 1998 ▶; Chen et al., 1987 ▶; Kersters-Hilderson et al., 1969 ▶; Myrbäck & Schilling, 1965 ▶; Brzozowski et al., 2000 ▶). This structural similarity is confirmed by superimposition of the RBcel1 structure on that of the exo-β-1,3-glucanase from Candida albicans (Exg) in complex with the transition-intermediate analogue castanospermine (Cutfield et al., 1999 ▶; Supplementary Fig. S3). Indeed, it is generally accepted that the pyranose ring is distorted and the transition intermediate(s) formed adopt(s) a half-chair or a boat conformation during the enzymic process. Cutfield and coworkers postulated that the distorted castanospermine bound in Exg results from relocation of the C2 and C1 atoms. This distorted conformation corresponds to a direct precursor to the transition state. It is therefore noteworthy that in the superimposed active sites several atoms of both inhibitor molecules, and more particularly C1 and C2 of Tris and the delocalized C1 and C2 of castanospermine, are located in similar positions (Supplementary Fig. S3).

4. Conclusion  

This is the first report of a three-dimensional structure of a family 5 glycoside hydrolase thought to be implicated in bacterial cellulose biosynthesis. Together, its overall structure, the close homology to other GH5_5 subfamily members, the conservation of the nature and position of functionally important residues and the interacting mode of the Tris molecule inside the active site strongly suggest that RBcel1 acts, as do the other GH5 family members, through a catalytic retaining mechanism involving two steps: glycosylation and deglycosylation. However, several discrepancies, notably around the aglycone-binding site, may be indicative of a change in the affinity which could result in the ability of RBcel1 to act as a transglycosylase. Undoubtedly, further investigations are required before any conclusions can be made. Nonetheless, this structure provides a starting point to better understand the involvement of GH5_5 subfamily members in bacterial cellulose synthesis.

Supplementary Material

PDB reference: RBcel1, 4ee9