la dinÁmica del crecimiento y producciÓn del sistema ligninolÍtica del hongo phanerochaete...

8/21/2019 LA DINMICA DEL CRECIMIENTO Y PRODUCCIN DEL SISTEMA LIGNINOLTICA DEL HONGO PHANEROCHAETE CHRY

1/9

Journal of Biotechnology 137 (2008) 5058

Contents lists available atScienceDirect

Journal of Biotechnology

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / j b i o t e c

Growth and ligninolytic system production dynamics of the PhanerochaetechrysosporiumfungusA modelling and optimization approach

J.A. Hormiga a, J. Vera b, I. Fras a, N.V. Torres Darias a,

a Biochemical Technology Group, Department of Biochemistry and Molecular Biology, University of La Laguna, 38306 La Laguna, Tenerife, Spainb Systems Biology and Bioinformatics Group, Department of Computer Science, University of Rostock, Rostock, Germany

a r t i c l e i n f o

Article history:

Received 4 April 2008

Received in revised form 27 June 2008

Accepted 2 July 2008

Keywords:

Mathematical modelling

Ligninolytic system

Growth

Optimization

Systems biotechnology

Phanerochaete chrysosporium

a b s t r a c t

The well-documented ability to degrade lignin and a variety of complex chemicals showed by the white-

rot fungus Phanerochaete chrysosporium has made it the subject of many studies in areas of environmental

concern, including pulp bioleaching and bioremediation technologies. However, until now, most of the

work in this field has been focused on the ligninolytic sub-system but, due to the great complexity of the

involved processes, less progress has been made in understanding the biochemical regulatory structure

that couldexplaingrowth dynamics, the substrate utilizationand the ligninolytic system productionitself.

In this work we want to tackle this problem from the perspectives and approaches of systems biology,

which have been shown to be effective in the case of complex systems. We will use a top-down approach

to the construction of this model aiming to identify the cellular sub-systems that play a major role in the

whole process.

We have investigated growth dynamics, substrate consumption and lignin peroxidase production of

the P. chrysosporium wild type under a set of definite culture conditions. Based on data gathered from

different authors and in our own experimental determinations, we built a model using a GMA power-law

representation, which was used as platform to make predictive simulations. Thereby, we could assess theconsistency of some current assumptions about the regulatory structureof theoverall process. Themodel

parameters were estimated from a time series experimental measurements by means of an algorithm

previously adapted and optimized for power-law models. The model was subsequently checked for qual-

ity by comparing its predictions with the experimental behavior observed in new, different experimental

settings and through perturbation analysis aimed to test the robustness of the model. Hence, the model

showed to be able to predict the dynamics of two critical variables such as biomass and lignin peroxi-

dase activity when in conditions of nutrient deprivation and after pulses of veratryl alcohol. Moreover,

it successfully predicts the evolution of the variables during both, the active growth phase and after the

deprivation shock. The close agreement between the predicted and observed behavior and the advanced

understanding of its kinetic structure and regulatory features provides the necessary background for the

design of a biotechnological set-up designed for the continuous production of the ligninolitycsystem and

its optimization.

2008 Elsevier B.V. All rights reserved.

1. Introduction

The group of white-rot wood-decaying basidiomycetes is the

mostefficient lignin degraders; theextracellular oxidative enzymes

thought to be involved in this process include an array of oxidases

and peroxidases (Vandenet al., 2006). These enzymes are responsi-

Corresponding author. Tel.: +34 922 318334; fax: +34 922 8354.

E-mail address:[email protected](N.V. Torres Darias).

URL:http://webpages.ull.es/users/sympbst/ (N.V. Torres Darias).

ble for generatinghighlyreactive (butnonspecific)free radicalsthat

affect lignin degradation. The nonspecific nature and extraordinary

oxidation potential of the peroxidases made them at first the focus

of considerable interest in the development of bioprocesses such

as fiber bleaching and the remediation of organopollutant contami-

nated soilsand effluents (Kirkand Farrell, 1987). However,currently

laccases constitute a preferredoptionfor technological applications

dueto thefact that they only require molecularoxygen forcatalysis

(Rodrguez and Toca, 2006; Baldrian, 2006).

One member of this group, the Phanerochaete chrysosporium,

is a white-rot capable of completely degrading all major compo-

0168-1656/$ see front matter 2008 Elsevier B.V. All rights reserved.

doi:10.1016/j.jbiotec.2008.07.1814
http://www.sciencedirect.com/science/journal/01681656mailto:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_3/dx.doi.org/10.1016/j.jbiotec.2008.07.1814http://localhost/var/www/apps/conversion/tmp/scratch_3/dx.doi.org/10.1016/j.jbiotec.2008.07.1814mailto:[email protected]://www.sciencedirect.com/science/journal/01681656


2/9

J.A. Hormiga et al. / Journal of Biotechnology 137 (2008) 5058 51

nents of plant cell walls including cellulose, hemicellulose and

lignin (Aitken et al., 1989; Gusse et al., 2006; Tekere et al., 2005).

Its ligninolytic system includes among others, a family of extra-

cellular lignin peroxidases (LiP) isoenzymes, for which relative

and absolute abundances are strongly influenced by growth con-

ditions (e.g., oxygen, temperature, nutrients and the presence

of some inducer compounds) (Doddapaneni and Yadav, 2005;

Sato et al., 2007). On the other hand, veratryl alcohol (VA;3,4-dimethoxybenzyl alcohol) is synthesized from glucose via

phenylalanine, 3,4-dimethoxycinnamyl alcohol and veratrylglyc-

erol by a secondary metabolic system in the first stages of the

secondary growth phase ofP. chrysosporium (Jensen et al., 1994;

Khindaria et al., 1996; Have and Teunissen, 2001), while the addi-

tion of VA to the culture leads to increased LiP production (Faison

and Kirk, 1983; Paszcynski et al., 1991; Collins et al., 1997).

As commented above, the ability to degrade lignin and a variety

of complex chemicals shown by this white-rot fungus has made

it the subject of many studies in areas of environmental concern.

Most of the research in this field deals mainly with the ligninolytic

system and a great deal of attention have been given to aspects

suchas the influence on the LiP production of the biomass, the pel-

let size, oxygen, temperature, etc. (Leisola and Schoemaker, 1988;

Ward et al., 2001). However, comparatively less attention has been

posed on the building up of a comprehensive and integrative view

of the underlying processes and regulatory structure of the growth,

substrate utilization and ligninolytic dynamics throughmathemat-

ical modelling. This happens in spite of the fact that this type of

approach becomes critical in such a complex process and it is cen-

tralto any rationalbiotechnological optimization aimedto improve

its economic feasibility (Torres and Voit, 2002).

In this work wepresent a case of top-down modelling of a com-

plex biological process aimed to describe the growth, substrate

consumption and LiP production of theP. chrysosporiumwild type

when in a set of definite culture conditions. Once the model was

built and its quality checked, it was used to predict its process

dynamics and subsequently applied to an experimental set-up of

biotechnological interest. Our results showthat the model is able topredict theobserved behaviour andit isthusused forthedesign and

optimization of a robust biotechnological set-up for the continuous

production of lignin degrading enzymes.

2. Materials and methods

2.1. Organism and culture conditions

P. chrysosporium wild type (MUCL 19343) was grown in a cul-

ture medium based onTien and Kirk (1988)modified byDosoretz

et al. (1993). The initial glucose concentration in the medium was

56 mM (10 g/liter). For the shock experiments, a pulse of veratryl

alcohol (VA; 3,4-dimethoxybenzyl alcohol; 2.6 mM) was added at180 h of culture time. Cultures were grown at 37C, without agi-

tation, in 500-ml Erlenmeyer flasks containing 150 ml of medium

(0.6 surface/volume specific ratio). Spore suspensions for inoculat-

ing the flasks were obtained by scraping the surface of 20 days old

cultures ofP. chrysosporiumgrown in solid medium and added to

10ml of sterile medium. The inoculums were gently shaken to lib-

erate spores to a liquid medium. Spore number was determined

by absorbing 660 nm in sterile water and approximately 106 ml1

were pooled and dispensed into each flask. The flasks were asep-

tically flushed (Millex-FG50 filter unit; Millipore) with filtered air

at 0.1l/min during all the experiments. Under these conditions,P.

chrysosporiumgrew largely as a mycelia mat.

After100 h fromthe start of inoculums theculturewas subjected

to a nutrient shock. It was induced by transferring the developed

mycelium into sterile water, being washed by gentle agitation, and

followedby incubation in sterile nutrient-freemedium (50mM 2,2-

dimethylsuccinate buffer, pH 4.5). After the VA pulse, more than

100U of peroxidase could be recovered in the 1015 h period fol-

lowing nutrient shock. The specific activity of the extra cellular

medium was510times higher thanthat obtainedwith standard6-

day-oldcultures. Checkingof specific properties of LiP wasobtained

after purification of enzyme as described inFras et al. (1995).

2.2. Biomass determination

Cultures were centrifuged a 3000g during 10 minand therecov-

ered mycelia were washed with 200 ml of deionized water, then

placed in tared metallic dishes, which were dried in an oven at

90 C until a constant weight was obtained (typically 2 days).

2.3. Assay of l ignin peroxidases

Lignin peroxidases (LiP) were collectedfrom culture medium by

centrifugationat 10,000gduring20 min at4 C. Supernants were

used for enzyme assays. LiP activity was determined spectropho-

tometrically at 310 nm recordingthe maximum rate of oxidation of

VA to deveraldehyde (310 =9300M1 cm1;Tien and Kirk, 1988).

Thereaction mixture contained 2 mM VA, 0.4mM H2O2and 1 nmol

of the enzyme( ofthe LiP being assumed tobe 162mM1 cm1) in50 mM 2,2-dimethylsuccinate buffer (pH 3.0). The reactants were

added to a test tube in the order given above. After the addition of

the H2O2, the reactants were vortexed and the reaction was moni-

toredfor up to 200s. Reactions were conductedat 26 C ina thermo

regulated cuvette holder. One unit (U) of activity is defined as the

amount of enzyme catalyzing the oxidation of 1mol veratryl alco-

hol min1 and activities were reported as U/L.

2.4. Mathematical modelling

The P. chrysosporium dynamics has been modelled using a

power-law representation (Voit, 2000) with the following struc-ture:

dXidt

=

j

cijj

pk=1

Xgjkk

, i = 1, . . . , nd (1)

whereXirepresents any of the nddependent variables of the model

(e.g., proteins or phospho-protein concentrations; levels of gene

expression; intermediary metabolites, etc.). Here, the biochemical

rate j is expanded as a product of a rate constant (j) and the pvariables of the system to characteristic kinetic orders (gjk), while

cijarethe stoichiometric coefficients of the system describing mass

conservation.

The maindifference between power-law models and other ODEs

models used in metabolic engineering is that kinetic orders canhave noninteger values. The use of noninteger kinetic orders relates

to the absence of data on the detailed reaction mechanisms, which

forces the modeller to condense several steps into simplified rep-

resentations (Vera et al., 2007; Savageau, 1998). Power-law models

allow the capture of complex dynamics (e.g., saturation behav-

ior, inhibition or cooperativity) by modulating the value of the

kinetic orders (Voit, 2000; Vera et al., 2007, 2008; Atkinson et

al., 2003). They have been used for long time and are currently

used in the modelling of different kinds of biochemical systems,

from metabolic systems (Alvarez-Vasquez et al., 2000, 2002, 2005;

Garcia et al., 2008) to cell signalling pathways (Vera et al., 2007,

2008) and gene networks (Atkinson et al., 2003), using published

kinetic data or quantitative time-courses of metabolites, proteins

and phospho-proteins. Since our model describes the interplay


3/9

52 J.A. Hormiga et al. / Journal of Biotechnology 137 (2008) 5058

between the regulatory and the metabolic level of the investigated

system, previous published experiences (Sevilla et al., 2005)sup-

portour choice of power-law models as a valid modellingapproach.

2.5. Parameter estimation

In recent times several optimization algorithms has been pre-

sented for model calibration in the contextof systems biology (e.g.,Polisetty et al., 2006; Balsa-Canto et al., 2008; Banga, 2008).In the

present paper a genetic algorithm was used for parameter estima-

tion. The algorithm has been adapted andoptimized for power-law

models (Vera et al., 2008).In the estimation process, each element

of the population of solutions represents a point in the parameter

value space. The initial populationof solutionsis generated through

a random exploration of the search space, which is defined using

feasible intervals of values for the variables. The best individuals

of the population are selected in the considered iteration based on

the value of the following objective function:

FObj =1

nvarntp

nvarj=1

ntpi=1

(Xj(ti) Xexp

j (ti))

2

j(ti)Xexp

j (ti)

2 (2)

where nvaris the number of variables monitored, ntpthe number of

time points where each variable was measured. In turn,Xj(ti) isthe

predicted value for thejth variable at the ith time point obtained

after numerical integration of the solution, while Xexp

j (ti) is the

value of thejth observable variable at theith time point measured

in the experiment, andj(ti) is the standarddeviation.Withthe aim

of improving the speed and finding a better value for the objective

function, the genetic algorithm previously described was comple-

mented with a additional fast-climbing stochastic method, which

allows local searches for the best solutions in eachgeneration, eval-

uating random points closeto eachsolution (ina inner fixedradius).

In order to avoida premature convergence, the amount and quality

of theseevaluations were selected to maintainthe greater diversity.

Thestoppingcriterion is based on either the previously establishedmaximum number of iterations or the minimum level of satisfac-

tion for the objective function. The computing time for each model

estimation ranged from 6 to 10h. The algorithm was implemented

in Matlab 7.1 (R14) (The Mathworks, Inc., Natick, Massachusetts,

USA) running under a standard PC computer (Pentium IV; 2.6 GHz,

1 GB RAM memory).

3. Results

3.1. Model development

Fig.1 shows the biochemical processes considered in our model.

Here, glucose (Gluc) serves for the synthesis of biomass (X, rate

equation V10) and to support other processes other from thebiomass synthesis (V12); monophenol (Mph) working as an alter-

native carbon source (V8). A term accounting for the biomass

degradation is also considered in the model (V11). Mph comes from

the biotransformation of diphenol (Dph), in a reaction catalyzed

by the activated lignin peroxidase (LiP*;Hammel et al., 1985)but

modulated by veratryl alcohol VA (V6). VA is treated as an oper-

ating variable in our model. Its consumption has been described

(Khindaria et al., 1995)but its concentration is more that 20 times

lesser than Gluc, thus making negligible its role as substrate. In

our model we therefore assume that the ratio of absorption or

degradation of VA by the biomass is small enough to consider its

concentration constant at least for the time interval of our exper-

iments. The lignin peroxidase (LiP) synthesis is activated by VA

(Khindaria et al., 1995)but inhibited by high levels of Gluc (V3;

Fig.1. Proposed kinetic model for the processes responsible of LiP growth and pro-

duction inP. chrysosporium. Dashed a rrows represent activation interactions, while

dashed lines ending in a bar represent inhibition ones. Solid lines represent rateprocesses. Thesymbols and refer to synthetic anddegradation processes respec-

tively. The clock symbol, , represents a time-delay in process which value is a

parameter estimated during the model calibration.

Barclay et al., 1993; Gaoet al., 2005). The natural degradation of the

proteinis also considered (V5). The lignin peroxidase (LiP) becomes

active (LiP*) through its interaction with H2O2 once in the extra

cellular medium (V2;Ollikka et al., 1998; Brck et al., 2003)and is

deactivated during the process of diphenol transformation that it

catalyzes (V6;DePillis and Ortiz de Montellano, 1989; Chung and

Aust, 1995b). The H2O2 used in the lignin peroxidase activation

is generated in a reaction catalyzed by the P. chrysosporium pro-

tein oxidase, Ox (V1). In this process O2 is constant and thus it is

not considered a system variable. Finally, the model also consid-ered the synthesis of Ox (V4) as well as its degradation (V7), being

the former activated by VA but inhibited by glucose (Belinky et al.,

2003). The processes involving the Ox, LiP and LiP* activities here

considered occurs in the extra cellular medium. In particular the

Ox and LiP reaction synthesis represented in the model includes

their excretion transportprocess (it is assumed thatthe transporta-

tion time is negligible when compared with the synthesis one).

Also, a time delay was considered in the synthesis of biomass from

glucose (Delay1; Delay2; V10) and therefore included in the cor-

responding rate equation. Some authors (Nikolov et al., 2007, in

press) suggest that the time delay in biochemical system models

can emerge as a consequence of either the intrinsic discrete time

that some processes take to be accomplished (for instance,the syn-

thesis of mRNA) or the modelling approach used, in which complexsequences of events, not represented in detail, provoke the emer-

gence of an apparenttime delay. In our model, time delay is related

to the second possibility and accounts for the simplified modelling

of processes such as the glucose diffusion in the system or the

biotransformation of glucose in biomass. In addition to these pro-

cesses, the model considers another reaction (V9; see below) that

describes the dynamics of the Gluc degradation itself, but having

a different meaning and value (9 being different from 10) fromV10. A similar situation happens withV3andV4that describes dif-

ferent biosynthetic processes that differs in is rate constant. Thesystems model is thus composed by 8 variables and 11 biochemi-

cal reactions and processes. Time delay in the synthesis of biomass

with glucose was modelled using a two-equation linear chain trick

(Macdonald,1978). The system appears highly regulated and there-


4/9


fore prone to show a counterintuitive behavior, thus making clear

the need of a quantitative and modelling-based approach, in order

to attain a proper understanding of it.

As a whole the model is a semi-structured one, in which pro-

cesses developing at several different levels are represented. It can

be seen that together with catabolic and anabolic reactions steps

there are phenomenological descriptions of protein synthesis and

degradation and growth processes. This modelling strategy seemsthe most suitable in the present case due to the complexity of the

system and also in view of the studys purposes. The correspond-

ing individual kinetic rate equations, in power-law expressions,

describing the above referred processes are as follows:

V1 = 1 Ox V7 = 7 Ox

V2 = 2 LiPH2O2 V8 = 8 Mphg7X

V3 = 3 Glucg1 VAX V9 = 9 GluTD

g6XV4 = 4 Gluc

g1 VAX V10= 10 GluTDg6X

V5 = 5 LiP V11= 11X2

V6 = 6 VAg2 Dphg3 LiPg4 V12= 12 Gluc

(3)

Depending on the available information and quantitative data,

some processes were described as simplified rate equations rep-

resenting several aggregated processes (e.g., protein synthesis,protein degradation or biomass synthesis) while in other cases a

precise mathematical description on the enzymatic catalyzedreac-

tions was possible (e.g., the LiP* and Ox catalyzed reactions). These

equations, combined withthe mass balance equations of the model

systems and the time delay expression, yield the equations for the

model:

dLiP

dt = V2 V6

dGluc

dt = V9 V12

dLiP

dt = V3 V2 V5 + V6

dMph

dt = 2(V6 V8)

d Ox

dt = V4 V7

dDph

dt = V6

dXdt

= V8 + V10 V11d H2O2

dt = V1 V2

dDelay1dt

= K(Gluc Delay1)

dDelay2dt

= K(Delay1 Delay2)

GlucTD = Delay2

(4)

The equations involving the delay termaccountfor a distributed

time-delay in the biosynthesis through glucose consumption. This

delay is modelled using the linear chain trick (Macdonald, 1978).

In this approach, the delay is described using time-dependent fic-

titious variables (in our case, Delay1 and Delay2); the features of

the distributed time-delay (average value and standard deviation)

depend on the number of fictitious variables used and the valueassigned to the rate constant K. The delayed value of glucose is

described by the variable GluTD and used in Eq.(3).We notice that

more sophisticated strategies are possible to introduce time-delay

in ODE models (see for example,Mocek et al., 2005).However, our

initial analysis indicated that the simple strategy used to model

time-delay was enough in our case study.

3.2. Parameter estimation

The process that finally leads to the formalized kinetic model

started with the analysis of the experimental data. The available

data were normalized with respect to the maximum values in

order to avoid numeric problems in the parameter estimation

procedure. In the data fitting and parameter estimation assays dif-

Fig. 2. Data fitting of the selected solution for the measured variables. The first

part (0 to 100h, arrow b) shows the time evolution of the biomass and glucose

concentrations(the overallR2 ofthisphasewas0.773).Thesecondpart(100to250 h)

shows the time courses of the variables involved in the induction and excretion of

LiP*. At 180 h (arrow b) a pulse of VA was added (the overallR2 of this phase was

0.824). Points andcontinuous linesdescribe theexperimentaldata whilethe dashed

ones represent the model predictions (error bars represent the standard error).

ferent time series of experimental measurements of the species

and intermediates represented were used. Parameters were esti-

mated using a genetic algorithm for those time intervals in which

convergence and numerical stability of the values of the param-

eters where achieved. A strategy for model selection was used

to decide on the most suitable structure for the model with a

reduced number of parameters. Along the model building process

we assayed different sets of kinetic configurations, differing both in

their reactionand regulatory structure. The generalstrategy used to

discriminate among them was toselect the onethat, with themini-

mum number of variables, reactions and interactions involved, was

the best in describing the observed behavior of our experimental

assays.Furthermore, we applied a strategy to reduce the number of

kinetic orders to be estimated from the quantitative data avoiding

thus identifiabilityissues. Towards this end, we didnot consider all

kinetic ordersto be noninteger atthe same time, butdefinedan iter-

ative process in whichan increasing number of kinetic orders were

allowed to be variable. We selected the first (simplest) model that

allows an appropriate fit to experimental data with no significant

improvement in the next (more complex) model tested.

The chosen solution was that one showing the best value of the

parameter estimation objective function, that is the set of reactions

and regulatory interactions and the parameters values yielding the

best fit with the time series data. This strategy of model building is

the standard one in Systems Biology.

The calculated parameter values were estimated by data fit ofthe experimental data available and are summarized inTable 1.

The model trajectories obtained with the chosen solution are

depicted inFig. 2where we compare the experimental data with

thedata fittingproduced by the model.The first part ofFig.2 shows

the concentration of biomass and glucose up to 100h time, before

the culture was washed (arrow a) as described in Section2.

It can be seen that the model is able to reproduce this stage

system behaviour. The second part of the Fig. 2shows the experi-

mental and model predicted evolution of the main variables from

100 to250h timeof culture time. At180 h time, oncethe initial glu-

cose concentration (56 mM) was depleted a VA pulse was added to

the medium. It can be seen that the model fitting and the observed

general pattern dynamics agrees quite well. In particular some crit-

ical qualitative features of the systems behaviour, such as previous


5/9


Table 1

Parameter values in the selected solution and the corresponding standard deviation

Parameter 1 2 3 4 5 6Value 0.2000.007 0.3150.078 0.0990.013 0.5210.116 0.0420.008 0.1030.021

Parameter 7 8 9 10 11 12Value 0.0660.017 0.1530.047 0.7250.001 0.5590.033 0.8140.118 0.0150.001

Parameter 1 2 3 g4 g5 g6Value 0.02750.004 1.1180.280 1.0130.201 1.9050.437 1.1760.031 0.5590.098

Parameter g7 K

Value 0.7050.113 0.6030.088

In the present case, given that the estimation algorithm is a genetic one, the standard deviation corresponds to the best solution set which accounts for the 10% of the total

data.i andKhave dimensions of time1 whilegiare dimensionless.

absence of LiP synthesis (P. chrysosporium produces LiP when in

conditions of nitrogen and carbon deprivation) and it appearance

after the VA pulse and the growth kinetics, shows close agreement.

Moreover, the dynamics of the glucose and LiP* concentration are

also well described by the model simulation.

We also tested the model predictions about biomass synthesis

by comparing themwith dataobtained in a continuous reactorwith

10g/l glucose input flux concentration (seeSheldon et al., 2008for

details). These conditions are significant in a biotechnological set-up and a have been assessed in many previous studies ( Barclay et

al., 1993; Gao et al., 2005).Fig. 3shows the results obtained.

It is observed a close agreement between both types of data.

Also, it is clear that biomass production shows a consistent delay

with respect to the glucose input until it reaches a stationary sit-

uation. In this state a balance between the glucose input and its

consumption is attained. These model predictions are in agree-

mentwith other observedsystemresponsewhen glucose is present

(Dosoretz et al., 1993)and with the influence of glucose on the P.

chrysosporiumgrowth (Barclay et al., 1993; Kirk et al., 1976; Gao et

al., 2005).

3.3. System dynamics

Once the model was built and its reliability assessed we car-

ried out some explorations aimed predict the system behaviour

when in conditions different from those used at model building.

These explorations were also aimed to define a set of experimental

conditions that would optimize the ligninolityc system produc-

Fig. 3. Comparison of predicted and measured biomass growth dynamics of P.

chrysosporium in a continuous culture with glucose as carbon source. Continuous

lines correspond to experimental growth curves obtained by Sheldon et al. (2008)

for initial spore concentrations of 0.5 (black circles) and 1 (blue squares) million

spores atan airflow rateof 2.8l/min anda glucose concentrationof 10g/l(error bars

represent the standard error; seeSheldon et al., 2008for details).

tion. Accordingly,the model wasrun in conditions simulating those

where LiP* is produced.

Figs. 4 and 5 show the evolution of the model variables in differ-

ent operating system conditions. InFig. 4Athe system was carried

out at a constant concentration of Dph (and VA) instead of glu-

cose as the main carbon source. The corresponding simulation

where the Dph input flux was kept constant is showed in Fig. 4B.

Another type of simulation in which simultaneous, constant fluxes

of Gluc and Dph were used to feed the system, is shown in Fig. 5.All together these predictions are qualitatively well in agreement

with the experimentally observed behaviour (seeDosoretz et al.,

1993; Barclay et al., 1993); particularly those regarding the dynam-

icsand the significant production of LiP*, biomass, LiPand H2O2.All

Fig. 4. Predicted variables dynamic in P. chrysosporium culture at constantDph flux

or concentration.A: Dynamics of themodel variablesat constantDph concentration

(0.5 normalized units). B: Dynamics of some model variables at constant Dph input

flux (104 normalized units/h). In both cases the experimental conditions included

a constant concentration of VA (0 and 5 normalized units, respectively).


6/9


Fig. 5. Predicted variables dynamic in P. chrysosporium culture at constant Dph

and glucose flux. The glucose and Dph input fluxes were kept constants at

(0.1 normalizedunits/h)and at constantconcentrationof VA (0.2normalized units).

of them reach a stationary state in the continuous operating sys-

tem; this feature being particularly useful for optimization studies.

Globally considered, these results constitute a pragmatic, a postericonfirmation of the model reliability.

3.4. Biotechnological design and optimization studies

Based on the presented model we were able to explore the opti-

mum conditions of a biotechnological set-up devised to efficiently

degrade lignin or alternatively, high pollutants compounds such

as xenobiotic compounds (Field et al., 1993),toxic effluents (Banat

et al., 1996) and biobleaching of Kraft pulp (Moreira et al., 1997).

For this purpose we have considered a model system, consisting

in fungal bioreactor where a vessel containing P. chrysosporiumis

allowed to grow in a continuous operation mode. In this system a

constant input/output flux guarantees a constant reaction volume;

the input flux carrying constantconcentrations of the suitable sub-strate (Gluc) as well as Dph as model compound, while the output

flux contains the steady state product concentrations, biomass and

Mph.

The corresponding model system equations have the following

form:

dLiP

dt = V2 V6

dGluc

dt = VinGluc V9 V12

dLiP

dt = V3 V2 V5 + V6

dMph

dt = 2 V6 V8

d Ox

dt = V4 V7

dDph

dt = VinDph V6

dX

dt = V8 + V10 V11d H2O2

dt = V1 V2

(5)

where VinGluc

andVinDph

represent the constant input fluxes. In this

biotechnological set-up it is assumed that the diffusion rates for

both,Dph and Gluc are high in relation to that of the systems oper-

ation. This is due to its typical dimensions and the flows dynamics.

We have explored in these conditions the influence of the glucose

(VinGluc

) and Dph (VinDph

) input fluxes on the biomass and ligninolityc

system production. Other relevant conditions of the system were

selected from the previous studies in order to guarantee a stable

stationary state and maximum ligninolityc expression, namely an

initial concentration of VA (0.2 normalized units) and a fermenta-

tion time long enough to allow the system to attain a steady state

(1000 h).

Fig. 6. Influence of the glucose (VinGluc

) and diphenol (VinDph

) input fluxes on the P.

chrysosporiumbiomass production (X) and Dph degradation rate (VDph

). A. Values

of biomass production (X). B. Values of diphenol degradation rate ( VinDph

). In both

cases the system was run at constant concentration of VA (0.2 normalized units)

and values computed at 1000 h of simulation time, once the system reached the

steady state.

Fig. 6shows that maximum levels of biomass (X) and degrada-

tion rate (VDph

, which is equal to V6; seeFig. 1and Eq.(4))can be

attainedwhen VinGluc

and VinDph

areat themaximumvalues(1 normal-

ized units/hour). It is worthto note however, that the VDph

remains

at very low values and almost unaffected for the whole range of

VinGluc

until theVinDph

reaches values between 102 to 101 units/per

hour (Fig. 6B). After this stage, where the VDph

is about 0.2 units,

it is when the increase of the glucose flux can cause the degrada-

tion rate to increase about two times.Fig. 6Ashows the evaluation

of the biomass production, showing a similar pattern. Here VinGluc

does not affectthe biomass synthesis forthe lower valueranges butwhen it is above 103, even for low values ofVin

Dphthe biomass can

reachalmost the maximum value. A significant increase of biomass

concentration is observed when theVinDph

is above 103 and then a

synergistic effect is observed for simultaneous values of bothfluxes

above 101.

From all the above it can be concluded that for low values of

VinGluc

in absence of glucose, it is necessary an intense VinDph

in order

to maintain a satisfactory concentration of biomass in the sys-

tem. Furthermore, under these conditions diphenol degradation

and biomass concentration saturate for values ofVinDph

higher than

0.1. Thus, in absence of glucose, the growthand the degradation rate

are quite limited.VinDph

is sensitive tothe glucose income(VinGluc

)only

for values of glucose flux higher than0.01. Furthermore,the produc-


7/9


tivity of the systemin termsof amount of degradeddiphenolwould

increase only for a high sustained influx of glucose (VinGluc

> 0.01).The ideal setting up of the system (as expected) is for high incomes

of both glucose and diphenol.

This information is useful in order to look for an optimal design

for an industrial bioreactor because it illustrates the keyparameters

and limiting factors of the system. The optimal design of the sys-

tem would be a trade-off between maximal diphenol degradation

Fig. 7. System sensitivities. A:S(VDph

,VinGluc

): influence ofVinGluc

on VDph

. B: S(VDph

,VinDph

): influence ofVinDph

on VDph

. C: S(X,VinGluc

): influence ofVinGluc

on Biomass. D:S(X,VinDph

):

influence ofVin

Dph on Biomass. E:S(V

Dph , VA): influence of VA on V

Dph .


8/9


withlow concentrations of soluble diphenoland minimal operation

costs of the plant (which implies minimization of theincoming flux

of Gluc and concentration of VA).

4. Discussion

Although extensive research has been done previously on the

biochemistry and the enzymatic activities of this fungus, very littleinformation is however available on this fungus growth kinetics

and nutrient consumption, within a continuously operated sys-

tem (see Sheldon et al., 2008). In this work we have built up a

model of the P. chrysosporium biomass and ligninolityc system

production. Through a top-down, iterative process of data anal-

ysis and parameter estimation we have developed a power-law

kinetic model in GMA version that was able to reproduce the sys-

tem dynamic features both qualitatively and quantitatively. Based

on this model description, we carried out an optimization study

aimed to unravel the conditions that, in a suitable, continuous

culture biotechnological set-up, allowed the system to attain a sta-

ble stationary state where the biomass and degradation rate can

reach a maximum. These conditions showed to be when there

is a simultaneous input flux of glucose and Dph in presence ofVA.

The proper design and operation of any microbial-based

biotechnological process requires the quantitative description of

thevariables relevantfor the kineticsof thesystem.Oncethis infor-

mation is available, it will be possible to derive an optimal process

design and to attain its optimal operation. Accordingly, and in order

to verify the quality and robustness of the model, we performed

a system sensitivity analysis (Siljak, 1969; Frank, 1978). We were

interested in the robustness of the system (changes in the biomass

andDhp concentrations as well as their degradation fluxes, V11and

V6, respectively) against changes in theVinGluc

or VinDph

(represented

by the parametersDphand Gluc).In this context, system sensitivities are defined as the ratio of a

relative changein a dependentvariableZi (concentrations or fluxes)to a relative change in a rate constant, j or a given variable. Theycan be determined by differentiation of the explicit steady state

solution:

S(Zi, j) =

Zij

jZi

0

=(logZi)

(log j)

where jstandsfor therate constants of the variable. The subscript0 refers to the steady state. Here robustness is especially important

in the case ofVinDph

: Dph is an industrial residue and therefore per-

turbations in the properties of its incoming flux are expected in a

real set-up of the bioreactor.Fig. 7illustrates the results of these

analyses.

What it is observed is that most of the computed sensitivitiesare rather small with values ranging from 0 to 1 within the inter-

val ofDph and Gluc analyzed values. This implies that the keyvariables and fluxes of the designed biotechnological system are

robust enough against perturbations in the values of the parame-

ters considered. Thus, the system proved to be robust and flexible

enough to maintain a fairly high efficiency, even when working at

conditions far from the optimal. AlsoFig. 7Eshows that the sys-

tem seems to be sensitive to changes in the concentration of VA.

Thus, the VA concentration is a key parameter to be controlled

in the industrial bioreactor set-up since changes in its value have

significant effects on the systems properties. Further studies will

refine this preliminary optimization approach by using a system-

atic, mathematically based optimization method (Marin-Sanguino

et al., 2007a,b).

Acknowledgements

The authors acknowledge discussions with Dr. Daniel Guebel

(Biotechnology Counseling Services, Buenos Aires, Argentina) and

Professor M.A. Falcn (Microbiology Department, University of La

Laguna). This work was supported by the Spanish Ministry of Edu-

cation and Science, research grant no. BIO-2002-04157-C02-02.

References

Alvarez-Vasquez, F., Gonzlez-Alcn,C., Torres,N.V., 2000. Metabolismof citric acidproductionbyAspergillus niger:modeldefinition, steady-stateanalysisand con-strained optimization of citric acid production rate. Biotechnol. Bioeng. 70 (1),82108.

Alvarez-Vasquez, F., Cnovas, M., Iborra, J.L., Torres, N.V., 2002. Modeling, optimiza-tion and experimental assessment of continuous l-()-carnitine production byEscherichia colicultures. Biotechnol. Bioeng. 80 (7), 794805.

Alvarez-Vasquez, F., Sims, K.J., Cowart, L.A., Okamoto, Y., Voit, E.O., Hannun, Y.A.,2005. Simulation and validation of modelled sphingolipid metabolism in Sac-charomyces cerevisiae. Nature 433 (7024), 425430.

Aitken, M.D., Venkatadri, R., Irvine, R.L., 1989. Oxidation of phenolic pollutants bya lignin degrading enzyme from the white-rot fungusPhanerochaete chrysospo-nium. Water Res. 23, 443450.

Atkinson, M.R., Savageau, M.A., Myers, J.T., Ninfa, A.J., 2003. Development of geneticcircuitry exhibiting toggle switch or oscillatory behavior in Escherichia coli. Cell

113 (5), 597607.Baldrian, P., 2006. Fungal laccases: occurrence and properties. FEMS Microbiol. Rev.30, 215242.

Balsa-Canto, E., Peifer, M., Banga, J.R., Timmer, J., Fleck, C., 2008 Mar 24. Hybridoptimization method withgeneralswitching strategyfor parameter estimation.BMC Syst. Biol. 2, 26.

Banat, I.M., Nigam, P., Singh, D., Marchant, R., 1996. Microbial decolourisation oftextile-dye-containing effluents: a review. Bioresource Technol. 58, 217227.

Banga, J.R., 2008. Optimization in computational systems biology. BMC Syst. Biol. 2,47.

Barclay, C.D.,Legge,R.L., Farquhar,G.F., 1993. Modellingthe growthkineticsofPhane-rochaetechrysosporium in submergedstaticculture.Appl. Environ.Microbiol.59,18871892.

Belinky, P.A., Flikshtein, N., Lechenko, S., Gepstein, S., Dosoretz, C.G., 2003. Reactiveoxygen species and induction of lignin peroxidase in Phanerochaete chrysospo-rium. Appl. Environ. Microbiol. 69, 65006506.

Brck, T.B., Gerini, M.F., Baciocchi,E., Harvey, P.J., 2003. Oxidation of thioanisole andp-methoxythioanisole by ligninperoxidase: kinetic evidence of a direct reactionbetween compound II and a radical cation. Biochem. J. 374, 761766.

Chung, N., Aust, S.D., 1995b. Inactivation of lignin peroxidase by hydrogen peroxideduring the oxidation of phenols. Arch. Biochem. Biophys. 316, 851855.

Collins, P.J., Field, J.A., Teunissen, P., Dobson, A.D.W., 1997. Stabilization of ligninperoxidases in white rot fungi by tryptophan. Appl. Environ. Microbiol. 63,25432547.

DePillis, G.D., Ortiz de Montellano, P.R., 1989. Substrate oxidation by the hemeedge of fungal peroxidases. Reaction ofCoprinus macrorhizus peroxidase withhydrazines and sodium azide. Biochemistry 28, 79477952.

Doddapaneni, H., Yadav, J.S., 2005. Microarray-based global differential expressionprofiling of P450 monooxygenases and regulatory proteins for signal transduc-tion pathwaysin thewhite rot fungus Phanerochaete chrysosporium. Mol. Genet.Genom. 274, 454466.

Dosoretz, C.G., Rothschild, N., Hadar, Y., 1993. Overproduction of lignin peroxidaseby Phanerochaete chrysosporium (BKM-F-1767)under nonlimiting nutrient con-ditions. Appl. Environ. Microbiol. 59, 19191926.

Faison, B., Kirk, T.K., 1983. Relation between lignin degradation and productionreduced oxygen species by Phanerochaete chrysosporium. Appl. Environ. Micro-biol. 46, 11401145.

Field, J.A., de Jong, E., Feijoo-Costa, G., de Bont, J.A.M., 1993. Screening for ligni-nolytic fungiapplicableto thebiodegradationof xenobiotics.TrendsBiotechnol.11, 4449.

Frank, P.M., 1978. Introduction to System Sensitivity Theory. Academic Press, NewYork.

Fras, I., Trujillo, J., Romero, J., Hernndez, J., Prez, J., 1995. Lignan models asinhibitors of Phanerochaete chrysosporium lignin peroxidase. Biochimie 77,707712.

Garcia, J., Shea, J., Alvarez-Vasquez, F., Qureshi, A., Luberto, C., Voit, E.O., Del Poeta,M., 2008. Mathematical modeling of pathogenicity ofCryptococcus neoformans.Mol. Syst. Biol. 4, 183.

Gao, D.W., Wen, XHand, Qian, Y., 2005. Effect of nitrogen concentration in culturemediums on growth and enzyme production ofPhanerochaete chrysosporium. J.Environ. Sci. (China) 17, 190193.

Gusse, A.C., Miller, P.D., Volk, T.J., 2006. White-rot fungi demonstrate first biodegra-dation of phenolic resin. Environ. Sci. Technol. 40, 41964199.

Hammel, K.E., Tien, M., Kalyanaraman, B., Kirk, T.K., 1985. Mechanism of oxidative Calpha-C beta cleavage of a lignin model dimer by Phanerochaete chrysosporiumligninase. Stoichiometry and involvement of free radicals. J. Biol. Chem. 260,

83488353.


9/9


Have, R., Teunissen,J.M., 2001. Oxidativemechanisms involvedin lignindegradationby white-rot fungi. Chem. Rev. 101, 33973413.

Jensen, K.A., Evans, K.M., Kirk, T.K., Hammel, K.E., 1994. Biosynthetic pathway forveratryl alcohol in the ligninolytic fungus Phanerochaete chrysosporium. Appl.Environ. Microbiol. 60, 709714.

Khindaria, A., Yamazaki, I., Aust, S.D., 1995. Veratryl alcohol oxidation by ligninperoxidase. Biochemistry 34, 1686016869.

Khindaria,A., Yamazaki, I.,Aust,S.D., 1996. Stabilizationof theveratryl alcohol cationradical by l ignin peroxidase. Biochemistry 35, 64186424.

Kirk,T.K., Connors,W.J.,Zeikus,J.G., 1976.Requirement fora growth substrateduring

lignin decomposition by two wood-rotting fungi. Appl. Environ. Microbiol. 32,192194.

Kirk, T.K., Farrell, R.L., 1987. Enzymatic combustion: the microbial degradation oflignin. Ann. Rev. Microbiol. 41, 46655505.

Leisola, M.S., Schoemaker, H.E., 1988. Lignin degradation. Trends Biochem. Sci. 13(3), 84.

Marin-Sanguino, A., Voit, E.O., Gonzalez-Alcon, C., Torres, N.V., 2007a. Optimizationof biotechnological systems through geometric programming. BCM Theor. Biol.Med. Model. 4, 3850.

Macdonald, N., 1978. Time Lags in Biological Models. Springer, Heidelberg.Moreira, M.T., Feijoo, G., Sierra-Alvarez, R., Lema, J.M., Field, J.A., 1997. Biobleaching

of oxygen delignified kraftpulp by several whiterot fungal strains.J. Biotechnol.53, 237251.

Nikolov, S., Vera, J., Kotev, V., Wolkenhauer, O., Petrov, V., 2007. Dynamic propertiesof a delayed protein cross talk model. Biosystems 91, 5168.

Nikolov, S., Vera, J., Rath, O., Kolch, W., Wolkenhauer, O., in press. The role ofinhibitory proteins as modulators of oscillations in NFkB signalling. IET Syst.Biol.,doi:10.1111/j.1472-8206.2008.00575.x.

Ollikka, P., Harjunp, T., Palmu, K., Mntsl, P., Suominen, I., 1998. Oxidation ofCrocein Orange G by lignin peroxidase isoenzymes. Kinetics and effect of H2O2.Appl. Biochem. Biotechnol. 75, 307321.

Paszcynski, A., Pasti, M.B., Goszczynski, S., Crawford, D.L., Crawford, R.L., 1991. Newapproach to improve degradation of recalcitrant azo dyes by Streptomyces spp.andPhanerochaete chrysosporium. Enzyme Microb. Technol. 13, 378384.

Polisetty,P.K., Voit,E.O.,Gatzke,E.P., 2006.Identificationof metabolicsystem param-eters using global optimization methods. Theor. Biol. Med. Model. 3, 4.

Rodrguez, Couto S., Toca Herrera, J.L., 2006. Industrial and biotechnological appli-cations of laccases: a review. Biotechnol. Adv. 24 (5), 50 0513.

Sato, S., Liu, F., Koc, H., Tien, M., 2007. Expression analysis of extracellular pro-teins from Phanerochaete chrysosporium grown on different liquid and solidsubstrates. Microbiology 153, 30233033.

Savageau, M.A., 1998. Development of fractal kinetic theory for enzyme-catalysedreactions and implications for the design of biochemical pathways. Biosystems47 (12), 936.

Sevilla, A., Vera, J., Daz, Z., Cnovas, M., Torres, N.V., Iborra, J.L., 2005. Design ofmetabolic engineering strategies for maximizing l-()-carnitine production byEscherichia coli. Integration of the metabolic and bioreactor levels. Biotechnol.Prog. 21 (2), 329337.

Marin-Sanguino, A., Voit, E.O., Gonzalez-Alcon, Torres, N.V., 2007b. Optimizationof biotechnological systems through geometric programming. BCM Theor. Biol.Med. Model. 4, 38.

Mocek, W.T., Rudnicki, R., Voit, E.O., 2005. Approximation of delays in biochemicalsystems. Math Biosci. 198 (2), 190216.

Sheldon, M.S., Mohammed, K., Ntwampe, S.K.O., 2008. An investigation of biphasicgrowth kinetics for Phanerochaete chrysosporium(BKMF-1767) immobilised ina membrane gradostat reactor using flow-cells. Enzyme Microb. Technol. 42,353361.

Siljak, D.D.,1969. NonlinearSystem. TheParameter Analysis and Design. Wiley,NewYork.

Tekere, M., Read, J.S., Mattiasson, B., 2005. Polycyclic aromatic hydrocarbonbiodegradation in extracellular fluids and static batch cultures of selected sub-tropical white rot fungi. J. Biotechnol. 115, 367377.

Tien,M., Kirk,T.K.,1988.Lignin peroxidase ofPhanerochaetechrysosporium. MethodsEnzymol. 161, 238 249.

Torres, N.V., Voit, E.O., 2002. Pathway Analysis and Optimization in Metabolic Engi-neering. Cambridge University Press.

Vanden, Wymelenberg A., Minges, P., Sabat, G., Martinez, D., Aerts, A., Salamov, A.,Grigoriev, I., Shapiro, H., Putnam, N., Belinky, P., Dosoretz, C., Gaskell, J., Kersten,P., Cullen, D., 2006. Computational analysis of the Phanerochaete chrysosporium

v2.0genomedatabaseand massspectrometryidentification of peptidesin ligni-nolyticculturesrevealcomplex mixturesof secretedproteins.FungalGenet.Biol.43, 343356.

Vera, J., Bachmann, J., Pfeifer, A.C., Becker, V., Hormiga, J.A., Torres Darias, N.V., Tim-mer, J., Klingmller, U., Wolkenhauer, O., 2008. A systems biology approach toanalyse amplification in the JAK2-STAT5 signalling pathway. BMC Syst. Biol. 2,38.

Vera, J., Balsa-Canto, E., Wellstead, P., Banga, J.R., Wolkenhauer, O., 20 07. Power-lawmodels of signal transduction pathways. Cell. Signal. 19, 15311541.

Voit, E.O., 2000. Computational Analysis of Biochemical Systems. A Practical Guidefor Biochemists and Molecular Biologists. Cambridge University Press.

Ward, G., Hadar, Y., Bilkis, I., Konstantinovsky, L., Dosoretz,C.G., 2001. Initial steps offerulic acidpolymerization bylignin peroxidase. J. Biol.Chem.276,1873418741.
http://dx.doi.org/10.1111/j.1472-8206.2008.00575.xhttp://dx.doi.org/10.1111/j.1472-8206.2008.00575.x

la dinÁmica del crecimiento y producciÓn del sistema ligninolÍtica del hongo phanerochaete...

Documents