Formulation of enzyme mixtures for effective hydrolysis of alkaline peroxide treated alfalfa hay

Glucose release from enzymatic hydrolysis of AP alfalfa

Compared to rumen enzymes alone, a combination of rumen enzymes (60%) with β-glucosidase (E-BGLUC; 20%) and β-xylosidase (E-BXSRB; 20%) resulted in a two fold increase (Figure 1a and Additional file 1a) in glucose release from AP alfalfa hay. These results suggest a positive synergetic interaction between rumen enzymes with β-glucosidase and β-xylosidase. The observed increase in glucose yield with supplementary β-glucosidase (E-BGLUC) and β-xylosidase (E-BXSRB) activity reflects possible feedback inhibition of rumen cellulase by oligomers released as results of enzymatic digestion. It has been well documented that the presence of sufficient β-glucosidase is important in reducing the inhibition of cellulose by cellobiose [13, 14]. The reason for suboptimal β glucosidase activity in the rumen enzyme preparation is unknown, but may reflect the fact that the enzyme mixture used originated from the fluid fraction of rumen contents. It is well known that significant quantities of beta-xylosidase and beta-glucosidase activity are associated with the particulate fraction of rumen contents. It is also possible that in the rumen, intact bacteria uptake cellobiose directly from rumen fluid thereby preventing it from accumulating to levels that impede cellulose digestion [15].

The model predicted optimized enzyme mixtures containing Accellerase 1500 (50%) with polygalacturonase (PGA28_ASPNG: 20%), endo-xylanase (XYN11A_THITE; 20%), feruloyl esterase (FAE 1a; 7%) and β-glucosidase (E-GLUC; 3%) resulting in higher glucose release as compared to control (100% Accellerase 1500) (Figure 1b and Additional file 1b). These results agree with a recent report identifying hemicellulose as an important physical barrier to enzymatic hydrolysis of cellulose, as it blocks enzyme access to the cellulose surface [16]. Hemicellulose, particularly in the form of xylan and its oligomers, has been documented to inhibit cellulase activity even more than the accumulation of glucose or cellobiose [16]. The present study indicates that more efficient cellulose hydrolysis would be achieved by formulating Accellerase XC (73%) with endoglucanase (EGL7A_THITE: 17%) and beta-xylosidase (E-BXSRB; 10%) activity (Figure 1c and Additional file 1c) as compared to Accellerase XC alone. Enhanced glucose release as a result of the addition of EGL7A may be attributed to additional cellulase activity or to the broad substrate specificity exhibited by members of GH7. These endoglucanases have been reported to readily act on xylan, arabinoxylan [17] and xyloglucan [18]. The phylogenetic analysis among the catalytic cores of GH7 enzymes in the CAZy database suggests that GH7 endoglucanases and xylanase may have diverged from a common ancestral gene, possible resulting in GH7 endoglucanases retaining remnants of xylanse activity [18]. These results can be explained by hypothesis that GH7 improves overall cell wall conversion through exposing more xylan trapped within the cell wall matrix, increasing the release of xylo-oligomers. It has been reported that xylo-oligomers are strong inhibitors of cellulase and inclusion of β-xylosidase (Trichoderma) reduces the inhibitory impact of these end products on cellulose hydrolysis [16]. Furthermore, xylanases (Thermoascus aurantiacus) alone have been shown to enhance the bioconversion of both cellulose and hemicellulose [19].

Xylose release from AP pre-treated alfalfa

A seven fold increase in xylose release from alfalfa (Figure 2a and Additional file 2a) was predicted when mixed rumen enzymes (60%) were complemented with β-glucosidase (E-GLUC; 20%) and β-xylosidase (E-BXSRB; 20%). Furthermore, the model predicted a four-fold increase in xylose release (Figure 2b and Additional file 2b) when Accellerase 1500 (75%) was complemented with β-glucosidase (E-GLUC; 25%). Interestingly, optimum xylose yield did not require the addition of xylanase, suggesting that there was sufficient xylanase already in Accellerase 1500. Of the commercial enzyme mixtures investigated, Accellerase XC contained the highest xylanase activity (2500–3800 acid birchwood xylanase units (ABXU/g), and consequently was expected to be most efficient at xylan hydrolysis. However, complementing Accellerase XC with endoglucanase (EGL7A_THITE; 21%) resulted in 1.6 fold higher xylose yield (Figure 2c and Additional file 2c) from AP pretreated alfalfa as compare to Accellerase XC alone. These results suggest improved xylan hydrolysis with increased cellulose solubility and reflect the layered structural complexity of cellulose and xylan in the plant cell wall.

Formulation of enzyme mixtures for effective hydrolysis of alkaline peroxide treated barley straw

Glucose release from AP pretreated barley straw

Maximum glucose yield (1.6 fold higher than control) from barley straw was observed (Figure 3a and Additional file 3a) when mixed rumen enzymes (70%) were combined with family 7 endoglucanase (EGL7A_THITE; 21%) and acetyl xylan esterase (AXE16A_ASPNG; 9%). Glucose yield from Accellerase 1500 was nominally (1.2 fold) higher when it was mixed with feruloyl esterase (FAE1a; 17%) and endoglucanase (EGL7A_THITE; 17%) (Figure 3b and Additional file 3b). Approximately 1.86 fold more glucose was released (Figure 3c and Additional file 3c) by supplementing Accellerase XC (50%) with feruloyl esterase (FAE1a; 25%) and acetyl esterase (AXE16A_ASPNG; 25%). These results demonstrate the importance of esterase activity for holocellulose hydrolysis in barley straw. Our results also resonate with an earlier report [20, 21] of conservation of acetyl and feruloyl ester linkages in hot water pretreated corn stover. It has also been documented that peroxide treatment at low concentrations (≤2.0%; as used in this study) did not result in notable lignin oxidation, whereas treatment with H₂O₂ at ≥3.0% resulted in significant cleavage of carboxyl groups [10]. Our results are consistent with the notion that ferulic acid linkages to lignin and /or other xylan/pectin chains are in part responsible for recalcitrance of cellulose microfibrils even after mild oxidative treatment. Cleavage of these linkages seems to be imperative in providing access of both cellulases and xylanases to the core of the cell wall matrix [10, 20].

Xylose release from pre-treated barley straw

Xylose release from barley straw was five-fold higher when mixed rumen enzymes (87%) were complemented with polygalacturonase (PGA28A_ASPNG; 11%) and feruloyl esterase (FAE1a; 2%) (Figure 4a and Additional file 4a). This suggests that cross-linked pectin/xylan limited the digestion of barley straw. Furthermore, an optimized mixture containing Accellerase 1500 (50%), arabinofuranosidase (ABF54B_ASPNG; 17%), 16% and 10% acetyl-xylan-esterase AXE16A_ASPNG and AXE16B_ASPNG respectively, and 7% E-BGLUC, was predicted to release 1.4 fold more xylose (Figure 4b and Additional file 4b) from barley straw than Accellerase 1500 alone. Likewise complementing Accellerase XC (66%) with feruloyl esterase (FAE1a; 16%) and acetyl-xylan-esterase (AXE16A_ASPNG; 18%) resulted in the highest xylose yield (Figure 4c and Additional file 4c) from barley straw as compared to Accellerase XC alone. These results suggest that cross-linked polymers with lignin are the primary recalcitrant structure inhibiting the conversion of barley straw to fermentable sugars. Synergistic interaction between feruloyl esterase and xylanase to increase the release of ferulic acid as well as synergetic interaction between feruloyl esterase, xylanase, cellulase, arabinofuranosidase and arabinanases to increase the release of reducing sugar has been reported by others [22].

Our results are consistent with previously reported correlations between lignin phenotype (lignin content and ferulate crosslinking) and enzymatic deconstruction of monocot (barley) and dicot (alfalfa) cell walls. With dicots (alfalfa), a strong negative correlation has been shown between initial lignin content and cellulolytic decomposition of dilute acid pretreated fibrous substrates [23]. Our model also suggests that efficient deconstruction of delignified AP alfalfa can be achieved by core enzymes alone, while feedback inhibition by released products was also an important factor controlling sugar release. With barley as a monocot, digestibility has been reported to not depend on initial lignin content, but rather the degree of ferulate bridges was strongly negatively correlated to digestibility [19, 24]. This observation was reflected by our results where esterase activity was shown to be critical for the comprehensive enzymatic deconstruction of barley straw.

To validate predictions made by statistical design, the profile of glucose and xylose released from AP pretreated barley and alfalfa as a function of incubation time and protein concentration were further studied. At 6 h, glucose release from AP treated barley as well as alfalfa was observed to be nearly linear for all four mixtures with increasing enzyme concentration (data not shown). For AP treated barley straw, optimized Accellerase 1500 mix was superior (p < 005) to Accellerase 1500 at 24 h (Figure 5a). Although both optimized Accellerase 1500 and Accellerase 1500 showed similar yields after 24 h, and 48 h of incubation with alfalfa hay, the optimized Accellerase 1500 mixture resulted in higher (p < 005) glucose yield from AP treated alfalfa (Figure 5b). At 48 h, with all four enzyme mixtures, sugar release tended to plateau at protein loads above 20 mg/g glucan, with optimized Accellerase 1500 exhibiting the highest glucose release from AP treated barley straw (Figure 5b). Similarly optimized Accellerase XC for barley straw and alfalfa hay, were also observed to enhance xylose release (p < 005) from both substrates (Figure 5c, d). Plotting data collected at 15 mg/g of glucan against time showed that optimized enzyme mixtures released more glucose and xylose than the original commercial enzyme preparations at terminal incubation (Figure 6a and b).

Modeled predictions were further validated with a scale up assay using (2% w/v) AP treated barley straw in a 2 ml reaction volume. We selected optimized Accellerase XC mix for the scale up validation as it predicted the same enzyme components (Accellerase XC, FAE 1a and AXE 16A) (Figures 3c and 4c) for enhanced glucose as well as xylose from barley straw. Superior performance of optimized Accellerase XC over Accellerase XC was confirmed in the scale up through the observation of increased glucose and xylose release (Figure 7a, b). A direct relationship was observed between xylan conversion (fraction of available xylan converted) and glucose conversion (fraction of available glucan conversion) for Accellerase XC as well as for optimized Accellerase XC (Figure 8a, and Additional file 5). However, the correlation between xylan and glucan digestion in optimized Accellerase XC (Figure 8b) was higher than conventional Accellerase XC, suggesting improved glucan conversion as a result of enhanced xylan saccharification.. Since FAE 1A and AXE 16A were important components of optimized Accellerase XC, we further analysed the release of acetate. Accellerase XC exhibited a weak correlation between xylose and acetate release, while there was a strong correlation between xylose and acetate release with optimized Accellerase XC. The synergetic relationship between Accellerase XC with FAE 1a and AXE 16A was evident by the direct relationship between xylose release and acetate release during incubation (Figure 8c, d). These results suggest that Accellerase XC lacked sufficient esterase activity to effectively, hydrolyze AP pretreated barley straw, a limitation that was overcome in part by the addition of complementary esterase activity.

Enzymes

Two acetyl xylan esterases, AXE16A_ASPNG and AXE16B_ASPNG; a polygalacturonase (PGA28A_ASPNG), a α-arabinofuranosidase (ABF54B_ASPNG); an endoglucanase GH7 (EGL7A_THITE) and an endoxylanase (XYN11A_THITE) were produced as recombinant proteins in A. niger. Recombinant feruloyl esterase (FAE 1a) was cloned into the pET30a expression vector and overexpressed in Escherichia coli BL21 (DE3). β-glucosidase (E-BGLUC) and β-xylosidase (E-BXSRB) were procured from Megazyme International (Bray, Ireland). Mixed rumen enzymes were prepared from rumen fluid collected from two rumen cannulated cows fed a mixed alfalfa hay (30%), barley silage (50%) and barley grain (20%; DM basis) diet. Rumen contents were collected 2 h after feeding from the reticulum, ventral, caudal and dorsal-ventral sac of the reticulo-rumen of each cow and thoroughly mixed. Contents were strained through 4 layers of cheesecloth and the collected fluid was centrifuged at 38,300 × g for 15 min. The supernatant (500 mL) was lyophilized and reconstituted in 50 mM sodium citrate (pH 5.0, containing 5 μg/mL tetracycline, 5 μg/mL cycloheximide and 0.02% sodium azide) and used as source of rumen mixed enzymes. Three batches of rumen mixed enzymes from three different collections of rumen fluid were prepared. The commercial enzyme preparations of Accellerase 1500 and Accellerase XC were obtained from Genencor (Rochester, NY, US). Accellerase 1500 contained endoglucanase (2200–2800 carboxymethycellulose (CMC U) U/g) and β-glucosidase activities (450–775 p-nitrophenyl-β-D-glucopyranoside (pNPG U) U/g) whilst Accellerase XC contained mainly endoglucanase (1000–1400 CMC U/g) and xylanase activity (2500–3800 acid birchwood xylanase units ABXU/g; Genencor, Rochester, NY, US).

Production of acetyl xylan esterase, polygalacturonase, α-arabinofuranosidase, endoglucanase, endoxylanase in A. niger

Nucleotide sequences corresponding to target genes were obtained from Mycocosm genome resource of the Joint Genome Institute (http://genome.jgi-psf.org/programs/fungi/index.jsf). Additional file 6: Table S1 outlines the origin of the cloned genes and the carbohydrases that they encode. Cloning was accomplished using the Gateway recombination method with Invitrogen enzymes (Life Technologies Inc., Burlington, ON). Forward and reverse primers for cloning possessed at their 3′ end 20–25 nucleotides identical to the N terminal and C-terminal portions of the coding region of the targeted ORF, along with Gateway BP reaction compatible recombination sites [25]. Complementary DNA prepared from poly (A) + RNA was amplified by polymerase chain reaction (PCR). The amplified PCR products were cloned into ANIp7G, a Gateway-compatible vector previously constructed from ANIp7 [26]. Protoplasts of A. niger strain N593 glaA::hisG were transformed with the target genes as previously described [27]. FAE 1a from Anaeromyces mucronatus was cloned, expressed in E. coli BL21 and purified as earlier described [28].

Biochemical characterization of recombinant enzymes

Recombinant proteins produced in A. niger were purified at room temperature by anion exchange chromatography using an ÄKTA chromatography system (Amersham Biosciences, Piscataway, NJ). Prior to chromatography, supernatants from A. niger cultures were concentrated using a Vivaspin ultrafiltration device (GE Healthcare Life Sciences, Baie d’Urfe, PQ) at a 10 kDa cut-off. The concentrated proteins were repeatedly washed in 20 mM Tris–HCl buffer, pH 8.0, by dilution and ultrafiltration, and the retentate was applied to a MonoQ HR 5/10 anion exchange column equilibrated with the same buffer. Bound proteins were eluted using a linear 0 to 1 M KCl gradient in 20 mM Tris–HCl buffer, pH 8.0. Fractions (1 mL) were collected at a flow rate of 1 mL/min and stored on ice until analyzed for enzyme activity. Purified protein was stored at -80°C for further use. Protein purity was checked by SDS-PAGE as previously outlined [29].

All enzyme reactions were carried out in triplicate. Glycoside hydrolase activity was determined using a reducing sugar assay performed in a 96-well micro plate format with an assay volume of 50 μL. Briefly, for a 50 μL assay, 10 μL of substrate (1%) was added to 30 μL of 50 mM Britton-Robinson buffer (50 mM boric acid, 50 mM acetic acid and 50 mM phosphoric acid) pH 5.0 and the reaction initiated by the addition of the appropriate enzyme diluted in 10 μL of Britton-Robinson buffer [30]. The reaction mixture was immediately incubated at 40˚C for 30 min. Reaction mixtures were placed on ice and 10 μL was withdrawn and mixed with 190 μL of ice-cold BCA reagent (Bicinchoninic acid assay, Sigma-Aldrich, Oakville, ON) and incubated at 80˚C for 40 min for colour development. The resultant mixture (160 μL) was loaded onto a flat bottom-micro plate and the optical density read at 562 nm. The monosaccharide subunits constituting the polysaccharide substrates were used to prepare standard curves. Polysaccharide substrates (all from Sigma-Aldrich unless otherwise specified) were: birchwood xylan (xylanases); carboxymethylcellulose (endoglucanase); and polygalacturonic acid, sodium salt (polygalacturonase). Arabinofuranosidase activity was determined using 4-nitrophenyl-L-arabinofuranoside as a substrate in 50 mM Britton-Robinson buffer, pH 5.0, as described previously [30]. Acetyl-esterase activity was assessed by measuring released acetic acid using a kit purchased from Megazyme.

Statistical design

Design-Expert® software (Version 8.0; Stat-Ease, Inc., Minneapolis, MN; http://www.statease.com) was used to create the simplex-lattice designs and to analyze responses. All experiments were performed as mixtures at 15 mg protein loadings per g cellulose of the pretreated alfalfa hay or barley straw. The number of mixtures in the simplex-lattice depended on both the number of components in the mixture and the degree of the polynomial. Using an augmented special quadratic design, ten component designs resulted in 66 separate assays. In order to optimize the hydrolysis efficiency of enzyme cocktails the relative abundance of each component was varied using the experimental design outlined in Additional file 7: Table S2. The lower and upper limits of each component were determined with an aim to analyze synergetic interaction among fungal enzymes (AXE16A_ASPNG, AXE16B_ASPNG, PGA28A_ASPNG, ABF54B_ASPNG, EGL7A_THITE, XYN11A_THITE, FAE 1a, E-BGLUC and E-BXSRB) with mixed rumen enzymes or with commercial enzyme preparation to achieve enhanced biomass conversion to fermentable sugar. Therefore, relative abundance of core enzymes (i.e., mixed rumen enzymes/Accellerase 1500/Accellerase XC) was set to vary from 50% to 100%, while upper limit and lower limits for fungal enzymes were set between 50% - 0% in assay mixtures.

Formulation of enzyme mixtures for effective conversion of alkaline peroxide pre-treated (AP) alfalfa hay and barley straw

Data analysis

Absorbance values from the glucose and xylose assays were converted to relative percent yields (relative to controls i.e., commercial enzyme or rumen enzyme mix only) and these values served as responses in the experimental design. For all experiments, ANOVA calculations of F value, P value, R², Adjusted R², Predicted R² and Adequate Precision were computed by the Design-Expert software as shown in Additional file 8: Table S3 and Additional file 9: Table S4). The F value indicated the effects (if any) of the individual components on the model. An F value close to 1 implied that the components of the mixture did not interact and hence had little effect on the model. Enzyme synergism was deemed to be present at a P value <0.0001. Adjusted R² and Predicted R² were estimated and if the difference between these values was > 0.2 the model was considered to be over-parameterized and a different order polynomial was tested and/or a backward or stepwise elimination regression was conducted with an ‘alpha out’ value set to 0.1 as suggested by Banerjee et al., 2010 [8]. This statistical approach eliminated all of the terms in the model that were insignificant (P > 0.0001) and the model that gave a difference between the Adjusted and Predicted R² values <0.2 was used to navigate the design space. Adequate precision estimated the signal-to-noise ratio, with a value >4 indicating adequate model discrimination. Once all the criteria for a robust model were fulfilled, the model was used to determine the enzyme mixtures that resulted in optimal glucose and xylose release. The predicted vs. actual estimates for each design are shown in Additional file 10.