Systematic deletions in the cellobiohydrolase (CBH) Cel7A from the fungus Trichoderma reesei reveal flexible loops critical for CBH activity

Glycoside hydrolase family 7 (GH7) cellulases are some of the most efficient degraders of cellulose, making them particularly relevant for industries seeking to produce renewable fuels from lignocellulosic biomass. The secretome of the cellulolytic model fungus Trichoderma reesei contains two GH7s, termed TrCel7A and TrCel7B. Despite having high structural and sequence similarities, the two enzymes are functionally quite different. TrCel7A is an exolytic, processive cellobiohydrolase (CBH), with high activity on crystalline cellulose, whereas TrCel7B is an endoglucanase (EG) with a preference for more amorphous cellulose. At the structural level, these functional differences are usually ascribed to the flexible loops that cover the substrate-binding areas. TrCel7A has an extensive tunnel created by eight peripheral loops, and the absence of four of these loops in TrCel7B makes its catalytic domain a more open cleft. To investigate the structure–function relationships of these loops, here we produced and kinetically characterized several variants in which four loops unique to TrCel7A were individually deleted to resemble the arrangement in the TrCel7B structure. Analysis of a range of kinetic parameters consistently indicated that the B2 loop, covering the substrate-binding subsites −3 and −4 in TrCel7A, was a key determinant for the difference in CBH- or EG-like behavior between TrCel7A and TrCel7B. Conversely, the B3 and B4 loops, located closer to the catalytic site in TrCel7A, were less important for these activities. We surmise that these results could be useful both in further mechanistic investigations and for guiding engineering efforts of this industrially important enzyme family.

Glycoside hydrolase family 7 (GH7) cellulases are some of the most efficient degraders of cellulose, making them particularly relevant for industries seeking to produce renewable fuels from lignocellulosic biomass. The secretome of the cellulolytic model fungus Trichoderma reesei contains two GH7s, termed TrCel7A and TrCel7B. Despite having high structural and sequence similarities, the two enzymes are functionally quite different. TrCel7A is an exolytic, processive cellobiohydrolase (CBH), with high activity on crystalline cellulose, whereas TrCel7B is an endoglucanase (EG) with a preference for more amorphous cellulose. At the structural level, these functional differences are usually ascribed to the flexible loops that cover the substratebinding areas. TrCel7A has an extensive tunnel created by eight peripheral loops, and the absence of four of these loops in TrCel7B makes its catalytic domain a more open cleft. To investigate the structure-function relationships of these loops, here we produced and kinetically characterized several variants in which four loops unique to TrCel7A were individually deleted to resemble the arrangement in the TrCel7B structure. Analysis of a range of kinetic parameters consistently indicated that the B2 loop, covering the substrate-binding subsites ؊3 and ؊4 in TrCel7A, was a key determinant for the difference in CBH-or EG-like behavior between TrCel7A and TrCel7B. Conversely, the B3 and B4 loops, located closer to the catalytic site in TrCel7A, were less important for these activities. We surmise that these results could be useful both in further mechanistic investigations and for guiding engineering efforts of this industrially important enzyme family.
The cellulases Cel7A and Cel7B are abundant both in the secretome of the cellulolytic fungus Trichoderma reesei and enzyme mixtures for the industrial deconstruction of lignocellulosic biomass (1)(2)(3)(4). These two enzymes are homologs and share the same overall fold with almost 50% amino acid sequence identity (1). Nevertheless, they attack their insoluble substrate in quite different manners. Cel7A from T. reesei (TrCel7A) is a typical processive cellobiohydrolase (CBH) 4 that targets the reducing end of a cellulose strand and subsequently makes consecutive cuts that release cellobiose as the enzyme moves along the strand (5)(6)(7)(8)(9). Conversely, TrCel7B is categorized as an endoglucanase (EG) that typically targets the strand internally and shows little or no processivity (10,11). TrCel7A and TrCel7B are also quite different with respect to their ability to hydrolyze different types of cellulose. Hence, TrCel7A is among the most efficient enzymes for the breakdown of crystalline cellulose. TrCel7B shows very limited activity against crystalline substrate, but effectively hydrolyzes amorphous parts of the cellulose (10,12,13). Moreover, in terms of kinetics, TrCel7A and TrCel7B display several opposite characteristics: TrCel7B has a much higher turnover frequency than TrCel7A (14), decreased product inhibition (15,16), and lower adsorption on cellulose compared with TrCel7A (17). These differences play a key role in the cellulolytic capacity of the fungal secretome as they promote conversion of heterogeneous substrates with different cellulose structure and give rise to a distinctive synergy between the two enzymes (12, 18 -20). In light of this, it is interesting to consider why structural homologs harbor these functional differences. The main structural differences between TrCel7A and TrCel7B are found in the flexible loops that cover the active site (21)(22)(23). TrCel7A has a total of eight such loops that form the tunnel. There are four loops on each side termed A1-A4 and B1-B4, respectively (24). In TrCel7B, four of these loops (A4, B2, B3, and B4) are truncated to an extent where they are essentially absent (23). This results in a more open and accessible substrate-binding cleft in TrCel7B, as illustrated in Fig. 1A. Several earlier studies have suggested that the functional differences between CBHs and EGs can be attributed to loop configuration (25). As an example, Meinke et al. (26) explored the importance of the loops for distinguishing between exo-and endolytic activity in a GH6 cellulase. In that case, one of the two main loops that covered the active site was deleted, which increased the endolytic activity. Furthermore, Von Ossowski et al. (27) showed that a deletion in the B3 loop (c.f. Fig. 1A) in TrCel7A caused a decrease in activity on crystalline cellulose but increased activity on amorphous substrate. This latter variant also showed decreased processivity and product inhibition and hence a more EG-like behavior. In addition to these experimental studies, several computational works (11, 24, 28 -30) have expanded our understanding of the loop properties in TrCel7A.
In the current work, we have investigated the functional role of TrCel7A loops. We focused on the four loops that are absent in TrCel7B and their role in the regulation of CBH-and EGtype activity. Specifically, we conducted comprehensive kinetic studies of the two WTs, TrCel7A and TrCel7B, and a group of TrCel7A variants in which these loops were trimmed individually.

Design and expression
Six variants were designed based on pairwise sequence and structural alignments of TrCel7A and TrCel7B. The variants were constructed by systematically deleting individual loops of TrCel7A to mimic the corresponding regions of TrCel7B, as specified in Fig. 1. Three variants with different degree of truncation of the B2 loop were made. One, henceforth termed ⌬B2-1, had a 14-amino acid deletion (⌬Trp 192 -Gly 205 ), whereas the two others had deletions of 13 (⌬Glu 193 -Gly 205 ) and 6 amino acids (⌬Ser 196 -Thr 201 ), termed ⌬B2-2 and ⌬B2-3, respectively. Two other variants, termed ⌬A4 and ⌬B3, had deletions in the A4 loop

Systematic loop deletions in T. reesei Cel7A
(⌬Asn 384 -Ser 388 ) or the B3 loop (⌬Gly 245 -Gly 253 ), respectively. The last variant, named ⌬B4, had a deletion of 4 amino acids (⌬Glu 335 -Phe 338 ). Although this area is not considered part of the B4 loop in the assignment proposed by Momeni et al. (24), sequence alignment indicates that deleting this region could create a variant more similar to TrCel7B, where an entire ␣ helix is absent (␣2 in Fig. S1). All variants and their names are listed in Fig. 1, and a more thorough account of the alignment may be found in Fig. S1.

Basic characterization
We first made a general characterization to assess whether the loop variants were pure, stable, and active. In this work, we used a soluble substrate analog. A more detailed characterization of the activity on real cellulosic substrate will be discussed below.
The purified enzymes (WTs and variants) showed a single band in SDS-PAGE, and their monomeric status was confirmed by size-exclusion chromatography (see Figs. S2 and S3). Next, we measured thermal stability by differential scanning calorimetry (DSC) and kinetic parameters for the hydrolysis of the substrate para-nitrophenyl ␤-D-lactopyranoside (pNPL). Results of this work are listed in Table 1. Stability and kinetic constants for WT TrCel7A were similar to previously reported values (31,32). DSC measurements showed that all variants had a small to moderate lowering of the transition temperature, T m . The most unstable variant was ⌬A4, with a loss of T m of about 10°C. Functionally, this variant was also severely impaired, and this most likely reflects that the tested truncation of the A4 loop led to major structural changes in the variant. By contrast, the three ⌬B2 variants showed moderate decrease in T m and only small changes in kinetic parameters on pNPL (k cat , K m , and cellobiose inhibition constant K i ; Table 1) compared with WT TrCel7A. The ⌬B3 variant had essentially unchanged stability but was different from the WT with respect to kinetic parameters on pNPL. Specifically, k cat , K m , and K i (cellobiose) were all increased by some 3-5-fold compared with WT TrCel7A. This kinetic behavior for a variant with truncated B3 loop has been reported before (27), and in this latter work, it was shown by crystallography that the deletion did not lead to structural disturbances. In light of this kinetic resemblance and small changes in T m , it appears that the ⌬B3 variant accommodated the mutation without major structural changes. Finally, the ⌬B4 variant showed kinetic parameters on pNPL and an inhi-bition constant very similar to that of the WT. This also supported the view of a mainly conserved overall fold of the variant, albeit with lowered thermostability (⌬T m ϭ 7.9°C).
In conclusion, the basic characterization suggested that five variants with truncations in either the B2, B3, or B4 loop could accommodate the deletions without major changes in their properties, and we consequently used these variants in the comparative analyses on real insoluble cellulose. The ⌬A4 variant, on the other hand, appeared severely changed and hence inapt for further comparisons of loop effects.

Binding and enzyme kinetics on Avicel
Microcrystalline cellulose Avicel, which has comparable fractions of crystalline and amorphous cellulose (10), was used as substrate for the primary characterization of all enzymes with respect to steady-state kinetics and adsorption. We first used conventional Michaelis-Menten (MM) kinetics, where the initial rate, v 0 , was estimated in experiments with low enzyme concentration and variable substrate loads. The results ( Fig. 2A) were analyzed with respect to the MM equation, using nonlinear regression, and the resulting parameters, conv V max /E 0 and conv K m , are listed in Table 2. The superscript "conv" specifies that these parameters were measured by the conventional approach (i.e. with excess of substrate). Kinetic parameters for TrCel7A on Avicel were in agreement with an earlier study (31). Results for the two WTs in Table 2 showed that the maximal specific rate and the Michaelis constant were an order of magnitude higher for TrCel7B ( conv V max /E 0 ϳ 2 s Ϫ1 and conv K m ϳ 44 g/liter) compared with TrCel7A (ϳ0.2 s Ϫ1 and ϳ4 g/liter). Interestingly, the specificity constant, was essentially the same for all enzymes (WTs and variants) ( Table 2). This implies that the catalytic efficacy was approximately the same and further supports the view that the variants accommodated the loop truncations without major structural disturbances. Returning to the kinetic parameters of the two WTs, the results were in line with the interpretation

Systematic loop deletions in T. reesei Cel7A
that conv V max is limited by slow dissociation from the substrate (5,33) and hence that the comparably low maximal rate for TrCel7A is linked to its strong binding to the substrate (34). This higher affinity of TrCel7A is reflected in a low conv K m in Table 2. TrCel7B, on the other hand, binds the substrate less tightly (10-fold higher conv K m ), and this allows faster dissociation and hence higher conv V max . In the current context, it is interesting to relate this relationship of rate and affinity to loop morphology. Thus, stronger substrate interaction of TrCel7A may rely, at least in part, on the extensive loop coverage of the substrate-binding tunnel, which enables numerous enzymeligand contacts that are not found in TrCel7B (29). The additional interactions in TrCel7A manifest themselves in the kinetic data, as argued above, and in the following, we will use this relationship to assess functional roles of the loops from the kinetic properties of the variants. First, we note that deletions in the B3 or B4 loops only brought about marginal changes in conv V max /E 0 and conv K m . This means that these variants maintained the high substrate affinity and low turnover number that is characteristic for TrCel7A and hence that the B3 and B4 loops were not critical for ligand interaction. By contrast, any of the three tested deletions in the B2 loop significantly increased both conv V max /E 0 and conv K m ( Table 2), suggesting that the B2 loop plays a key role in substrate binding. These observations suggest that loss of the B2 loop made TrCel7A much more "EG-like" than loss of either the B3 or B4 loop. In the following, we will analyze other functional data along the same lines, to further elucidate how the investigated loops control the balance between typical EG and CBH behavior of TrCel7 enzymes. Whereas the conventional MM equation can be applicable to the current system (35,36), it has been suggested that a so-called inverse MM approach offers more robust steady-state analysis for cellulases (37,38). In this inverse approach, the roles of enzyme and substrate are swapped, so that enzyme is in excess in the experiments and saturation represents a situation where all sites on the solid surface are occupied (with additional free enzyme in the aqueous phase). To perform this type of analysis, we measured the initial rate at low substrate load and a range of enzyme concentrations. Results in Fig. 2B were analyzed with respect to the inverse MM equation (39), using nonlinear regression, and the resulting parameters inv V max /S 0 and inv K m are listed in Table 2. Interestingly, these inverse parameters confirm the overall picture from the conventional MM, in the sense that the parameters for all ⌬B2 variants shifted strongly toward the EG, whereas ⌬B3 and ⌬B4 only shifted slightly. This again underscores the importance of the B2 loop for the function of TrCel7A. To illustrate the molecular meaning of the inverse parameters, we note that the inverse maximal rate, inv V max , will scale with the number of attack sites on the surface of the cellulose particle. The more sites on which the enzyme can bind and form a productive complex, the higher the inv V max . One consequence of this is that the number of attack sites per gram of cellulose, ⌫ attack , can be roughly estimated as the ratio of maximal specific rates from the inverse and conventional MM analysis (37).

Systematic loop deletions in T. reesei Cel7A
Results for ⌫ attack in Table 2 reveal distinctive differences between TrCel7A and TrCel7B. The number of attack sites recognized per gram of Avicel is some 50-fold higher for TrCel7A compared with TrCel7B. This ability of TrCel7A to attack a broad range of structures on the heterogeneous surface has been reported before (40), and the distinctively reduced value for the homologous TrCel7B points toward loop morphology as a key determinant of this ability. Although molecular interpretations of this difference await further experimental work, we conclude that TrCel7A is promiscuous with respect to the conformation of the cellulose strand it attacks, whereas TrCel7B is a selective enzyme, which is only capable of forming catalytically competent complexes with a small subset of sites on Avicel. When considering ⌫ attack for the variants, it appeared that different deletions in the B2 loop consistently lowered ⌫ attack by an order of magnitude ( Table 2). This suggested that the B2 loop is vital for the ability of TrCel7A to attack variable surface structures. Conversely, deletions in the B3 and B4 loops only produced negligible changes in ⌫ attack , and we deduce that these loops are of minor importance for the ability to thread and convert different structures on the surface. It is interesting to compare these attack site densities with the maximal adsorption capacities, ⌫ max , derived from the binding isotherms in Table 2 and Fig. 2C. Unlike ⌫ attack , we found that ⌫ max varied only moderately between the investigated enzymes, and this suggests that the studied loops are relatively unimportant for the enzymes' ability to adsorb onto Avicel. Taken together, these results show that the B2 loop is crucial for the ability to form catalytically competent complexes on different surface structures, but not for adsorption per se. Overall, both conventional and inverse steady-state analyses highlighted the B2 loop as a key determinant of TrCel7A's function as a CBH. Conversely, kinetic parameters of the ⌬B3 and ⌬B4 variants were only marginally changed compared with WT TrCel7A, and we conclude that these loops played minor roles for the distinction between CBH-and EG-like function.

Activity on other cellulosic substrates
CBHs and EGs differ with respect to their preference for different types of cellulosic substrates. Typically, EGs are most active on amorphous substrates, whereas CBHs are particularly effective degraders of crystalline cellulose (10). To assess these differences for the loop variants, we made end-point activity measurements on regenerated amorphous cellulose (RAC; predominantly amorphous cellulose) and bacterial microcrystalline cellulose (BMCC; predominantly crystalline cellulose). Results shown in Fig. 3 confirmed the typical preferences of CBH and EG WTs mentioned above. More importantly, we found that the loop variants showed intermediate preferences.
Again, these results followed the picture where the three ⌬B2 variants shifted distinctively toward a more EG-like behavior with a 4 -5-fold increase in the activity against RAC and a loss of activity against BMCC. The analogous changes were small for the ⌬B3 and ⌬B4 variants, and these loops hence did not appear to be of key importance for the differences in substrate specificity.

Endolytic activity
The defining property of an EG is its ability to perform endolytic cleavage. To assess this, we investigated the activity against the synthetic insoluble substrate azurine-cross-linked hydroxyethylcellulose (AZCL-HE-cellulose), which is designed to elucidate endoactivity. As expected, the results in Table 3 show much higher activity on this substrate of TrCel7B compared with TrCel7A (about 300 times). For the variants, we found that the three ⌬B2 variants had become more EG-like with about 4 times higher activity than TrCel7A (Table 3). For the ⌬B3 and ⌬B4 variants, we saw a marginal increase in the activity on AZCL-HE-cellulose. These results clearly follow the picture from kinetic data on Avicel in as much as the B2 loop appears to be particularly important for functional distinction. We note, however, that even the most active variant (⌬B2-3) was some 70 times less efficient on AZCL-HE-cellulose than TrCel7B. This clearly suggests that full endolytic activity as in TrCel7B cannot be established through the deletion of only one loop, and this probably reflects limitations in the accessibility of the substrate binding region exerted by the remaining loops and their mutual interactions. When comparing with the other kinetic data, we note that the shift toward EG-like behavior of the ⌬B2 variants was less pronounced for AZCL-HE-cellulose activity (Table 3) than for kinetic parameters on real cellulose ( Table 2). This may be related to the cross-linked nature of the artificial AZCL-HE substrate as well as its bulky chromogenic group, which could make this substrate particularly sensitive to reduced accessibility.

Table 3 Relative values of endolytic activity, normalized with respect to TrCel7A
Endolytic activity was assessed on the endoglucanase-specific substrate AZCL-HEcellulose.

Enzyme
Relative activity on AZCL-HE-cellulose

Conclusions
The kinetic parameters of TrCel7A and TrCel7B WTs in Table 2 unveil some interesting differences and similarities of these homologous enzymes. Thus, when acting on the real insoluble substrate, the EG (TrCel7B) had a maximal turnover defined in the conventional way ( conv V max /E 0 ), which is an order of magnitude higher than the CBH (TrCel7A). The (conventional) Michaelis constant, conv K m , was also an order of magnitude higher for TrCel7B, and it follows that the kinetic efficiency (or specificity constant), conv V max /E 0 / conv K m , was identical for the two WTs (Table 2). We used the differences in kinetic parameters for variants to elucidate functional roles of the loops. We found that the B2 loop, which covers the region around pyranose subsites Ϫ3 and Ϫ4 quite far from the catalytic residues (24,29,41), was particularly important for CBHlike behavior. Hence, variants with deletion of either 14, 13, or 6 residues in this loop consistently showed strong shifts toward more EG-like function. By contrast, deletions in the B3 or B4 loops were of minor importance, and these variants essentially maintained characteristic CBH kinetics as well as the ability to attack a broad range of cellulose conformations on the substrate surface. For the ⌬B4 variant, the observation of moderate kinetic changes could be connected to the small size of the loop, but interestingly, the sizable deletion of 9 residues in the larger B3 loop did not have much impact on the kinetics either. These observations clearly point toward a special role of the B2 loop as a regulator of kinetic properties, and further studies of this loop appear promising for both mechanistic understanding and guidance for enzyme engineering of the industrially important GH7 family.

Design of variants
Structures of TrCel7A (PDB entry 4C4C) and TrCel7B (PDB entry 1EG1) were visualized and structurally aligned using the PyMOL Molecular Graphics System, version 2.0 (Schrödinger, LLC). The pairwise alignment was made with Clustal Omega (42) and analyzed with the ESPript 3.0 web server using default parameters (43). Details about the alignment can be found in Fig. S1.

Thermal stability
All enzyme characterization was performed in triplicates and with a buffer of 50 mM sodium acetate, pH 5.0, henceforth referred to as standard buffer. All chemicals were purchased at Sigma-Aldrich. Physical stability of all investigated enzymes was evaluated by DSC using MicroCal VP-Capillary DSC from Malvern Panalytical. The enzymes were buffer-changed by Vivaspin 20, 20,000 molecular weight cut-off (Sartorious, Stonehouse, Gloucestershire, UK) and suspended with standard buffer and subsequently diluted to a concentration of 0.5 mg/ml. The enzyme stability was tested with heating scans from 20 to 95°C with a scan rate of 3.3°C. For all enzymes, a distinct thermal transition was evident, whereas no transition was observed for the buffer scans. Data were collected and analyzed with Origin version 7 software (OriginLab, Northampton, MA), and the buffer scan was subtracted from each measurement. T m was calculated after fitting and baseline subtraction of the obtained thermograms.

Analytical size-exclusion chromatography
To determine the structural integrity and oligomerization state of the enzymes, size-exclusion chromatography was performed. The enzymes were prepared to concentrations above 1 mg/ml and subjected to a Superdex TM 75 Increase 10/300 GL column (GE Healthcare) and eluted isocratically at a flow of 0.5 ml/min with 25 mM MES and 200 mM NaCl, pH 6, buffer. Gel filtration calibration kit LMW (GE Healthcare) was used for determination of the void volume and for calibration to determine the molecular weight of the enzymes.

Activity on para-nitrophenol-␤-lactoside and estimation of inhibition constant (K i ) for cellobiose
Kinetic parameters on the soluble synthetic substrate pNPL and K i on cellobiose were estimated at 25°C. Michaelis-Menten curves with 11 different pNPL concentrations ranging between 0.16 and 5 mM were made, in the presence or absence of cellobiose, at two concentrations: 100 and 200 M for TrCel7A and the deletion variants, 2 and 4 mM for TrCel7B. The final enzyme concentration was 0.5 M in all cases. The reactions were started by mixing 90 l of pNPL at different concentrations with 60 l of enzyme in a 96-well microtiter plate, followed by sealing and incubation at 25°C in a Thermo-Mixer operating at 1,100 rpm for 30 min. The reactions were quenched by the addition of 150 l of 1 M Na 2 CO 3 . 150 l of the mixtures were then transferred to a new microtiter plate, and the concentration of the product para-nitrophenol (Thermo-Fisher, Kandel, Germany) was quantified in a spectrophotometer by measuring the absorbance at 405 nm and using a calibration curve of para-nitrophenol at six different concentrations (15-500 M). Appropriate blanks were subtracted from all measurements. The experimental curves were analyzed by the software OriginPro using nonlinear regressions. To determine which inhibition mech-

Systematic loop deletions in T. reesei Cel7A
anism better described the experimental data, all curves were fitted with a global fit of different inhibition models by using the smallsample corrected Akaike information criterion (31,47,48).

Activity on microcrystalline cellulose, conventional Michaelis-Menten ( conv MM)
The activity of the TrCel7A and TrCel7B WT and the six variants was tested on Avicel PH101. The substrate was washed six times in MilliQ water and three times in standard buffer. Aliquots of 230 l of washed Avicel with loads between 0.5 and 110 g/liter were then transferred to 96-well microtiter plates (96F 26960 Thermo Scientific, Waltham, MA), and the enzymatic reaction was started by adding 20 l of enzyme stock prepared in standard buffer to a final concentration of 50 nM. Each plate was sealed and mixed at 1,100 rpm in a Thermo-Mixer equipped with a ThermoTop (Eppendorf, Hamburg, Germany). The enzyme-substrate contact time was 60 min at 25°C. The reaction was stopped by a short 3-min centrifugation at 2,000 ϫ g. Volumes of 60 l of the supernatant were retrieved and analyzed for its content of reducing sugars by using the 4-hydroxybenzoic acid hydrazide (PAHBAH) method (49). Specifically, a solution of 15 g/liter PAHBAH was dissolved in 0.177 M potassium sodium tartrate tetrahydrate and 0.5 M NaOH. 90 l of this solution were mixed with the reaction supernatant, followed by heating at 95°C for 10 min in a T100 TM PCR cycler (Bio-Rad) and cooling to 5°C for 5 min. Finally, 100 l were transferred into a microtiter plate, and the absorbance at 405 nm was measured by a plate reader (Spectra-Max M2e, Molecular Devices, Sunnyvale, CA) using the software SoftMax Pro version 6.2, Molecular Devices. A standard curve of 1,000 to 32.15 M of cellobiose (Fluka) dissolved in standard buffer was included in each plate.

Binding isotherms and inverse Michaelis-Menten ( inv MM)
To quantify enzyme adsorption on microcrystalline cellulose, a constant substrate load of Avicel was used, and the enzyme concentration varied between 0.05 and 6 M. The enzymatic reaction was started by adding 60 l of enzyme at different concentrations to microtiter plates containing 190 l of Avicel so the final substrate concentration was 8 g/liter. The samples were mixed in a ThermoMixer for 1 h at 25°C and then centrifuged at 2,000 ϫ g for 3 min to separate free enzyme from substrate-bound enzyme. 60 l of supernatant were mixed with 90 l of standard buffer in a black microtiter plate (Greiner bio-one655079),andtheintrinsicproteinfluorescencewasmeasured in a plate reader by using excitation and emission wavelengths of 280 and 345 nm, respectively. The free enzyme concentration was quantified by comparing the fluorescence signal with a calibration curve made with enzyme dissolved in standard buffer with known concentrations ranging from 0.05 to 6 M. In the inverse MM approach, the reaction condition was the same as for Langmuir isotherms. The reactions were incubated and analyzed for the reducing sugar content as described for the conv MM approach. Because increasing enzyme concentration affects the absorbance signal, enzyme blanks for each concentration were included.

Activity on RAC and BMCC
RAC was prepared from Avicel according to a protocol described previously (50). 230 l of RAC suspensions dissolved in standard buffer were mixed with 20 l of TrCel7A, TrCel7B, or the deletion variants in microtiter plates to a final enzyme concentration of 50 nM and a final RAC concentration of 4 g/liter. BMCC was prepared from HCl treatment of bacterial cellulose, using a protocol described previously (51). 190 l of BMCC suspension in standard buffer were mixed with 60 l of TrCel7A, TrCel7B, or the deletion variants in microtiter plates to a final concentration of 100 nM and a final BMCC concentration of 4 g/liter. The enzymatic reactions on RAC and BMCC were carried out at 25°C in ThermoMixers operating at 1,100 rpm, with a contact time of 1 h. The reactions were stopped by centrifugation, and the reducing sugar content was analyzed using PAHBAH, as described previously.

Endolytic activity on AZCL-HE-cellulose
Endolytic activity of the enzymes was estimated with the chromogenic insoluble substrate AZCL-HE-cellulose (Megazyme, Bray, Ireland). A stock suspension of AZCL-HE-cellulose in standard buffer at a concentration of 5 g/liter was prepared, and aliquots of 90 l were transferred to transparent microtiter plates. The enzyme reaction was started by adding 60 l of enzyme stock to a total final enzyme concentration of 5 M. The plates were incubated at 25°C in a ThermoMixer with ThermoTop. For TrCel7A and deletion variants, the reaction time was 2 h, and for TrCel7B, it was 15 min. The reactions were stopped by adding 150 l of NaOH 0.1 M, followed by centrifugation for 3 min at 2,000 ϫ g. 100 l of the supernatant was transferred to microtiter plates, and the adsorption at 595 nm was measured in a plate reader. Activity was expressed as adsorption difference between reaction and a substrate blank divided by the reaction time in minutes (⌬A 595 /min).