G. Johansson: Department of Biochemistry, University of Uppsala, PO Box576, S-75 123 Uppsala, Sweden. Fax:. + 46 18 552139, Tel.: + 46 18 4714477, E-mail: Gunnar.Johansson@biokem.uu.se
A comprehensive experimental study of substrate inhibition in cellulose hydrolysis based on a well defined system is presented. The hydrolysis of bacterial cellulose by synergistically operating binary mixtures of cellobiohydrolase I from Trichoderma reesei and five different endoglucanases as well as their catalytic domains displays a characteristic substrate inhibition. This inhibition phenomenon is shown to require the two-domain structure of an intact cellobiohydrolase. The experimental data were in accordance with a mechanism where cellobiohydrolases previously bound to the cellulose by means of their cellulose binding domains are able to find chain ends by lateral diffusion. An increased substrate concentration at a fixed enzyme load will also increase the average diffusion distance/time needed for cellobiohydrolases to reach new chain ends created by endoglucanases, resulting in an apparent substrate inhibition of the synergistic action. The connection between the binding properties and the substrate inhibition is encouraging with respect to molecular engineering of the binding domain for optimal performance in biotechnological processes.
Cellulose, the major component of plant cell walls, is the most abundant polysaccharide in nature and a virtually inexhaustible source of renewable bioenergy. In nature, the cellulose is hydrolysed to glucose by microorganisms, mainly fungi and bacteria, that produce a set of extracellular enzymes, the cellulolytic system. A characteristic feature of most cellulolytic enzymes is a structural organization based on a catalytic domain, ‘core’, linked by a flexible region to a cellulose-binding domain (CBD) [1–3]. It is well documented that CBDs are required for efficient hydrolysis of crystalline substrates, but not of soluble ones, although their exact roles are not yet well understood . In their theoretical study, Adam and Delbrück  demonstrated that the change from three- to two-dimensional diffusion leads to a reduced mean diffusion time for the ligand to reach its target when three dimensional diffusion in space is gradually replaced by a two-dimensional diffusion on the surface. It has been proposed that binding of cellobiohydrolase I (CBH I) to crystalline cellulose is mediated by cooperative binding of both domains, giving rise to the two-dimensional diffusion of the enzyme on the cellulose surface . High cooperativity between domains was demonstrated in the binding of a double CBD construct to bacterial microcrystalline cellulose (BMCC) . At the same time, the binding of intact CBH I , as well as its separate CBD , have been demonstrated to be fully reversible. Surface diffusion rates for fluorescence-labeled bacterial cellulases from Cellulomonas fimi on crystalline Valonia cellulose have been measured by direct observation of the process in a confocal microscope . The importance of the length of the linker peptide of CBH I in binding to and hydrolysis of crystalline cellulose has also been demonstrated .
A kinetic property applicable to insoluble cellulose hydrolysis that has not yet been adequately characterized is an apparent substrate inhibition, which was reported originally in the 1970s [12,13]. The substrate inhibition depends on the cellulose substrate used. Ryu and Lee  observed substrate inhibition of the Trichoderma reesei cellulase system with hammer-milled Solka Floc SW-40 but not with ball-milled Solka Floc BW-200. Liaw and Penner  demonstrated substrate inhibition of the Trichoderma viride cellulase system on Avicel but not on Solka Floc BW-200. A systematic shift of the apparent optimum towards higher substrate concentrations in parallel with higher enzyme concentrations usedQ2 was demonstrated for the T. viride and T. reesei cellulase systems [15,16]. As all results reported to date were obtained from studies with crude cellulase mixtures and poorly defined substrates no conclusions have been drawn about the underlying mechanism.
A theoretical explanation for substrate inhibition was recently given by Fenske et al.. They used simple individual-based Monte Carlo simulation of the hydrolysis of the crystalline substrate and found that the apparent substrate inhibition can be ascribed to the loss of synergism caused by the surface dilution of enzymes occurring as the enzyme-to-substrate ratio is increasedQ3.
In this study we investigated the mechanism of substrate inhibition in cellulose hydrolysis using intact CBH I cellobiohydrolase (Cel 7A)  or its corresponding catalytic domain isolated from T. reesei, endoglucanases endoglucanase I (EG I; Cel 7B), endoglucanase II (EG II; Cel 5A), endoglucanase III (EG III; Cel 12A) from T. reesei, and endoglucanase 38 (EG 38; tentatively family 5 from Phanerochaete chrysosporium) and well defined substrates such as bacterial cellulose (BC) and bacterial microcrystalline cellulose (BMCC). The choice of a family 7 cellobiohydrolase is natural, as these quantitatively dominate in cellulose degrading organisms such as T. reesei and Phanerochaete chrysosporium. Similarly, EG I and EG II are the dominant endoglucanases in T. reesei, and presumably acting as important partners to CBH I in nature. EG 38 is expected to be one of the endoglucanases with a similar role in P. chrysosporium, whereas EG III, which lacks a cellulose binding domain, is chosen as it represents a different structural and functional type of endoglucanase. The cellulose produced by Acetobacter xylinum has recently become a substrate of choice for cellulase studies [19–22]. This bacterial cellulose has several advantages in comparison to that of plant origin: it has a more homogeneous structure, higher crystallinity and is available in a never-dried form. Hydrolysis of BC with hydrochloric acid yields a product with a reduced degree of polymerization and higher crystallinity: BMCC .
CBH I, EG I, EG II and EG III were purified from the culture filtrate of T. reesei strain QM 9414 as described previously . The EG II and CBH I core proteins were obtained by papain cleavage of the purified intact enzymes. Core proteins and CBDs were subsequently isolated by gel chromatography . EG 38 was purified from culture filtrates of P. chrysosporium. The catalytically inactive E212Q mutant of CBH I  was generous gift from J. Ståhlberg (Department of Molecular Biology, University of Uppsala, Sweden). The purity of the enzymes was confirmed by SDS/PAGE.
Preparation of cellulose substrates
Bacterial cellulose and bacterial microcrystalline cellulose were prepared from commercially available Acetobacter xylinum cellulose (‘CHAOKOH® coconut gel in syrup’ from Thep. Padung Porn Coconut Co. Ltd., Bangkok, Thailand). For this, the cubes of commercial product were rinsed with running tap water for 2 days in order to remove the sweet syrup. Rinsed cubes (1.6 g) were boiled with 1% NaOH in 1.8 L total volume under continuous stirring (1200 r.p.m.) for 2 days, replacing the hydroxide when it turned yellow. After this alkali treatment, the cellulose cubes were washed with distilled water until neutral, and then homogenized in a Waring blender. The homogenized suspension of the BC was washed several times with 0.05 m sodium acetate buffer, pH 5.0 and stored at 4 °C. BMCC was prepared from the BC as described in .
Concentrations of the enzymes
Enzyme concentrations were determined from absorbance measurements at 280 nm using the molar extinction coefficients 78 800; 67 000; 78 000; 64,000, 38,200, and 57 200 m−1·cm−1 for CBH I, EG I, EG II, EG II core protein, EG III, and EG 38, respectively.
This was determined by the anthrone/sulfuric acid method for total sugar using cellobiose as standard and absorption measurements at 585 nm .
Enzymatic hydrolysis by individual enzymes
Experiments were carried out in 1.5 mL microtubes by incubating cellulose suspensions in 0.05 m sodium acetate buffer, pH 5.0 with 1 µm enzyme at 25 °C without agitation. The reaction was initiated by addition of enzyme followed by vortex mixing for 5 s and stopped after 1 h by addition of 55 µL 1.0 m NaOH to a final pH of 12.5. The cellulose residue was pelleted by centrifugation (16 000 g, 5 min) and the total concentration of sugar in the supernatant was determined. Each data point was performed at least in triplicate and the results are expressed as mean values.
Enzymatic hydrolysis by binary mixtures
Here, the reaction was initiated by addition of a binary mixture of enzymes giving a final concentration of 1 µm for CBH I (if not indicated otherwise) and various concentrations of the second enzyme component. Hydrolysis conditions were as in the previous section.
Preparation and hydrolysis of cellulose with different amount of chain ends on the surface
As endoglucanase attack takes place at random positions on the cellulose chains in the outer layers of the crystal, the pretreatment of the BC by EG for various time results in a product with different chain end densities on the crystal surface. Intact EG 38 from P. chrysosporium was chosen due to its high efficiency in this process. BC (1.7 mg·mL−1) in 10-mL total volume was incubated with 0.025 µm EG 38 for 1min and 1 h in 0.05 m sodium acetate buffer, pH 5.0 at 25 °C without agitation. The reaction was stopped by addition of 1.1 mL 1 m NaOH followed by neutralization with 0.5 m sodium acetate buffer. An aliquot of this pretreated cellulose was used for determination of reducing ends on cellulose. For this purpose, the cellulose was washed thoroughly with 100 µm cellobiose (cellobiose was added as external standard to work in the linear concentration range of the assay) through repeated centrifugation (5min, 10 000 g) and resuspension steps. The amount of reducing ends on the cellulose was determined by a modified version of the Somogyi–Nelson method for reducing sugars [27,28] where the boiling time was extended from 20 min to 1 h . The cellulose concentration in the sample was kept at 5 mg·mL−1. All samples were centrifuged (16 000 g, 5 min) before absorbance measurements at 510 nm and calibration curves were made using cellobiose as standard. Each data point is the average of five measurements. In the case of nonpretreated BC, the addition of NaOH preceded the addition of EG 38, with the rest of the procedure being as above.
The remaining fraction of the EG 38-pretreated cellulose was washed similarly, but the cellobiose solution was replaced with 0.05 m sodium acetate buffer, pH 5.0 and used in the following hydrolysis under the conditions described above.
Pretreatment of BC with EG 38 in the presence of catalytically inactive CBH I mutant E212Q
BC (0.5 mg·mL−1) was hydrolysed with 0.025 µm EG 38 alone or in the presence of 1 µm E212Q in 0.05 m sodium acetate buffer, pH 5.0 at 25 °C. The hydrolysis reaction was stopped after the specified time by NaOH and washed with 0.05 mm sodium acetate buffer, pH 5.0, as described above. The washed cellulose (0.5 mg·mL−1) was used in the following hydrolysis with 1 µm CBH I in 0.05 m sodium acetate buffer, pH 5.0 at 25 °C.
All hydrolysis experiments additionally had zero time data points which were treated similarly to other data points with the exception that the addition of NaOH preceded the addition of enzymes. Results from the set of control experiments performed in the presence of a large excess of β-glucosidase activity indicated that product inhibition by cellobiose had no significant influence on the rate data so the β-glucosidase was not used further.
Adsorbtion of CBH I to bacterial cellulose
BC (1 mg·mL−1) was incubated with CBH I (0.5–15 µm) in 0.05 m sodium acetate buffer, pH 5.0 at 25 °C for 30 min. The cellulose was pelleted by centrifugation (5min, 16 000 g) and the concentration of free CBH I in the supernatant was determined from the activity on paranitrophenyl lactoside as described previously .
The activity of individual enzymes at different substrate concentrations was approximated to the Michaelis–Menten type saturation function:
where [S] is substrate concentration (mg·mL−1), V is maximum activity of enzyme (mm cellobiose·h−1), and K represents the substrate concentration (mg·mL−1) needed for half maximum activity under the assay conditions.
The activities of binary mixtures at different substrate concentrations were approximated to the following function:
where a, b, c, and d are arbitrary constants. The first term may in principle represent the contribution from a single enzyme action, whereas the second term represents substrate-inhibited synergistic action.
Hydrolysis of cellulose by binary mixtures of CBH I with different endoglucanases at different substrate concentration
Here, the hydrolysis of BC was carried out by 1 µm CBH I in a binary mixture with different endoglucanases (0.1 µm) at different substrate concentrations. The endoglucanases tested were EG I, EG II, EG II core protein and EG III from T. reesei and EG 38 from P. chrysosporium. EGI and II from T. reesei are the dominating endoglucanases from that organism, suggesting that the synergism observed between them and CBH I is biologically relevant. EG 38 is one of the endoglucanases expected to play a corresponding role in P. chrysosporium. The EG II core protein and EG III was included in order to reveal the effect of the CBD present in the three other endoglucanases. Substrate inhibition of hydrolysis was observed for all binary mixtures used (Fig. 1). Here, the hydrolysis by individual enzyme components followed the ‘classical’ pattern with asymptotic saturation at higher substrate concentrations, indicating that substrate inhibition is related to the synergistic action. However, the hydrolysis of bacterial microcrystalline cellulose by the binary mixture of CBH I and EG 38 does not show the inhibition by substrate observed for BC (Fig. 1, inset) stressing that the substrate inhibition also depends on the nature of the substrate. The synergistic effects between CBH and EG on BMCC are also far less pronounced than those observed on BC. The magnitude of synergism is shown in Fig. 1 by comparing the activities of binary mixture and individual 1 µm CBH I because the activity of individual 0.1 µm EG remains below the limit of the detection. The most effective binary mixtures on BC were those containing the intact EG-type enzyme. Here, the mixtures containing EG II and EG 38 were most active and indistinguishable from each other. The activities of binary mixtures containing CBD-less EG III and EG II core protein were notably weaker, but the general pattern of substrate inhibition remained the same. Regardless of the EG component used, the apparent velocity maximum is found at a BC concentration of about 0.5 mg·mL−1. However, the position of the apparent optimum substrate concentration was distinctly dependent on the concentration of CBH I as well as the EG component. Using higher CBH I concentrations at constant EG 38 concentration shifts the apparent optimum towards higher BC concentrations (Fig. 2A). A similar trend was observed when higher EG II concentrations were used at constant CBH I concentration (Fig. 2B).
No substrate inhibition was observed upon hydrolysis of BC with a binary mixture containing CBH I core domain and EG 38, indicating that the two-domain organization of the CBH component may indeed be the crucial factor for the inhibition pattern (Fig. 2A).
Figure 3 shows the influence of temperature on the substrate inhibition. When BC is hydrolysed by 1 µm CBH I in the binary mixture with 0.01 µm EG II at 25 °C, inhibition by substrate clearly occurs. Raising the temperature to 50 °C not only increases the overall activity of the binary mixture but also virtually eliminates the substrate inhibition within the observed range of BC concentration.
Preparation and hydrolysis of bacterial celluloses with different chain end densities on the surface
For this purpose the BC was pretreated with EG 38 for selected times. Upon the removal of EG and thorough washing the number of additional reducing ends generated on the cellulose was determined (see Experimental procedures). We measured the initial amount of chain ends on BC to be 4.8 ± 0.4 µmol·g−1. The additional amount of chain ends generated (EG generated additional chain ends can not be located at the end of the crystal and are so called ‘internal’ chain ends) was measured to be 2.0 ± 0.4 and 6.0 ± 0.2 µmol·(g cellulose)−1 after 1 min and 1 h endoglucanase treatment, respectively. This results in at most a threefold difference in density of chain ends on the surface, or probably even less, considering that intact BC also contains some amount of ‘internal’ chain ends. The pretreated cellulose samples were subjected to hydrolysis with 1 µm CBH I at different substrate concentrations. The data in Fig. 4 demonstrate that the hydrolysis of the different pretreated celluloses reaches a plateau at different levels. The maximum activities of CBH I were calculated from a Michaelis–Menten type equation (Eqn 1) to be 0.20 ± 0.01 (mm cellobiose·h−1) and 0.27 ± 0.01 (mm cellobiose·h−1) preparations pretreated for 1 min and 1 h, respectively. The arrows in Fig. 4 connect the positions where the absolute numbers of newly created chain ends on both pretreated substrates are equal, assuming the threefold difference discussed above.
Pretreatment of bacterial cellulose with endoglucanase in the presence of catalytically inactive CBH I mutant
Here, the BC (0.5 mg·mL−1) was pretreated with 0.025 µm EG 38 alone and in the presence of 1 µm catalytically inactive CBH I E212Q . Pretreated cellulose samples were subjected to thorough washing and the effect of pretreatment was monitored by the activity of CBH I on pretreated substrates. The data in Fig. 5 demonstrate that there was no difference between pretreatment of BC with EG 38 alone or in the presence of E212Q. Pretreatment of BC with 1 µm E212Q alone did not have any influence on the following hydrolysis by CBH I (data not shown).
The CBH I is generally regarded as a classical exoenzyme, meaning that its activity on cellulose depends on the number of available chain ends . Due to the high degree of polymerization, the number of free chain ends in BC is initially very small. We measured the total amount of reducing ends on our BC sample to be 4.8 ± 0.4 µmol·g−1 and the maximum adsorption capacity for CBH I to be 6.8 µmol·g−1. For steric reasons, it is very unlikely that each chain end may carry an enzyme molecule  the adsorption capacity must obviously be ascribed to the two-domain structure of cellulases, which causes most of the CBH I molecules to be anchored via the CBD in a nonproductive mode without access to chain ends. When new/additional chain ends are created by a randomly acting EG component they may be captured by available CBH I molecules to start processive action at additional sites along the cellulose chains. The result is a strong synergism between the enzymes. Cellulases have been proposed to be capable of lateral diffusion on the cellulose surface, and this phenomenon is well documented for Cellulomonas fimi exoglucanase Cex and endoglucanase CenA on a crystalline cellulose surface . Lateral diffusion gives a nonproductively bound enzyme a ‘standby’ status, because it may migrate to find a free chain end and become productive without dissociating from the cellulose surface. The probability, and thus average time needed for this migration is, however, dependent on the average distance between available chain ends and nonproductively bound cellobiohydrolases. This gives the rationale for substrate inhibition as proposed by Fenske et al..
At the concentrations of enzyme and cellulose studied, the surface coverage correlates strongly with the enzyme/substrate ratio. A high ratio will thus give a similarly high surface coverage, leading to a reduction of the average distance between ‘stand by’ cellobiohydrolases and available chain ends created by endoglucanases. The observed decrease in activity observed at higher substrate concentrations is in agreement with such a model, where the increased availability of substrate does not outweigh the negative effect on the rate caused by the increase in average time for the cellobiohydrolases to reach the chain ends.
The hypothesis above can be tested if one allows the cellobiohydrolase only to act on endoglucanase-pretreated cellulose samples with different chain end densities on the surface. If the activity of CBH I depended only on the absolute number of chain ends on the cellulose surface, but not on their density, then the lower activity of CBH I on BC pretreated for 1 min compared to that on BC pretreated for 1 h, should be able to compensate by adjusting the a cellulose concentration to increase the bulk concentration of chain ends. However, as the two upper curves in Fig. 4 reach the plateau at different levels, a lower surface density of chain ends cannot be compensated for by a higher absolute number of chain ends achieved by increasing the total substrate concentration. Therefore, the apparent substrate inhibition can be accounted for simply by the increased average distance (i.e. lateral diffusion time) between stand-by CBH I molecules and available chain ends due to surface dilution. The same situation arises in the hydrolysis of BC by endoglucanase/cellobiohydrolase mixtures at lower surface coverages. A higher endoglucanase content in the binary mixtures is expected to provide same surface density of chain ends at higher cellulose concentrations and does indeed give a shift of ‘optimum substrate concentration’ towards higher BC concentrations (Fig. 2B), in accordance with the model. Similarly, higher CBH I concentrations used, which also should maintain a certain surface coverage at higher BC concentrations, shift the maximum towards higher BC concentration (Fig. 2A). Actually, the total picture suggests that the activity maximum occur at a rather constant, characteristic cellulase/cellulose ratio.
The characteristic substrate inhibition discussed above was not observed when the intact CBH I was replaced by CBH I core, i.e. the catalytic unit without a CBD in the binary mixture with EG 38, stressing that the two-domain nature of the cellobiohydrolase, which allows standby adsorption to the cellulose, is crucial (Fig. 2A). As pointed out by Adam and Delbrück , the reduction in dimensionality of diffusion, i.e. going from three-dimensional diffusion in solution to two-dimensional (lateral) diffusion on the cellulose microparticles leads to reduced mean diffusion times and gives an advantage for such enzymes that concentrate on the polymeric substrate surface if the binding mechanism allows lateral diffusion. The loss of the substrate inhibition pattern observed at higher temperatures suggests that the lateral diffusion rate is enhanced by the temperature increase to such an extent that it no longer is a prominent rate limiting factor. It is also possible that higher temperature allows a larger relative contribution from degradation modes based on three-dimensional diffusion.
However, one can speculate that the activity of the EG component is somehow enhanced, resulting in a higher number of newly generated chain ends, at a higher total surface coverage. We controlled the possible influence of total surface coverage on EG activity by pretreatment of BC with EG 38 alone and in the presence of 1 µm CBH I E212Q, respectively. The E212Q mutant is catalytically inactive but binds still to cellulose . However, as no influence of E212Q on the pretreatment of BC by EG 38 was observed (Fig. 5), the EG action was not facilitated by a higher surface coverage.
It has also been reported that substrate inhibition does not occur with on all cellulosic substrates . Hydrolysis of BMCC with a binary mixture of CBH I and EG 38 does, indeed, display a saturating character (Fig. 1, inset), probably due to the restricted activity of EG on highly crystalline cellulose, where the amorphous parts, which are supposed to serve as important targets for endoglucanases in the synergistic action, are removed by acid hydrolysis employed for BMCC preparation. Also, pretreatment of BMCC with EG had no effect on the subsequent hydrolysis by CBH I, indicating the different modes of synergistic action on BC and BMCC-like substrates .
It is important to note, however, that substrate-inhibited synergistic action probably always exists in parallel with an action which obeys normal saturation kinetics. The presence/absence of a pronounced maximum at a particular substrate concentration should thus depend on the relative contributions of the different modes of action. In conclusion, the adaptation of data to Eqn (2) should reveal contributions from substrate-inhibited (synergistic) action when no velocity maximum is observed.Q6 Here, it can be assumed that the ratio between the constants c and a in Eqn 2. should correlate with the relative contribution from the synergistic mode of action.
We thank Anu Nutt for purification of the enzymes and Dr David Eaker for linguistic revision. This work was supported by grants from Royal Swedish Academy of Sciences and Wood Ultrastructure Research Centre (NUTEK).