Biochemical Engineering Journal 99 (2015) 156–166

Contents lists available at ScienceDirect

Biochemical Engineering Journal

journa l homepage: www.e lsev ier .com/ locate /be j

Kinetic study of butanol production from various sugars byClostridium acetobutylicum using a dynamic model

Francesca Raganati a, Alessandra Procentesea, Giuseppe Olivieri a,b,∗, Peter Götzc,Piero Salatinoa, Antonio Marzocchellaa

a Dipartimento di Ingegneria Chimica, Dei Materiali e Della Produzione Industriale – Università Degli Studi di Napoli Federico II, P.le V. Tecchio 80, 80125Napoli Italyb Bioprocess Engineering, Wageningen University, AlgaePARC, P.O. Box 16, 6700 AA Wageningen, The Netherlandsc Department of Life Sciences and Technology/Bioprocess Engineering, Beuth University of Applied Sciences Berlin, Seestrasse 64, 13347 Berlin Germany

a r t i c l e i n f o

Article history:Received 27 July 2014Received in revised form 25 February 2015Accepted 1 March 2015Available online 28 March 2015

Keywords:Substrate inhibitionProduct inhibitionClostridium acetobutylicumDynamic modellingBiokineticsCOPASI

a b s t r a c t

This paper presents a kinetic dynamic model of acetone–butanol–ethanol production by Clostridiumacetobutylicum DSM 792 developed with the biochemical networks simulator COPASI. This model isan evolution of previous models described in the literature, updated by including various mono-, di-,hexose and pentose sugars: glucose, mannose, fructose, sucrose, lactose, xylose and arabinose. The kineticrelationships of uptake of substrate, butanol production, cell growth and cell death are also included.

The batch fermentation tests were carried out at an initial sugar concentration ranging from 5 to 100 g/L.The data from the batch tests were used to assess the kinetic parameters of the model. This model gavesatisfactory results for each sugar, both in terms of simulation of fermentation – the square correlationcoefficient of metabolite concentrations, calculated by comparing experiments and simulations, rangedbetween 0.87 and 0.925 – and of comparison with the models reported in the literature.

The effects of mono-, di-, hexose and pentose sugars on the growth and production of metabolites,including acids and solvents, were reviewed according to the proposed model. The low fermentationperformance measured for xylose and lactose were interpreted taking into account the sugar uptake, theacid production and the hydrolysis path.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

Butanol is an energy carrier with remarkable features –hydrophobicity, high energy density, possibility to be used in thecurrent internal combustion engines without any upgrade, dis-tribution by the current infrastructures – and that has alreadybeen proposed as a substitute/supplement of gasoline [1–3].Acetone–butanol–ethanol (ABE)-producing Clostridia produce sol-vents by fermenting several biomasses, such as palm oil waste [4],agro-industrial waste [5], and agricultural crops [6,7]. Clostridiumsaccharoperbutylacetonicum, Clostridium acetobutylicum, Clostrid-ium beijerinckii, and Clostridium aurantibutyricum can metabolize agreat variety of substrates: pentoses, hexoses, mono-, di- and poly-saccharides [8]. Under batch conditions the fermentation processof solvent-producing Clostridium strains proceeds with the produc-

∗ Corresponding author at: Bioprocess Engineering, Wageningen University,AlgaePARC, P.O. Box 16, 6700 AA Wageningen, The Netherlands.Tel.: +31 6 19304246.

E-mail addresses: giolivie@unina.it, giuseppe.olivieri@wur.nl (G. Olivieri).

tion of cells, hydrogen, carbon dioxide, acetic acid and butyric acidduring the initial growth phase (acidogenesis) [8]. As the acid con-centration increases (pH decreases), the cell metabolism shifts tosolvent production (solventogenesis). The acidogenic cells – able toreproduce themselves – enter the solventogenesis state undergoinga morphological change [8]. During solventogenesis, the active cellsbecome endospores unable to reproduce themselves. Therefore,two physiological states must be taken into account for Clostridia:one for the acidogenic phase, and one for the solventogenic phase.Considerable research has been carried out on the ABE fermentationsystems to enhance butanol production [9–11]. Yet several ques-tions are still open as to how to optimise the industrial processesto produce butanol by fermentation.

The number of models reported in the literature is limited [12],which confirms the complexity of the metabolic pathway involvedin ABE production. Papoutsakis [13] developed a stoichiometricmodel: it may be used to calculate or estimate the rates of the reac-tions occurring in the pathway in several ABE-producing Clostridia.Votruba et al. [14] developed a mathematical model of batch cul-tures of C. acetobutylicum based on the biochemistry and physiology

http://dx.doi.org/10.1016/j.bej.2015.03.0011369-703X/© 2015 Elsevier B.V. All rights reserved.

F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166 157

Nomenclature

[AACoA] Acetoacetyl-CoA concentration (mM)ABE Acetone–butanol–ethanol[Acet] Acetate concentration (mM)[A] Acetone concentration (mM)[ACoA] Acetyl-CoA concentration (mM)Ar Arabinose[BCoA] Butyrl-CoA concentration (mM)[Biomass] Biomass concentration (mM)[B] Butanol concentration (mM)[Butyr] Butyrate concentration (mM)[E] Ethanol concentration (mM)F Switching factor of on-off mechanism[F6P] Fructose 6-phosphate concentration (mM)Fr FructoseG Glucose[G3P] Glyceraldehyde 3-phosphate concentration (mM)j Number of the corresponding reaction in Fig. 1A-B[Pyr] Pyruvate concentration (mM)Kaj Activation constant for activator (mM)Kiij Inhibition constant for inhibitor (mM)Kisj Inhibition constant for substrate (mM)Kmj Concentration of metabolite where the rate is equal

to half the value ov Vmax (mM)Kmsj Specific activation constant (mM)L LactoseM Mannoserj Rate equation of metabolic reactionVmaxj Maximum reaction rate (h−1)S Sucrose[Sugar] Sugar concentration (mM)YE Yeast extractX Xylose

of metabolite growth and synthesis. Desai et al. [15] analysed thecontribution of acid formation pathways to the metabolism of C.acetobutylicum ATCC824T according to the metabolic flux analysis(MFA). Shinto et al. [16,17] reported a kinetic simulation model todescribe the dynamic behaviour of metabolites in ABE productionby C. saccharoperbutylacetonicum N1-4 ATCC13564 using glucose orxylose as carbon source. Kim et al. [18] developed a kinetic modeldescribing the metabolism in acetone–butanol–ethanol (ABE)-fermentation by C. acetobutylicum ATCC 824; in particular theyused an optimization algorithm combining a genetic algorithmand the Levenberg–Marquardt algorithm in order to estimate thekinetic parameters of the model. Millat et al. [19] presented amethod for linking the pH shift, the clostridial growth and theacetone–butanol–ethanol fermentation metabolic network sys-tematically into a model that combines the dynamics of the externalpH and optical density with a metabolic model. Mayank et al. [12]pointed out the necessity to develop dynamic models to simu-late steady state fermentations as well as transient behaviours tosupport the development and optimization of butanol productionprocesses. Moreover, Mayank et al. [12] concluded that the effi-cient design of large-scale fermenters requires the combinationof reliable dynamic models and the description of the bioreactorhydrodynamics, in particular when immobilized cells are used.

The sugars present in the most promising feedstock (raw andtreated) for the ABE fermentation are widely assorted (glucose,arabinose, mannose, xylose, fructose, sucrose, lactose, etc.). Thereare several experimental studies on the effects of these sugars onfermentation performance [20–22], but the theoretical investiga-tions are quite limited. To the authors’ knowledge, the models

described in the literature are generally focused on individual sug-ars (glucose and xylose). No effort has been deployed to developmodels able to simulate the fermentation of the wide spectrum ofsugars that can be used as carbon source [12].

This study proposes a kinetic dynamic model to investigatethe effects of the carbon source on the ABE production by C. ace-tobutylicum DSM 792. The model was applied to several sugars– glucose, mannose, fructose, sucrose, lactose, xylose, and arabi-nose – using the specific metabolic pathway of each sugar: theEmbden–Meyerhof–Parnas (EMP) pathway equations for hexoseand disaccharide sugars; and the pentose phosphate (PP) pathwayequations for pentose sugars. Two main issues were addressed: (1)the dynamic behaviour of the metabolites involved in ABE produc-tion, and (2) the inhibitory and activatory mechanisms. The modelwas aimed to reproduce the main features of batch fermentations,characterized by a transient behaviour dominated by the accumu-lation of the inhibiting metabolites. The model may also supportthe simulation of biofilm behaviour where non-homogenous con-centration of metabolites is expected throughout the biofilm.

2. Materials and methods

2.1. Microorganism and media

C. acetobutylicum DSM 792 was supplied by DSMZ (Braun-schweig, Germany). Stock cultures were reactivated accordingto the DSMZ procedure. The reactivated cultures were stored at−80 ◦C. The thawed cells were inoculated into 12 mL syntheticmedium containing glucose (30 g/L) and yeast extract (5 g/L) in15 mL Hungate tubes (pre-cultures). The cells were grown underanaerobic conditions for 48 h at 37 ◦C. Then they were transferredinto fermentation bottles. Each test was carried out in duplicateand the mean values are reported as results. The error was typicallywithin 5%.

The fermentation medium consisted of 5 g/L yeast extract, 2 g/Lammonium chloride (N-source) and 5 g/L CaCO3 supplementedto a P2 stock solution: buffer) 0.25 g/L KH2PO4, 0.25 g/L K2HPO4;mineral) 0.2 g/L MgSO4·7H2O, 0.01 g/L MnSO4·7H2O, 0.01 g/LFeSO4·7H2O [22]. The medium was sterilized in an autoclave beforethe carbon source addition. Chemicals (CaCO3, KH2PO4, K2HPO4,MgSO4·7H2O, MnSO4·7H2O, and FeSO4·7H2O) and yeast extractwere from Sigma–Aldrich.

The sugars investigated were: glucose, mannose, fructose,sucrose, lactose, arabinose and xylose (Sigma–Aldrich). Theconcentrated sugar solutions were sterilized by filtration andsupplemented to the autoclaved medium. The stock solution con-centration of each investigated sugar was 300 g/L.

2.2. Analytical methods

The pH was measured off-line using a pH-meter (Hanna Instru-ments). The samples taken from the cultures were analysed tomeasure the liquid phase concentration of biomass, sugars andmetabolites. The cell density was measured as optical absorbanceat 600 nm (OD600) using a spectrophotometer (Cary 50 Varian). Thecalibration tests for C. acetobutylicum dried mass indicated that 1OD600 = 0.4 gDM/L [10].

The concentration of soluble species was measured in the liq-uid supernatant after cell separation by centrifugation. The sugarconcentration was measured by high performance liquid chro-matography (HPLC) using an Agilent 1100 system (Palo Alto, CA).The sugars were separated on a 8 �m Hi-Plex H, 30 cm × 7.7 mmat room temperature and detected with a refractive indexdetector. Deionized water was used as mobile phase at a flowrate of 0.6 mL/min. Acetone, butanol, ethanol, acetic and butyric

158 F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166

acid concentrations were measured by means of a GC apparatusequipped with a FID, and outfitted with a capillary column poraplotQ (25 m × 0.32 mm). An internal standard (hexanoic acid) was usedto assess acids and alcohols and their concentrations.

2.3. Operating conditions and procedures

Screw-cap Pyrex bottles (100 mL) containing 75 mL mediumwere used as fermenters. All the cultivations were carried out at37 ◦C and without pH control. The cultures were agitated by meansof a rotary shaker at 110 rpm. The medium was inoculated with a6.25% (v/v) suspension of actively growing pre-cultures. The initialcell concentration was set at 150 mg/L. Three millilitres of cultureswere sampled periodically for cell/metabolites characterization.

The initial cell concentration corresponded to 1.5 mM when thecell molecular weight was assumed equal to 101 g/mol [23,24]. Theinitial concentration of the investigated sugar in each batch testranged between 5 and 100 g/L.

3. Theoretical framework

3.1. Model

A kinetic simulation model was developed using the biochemi-cal networks simulator software COPASI [25]. COPASI supports thedevelopment and the analyses of a reaction network and includesapproximate velocity functions of enzyme kinetics.

The proposed model includes the substrate utilization rate, theproduction rate and the cell growth rate and is based on themetabolic pathways of C. acetobutylicum [16,17]. The pathwaysreported in Fig. 1A (ModelHex/Disacc) were used when glucose, man-nose, fructose, sucrose, and lactose were the carbon source. Thepathways reported in Fig. 1B (ModelPent) were used when xyloseand arabinose were the carbon source. Glucose, fructose, mannose,

sucrose and lactose are metabolized via the EMP pathway, andxylose and arabinose are metabolized via the PP pathway [16,17].The models are detailed hereinafter.

ModelHex/Disacc – the single step of the metabolic pathwaydescribed in Fig. 1A – hexose and disaccharide sugars – is reportedin Table 1 as Reaction r1 through r2.

ModelPent – the first steps of xylose and arabinose metabolismare reported in Table 1 as Reaction r20 through r25. Reaction r13through r16 also hold for the conversion of xylose and arabinoseprovided that G3P is produced.

The main assumptions made in the model by Shinto et al. [16,17]are as follows:

• A Michaelis–Menten type kinetics characterized by butanol non-competitive inhibition is assumed to exist for the rate equationof cell growth r∗

15.• According to the Eq. (1.16), the death rate of the cell – r16 – is

assumed to be a first order kinetics in the biomass concentration.• The kinetic rates of Eqs. (1.1) and (1.20) – r∗

1 and r∗20, respec-

tively – are obtained by combining the substrate-sugar inhibitionand the butanol-uncompetitive inhibition behaviour, accordingto the known effects of inhibition by substrate and solvent on thesubstrate conversion rate [8].

• The kinetic rate of Eq. (1.10) – r∗10 – is obtained by combining

the butyrate activation and the butanol-uncompetitive inhibitionbehaviour, according to the enhancement effects of butyrate onthe butanol production rate reported by Shinto et al. [16]

The assumptions made in the kinetic model proposed (herein,referred to as proposed model) are summarised in the following.The kinetic relationships are reported without the superscript “*”.

• The kinetic expressions proposed by Shinto et al. [16,17] for sugaruptake (r∗

1 and r∗20), butanol production (r∗

10) and cell growth

Fig. 1. (A) Metabolic pathways of C. acetobutylicum. Carbon source: glucose, mannose, fructose, sucrose, and lactose. Enzymes (bold style): PTA, phosphotransacetylase; AK,acetate kinase; CoAT, CoA transferase; PTB, phosphotransbutyrylase; BK, butyrate kinase; BADH, butyraldehyde dehydrogenase; BDH, butanol dehydrogenase. (B) MetabolicPP pathways of C. acetobutylicum. Carbone source: xylose and arabinose. Enzymes (bold style): TA, transaldolase; TK, transketolase.

F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166 159

Table 1Reaction steps of the EMP and PP metabolic pathways and associated kinetics.

Name Reactions Kinetics Refs. Eqs.

r1 G/M/F → F6P or r1 =Vmax1[S]

Km1+[S]+Km1

([S]

Kis1

)2

(1 − [B]

BMAX1

)nB1F

a (1.1)

S → 2 F6P or b

L → F6P + 2G3P

r∗1 = Vmax1[S]

Km1+Km1([S]/Kis1)2+[S](1+[B]/Kii1)F a

r2 F6P → 2 G3P r2 = Vmax2[F6P]Km2+[F6P] F a (1.2)

r3 G3P → Pyr r3 = Vmax3[G3P]Km3+[G3P] F a (1.3)

r4 Pyr → ACoA r4 = Vmax4[Pyr]Km4+[Pyr] F a (1.4)

r5 ACoA → Acetate r5 = Vmax5[ACoA]Km5+[ACoA] F a (1.5)

r6 Acetate → ACoA r6 = Vmax6[Acet]Km6+[Acet] F a (1.6)

r7 ACoA → E r7 = Vmax7[ACoA]Km7+[ACoA] F a (1.7)

r8 ACoA → ½ AACoA r8 = Vmax8[ACoA]Km8+[ACoA]

a (1.8)

r9 AACoA → BCoA r9 = Vmax9[AACoA]Km9+[AACoA] F a (1.9)

r10 BCoA → B r10 = Vmax10[BCoA]Km10(1+Ka10/[Butyr])+S

(1 − [B]

BMAX10

)nB10F b (1.10)

r∗10 = Vmax10[BCoA]

Km10(1+Ka10/[Butyr])+[BCoA](1+[B]/Kii10) F a

r11 Acet + AACoA → A + ACoA r11 = Vmax11

(1

1+Km11A/[Acet]

)(1

1+Km11B/[AACoA]

)a (1.11)

r12 Butyr + AACoA → A + BCoA r12 = Vmax12

(1

1+Km12A/[Butyr]

)(1

1+Km12B/[AACoA]

)a (1.12)

r13 BCoA → Butyr r13 = Vmax13[BCoA]Km13+[BCoA] F a (1.13)

r14 Butyr → BCoA r14 = Vmax14[Butyr]Km14+[Butyr] F a (1.14)

r15 ACoA → Biomass r15 = Vmax15[ACoA]Km15+[ACoA]

(1 − [Acet]

AcetMAX

)nAcetate(

1 − [Butyr]ButyrMAX

)nButyrate(

1 − [A]AMAX

)nA(

1 − [E]EMAX

)nE(

1 − [B]BMAX15

)nB15 b (1.15)

r∗15 = Vmax15[ACoA]

Km15(1+[B]/Kii15)+[ACoA](1+[B]/Kii15)a

r16 Biomass → Inactive Cells r16 = Vmax16[Biomass][B]Kms16×Ka16+(Kms16+[Biomass])[B]

b (1.16)

r∗16 = Vmax16 [Biomass] a

r20 X/Ar → X5P r20 = Vmax20[S]

Km20+[S]+Km20

([S]

Kis20

)2

(1 − [B]

BMAX20

)nB20F b (1.17)

r∗20 = Vmax20[S]

Km20+Km20([S]/Kis20)2+[S](1+[B]/Kii20)F a

r21 X5P → R5P r21 = Vmax21[X5P]Km21+[X5P]

a (1.18)

r22 R5P → X5P r22 = Vmax22[R5P]Km22+[R5P]

a (1.19)

r23 R5P + X5P→G3P + S7P r23 = Vmax23

(1

1+Km23A/[R5P]

)(1

1+Km23B/[X5P]

)a (1.20)

r24 G3P + S7P → E4P + F6P r24 = Vmax24

(1

1+Km24A/[S7P]

)(1

1+Km24B/[G3P]

)a (1.21)

r25 E4P + X5P → F6P + G3P r25 = Vmax25

(1

1+Km25A/[X5P]

)(1

1+Km25B/[E4P]

)a (1.22)

a [16,17].b [27].

(r∗15) tend to zero as butanol concentration approaches infi-

nite. However, this behaviour does not fit the metabolism of C.acetobutylicum that is characterized by full inhibition as the con-centration of inhibitor metabolites approaches a critical value[23,24,26]. Moreover, the rate of the cell death (1.16) is acti-vated by butanol [27]. In the proposed model no reactivationof death cells (spores) was taken in consideration and a modi-fied set of reaction rates was used to substitute the uptake ratesr∗1 and r∗

20, the butanol production rate r∗10, the growth rate r∗

15,the cell death rate r∗

16. Table 1 reports the proposed reactionrate as: r1 and r20, complete inhibition as butanol concentra-tion approaches the critical value BMAX was included; r15, aninteractive multiproduct-inhibited model was used for the cell

growth kinetics; r16, a specific butanol activation expression wasincluded.

• Acetoacetyl-CoA transferase (CoAT) has a broad carboxylic acidspecificity and can catalyse the transfer of CoA to either acetateand butyrate [16,28]. According to this observation, a specificexpression of the reaction rate of CoAT was proposed for eachsubstrate in agreement with the random bi bi model [27]: r11and r12.

• The reaction rate equations of TK and TA consisted of randombi–bi mechanisms (Eqs. (1.20)–(1.22)).

• Several metabolic reactions of ABE fermentation require ATP orNADH (Fig. 1A). These reactions are not expected to happen ifenergy source depletion occurs – e.g. sugar exhaustion – and

160 F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166

an on–off mechanism was used. A switch-factor F was intro-duced and its value may be 1 or 0 depending on the substrateconcentration in the broth. According to the observation byOkamoto et al. [29], the threshold value of substrate concentra-tion at which F switches from 1 to 0 is larger than zero: F is set to1 for substrate concentrations larger than 1.00 mM. The switch-factor was used for the rate of reactions that include ATP, ADP,NADH, and NAD+ amongst the substrates: r1–r7, r9, r10, r13, r14and r20.

Although there is experimental evidence that the sugar concen-tration is not the key to cause and effect in regulation, this approachis well suited to represent batch fermentations, where sugar deple-tion and regulatory onset of solventogenesis are coupled.

3.2. Assessment of the model parameters

The main assumptions made to assess the model parameters arereported hereinafter.

• According to the kinetics of enzymatic reactions, the maximumreaction rate is proportional to the effective concentration of theenzymes. In turn, this effective concentration of the enzymesdepends on the sugar type. Therefore, it is reasonable to assumethat the maximum reaction rate depends on the sugar type;

• The “affinity” constants – Kaj, Kiij, Kisj, Kms16 and Kmj (except forj = 1 and 20) – do not depend on the sugar because they depend onthe enzyme responsible for the reaction step but not on its con-centration. The values of the “affinity” constants were assumedto be constant for all the investigated sugars and were assessedfor glucose;

• The parameters of the sugar uptake Reactions (r1 and r20) wereassessed for each sugar. Indeed, Servinsky et al. [30] reportedthat C. acetobutylicum has sugar-specific mechanisms for thetranscriptional regulation of transport and metabolism genes. Inparticular, C. acetobutylicum utilizes: (a) symporters and ATP-binding cassette (ABC) transporters for the uptake of pentosesugars; (b) phosphotransferase system (PTS) transporters and agluconate – H+ (GntP) transporter – for the uptake of disaccha-rides and hexoses. Moreover, Servinsky et al. [30] reported thatthe transcription of some transporter genes is induced by specificsugars. Sugar-specific transport roles are suggested for varioustransporters of the PTS and of the ABC superfamily [30].

The assessment of the model parameters was carried out accord-ing to the following procedure.

i.) The proposed model was applied to the data measured dur-ing the glucose fermentation tests. The values of the kineticparameters were estimated by fitting the experimental datameasured during the batch cultures of C. acetobutylicum inglucose-based medium with glucose initial concentration rang-ing between 5 and 555 mM;

ii.) The kinetic parameters for the hexoses (except glucose) anddisaccharides (sucrose and lactose) were assessed. Except forKaj, Kiij, Kisj, Kmj, Kms16 (i and j ranging between 2 and 16), themaximum reaction rate Vmaxj of each reaction step and thevalue of BMAXj, AcetMAX, ButyrMAX, AMAX, EMAX, nBJ, nAcet, nButyr,nA, and nE were assessed. The parameters of the reaction rater1 were also estimated for each sugar.

iii.) The kinetic parameters of the xylose pathway (r20–r25) wereassessed. Except for Kaj, Kiij, Kisj, Kmj, Kms16 (i and j rangingbetween 2 and 16), the maximum reaction rate Vmaxj of eachreaction step and the value of BMAXj, AcetMAX, ButyrMAX, AMAX,EMAX, nBJ, nAcet, nButyr, nA, and nE were assessed. Ta

ble

2K

inet

icp

aram

eter

sas

sess

edby

pro

cess

ing

exp

erim

enta

ldat

aac

cord

ing

toth

ep

rese

nte

dan

dSh

into

etal

.’sm

odel

.Car

bon

sou

rce:

glu

cose

.

Rea

ctio

ns

Vm

axj

Km

jK

isj

Kiij

Kaj

Km

sjK

mjA

Km

jBA

MA

XB M

AX

jE M

AX

Ace

t MA

XB

uty

r MA

Xn A

n Bj

n En A

cet

n Bu

tyr

(h−1

)(m

M)

(mM

)(m

M)

(mM

)(m

M)

(mM

)(m

M)

(mM

)(m

M)

(mM

)(m

M)

(mM

)

ab

ab

ab

ba

ba

ab

ab

aa

aa

aa

aa

aa

r 16.

796.

800.

0727

7.75

135

299

220

0.71

r 295

114

0.11

0.39

r 330

526

521

227

3r 4

276

254

286

145

r 58.

209.

447.

2023

9r 6

5.23

0.59

0.14

0.59

r 70.

480.

460.

090.

29r 8

310

297

162

204

r 924

829

29.

563.

24r 1

092

790.

420.

3728

77.

4112

.519

30.

25r 1

16.

677.

490.

001

0.00

0227

629

9r 1

20.

340.

0524

121

719

725

0r 1

372

270

90.

050.

07r 1

420

719

84.

137.

32r 1

52.

432.

610.

0004

0.65

299

968

180

712

124.

517

71

0.22

10.

150.

13r 1

61.

890.

044.

3710

.5

aPr

esen

ted

Mod

el.

bSh

into

-par

amet

erse

t.

F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166 161

Fig. 2. Experimental time-course data and simulation results of target metabolites for glucose, mannose, fructose, sucrose, lactose, arabinose and xylose using the proposedmodel. Initial sugar concentration 60 g/L.

Table 3Kinetic parameters assessed by processing experimental data according to the presented and Shinto’s model. Carbon source: xylose.

Reactions Vmax Kmi Kis Kii KmA KmB BMAX nB

(h−1) (mM) (mM) (mM) (mM) (mM) (mM)

a b a b a b b a b a b a a

r20 1.50 4.32 54 0.40 293 9.3 14.8 172 1.79r21 296 299 27 186r22 218 129 43 210r23 216 149 0.08 69 3.74 25.6r24 205 61 5 212 22 0.36r25 298 113 0.21 127 2.98 1.14

(a) Presented Model.(b) Shinto-parameter set.

iv.) The kinetic parameters for the arabinose pathway wereassessed: the parameters of the reaction rate r20, Vmaxj of eachreaction step, and BMAXj, AcetMAX, ButyrMAX, AMAX, EMAX, nBJ,nAcet, nButyr, nA, and nE. All the other parameters were set tothe values measured for the xylose pathway.

The simulated annealing (SA) method – an optimization algo-rithm of COPASI – was used for the parameter fitting.

The study also included the assessment of the kinetic parame-ters of the model by Shinto et al. [16,17] based on the experimentaldata collected in the present investigation (hereinafter, referred to

162 F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166

Fig. 3. (A) Time-resolved concentration of glucose, biomass and metabolites. Experimental data vs. simulation results. (B) Time-resolved concentration of xylose, biomassand metabolites. Experimental data vs. simulation results. Initial sugar concentration 60 g/L.

Table 4Average squared correlation coefficients (r2) between simulation results and experimental data.

r2 Glucose Mannose Fructose Sucrose Lactose Arabinose Xylose

Shinto et al.’s simulation a 0.855 0.812 0.820 0.800 0.904 0.848 0.830Presented Model 0.894 0.887 0.870 0.880 0.925 0.904 0.890

a [16,17].

as Shinto-parameter set). The Shinto-parameter set was assessedfor C. acetobutylicum and for all the seven sugars investigated.

The soundness of the model was tested according to the follow-ing two procedures:

• The assessment of the average squared correlation coefficients(r2) between the simulation results and the experimental data.

• The comparison of the results of the proposed model with thoseof the model by Shinto et al. [16,17] using the Shinto-parameterset.

4. Results and discussion

4.1. Batch cultures

The batch cultures were carried out at initial sugar concentra-tions ranging between 5 and 555 mM for hexose sugars (glucose,mannose, and fructose), 14 and 278 mM for disaccharide sugars

(sucrose and lactose), 33 and 667 mM for pentose sugars (xyloseand arabinose). The time resolved concentration of sugar andmetabolites was measured. As expected, the pH (data not reported)decreased with the time since the beginning of the fermentationand stabilized at about 4.7 under the solventogenic phase.

The experimental data reported in Fig. 2 refer to the batch fer-mentation tests carried out at an initial sugar concentration of335, 167 and 400 mM for hexose, disaccharide and pentose sugars,respectively.

The data reported in Fig. 2 highlight that solvent productionis sugar specific: the glucose fermentation was characterized bythe highest butanol concentration (169 mM); the performanceof mannose, fructose, and sucrose fermentations was slightlylower (butanol concentration 140, 134, and 135 mM, respectively);the performance of arabinose and xylose fermentations was thelowest (116, and 112 mM, respectively); the lactose fermenta-tion was characterized by the lowest final butanol concentration(19 mM).

F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166 163

Table 5Sugar uptake (r1/r20), butanol production (r10) and cell growth (r15) kinetic parameters for mannose, fructose, sucrose, lactose, arabinose and xylose.

Reaction Parameters Mannose Fructose Sucrose Lactose Arabinose Xylose

r1 Vmax (h−1) 4.71 4.64 3.4 1.11 2.7 1.5Km (mM) 0.1 0.04 0.49 40.8 46 54Kis (mM) 9 6 20 100 200 210BMAX(mM) 199 199 199 153 186 171nB 0.94 0.95 1.13 1.85 1.79 1.79

r10 BMAX (mM) 180 181 172 135 163 158nB 0.46 0.47 0.58 1.62 0.62 0.64

r15 AcetMAX (mM) 116 124 114 119 100 82ButyrMAX (mM) 179 150 141 141 137 127AMAX (mM) 991 920 1096 972 875 810BMAX (mM) 151 148 140 110 136 137EMAX (mM) 722 760 765 725 689 661nAcet 0.72 0.79 0.93 1.39 0.91 1.21nButyr 0.26 0.28 0.37 0.87 0.72 0.87nA 1 1 1 1 1 1nB 0.76 0.73 0.83 1.93 1.18 1.85nE 1 1 1 1 1 1

4.2. Effect of inhibition and activation terms

The data reported in Fig. 2 were processed according to theproposed model to assess the kinetic parameters of the metabolicpathways of C. acetobutylicum for hexoses/disaccharide (Fig. 1A)and pentose sugars (Fig. 1B). The procedure is reported in Section3.2.

The kinetic parameters assessed by processing the experimen-tal data measured during the glucose fermentation are reportedin Table 2. Fig. 3A reports the experimental data and the resultsof the plots from both the proposed model and the Shinto et al.’smodel [16,17]. The Shinto-parameter set was used for the simula-tions made by means of the Shinto et al.’s model. The agreementof the proposed model with the experimental data is very satis-

Fig. 4. Deviation of Vmax of Reaction r1 through r16 assessed for hexoses/disaccharides/pentoses with respect to the value assessed for glucose as a function of the sugar.

164 F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166

Fig. 5. Deviation of Vmax of Reaction r20 through r25 assessed for arabinose withrespect to the value assessed for xylose as a function of the sugar.

factory. The average correlation coefficient (r2) reported in Table 4highlights that the proposed model is more likely than (r2 = 0.894)than the Shinto et al.’s simulation (0.855).

Table 3 reports the kinetic parameters of the PP pathway (r20through r25) assessed by evaluating data measured during xylosefermentation. Fig. 3B reports the experimental data and the resultsof the plots from both the proposed model and the Shinto et al.’smodel [16,17]. The Shinto-parameter set was used for the simula-tions made by means of the Shinto et al.’s model. The agreementof the proposed model with the experimental data is very satis-factory. The average correlation coefficient (r2) reported in Table 4highlights that the proposed model is more likely (r2 = 0.89) thanthe Shinto et al.’s simulation (0.83).

Tables 2 and 3 also report the Shinto-parameter set assessedfor the Shinto et al.’s model of the C. acetobutylicum (b-data groupin both tables). The direct comparison of the parameters of thesame reaction in the two models cannot be carried out withouttaking into account the overall structure of the models. The mainparameter changes were recorded for the reactions characterizedby different relationships: Reactions r1, r10, r15 and r16. The Vmax

assessed for the reactions does not change regardless of the modelused (present vs. Shinto’s). The Vmax of Reaction r16 (cell death rate)does depend on the model because the Shinto’s version does nottake into account the expected activation by butanol.

The kinetic parameters of the proposed model as applied tomannose, fructose, sucrose, lactose, and arabinose were assessedaccording to the procedure reported in Section 3.2. The simulationsof the proposed model are plotted in Fig. 2 for all the investigatedsugars: the experimental dynamic behaviour of target metabolitesis also reported. Table 5 reports the kinetic parameters of the sugaruptake (r1 and r20), the butanol production (r10) and the cell growth(r15) assessed for the fermentation of mannose, fructose, sucrose,lactose, arabinose, and xylose. The average r2 for each sugar withreference to the proposed model and Shinto et al.’s simulation[16,17] were calculated and are reported in Table 4. The analysisof the table suggests that in the proposed model r2 increases ascompared to Shinto et al.’s simulation [16,17], whatever the sugartested. The results confirm that the structure of the proposed modelimproves the simulation results. The results of this model are morelikely than Shinto et al.’s model, which is particularly useful whenthe kinetic model is included in an overall model of the conversionprocess (e.g. biofilm reactors). The main updates with respect toShinto et al.’s model [16,17] are: full inhibition term for butanolin the uptake kinetics (r1 and r20) and in the butanol productionkinetics (r10); an interactive multiproduct-inhibited model for thecell growth kinetics (r15); a specific butanol activation expressionfor the cell death kinetics (r16).

The results of the proposed simulations were also analysed toinvestigate the sensitivity of each reaction step to every sugar. Fig. 4

reports the variation of the maximum reaction rate (Vmax) of thereaction steps with respect to the value assessed for glucose fer-mentation, for the hexose/disaccharide tested. Fig. 5 reports thevariation of Vmax of the reaction steps of the arabinose PP path-way with respect to the value assessed for xylose fermentation.The main observations are reported hereinafter.

(a) The sugar uptake Reactions – r1 and r20 – stronglydepend on the sugar type. The values of the parameterssuggest that the C. acetobutylicum preference scale is: glu-cose > mannose/fructose > sucrose > arabinose > xylose > lactose.This scale agrees with the literature on sugar conversion[20–22].

(b) Except for the cell death Reaction (r16), the Vmax of each reac-tion step of hexose/disaccharides is smaller than that of thehomologous step of glucose fermentation. The Vmax16 assessedfor glucose fermentation is smaller than that measured for thehexose/disaccharides investigated.

(c) The reaction steps r2, r3, r5, r6, r8, r9, r13, r15 for mannose andfructose fermentations are just barely affected by the sugartype: the parameters differ by less than 20% with respect to thehomologous parameters measured for glucose fermentation.

(d) For sucrose fermentation the reaction steps r3, r5, r6, r8, r9 arejust barely affected by the sugar type: the parameters differby less than 20% with respect to the homologous parametersmeasured for glucose fermentation.

(e) Except for r3, when lactose, arabinose, and xylose are usedas carbon source the reaction steps of the C. acetobutylicumfermentation are strongly affected by the sugar type: the param-eters differ by more than 20% with respect to the homologousparameters measured for glucose fermentation.

(f) Except for the sugar uptake reaction r20, the reactions of thePP pathway are not affected by the sugar. The kinetics of thereaction r20 strongly depends on the sugar: the arabinose uptakeis significantly faster than xylose.

Based on these observations it is possible to discuss the effectsof the sugars on the proposed model.

Observation (a) would confirm the first assumption reported inSection 3.2. The concentration of the enzymes responsible for theinvestigated reaction steps is expected to be high when glucose isused as carbon source.

The fermentation of the hexoses does not strongly depend onthe sugar type. Although the glucose fermentation is characterizedby faster reaction rates, the difference of the rate assessed for theother hexoses is within 20%.

As regards the fermentation of the di-saccharide sucrose andlactose, the results may be interpreted in terms of the differenthydrolysis path of the two sugars. Both sugars are transported intothe cells according to the same PTS mechanism, and hydrolysedinto simple sugars. The sucrose is hydrolysed into fructose-6-P and glucose-6-P: both sugars can be metabolized via theEmbden–Meyerhof pathway [31]. The lactose is hydrolysed intoglucose and galactose-6-P: “the glucose may be phosphorylatedand incorporated into the glycolytic pathway whereas galactose-6-P is generally metabolized by the tagatose 6-P pathway” [32].Therefore, the lower performance of the lactose and sucrose withrespect to the glucose may be due to the sugar transport across thecell membrane. Moreover, the lower performances of the lactosewith respect to the sucrose may be due to the bottleneck of thegalactose-6-P pathway.

The acidogenic phase length is not affected by the sugar typeexcept for lactose and xylose. These two sugars are characterizedby a longer acidogenic phase (Fig. 2). This observation agrees withthe Vmax value assessed for acid production (Reactions r5 and r13)that is smaller than that assessed for glucose (see Fig. 4e and m). The

F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166 165

observed behaviour of lactose is in agreement with the lower sugaruptake discussed in the previous paragraph. A peculiar behaviourhas been observed when comparing arabinose and xylose, twosugars characterized by slow fermentation. Both pentoses are char-acterized by a sugar uptake rate lower than that of glucose but thatof xylose is even lower. Moreover, the acid production for xylose islower than that assessed for arabinose. The combination of thesetwo steps produces an acidogenic phase longer than that of arabi-nose.

The proposed model may be used as an update of the dynamicmodel proposed by Shinto et al. [16,17] to support the design andoptimization of fermenters dedicated to butanol production. Themodel reproduces the main features of batch fermentations thatare generally characterized by a transient behaviour dominatedby the accumulation of the inhibiting metabolites. Furthermore,this model is expected to support the simulation of continuousfermenters that are aimed at producing butanol at the maximumpossible concentration. The model may also support the simula-tion of biofilm behaviour where non-homogenous concentration ofmetabolites is expected and in particular when butanol concentra-tion is expected to reach maximum/inhibiting values in the innerpart [12]. Of course, the development of a biofilm reactor modelwould benefit from a more detailed dynamic model [33,34] butsuch a detailed model would be so complex that the simulationswould hardly be feasible.

5. Conclusions

A kinetic dynamic model of acetone–butanol–ethanol (ABE)production by C. acetobutylicum DSM 792 developed with the bio-chemical networks simulator COPASI is illustrated in this paper.

It is an updated version of the model by Shinto et al. [16,17]. Thekinetic relationships included in this model are more representa-tive of the fermentation behaviour. Full inhibition by metabolites isincluded in the kinetics of sugar uptake, of butanol production, andcell growth rate. The butanol activation is used for the cell deathkinetics. The proposed kinetics have a key role when it is includedin a model of butanol continuous production in a structured fer-menter such as a biofilm reactor. The knowledge of this kinetics isfundamental for the reliability of the models aimed at supportingthe operation of the fermenters and at investigating the dynamicsof the conversion process. The operation of this reactor typology isusually aimed at producing high butanol concentration stream andthe inhibition role of this solvent is crucial for the success of theprocess.

The effects of the substrate were studied by applying the modelto several sugars: glucose, mannose, fructose, sucrose, lactose,xylose, and arabinose. The parameters assessed for the glucosewere used for the simulations with the other sugars except for thevalue of the sugar uptake Reactions (r1 and r20) and the maximumreaction rates of all reaction steps.

The model gave satisfactory results: the squared correlationcoefficient (r2) between experimental and calculated time-courseof metabolites ranges between 0.87 and 0.925.

Acknowledgements

The Università Degli Studi di Napoli Federico II is acknowl-edged for the grant of F. Raganati to be at the University of Berlin.The authors thank the Ministero dello Sviluppo Economico (Italy)for the financial support to the project EuroTransBio ETB-2012-16OPTISOLV (Development, optimization and scale-up of biologicalsolvent production) (C01/0878/01-03/X21) as well as the BMBF(Germany) for supporting the projects COSMIC2 (FKZ: 0315782B)and OPTISOLV (FKZ: 031A231C).

References

[1] P. Dürre, Biobutanol: an attractive biofuel, Biotechnol. J. 2 (2007)1525–1534.

[2] R. Cascone, Biobutanol – a replacement for bioethanol? Chem. Eng. Prog. 104(8) (2008) S4–S9.

[3] B. Chen, F. Chuang, C. Lin, J. Chang, Deciphering butanol inhibition toclostridial species in acclimatized sludge for improving biobutanolproduction, Biochem. Eng. J. 69 (2012) 100–105.

[4] T.M. Lee, A. Ishizaki, S. Yoshino, K. Furukawa, Production of acetone butanol,and ethanol from palm oil mill effluent using C. saccharoperbutylacetonicumN1-4, Biotechnol. Lett. 17 (1995) 649–654.

[5] F. Raganati, G. Olivieri, A. Procentese, M.E. Russo, A. Marzocchella, P. Salatino,Butanol production by bioconversion of cheese whey in a continuous packedbed reactor, Bioresour. Technol. 138 (2013) 259–265.

[6] N. Qureshi, M.M. Meagher, J. Huang, R.W. Hutkins, Acetone butanol ethanol(ABE) recovery by pervaporation using silicalite–silicone compositemembrane from fed-batch reactor of Clostridium acetobutylicum, J. Membr.Sci. 187 (2001) 93–102.

[7] K. Gao, L. Rehmann, ABE fermentation from enzymatic hydrolysate ofNaOH-pretreated corncobs, Biomass Bioenergy 66 (2014) 110–115.

[8] D.T. Jones, D.R. Woods, Acetone–butanol fermentation revisited, Microbiol.Rev. 50 (4) (1986) 484–524.

[9] H. Thi, M. Tran, B. Cheirsilp, B. Hodgson, K. Umsakul, Potential use of Bacillussubtilis in a co-culture with Clostridium butylicum foracetone–butanol–ethanol production from cassava starch, Biochem. Eng. J. 48(2010) 260–267.

[10] F. Napoli, G. Olivieri, M.E. Russo, A. Marzocchella, P. Salatino, Butanolproduction by Clostridium acetobutylicum in a continuous packed bed reactor,J. Ind. Microbiol. Biotechnol. 37 (2010) 603–608.

[11] F. Napoli, G. Olivieri, M.E. Russo, A. Marzocchella, P. Salatino, Optimization ofsolvent recovery in the production of butanol by fermentation, Environ. Eng.Manage. J. 11 (8) (2012) 1499–1504.

[12] R. Mayank, A. Ranjan, V.S. Moholkar, Mathematical models of ABEfermentation: review and analysis, Crit. Rev. Biotechnol. 33 (2013)419–447.

[13] E.T. Papoutsakis, Equations and calculations for fermentations of butyric acidbacteria, Biotechnol. Bioeng. 26 (1984) 174–187.

[14] J. Votruba, B. Volesky, L. Yerushalmi, Mathematical model of a batchacetone–butanol fermentation, Biotechnol. Bioeng. 28 (1986) 247–255.

[15] R.P. Desai, L.M. Harris, N.E. Welker, E.T. Papoutsakis, Metabolic flux analysiselucidates the importance of the acid-formation pathways in regulatingsolvent production by Clostridium acetobutylicum, Metab. Eng. 1 (1999)206–213.

[16] H. Shinto, Y. Tashiro, M. Yamashita, G. Kobayashi, T. Sekiguchi, T. Hanai,Kinetic modeling and sensitivity analysis of acetone–butanol–ethanolproduction, J. Biotechnol. 131 (2007) 45–56.

[17] H. Shinto, Y. Tashiro, M. Yamashita, G. Kobayashi, T. Sekiguchi, T. Hanai,Kinetic study of substrate dependency for higher butanol production inacetone–butanol–ethanol fermentation, Process Biochem. 43 (12) (2008)1452–1461.

[18] W. Kim, M. Eom, S. Lee, J. Choi, S. Park, Kinetic study of metabolic pathways inClostridium acetobutylicum, in: 8th IFAC Symposium on Advanced Control ofChemical Processes Proceedings, Singapore July 10–13, 2012.

[19] T. Millat, H. Janssen, G.J. Thorn, J.R. King, H. Bahl, R.J. Fischer, O. Wolkenhauer,A shift in the dominant phenotype governs the pH-induced metabolic switchof Clostridium acetobutylicum in phosphate-limited continuous cultures, Appl.Microbiol. Biotechnol. 97 (14) (2013) 6451–6466.

[20] T. Ezeji, N. Qureshi, H. Blaschek, Butanol production from agriculturalresidues: impact of degradation products on Clostridium beijerinckii growthand butanol fermentation, Biotechnol. Bioeng. 97 (2007) 1460–1469.

[21] T.C. Ezeji, H.P. Blaschek, Fermentation of dried distillers’ grains and solubles(DDGS) hydrolysates to solvents and value-added products by solventogenicClostridia, Bioresour. Technol. 99 (12) (2008) 5232–5242.

[22] F. Raganati, S. Curth, P. Götz, G. Olivieri, A. Marzocchella, Butanol productionfrom lignocellulosic-based hexoses and pentoses by fermentation ofClostridium acetobutylicum, Chem. Eng. Trans. 27 (2012) 91–96.

[23] A.P. Zeng, A. Ross, H. Biebl, C. Tag, B. Gunzel, W.D. Deckwer, Multiple productinhibition and growth modeling of Clostridium butyricum and Klebsiellapneumoniae in glycerol fermentation, Biotechnol. Bioeng. 44 (1994)902–911.

[24] F. Napoli, G. Olivieri, M.E. Russo, A. Marzocchella, P. Salatino, Continuouslactose fermentation by Clostridium acetobutylicum – assessment of energeticsand product yields of the acidogenesis, Enzyme Microb. Technol. 50 (3)(2012) 165–172.

[25] S. Hoops, S. Sahle, R. Gauges, C. Lee, J. Pahle, N. Simus, M. Singhal, L. Xu, P.Mendes, U. Kummer, COPASI – a complex pathway simulator, Bioinformatics22 (2006) 3067–3074.

[26] F. Napoli, G. Olivieri, M.E. Russo, A. Marzocchella, P. Salatino, Continuouslactose fermentation by Clostridium acetobutylicum – assessment ofacidogenesis kinetics, Bioresour. Technol. 102 (2011) 1608–1614.

[27] R. Gheshlaghi, J.M. Scharer, M. Moo-Young, C.P. Chou, Metabolic pathways ofClostridia for producing butanol, Biotechnol. Adv. 27 (6) (2009) 764–781.

[28] Z.L. Boynton, G.N. Bennett, F.B. Rudolph, Intracellular concentrations ofcoenzyme A and its derivatives from Clostridium acetobutylicum ATCC 824 and

166 F. Raganati et al. / Biochemical Engineering Journal 99 (2015) 156–166

their roles in enzyme regulation, Appl. Environ. Microbiol. 60 (1994)39–44.

[29] M. Okamoto, T. Sakai, K. Hayashi, Biochemical switching device realizingMcCulloch–Pitts type equation, Biol. Cybern. 58 (1988) 295–299.

[30] M.D. Servinsky, J.T. Kiel, N.F. Dupuy, C.J. Sund, Transcriptional analysis ofdifferential carbohydrate utilization by Clostridium acetobutylicum,Microbiology 156 (2010) 3478–3491.

[31] M. Tangney, C. Rousse, M. Yazdanian, W.J. Mitchell, Note: sucrose transportand metabolism in Clostridium beijerinckii NCIMB 8052, J. Appl. Microbiol. 84(1998) 914–919.

[32] Y. Yu, M. Tangney, H.C. Aass, W.J. Mitchell, Analysis of the mechanism andregulation of lactose transport and metabolism in Clostridium acetobutylicumATCC 824, Appl. Environ. Microbiol. 73 (2007) 1842–1850.

[33] R.D. Li, Y.Y. Li, L.Y. Lu, C. Ren, Y.X. Li, L. Liu, An improved kinetic model for theacetone–butanol–ethanol pathway of Clostridium acetobutylicum andmodel-based perturbation analysis, BMC Syst. Biol. 5 (Suppl. 1) (2011)S12.

[34] L. Yerushalmi, B. Volesky, J. Votruba, Modeling of culture kinetics andphysiology for C. acetobutylicum, Can. J. Chem. Eng. 64 (1986)607–616.