Inter-domain synergism is required for efficient feeding of cellulose chain into active site of cellobiohydrolase Cel7A

Structural polysaccharides like cellulose and chitin are abundant and their enzymatic degradation to soluble sugars is an important route in green chemistry. Processive glycoside hydrolases (GHs), like cellobiohydrolase key peptide. The obscured The Trp-38 to Ala substitution resulted in increased off-rates and decreased processivity. The effect of the Trp-38 to Ala substitution on on-rates was strongly dependent on the presence of the CBM-linker. This compensation between CBM-linker and Trp-38 indicates synergism between CBM-linker and CD in feeding the cellulose chain into the active site. The inter-domain synergism was pre-requisite for the efficient degradation of cellulose in the presence of endoglucanase.


Introduction
Structural polysaccharides represent a vast resource available for clean energy and chemistry (1,2). A lot of effort has been put into developing the industrial approaches for optimal deconstruction and valorization of structural polysaccharides including cellulose (the homopolymer of β-1,4-linked glucose units). The highly recalcitrant and insoluble crystalline matrix of cellulose is challenging for enzymatic degradation (2). In nature the most efficient degraders of cellulose are fungi, which secrete a myriad of cellulases for efficient decomposition of cellulose. For biotechnological applications the best known fungal isolate is the filamentous fungus Trichoderma reesei (teleomorph Hypocrea jecorina) (3). The key component of T. reesei's enzyme system is a glycoside hydrolase (GH) (4) family 7 enzyme TrCel7A a processive cellobiohydrolase (CBH) acting from the reducing end of the cellulose chain (5-7). Most T. reesei cellulases including TrCel7A have a two-domain structure consisting of a non-catalytic cellulose binding module (CBM) connected by a flexible glycosylated linker peptide to the catalytic domain (CD) that is responsible for catalysis of the glycosidic bond hydrolysis (3,8,9) ( Figure  1A). The CBM has been shown to have a role in targeting (10)(11)(12), but also in disrupting the crystalline cellulose aggregates (13). A role of the linker peptide in activity and binding has also been recognized (9,(14)(15)(16). The CBM of TrCel7A belongs to the CBM family 1 (8), a family of CBMs that primarily targets crystalline cellulose (17,18). In line with this, the removal of the CBM has a more negative effect on the hydrolysis of crystalline cellulose compared to amorphous substrates (19). The CBM is beneficial for activity at low substrate concentrations, whereas at high substrate loads it is less important or even a disadvantage (20)(21)(22). The catalytic domain of TrCel7A contains a 5 nm long tunnel shaped active site (23,24). The tunnel is roofed by four surface loops and accommodates 10 glucose unit binding sites (-7 to +3), 7 substrate (denoted withsign) and 3 product (denoted with + sign) binding sites (23,24) (Figure 1B). Characteristic for GHs, the active site of TrCel7A contains several aromatic residues that stack with sugar rings through π-interactions. Two of these, Trp-40 and Trp-38, locate at the entrance (-7) and in the middle (-4) of the tunnel, respectively. Another two, Trp-367 and Trp-376 are located near the scissile bond in binding sites -2 and +1, respectively ( Figure 1B). The negative effects of single aromatic substitutions on hydrolysis of crystalline substrates have been demonstrated for TrCel7A (7,22,25) as well as for bacterial cellulases (26)(27)(28) and chitinases (29)(30)(31). On the other hand the deficiency of Trp to Ala variants in hydrolysis of crystalline substrates has often accompanied by the increased activity on amorphous and soluble substrates (29,31). The hydrolysis of recalcitrant polysaccharides by processive enzymes is a complex process consisting of at least three steps (i) adsorption, recognition and initial sliding of cellulose chain into the active site (3,32); in this work we will collectively refer to this as feeding, (ii) processive sliding with concomitant hydrolysis of polymer chain, and (iii) dissociation ( Figure 1C). Computational studies have suggested that Trp-40 and Trp-38 of TrCel7A participate in the substrate binding whereas Trp-367 and Trp-376 are directly involved in catalysis of glycosidic bond hydrolysis by stabilizing the transition state (33,34). Substitution of  to Ala at the tunnel entrance (W40A) reduces the hydrolysis rates of highly crystalline cellulose by TrCel7A apparently through impairing the cellulose chain acquisition (7,25). Substitution of Trp-38 to Ala (W38A) in TrCel7A has been shown to decrease the affinity of the variant to crystalline cellulose, Avicel, indicating the importance of Trp-38 in binding (22). In this study it was concluded that the decreased affinity was primarily governed by the increased off-rate constant (koff) rather than decreased on-rates (22). Increased koff of W38A variant compared to the wild type (WT) enzyme was also proposed to be responsible for its higher steady-state hydrolysis rates at saturating substrate concentrations (22). This conclusion stems from the obstacle model of cellulose hydrolysis whereby the steady state rate of hydrolysis by CBH acting in isolation is governed by the koff and apparent processivity (P app ) of CBH (19,22,(35)(36)(37)(38)(39). Although the importance of koff and P app is well recognized the experimental measurements of the values of these parameters are scarce. Furthermore the reported koff and P app values not only depend on the cellulose substrate but also the method used. The processivity values of TrCel7A measured on different cellulose substrates using different methods fall in-between 10 and 70 (6,19,35,(40)(41)(42)(43). The differences in koff values measured for TrCel7A using different methods are even more prominent ranging from 10 -6 to 10 -1 s -1 (22,35,39,(43)(44)(45).
Although several studies have supported the obstacle model, slow complexation (46) or a combination of slow complexation and slow dissociation (47) has been proposed as ratelimiting events. Furthermore, the inner processive cycle has been proposed to be ratelimiting when acting in synergy with endoglucanases (EGs) (37). These different proposals do not exclude one another as the rate-limiting step can change dependent on the experimental conditions. It stresses, however, the importance of the experimental conditions. Therefore, parallel assessment of binding, complexation, processivity, dissociation and synergism is needed to elucidate the role of the CBM or aromatic residues in active site.
Here we studied the role of CBM-linker and Trp-38 in TrCel7A by measuring their effect on total-and active site mediated binding, offrates, on-rates, processivity, and synergism with EG. A novel method relying on measuring the exchange rate of the enzyme between non-labeled and 14 C-labeled cellulose was introduced for measuring the dissociation of TrCel7A. We found that the CBM-linker had more influence in determining on-rates than off-rates. The Trp-38 to Ala substitution resulted in decreased processivity and approximately 2 -3 fold increase in off-rate constant on both crystalline and amorphous cellulose. The effect of Trp-38 to Ala substitution on on-rates was strongly dependent on the presence of CBM-linker indicating compensation between CBM-linker and Trp-38. The presence of both CBM-linker and Trp-38 was required for the efficient degradation of cellulose in the presence of EG.

Measuring the off-rates of TrCel7A with Substrate Exchange Experiments
Recently a new method, referred to as substrate exchange experiment (SEE), for measuring koff for dissociation of chitinase from chitin was developed (31). The method is based on measuring the exchange rate of the enzyme between a non-labeled (substrate of interest) and a 14 C-labeled polymeric substrate (reference substrate). The method does not rely on the assumptions about the rate-limiting step, though the experiment conditions must be adjusted so that both substrates are at saturating concentrations for the enzyme. Here we adapted the SEE for measuring the koff values of cellulase TrCel7A. Besides intact WT enzyme its catalytic domain lacking the CBM and linker peptide (WTCD) was studied. To reveal the contribution of W38 at binding site -4 the W38A mutants of both variants, W38A and W38ACD, respectively, were also included. Because of its high binding capacity 14 C-labeled amorphous cellulose ( 14 C-AC) was used as a reference substrate. The koff values were measured for the dissociation from Avicel, the wood derived model cellulose often used in cellulase studies. Comparison of the reference time curves (Avicel and 14 C-AC were mixed together before the addition of the enzyme) with the time curves of the hydrolysis of 14 C-AC present alone reveals that 14 C-AC at 2 mg ml -1 effectively outcompetes Avicel at 50 mg ml -1 (Fig 2). In SEE Avicel at 100 mg ml -1 was pre-incubated with 0.4 µM TrCel7A for 1 h before an equal volume of 14 C-AC was added. With W38ACD as an exception a lag phase in the formation of soluble 14 C-product (expressed in 14 C-cellobiose ( 14 C-CB) equivalents) characteristic to SEE was observed. The koff values were found by comparison of SEE with the reference time curves (see experimental section) and are listed in Table 1. The absence of clear lag-phase with W38ACD is apparently a mixed effect of fast dissociation and non-complete saturation with Avicel before the addition of 14 C-AC (the Km for W38ACD for Avicel is 24.3 mg ml -1 (22)). Working with Avicel at concentrations significantly higher than 100 mg ml -1 was not possible for technical reasons. Unfortunately the SEE attempts, using bacterial cellulose (BC) or amorphous cellulose (AC) as a substrate, were not successful. The absence of lag-phase with BC can be caused by the difficulties in obtaining saturating conditions with BC and/or multiple binding modes of TrCel7A on this substrate (48,49). Using AC, we found that 14 C-AC was not efficient enough to compete with AC at saturating concentration for the enzymes.

Apparent processivity and rate constant of the initiation of processive runs
Since the koff is directly related to the enzymes intrinsic processivity (50) the enzyme variants with increased off-rates are expected to have also lower processivity. Therefore we also measured the processivity of TrCel7A and its variants. Here we used the method that relies on the hydrolysis of reduced cellulose under single-hit conditions (35,51). The apparent processivity is given by the ratio of the number of processive cuts (Ncatal) and the number of initiations of processive runs (Ninit). It has been shown that for the enzymes employing reducing-end exo and/or endo-mode initiation the Ninit is given by the number of insoluble reducing ends (IRG) generated on reduced cellulose. The Ncatal is represented by the total amount of enzyme generated reducing groups (RGtot). Therefore, P app =[RGtot]/[IRG]. The concentration of IRGs was measured by fluorescence labeling of enzyme generated reducing ends with anthranilic acid (AA).
First the P app was measured using reduced bacterial cellulose (rBC) as a substrate. Wild type enzymes, WT and WTCD, produced more soluble reducing groups (SRGs) compared to their W38A variants (Fig 3A). At shorter hydrolysis times the highest amount of IRGs was produced by W38A. However, this high initial activity was followed by a rapid slowdown. The slowdown in IRG production was most evident in case of W38ACD where the increase in IRGs after the first 10 min remained below the limit of detection. The P app was found as the slope of the linear regression line to the data plotted in coordinates [RGtot] versus [IRG] (Fig 3C). The highest P app of 70 ± 10 was measured for WT. This figure is in a good agreement with the P app of 61 ± 14 measured for TrCel7A with rBC substrate using diaminopyridine (DAP) labeling of IRGs (35). The W38A had 2.3 fold lower P app compared to the WT. The P app of WTCD remained in-between corresponding figures of WT and W38A ( Table 1). Because of the very low amount of IRGs we could not measure the P app of W38ACD.
The P app values were also measured using reduced amorphous cellulose (rAC) as a substrate. Contrary to rBC the W38A variants outperformed the WT and WTCD in both, the production of SRGs and IRGs (Fig 4 A & B). However, the P app of W38A variants was lower than that of their wild type counterparts (Fig 4 C, Table 2) indicating that P app is not the sole determinant of activity. With all enzymes the P app measured with rAC was significantly lower than that found with rBC. The P app 17 ± 2 found for WT on rAC was within the error limits with P app of 20.8 ± 1.2 measured for the same system using DAP labeling of reducing ends before (35).
Besides measuring P app the production of IRGs on reduced cellulose can be used as a measure of koff (31,35). As originally developed the rate constant of IRG production (kIRG) was taken equal to the koff using the following relation koff ≈ vIRG/[E]tot. Using the total concentration of enzyme ([E]tot) instead of the concentration of active site bound enzyme assumes saturating substrate concentration. Furthermore, it assumes that the dissociation is rate limiting for the enzyme recruitment so that the rate of the initiation of processive runs is governed by koff. However, kIRG represents the dissociation rate constant of a polymer chain from the active site if [E]tot is replaced with the concentration of the enzyme with active site occupied by the polymer chain ([Enzyme]bound- [Enzyme]bound-OA is measured in parallel with the rate of IRG formation (vIRG) by measuring the inhibition of the hydrolysis of low molecular weight reporter molecule like 4methylumbelliferyl-β-lactoside (MUL) by reduced cellulose. In this way we found that at 100 nM total enzyme concentration and rBC at 1 mg ml -1 the concentration of [WT]bound-OA, [WTCD]bound-OA, and [W38A]bound-OA were 94.9 ± 0.8 nM, 52.9 ± 6.6 nM, and 56.7 ± 8.9 nM, respectively. The level of active site bound enzyme was constant within the studied time interval (10 min -60 min) (data not shown). In the case of W38ACD the concentration of active site bound enzyme was below the limit of detection suggesting that the low rate of IRG production was caused by inefficient binding of the cellulose chain in the active site. Provided with the concentration of enzyme with occupied active site and rates of IRG formation the kIRG values on rBC were calculated for WT and WTCD ( Table 1). Because of the nonlinear time curve of IRG formation the kIRG for W38A was not found. Notably, the kIRG values of WT and WTCD on rBC were close to their koff values measured with Avicel using SEE. The consistency between koff and kIRG is expected when (i) exomode initiation from the non-reducing end is negligible, and (ii) the lateral diffusion on cellulose surface is not the prevalent mode of targeting the hydrolysis initiation sites. If the former assumption is violated, kIRG will underestimate koff. If the latter assumption does not hold kIRG will overestimate koff since enzyme makes the number of initiations (revealed by kIRG) before the full dissociation from the surface takes place (revealed by SEE koff). The latter scenario was recently demonstrated for chitinase ChiA from bacterium Serratia marcescens (SmChiA) (31). However, lateral diffusion was suggested to be unimportant for TrCel7A (49).
With rAC (1 mg ml -1 ) the concentrations of active site bound enzymes was close to the total enzyme concentration (100 nM). An exception here was W38ACD with active site bound enzyme concentration of 42.4 ± 3.7 nM.
The kIRG values on rAC (

Binding to BC and AC
Different levels of binding to rBC at 1 mg ml -1 prompted us to study the binding in more detail. For that the binding to BC was studied at different cellulose concentrations. Furthermore, the binding was studied on the level of total bound enzyme ([Enzyme]bound-tot) as well as on the level of active site bound enzyme. The difference between [Enzyme]bound-tot and [Enzyme]bound-OA represents the population of bound enzyme with active site free from cellulose chain ([Enzyme]bound-FA). The binding of core domains was too weak to be saturated at highest practical possible BC concentration. Hence the binding studies to BC were conducted only with the intact enzymes. The binding of WT to BC was stronger than that of W38A. Another major difference between WT enzyme and W38A variant is that the variant has significant population of bound enzyme with free active site. Still, the concentration of active site bound W38A seems to level-off at the value of total enzyme concentration with increasing BC concentration (Fig 5). Because the binding of TrCel7A involves different binding modes which relative contribution depends on the enzyme to substrate ratio (48) a more thorough quantitative analysis of binding to BC was omitted.
To get insight into the binding of core domains we also measured the binding to AC, a substrate with high binding capacity. Here, only the active site level bound enzyme was measured ( Fig 6A). The binding curves were analyzed using simple hyperbolic relation between [Enzyme]bound-OA and the substrate concentration ([S]):  Table 2). The most remarkable is the different effect of the CBM-linker to the bindings of WT and W38A enzymes. In case of the WT, removal of the CBM-linker resulted in about twofold increase in [S]0.5, whereas the corresponding effect to the W38A was an order of magnitude higher (Table 2). We also rearranged the data to show a conventional binding isotherm that implies the saturation of the substrate with the enzyme (Fig 6B). The data were analyzed according to the one binding site Langmuir isotherm. Because of the weak binding the concentration of CDs was far from reaching saturating. For this reason only the binding of intact enzymes was analyzed. The WT had about two-fold higher binding capacity compared to W38A, 2.2 ± 0.1 µmol g -1 and 1.0 ± 0.1 µmol g -1 , respectively. The half-saturating free enzyme concentrations of WT and W38A were 1.5 ± 0.1 nM and 1.9 ± 0.6 nM, respectively.

Synergistic hydrolysis of BC
It is well known that efficient and complete degradation of recalcitrant polysaccharides can be achieved only by using synergistic enzyme mixtures instead of individual enzymes. Therefore we assessed the performance of W38A in synergistic hydrolysis of BC. It has been shown that, at optimal substrate concentration BC is very efficiently degraded by the mixture of just three enzyme components, TrCel7A, EG and β-glucosidase (BG) (37). The activity of the W38ACD variant was too low to be measured accurately and the performance of WTCD in synergistic hydrolysis of BC has been described before (37). Therefore, only intact enzymes were included in the studies of synergism here. First we followed the release of 14 C-product in hydrolysis of 14 C-BC by WT and W38A supplemented with EG TrCel5A and BG. 14 C-BC was used at low concentration (0.25 mg ml -1 ), a conditions where the highest synergism between TrCel7A and EG has been reported (37). Indeed, the activity of synergistic mixtures containing WT was very high with more than 50% of 14 C-BC solubilized during 1 h of hydrolysis. In a sharp contrast the mixture containing W38A showed only a very low activity that abruptly decayed after first few minutes of hydrolysis (data not shown). In tracking the causes we found that the low activity of W38A was caused by an inefficient binding through active site, which remained below the limit of detection. Next, we assessed the synergistic effect at different BC concentrations. As seen in Figure 7A the deficiency of W38A compared to WT decreased with the increasing BC concentration. This was also reflected in the degree of synergistic effect (DSE), a parameter showing the ratio of the activity of an enzymes mixture over the sum of the activities of the individual components. While the DSE of WT decreased with increasing BC concentration the DSE of W38A showed little dependence (Table S1). We note that the performance of W38A in synergistic hydrolysis of BC is similar to that of the WTCD described before (37). Since the active site mediated binding of WTCD has been shown to be negatively affected by the presence of EG we also measured the concentration of W38A bound through the active site in the presence of EG. Binding was measured after 20 min of hydrolysis at different BC concentrations. The active site mediated binding of W38A was clearly depressed by the presence of EG ( Fig  S1). This is in contrast to WT binding, which has been shown to be stimulated by EG (37). Finally we tested the effect of pre-treatment of BC with EG to the following hydrolysis by TrCel7A. For that the BC (1 mg ml -1 ) was incubated with 0.5 µM TrCel5A for 1 h. After removal of the EG by washing with alkali, the pre-treated BC was used as substrate for individual WT-or W38A enzymes. As also shown in earlier reports (52,53) the pretreatment of BC with EG stimulated the following hydrolysis by TrCel7A (Fig 7B). The stimulating effect of EG pre-treatment for WT was about two-fold and was independent of the BC concentration. In contrast the EG pre-treatment of BC had a negative effect to the following hydrolysis by W38A at low BC concentrations and a slightly stimulating effect at highest BC concentration used (Fig 7B).

Discussion
Processive GHs utilize a complex multi-step catalytic mechanism ( Figure 1C) and the identification of the rate-limiting step of these enzymes has been in a focus of intensive research. In this study we zoomed in on different parts of these molecular steps by varying the experimental conditions and the selected parts of the enzyme. More specifically the enzyme was changed by removing the CBM-linker and changing a Trp-38 in the initial part of the tunnel to an Ala. The combined effect of W38A and CBM-linker was also tested to see if an inter-domain synergy existed between the interaction of the CBM-linker and the CD.
Dissociation (off-rate) -The importance of koff in determining the hydrolysis rate of recalcitrant polysaccharides by processive GHs is well recognized (19,31,35,36,38,54). The results of many studies support the obstacle model whereby the hydrolysis rate is directly related to the P app and koff according to vCB = koff P app [Enzyme]bound-OA (36). For TrCel7A this simple relationship is supported by measurements showing that dissociation is much slower than the processive run with concomitant release of cellobiose product (37). Since the P app has been shown to be limited by the length of the obstacle free path on the substrate (35) most of the population of [Enzyme]bound-OA is expected to be nonproductive, e.g. stalled behind obstacles (38, 55, 56) (see Figure 1C). In line with this the CBH variants with higher off-rates have been shown to have higher maximal rates (22). The positive correlation between activity and offrates of GHs is most prominent on amorphous (31) and soluble polymeric substrates (29,30). However, in most of the studies the higher offrates of enzyme variants were suggested rather than measured directly. Here, we adapted the SEE for measuring the koff values of dissociation of TrCel7A from Avicel. The TrCel7A variants with higher expected offrates, WTCD, W38A, and W38ACD were also included. The pVmax values measured with Avicel have been published before (22) and they increased in the order of WT < WTCD < W38A < W38ACD. Although we could not measure the koff for W38ACD, the same sequence seems to hold also for the koff values (Table 1). Higher activity of the variants with higher off-rates is also consistent with the hydrolysis of rAC. Even though their P app was reduced ( Table 2) the W38A variants clearly outperformed their WT counterparts in production of SRGs (Fig 4A, note that in this experiment less than 50% of W38ACD was bound to rAC while other enzymes were saturated). Unfortunately we were not able to measure the koff values using SEE with AC and BC. Therefore, with rAC and rBC as substrates the kIRG was used as the measure of koff. The good consistency between the koff and kIRG values measured on crystalline substrates Avicel and rBC (Table 1) suggests that for TrCel7A these two approaches provide equivalent rate constants. For separate domains of TrCel7A only the dissociation of its CBM from bacterial microcrystalline cellulose has been measured before with the resulting koff value of 0.029 s -1 (57). This figure is more than an order of magnitude higher than the koff value of WTCD measured here ( Table 1), suggesting that dissociation of the WT should be governed by the dissociation of the CD. This interpretation also parallels the conclusions in a recent, independent study (49). The 2.3 fold higher koff value of the WTCD compared to that of the WT measured here may indicate to the effect of physical linkage between two domains on dissociation/association. Indeed, the synergistic effect of the physical linkage between two CBMs in the binding to different celluloses has been demonstrated (58). However, one must bear in mind that the koff of 0.029 s -1 has been measured for TrCel7A CBM having no glycosylated linker (57). The presence of the glycosylated linker has shown to increase the affinity of the binding of TrCel7A CBM to BC by a factor of 10 (9). Unfortunately the relative contribution of onand off-rate components in a binding affinity of the glycosylated linker is not known.

Feeding of cellulose chain (on-rate) -
Dissecting the [S]0.5 values for binding to AC into its on-and off-rate components reveals that the CBM-linker had little effect on the dissociation from the amorphous cellulose ( Table 2, Table 3). In contrast the CBM-linker had a prominent effect on the on-rates especially in the case of the W38A variant where the removal of the CBM-linker resulted in a 100 fold decrease in the kon value ( Table  3). Comparison of the effects on the on-rates reveals that the absence of Trp-38 can be compensated in a large part by the CBM-linker and vice versa. Contrary to what was observed with on-rates the increase in off-rate (about 2.5 fold) caused by the Trp-38 to Ala substitution was independent on the presence of the CBMlinker (Table 3). With crystalline substrates, Avicel and BC, we do not have full data-set to analyze the effect of the CBM-linker and the Trp-38 on both, on-and off-rates. A moderate effect (about 2-3 fold increase) of the CBMlinker (in WT enzyme) and the Trp-38 (in intact enzyme) on the off-rates (Table 1) in parallel with the drastically reduced binding affinity of the W38ACD (too low to be measured with BC) suggest that the CBMlinker/Trp-38 compensation in on-rates holds also with crystalline substrates. Compensation of the reduced binding affinity upon W40A substation in the binding site -7 of TrCel7A by CBM has been reported before (25). Altogether these results suggest that there is a strong synergistic effect between the CBMlinker and the entrance region of the active site tunnel in feeding the cellulose chain into the active site. About two-fold lower maximum adsorption capacity to AC of W38A compared to WT (Fig 6B) suggests a role of Trp-38 in chain end recognition. A similar effect of Trp38 was reported also with crystalline Avicel (22). Apart from this suggestion, the data presented here does not allow us to single out which elementary step in feeding is synergistically targeted.
Synergism -We started this discussion with demonstrating the consistency between the experimental data and the off-rate limited hydrolysis by individual CBHs. The situation may be different with optimal synergistic enzyme mixtures where the rate of cellulose hydrolysis may be limited by the rate of the glycosidic bond hydrolysis (37). Several studies have pointed to that the CBM has no effect on the kcat of the glycosidic bond hydrolysis (7,37,42). Likewise, analysis of the rates of Avicel hydrolysis measured in both, steady and pre-steady state regimes suggests that the W38A substitution has no significant effect on kcat (Fig S2). The poor performance of WTCD (37) and W38A (Fig 7) in synergistic hydrolysis of BC under conditions that are optimal for WT apparently reflects deficiency in feeding of a cellulose chain. The deficiency of W38A in feeding of a cellulose chain is much more evident in synergistic hydrolysis than when the enzyme acts in isolation. This suggest that the synergistic interaction between CBM-linker and tunnel entrance region is pre-requisite in feeding EG generated chain ends. A possible mechanistic interpretation is that EG generates so called blunt chain ends (59) that are more tightly attached to the bulk cellulose crystal and are more difficult to feed into the active site.
Based on literature data and observations made here, the role of the active site aromatic residues near the tunnel entrance and the CBM-linker of TrCel7A can be envisaged. The primary role of Trp residue at the tunnel entrance, Trp-40 is chain end recognition (7,25,60). A similar role of the Trp at the tunnel entrance has been suggested for other GHs like TrCel6A (61) and TrCel7B (62), and chitinases (63). Our data point towards the importance of Trp-38 at subsite -4 for progression of the cellulose chain end in the active site tunnel during the feeding. Trp-38 locates in the active site tunnel just before the cellulose chain is twisted (Fig 1B) (24,33). In this position the medium affinity region of the active site tunnel borders with the low affinity binding region (64). Therefore one can speculate that Trp-38 is needed for holding the cellulose chain end in the position, long enough to proceed through the region of twist. Once the chain end has proceeded through the region of twist the further sliding and processive hydrolysis is not affected by W38A. This is supported by a limited effect of Trp-38 to Ala substitution on kcat of glycosidic bond hydrolysis. Since P app is independent on the on-rate the reduced P app of the W38A variants apparently reflects the increased koff values. Computational studies have revealed that Trp-38 to Ala substitution in TrCel7A results in the decrease in binding free energy by 3.8 kcal mol -1 (62). The similar conclusions about the role of the Trp-167 (analogous to the Trp-38 in TrCel7A) in determining the binding free energy (63), P app and koff values (31) can be made for chitinase, SmChiA. Thus, the Trp residue in binding site -4/-3 appears to be functionally equivalent in two enzymes with different active site architectures, i.e tunnel shaped active site of TrCel7A versus deep cleft shaped active site of SmChiA. The role of the CBM in increasing the enzymes affinity for the substrate has been recognized for a long time (65) and we suggest that this is mainly due to increased on-rate.
Conclusion -Taken together, the primary role of aromatic residues near active site entrance is in feeding of the cellulose chain into the active site tunnel. They also contribute to processivity via a reduction of the off-rate. Decreasing the binding strength in the active site entrance region leads to increased off-rates. This may be reflected in an increased overall activity under conditions where the availability of the chain ends compensates for the deficient feeding, i.e at saturating substrate concentrations. The inefficient feeding of the chain end by the Trp to Ala substitution near the tunnel entrance can be compensated in a large part by the CBMlinker. It is therefore possible that, analogous to the Trp-38, the CBM-linker is required for the efficient chain end feeding into the active site.

Enzymes
Intact TrCel7A (WT) and its W38A variant were heterologously expressed in Aspergillus oryzae and purified as described previously (22). W38ACD was constructed as described in Kari 2014. For WTCD the intact TrCel7A was first purified from the culture filtrate of T. reesei QM 9414 (66) and WTCD was prepared by limited proteolysis of intact TrCel7A with papain (67). Aspergillus BG was purified from Novozyme ® 188 (C6105, Sigma) as described before (68).

Celluloses
BC was prepared by laboratory fermentation of the Gluconobacterium xylinum strain ATCC 53582 as described elsewhere (69). 14 C-BC was prepared as BC but the glucose carbon source was supplemented with 14 C-glucose as described Jalak 2012. The specific radioactivity of 14 C-BC preparation was 6.4 x 10 5 dpm mg -1 . AC was prepared from Avicel as described before (69). 14 C-AC was prepared from 14 C-bacterial microcrystalline cellulose by dissolution in and regeneration from phosphoric acid (69) and had specific radioactivity of 6.4 x 10 5 dpm mg -1 . rBC and rAC were prepared from BC and AC respectively using reduction with sodium borohydride (69).

Substrate exchange experiment (SEE) -In SEE,
Avicel (100 mg ml -1 ) was pre-incubated with 0.4 µM TrCel7A and 50 nM BG in 50 mM sodium acetate buffer pH 5 (containing BSA 0.2 mg ml -1 ) at 25 °C for 1 h with magnetic stirring. 1 ml of the above reaction mixture was added to an equal volume of 14 C-AC and the release of radioactivity in time was followed. Final concentrations of Avicel, 14 C-AC, TrCel7A and BG were 50 mg ml -1 , 2 mg ml -1 , 0.2 µM and 25 nM, respectively. After mixing with 14 C-AC the stirring was omitted but the reaction mixture was gently suspended with pipette before each sampling. For time points 0.2 ml aliquots were withdrawn and stopped by adding NaOH to 0.1 M. For zero time points NaOH was added before enzymes. Cellulose was separated by centrifugation (5 min 10 4 × g) and radioactivity in the supernatant was quantified using a liquid scintillation counter. The released radioactivity was expressed in 14 C-CB equivalents. The reference curves were made using the same experiment conditions as in SEE (see above) but Avicel and 14 C-AC were mixed together before the addition of the enzymes. Hydrolysis time courses of 14 C-AC present as a sole substrate were also made. In this case the experiment setup was identical to that used to generate reference curves but Avicel was replaced with the equal volume of buffer.
The koff values were found by comparing the release of radioactivity in reference curves and SEE as described before (31). In short, the reference curves were analyzed using nonlinear regression according to [ 14 C-CB] = A*t b where t is time and A, and b are empirical constants (70). The SEE time curves were analyzed using non-linear regression according to [ 14 C-CB]=A*(1-e -t*koff )t b . In latter regression the value of the constant b was fixed to the value found from the analysis of reference curves so that only the values of koff and A were let free (31) Measuring Apparent Processivity and rate constant of the initiation of processive runs -rBC or rAC (1 mg ml -1 ) was incubated with 0.1 µM TrCel7A in 50 mM sodium acetate buffer pH 5 (containing BSA 0.1 mg ml -1 ) at 25 °C. At selected times the reaction was stopped by the addition of NaOH to 0.2 M. For zero time points NaOH was added before the enzyme. Cellulose was separated by centrifugation (2 min 10 4 × g) and supernatant was used for measuring SRGs using 3-methyl-2benzothiazolinone hydrazone hydrochloride (MBTH) method, as described in Horn et al. (71). The amount of IRGs was measured using fluorescence labeling of cellulose with AA (69). First cellulose was washed 2x with 1 ml water, 1x with 1 ml sodium acetate buffer and 1x with 1 ml water. All washings were performed through centrifugation (2 min 10 4 × g) and re-suspension steps. After washing cellulose was suspended in 0.2 ml water and AA labeling was carried in 80% buffered methanol at 80 o C, 2 h in the presence of 0.5 M NaCNBH3 and 50 mM AA (69). Cellulose concentration was 1.6 mg ml -1 . After AA labeling the cellulose was washed 3x with 1 ml water, incubated 1 h at RT with 0.2 M sodium hydroxide to remove the unspecific labeling. After alkali treatment the AA-labeled cellulose was further washed 3x with 50 mM NaAc buffer pH 5. Before fluorescence measurement the AA-labeled cellulose was solubilized by incubating overnight with crude mixture of T. reesei cellulases. The fluorescence of the hydrolysate was measured at excitation and emission wavelengths of 330 nm and 435 nm, respectively.
Measuring the total and the active site mediated binding of TrCel7A with BC -For quantifying Cel7Abound-OA the initial rates of the MUL hydrolysis by TrCel7A in the presence of BC were measured (48). For that the BC (0.1 -5 mg ml -1 ) was incubated with 0.1 µM TrCel7A and 0.1 µM β-glucosidase in 50 mM sodium acetate buffer pH 5 (containing BSA 0.1 mg ml -1 ) at 25 °C. After 20 min MUL was added to a final concentration of 5 µM and after further incubation for 2 min the reaction was quenched by the addition of ammonium hydroxide to 0.1 M. After separation of cellulose (2 min 10 4 × g) the concentration of released MU was found by measuring the fluorescence in the supernatant. The concentration of TrCel7A with free active site was found from the rate of MUL hydrolysis, and the concentration of Cel7Abound-OA was found as a difference of total concentration of TrCel7A and TrCel7A with free active site (48).
For quantifying the concentration of Cel7Abound-tot BC (0.1 -5 mg ml -1 ) was incubated with 0.1 µM TrCel7A and 0.1 µM BG in 50 mM sodium acetate buffer pH 5 (containing BSA 0.1 mg ml -1 ) at 25 °C. After 20 min the cellulose was separated by filtration through glass microfiber filter. Filtrate was further centrifuged (2 min 10 4 × g) and the concentration of TrCel7A free from cellulose was measured by following the MUL (5 µM) hydrolyzing activity in the supernatant using appropriate calibration curves (48). The concentration of Cel7Abound-tot was found as the difference between the total concentration of TrCel7A and the concentration of the TrCel7A free from cellulose.

Measuring the active site mediated binding of TrCel7A with AC
[Cel7A]bound-OA with AC was determined analogously to that with BC. The AC (0.001 -5 mg ml -1 ) was incubated with 5 nM TrCel7A and 10 nM β-glucosidase in 50 mM sodium acetate buffer pH 5 (containing BSA 0.1 mg ml -1 ) at 25 °C. After 20 min the MUL was added to a final concentration of 5 µM and after further incubation for 1 h the reaction was quenched by the addition of ammonium hydroxide to 0.1 M. AC was separated by centrifugation (2 min 10 4 × g) and the concentration of Cel7Abound-OA was measured as described for BC (see above). http://www.jbc.org/ Downloaded from separated by centrifugation (5 min 10 4 × g) and the concentration of hydrolysis products was measured from the radioactivity in the supernatant using liquid scintillation counter.

Pre-treatment of BC with EG and further hydrolysis with Cel7A
For the pre-treatment BC (1 mg ml -1 ) was incubated with 0.5 µM EG Cel5A for 1 h in 50 mM sodium acetate buffer pH 5 at 25 °C. The reaction was stopped by adding NaOH up to 0.1 M. To remove EG the pre-treated BC was thoroughly washed with 0.1 M NaOH, water and 50 mM sodium acetate buffer, pH 5 through centrifugation (10 min 4500 × g) and re-suspension steps. For the following hydrolysis with TrCel7A the pre-treated BC (0.25 -2.5 mg ml -1 ) was incubated with 0.1 µM TrCel7A in 50 mM sodium acetate buffer pH 5 (containing BSA 0.1 mg ml -1 ) at 25 °C for 1 h. The reaction was stopped by the addition of NaOH to 0.2 M. After separation of cellulose by centrifugation (2 min 10 4 × g) the concentration of reducing sugars in the supernatant was measured using MBTH method (71). Experiments without EG pretreatment were performed exactly as described above (including washing steps) but the EG was omitted. The koff values measured using SEE (Fig. 2). b data from Kari et. al. 2014. c kon values are calculated according to kon = koff/Km using listed koff and Km values. d kIRG represents the off-rate constant measured from the generation of IRGs on rBC (Fig 3B). d P app is found as a slope of the linear regression line of the data in coordinates [RG]tot vs [IRG] (Fig 3  C).

TABLE 2 Kinetic parameters of TrCel7A and its variants measured with amorphous cellulose
a kIRG represents the off-rate constant measured from the generation of IRGs on rAC (Fig 4B). b P app is found as a slope of the linear regression line of the data in coordinates [RG]tot vs [IRG] (Fig 4  C). c [S]0.5 was found by the non-linear regression analysis of the data in Fig 6A according

TABLE 3
The effect of the Trp-38 to Ala substitution and CBM-linker to off-and on-rates on AC Figure 1. Structure and mechanism of TrCel7A. A) Cartoon representation (light blue) of catalytic domain, linker and CBM for TrCel7A in complex with a cellulose strand. The O-glycosylation of the linker is shown as a surface representation and cellulose strand is representated as green stricks. The image was made using the crystal structure of the catalytic domain (Protein Data Bank code 8CEL) and CBM (Protein Data Bank code 1CBH). B) Positions of the Trp residues in the active site tunnel of TrCel7A in complex with a cellodextrin chain (green). Numbers refers to the different binding subsites in the tunnel with -7 at the entrance and -1/+1 being the position of the scicile bond. Trp-38 investigated in this study is highlited in maganta. C) Molecular steps in the hydrolysis of cellulose for a processive enzyme. Steps leading from the free enzyme in the solution to the enzyme with reducing end of cellulose chain in the binding site -1 are collectively referred to as feeding. Processive catalysis includes the formation of Michaelis complex (by sliding the chain end from binding site -1 to +2), hydrolysis of glycosidic bond, and expulsion of cellobiose (green ellipces). Processive catalysis is repeated until the enzyme meets an obstacle (depicted here as upper cellulose fibril) or happens to dissociate. In SEE Avicel (100 mg ml -1 ) was preincubated with 400 nM TrCel7A for 1 h, after which an equal volume of 14 C-AC (final concentration 2 mg ml -1 ) was added, and the release of radioactivity (expressed in 14 C-CB equivalents) in time was followed (◊). In reference time curves (♦) the same conditions were used as in SEE but Avicel was mixed with 14 C-AC before the addition of TrCel7A. Control experiments (□) of hydrolysis of 14 C-AC (2 mg mL -1 ) in the absence of Avicel are also shown (made in one parallel). Error bars show S.D. and are from two independent experiments. Solid lines represent the best fit from non-linear regression analysis (see experimental procedures). The TrCel7A variants used were: A, WT; B, W38A; C, WTCD; D, W38ACD.