Synthetic Enzyme Mixtures for Biomass Deconstruction ... · ARTICLE Synthetic Enzyme Mixtures for Biomass Deconstruction: Production and Optimization of a Core Set Goutami Banerjee,1,2

ARTICLE

Synthetic Enzyme Mixtures for BiomassDeconstruction: Production and Optimizationof a Core Set

Goutami Banerjee,1,2 Suzana Car,1,2 John S. Scott-Craig,1,2 Melissa S. Borrusch,1,2

Nighat Aslam,1,2 Jonathan D. Walton1,2

1Department of Energy Great Lakes Bioenergy Research Center,

Michigan State University, E. Lansing Michigan 48824; telephone 517-353-4885;

fax 517-353-9168; e-mail: [email protected] of Energy Plant Research Laboratory, Michigan State University,

E. Lansing Michigan 48824

Received 11 January 2010; revision received 5 March 2010; accepted 15 March 2010

Published online 26 March 2010 in Wiley InterScience (www.interscience.wiley.com)
. DOI 10.1002/bit.22741
ABSTRACT: The high cost of enzymes is a major bottleneckpreventing the development of an economically viable lig-nocellulosic ethanol industry. Commercial enzyme cocktailsfor the conversion of plant biomass to fermentable sugars arecomplex mixtures containing more than 80 proteins ofsuboptimal activities and relative proportions. As a steptoward the development of a more efficient enzyme cocktailfor biomass conversion, we have developed a platform,called GENPLAT, that uses robotic liquid handling andstatistically valid experimental design to analyze syntheticenzyme mixtures. Commercial enzymes (Accellerase 1000þ/� Multifect Xylanase, and Spezyme CP þ/� Novozyme188) were used to test the system and serve as comparativebenchmarks. Using ammonia-fiber expansion (AFEX) pre-treated corn stover ground to 0.5mm and a glucan loadingof 0.2%, an enzyme loading of 15mg protein/g glucan, and48 h digestion at 508C, commercial enzymes released 53%and 41% of the available glucose and xylose, respectively.Mixtures of three, five, and six pure enzymes of Trichodermaspecies, expressed in Pichia pastoris, were systematicallyoptimized. Statistical models were developed for the opti-mization of glucose alone, xylose alone, and the average ofglucoseþ xylose for two digestion durations, 24 and 48 h.The resulting models were statistically significant(P< 0.0001) and indicated an optimum composition forglucose release (values for optimized xylose release are inparentheses) of 29% (5%) cellobiohydrolase 1, 5% (14%)

Abbreviations used: Glu, glucose; Xyl, xylose; CBH, cellobiohydrolase; CBH1, cello-

biohydrolase 1 of Trichoderma reesei (GenBank CAA49596); CBH2, cellobiohydrolase 2

of T. reesei (GenBank P07987); BG, b-glucosidase; EG, endo-b1,4-glucanase; EG1,

endo-b1,4-glucanase 1 from T. reesei (GenBank AAA34212); EX, endo-b1,4-xylanase;

EX2, endo-b1,4-xylanase 2 from T. reesei (GenBank AAB29346); EX3, endo-b1,4-

xylanase 3 from T. reesei (GenBank BAA89465); BX, b-xylosidase; MW, molecular

weight; AFEX, ammonia fiber expansion.

Nighat Aslam’s present address is Department of Chemistry and Biochemistry,

Molecular Biochemistry Lab, University of Agriculture, Faisalabad 38040, Pakistan.

Correspondence to: J.D. Walton

Additional Supporting Information may be found in the online version of this article.

� 2010 Wiley Periodicals, Inc.

cellobiohydrolase 2, 25% (25%) endo-b1,4-glucanase 1,14% (5%) b-glucosidase, 22% (34%) endo-b1,4-xylanase3, and 5% (17%) b-xylosidase in 48 h at a protein loading of15mg/g glucan. Comparison of two AFEX-treated cornstover preparations ground to different particle sizes indi-cated that particle size (100 vs. 500mm) makes a largedifference in total digestibility. The assay platform andthe optimized ‘‘core’’ set together provide a starting pointfor the rapid testing and optimization of alternate coreenzymes from other microbial and recombinant sourcesas well as for the testing of ‘‘accessory’’ proteins for devel-opment of superior enzyme mixtures for biomassconversion.

Biotechnol. Bioeng. 2010;106: 707–720.

� 2010 Wiley Periodicals, Inc.

KEYWORDS: lignocellulose; Trichoderma reesei; Pichiapastoris; cellulase; xylanase; b-glucosidase; b-xylosidase;corn stover

Introduction

Current scenarios for the conversion of lignocellulosicbiomass to ethanol and other liquid transportation fuelsutilize enzymes to depolymerize polysaccharides to theirconstituent fermentable sugars (Dashtaban et al., 2009;Farrell et al., 2006; Wilson, 2009). However, the high cost ofenzymes is a major hindrance to the development of a viablelignocellulosic ethanol industry (Himmel et al., 2007; Lyndet al., 2008).

A multitude of enzymatic activities are required for theconversion of lignocellulosic biomass into useful (fermen-table) products. Known essential activities to degradecrystalline cellulose include cellobiohydrolase (CBH),

Biotechnology and Bioengineering, Vol. 106, No. 5, August 1, 2010 707

endo-b1,4-glucanase (EG), and b-glucosidase (BG). Alarger suite of enzymes is necessary to depolymerizehemicelluloses, including endo-b1,4-xylanase (EX),b-xylosidase (BX), a-arabinosidase, a-glucuronidase, andesterase. Currently available commercial enzyme prepara-tions for the depolymerization of lignocellulosic materialsare partially defined complex mixtures of the secretedproteins from filamentous fungi such as Trichoderma reeseithat have been grown in the presence of inducers such assophorose or lactose. Commercial enzyme ‘‘cellulase’’mixtures contain between 80 and 200 proteins (Banerjeeet al., 2010; Nagendran et al., 2009). Except for a few of thebetter-characterized cellulases and hemicellulases, the rolesof most of these proteins in lignocellulose deconstructionare poorly understood.

The way forward for the development of more efficientlignocellulose-degrading enzyme cocktails will requiredeeper and more precise knowledge about the specificenzymes that are involved in the degradation of lignocellu-lose. However, it is not possible to gain this knowledgeworking only with partially defined complex mixtures(Berlin et al., 2007; Decker et al., 2009; Kumar and Wyman2009; Zhang et al., 2006). The precise contribution of anindividual enzyme can be established only by working with itin a purified state (Kim et al., 1998).

Previous efforts to construct defined enzyme mixtureshave incorporated one or more of the following desirablecharacteristics: use of verified purified proteins produced insufficiently large quantities to permit comparative studiesover many months or years; a sufficiently large number ofenzymes to achieve a minimum reasonable release of at leastglucose (Glu) and xylose (Xyl), the major sugars in grassstover; statistically valid experimental design; moderate tohigh throughput through the use of microplates and liquidhandling robots; and the use of realistic lignocellulosicsubstrates (Berlin et al., 2006; Chundawat et al., 2008; Gaoet al., 2010; Kim et al., 1998; King et al., 2009; Irwin et al.,1993; Meyer et al., 2009; Rosgaard et al., 2007; Xiao et al.,2004).

Here we describe an enzyme production and assayplatform, called Great Lakes Bioenergy Research CenterEnzyme Platform (GENPLAT), that satisfies these experi-mental conditions. Six core enzymes were produced athigh levels by expression in the yeast Pichia pastoris,which has no interfering enzymatic background. Theenzymes were combinatorially assayed using automatedpipetting of corn stover slurry, enzyme additions, andsugar (Glu and Xyl) determinations. The system providesa robust platform that can be used for the identificationof superior forms of the core enzymes and for the discovery,testing, and optimization of accessory enzymes derivedfrom T. reesei and other microorganisms. The combina-tion of a library of purified enzymes and a highthroughput assay platform can also be employed for rapidoptimization of enzyme mixtures tailored to differentpretreatments, feedstocks, and pretreatment/feedstockcombinations.

708 Biotechnology and Bioengineering, Vol. 106, No. 5, August 1, 2010

Materials and Methods

Biomass Pretreatment and Grinding

Corn (Zea mays L.) stover cut into �1-cm pieces wassubjected to ammonia fiber expansion (AFEX) as described(Balan et al., 2009). For ‘‘GLBRC stover,’’ the AFEXconditions were 1408C, 1:1 ammonia to biomass loading,60% moisture, and 30min residence time. For ‘‘Kramerstover,’’ the AFEX conditions were 1308C, 1:1 ammonia tobiomass loading, 60% moisture, and 15min residence time(Gao et al., 2010). The GLBRC stover was passed through aChristy & Norris 8-inch Lab Mill using a 1-mm screen(Christy Turner, Ltd, Suffolk, United Kingdom). Followingthe initial grinding, the material was passed through a0.5-mm mesh screen. The material retained by the screen(�1/3 of the original) was re-ground in a Wiley Mill(Standard Model #3, Arthur H. Thomas, Philadelphia, PA)using a 0.5-mm screen. The first and second grindings werecombined; none of the original corn stover was discarded.The Kramer stover was milled to <100mm particle size asdescribed (Chundawat et al., 2008; Gao et al., 2010).

Both the GLBRC stover and the Kramer stover had a totalGlu content of 34.4% and Xyl content of 22.4%, asdetermined by complete acid hydrolysis and quantitation ofthe alditol acetate sugar derivatives by gas chromatography(Blakeney et al., 1983).

Sources of Enzymes

The individual enzymes used in this paper were produced inPichia pastoris from native Trichoderma reesei genes, with theexception of CBH1 (Cel7A), which was fromT. longibrachiatum (Megazyme, Ltd, Bray, Ireland). TheGenBank accession numbers for the enzymes used in thisstudy are CBH2 (also known as Cel6A), P07987; EG1(Cel7B), AAA34212; BG, AAA18473; EX3, BAA89465; BX,CAA93248. All of these are abundant proteins in thesecretome of T. reesei RUTC30 grown on AFEX-treated cornstover and in Spezyme CP (Nagendran et al., 2009).

For cloning of the genes, T. reesei was grown in stillculture on corn stover as described (Nagendran et al., 2009).Total RNA was extracted from 12-day mycelium using anRNeasy Plant Mini Kit (Qiagen, Valencia, CA). cDNA fromtotal RNA was synthesized using SuperScript III reversetranscriptase (Invitrogen, Carlsbad, CA). This cDNA wasused as the template to amplify all of the core set genes(except CBH1). Gene-specific primers (SupplementaryTable S1) were designed with linkers to facilitate integrationof the genes into the Pichia pastoris expression vector pPICZ.The constructs were transformed into P. pastoris strain X33following the manufacturer’s instructions (Invitrogen).In an earlier study, a suite of cell-wall-active enzymesfrom Aspergillus nidulans were expressed in P. pastorisusing a combination of the native signal peptide andthe yeast a-factor signal peptide of the vector pPICZa

(Bauer et al., 2006). In contrast, secretion of the enzymes inthis paper was driven only by the native T. reesei signalpeptide.

P. pastoris transformants were selected on plates contain-ing YPDS (1.34% yeast nitrogen base, 1% glycerol, 2% Glu,1M sorbitol, 2% agar)þ 100mg/mL zeocin (phleomycin).Positive clones were confirmed by colony PCR using gene-specific primers, streaked onto plates containing YPDS þzeocin, and incubated at 308C for 3 days. Single colonieswere grown in 250-mL baffle flasks (Bellco Biotechnology,Vineland, NJ) containing 50mL BMGY medium (100mMpotassium phosphate, pH 6.0, 1.34% yeast nitrogen base, 2%peptone, 1% yeast extract, 0.4mg/mL biotin, 1% glycerol).The culture was incubated 16 h at 308C with shaking at250 rpm. The entire culture was used to inoculate 450mLBMGYmedium which was then grown with shaking in a 2-Lbaffle flask at 250 rpm for 3 days. On the fourth day, theculture was collected in 500-mL sterile centrifuge bottles andcentrifuged for 5min at 1,500g. The pellet was washed with100mL of 1M sorbitol and recentrifuged for 5min at 1,500g.The pellet was resuspended in 100mL BMM medium(100mM potassium phosphate, pH 6.0, 1.34% yeastnitrogen base, 0.4mg/mL biotin) and transferred to a 2-Lbaffle flask. The flask was incubated at 308C for 4 days withshaking at 250 rpm. To induce protein expression, 500mL ofmethanol was added every 24 h starting at time zero.

Each enzyme was produced in batches of 300mL (i.e.,three 2-L flasks each containing 100mL). The contents ofthe flasks were combined in one 500-mL centrifuge bottleand centrifuged for 5min at 1,500g. The supernatants wereconcentrated to 100mL using a tangential flow filtrationsystem with a 10 kDa-cutoff membrane (Vivaflow,Sartorius, Bohemia, NY), buffer exchanged with fivevolumes (500mL) of 25mM sodium acetate, pH 5.0, andthen concentrated further to 25mL, all in the samefiltration system. After concentration, sterile glycerol wasadded from an 80% (v/v) stock to a final concentration of20% (v/v). Concentrated desalted enzymes were aliquotedinto 1.5mL tubes (500mL per tube) and stored at �808C.Final stock enzyme concentrations ranged from 0.2 to1mg/mL.

Protein concentrations were determined by thebicinchoninic acid (BCA) method (Pierce Biotechnology,Rockford, IL) using bovine serum albumin as standard(Smith et al., 1985). Background protein concentrationsin supernatants of P. pastoris transformed with emptyvector and grown under identical conditions weresubtracted.

Spezyme CP (lot number 301-05330-206), MultifectXylanase (lot number 4900667792), and Accellerase 1000(lot number 1600844643) were a gift of Genencor, Inc.(Rochester, NY) Novozyme 188 (catalog C6105; batchnumber 027K0698) was purchased from Sigma–Aldrich (St.Louis, MO). Protein concentrations for the preparationsused in this paper were 84mg/mL for Accellerase 1000,123mg/mL for Spezyme CP, 85 mg/mL for Novozyme 188,and 35mg/mL for Multifect Xylanase.

Enzymatic Hydrolysis

Design-Expert1 software (Version 7.0 or 8.0; Stat-Ease, Inc.,Minneapolis, MN; http://www.statease.com) was used tocreate the simplex-lattice designs and to analyze responses.Most experiments were performed as mixtures at fixedprotein loadings per g glucan in the corn stover. The numberof mixtures in the simplex-lattice depended on both thenumber of components in the mixture and the degree of thepolynomial. Using an augmented special cubic model, the 3-component experiment had 13 separate assay mixtures, thefive-component experiment had 41, and the six-componentmixture had 63. All assays were replicated. Based onpreliminary experiments (see Results Section), a minimumenzyme proportion of 5% was used as a lower limitconstraint in all experiments, which thereby automaticallydetermined the upper limit for each enzyme.

All assays were performed on a Biomek FXP LaboratoryAutomation Workstation equipped with dual arms (Span 8with disposable tip module and a fixed AP96 module)(Beckman-Coulter, Fullerton, CA). Tip-touch functionalitywas used throughout.

AFEX-treated corn stover (‘‘GLBRC stover’’), ground to0.5mm particle size, was suspended at a final concentrationof 1.47% total biomass (equivalent to 0.5% glucan) in50mM sodium citrate, pH 4.7, containing 5mg/mLtetracycline and 5mg/mL cycloheximide. The slurry waskept in suspension using a paddle reservoir designed fordispensing pharmaceutical beads on the Biomek FXP(Model VP 756C-1P100, V&P Scientific, Inc., San Diego,CA). Throttle speed was set to 85. At the start of eachexperiment, the reservoir was filled to a volume of 100mLand replenished when the level fell below 70mL. The last5mm of the Span 8–1,000mL tips were cut off in order topipet the stover slurry. Prior to aspirating, 200mL of theslurry was mixed once at 100mL/sec and then the samevolume was aspirated 19mm below the liquid level anddispensed into 1.5-mL deep-well plates (Abgene, FisherScientific, Epson, United Kingdom) 2mm from the bottomof the wells. The final glucan loading was 0.2% (2.9mg totalbiomass) in a reaction volume of 500mL.

Accuracy and precision of the biomass dispensing wastested by drying and weighing a series of dispensed aliquots,each of 0.2mL (nominally containing 2.90mg). Themean� 1 SD was 3.00� 0.1mg (n¼ 11).

Worksheets specifying source labware, source wells,destination labware, destination wells, and transfer volumeswere generated in Microsoft Excel and imported into theBiomek FXP software, version 3.3, using the transfer-from-file function. Water was dispensed 4mm below the liquidlevel. Enzymes were aspirated from 1.5-mL centrifuge tubes4mm below the liquid level and dispensed into the 96-wellplates 3mm below the liquid level.

On completion of the dispensing program, the plates werevacuum-sealed with pierceable capmats (USA Scientific,Ocala, FL), and incubated at 508C for the indicated time(either 24 or 48 h) in a hybridization oven rotating at 10 rpm

Banerjee et al.: Biomass Enzyme Optimization 709

Biotechnology and Bioengineering

(Model 5420, VWR). The plates were then centrifuged at1,500g for 3min in a swinging bucket centrifuge (Eppendorf,Hamburg, Germany, Model 5417R) to separate the solidresidue from the digested biomass. The supernatants(100mL) were transferred into Costar 96-well plates(Abgene, Fisher Scientific) using the AP96 pod of theBiomek and then heated at 1008C for 10min to inactivatethe enzymes. Each enzyme combination was run induplicate, sampled twice, and the supernatants assayedtwice for Glu and Xyl. Therefore, n¼ 8 for all assays.

Glu and Xyl Assays

Free Glu and Xyl were assayed colorimetrically using kitssupplied by Megazyme, Ltd (catalog K-GLUC and K-XYLOSE, respectively). Sugar assays were done in 384-wellplates (BD Falcon, BD Biosciences, San Jose, CA) using64mL of assay reagent and 4mL of sample. The plates wereincubated at 508C for 20min (Santoro et al., 2010). Theadvantages of using enzyme-coupled assays to measure freeGlu and Xyl include higher specificity compared to reducingsugars assays (King et al., 2009), speed compared to HPLCor gas chromatography (Berlin et al., 2006), and adaptabilityto multiwell plates.

Data Analysis

Absorbance values from the Glu and Xyl assays wereconverted to mol of Glu and Xyl and then to percenttheoretical yields, which values then served as the responsesin the experimental design. For all experiments, we show theANOVA calculations of F-value, P-value, R2, Adjusted R2,Predicted R2, and Adequate Precision as calculated by theDesign-Expert software. The F-value determines the effects,if any, of the individual components on the model. AnF-value close to 1 implies that the components of themixture do not interact and hence have little effect on themodel. The P-value tests whether the F-value determined bythe model is significantly different from background; aP-value <0.05 is considered to be statistically significant. R2

is the measure of the amount of variation around the meanexplained by the model, whereas Adjusted R2 compensatesfor the number of terms or components in the mixture.Predicted R2 is a measure of the amount of variation in newdata explained by the model. Adjusted R2 and Predicted R2

should be within 0.2 of each other for the model to besignificant. A difference>0.2 implies a greater than expectedinability to predict new data, often due to model over-parameterization, in which case we chose a different order ofpolynomial and/or a backward or stepwise eliminationregression with an ‘‘alpha out’’ value set to 0.1. This helpedeliminate all of the terms in the model that wereinsignificant, that is, P> 0.05. The best model that gave adifference between the Adjusted and Predicted R2 values<0.2 was used to navigate the design space. AdequatePrecision estimates the signal-to-noise ratio, a value


>4 indicating adequate model discrimination. Once allthe criteria for a robust model were fulfilled, the model wasused to determine the enzyme mixtures that generate theoptimal Glu/Xyl release (Anderson and Whitcomb, 2004).

Enzyme Assays

Glycosidase activities were assayed using 4-nitrophenyl(pNP) sugar derivatives (Sigma, St. Louis, MO). Enzyme in avolume of 2–25mL was mixed with 100mL 100mM sodiumacetate, pH 5, and water was added to a final volume of300mL. p-NP substrate (10mM in water) was added and thereactions incubated 20–30min at 508C. The reaction wasterminated by addition of 600mL of 1M sodium carbonateand the absorbance read at 400 nm. An extinction coefficientof 1.3 AUmol�1mL was used (Chaplin and Kennedy, 1986).

Stock solutions of polysaccharides (1% birchwoodxylan, 1% carboxymethyl [CM]-cellulose, and 0.5% barleyb-glucan) (Sigma) were made in water. Polysaccharide stock(100mL) plus enzyme (10–25mL) plus water to 300mL wereincubated for 60min at 378C. An aliquot (25mL) was mixedwith 1.5mL p-hydroxybenzoic acid hydrazide workingsolution. After heating for 10min at 1008C, the absorbancewas measured at 410 nm (Lever, 1972). Time zero valueswere subtracted from the 60min values. Glu was used as thestandard.

Results

Performance of Commercial Enzymes on GENPLAT

The assay platform was tested using standard commercialenzyme mixtures: Spezyme CP with and without supple-mentation with Novozyme 188, and Accellerase 1000 withand without supplementation with Multifect Xylanase.These experiments allowed comparisons to be madebetween GENPLAT and previous automated enzyme assays,and also provided a benchmark against which to evaluatesynthetic enzyme mixtures subsequently developed. Resultswith commercial enzymes for 48 h digestion are shown inTable I. In this and other experiments, yields at 24 h werelower than at 48 h but the predicted and experimentallydetermined optimal ratios were similar (SupplementaryTable S2). The average coefficient of variance for Glu was0.8% and for Xyl 0.3%, indicating that GENPLAT has anacceptable level of precision. Maximum Glu and Xyl release(64.3% and 44.7%, respectively) were achieved with 15mgAccellerase 1000þ 15mg Multifect Xylanase, but severalother combinations were almost as effective (Table I). Amixture of Spezyme CP and Novozyme 188 was slightlysuperior to Spezyme CP alone for Glu (1–4% improve-ment), but made a larger difference for Xyl (5–10%improvement). Novozyme 188 alone was ineffective(releasing a maximum of 7.6% and 9.6% of Glu and Xyl,respectively). Multifect Xylanase by itself was clearly

Table I. Digestion of AFEX-pretreated corn stover with commercial enzymes, 48 h digestion.

Spezyme CP Novozyme 188 Accellerase 1000 Multifect-xylanase Total protein concentration % Glucan conversion % Xylan conversion

0 0 0 0 0 0.78� 0.3 0.7� 0.2

7.5 0 0 0 7.5 35.5� 0.8 26.4� 0.1

0 7.5 0 0 7.5 6.1� 0.2 4.3� 0.3

3.75 3.75 0 0 7.5 42.0� 0.9 26.5� 0.5

15 0 0 0 15 48.8� 0.1 28.8� 0.4

0 15 0 0 15 6.9� 0.4 6.7� 0.1

7.5 7.5 0 0 15 49.3� 0.7 32.1� 0.0

30 0 0 0 30 56.5� 0.9 34.6� 0.8

0 30 0 0 30 7.6� 0.1 9.6� 0.1

15 15 0 0 30 57.4� 3.0 38.6� 0.5

0 0 7.5 0 7.5 44.5� 1.8 20.3� 0.2

0 0 0 7.5 7.5 7.8� 0.5 34.3� 0.3

0 0 3.75 3.75 7.5 48.0� 2.9 39.3� 0.3

0 0 15 0 15 50.5� 1.5 28.8� 0.2

0 0 0 15 15 7.8� 0.1 37.6� 1.0

0 0 7.5 7.5 15 52.8� 0.8 40.6� 0.9

0 0 30 0 30 54.3� 1.7 32.7� 0.3

0 0 0 30 30 9.1� 0.3 42.9� 0.2

0 0 15 15 30 64.3� 2.5 44.7� 1.1

Enzyme concentrations are expressed asmg protein/g total stover glucan. ‘‘% conversion’’ indicates free Glu and Xyl present in the reaction supernatants atthe end of the incubation, as a percentage of the total Glu and Xyl content of the input stover. Results are expressed as the mean� 1 SD (n¼ 8). For 24 hdigestion results see Supplementary Table S2.

superior (increase of 9–14%) to Accellerase alone for Xylrelease at all concentrations tested. Addition of MultifectXylanase to Accellerase 1000 had a modest effect on Gluyields (1–10%) but a larger effect on Xyl yield (12–19%improvement). These results indicate that Accellerase 1000is sub-optimal for some critical activity or activities suppliedby Multifect Xylanase, possibly EX itself. This might berelated to the fact that Accellerase 1000 has been developedfor acid-pretreated corn stover, whereas alkaline pretreat-ments such as AFEX, by leaving intact the majority of thehemicellulose, generate material with higher Xyl levels(Kumar and Wyman, 2009).

Figure 1. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-

PAGE) of the core set of enzymes. These samples are from the same batches of

enzymes used in all of the experiments described in this paper. Three micrograms

were loaded in each lane. Lane 1: CBH1; Lane 2: BG; Lane 3: EG1; Lane 4: BX; Lane 5:

EX3; Lane 6: CBH2. M, molecular weight markers, indicated in kilodaltons (kDa)

(Precision-Plus, Bio-Rad, Richmond, CA).

Expression of Core Enzymes

As a starting point for the development of a completelysynthetic enzyme mixture, a core set should include thoseenzymes required for a minimum reasonable release of Gluand Xyl from a given biomass. For cellulose, this stipulates atleast one cellobiohydrolase (CBH), one endo-b1,4-gluca-nase (EG), and one b-glucosidase (BG). For hemicelluloses,in which Xyl is the dominant sugar, a core set requires anendo-b1,4-xylanase (EX) and a b-xylosidase (BX). T. reesei,like many fungi, makes multiple enzymes with theseactivities. The enzymes used in this paper are among themost abundant in the secretome of T. reesei RUTC30 grownon corn stover and in Spezyme CP (Nagendran et al., 2009).The enzymes were produced by expression in P. pastoris,except for CBH1. This enzyme did not express well inP. pastoris and instead a commercially available enzymefrom T. longibrachiatum, a species closely related to T. reesei,was used (Fig. 1). Analysis of the specific activities of the six

enzymes on a variety of substrates indicated that theactivities were comparable to those found by others for thenative enzymes, and that there was no evidence ofcontaminating activities (Table II). Most importantly,CBH1 had no detectable activity on CMC-cellulose, whichwould indicate EG contamination (Table II). The activity ofEX3 on pNP-cellobioside is consistent with published results(see below).

The molecular weights (MWs) by SDS-PAGE of theP. pastoris-expressed enzymes were within 10% of thetheoretical values predicted from the gene sequences (i.e.,without the signal peptides or any post-translational



Table II. Specific activities of enzymes.

Substrates CBH1 EG1 BG EX3 BX CBH2

pNP-cellobioside 0.007 2.50 7.7 16.0 0.0 0.0

pNP-glucoside 0.0 0.0 245.3 0.0 3.0 0.0

pNP-xyloside 0.0 0.0 0.63 0.0 76.8 0.0

Birchwood xylan 0.0 1.99 0.0 12.7 0.0 0.0

CM-cellulose 0.0 2.36 0.0 0.0 0.0 0.0

Barley b-glucan 0.0 4.13 0.0 0.0 0.0 0.0

The same batches of enzymes were used in all of the experimentsdescribed in this paper. pNP, p-nitrophenyl. Units for pNP derivativesaremmol pNP/mg enzyme/min. Units for the other substrates aremmol Gluequivalent/mg enzyme/min.

modifications), with the exception of EG1 and EX3. EG1 hasa predicted MW of 46 kDa but the protein expressed inP. pastoris had an apparent MW of �60 kDa (Fig. 1). Thesame protein expressed in Saccharomyces cerevisiae had aMW of�97 kDa due to hyperglycosylation (Lu et al., 2009),and hyperglycosylation might also account for the higherMW of EG1 in P. pastoris. EX3 expressed in P. pastoris ran asa tight doublet with apparent MW’s of 30–32 kDa (Fig. 1)whereas the predicted MW is 36 kDa. However, Xu et al.(1998) reported that the native protein (which is notglycosylated) behaved as a 32-kDa protein on SDS-PAGE.This is close to the value for the P. pastoris-expressed protein(Fig. 1).Why EX3 expressed in P. pastoris runs as a doublet isnot clear, but might be caused by differential glycosylationor some other post-translational modification. Ion exchangechromatography resolved the doublet into two peaks, bothof which had xylanase activity (data not shown).

Initial Studies With Pure Enzymes

A five-enzyme core set (comprising CBH1, EG1, BG, EX3,and BX) was first tested in a full-factorial experiment at zero,low (0.8mg/0.5mL), and high (3.2mg/0.5mL) concentra-tions, for a total sample size of 243 (35). In contrast to thesubsequent experiments discussed below, this is not amixture experiment. Figure 2 presents the Glu yield as afunction of total protein loading, which ranged from zero to16mg protein/g glucan. The maximum Glu yield at anyparticular protein loading is indicated by the dashed lines,and the average Glu yield is indicated by the solid lines.From this study, several preliminary conclusions could bedrawn. First, even at the highest concentration tested in thisexperiment, CBH1 was sub-optimal. Second, an

Figure 2. Percent glucan conversion (percentage of the total glucan in the

samples converted to free Glu) as a function of total enzyme concentration. The

experiment was a full-factorial design using five enzymes each at three concentra-

tions (zero, medium, and high). The 81 different combinations of all of the other four

core enzymes are plotted for each individual enzyme concentration. The dashed lines

connect the maximum glucan conversion obtained at each total enzyme loading. The

solid lines represent third-order polynomial regression curves calculated using

Microsoft Excel.


intermediate concentration of EG was close to optimal.Third, an intermediate level of BG was necessary andadequate. Fourth, EX weakly promoted Glu release. Fifth,BX made almost no difference at the low concentration and

was slightly inhibitory at the high concentration. Theseresults suggest that for optimum Glu release one shouldincrease CBH1 concentration, maintain BG and EG atmoderate levels, and include a small amount of EX and BX.These conclusions formed a preliminary basis for sub-sequent experiments performed at fixed protein loadings.

Construction and Optimization of a Six-ComponentSynthetic Enzyme Set

The three core cellulose-degrading enzymes (CBH1, EG1,and BG) were initially optimized by themselves on AFEX-pretreated stover using a mixture experimental design. Thisexperiment provided a benchmark for cellulases in theabsence of xylanases or accessory enzymes. On the basis ofthe experiment shown in Figure 1, the minimum proportionof any one component was fixed at 5%, which automaticallyestablished the maximum concentration of the othercomponents, for example, in a five-component mixtureno one component exceeded 80%. The experimental resultsfor the three-component mixtures at 48 h are given inTable III. The results for 24 h digestion, and the ternary

Table III. Three-component results (CBH1 or CBH2, EG1, BG), 48 h

digestion.

Enzyme proportion 48 h

CBH1 CBH2 EG1 BG % Glucan conversion % Xylan conversion

0.62 0 0.05 0.33 22.7� 0.6 6.2� 1.2

0.62 0 0.33 0.05 28.7� 1.8 3.6� 0.5

0.33 0 0.05 0.62 19.3� 0.9 6.8� 1.1

0.05 0 0.90 0.05 18.4� 0.8 4.2� 0.9

0.90 0 0.05 0.05 24.3� 1.9 5.1� 1.6

0.19 0 0.62 0.19 24.9� 1.7 4.1� 1.6

0.05 0 0.62 0.33 19.1� 1.6 5.0� 0.4

0.62 0 0.19 0.19 29.4� 1.3 6.7� 1.1

0.33 0 0.62 0.05 26.5� 0.9 4.0� 1.2

0.19 0 0.19 0.62 21.5� 1.4 6.2� 1.1

0.05 0 0.05 0.90 13.8� 0.2 4.8� 1.5

0.33 0 0.33 0.33 26.5� 2.0 5.7� 1.4

0.05 0 0.33 0.62 18.0� 0.4 7.2� 1.5

0 0.62 0.05 0.33 13.3� 0.6 3.2� 0.2

0 0.62 0.33 0.05 15.1� 0.7 2.3� 0.4

0 0.33 0.05 0.62 13.4� 0.5 4.9� 0.8

0 0.05 0.90 0.05 14.9� 0.7 4.5� 1.6

0 0.90 0.05 0.05 12.1� 0.6 2.5� 1.0

0 0.19 0.62 0.19 17.3� 0.5 0.2� 0.0

0 0.05 0.62 0.33 8.6� 0.2 5.5� 0.4

0 0.62 0.19 0.19 15.0� 0.9 1.7� 1.3

0 0.33 0.62 0.05 17.6� 1.9 4.1� 1.6

0 0.19 0.19 0.62 8.8� 0.3 5.6� 0.9

0 0.05 0.05 0.90 7.3� 0.4 4.1� 1.5

0 0.33 0.33 0.33 16.7� 0.9 3.5� 0.9

0 0.05 0.33 0.62 14.3� 0.5 7.3� 0.1

‘‘% conversion’’ indicates the percentage of the total glucan or xylanconverted to free Glu or Xyl, respectively, expressed as the mean� 1 SD(n¼ 8). Twenty-four hours digestion results are shown in Table S3. Allprotein loadings were 15mg/g glucan. Ternary diagrams and statisticalanalyses are shown in Figure S1.

diagrams and statistical analyses of the models for 24 and48 h, are given in Supplementary Table S3 and Figure S1. At48 h digestion, a mixture of 57% CBH1, 38% EG1, and 5%BG gave the highest predicted yield, which was 29.4% of thetotal Glu content (Fig. S1). The optimum proportions at24 h were similar (Fig. S1). In the absence of EX and BX, Xylyields were too low (less than 8%) to predict a statisticallysignificantmodel for xylan conversion to free Xyl (Table III).

T. reesei makes two cellobiohydrolases, CBH1 (Cel7A)and CBH2 (Cel6A). In a three-component mixture contain-ing CBH2 instead of CBH1, CBH2 did contribute to Glurelease but was inferior to CBH1 (Tables III and S3). Acombination of CBH1 and CBH2 was later shown to besuperior to either alone (see below).

Addition of EX3 and BX to CBH1, EG1, and BG (a five-component mixture) caused a slight increase in the maximalGlu yield at 48 h (from 29% to 31%) but a large increase inXyl yields (from 3% to 22%) (Table IV). Experimentalresults for the five-component digestions for 24 and 48 h areshown in Tables S4 and S5, respectively. When Glu releasewas chosen to be optimized, addition of these two enzymescaused a shift in the optimum proportion of CBH1 from57% to 38% without altering the proportions of EG1 or BG(Table IV and Fig. S2). This might be related to the fact thatEX3, like CBH1, is active on pNP-cellobiose (Table II). Theenzyme mixture optimized for Xyl release at 48 h contained5%CBH1, 24% EG1, 5% BG, 55% EX3, and 11% BX, that is,EX3 replaced CBH1 as the major component when themixture was optimized for Xyl instead of Glu. A mixturemodel optimized to release a balance of Glu and Xylcontained intermediate proportions of CBH1 and EX3(Table IV).

Addition of CBH2 to the other five enzymes (a six-component mixture) resulted in enhanced Glu yield at 48 hfor the same protein loading; Glu yield showed a significantincrease from 30.8% to 38.7% (Tables IV and V). The effectof CBH2 on Xyl yield was small but statistically significant,from 23.3% to 26.4% (Table IV). Thus, although CBH2cannot substitute for CBH1, a mixture of the two is superiorto either one alone. Results for the six-component assays at48 h are shown in Tables IV and V, six-component results at24 h are shown in Tables S6 and S7, and ternary diagrams at48 and 24 h are shown in Figures 3 and S3, respectively.

Optimizing for Glu instead of Xyl had only a small effecton Xyl yields in either the five-component or the six-component experiment (a decrease of 0.9–3.1%) (Table IV).On the other hand, there was a large Glu penalty foroptimizing for Xyl in the five-component experiment (adecrease of 10.4%) but a lower penalty in the six-componentexperiment (a decrease of 3.4%).

There were several trends going from three componentsto six (Table IV). First, the relative proportion of CBH1declined. This might be because the activities of CBH2 andEX3 overlap with CBH1. Second, the optimum proportionof BG was not constant, but varied from 5% (the minimumtested) to 14% in the six component mixture optimized forGlu. This might reflect the fact that as glucan conversion



Table IV. Model predictions and experimental (‘‘Exptl.’’) values for three, five, or six component mixtures optimized for Glu alone, Xyl alone, or GluþXyl.

Number of

compo-nents

Model

optimized for

Enzyme proportions % Glucan conversion % Xylan conversion

CBH1 EG1 BG EX3 BX CBH2

Model

prediction Exptl.

Model

prediction Exptl.

3 Glu 0.57 0.38 0.05 — — — 29.0 28.7� 0.4 — 3.1� 0.8

5 Glu 0.38 0.38 0.05 0.13 0.05 — 30.8 30.8� 1.6 — 22.4� 1.2

5 Xyl 0.05 0.24 0.05 0.55 0.11 — — 20.4� 0.6 25.0 23.3� 0.5

5 GluþXyl 0.33 0.30 0.05 0.27 0.05 — 30.2 31.6� 0.6 23.0 22.9� 0.8

6 Glu 0.29 0.25 0.14 0.22 0.05 0.05 40.0 38.7� 0.5 — 23.3� 2.0

6 Xyl 0.05 0.25 0.05 0.34 0.17 0.14 — 35.3� 0.3 27.8 26.4� 0.5

6 GluþXyl 0.22 0.22 0.15 0.30 0.06 0.05 39.0 39.9� 1.0 25.7 24.8� 0.6

All protein loadings were 15mg/g glucan and all digestions were for 48 h. The model predictions for maximum Glu and Xyl release are derived from thedata shown in Tables III and V and Table S5. The experimental values are the actual Glu and Xyl release observed using the model prediction. Xyl release fromthe three-component mixture was too low to produce a statistically significant model for Xyl or GluþXyl release. Twenty-four hours results are shown inSupplementary Table S6.

rises, more BG is needed to hydrolyze the cellobioseproduced by the combined action of CBH1, EG1, EX3, andCBH2. It also illustrates the importance of not omitting anycomponent from the experimental design, as is sometimesdone to simplify the experimental design (Gao et al., 2010).Third, the optimized mixtures for Xyl release can be quitedifferent from the optimum mixtures for Glu. EX3 and BXare relatively less important for Glu release than CBH1 andEG1, whereas CBH1 is relatively less important than EX3 orCBH2 for Xyl release. This is consistent with the knownactivities of these enzymes, that is, Glu release requiresenzymes active on glucan and Xyl release requires enzymesactive on xylan. Fourth, of all combinations tested, the six-component mixtures were superior to three- or five-component mixtures for Glu. A six-component mixtureoptimized for Glu or GluþXyl gave the highest yields of Glu(38.7% and 39.9%, respectively; the difference between thetwo was statistically insignificant). For Xyl, the best yieldscame from a six-component mixture optimized for Xyl(26.4% Xyl yield), although the Xyl differences were small(total range 4%) for five-component and six-componentmixtures optimized for Glu, Xyl, or GluþXyl.

The results for Glu in the three-, five-, and six-componentmixtures, and for Xyl in the five- and six-componentmixtures were plotted as ternary diagrams (Figs. 3, S1, S2,and S3). Accompanying the diagrams are the statisticalanalyses of the mixtures; the statistical parameters indicatedthat all of these models are highly significant (P< 0.0001).The ANOVA results for the six-component experiment areshown in Table S8.

Comparison of Two Biomass Feedstocks Differing inParticle Size

While the current work was in progress, Gao et al. (2010)published an optimization of a six-component enzymemixture that contains activities that are largely similar (butnot identical – see Discussion Section) to those used here,and they also derived similar optimal proportions (see


Discussion Section). Many conditions were similar betweenthe two studies, for example, both laboratories used AFEX-treated corn stover, glucan loadings of 0.2% delivered asslurries, incubations at 508C for 24 h (although 48 h is ourstandard condition), and measurement of Glu and Xyl byenzyme-coupled assays.

Despite the similarities in experimental protocols, theoptimized six-component mixture (at a protein loading of16.5mg/gm glucan) of Gao et al. (2010) yielded 67% ofmaximum Glu and 48% of maximum Xyl, whereas theoptimal enzyme mixtures in this study yielded only 39% Gluand 26% Xyl (Table IV). Of several differences in protocols,a potentially significant one was the comparative particlesizes of the stovers used in the two sets of experiments. Gaoet al. (2010) used stover (‘‘Kramer stover’’) that had beenground to <100mm particle size (equivalent to 140–170mesh), whereas our stover (‘‘GLBRC stover’’) wasground to <0.5mm particle size (35 mesh). Below a certainthreshold, more finely ground biomass is known to be moredigestible (Decker et al., 2009). To test whether this couldaccount for the differences in apparent digestibility, weobtained a sample of Kramer stover from the laboratory ofBruce Dale (Department of Chemical Engineering, MSU)and compared the two in parallel on GENPLAT usingvarious concentrations of two commercial enzymes. Tomake the comparisons more precise, we included the sameenzyme loadings used by Gao et al. (16.5mg/g glucaninstead of our normal 15mg/g glucan) and measured Glurelease after 24 h as well as 48 h. As shown in Table VI, glucanconversion was up to 31% higher for Kramer compared toGLBRC stover. This trend was seen for both 24 and 48 hdigestions.

Discussion

The GENPLAT Enzyme Optimization Platform

In this paper, we describe the development, validation, andapplication of a highly automated platform, named

Table V. Six-component results, 48 h digestion.

Enzyme proportion 48 h

CBH1 EG1 BG EX3 BX CBH2 % Glucan conversion % Xylan conversion

0.75 0.05 0.05 0.05 0.05 0.05 28.5� 0.2 17.2� 0.20.05 0.05 0.75 0.05 0.05 0.05 21.2� 0.2 15.4� 0.10.05 0.75 0.05 0.05 0.05 0.05 26.3� 0.2 17.0� 0.50.05 0.05 0.05 0.05 0.75 0.05 19.1� 0.2 15.4� 1.60.05 0.05 0.05 0.75 0.05 0.05 23.8� 0.5 25.5� 0.00.05 0.05 0.05 0.05 0.05 0.75 23.1� 0.0 8.9� 0.00.52 0.05 0.28 0.05 0.05 0.05 28.0� 0.0 15.5� 0.00.52 0.28 0.05 0.05 0.05 0.05 30.9� 0.1 16.0� 0.00.52 0.05 0.05 0.05 0.28 0.05 27.5� 0.1 16.1� 0.00.52 0.05 0.05 0.28 0.05 0.05 32.4� 0.2 24.1� 1.20.52 0.05 0.05 0.05 0.05 0.28 29.2� 0.2 17.8� 0.70.28 0.05 0.52 0.05 0.05 0.05 28.6� 0.0 19.9� 0.70.28 0.28 0.28 0.05 0.05 0.05 36.3� 0.1 21.6� 0.80.28 0.05 0.28 0.05 0.28 0.05 29.1� 0.8 18.7� 0.00.28 0.05 0.28 0.28 0.05 0.05 35.0� 0.2 22.4� 1.40.28 0.05 0.28 0.05 0.05 0.28 31.3� 0.4 14.8� 0.00.28 0.52 0.05 0.05 0.05 0.05 35.9� 0.1 19.6� 0.50.28 0.28 0.05 0.05 0.28 0.05 32.6� 0.9 21.0� 1.40.28 0.28 0.05 0.28 0.05 0.05 37.6� 0.2 23.1� 0.70.28 0.28 0.05 0.05 0.05 0.28 34.7� 1.0 17.8� 0.90.28 0.05 0.05 0.05 0.52 0.05 29.0� 0.5 18.0� 0.00.28 0.05 0.05 0.28 0.28 0.05 32.3� 0.7 25.8� 1.80.28 0.05 0.05 0.05 0.28 0.28 29.9� 0.2 22.2� 1.50.28 0.05 0.05 0.52 0.05 0.05 33.8� 0.2 22.8� 0.00.28 0.05 0.05 0.28 0.05 0.28 33.2� 0.2 20.7� 0.10.28 0.05 0.05 0.05 0.05 0.52 28.9� 1.0 14.8� 0.10.05 0.28 0.52 0.05 0.05 0.05 28.7� 0.2 18.8� 0.90.05 0.05 0.52 0.05 0.28 0.05 23.2� 0.3 17.0� 0.00.05 0.05 0.52 0.28 0.05 0.05 25.1� 0.5 20.3� 0.60.05 0.05 0.52 0.05 0.05 0.28 25.7� 0.4 15.5� 0.20.05 0.52 0.28 0.05 0.05 0.05 31.5� 0.1 18.5� 1.10.05 0.28 0.28 0.05 0.28 0.05 28.8� 0.7 19.3� 2.50.05 0.28 0.28 0.28 0.05 0.05 32.1� 1.7 23.8� 1.10.05 0.28 0.28 0.05 0.05 0.28 31.8� 1.3 19.5� 1.20.05 0.05 0.28 0.05 0.52 0.05 22.4� 0.4 22.2� 2.00.05 0.05 0.28 0.28 0.28 0.05 26.7� 0.2 23.5� 0.00.05 0.05 0.28 0.05 0.28 0.28 26.2� 0.3 20.2� 0.70.05 0.05 0.28 0.52 0.05 0.05 27.8� 0.5 23.4� 1.00.05 0.05 0.28 0.28 0.05 0.28 29.8� 0.0 17.9� 0.00.05 0.05 0.28 0.05 0.05 0.52 27.3� 1.6 12.9� 0.00.05 0.52 0.05 0.05 0.28 0.05 29.1� 0.5 20.8� 0.00.05 0.52 0.05 0.28 0.05 0.05 33.3� 0.5 22.1� 0.00.05 0.52 0.05 0.05 0.05 0.28 32.3� 2.0 17.4� 0.50.05 0.28 0.05 0.05 0.52 0.05 27.6� 0.7 22.1� 0.00.05 0.28 0.05 0.28 0.28 0.05 28.6� 0.0 25.1� 0.30.05 0.28 0.05 0.05 0.28 0.28 29.8� 0.9 24.2� 0.00.05 0.28 0.05 0.52 0.05 0.05 31.5� 0.3 28.0� 2.00.05 0.28 0.05 0.28 0.05 0.28 34.0� 0.7 24.3� 2.10.05 0.28 0.05 0.05 0.05 0.52 27.4� 0.0 15.6� 0.00.05 0.05 0.05 0.28 0.52 0.05 23.8� 0.2 23.7� 0.60.05 0.05 0.05 0.05 0.52 0.28 25.4� 0.8 19.9� 0.00.05 0.05 0.05 0.52 0.28 0.05 26.2� 0.3 25.8� 7.50.05 0.05 0.05 0.28 0.28 0.28 27.9� 0.1 23.2� 4.30.05 0.05 0.05 0.05 0.28 0.52 23.7� 1.1 17.9� 0.00.05 0.05 0.05 0.52 0.05 0.28 28.1� 0.8 19.5� 0.00.05 0.05 0.05 0.28 0.05 0.52 26.6� 0.0 14.9� 0.00.46 0.11 0.11 0.11 0.11 0.11 39.3� 0.9 22.8� 0.00.11 0.11 0.46 0.11 0.11 0.11 33.0� 0.2 23.4� 0.10.11 0.46 0.11 0.11 0.11 0.11 34.9� 0.2 25.0� 1.60.11 0.11 0.11 0.11 0.46 0.11 31.5� 0.4 25.5� 0.60.11 0.11 0.11 0.46 0.11 0.11 36.1� 0.1 27.4� 0.50.11 0.11 0.11 0.11 0.11 0.46 32.5� 1.2 20.7� 0.20.17 0.17 0.17 0.17 0.17 0.17 36.9� 1.1 26.2� 1.8

All protein loadings were 15mg/g glucan. Twenty-four hours digestion results are shown in Table S7.



Figure 3. Ternary diagrams and statistical analyses of the 48-h digestion results for six-component enzyme mixtures. For the 24-h results, see Figure S3. A: Glucose. Model

prediction: CBH1, 0.29; EG1, 0.25; BG, 0.14; EX3, 0.22, BX, 0.05; CBH2, 0.05. B: Xylose. Model prediction: CBH1, 0.05; EG1, 0.25; BG, 0.05; EX3, 0.34; BX, 0.17; CBH2, 0.14. [Color figure can

be seen in the online version of this article, available at www.interscience.wiley.com.]

GENPLAT, for the development of enzyme mixtures forbiofuel applications. We combine GENPLAT with the use ofthe heterologous host Pichia pastoris as a means to producequantities of enzymes that are sufficient for hundreds tothousands of assays and that are free of contaminatingactivities. Key features of GENPLAT are the use of a beadmixing chamber for accurate and rapid pipetting of stoverslurries of � 0.5mm particle size at a concentration of0.5%; the use of 96-deep well plates; automated liquidhandling for dispensing enzymes, slurry, and buffers; gentle


mixing during digestion by end-over-end rotation; andautomated colorimetric determination of Glu and Xyl.Several of these features have been incorporated intoother automated platforms (Berlin et al., 2006; Chundawatet al., 2008; Gao et al., 2010; King et al., 2009). GENPLATcan be used for several types of enzyme studies, suchas evaluating alternative versions of the core enzymes,testing of accessory enzymes, and optimizing enzymemixtures for different pretreatments and biomass feedstockcombinations.

Table VI. Comparative digestibility of Kramer corn stover and GLBRC corn stover by commercial enzymes in GENPLAT, after 24 or 48 h.

Protein loading (mg/g glucan) % Glucan conversion (24 h) % Glucan conversion (48 h)

Spezyme CP Accellerase 1000 Kramer stover GLBRC stover D Kramer stover GLBRC stover D

0 0 2.0� 0.3 0.1� 0.1 0.9 1.3� 0.3 0.9� 0.5 0.4

8.25 0 34.5� 2.5 29.9� 0.6 4.6 59.5� 3.5 42.0� 1.0 17.5

16.5 0 54.5� 2.2 44.4� 0.8 10.1 83.1� 2.8 55.9� 2.6 27.2

33 0 72.6� 3.8 58.2� 1.9 14.4 94.9� 3.0 63.9� 3.0 31.0

0 8.25 44.6� 3.4 35.1� 2.3 9.5 68.6� 3.9 46.6� 1.0 22.0

0 16.5 64.0� 3.5 50.4� 3.0 13.6 83.6� 1.8 55.7� 3.3 27.9

0 33 76.1� 2.8 58.6� 2.5 17.5 92.0� 3.8 61.8� 0.8 30.2

Both stover samples had been pretreated by AFEX before grinding. D, difference in % Glu yield between Kramer and GLBRC corn stover.

Choice of Enzyme Source

There were several justifications for using the enzymes ofT. reesei in initial optimization experiments. One is thatmore of the individual biomass-relevant enzymes of thisfungus have been characterized than of any other organism(Nagendran et al., 2009). Working with enzymes ofestablished biochemical function avoids the potential errorsthat can arise from deducing function by homology, that is,by assuming that amino acid sequence identity equates withcatalytic identity. For example, the genome of Aspergillusnidulans contains 20 genes in Glycosyl Hydrolase (GH)family 3. Bauer et al. (2006) expressed five of these inP. pastoris; three had no recorded activity, two were active onpNP-glucoside, and none were tested against the criticalsubstrate, cellobiose. On the other hand, the GH family 3enzyme of T. reesei used in this paper has been shown tohydrolyze cellobiose (Mach et al., 1995), and to contributeto Glu release from corn stover (Fig. 2). Another, morespeculative, justification for using enzymes from a singleorganism is that they are more likely to be compatible witheach other because they have co-evolved in the same genome(Banerjee et al., 2010).

As the source of our enzymes, we expressed the T. reeseigenes in P. pastoris. This yeast-like organism has been usedto produce numerous active proteins from plants, animalsand fungi, including many secreted fungal enzymes (e.g.,Bauer et al., 2006; Jeya et al., 2009; Li et al., 2009). For ourexperiments, there were several justifications for usingP. pastoris. First, obtaining the necessary enzymes bypurification, even from concentrated commercial sources, islaborious, must be repeated frequently, and meticulousremoval of contaminating activities is essential (Irwin et al.,1993). Second, expression in a heterologous host such asP. pastoris provides high titer (in our hands, 16–83mg/L).This is significant because every individual batch of enzymemust be validated for protein concentration, activity, andlack of interfering activities. One batch of enzyme made in a300-mL culture of P. pastoris is sufficient for hundreds tothousands of assays on GENPLAT. Third, compared toover-expression in another filamentous fungus, includingT. reesei itself, P. pastoris lacks background contaminatingactivities and therefore extensive purification is not

necessary (see Supplementary Table S2 in De Schutteret al., 2009).

The major potential disadvantage of using P. pastoris toproduce heterologous proteins is abnormal glycosylation.There is some evidence that abnormal glycosylation has adetrimental impact on the activity of cellulases (Baker et al.,1998; Boer et al., 2000). Based on large increases inmolecular weight, some proteins expressed in P. pastoris areprobably hyperglycosylated. By this criterion, however, nomore than one of our core six proteins is hyperglycosylated(Fig. 1). Furthermore, it has been shown that ‘‘normal’’glycosylation in T. reesei itself depends on growth conditionsand strain, and therefore if glycosylation affects activity thenone cannot strictly compare the same enzyme producedunder different growth conditions or by different strains ofT. reesei (Harrison et al., 1998; Stals et al., 2004a,b).

An alternative strategy to expressing proteins in P. pastorisis over-expression in T. reesei itself. In this case, however,there is more possibility of contamination by competingactivities. Especially problematic for studying the effects ofindividual component enzymes is contamination of CBHwith EG because of the synergistic effects that a smallamount of one can have on the apparent activity of the other(Irwin et al., 1993). The problem of enzyme purity ofproteins overexpressed in T. reesei can, in principle, beminimized by adding protein tags, such as 6xHis, tofacilitate rapid and rigorous purification, but some proteinsdo not tolerate tags well, not all His-tagged proteins bind tonickel columns, and some untagged proteins do bind. Takentogether, in our opinion the potential problem of abnormalglycosylation is outweighed by the other advantages of usingP. pastoris expression as a source of enzymes for GENPLAT.

Design of Experiment

Optimization of enzyme cocktails requires statistically validexperimental design. We used a program called Design-Expert, which is rigorous and user-friendly. One attractivefeature of Design-Expert is that it allows optimization formultiple outcomes, in our case Glu, Xyl, or GluþXyl. Inthis paper our results for GluþXyl optimization gave equalweight to each, but it is also possible to weight one or the



other, depending on the desired outcome. For example,considering that Glu is a preferred substrate for fermentativeorganisms such as S. cereviseae (Lau and Dale, 2009), itmight be preferable to weight the optimization programtoward release of Glu.

Comparison of GENPLAT With Other OptimizationPlatforms

There have been a number of efforts to automate one ormore of the steps in enzyme biomass digestion with the goalof facilitating the discovery of better enzymes and enzymemixtures. Past solutions to the problem of uniformlydispensing insoluble biomass have included paper hand-sheets, solids-dispensing robots, and slurry pipeting (Berlinet al., 2006; Gao et al., 2010; Santoro et al., 2010). We foundthat it is possible to rapidly and accurately pipet slurries upto 0.5% concentration from a mixing chamber designed tomaintain pharmaceutical beads in suspension.

While the current work was in progress, our colleagues inthe Great Lakes Bioenergy Research Center published theresults of a six-enzyme optimization (Gao et al., 2010), basedon the high-throughput platform of Chundawat et al.(2008). There are some instructive similarities anddifferences between the conclusions of the two studies.The enzyme proportions for maximal Glu release found byGao et al. (2010) are quite similar to those reported in thispaper. At a total protein loading of 16.5mg/g glucan(compared to our 15mg/g glucan), Gao et al. (2010)deduced that the optimum proportions for CBH1, CBH2,EG1, EX, and BX were 30%, 19%, 32%, 15%, and 4%,respectively. BG was kept fixed at 10% and they did notindependently optimize for Xyl release. In our six-component experiment, the optimum proportions of thesefive enzymes was 29%, 5%, 25%, 22%, and 5%. BG wasoptimal at 14% (Table IV). Except for CBH2, these resultsare very similar, that is, the two studies independentlyconfirm the reproducibility of each other. In regard to whyour conclusions about the optimum proportion of CBH2are so different (19% vs. 5%), one possible explanation isthat Gao et al. (2010) used EX2 (GH family 11) whereas weused EX3 (GH10) as the core endo-b1,4-xylanase. EX2is specific for b1,4-xylan, whereas EX3 cleaves both xylanand pNP-cellobioside and short cello-oligosaccharides(Table II; Collins et al., 2005). Therefore, EX3 might beable to partially substitute for CBH2, at least when CBH1and BG are also present. The difference in the CBH2optimum might also be related to the higher proportion ofBG in our mixtures (14% vs. 10%), because CBH2 isinhibited by cellobiose, the substrate of BG (Harjunpaaet al., 1996).

Other differences in protocols between the two labsinclude the source of enzymes, the protein loadings(expressed as mg protein/g glucan), and the particle sizeof the stover. In regard to enzyme sources, our study usedpurified CBH1 from T. longibrachiatum, and CBH2, EG1,


BG, EX3, and BX from T. reesei expressed in P. pastoris. Gaoet al. (2010) used CBH1, EG1, and EX2 purified fromcommercial T. reesei preparations (Spezyme CP or MultifectXylanase), BG purified from a commercial Aspergillus nigerpreparation (Novozyme 188), and BX of A. nidulansexpressed in P. pastoris. Unlike the known differencesbetween EX2 versus EX3, we are not aware of any evidencethat the BG’s and BX’s of T. reesei, A. niger, and A. nidulanshave significantly different catalytic properties. In regard toprotein loadings, in our study we used 15 mg/g glucanthroughout (except for the experiments shown in Fig. 1 andTable VI), whereas Gao et al. (2010) typically used loadingsof 8.25, 16.5, and 33mg/g glucan. These loading differencesresult from the fact that Gao et al. (2010) did not include BGoptimization in their experimental design, but instead keptit at a constant 10% addition. However, it is unlikely that a10% difference in total protein loading (15mg/g glucan vs.16.5mg/g glucan) could account for the large differences intotal Glu yield.

In regard to particle size, Gao et al. (2010) used stoverground to <100mm, whereas we used stover ground to<500mm. As shown by a side-by-side comparison(Table VI), the stover used by Gao et al. (2010) isconsiderably more sensitive to enzymatic digestion. Severalstudies indicate that sensitivity to enzymatic digestion isinversely correlated with particle size below 20–80 mesh(177–841mm) (Decker et al., 2009). Corn fiber that waseither unground or ground to 40 mesh (<420mm) beforeAFEX pretreatment was equally digestible (Moniruzzamanet al., 1997). However, grinding stover to a size below about150mm was found to constitute a form of mechanicalpretreatment that enhanced digestibility (Chundawat et al.,2007). Therefore, the most plausible explanation for thedifferences in total Glu and Xyl yields obtained by Gao et al.(2010) and by us is the difference in particle sizes of the cornstovers used in the two studies. Considering the energeticexpense of grinding, the particle size used in this paper(<500mm) is probably more relevant to an eventualindustrial process (Decker et al., 2009).

Conclusions

In our six-component experiments, we deduced the optimalratios for the different enzymes for Glu, Xyl, and GluþXyl(Table IV). In the mixture optimized for Glu, CBH1was thedominant enzyme, closely followed by EG1 and EX3. Theimportance of CBH1 and EG1 is reasonable from the factthat these two enzymes act on cellulose, the major glucan incorn cell walls. It is less obvious why EX3 is so important.Possible reasons include hindrance by xylans of access tocellulose by the true cellulases in corn stover or that EX3 alsoacts on cello-oligosaccharides (Table II). While not apparentfrom the experiments summarized in Table V (in which noenzyme was proportionately less than 5%), a smallproportion of BG and BX did affect Glu and Xyl yield, asseen in Figure 2.

In the six-component mixture optimized for Xyl, EX3instead of CBH1 was the most important component,followed by EG1. Higher proportions of BX and CBH2 werenecessary for optimal Xyl release compared to optimal Glurelease. The requirement for a high level of EX3 and BX isreasonable given that without both of these enzymes there isnegligible release of Xyl (three component results, Table IV),that is, a mixture of CBH1, EG1, and BG does not degradexylan.

CBH1 and CBH2 are both cellobiohydrolases, one actingat the reducing end and the other from the nonreducing endof cellulose polymers. When tested by themselves togetherwith EG1 and BG, CBH2 was not as effective as CBH1(Table III), but a combination of the two was more effectivethan either one alone (Table IV). CBH2 had negligible effecton the release of Xyl.

Our synthetic mixtures were compared against commer-cial enzymes under the identical GENPLAT conditions,including equal protein loadings. At 15mg/mL, the bestcommercial mixture (Accellerase 1000þMultifectXylanase) released 52.8% of the available Glu and 40.6%of the free Xyl (Table I). At 15mg/mL, the best six-component mixtures released 39.9% of the Glu and 26.4% ofthe Xyl (Table IV). That is, no six-component mixture wasas effective as a commercial mixture. This indicates that oneor more proteins are required for optimal corn stoverconversion in addition to the core six. Candidate required‘‘accessory’’ proteins include others that are abundant in thesecretome of T. reesei, such as xyloglucanase, swollenin,Cip1, Cel5A, or Cel12A (Nagendran et al., 2009).

As a generalized and versatile platform for the develop-ment of improved enzyme mixtures, we foresee severalapplications of GENPLAT. First, it can be used tosystematically analyze the role in lignocellulose digestionof the more than 80 enzymes and proteins that are secretedby T. reesei in addition to the core set of cellulases andxylanases (Nagendran et al., 2009). Second, it can be used todevelop optimal enzyme mixtures for feedstocks andfeedstock/pretreatment combinations other than AFEX-treated corn stover, such as switchgrass, acid-pretreatedcorn stover, or dried distiller’s grains (DDG). Third, it canbe used to screen for naturally occurring alternative coreenzymes or to test engineered enzymes in order to findsuperior versions of those from T. reesei. With GENPLAT itis possible to test alternative enzymes in a realistic multi-enzymic context on industrially relevant biomass substrates(Banerjee et al., 2010; Heinzelman et al., 2009).

This work was funded by the Department of Energy Great Lakes

Bioenergy Research Center (DOE Office of Science BER DE-FC02-

07ER64494). We thank Nick Santoro and Shane Cantu (Great Lakes

Bioenergy Research Center) for advice on robotic liquid handlers and

automated sugar assays; the laboratory of Bruce Dale (Venkatesh

Balan, Shishir Chundawat, and Dahai Gao) for the Kramer stover and

for many fruitful discussions; and Cliff Foster (Great Lakes Bioenergy

Research Center) for the stover analyses. Pat Whitcomb at Stat-Ease,

Inc. (Minneapolis, MN) gave valuable advice on experimental design.

References

Anderson MJ, Whitcomb PJ. 2004. RSM simplified: Optimizing processes

using response surface methods for design of experiments. Productivity

Press. Florence, KY.

Baker JO, Ehrman CI, Adney WS, Thomas SR, Himmel ME. 1998. Hydro-

lysis of cellulose using ternary mixtures of purified cellulases. Appl

Biochem Biotechnol 70–72:395–403.

Balan V, Bals B, Chundawat SP, Marshall D, Dale BE. 2009. Lignocellulosic

biomass pretreatment using AFEX. Meth Mol Biol 581:61–77.

Banerjee G, Scott-Craig JS, Walton JD. 2010. Improving enzymes for

biomass conversion: A basic research perspective. Bioenerg Res

3: 82–92.

Bauer S, Vasu P, Persson S, Mort AJ, Somerville CR. 2006. Development

and application of a suite of polysaccharide-degrading enzymes for

analyzing plant cell walls. Proc Natl Acad Sci USA 103:11417–

11422.

Berlin A,Maximenko V, Bura R, Kang KY, Gilkes N, Saddler J. 2006. A rapid

microaassay to evaluate enzymatic hydrolysis of lignocellulosic sub-

strates. Biotechnol Bioeng 93:880–886.

Berlin A, Maximenko V, Gilkes N, Saddler J. 2007. Optimization of enzyme

complexes for lignocellulose hydrolysis. Biotechnol Bioeng 97:287–296.

Blakeney AB, Harris PJ, Henry RJ, Stone BA. 1983. A simple and rapid

preparation of alditol acetates for monosaccharide analysis. Carbohydr

Res 113:291–299.

Boer H, Teeri TT, Koivula A. 2000. Characterization of Trichoderma reesei

cellobiohydrolase Cel7A secreted from Pichia pastoris using two dif-

ferent promoters. Biotechnol Bioeng 69:486–494.

Chaplin MF, Kennedy JF. 1986. Carbohydrate analysis: A practical

approach. Oxford: IRL Press.

Chundawat SP, Balan V, Dale BE. 2007. Effect of particle size based

separation of milled corn stover on AFEX pretreatment and enzymatic

digestibility. Biotechnol Bioeng 96:219–231.

Chundawat SP, Balan V, Dale BE. 2008. High-throughput microplate

technique for enzymatic hydrolysis of lignocellulosic biomass. Biotech-

nol Bioeng 99:1281–1294.

Collins T, Gerday C, Feller G. 2005. Xylanases, xylanase families and

extremophilic xylanases. FEMS Microbiol Rev 29:3–23.

Dashtaban M, Schraft H, Qin W. 2009. Fungal bioconversion of lignocel-

lulosic residues; opportunities & perspectives. Int J Biol Sci 5:578–595.

De Schutter K, Lin Y-C, Tiels P, Van Hecke A, Glinka S, Weber-Lehmann J,

Rouze P, Van de Peer Y, Callewaert N. 2009. Genome sequence of the

recombinant protein production host Pichia pastoris. Nature Biotech-

nol 27:561–566.

Decker SR, Brunecky R, Tucker MP, Himmel ME, Selig MJ. 2009. High-

throughput screening techniques for biomass conversion. Bioenerg Res

2:179–192.

Farrell AE, Plevin RJ, Turner BT, Jones AD, O’Hare M, Kammen DM. 2006.

Ethanol can contribute to energy and environmental goals. Science

311:506–508.

Gao D, Chundawat SP, Krishnan C, Balan V, Dale BE. 2010. Mixture

optimization of six core glycosyl hydrolases for maximizing sacchar-

ification of ammonia fiber expansion (AFEX) pretreated corn stover.

Bioresour Technol 101: 2770–2781.

Harjunpaa V, Teleman A, Koivula A, Ruohonen L, Teeri TT, Teleman O,

Drakenberg T. 1996. Cello-oligosaccharide hydrolysis by cellobiohy-

drolase II from Trichoderma reesei. Eur J Biochem 240:584–591.

Harrison MJ, Nouwens AS, Jardine DR, Zachara NE, Gooley AA, Neva-

lainen H, Racker NH. 1998. Modified glycosylation of cellobiohydro-

lase I from a high cellulase-producing mutant strain of Trichoderma

reesei. Eur J Biochem 256:119–127.

Heinzelman P, Snow CD, Wu I, Nguyen C, Villalobos A, Govindarajan S,

Minshull J, Arnold FH. 2009. A family of thermostable fungal cellulases

created by structure-guided recombination. Proc Natl Acad Sci USA

106:5610–5615.

Himmel ME, Ding SY, Johnson DK, Adney WS, Nimlos MR, Brady JW,

Foust TD. 2007. Biomass recalcitrance: Engineering plants and enzymes

for biofuels production. Science 315: 804–807.



Irwin DC, Spezio M, Walker LP, Wilson DB. 1993. Activity studies of eight

purified cellulases: Specificity, synergism, and binding domain effects.

Biotechnol Bioeng 42:1002–1013.

JeyaM, Thiagarajan S, Lee JK, Gunasekaran P. 2009. Cloning and expression

of GH11 xylanase gene from Aspergillus fumigatus MKU1 in Pichia

pastoris. J Biosci Bioeng 108:24–29.

Kim E, Irwin DC, Walker LP, Wilson DB. 1998. Factorial optimization of a

six-cellulase mixture. Biotechnol Bioeng 58:494–501.

King BC, Donnelly MK, Bergstrom GC, Walker LP, Gibson DM. 2009.

An optimized microplate assay system for quantitative evaluation of

plant cell wall-degrading enzyme activity of fungal culture extracts.

Biotechnol Bioeng 102:1033–1044.

Kumar R, Wyman CE. 2009. Effect of xylanase supplementation of cellulase

on digestion of corn stover solids prepared by leading pretreatment

technologies. Bioresour Technol 100:4203–4213.

Lau MW, Dale BE. 2009. Cellulosic ethanol production from AFEX-treated

corn stover using Saccharomyces cerevisiae 424A(LNH-ST). Proc Natl

Acad Sci USA 106:1368–1373.

Lever M. 1972. A new reaction for colorimetric determination of carbohy-

drates. Anal Biochem 47:273–279.

Li YL, Li H, Li AN, Li DC. 2009. Cloning of a gene encoding thermostable

cellobiohydrolase from the thermophilic fungus Chaetomium thermo-

philum and its expression in Pichia pastoris. J ApplMicrobiol 106:1867–

1875.

Lu Y, Wang TH, Ding XL. 2009. Induction of production and secretion

beta(1-4) glucanase with Saccharomyces cerevesiae by replacing the

MET10 gene with egl1 gene from Trichoderma reesei. Lett Appl Micro-

biol 49: 702–707.

Lynd LR, Laser MS, Bransby D, Dale BE, Davison B, Hamilton R, Himmel

M, Keller M, McMillan JD, Sheehan J, Wyman CE. 2008. How biotech

can transform biofuels. Nature Biotechnol 26:169–172.

Mach RL, Seiboth B, Myasnikov A, Gonzalez R, Strauss J, Harkki AM,

Kubicek CP. 1995. The bgl1 gene of Trichoderma reesei QM 9414

encodes an extracellular, cellulose-inducible beta-glucosidase involved

in cellulase induction by sophorose. Mol Microbiol 16:687–697.

Meyer AS, Rosgaard L, Sørensen HR. 2009. The minimal enzyme cocktail

concept for biomass processing. J Cereal Sci 50:337–344.


Moniruzzaman M, Dale BE, Hespell RB, Bothast RJ. 1997. Enzymatic

hydrolysis of high-moisture corn fiber pretreated by AFEX and recovery

and recycling of the enzyme complex. Appl Biochem Biotechnol

67:113–126.

Nagendran S, Hallen-Adams HE, Paper JM, Aslam N, Walton JD. 2009.

Reduced genomic potential for secreted plant cell-wall-degrading

enzymes in the ectomycorrhizal fungus Amanita bisporigera, based

on the secretome of Trichoderma reesei. Fungal Genet Biol 46:427–435.

Rosgaard L, Pedersen S, Langston J, Akerhielm D, Cherry JR, Meyer AS.

2007. Evaluation of minimal Trichoderma reesei cellulase mixtures on

differently pretreated barley straw substrates. Biotechnol Prog 23:1270–

1276.

Santoro N, Cantu SL, Tornqvist CI, Falel TG, Bolivar JL, Paterson SE, Pauly

M, Walton JD. 2010. A high throughput platform for screening

milligram quantities of plant biomass for lignocellulose digestibility.

Bioenerg Res 3: 93–102.

Smith PK, Krohn RI, Hermanson GT, Mallia AK, Gartner FH, Provenzano

MD, Fujimoto EK, Goeke NM, Olson BJ, Klenk DC. 1985. Measure-

ment of protein using bicinchoninic acid. Anal Biochem 150:76–85.

Stals I, Sandra K, Geysens S, Contreras R, Van Beeumen J, Claeyssens M.

2004a. Factors influencing glycosylation of Trichoderma reesei cellu-

lases. I: Postsecretorial changes of the O- andN-glycosylation pattern of

Cel7A. Glycobiology 14:713–724.

Stals I, Sandra K, Devreese B, Van Beeumen J, Claeyssens M. 2004b. Factors

influencing glycosylation of Trichoderma reesei cellulases. II: N-glyco-

sylation of Cel7A core protein isolated from different strains. Glyco-

biology 14:725–737.

Wilson DB. 2009. Cellulases and biofuels. Curr Opin Biotechnol 20:295–

299.

Xiao Z, Storms R, Tsang A. 2004. Microplate-based assay to measure total

cellulase activity. Biotechnol Bioeng 88:832–837.

Xu J, Takakuwa N, Nogawa M, Okada H, Morkikawa Y. 1998. A third

xylanase from Trichoderma reesei PC-3-7. Appl Microbiol Biotechnol

49:718–724.

Zhang Y-HP, Himmel ME, Mielenz JR. 2006. Outlook for cellulase

improvement: Screening and selection strategies. Biotechnol Adv 24:

452–481.

Documents

Synthetic Enzyme Mixtures for Biomass Deconstruction ... · ARTICLE Synthetic Enzyme Mixtures for Biomass Deconstruction: Production and Optimization of a Core Set Goutami Banerjee,1,2