Reassembly and co-crystallization of a family 9 processive endoglucanase from its component parts: structural and functional significance of the intermodular linker

Cloning of the GH9, CBM3cL and CBM3cNL proteins

Cloning of the DNA fragments encoding the C-terminally His-tagged CBM3c with the linker and the untagged GH9 module from Cel9I of C. thermocellum (GenBank accession code L04735) was described earlier (Burstein et al., 2009; Gilad et al., 2003). C-terminally His-tagged CBM3c without the linker connecting it to the GH9 was amplified using the same procedure and the following primers: F′-5′ CCATGGGCGAAGTTCCGGAGGATGAAATA and R′-5′ CTCGAGCGGTTCCCTTCCAAATACCAG. The PCR products were purified and cleaved with restriction enzymes NcoI and XhoI and inserted into the pET-28a(+) expression vector (Novagen, Madison, WI, USA).

Expression and purification of recombinant proteins

The GH9 and CBM3c modules both with (GH9L, CBM3cL) and without (GH9NL and CBM3cNL) the linker regions were expressed independently by the identical expression procedure. Escherichia coli strain BL21(DE3)RIL harboring the plasmids was aerated at 310 K in 3-liters Terrific Broth supplemented with 25 mg ml⁻¹ kanamycin. After 3 h, the culture reached an A₆₀₀ of 0.6; 0.1 mM isopropyl-β-D-1-thiogalactopyranoside was added to induce gene expression, and cultivation was continued at 310 K for an additional 12 h. Cells were harvested by centrifugation (5,000 × g for 15 min) at 277 K and were subsequently re-suspended in 50 mM NaH₂PO₄, pH 8.0, containing 300 mM NaCl at a ratio of 1 g wet pellet to 4 ml buffer solution. A few micrograms of DNase powder were added prior to the sonication procedure. The suspension was kept on ice during sonication, after which it was centrifuged (20,000 × g at 277 K for 20 min), and the supernatant was collected.

The soluble expressed His-tagged CBM3c modules with or without the linker, according to the type of the experiment, were applied batchwise to Ni-IDA resin during 1-h incubation with gentle stirring at 4 °C. Non-specifically bound proteins were washed with a buffer containing 50 mM NaH₂PO₄ pH 6, 300 mM NaCl, 10% glycerol and 10 mM imidazole. Crude extract supernatant fluids, containing the expressed GH9 module, were added to the CBM3c-bound Ni-IDA resin, and the mixture was incubated overnight with gentle stirring at 4 °C. The adsorbed protein complexes were eluted with 300 mM imidazole and subjected to further purification by size-exclusion chromatography. Fast protein liquid chromatography (FPLC) was performed using a Superdex 75pg column and ÄKTA Prime system (GE Healthcare, Piscataway, New Jersey, USA) to further purify the complex. One peak, corresponding approximately to 70 kDa, matching the predicted molecular weight of the GH9-CBM3c complex, was observed in the chromatogram. The 15 amino-acid linker sequence (about 1.5 kDa) did not significantly affect the elution volume, compared to that of the complex without the linker, presumably due to the limited resolution of the column. The relevant fractions (the purified complexed proteins) were analyzed by 15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) with Coomassie brilliant blue staining. Two clear bands, of about 52 and 19.5 kDa were observed. The rearranged modules were concentrated to 6 mg ml⁻¹ using Centriprep YM-3 centrifugal filter devices YM-3 (Amicon Bioseparation, Millipore Corporation, Bedford, Massachusetts, USA). Protein concentration was determined by measuring UV absorbance at 280 nm.

The full-length Cel9I was purified by affinity chromatography on Avicel as reported earlier (Burstein et al., 2009; Gilad et al., 2003).

Microcalorimetric analysis

Isothermal titration calorimetry (ITC) experiments were carried out using a VP-ITC MicroCalorimeter (MicroCal, LLC, Northampton, Massachusetts, USA) at 298 K. About 300 µM solution of CBM3cNL was injected into a 65 µM solution of GH9. The reaction was performed in a buffer containing 50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.05% sodium azide. Heats of dilution of the titrants were subtracted from the titration data, and the corrected data were analyzed using the Origin ITC analysis software package supplied by MicroCal. Thermal titration data were fit to the one binding site model, and enthalpy (ΔH), entropy (ΔS), association constant (Ka) and stoichiometry of binding (N) were determined. In all cases, the calculated stoichiometry (N) was lower than one, most likely due to the fact that the CBM3 proteins lost their native functionality with time. For the analysis, the CBM3 protein concentrations were corrected as to provide a stoichiometry of one. Two titrations were performed to evaluate reproducibility.

Enzyme activity assay

Reactions were performed at 333 K, in 50 mM citrate buffer (pH 6.0). The soluble cellulolytic substrate was carboxymethyl cellulose (CMC, Sigma Chem. Co. St. Louis, Missouri, USA). The amount of reducing sugars released from the substrate was determined with the 3,5-dinitrosalicylic acid (DNS) reagent as described by Miller et al. (Miller, 1959). Activity was defined as the amount (micromole) of reducing sugar released after 10 min of reaction.

Crystallization

Initially the protein samples containing 6 mg/ml protein solution in 1.2 mM Tris–HCl pH 7.5, 1.5 mM sodium chloride, 0.025% sodium azide, were screened, using the microbatch crystallization method under 1:1 mixture of silicon and paraffin oil (Chayen et al., 1990), using 288 conditions from the Hampton Research HT screens (SaltRx, Index HT, and Crystal Screen HT; Hampton Research, Aliso Viejo, California, USA) and 96 conditions of the Wizard Crystallization kit from Emerald BioSystems (Rigaku Reagents, Bainbridge Island, Washington, USA). The dyad of GH9 and CBM3cNL did not yield any crystals. Screening of the GH9-CBM3cL resulted in plate-like crystals that appeared after several days under several conditions, all of which contained PEG 3350 and 4000. The best crystals were obtained in 30% PEG (both 3350 and 4000), 0.2 M magnesium chloride, and 0.1 M Hepes, pH 7.5. Attempts to optimize this condition using microbatch, hanging-drop, and sitting drop methods were unsuccessful, as the crystals remained very thin and fragile. The superfine Eyelash (Ted Pella, Inc, Redding, California, USA) was used to touch these crystals and consequently to streak the sitting drops, composed of 5 µl of the protein solution and 5 µl of the precipitating solution (24% PEG 3350, 0.2 M magnesium chloride, 0.1 M Hepes, pH 8.0). After one day, crystals of different morphology, with maximum size of about 0.05 mm, appeared in the drop.

Data collection and crystallographic analysis

The crystals of the GH9-CBM3cL complex were harvested from the crystallization drop using a nylon cryo-loop (Hampton Research, Aliso Viejo, California, USA). For data collection, crystals were mounted on the MiTeGen stiff micro-mount (MiTeGen, Ithaca, New York, USA) made of polyimide and flash-cooled in a nitrogen stream produced by Oxford Cryostream low temperature generator (Cosier & Glazer, 1986) at a temperature of 100 K. Mother-liquor of the crystals served for cryo-protection during the cooling in liquid nitrogen.

Diffraction data from the GH9-CBM3cL crystals were measured using the ID23-2 beam line at ESRF, Grenoble, France. A MAR CCD 225 area detector and X-ray radiation of 0.873 Å wavelength were used. Diffraction data of 480 images with 0.5°oscillation per image were collected. Data were processed with DENZO and scaled with SCALEPACK as implemented in HKL2000 (Otwinowski & Minor, 1997). The crystals diffracted to 1.68 Å resolution and belong to the orthorhombic space group P2₁2₁2₁, with unit cell parameters a = 70.4, b = 88.5, c = 106.5 Å. There is one GH9-CBM3cL complex per asymmetric unit with a Matthews density V_M of 2.37 ų Da⁻¹, corresponding to a solvent content of 48.15% (Matthews, 1968). The X-ray data analysis statistics are presented in Table 1 (Stout & Jensen, 1968).

Molecular replacement was carried out with MOLREP (Vagin & Teplyakov, 1997), using the coordinates of the GH9 and CBM3c modules of endoglucanase 9G from Clostridium cellulolyticum (PDB code 1G87, 66 and 51% sequence identity, respectively), as a search model. The MOLREP calculations with the GH9 domain converged into a clear solution with 1 molecule in the asymmetric unit with an R-factor of 0.533 and correlation coefficient of 0.567. This solution was inserted into MOLREP calculations as a fixed molecule and the coordinates of CBM3c module were used for the search producing a solution with an R_cryst of 0.505, and correlation coefficient of 0.582. The resulting model with 5% of reflections forming test set (Brünger, 1992) was subjected to 10 cycles of restrained refinement using anisotropic B-factors, yielding the R_cryst and R_free 0.329 and 0.359, respectively (REFMAC5) (Murshudov, Vagin & Dodson, 1997). Automated model building by ARP/wARP (Perrakis, Morris & Lamzin, 1999) produced a complete structure with R_cryst and R_free of 0.218 and 0.243, respectively. The model was manually corrected using COOT (Emsley & Cowtan, 2004) and refined using REFMAC5 (Murshudov, Vagin & Dodson, 1997). The R_cryst and R_free improved to 0.184 and 0.228, respectively. Solvent atoms were built using ARP/warp (Perrakis, Morris & Lamzin, 1999). Refinement of TLS (rigid body translation/libration/screw motions) parameters was performed (Winn, Isupov & Murshudov, 2001; Winn, Murshudov & Papiz, 2003). The model was subjected to several additional cycles of manual rebuilding and refinement. The model converged to final R_cryst and R_free factors of 0.144 and 0.176, respectively.

The refinement statistics of the structure are summarized in Table 2. The structure was validated using MolProbity (Davis et al., 2007).

Protein sequence analysis

Sequence alignments were performed using CLUSTALW (Larkin et al., 2007) and the coloring of residues (representing degree of conservation) using ProtSkin (Deprez et al., 2005). Sources of the sequences used in this work are as follows: Clostridium thermocellum Cel9I GH9 module, CBM3c and CBM3b (AAA20892.1); Clostridium cellulolyticum Cel9G GH9 module, CBM3c (AAA73868.1); Thermobifida fusca Cel9A GH9 module and CBM3c (AAB42155.1); Cellulomonas fimi Ce9A CBM3c (AAA23086.1); Clostridium cellulovorans EngH CBM3c (AAC38572.2) and CbpA CBM3a (AAA23218.1); Clostridium stercorarium CelZ CBM3c and CBM3b (CAA39010.1) and CelY CBM3b (CAA93280.1); Clostridium thermocellum CipA CBM3a (CAA48312.1), CelQ CBM3c (BAB33148.1), Cel9V CBM3c’ and CBM3b’ (CAK22315.1), Cel9U CBM3c’ and CBM3b’ (CAK22317.1) and Cbh9A CBM3b (CAA56918.1); Clostridium cellulolyticum CipC CBM3a (AAC28899.2) and CelJ CBM3c (AAG45158.1); Acetivibrio cellulolyticus Cel9B CBM3c’ and CBM3b’ (CAI94607.1) and CipV (ScaA) CBM3b (AAF06064.1); Clostridium josui CipA (CipJ) CBM3a (BAA32429.1); Bacteroides cellulosolvens ScaA CBM3b (AAG01230.2); Bacillus subtilis CelA CBM3b (AAA22307.1); Pectobacterium atrosepticum CelVI CBM3b (X79241.2); Bacillus licheniformis CelA CBM3b (CAJ70714.1).

Cloning, expression and purification of Cel9I and its modular components

The full-length C. thermocellum Cel9I enzyme and its individual component parts were over-expressed in Escherichia coli, according to Burstein et al. (2009), in order to investigate the contribution of the ancillary modules and their linkers to the catalytic activity of the enzyme. These include the isolated GH9 module with and without a His tag, the His-tagged CBM3c module together with its adjacent N-terminal linker that connects it to the GH9 module (CBM3cL) and His-tagged CBM3c module without the N-terminal linker (CBM3cNL). For details, see Fig. 1. Following purification procedures, all recombinant proteins showed a single band in SDS-PAGE of the anticipated molecular masses.

Recovery of endoglucanase activity upon association of CBM3cNL and GH9 compared to CBM3cL and GH9

Previous works (Burstein et al., 2009; Gilad et al., 2003) demonstrated that the Cel9I catalytic module alone has no detectable activity on CMC (carboxymethyl cellulose) and that adding the CBM3cL to form the Cel9I-CBM3cL dyad serves to recover up to 70% of the lost activity. To further examine the importance of the linker connecting the GH9 and the CBM3c modules, we tested the ability of CBM3cNL to recover the CMCase activity of GH9. A fixed amount of the catalytic module (70 pmol in 400 µl) was mixed with increasing amounts of CBM3cL or CBM3cNL. The activity of the intact Cel9I enzyme was defined as 100%, and the activity of the reconstituted complexes was measured relative to that of Cel9I. The results indicated that GH9-CBM3cNL exhibit only about 10% of the intact Cel9I activity towards CMC, whereas the reassembled GH9-CBM3cL provided up to 50% of the activity (Fig. 2). The fact that a higher than one molar ratio was required to obtain maximum activity can be explained by the fact that the CBM protein was only partly functional as was also observed in the ITC experiments described below. Overall, the results suggest that the linker is required for better fitting of the reconstituted CBM3c which results in better recovered activity.

Overall structure of the reassembled GH9-CBM3c

Initially, we tried to overexpress and purify the covalently linked GH9-CBM3c, however the full-length protein was unstable and proteolysis occurred during the overexpression and purification stages, partially resulting in separate GH9 and CBM3c modules. Therefore, the obtained protein samples were not homogenous and were not suitable for crystallization trials. Instead, we employed an alternative approach where we expressed the two domains separately and combined them in vitro. Surprisingly, the combined modules crystallized and formed a structure similar to those of the known GH9 cellulases.

The crystal structure of the reassembled C. thermocellum Cel9I GH9-CBM3cL dyad was determined by molecular replacement and the coordinates are deposited in Protein Data Bank with code 2XFG. Data collection and refinement statistics are given in Tables 1 and 2. The catalytic GH9 and the ancillary CBM3c modules reassembled in vitro to form a dyad (Fig. 3A) similar in structure to the intact tandem GH9-CBM3c modules of the orthologous endoglucanases: Cel9G from C. cellulolyticum (1G87) and Cel9A (previously termed cellulase E4) from Thermobifida fusca (1TF4), with an RMS deviation of 0.783 Å over 468 Cα atoms with Cel9G and 0.757 Å with Cel9A (Fig. 3B).

Structure of the GH9 module

The catalytic module of the Cel9I enzyme consists of residues 1–446, comprising 15 α-helices, whereby the twelve longest ones form the (α/α)₆-barrel (Fig. 4A). The hydrophobic core of the GH9 module is formed by 118 hydrophobic and aromatic amino acids, the vast majority of which are also conserved in the GH9 modules from C. cellulolyticum Cel9G and T. fusca Cel9A. Hydrophobic and aromatic cores have been proposed to play an important role in the formation of (α/α)₆-barrels (Mandelman et al., 2003). The GH9 module of Cel9I thus shows high structural similarity with the two latter GH9 structures: C. cellulolyticum Cel9G (0.367 Å RMS deviation over 349 C-alpha atoms) and T. fusca Cel9A (0.532 Å RMS deviation over 359 C-alpha atoms).

The catalytic site of the GH9 module is located at the depression in the flat surface, formed by the loops connecting the N termini of the barrel helices (Fig. 4B). The flat face is rich in charged and polar residues (Fig. 4B), highly conserved also in Cel9G (1G87) and Cel9A (1TF4). The GH9 modules of these cellulases (Mandelman et al., 2003; Sakon et al., 1997; Zhou et al., 2004) exhibit similar flat faces and clefts, and these conserved residues (His 126, Trp 129, Phe 205, Tyr 206, Trp 209, Trp 256, Asp 261, Asp 262, Trp 314, Arg 318, His 376, Arg 378, and Tyr 419) have been shown to bind natural and synthetic oligosaccharides (Fig. 4C). In the present structure, as in the other known GH9-CBM3c bimodular structures, one end of this cleft is blocked by a loop formed by residues 243–254 and the other end is fused with the flat surface of the CBM3c module (Fig. 4B). Details of the catalytic cleft are presented in Fig. 4C.

One calcium ion is found near the catalytic cleft of the GH9 module of Cel9I and is coordinated by a Ser 210 (OG) 2.6 Å, Gly 211 (O) 2.4 Å, Asp 261 (O) 2.4 Å, Asp 214 bifurcated (OD1, OD2) 2.5 Å, and Glu 215 bifurcated (OD1, OD2) 2.5 Å (Fig. 4D). Despite some minor changes in the residues of coordination this ion seems to be structurally equivalent to those of T. fusca Cel9A (RMS deviation 0.160 Å over 5 Cα atoms of the coordinating residues), and C. cellulolyticum Cel9G (RMS deviation 0.503 Å over 4 Cα atoms). In all three cases the calcium ion draws together the N-terminal ends of α-helixes 8 and 10.

Structure of the CBM3c module

The CBM3c module consists of 150 amino acids arranged in an eight β-stranded sandwich motif homologous to other known family 3 CBM structures (Gilbert, Knox & Boraston, 2013; Mandelman et al., 2003; Petkun et al., 2010b; Sakon et al., 1997; Shimon et al., 2000; Tormo et al., 1996; Yaniv et al., 2014; Yaniv et al., 2012b; Yaniv et al., 2011). The “lower” face of the sandwich is formed by β-strands 1, 2, and 7; the “upper” face is formed by β-strands 3, 3′, 6, 8, and 9 (Fig. 5A). The structure of Cel9I CBM3c is particularly similar to the structures of the other two previously described CBM3c structures (RMS deviation 0.734 Å over 116 C-alpha atoms with CBM3c from C. cellulolyticum Cel9G; RMS deviation 0.829 Å over 113 atoms with CBM3c from T. fusca Cel9A). Only 31% of amino acids are located in β-strands of the CBM3c module from Cel9I; others are found in the loop regions.

One calcium ion was found in the upper β-sheet of the CBM3c molecule (Fig. 5B) and is coordinated by a water molecule and five residues from the upper β-sheet: Asn 500 (O), Glu 503 bifurcated (OE1, OE2), Asn 573 (O), Asn 576 (OD1), Asp 577 (OD1). This calcium atom is in a similar location as in Cel9A and Cel9G, and probably plays a structural role for most CBM3 modules, as was suggested previously (Tormo et al., 1996).

The lower sheet forms a flat platform conserved between the CBM3c modules and the other two molecular structures. This flat surface is rich in charged and polar conserved surface residues: Asn 466, Glu 474, Lys 476, Ser 518, Tyr 520, Glu 559, Gln 561, and Arg 563 (Fig. 5C). The planar region of the CBM3c modules in all three enzymes is particularly aligned in continuation of the catalytic cleft of the catalytic modules, and has been proposed to bind single chains of cellulose and guide them to the cleft (Mandelman et al., 2003; Sakon et al., 1997).

The CBM3c possesses a very interesting surface structure, formed by the β-strands on the opposite side of the flat surface, called the “shallow groove” (Shimon et al., 2000; Tormo et al., 1996). The “shallow groove” is lined by four aromatic rings (Phe 498, Tyr 538, Tyr 578 and Tyr 597), two charged or polar residues (Arg 496, and Glu 540), Leu 602, Pro 595 and Pro 608. These residues are also conserved in other CBM3 modules regardless of their subgroup relation (a, b, or c), their cellulose-binding ability and their effect on the activity of the catalytic module. Figure 5D shows the shallow groove of the CBM3c module from the Cel9I enzyme colored according to the extent of the conservation of the residues in other CBM3a, b and c modules (darker blue represents more conservation). The alignment was performed over 25 CBM3 sequences (11 CBM3c, 12 CBM3b and CBM3b’, and 4 CBM3a). Conservation of this surface structure, regardless of the particular known function of the CBMs, implies that this site has some kind of “generic” function. This shallow groove may serve to bind to single oligosaccharide chains or to peptide chains, such as the intermodular linkers common to cellulases or cellulosomal scaffoldin subunits. There is evidence that the shallow groove interacts with a linker region (Petkun et al., 2010a; Shimon et al., 2000; Yaniv et al., 2012a).

Contact residues between the GH9, linker and CBM3c

The in vitro reassembled GH9-CBM3cL complex has a large intermodular interface, the contact area of which is 1,108.3 Ų, corresponding to 12.3% of the total surface-exposed area of the CBM3c module and 6.2% of the exposed GH9 module (Krissinel & Henrick, 2007). The GH9 and the CBM3cL modules of Cel9I are assembled into the reconstituted GH9-CBM3c complex by 31 hydrogen bonds (4 main chain-main chain, 19 main chain-side chain, and 8 side chain-side chain), 14 hydrophobic, 3 aromatic interactions, and 3 ionic bonds (http://pic.mbu.iisc.ernet.in/index.html) (Tina, Bhadra & Srinivasan, 2007). Sixteen residues from the GH9 module and seventeen residues from the CBM3c participate in these interactions (contact residues are shown in Fig. 6A). The vast majority of the contact residues and contacts are similar to those of C. cellulolyticum Cel9G and of T. fusca Cel9A (Fig. 6B).

Conserved residues of the linker (which spans from Gly447 to Asp462) make numerous contacts with conserved residues of the GH9 module, emphasizing the importance of the linker in this interaction (Fig. 6). A conserved Gly447 of the linker interacts via hydrogen bonds with Gly444 and Tyr440 of the GH9 module. Pro449 forms hydrophobic interactions with Tyr440, and Asp450 forms hydrogen bonds with Gln11. Another linker residue, Phe453, forms hydrophobic and aromatic interactions with two aromatic residues of the GH9 module, i.e., Phe15 and Tyr399. Gly455 forms hydrogen bonds with Asn33 and Arg31. Glu457 forms intricate interactions with a variety of residues of the GH9 module, which include hydrogen bonding with Asn33, Val397, Arg395, as well as hydrogen and ionic interactions with His396. Glu461 exhibits hydrogen and ionic interactions with Arg395. Additionally, linker residues Glu457 and Glu461 form hydrogen bonds and salt bridges with Trp490 and Lys487, respectively. The latter belong to the CBM3c module and are part of a loop (486–498), which protrudes towards and forms interactions with several amino acids of the GH9 module. There are also many hydrogen bonds formed between the neighboring amino acids of the linker, thus contributing to its defined conformation. Altogether the multiple, well-conserved interactions between the linker, the GH9 and the CBM3c modules stabilize the spatial orientation of the modules towards one another and contribute to the structural rigidity of the entire molecule, resulting in an active enzyme structure.

As mentioned above, the mutual spatial orientation of the GH9 and CBM3c modules is very similar to that in the native, intact bimodular pairs from Cel9G and Cel9A leading to the overall similarity in structures. The remarkable conservation of the overall architecture in the reassembled in vitro complex together with the striking conservation of the contact residues implies its high functional importance. In all of these structures (Cel9G, Cel9A, and the reassembled GH9-CBM3cL from Cel9I), the flat surface of the CBM3c module is aligned in continuation with the catalytic cleft of the GH9 module, making an extended platform (Fig. 4B). This platform is rich in charged and polar surface residues that are highly conserved throughout the family 3 CBMc’s.

Microcalorimetric analysis of the GH9-CBM3c complex formation

The binding constants of GH9 and the CBM3c were obtained by performing isothermal titration calorimetry (ITC) experiments in which a solution of GH9 was titrated with a solution of CBM3c with or without the linker (Fig. 7). Control experiments for each of the components alone were conducted and subtracted from the titration data. In both cases the titration curve could be fitted to a one-site binding model although the calculated stoichiometry was less than one. The low stoichiometry is probably a result of the fact that the soluble CBM module lost its functionality with time and its true active concentration was less than the measured protein concentration. To estimate the binding constants for the two CBM3c forms, the CBM3c concentrations were corrected to provide a stoichiometry of one. In all cases, the binding reactions were enthalpy driven with a negative entropy contribution. CBM3cL provided binding constants (K_d) between 1.3–2.0 × 10⁻⁶ M, whereas CBM3cNL exhibited stronger binding constants, K_d between 2.9–4.3 × 10⁻⁷ M. Thus, the linker may serve as a mitigating factor for the binding process, ensuring specific binding orientation. This is consistent with the structural data and the activity assays, which emphasizes the important role of the linker in enzyme functioning. In the case of CBM3cNL, the binding process may occur faster in the absence of linker, but may also lead to unspecific binding and aggregation of the modules.