You are on page 1of 8

698 Biotechnol. Prog.

2004, 20, 698−705

Development and Validation of a Kinetic Model for Enzymatic


Saccharification of Lignocellulosic Biomass

Kiran L. Kadam,*,† Eric C. Rydholm,‡ and James D. McMillan†

National Renewable Energy Laboratory, National Bioenergy Center, Biotechnology Division for Fuels and
Chemicals, 1617 Cole Blvd., Golden, Colorado 80401, and Department of Chemical Engineering,
University of Colorado, Boulder, Colorado 80309

A multireaction kinetic model was developed for closed-system enzymatic hydrolysis


of lignocellulosic biomass such as corn stover. Three hydrolysis reactions were modeled,
two heterogeneous reactions for cellulose breakdown to cellobiose and glucose and
one homogeneous reaction for hydrolyzing cellobiose to glucose. Cellulase adsorption
onto pretreated lignocellulose was modeled via a Langmuir-type isotherm. The sugar
products of cellulose hydrolysis, cellobiose and glucose, as well as xylose, the dominant
sugar prevalent in most hemicellulose hydrolyzates, were assumed to competitively
inhibit the enzymatic hydrolysis reactions. Model parameters were estimated from
experimental data generated using dilute acid pretreated corn stover as the substrate.
The model performed well in predicting cellulose hydrolysis trends at experimental
conditions both inside and outside the design space used for parameter estimation
and can be used for in silico process optimization.

Introduction challenges faced in efficiently hydrolyzing biomass feed-


stocks. A kinetic model based on observable, macroscopic
Abundantly available lignocellulosic biomass poten- properties of the overall system can be helpful in the
tially can be converted to ethanol; this approach, besides design and economic evaluation of processes for sugar
yielding a renewable liquid fuel, offers concomitant and ethanol production. Experimental data generation
environmental benefits (1). Corn stover is an example is labor-intensive and analytically challenging; a kinetic
biomass feedstock that can generate more than three model is a useful forecasting tool in this context.
billion gallons per year of ethanol in the United States
Previous Modeling Efforts. Cellulose hydrolysis is
(2). Sugar production from cellulose is a major step in
effected via a synergistic action of endo-β-1,4-glucanase
the biomass-to-ethanol process. The enzymatic hydrolysis
(EG), exo-β-1,4-cellobiohydrolase (CBH), exo-β-1,4-glucan
of cellulose offers advantages over other chemical conver-
glucohydrolase, and β-glucosidase (12-15). The first
sion routes of higher yields, minimal byproduct forma-
three enzymes produce cellobiose and glucose; cellobiose
tion, low energy requirements, and mild operating con-
is hydrolyzed to glucose by the action of β-glucosidase.
ditions (3-6). Although the enzymatic route has the
Both sugars cause end-product inhibition; however, cel-
highest current costs, it has long-term potential for cost
lobiose is a much stronger inhibitor than glucose.
reductions compared to other more established routes
such as concentrated acid and two-stage dilute acid Previously reported saccharification models depict
hydrolysis (7). Simultaneous saccharification and fer- enzymatic hydrolysis of lignocellulose in a variety of ways
mentation (SSF) is one of several process configurations and can be broadly characterized as being theoretical
by which ethanol can be produced from biomass; organ- (16-20), empirical (21-24), or based on SSF (25-27).
isms and configurations have been reviewed by Wyman The most relevant previous modeling efforts are sum-
(8). In simultaneous saccharification and cofermentation marized below.
(SSCF), a variant of SSF, sugars from lignocellulosic The kinetic model of enzymatic hydrolysis of cellulose
biomass are produced and converted in situ to ethanol; by Wald et al. (19) incorporates enzyme adsorption,
hemicellulosic sugars already present in the hydrolyzate product inhibition, and a multiple enzyme system. The
are also fermented to ethanol (9). simultaneous adsorption of cellulase components, which
Irrespective of processing schemes, enzymatic saccha- have different adsorption characteristics, is responsible
rification of lignocellulose is a key cost center in the for hydrolysis; enzyme adsorption had not been ad-
overall bioconversion process. Gregg et al. (10) and equately addressed before this effort. The Michaelis-
Esteghlalian et al. (11) provide an excellent overview of Menten model used in earlier modeling attempts is
applicable to low substrate concentrations when the
substrate is not close to maximum loading. In the Wald
* To whom correspondence should be addressed. Fax: 303 384 et al. (19) representation, enzyme adsorption is a function
6877. Email: kiran_kada@nrel.gov. Corresponding author’s cur- of available sorption sites and, thus, of accessible surface
rent address: Quintessence Technologies LLC, 16696 W. Bayaud
Dr., Golden, CO 80401. Fax: 303 986 2530. E-mail: kkadam1@ area via a Langmuir-type isotherm relationship. This
yahoo.com. model does a good job of recognizing and representing
† National Bioenergy Center. surface phenomenon and is a first attempt to realistically
‡ University of Colorado. portray enzymatic hydrolysis. Although the model does
10.1021/bp034316x CCC: $27.50 © 2004 American Chemical Society and American Institute of Chemical Engineers
Published on Web 04/23/2004
Biotechnol. Prog., 2004, Vol. 20, No. 3 699

not consider glucose end-product inhibition, it was ca- lignocellulosic biomass, but the parameters should be
pable of simulating saccharification of rice straw ligno- based on observable phenomena to the extent possible.
cellulose at high substrate (up to 333 g/L) and high The model should be grounded in the biochemistry of
enzyme (up to 9.2 FPU/mL) concentrations. enzymatic hydrolysis and should include adsorption of
Gusakov et al. (28, 29) mathematically describe the cellulase components onto and desorption off of the
kinetics of glucose and cellobiose formation from cellulose lignocellulosic substrate, thermal effects, and end-product
for batch and plug-flow processes for enzymatic cellulose inhibition. However, not all system complexities must be
hydrolysis and take into account the composition of the incorporated to develop an effective model; the general
cellulase complex, the structural complexity of cellulose, preference is to minimize the number of model param-
inhibition by reaction products, and enzyme inactivation eters.
during enzymatic hydrolysis. Inactivation of enzymes as Substrate Reactivity. There are several factors
modeled is limited to that due to unproductive adsorption besides end-product inhibition that reduce hydrolysis
to substrate. Thus, while these models are realistic and rates as saccharification progresses including lower
sophisticated, thermal inactivation is not modeled and substrate reactivity (variations in crystal structure,
the representation of cellulase adsorption is rudimen- degree of polymerization, substrate accessibility, etc.),
tary: [enzyme] + [substrate] S [enzyme-substrate]ads. enzyme inactivation, and loss of enzyme due to adsorp-
Nonetheless, the reported models satisfactorily predicted tion on lignin. Cellulose contains both amorphous and
batch glucose and cellobiose accumulation over the crystalline regions. In studies with pure celluloses,
substrate concentration range of 5-100 g/L and the amorphous cellulose is degraded 5-10 times more rapidly
enzyme concentration range of 5-60 g/L. when compared to highly crystalline celluloses by both
The field was further advanced by South et al. (26) and fungal enzymes and ruminal bacteria (12, 31-33). It has
Philippidis and Hatzis (27), who developed SSF-based been suggested that the high initial rates are due to
models to describe lignocellulose conversion to ethanol. preferential hydrolysis of the amorphous region and that
The model by South et al. (26), which applies to batch the rate decreases later as the enzyme encounters the
and continuous operations, incorporates a hydrolysis rate more recalcitrant crystalline region (34). However, as
equation (for cellulose to cellobiose conversion) and uses mentioned above several researchers have observed no
the following equation for the adsorption of cellulase to substantial change in crystallinity as the saccharification
cellulose: [bound enzyme] ) constant × [free enzyme] progresses (35-37) or obtained equivocal results. Also,
× [substrate]. Adsorption of cellulase to the lignin portion Wald et al. (19) and Desai and Converse (38) argue that
of the biomass is similarly modeled. The authors report the two-substrate hypothesis suffers from making an
that the model accurately describes batch SSF data from assumption, i.e., the easier to digest substrate is not
well-mixed bioreactors over a wide range of substrate and shrouded or otherwise blocked by the recalcitrant one,
enzyme concentrations. In the SSF model reported by which is phenomenologically not sound. Thus, changing
Philippidis and Hatzis (27), cellulose hydrolysis rate intrinsic substrate reactivity may be a more practical way
equations include cellulose-to-cellobiose conversion as to model the biphasic behavior than resorting to an
well as direct cellulose-to-glucose conversion. This model assumed dichotomy between amorphous and crystalline
does not include enzyme-substrate adsorption equilibria, regions. Hence, although structure in cellulose can be
and cellobiose-to-glucose conversion is assumed to be introduced by dividing total cellulose into amorphous and
independent of enzyme concentration. Moreover, enzyme crystalline fractions, we chose to represent cellulose as
inactivation is modeled as an exponential function of a uniform substrate. This approach simplifies cellulase
cellulose conversion rather than time, making the model adsorption modeling and obviates the need to indepen-
less general. This diminution in generality stems from dently measure distinct enzyme adsorption equilibria for
the fact that enzyme inactivation is known to be related the putative amorphous and crystalline cellulose.
to the duration of exposure to the temperature and is Assuming that the normalized initial saccharification
usually modeled as a first-order kinetics, which leads to rates during secondary hydrolysis of residual substrates
a mechanism of exponential decay with respect to time (see Materials and Methods for details) are indicative of
(16, 30). However, the model derived by the above substrate reactivity, the following correlation is then
authors was able to predict SSF performance well. proposed:
Together, these previous efforts incorporate many of the
S
important phenomena that a kinetic model of enzymatic RS ) R (1)
cellulose hydrolysis needs to describe. S0
Study Objective. Kinetic modeling of enzymatic hy-
drolysis is complicated by the heterogeneous nature of where RS is a substrate reactivity parameter (dimension-
the substrate and multiple enzyme activities. Rigorous less), R is a constant (dimensionless), S is the substrate
models must consider the phenomena of enzyme adsorp- concentration at a given time (g/kg), S0 is the initial
tion, sugar inhibition, temperature effects, and substrate substrate concentration (g/kg), and S/S0 is the dimen-
reactivity. Our objective was to formulate and validate sionless substrate concentration at a given time.
a kinetic model that incorporates these features. This Enzyme Adsorption. Both Langmuir and Freundlich-
paper describes the development and validation of a Langmuir adsorption models have been used to describe
kinetic model for enzymatic cellulose hydrolysis that is enzyme adsorption onto solid lignocellulosic substrates
capable of predicting performance over a range of operat- (39-41). Although the Langmuir model has been used
ing conditions encompassing various background sugar for describing enzyme adsorption in this heterogeneous
concentrations, temperatures, and mixing regimes. system, it should be noted that the underlying assump-
tions for the Langmuir model, uniform binding sites and
Model Development no interaction between the adsorbing molecules, are not
valid for cellulase adsorption onto cellulose. Nevertheless,
The general modeling philosophy followed in this work the Langmuir formulation remains useful for mathemati-
is that a kinetic model should be sophisticated enough cally describing the phenomenon of enzyme adsorption,
to describe the complexities of enzymatic hydrolysis of and we have used it in this study.
700 Biotechnol. Prog., 2004, Vol. 20, No. 3

temperature on saccharification is captured via the


Arrhenius rate relationship indicated in eq 10; this
equation is only valid at temperatures below the enzyme’s
temperature maximum. Although we are currently using
a mesophilic enzyme preparation with an optimal tem-
perature of 50 °C, being able to describe the temperature
dependence of enzymatic hydrolysis over a broad tem-
perature range will be very important for modeling and
optimizing processes based on the use of more thermo-
stable enzymes currently being developed.
Enzyme Adsorption

Eimax KiadEiFS
Langmuir isotherm EiB ) (2)
1 + KiadEiF

Figure 1. Reaction scheme for modeling cellulose hydrolysis. Cellulose-to-Cellobiose Reaction with Competi-
Enzymes involved in r1: endo-β-1,4-glucanase (EC 3.2.1.4) and tive Glucose, Cellobiose and Xylose Inhibition
exo-β-1,4-cellobiohydrolase (CBH) (EC 3.2.1.91). Enzymes in-
volved in r2: exo-β-1,4-cellobiohydrolase (CBH) (EC 3.2.1.91) and k1rE1BRSS
exo-β-1,4-glucan glycohydrolase (EC 3.2.1.74). Enzymes involved r1 ) (3)
in r3: β-bludosidase or cellobiases (EC 3.2.1.21). G2 G X
1+ + +
K1IG2 K1IG K1IX
Proposed Kinetic Model. Figure 1 depicts the sim-
plified cellulose hydrolysis reaction sequence that forms Cellulose- to-Glucose Reaction with Competitive
the basis for the model. Each enzymatic reaction is Glucose, Cellobiose and Xylose Inhibition
potentially inhibited by the sugar it generates or by the
six sugars already present in the system, i.e., glucose, k2r(E1B + E2B)RSS
cellobiose, galactose, mannose, xylose, and arabinose. To r2 ) (4)
G2 G X
simplify model development, the sugar system was 1+ + +
consolidated to three sugars: cellobiose, glucose, and K2IG2 K2IG K2IX
xylose. In terms of inhibition capacity, xylose was used
as the surrogate for both pentoses (xylose and arabinose) Cellobiose- to-Glucose Reaction with Competi-
and glucose was used as the surrogate for all hexoses tive Glucose and Xylose Inhibition
(glucose, galactose, and mannose).
The potential for complicated sugar inhibition phe- k3rE2FG2
r3 )
( )
nomena coupled with the presence of both heterogeneous (5)
G X
and homogeneous reaction types necessitates a sophis- K3M 1+ + + G2
K3IG K3IX
ticated model to effectively simulate the system. The
Langmuir adsorption model is shown in eq 2. Rate
Mass Balances
equations based on a competitive mode (42) of sugar
inhibition are shown below. Equation 3 describes the dS
conversion of cellulose to cellobiose, eq 4 the conversion cellulose: ) - r1 - r2 (6)
dt
of cellulose to glucose, and eq 5 the conversion of
cellobiose to glucose. The competitive mode of sugar dG2
inhibition assumes that inhibitor sugars are substrate cellobiose: ) 1.056r1 - r3 (7)
analogues and bind competitively to the active site, dt
thereby retarding the formation of enzyme-substrate
complex. In the noncompetitive mode of inhibition, dG
glucose: ) 1.111r2 + 1.053r3 (8)
inhibitors do not bind to the active site but bind rather dt
at a remote positions causing the maximum specific
reaction rate to diminish. Unlike competitive inhibition, enzyme: ETi ) EFi + EBi (9)
noncompetitive inhibition is irreversible (42). We chose
to model competitive inhibition as it is mechanistically Temperature Dependence
more realistic.
These rate equations assume that (1) enzyme adsorp- Arrhenius eq: kir(T2) ) kir(T1) e- Eai/R{1/T1-1/T2};
tion follows a Langmuir-type isotherm with the first- 30 °C e T e 55 °C (10)
order reactions (r1 and r2) occurring on the cellulose
surface; (2) the cellulose matrix is uniform in terms of The proposed model distinguishes between the β-glu-
its susceptibility to enzymatic attack (i.e., no provision cosidase and CBH/EG enzymes, representing their ad-
was made to include separately more reactive amorphous sorption via a Langmuir-type isotherm. This more struc-
and more recalcitrant crystalline cellulose fractions); (3) tured formulation permits the model to better describe
enzyme activity remains constant; and (4) conversion of the kinetics of cellulase preparations having different
cellobiose to glucose occurs in solution and follows proportions of these components. Beyond this, the model
classical Michaelis-Menton kinetics. incorporates potential inhibition by xylose, a prominent
The mass balances for cellulose, cellobiose, glucose, and sugar in hydrolyzates of dilute acid pretreated biomass;
cellulase enzymes are listed in eqs 6-9, respectively, and inhibition of cellulases by xylose has not been considered
are based on fundamental principles of mass conserva- previously.
tion; these equations will not change if modifications are Computational Methodology. The kinetic model
made in the kinetic rate expressions. The effect of was coded and solved using Matlab programming envi-
Biotechnol. Prog., 2004, Vol. 20, No. 3 701

ronment (The MathWorks, Natick, MA). The Matlab Table 1. Estimated Model Parameters
optimization function “lsqnonlin” was used to simulta- parameter value
neously estimate the various model parameters. Param-
Independently Established Parameters
eters characterizing the Langmuir, Arrhenius, and sub- K1ad (g protein/g substrate) 0.4
strate reactivity correlations were determined indepen- K2ad (g protein/g substrate) 0.1
dently and served as inputs. E1max (g protein/g substrate) 0.06
E2max (g protein/g substrate) 0.01
Materials and Methods Ea (cal/mole) -5540
Rs RS/S0, R ) 1
The substrate used in the enzymatic hydrolysis studies Parameters Obtained by
was pretreated corn stover (composed of approximately Regression of Saccharification Data
60% cellulose) that was water-washed to remove en- k1r (g/mg‚h) 22.3
trained soluble sugars. Corn stover was pretreated with K1IG2 (g/kg) 0.015
dilute sulfuric acid in an engineering-scale countercur- K1IG (g/kg) 0.1
rent pretreatment reactor system, a 200 kg dry biomass/ K1IX (g/kg) 0.1
day capacity Sunds hydrolyzer (Metso Paper USA Inc., k2r (g/mg.h) 7.18
K2IG2 (g/kg) 132.0
Norcross, GA) using 1.4% acid concentration, 8-min K2IG (g/kg) 0.04
residence time, and 190 °C temperature. The details of K2IX (g/kg) 0.2
the hydrolyzer setup and operation are described by k3r (h-1) 285.5
Nguyen et al. (43) and Tucker et al. (44). The pretreated K3M (g/kg) 24.3
corn stover slurry was separated into liquor and solid K3IG (g/kg) 3.9
fractions, and the solid fraction was washed (45) and used K3IX (g/kg) 201.0
in the experiments.
For shake flasks the standard operating conditions Results and Discussion
were as follows: 100 g total charge in 250-mL baffled The basic approach followed was to obtain experimen-
Erlenmeyer flasks; 10% corn stover solids by weight (dry tal data, estimate model parameters, and conduct simu-
basis); enzyme loading of 45 mg protein/g cellulose (CPN lations using these parameters. Model predictions were
commercial cellulase, Iogen Corp., Ottawa, Ontario, compared with experimental data to validate the model
Canada), equivalent to approximately 15 FPU/g cellulose and assess its utility in predicting hydrolytic performance
(no β-glucosidase supplementation); 45 °C temperature beyond the range of experimental conditions employed
of incubation; shaker (Innova 4000, New Brunswick in parameter estimation. In the ensuing figures, experi-
Scientific, Edison, NJ) speed of 130 rpm; initial pH of mental data are presented on a weight/weight basis
4.8 (0.05 M citrate buffer added, pH uncontrolled). For because of the heterogeneous, two-phase nature of the
stirred-tank reactors (BioFlo III, New Brunswick Scien- system. Doing so also readily allows one to calculate
tific, Edison, NJ) the standard operating conditions were conversion yields. Sugar and substrate concentrations are
same as above, except as follows: 700 g total charge and expressed as g/kg, i.e., grams of entity per kilogram of
impeller speed of 250 rpm. total system mass.
Liquid samples were analyzed for glucose, xylose, and Parameter Estimation. There are two types of model
cellobiose using HPLC (Agilent Technologies, Palo Alto, parameters: independently established parameters that
CA; Bio-Rad Laboratories, Hercules, CA). For character- are derived through separate experimentation and ki-
izing Langmuir adsorption behavior, free enzyme was netic parameters that are derived via Matlab regression
assumed to be the protein concentration in the superna- of hydrolysis data; both are listed in Table 1.
tant and was measured using the Bradford assay using Adsorption. Langmuir parameters estimated from
Coomassie blue dye (Pierce Biotechnology, Rockford, IL). enzyme adsorption studies are as follows (in g protein/g
Bound enzyme was calculated by subtracting the free substrate): for EG/CBH E1max ) 0.06 and K1ad ) 0.4, and
enzyme concentration from the initial protein concentra- for β-glucosidase E2max ) 0.01 and K2ad ) 0.1.
tion charged to each flask. Adsorption of one cellulase Substrate Reactivity. The initial hydrolysis rates
preparation (CPN cellulase) and one β-glucosidase prepa- during secondary enzymatic hydrolysis were plotted as
ration (Novo 188 from Novozymes Biotech, Davis, CA) a function of S/S0 (Figure 2) to arrive at substrate
were tested. reactivity. Given that the best-fit value for R is close to
The issue of substrate reactivity was studied using unity (0.97 with R2 ) 0.97, data not shown), the substrate
washed pretreated corn stover solids and CPN cellulase reactivity correlation of eq 1 simplifies to
preparation. During hydrolysis, two flasks were har-
vested at each time point, and the residual substrate was S
RS ) (11)
then subjected to secondary enzymatic hydrolysis for an S0
additional 4 h using fresh enzyme added at a constant
enzyme/substrate ratio. The residual cellulose was esti- The experimental 0-4 h rate data can be described
mated from net glucose and cellobiose produced. Washing well by this simplified substrate reactivity correlation
the substrate removes end-product inhibition, and adding (Figure 2).
fresh enzyme eliminates enzyme inactivation as a factor. Temperature Dependence. To predict the effect of
Hence, this assay attempts to isolate intrinsic substrate temperature on cellulose saccharification, a single Ar-
reactivity. It should be noted, however, that the washing rhenius relationship was assumed for all reactions. It
step does not necessarily dislodge bound enzyme. should be noted that the Arrhenius relationship would
The standard conditions using the shake-flask system be different for each enzyme component because of their
described above were used for parameter estimation. This varying degrees of thermostability, with CBH being the
defines the standard design space. Experiments for model most thermostable. Hence, the above assumption is a
validation were conducted using conditions outside the simplification of reality. This equation was used in model
design space with different background sugar concentra- validation experiments performed at temperatures be-
tions, temperatures, and mixing regimes. sides 45 °C, the standard operating temperature used.
702 Biotechnol. Prog., 2004, Vol. 20, No. 3

Figure 2. Relative overall glucose production rates during Figure 3. Arrhenius plot based on initial glucose production
secondary hydrolysis using constant enzyme/substrate ratio. rates.

The best-fit activation energy, Ea, based on initial rate


data was estimated to be -5540 cal/mol for the CPN
enzyme (Figure 3). We expect thermostable enzymes to
be eventually employed, and the activation energy for
these enzymes will be different.
Kinetic Parameters. The independently established
parameters were used as input to the model, and the
kinetic parameters were subsequently regressed; Table
1 lists best-fit estimates of these parameters. A compari-
son was attempted between the parameters listed in
Table 1 and those reported by Oh et al. (46), who
determined the competitive glucose inhibition parameter
for r3 to be 0.33 g/L. Their method was to determine the
initial rate of cellobiose consumption during SSF and
back calculate the cellobiose-to-glucose conversion rates.
This approach makes it more difficult to pinpoint the true
effect of glucose on β-glucosidase activity. Oh et al. (46) Figure 4. Enzymatic cellulose hydrolysis was conducted in the
also determined K1IG2 ) 6.1 g/kg and K1IG ) 12.4 g/kg for shake-flask system and a stirred-tank reactor using 10% w/w
corn stover solids with no added background sugars.
competitive inhibition. The lower inhibition they found
for reaction r1 may be balanced by the higher inhibition
they observed for the cellobiose-to-glucose reaction (r3). ducted in stirred-tank reactors using 10% w/w corn stover
Oh et al. (46) did not model the direct conversion of solids. Results showed that the model does a fairly good
cellulose to glucose and report that, at a glucose concen- job of predicting glucose production in a shake flask or a
tration of 20 g/L, β-glucosidase was practically inactive. stirred-tank reactor (Figure 4). Hence, the parameters
Our results using CPN cellulase enzyme show a less estimated on the basis of shake-flask data appear ap-
dramatic reduction of β-glucosidase activity above 20 g/L plicable in predicting hydrolysis in a stirred-tank reactor
glucose. When 30 g/kg glucose (our concentrations are exhibiting different hydrodynamics.
on a w/w basis) is added initially, the cellobiose conver- Validation tests also show that the model does a sound
sion is slower compared to the case with no added job of predicting hydrolysis behavior at high initial
glucose, but cellobiose conversion still occurs. These glucose, xylose, and cellobiose concentrations (Figures
findings illustrate how different reaction schemes and 5-7, respectively). Thus, inhibition by these three sugars
model structures can result in different parameter values is well captured by the model. Xylose inhibition as
and make direct comparisons of different models difficult predicted by the model is authentic in the sense that the
(see also ref 47). Such comparisons are further con- 24-h hydrolysis efficiencies at 0 and 50 g/kg background
founded by different researchers using different enzyme xylose are 53% vs 46%, respectively, suggesting signifi-
mixtures, different enzyme and substrate loading ranges, cant xylose inhibition. The high cellobiose concentration
different hydrolysis temperatures, etc. used is only for model validation given that cellobiose
Model Validation. As shown in Figure 4, which is concentrations typically peak at much lower levels (10-
generated using standard conditions, the proposed model 12 g/L); cellobiose concentrations are mainly dependent
predicts hydrolysis behavior with reasonable accuracy. on the β-glucosidase content of a given enzyme prepara-
However, the kinetic model for cellulose saccharification tion.
needs to be validated using datasets not used to estimate Model Predictions of Sugar Inhibition. Model
model parameters. Simulations were run using the fidelity can be further assessed by examining if it
parameters listed in Table 1 to predict cellulose hydroly- correctly describes sugar inhibition. Inhibition by each
sis performance at background sugar concentrations, sugar is inversely proportional to its respective inhibition
temperatures, and mixing regimes different from those constant. On the basis of this fact, inhibition of reaction
in the standard design space. r1 by cellobiose is more potent than that by glucose, which
The data used for parameter estimation were gener- is expected. Conversely, inhibition of reaction r2 by
ated using 10% w/w insoluble solids in a shake-flask glucose is much higher than that by cellobiose because
system. Enzymatic cellulose hydrolysis was also con- glucose is the reaction product obtained directly from
Biotechnol. Prog., 2004, Vol. 20, No. 3 703

Figure 5. Kinetic model validation. Enzymatic cellulose hy- Figure 8. Sugar inhibition patterns for cellulose-to-cellobiose
drolysis was conducted in the shake-flask system using 10% w/w reaction (r1). Enzymatic cellulose hydrolysis was conducted in
corn stover solids with initial background glucose of 30 or 50 the shake-flask system using 10% w/w corn stover solids with
g/kg. initial background xylose of 30 g/kg. The overall hydrolysis
profile was similar to that shown in Figure 5.

Figure 6. Kinetic model validation. Enzymatic cellulose hy-


drolysis was conducted in the shake-flask system using 10% w/w Figure 9. Kinetic model validation. Enzymatic cellulose hy-
corn stover solids with initial background xylose of 40 g/kg. drolysis was conducted in the shake-flask system using 10% w/w
corn stover solids at temperatures of 40, 50, and 55 °C.

respective inhibition constant, was calculated for reaction


1. The relative contribution of each sugar to the total to
inhibition was then calculated. Figure 8 depicts the
resulting sugar inhibition patterns along with predicted
reaction rates normalized to the time-zero reaction rate.
As expected, normalized reaction rates decrease with
time. The model predicts that inhibition due to cellobiose
and xylose plays a major role during the early phase.
However, as glucose concentration builds up, its contri-
bution to total inhibition also rises. The overall inhibition
pattern mechanistically makes sense.
Model Predictions of Temperature Effects. The
model was also used to predict hydrolytic performance
at different temperatures using eq 10. The model predicts
hydrolysis behavior reasonably well for 40° and 50 °C
Figure 7. Kinetic model validation. Enzymatic cellulose hy- for the first 96 h but falters after 24 h for 55 °C (Figure
drolysis was conducted in the shake-flask system using 10% w/w 9). This is explained by the fact that the temperature
corn stover solids with initial background cellobiose of 30 g/kg. optimum for the CPN enzyme is near 50 °C, and the
cellulose. Xylose inhibition of reaction r1 is of the same enzyme is slowly inactivated at higher temperatures.
order of magnitude as that by glucose. Inhibition of Therefore, this feature of the model needs more fine-
reaction r3 is dominated by glucose, with inhibition by tuning. Obviously, more thermostable second-generation
xylose being two orders of magnitude less than for enzymes should be available before putting an extensive
glucose. Hence, the model is consistent with expected effort into modeling enzyme inactivation.
inhibition by glucose and cellobiose, and the predicted
role for xylose seems reasonable as well. To further Conclusion
illustrate sugar inhibition patterns, inhibition by each We have developed and validated a kinetic model for
sugar, which is the ratio of sugar concentration and its batch enzymatic saccharification of dilute acid pretreated
704 Biotechnol. Prog., 2004, Vol. 20, No. 3

corn stover that incorporates structure in how the cel- Nehoda KC: Environmental Life Cycle Implications of Fuel
lulase enzyme system is represented and considers Oxygenates Production from California Biomass, NREL/TP-
feedback inhibition by biomass-derived sugars. The model 580-25688; National Renewable Energy Laboratory: Golden,
is capable of predicting hydrolysis performance beyond CO, 1999.
(5) Mandels, M, Hontz, L, Nystrom, J. Enzymatic hydrolysis
the range of experimental conditions used to estimate the
of waste cellulose. Biotechnol. Bioeng. 1974, 16, 1471-1493.
model parameters. The model differs from previous (6) Ghose, T. K.; Ghosh, P. Bioconversion of cellulosic sub-
models in that it considers inhibition by xylose, a major stances. J. Appl. Chem. Biotechnol. 1978, 28, 309-320.
sugar in hemicellulose-derived hydrolyzates. Moreover, (7) Lynd, L. R.; Cushman, J. H.; Nichols, R. J.; Wyman, C. E.
the expected benefit of β-glucosidase supplementation is Fuel ethanol from cellulosic biomass. Science 1991, 251,
captured by adding structure, thereby rendering the 1318-1323.
model potentially useful for predicting the hydrolysis (8) Wyman, C. E. Ethanol from lignocellulosic biomass-
performance of cellulase preparations having different Technology, economics, and opportunities. Bioresource Tech-
levels of β-glucosidase. A key limitation identified is that nol. 1994, 50, 3-16.
the effect of temperature is not sufficiently well repre- (9) McMillan, J. D.; Newman, M. M.; Templeton, D. W.;
sented in the model. Nevertheless, the model provides a Mohagheghi, A. Simultaneous saccharification and cofermen-
tation of dilute-acid pretreated yellow poplar hardwood to
solid foundation for future model refinements and has ethanol using xylose-fermenting Zymomonas mobilis. Appl.
the potential to be used for in silico optimization of the Biochem. Biotechnol. 1999, 77-9, 649-665.
hydrolysis portion of a bioethanol process. (10) Gregg, D. J.; Boussaid, A.; Saddler, J. N. Techno-economic
evaluations of a generic wood-to-ethanol process: Effect of
Acknowledgment increased cellulose yields and enzyme recycle. Bioresource
This work was funded by the Office of the Biomass Technol. 1998, 63, 7-12.
(11) Esteghlalian, A. R.; Srivastava, V.; Gilkes, N.; Gregg, D.
Program of the U.S. Department of Energy. We thank
J.; Saddler, J. N. An overview of factors influencing the
Jeffrey S. Knutsen of the University of Colorado, Boulder, enzymatic hydrolysis of lignocellulosic feedstocks. ACS Sum-
CO, for his considerable help with enzyme adsorption posium Series 769; American Chemical Society: Washington,
experiments. DC, 2000; pp 100-111.
(12) Klyosov, A. A. Trends in biochemistry and enzymology of
Notation cellulose degradation. Biochemistry 1990, 29, 10577-10585.
Ea activation energy (cal/mol) (13) Selby, K.; Maitland, C. C. The cellulase of Trichoderma
viride. Separation of the components involved in the solubi-
ET total enzyme concentration (g/kg) lization of cotton. Biochem. J. 1967, 104, 716-724.
EB bound enzyme concentration (g/kg) (14) Li, L. H.; Flora, R. M.; King, K. W. Individual roles of
EF free enzyme concentration (g/kg) cellulase components derived from Trichoderma viride. Arch.
E1B bound concentration of CBH and EG (g/kg) Biochem. Biophys. 1965, 111, 439-447.
(15) Maitland, C. Mechanism of cellulase action. Nature 1965,
E2B bound concentration of β-glucosidase (g/kg) 207, 27.
E2F concentration of β-glucosidase in solution (g/kg) (16) Caminal, G.; Lopez-Santin, J.; Sola, C. Kinetic modeling
Emax maximum mass of enzyme that can adsorb onto of the enzymatic hydrolysis of pretreated cellulose. Biotechnol.
a unit mass of substrate (g protein/g cellulose) Bioeng. 1985, 27, 1282-1290.
G glucose concentration (g/kg) (17) Huang, A. A. Kinetic studies on insoluble cellulose-cellulase
system. Biotechnol. Bioeng. 1975, 17, 1421-1433.
G2 cellobiose concentration (g/kg)
(18) Moldes, A. B.; Alonso J. L.; Parajo, J. C. Cogeneration of
Kad dissociation constant for the enzyme adsorption/ cellobiose and glucose from pretreated wood and bioconver-
desorption reaction (g protein/g cellulose) sion to lactic acid: A kinetic study. J. Biosci. Bioeng. 1999,
kir reaction rate constants (kg/g.h) 87, 787-792.
KiIG inhibition constants for glucose (g/kg) (19) Wald, S.; Wilke, C. R.; Blanch, H. W. Kinetics of the
KiIG2 inhibition constants for cellobiose (g/kg) enzymic hydrolysis of cellulose. Biotechnol. Bioeng. 1984, 26,
221-230.
KiIX inhibition constants for xylose (g/kg) (20) Converse, A. O.; Optekar, J. D. A synergistic kinetics model
K3M substrate (cellobiose) saturation constants (g/kg) for enzymatic cellulose hydrolysis compared to degree-of-
ri reaction rate (g/kg‚h) synergism experimental results. Biotechnol. Bioeng. 1993, 42,
R universal gas constant (cal/mol‚K) 145-148.
(21) Kurakake, M.; Shirasawa, T.; Ooshima, H.; Converse, A.
Rs substrate reactivity
O.; Kato, J. An extension of the Harano-Ooshima rate
S substrate concentration (g/kg) expression for enzymatic hydrolysis of cellulose to account
T temperature (K) for changes in the amount of adsorbed cellulase. Appl.
X xylose concentration (g/kg) Biochem. Biotechnol. 1995, 50, 231-241.
R constant relating substrate reactivity with de- (22) Koullas, D. P.; Christakopoulos, P.; Kekos, D.; Macris, B.
gree of hydrolysis J.; Koukios, E. G. Correlating the effect of pretreatment on
the enzymatic hydrolysis of straw. Biotechnol. Bioeng. 1992,
39(1), 113-116.
(23) Parajo, J. C.; Alonso, J. L.; Santos, V. Development of a
References and Notes generalized phenomenological model describing the kinetics
(1) Lynd, L. R. Overview and evaluation of fuel ethanol from of the enzymatic hydrolysis of alkaline-treated pine wood.
cellulosic biomass: Technology, economics, the environment, Appl. Biochem. Biotechnol. 1996, 56, 289-299.
and policy. Ann. Rev. Energy Environ. 1996, 21, 403-465. (24) Holtzapple, M. T.; Caram, H. S.; Humphrey, A. E. A
(2) Kadam, K. L.; McMillan, J. D. Availability of corn stover as Comparison of 2 empirical-models for the enzymatic-hydroly-
a sustainable feedstock for bioethanol production. Bioresource sis of pretreated poplar wood. Biotechnol. Bioeng. 1984, 26,
Technol. 2003, 18, 17-25. 936-941.
(3) Van Wyk, J. P. H. Biotechnology and the utilization of (25) Asenjo, J. A. Modeling the bioconversion of cellulose into
biowaste as a resource for bioproduct development. Trends microbial products: rate limitations. Process Biochem. 1984,
Biotechnol. 2001, 19, 172-177. 19, 217-224.
(4) Kadam, K. L.; Camobreco, V. J.; Glazebrook, B. E.; Forrest. (26) South, C. R.; Hogsett, D. A. L.; Lynd, L. R. Modeling
L. H.; Jacobson, W. A.; Simeroth, D. C.; Blackburn, W. J.; simultaneous saccharification and fermentation of lignocel-
Biotechnol. Prog., 2004, Vol. 20, No. 3 705

lulose to ethanol in batch and continuous reactors. Enzyme (39) Kim, D. W.; Kim, T. S.; Jeong, Y. K.; Lee, J. K. Adsorption
Microb. Technol. 1995, 17, 797-803. kinetics and behaviors of cellulase components on microcrys-
(27) Philippidis, G. P.; Hatzis, C. Biochemical engineering talline cellulose. J. Ferment. Bioeng. 1992, 73, 461-466.
analysis of critical process factors in the biomass-to-ethanol (40) Velkovska, S.; Marten, M. R.; Ollis, D. F.; Kinetic model
technology. Biotechnology Prog. 1997, 13, 222-231.
for batch cellulase production by Trichoderma reesei RUT
(28) Gusakov, A. V.; Sinitsyn, A. P.; Klyosov, A. A. Kinetics of
C30. J. Biotechnol. 1997, 54, 83-94.
the enzymatic hydrolysis of cellulose: 1. A mathematical
model for a batch reactor process. Enzyme Microb. Technol. (41) Ooshima, H.; Burns, D. S.; Converse, A. O. Adsorption of
1985, 7, 346-352. cellulase from Trichoderma reesei on cellulose and lignacious
(29) Gusakov, A. V.; Sinitsyn, A. P.; Klyosov, A. A. Kinetics of residue in wood pretreated by dilute sulfuric acid with
the enzymatic hydrolysis of cellulose: 2. A mathematical explosive decompression. Biotechnol. Bioeng. 1990, 36, 446-
model for the process in a plug-flow column reactor. Enzyme 452.
Microb. Technol. 1985, 7, 383-388. (42) Atkinson, B.; Mavituna, F. Biochemical Engineering and
(30) Sadana, A. Models of enzyme deactivation. In Thermosta- Biotechnology Handbook; Stockton Press: New York, 1991.
bility of Enzymes; Gupta, M. N., Ed.; Narosa: New Delhi,
1992; pp 84-93. (43) Nguyen, Q. A.; Dickow, J. H.; Duff, B. W.; Farmer, J. D.;
(31) Weimer, P. J.; French, A. D.; Calamari, T. A. Differential Glassner, D. A.; Ibsen, K. N.; Ruth, M. F.; Schell, D. J.;
fermentation of cellulose allomorphs by ruminal cellulolytic Thompson, I. B.; Tucker, M. P. NREL/DOE ethanol pilot-
bacteria. Appl. Environ. Microbiol. 1991, 57, 3101-3106. plant: current status and capabilities. Bioresource Technol.
(32) Gama, F. M.; Teixeira, J. A.; Mota, M. Cellulose morphol- 1996, 58, 189-196.
ogy and enzymatic reactivity: a modified solute exlusion (44) Tucker, M. P.; Farmer, J. D.; Keller, F. A.; Schell, D. J.;
technique. Biotechnol. Bioeng. 1994, 43, 381-387. Nguyen, Q. A. Comparison of yellow poplar pretreatment
(33) Lynd, L. R.; Weimer, P. J.; van Zyl, W. H.; Pretorius, I. S:. between NREL digester and Sunds hydrolyzer. Appl. Bio-
Microbial cellulose utilization: fundamentals and biotech- chem. Biotechnol. 1998, 70/72, 25-35.
nology. Microbiol. Mol. Biol. Rev 2002, 66, 506-577.
(34) Jackson, L. S.; Heitmann, J. A.; Joyce, T. W. Enzymatic (45) Dowe, N.; McMillan. J. SSF Experimental Protocols: Li-
modifications of secondary fiber. TAPPI J. 1993, 76, 147- gnocellulosic Biomass Hydrolysis and Fermentation, LAP-008
154. (Laboratory Analytical Procedure); National Renewable En-
(35) Puls, J.; Wood, T. M. The degradation pattern of cellulose ergy Laboratory: Golden, CO, 2001.
by extracellular cellulases of aerobic and anaerobic microor- (46) Oh, K. K.; Kim, S. W.; Jeong, Y. S.; Hong, S. I. Bioconver-
ganisms. Bioresource Technol. 1991, 36, 15-19. sion of cellulose into ethanol by nonisothermal simultaneous
(36) Lenz, J.; Esterbauer, H.; Sattler. W.; Schurz, J.; Wrents- saccharification and fermentation. Appl. Biochem. Biotechnol.
chur, E. Changes of structure and morphology of regenerated 2000, 89, 15-30.
cellulose caused by acid and enzymatic hydrolysis. J. Appl.
Polym. Sci. 1990, 41, 1315-1326. (47) Sattler, W.; Esterbauer, H.; Glatter, O.; Steiner, W. The
(37) Ohmine, K.; Ooshima, H.; Harano, Y. Study on enzymatic effect of enzyme concentration on the rate of the hydrolysis
hydrolysis of cellulose by cellulase from Trichoderma viride. of cellulose. Biotechnol. Bioeng. 1989, 33, 1221-1234.
Biotechnol. Bioeng. 1983, 25, 2041-2053.
(38) Desai, S. G.; Converse, A. O. Substrate reactivity as a Accepted for publication January 28, 2004.
function of the extent of reaction in the enzymic hydrolysis
of lignocellulose. Biotechnol. Bioeng. 1997, 56, 650-655. BP034316X

You might also like