-
Contents lists available at ScienceDirect
Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech
Targeted poly(3-hydroxybutyrate-co-3-hydroxyvalerate)
bioplasticproduction from carbon dioxide
Stef Ghyselsa,⁎, Md. Salatul Islam Mozumderb, Heleen De Weverc,
Eveline I.P. Volckea,Linsey Garcia-Gonzalezc
aGhent University, Department of Biosystems Engineering, Coupure
Links 653, 9000 Gent, Belgiumb Shahjalal University of Science and
Technology, Department of Chemical Engineering and Polymer Science,
Sylhet, Bangladeshc Flemish Institute for Technological Research
(VITO), Business Unit Separation and Conversion Technology,
Boeretang 200, 2400 Mol, Belgium
G R A P H I C A L A B S T R A C T
A R T I C L E I N F O
Keywords:Gas
fermentationPoly(3-hydroxybutyrate-co-3-hydroxyvalerate)Carbon
capture and utilizationModelling
A B S T R A C T
A microbial production process was developed to convert CO2 and
valeric acid into tailored
poly(3-hydro-xybutyrate-co-3-hydroxyvalerate) (PHBV) bioplastics.
The aim was to understand microbial PHBV production inmixotrophic
conditions and to control the monomer distribution in the polymer.
Continuous sparging of CO2with pulse and pH-stat feeding of valeric
acid were evaluated to produce PHBV copolyesters with
predefinedproperties. The desired random monomer distribution was
obtained by limiting the valeric acid concentration(below 1 −g L
1). 1H-NMR, 13C-NMR and chromatographic analysis of the PHBV
copolymer confirmed both themonomer distribution and the
3-hydroxyvalerate (3HV) fraction in the produced PHBV. A
physical-based modelwas developed for mixotrophic PHBV production,
which was calibrated and validated with independent ex-perimental
datasets. To produce PHBV with a predefined 3HV fraction, an
operating diagram was constructed.This tool was able to predict the
3HV fraction with a very good accuracy (2% deviation).
1. Introduction
The 2030 framework for climate and energy policies contains
abinding target to cut greenhouse gas emissions in the EU by at
least40% below 1990 levels by 2030 and has the ambition to further
reduce
them by 80–95% by 2050 (Delbeke and Vis, 2016). As theoretical
limitsof efficiency are being reached and process-related emissions
aresometimes inevitable, there is an urgent need to develop
efficientcarbon capture systems (Pachauri and Meyer, 2014). In the
past, mostresearch focused on the capture and storage of CO2, also
referred to as
https://doi.org/10.1016/j.biortech.2017.10.081Received 28 August
2017; Received in revised form 18 October 2017; Accepted 20 October
2017
⁎ Corresponding author.E-mail address: [email protected] (S.
Ghysels).
Bioresource Technology 249 (2018) 858–868
0960-8524/ © 2017 Elsevier Ltd. All rights reserved.
MARK
http://www.sciencedirect.com/science/journal/09608524https://www.elsevier.com/locate/biortechhttps://doi.org/10.1016/j.biortech.2017.10.081https://doi.org/10.1016/j.biortech.2017.10.081mailto:[email protected]://dx.doi.org/10.1016/j.biortech.2017.10.081http://crossmark.crossref.org/dialog/?doi=10.1016/j.biortech.2017.10.081&domain=pdf
-
Carbon Capture and Storage (CCS). Alternatively, CO2 could be
re-cognized as a valuable resource, which can be utilized for the
produc-tion of carbon-based chemicals with current or future demand
(Liuet al., 2015). This is known as carbon capture and utilization
(CCU) andcan provide much needed additional capacity in the move
towards alow-carbon economy (Cheah et al., 2016).
Producing plastics from CO2 instead of oil constitutes an
interestingcase for CCU. Polyhydroxyalkanoate (PHA) is a class of
biodegradablebioplastics (microbial polyesters) produced from
renewable resources(Doi, 1991). A two-phase fermentation process is
typically applied,comprising biomass growth, followed by PHA
accumulation undernutrient-limiting conditions (Johnson et al.,
2010; Grousseau et al.,2014). PHAs can be produced by various
prokaryotic species, the mostwidely studied one being Cupriavidus
necator (formerly referred to asRalstonia eutropha, Alcaligenes
eutrophus and Wautersia eutropha). C.necator has the capacity to
grow and accumulate PHA autotrophically;using CO2 as the sole
carbon source and H2 as energy source (Tanakaet al., 1995; Ishizaki
and Tanaka, 1991). Deploying C. necator for PHAproduction can thus
bridge CCU and bioplastic production (Kumaret al., 2017). So far,
most experimental work on PHA production fromCO2 has been conducted
in view of optimizing the production of thehomopolymer
poly(3-hydroxybutyrate) (PHB) (Mozumder et al.,
2015;Garcia-Gonzalez et al., 2015; Ishizaki and Tanaka, 1991).
Despite the interesting properties of PHB, its use as
commodityplastic is hampered by its stiffness, brittleness and low
impact strength.These physical properties can be improved by the
inclusion of othermonomers in the polymer (Kachrimanidou et al.,
2014; Ray and Kalia,2017). Indeed, copolymers such as
poly(3-hydroxybutyrate-co-3-hy-droxyvalerate) (PHBV) display
greater ductility and toughness com-pared to PHB (Pan and Inoue,
2009). Producing targeted PHBV
copolyesters could thus extent the range of PHA bioplastic
propertiesand applications. Targeted PHBV copolyester production
can beachieved by altering (i) the monomer distribution (e.g. block
copoly-mers or random copolymers) and (ii) the comonomer fractions
in thePHA copolymer (Wang et al., 2001; McChalicher and Srienc,
2007).
Copolymer production with organic substrates is well
known.However, very few reports investigated mixotrophic PHA
production inwhich CO2 is supplied in combination with an organic
co-substrate(Park et al., 2014; Volova and Kalacheva, 2005; Volova
et al., 2008;Volova et al., 2013). The upper part of Table 1
summarizes the state-of-the art on mixotrophic or heterotrophic PHA
copolyester productioncomprising 3HB and 3HV monomers, both in
therms of experimentaland modelling efforts. Table 1 indicates that
valeric acid is the com-monly investigated precursor of 3HV, while
CO2 is the precursor for3HB in mixotrophic conditions. Park et al.
(2014), Volova andKalacheva (2005), Volova et al. (2008) and Volova
et al. (2013) ex-clusively applied pulse feeding the various
organic co-substrates toevaluate e.g. their effect on the PHA
production and composition.Mixotrophic PHA copolymer production,
however, comprises a numberof process steps in which a lot of
process variables and other influencingfactors are involved
(Penloglou et al., 2012).
Modelling and simulation are very useful tools to understand
thedynamic process behavior, the underlying mechanisms and to
developcontrol strategies for maximizing PHA production (Novak et
al., 2015).The production of PHA copolyesters consisting of 3HB and
3HVmonomers from exclusively organic substrates was modelled
bySp̆oljarić et al. (2013) and Koller et al. (Dec 2006) (Table 1).
Bothstudies assumed no influence of the consumption of one
substrate to theconsumption rate of the other, while the results of
Park et al. (2014)indicated that such influences do take place
between CO2 and valeric
Table 1State-of-the-art of experimental and modelling efforts on
PHBV production, deploying either mixotrophic conditions, or
multi-substrate heterotrophic conditions. FAME: fatty acidsmethyl
esters, 3HB: 3-hydroxybutyrate, 3HV: 3-hydroxyvalerate, 4HB:
4-hydroxybutyrate, 3HHx: 3-hydroxyhexanoate, 3HHp:
3-hydroxyheptanoate, 3HO: 3-hydroxyoctanoate, VA:valeric acid,
PHBV: poly(3-hydroxybutyrate-co-3-hydroxyvalerate).
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
859
-
acid during mixotrophic fermentation. However, up until now no
modelexists for mixotrophic PHA copolymer production, which
requires in-clusion of such substrate interaction. This, together
with the improvedproperties of PHBV compared to PHB, prompted us to
focus on ex-perimentation and modelling of substrate interaction
and randomPHBV production with alterable 3HV content, which can be
set ac-cording to the envisaged applications.
This work is thus the first to study and model a mixotrophic
pro-duction process for the production of PHBV copolymers with
predefinedcomposition. Experiments consisted of a first stage of
cell mass growthusing glucose as substrate, followed by PHBV
copolymer productionusing CO2 and valeric acid as carbon sources
(Fig. 1). A pH-stat strategywas applied for the first time
throughout the whole mixotrophic PHBVcopolymer production phase.
The developed model was calibrated andvalidated using data from
independent experiments conducted in thisstudy. 1H-NMR, 13C-NMR and
chromatographic analyses of the PHBVcopolymers were performed to
assess whether the monomer distribu-tion and 3HV fraction of PHBV
were correctly predicted by the model.
2. Materials and methods
2.1. Experimental set-up
2.1.1. Organism and inoculumC. necator, strain DSM 545
(Leibniz-Institut, DSMZ GmbH,
Germany) was deployed as microorganism. Stock cultures were
storedat −20 °C in 2mL cyrovials containing 0.5 mL glycerol (85%,
Merck,Germany) and 1mL of a late exponential-phase liquid culture
in Lennoxbroth (LB) medium (Invitrogen, Life Technologies Europe
B.V.,Belgium). These stock cultures were used to inoculate
preculture 1 bytransferring 200 μL to 5mL of LB-medium in 15mL
test-tubes. Thepreculture was cultivated in an orbital shaker
(Innova 42, Eppendorf,USA) for 24 h at 30 °C and 200 rpm.
Subsequently, 2 mL of the strainwas sub-cultured during 24 h at 30
°C and 180 rpm in 100mL of pre-culture 2 seeding medium in 500mL
baffled flasks. Finally, the seedculture was used to inoculate the
bioreactor (12.5% v/v inoculum).Compositions of the culture and
fermentation media are specified inMozumder et al. (2014).
2.1.2. Bioreactor set-up and controlA 7 L, double jacked,
lab-scale bioreactor unit with EZ-control
system (Applikon Biotechnology, the Netherlands) for on-line
mon-itoring and controlling of the stirring speed, dissolved oxygen
(DO),foam formation, pH and temperature was used. Foam formation
wasmeasured through a level contact (conductivity) sensor and was
con-trolled by the addition of 30% antifoam C emulsion
(Sigma–AldrichChemie, GmbH, Germany). The process temperature was
measured by aplatinum resistance thermometer sensor (PT 100) and
kept constant at30 °C. The head space pressure of the bioreactor
was controlled by apressure transmitter (Keller, PR-35XHT) and a
pneumatic control valve(Badger Meter, ATC type 755) as back
pressure control valve.
Gas from the bioreactor outlet was continuously withdrawn via
aheated traced tubing using a gas sample pump (Bühler, PS2 Eexd) at
a
minimal flow rate of 164 Lmin−1 and dried by a gas cooler
(Bühler,EGK2 Ex). The condensate was returned to the bioreactor.
The gas wassplit in two streams. One stream was pumped through a
gas filter(Swagelok in-line filter, F-Series, 0.5μm) and a variable
area flowmeter(Krohne, DK 800/R/k1) to an on-line gas chromatograph
(GC)(MicroSAM, Siemens) to determine the gas composition (H2, O2,
CO2and N2) of the head space of the bioreactor. The GC was equipped
withthree micro thermal conductivity detectors and argon was used
as thecarrier gas. A second stream was resupplied to the bioreactor
through agas return line and a variable area flowmeter (Krohne, DK
800/R/k1).Dependent on the (over) pressure (setpoint) of the head
space in thebioreactor, a part of this stream was discharged to the
atmosphere. Thisvent was connected with the a gas counter
(Schlumberger, Gallus 2000)to monitor the gas exit.
Both phases of the heterotrophic-mixotrophic fermentation
processwere operated differently. The heterotrophic cell growth
phase was notsubject to further study in this work.
2.1.3. Operating conditions of the heterotrophic phaseThe first
phase was carried out according to Garcia-Gonzalez et al.
(2015). The temperature was set to 30 °C, the agitation speed
950 rpmand the pressure 1 bar. The DO concentration was maintained
around55% of air saturation during heterotrophic growth using a
cascadecontrol strategy consisting of the agitation speed (950
rpm), air and/oroxygen flow. These relatively high DO levels were
chosen to ensure thatheterotrophic growth was not limited by the O2
concentration. The pHwas controlled at 6.80 by adding acid (2M
H2SO4, 95–97%, Merck,Germany) or base (NH4OH 28.0–30.0% NH3 basis,
Sigma–AldrichChemie GmbH, Germany). The ammonium concentration was
main-tained between 0.60 and 0.71 g N L−1. Glucose (650 g L−1
Merck,Germany) was fed exponentially the first 10 h, followed by a
feedingregime based on alkali-addition to control the glucose
concentration at12 g L−1 (Mozumder et al., 2014).
2.1.4. Operating conditions of the mixotrophic phaseThe
mixotrophic phase was initiated once the biomass concentra-
tion reached approximately 15 −g L 1 by ceasing glucose addition
andreplacing NH4OH by NaOH for pH control. Once the nitrogen
con-centration dropped below 100mg L−1 due to consumption, the
agita-tion speed was increased to 1200 rpm and CO2 (industrial
X50S, AirProducts, Belgium), H2 (technical X50S, Air Products,
Belgium) and O2(industrial X50S, Air Products, Belgium) were
continuously spargedinto the bioreactor, keeping a constant gas
composition in the head-space of H2:O2:CO2= 84:2.8:13.2 vol% at
80mbar overpressure(Garcia-Gonzalez et al., 2015). Under such
conditions, nitrogen andoxygen became limited, triggering
biopolymer synthesis. Two valericacid feeding strategies were
tested: (i) pulse-feeding to study mixo-trophic conditions and
perform model calibration and (ii) semi-con-tinuous addition by a
pH-stat to produce targeted PHBV copolyestersand perform model
validation. Samples were taken at regular time in-tervals for
analysis.
Pulse-feeding was performed during a first experiment by
spikingvaleric acid (99%, Sigma–Aldrich Chemie, GmbH, Germany) into
the
Fig. 1. Schematic representation of thetested and modelled
set-up in this work.Carbon dioxide (CO2) and hydrogen gas (H2)are
supplied in aerobic conditions with va-leric acid to produce
specific poly(3-hydro-xybutyrate-co-3-hydroxyvalerate) (PHBV)via
fermentation with Cupriavidus necator.
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
860
-
bioreactor four times during the biopolymer production phase.
Threepulses at 2, 24 and 47 h were given to reach a valeric acid
concentrationof 1.5 g L−1. The fourth pulse was added after 68 h in
the biopolymerproduction phase, envisaging a valeric acid
concentration of 0.5 g L−1
and ensuring all valeric acid was consumed at the end of the
fermen-tation run. 2M H2SO4 was used for pH control.
Semi-continuous addition of valeric acid through a pH-stat
ap-proach (Huschner et al., 2015) was studied in a second
experiment,using 920.7 g L−1 valeric acid for pH control. At the
start of the mix-otrophic phase, 0.741mL valeric acid was added to
reach a mediumconcentration of 0.239 g L−1 valeric acid. The pH was
corrected to 6.80with NaOH. Consumption of valeric acid for 3HV
production initializedthe pH-stat cycle: when the pH of the mineral
medium increased uponvaleric acid consumption, extra valeric acid
was added to maintain thepH at optimum level. By doing so, the pH
and valeric acid inflow ratewere aimed constant.
2.1.5. Analytical proceduresThe concentrations of glucose,
ammonium-nitrogen ( +NH4 -N), cell
dry weight and 3HB concentrations were determined as in
Mozumderet al. (2014). To determine the 3HV concentrations, a
standard of PHBVwith 12mol% 3HV (Sigma-Aldrich Chemie, GmbH,
Germany) was in-cluded. The valeric acid concentration in the
medium was analyzed byHPLC (Agilent technology 1200 series,
Belgium) using a RefractiveIndex Detector (RID-10A Shimadzu,
Japan). Separation was achievedon a Agilent Hi-Plex H 8 μcolumn
Belgium (7.7×300mm) using
0.01M H2SO4 as mobile phase at a flow rate of 1.0mLmin−1 and
atemperature of 60 °C.
PHBV extraction was performed as in Mozumder et al. (2014).
Ex-tracted PHBV was subsequently dissolved in deuterated chloroform
(D-CCl3 99.8 atom% D, contains 0.03% (v/v) tetramethylsilane,
Sigma-Aldrich Chemie, GmbH, Germany) for 1H-NMR and 13C-NMR
analysis(Bruker Avance-III NanoBay NMR spectrometer, Germany).
The400MHz 1H-NMR spectra were observed at 30° C in D-CCl3 (10–12 −g
L 1
PHBV in D-CCl3) with a 4-s pulse repetition, 6000-Hz spectra
width,65 K data points, and 8 accumulations. The 1H noise decoupled
13C-NMR spectra were recorded at 30 °C in D-CCl3 (30–37 −g L 1)
with a 5-spulse repetition. 24,000 spectral width, 65 K data
points, and 1098accumulations.
2.2. PHBV production model
Four bioconversions were considered in modelling the
mixotrophicphase: (i) 3HB production on CO2, O2 and H2, (ii) 3HV
production onvaleric acid and O2, (iii) aerobic maintenance on 3HB
and (iv) aerobicmaintenance on 3HV.
2.2.1. Bioconversion stoichiometry and kineticsThe stoichiometry
and kinetics for these conversions are summar-
ized in Table 2 and Table 3, respectively. The stoichiometry for
3HBproduction on CO2, O2 and H2 was reported by Ishizaki and
Tanaka(1991):
Table 2Stoichiometric matrix for relevant bioconversions in
poly(3-hydroxybutyrate-co-3-hydro-xyvalerate) (PHBV) production.
Conversion 1: 3-hydroxybutyrate (3HB) production onCO2, O2 and H2.
Conversion 2: 3-hydroxyvalerate (3HV) production on valeric acid
and O2.Conversion 3: aerobic maintenance on 3HB. Conversion 4:
aerobic maintenance on 3HV.The first two columns are the carbon
sources, column three and four are the monomers andlast thee are
the other components. The components in gray were not included as
statevariables.
Table 3Production rates of 3HB and 3HV and maintenance rates on
3HB and 3HV with corresponding kinetics. Conversion 1:
3-hydroxybutyrate(3HB) production on CO2, O2 and H2. Conversion 2:
3-hydroxyvalerate (3HV) production on valeric acid and O2.
Conversion 3: aerobicmaintenance on 3HB. Conversion 4: aerobic
maintenance on 3HV.
Bioconversion Reaction rate
1
=ρ μ X3HB 3HB , with =μ3HB ⎛⎝
⎞⎠
⎛
⎝
⎜⎜⎜
⎞
⎠
⎟⎟⎟
⎛⎝
⎞⎠+ + +
+ + +( )μ HKH H OKO O
OKIO
COKCO CO
KPINN KPIN c n3HB
max 22 2
2
2 222
2
22 2
11 ·Val
2
=ρ μ X ,3HV 3HV with =⎛
⎝
⎜⎜⎜
⎞
⎠
⎟⎟⎟
⎛⎝
⎞⎠+ +
+ +μ μ O
KO OO
KIO
KKPIN
N KPIN3HV 3HBmax 2
2 222
2
Val
Val Val
3=ρ ms X ,ms,3HB 3HB with = + +( )( )ms ms K3HB 3HBmax 3HB3HB
3HB 3HB3HB 3HV
4=ρ ms X ,ms,3HV 3HV with = + +( )( )ms ms K3HV 3HVmax 3HV3HV
3HV 3HV3HB 3HV
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
861
-
+ + → +33H 12O 4CO C H O 30H O,2 2 2 4 6 2 2 (1)
where C H O4 6 2 represents the elemental composition of a
3HBmonomer. The production rate of 3HB (ρ3HB) from Mozumder et
al.(2015) was adapted for mixotrophic conditions. Limitation of the
3HBproduction rate by H2 and CO2 was modelled through Monod
kinetics.For O2, both limitation and inhibition were taken into
account (throughHaldane kinetics) as the O2 concentration exceeded
the inhibitionconstant from Tanaka et al. (1995). The inhibition of
3HB production inthe presence of a nitrogen source was accounted
for through a non-competitive inhibition equation (Sp̆oljarić et
al., 2013; Mozumder et al.,2014). The decrease of 3HB synthesis
from CO2 in the presence of non-limiting valeric acid
concentrations, observed by Park et al. (2014) wasaccounted for
through a generalized non-competitive inhibition equa-tion (Kwon
and Engler, 2005).
The conversion of valeric acid to a 3HV monomer was modelled
inthis work, using the generic conversion of organic substrate to
amonomer from Akiyama et al. (2003). However, both
experimentsconsistently indicated that one mole of valeric acid did
not yield onemole 3HV. Only a fraction f of valeric acid was
directly converted to3HV. The remainder ( −f1 ) was assumed to be
oxidized with O2 to CO2and H2O, in line with Sp̆oljarić et al.
(2013) (Table 1) and Akiyamaet al. (2003). The overall conversion
is given by Eq. (2):
+ ⎛⎝
− ⎞⎠
→ + − + −f f f fC H O 132
6 O C H O (5 4 )H O (5 5 )CO .5 10 2 2 5 8 2 2 2 (2)
Conversion of valeric acid to 3HB was assumed negligible
andtherefore not modelled. 3-Ketovaleryl-CoA, an intermediate in
theconversion of valeric acid to 3HV, can result in 3HB if it were
convertedto propionyl-CoA and subsequently decarboxylated to
acetyl-CoA inoxygen-unlimited conditions (Doi et al., Jun 1988).
Such a conversionof valeric acid to 3HB was however assumed
negligible, as oxygen waslimited during bioplastic accumulation in
this study, which limits oxi-dative decarboxylation of 3HV
precursors to 3HB precursors (Lefebvreet al., Mar. 1997). Also, CO2
was provided in high concentrations as co-substrate, which
thermodynamically does not favor decarboxylation of3HV precursors
to 3HB precursors. The effect of oxygen and nitrogen on3HV
production was assumed to be similar as for 3HB production.
Forvaleric acid, only limitation effects were taken into account
throughMonod kinetics, since the established concentrations were
sufficiently
low to prevent substrate inhibition.Maintenance denotes
substrate conversion to maintain basic cell
functions. In aerobic conditions, 3HB and 3HV are co-converted
withO2 to CO2 and H2O for energy:
+ → +C H O 92
O 4CO 3H O.4 6 2 2 2 2 (3)
+ → +C H O 6O 5CO 4H O.5 8 2 2 2 2 (4)
Limitations of 3HB and 3HV were modelled through Monod
ki-netics. In addition, the specific conversion rate of 3HB and 3HV
weremodelled proportional to the fraction of 3HB and 3HV in
PHBV.
2.2.2. Mass balancesMass balances were set up over the liquid
phase of the bioreactor,
which was assumed completely mixed. The fermentation
mediumdensity was assumed constant in time and equal to that of
water. PHBVwas modelled as two state variables, 3HV and 3HB
(Sp̆oljarić et al.,2013). The residual biomass concentration was
not included as statevariable, for no residual biomass growth
occurs in nutrient-limitingconditions. Medium concentration of the
gaseous substrates CO2, O2and H2 were also not modelled as state
variables, since their con-centrations were quasi constant during
the whole process. Detailedderivation of the mass balances is
provided in Supplementary in-formation. The final accumulation
rates of valeric acid, liquid volumeand monomers over time are:
= − −d tdt
F tV t
F tV t
tμ
X tVal( ) ( )Val( )
( )( )
Val( )Y
( )F combVal 3HV,3HVVal (5)
= =−
+ +dV tdt
F t Fρ
ρF
ρρ
Fρρ
( ) ( )ValF F
w
F
w
F
wVal acid alk
Val acid alk
(6)
= − + −d tdt
F tV t
μ X t ms X t3HB( ) ( )( )
3HB ( ) ( )3HB 3HB (7)
= − + −d tdt
F tV t
μ X t ms X t3HV( ) ( )( )
3HV ( ) ( )3HV 3HV (8)
t V t tVal( ), ( ), 3HB( ) and t3HV( ) represent the valeric
acid concentra-tion, volume, 3HB concentration and the 3HV
concentration respec-tively. X represents the residual biomass
concentration. Y3HVVal re-presents the 3HV production yield over
valeric acid (100f/102 fromTable 3). F F,Val acid and Falk denote
the flow rate of valeric acid, acid(H2SO4) and alkali (NaOH) into
the bioreactor. ρ ρ ρ, ,F F FVal acid alk and ρwrepresent the
density of the feeding solution for valeric acid, H2SO4,NaOH and
the medium respectively. ValF represents the valeric
acidconcentration of the feeding solution, while F t( ) represents
the totalinflow rate.
2.2.3. Model calibration and validationAn objective function J
θ( ( )) for model calibration (parameter esti-
mation) was defined in Eq. (9) by means of the sum of squared
errors:
∑= −=
J θ y t y t θ( ) ( ( ) ( , )) .i
N
i im
1
2
(9)
y t( )i represents a vector with the model outputs at time t, in
this casethe measured experimental valeric acid concentration, 3HB
con-centration and 3HV concentration. y t θ( , )i
m is a vector with the simu-lated values by the model for a
parameter set θ at time t. N is thenumber of experimental data.
During model calibration, the optimalparameter set is identified as
the one resulting in the lowest sum ofsquared errors: =θ J θargmin{
( )}opt . This minimization problem wasimplemented in MATLAB 6.1
(The MathWorks Inc.) through the build-in function fmincon, based
on the interior-point algorithm. The esti-mated parameters are the
kinetic parameters indicated in Table 4 aswell as the medium
concentration of the gaseous substrates CO2, O2 andH2 as in
Mozumder et al. (2015). This is motivated by the fact that
their
Table 4Model parameter values from parameter estimation,
calculations and assumptions (θopt).
Estimated parameter Value Unit
μ3HBmax 0.01808 − −[g 3HB g X h ]1 1
μ3HVmax 0.1243 − −[g 3HV g X h ]1 1
ms3HBmax 0.001123 − −[g 3HB g X h ]1 1
ms3HVmax 0.006361 − −[g 3HV g X h ]1 1
H2 9.5758e−05 −[g H L ]2 1
O2 0.004283 −[g O L ]2 1
CO2 9.7086e−4 −[g CO L ]2 1
KVal 9.0447 −[g Val L ]1
K3HV 5.4288 −[g 3HV L ]1
K3HB 7.5534 −[g 3HB L ]1
C 99.0287 [ ]n 1.0044 [ ]
Calculated parameters Value Unit
Y3HVVal = 0.64f100102
−[g 3HV g Val ]1
ValF 920.7 −[g Val L ]1
f 0.65 [ ]
Value-assumed parameters Value Unit
N 0 −[g N L ]1
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
862
-
concentrations were quasi constant during the whole
process.Model validation was performed with independent data from
a
distinct experiment by calculating validation errors, i.e. the
absolutedeviation of new experimental data for the valeric acid,
3HV and 3HBconcentrations from values predicted with the calibrated
model:
−y t y t θ| ( ) ( , )|i im .
2.2.4. Calculation of process performance parametersThe residual
cell concentration (RCC) is defined as:
= − −RCC CDW 3HB 3HV, where CDW,3HB and 3HV represent the
celldry weight and monomer concentrations respectively. The
fraction ofcomonomer 3-hydroxyvalerate at time t is calculated
as:
=+
F t tt t
( ) 3HV( )3HV( ) 3HB( )
.3HV(10)
Similarly, the PHBV content FPHBV (polymer fraction in the
micro-organisms) was calculated from the PHBV concentration and the
re-sidual biomass concentration (X):
= ++ +
F 3HV 3HB3HV 3HB X
.PHBV
The synthesis rate of PHBV (productivity) − −[g PHBV L . h ]1 1
wascalculated from the final PHBV concentration (PHBV) −[g PHBV L
]1 andthe duration of the process tΔ [h]: PHBV/ tΔ . The 3HB and
3HVsynthesis rate were calculated in a similar from the monomer
con-centrations and process duration.
The experimental yield of 3HV over valeric acid −[g 3HV g Val ]1
wascalculated from the accumulated 3HV concentration 3HV −[g 3HV L
]1after consumption of valeric acid Val −[g Val L ]1 : 3HV/Val. The
theore-tical maximal 3HV yield over valeric acid −[g 3HV g Val ]1
is 0.98 (massbased), obtained for =f 1 in f100 /102 from Table 2.
The experimentallymeasured fraction f from the maximal 3HV yield
over valeric acid was
calculated as the ratio of the experimental 3HV yield over
valeric acidand the maximal 3HV yield over valeric acid.
The dyad sequence distribution D and degree of randomness R
toanalyse the polymers microstructure were calculated according
toZagar et al. (2006).
3. Results and discussion
Two heterotrophic-mixotrophic fermentation experiments
wereconducted for PHBV production from CO2 and valeric acid. Data
from afirst experiment with valeric acid pulses and CO2 sparging
were used toestablish the stoichiometric conversion from valeric
acid to 3HV, tocalibrate the model and assess substrate
interaction. In a second ex-periment, a pH-stat mediated valeric
acid addition strategy was de-signed to produce predefined PHBV
copolyesters in mixotrophic con-ditions. Data from this experiment
were used for model validation.
3.1. Pulse-feeding valeric acid in mixotrophic conditions
The experimental results for PHBV synthesis, show a total
3HBconcentration of 14.4 g L−1 after 70 h in the mixotrophic phase
(Fig. 2),while the 3HV concentration amounted to 1.5 g L−1 (Fig.
2), corre-sponding to a PHBV content of 51.5% ( =X 15 g L−1).
3.1.1. 3HV conversion stoichiometryThe experimental 3HV yield
over valeric acid was 0.64
−
−( )0.81g 3HV g Val1.27g Val L 11 . The corresponding fraction
f, i.e. the ratio of experi-mental 3HV yield (0.64) over the
maximal 3HV yield over valeric acid(0.98), amounts to 0.65.
Substituting this value in Eq. (2) results in thestoichiometry for
the conversion of valeric acid to 3HV:
2 24 47 68Time (h)
0
0.5
1
1.5
2
Val
eric
aci
d co
ncen
trat
ion
(g.L
-1)
Model outputExperimental data
2 24 47 68Time (h)
0
0.5
1
1.5
2
3-H
ydro
xyva
lera
te c
once
ntra
tion
(g.L
-1)
Model outputExperimental data
2 24 47 68Time (h)
0
5
10
15
20
3-H
ydro
xybu
tyra
te c
once
ntra
tion
(g.L
-1)
Model outputExperimental data
2 24 47 68Time (h)
0
5
10
15
20
PH
BV
con
cent
ratio
n (g
.L-1
) Model outputExperimental data
(a) (b)
(c) (d)
Fig. 2. Comparison of concentration measurements with simulation
results of the calibrated model for (a) valeric acid, (b)
3-hydroxybutyrate, (c) 3-hydroxyvalerate and (d) PHBV
(poly(3-hydroxybutyrate-co-3-hydroxyvalerate)). Time zero denotes
the start of the mixotrophic phase.
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
863
-
+ → + +C H O 2.6O 0.65C H O 2.4H O 1.75CO .5 10 2 2 5 8 2 2 2
(11)
The obtained molar 3HV yield over valeric acid ( =f 0.65) is
ap-proximately twice as high as those obtained by Gahlawat and
Soni(2017), which were 0.28 −mol 3HV mol Val 1 and 0.30
−mol 3HV mol Val 1 for 2 −g L 1 and 4 −g L 1 valeric acid
respectively. Weassume this can be explained by the more equal
substrate preferencebetween the two organic substrates glycerol and
valeric acid in Gahlawatand Soni (2017). Compared to CO2, which is
converted through the en-ergy-demanding Calvin-Benson-Bassham (CBB)
cycle, valeric acid can beregarded as a more favorable substrate,
as will be motivated further inthis section. Therefore, the valeric
acid yield towards 3HV would behigher in mixotrophic conditions.
Results of (Park et al., 2014) supportthis reasoning. Indeed, the
obtained molar 3HV yield over valeric acid inmixotrophic condtions
varies between 0.53 −mol 3HV mol Val 1 and 0.57
−mol 3HV mol Val 1, which is closer to the values obtained in
our study.The deviation between our findings and those of Park et
al. (2014) maybe attributed to other factors, such as a different
gas composition. Parket al. (2014) applied
H2:O2:CO2=77.78:11.11:11.11 vol% as opposed
to(H2:O2:CO2=84:2.8:13.2 vol%) applied in this study, which also
notlays in the explosion range of hydrogen gas.
3.1.2. Model calibrationThe estimated parameters obtained upon
minimizing the objective
function (θopt) are represented in Table 4. The coefficients of
determi-nation, R2, were determined to compare the calibrated model
outputwith the experimental observations. R2 was 0.8336 for the
valeric acidconcentration, 0.4875 for the 3HV concentration and
0.9360 for the3HB concentration. The predicted PHBV concentration
is the sum of itsmonomer concentrations (3HB+3HV) and had a R2 of
0.9385. Thegood agreement between the simulation results and
experimentalmeasurements for the valeric acid concentrations, 3HB
and PHBVconcentrations is also apparent in Fig. 2. Only for the
last four 3HVconcentrations a small deviation of the predictions
was observed.Overall, the model captured most essential microbial
processes and canreasonably explain the experimental
observations.
3.1.3. Substrate interaction between CO2 and valeric acidThe
interaction between both substrates was assessed by focusing on
the predicted and experimental 3HB and 3HV concentrations just
be-fore and after the first valeric acid pulse (see Fig. 3). Before
the firstpulse, neither valeric acid, nor 3HV was present in the
reactor, while3HB accumulated at an apparent constant rate. The 3HB
consumptionrate for maintenance was so low (0.001123 − −g 3HB g X
h1 1, Table 4),compared to the 3HB production rate (0.01808 − −g
3HB g X h1 1,Table 4), that its contribution was not visible.
Immediately after thepulse, the 3HB production rate decreased
dramatically and the con-sumption rate of 3HB for maintenance
became visible (as a slightlynegative 3HB accumulation rate). As
valeric acid is converted, the 3HVconcentration increased.
However, at a valeric acid concentration of 1.14 g L−1
(indicated bythe arrows in Fig. 3), the 3HB concentration again
increased togetherwith 3HV, until all valeric acid was consumed. In
other words, themodel indicates the simultaneous generation of both
monomers be-tween 0 and 1.14 g L−1 valeric acid. This value was
therefore denotedas the ‘critical valeric acid concentration’.
Deriving the critical valericacid concentration after model
calibration guaranteed that all data wastaken into account in its
calculation. Also, the experimental measure-ments were insufficient
to adequately annotate a critical valeric acidconcentration, due to
the very fast valeric acid conversion. The criticalvaleric acid
concentration was identical for the other pulses(Supplementary
Information). For the subsequent experiment, the va-leric acid
concentration was therefore kept below 1 g L−1 to obtainrandom
PHBV.
Although this is not explicitly mentioned, the results of Park
et al.(2014) also indicated the existence of a critical valeric
acid con-centration, which in their study amounted to 0.46 −g L 1
valeric acid.This is about three times less than that obtained in
this study. Thisdifference may be attributed to the different
operating conditions.
Our findings and those of Park et al. (2014), thus indicate that
au-totrophic and heterotrophic monomer synthesis pathways can
coexist,but only if heterotrophic substrate is limiting. This is in
accordance withSchwartz et al. (2009) and Shimizu et al. (2015),
who evidenced the
620
Time (h)
0
0.5
1
1.5
2
2.5
Mon
omer
con
cent
ratio
n (g
.L-1
)
0
1.14
2.5
Val
eric
aci
d co
ncen
trat
ion
(g.L
-1)
3HB (predicted)3HB (experimental)3HV (predicted)3HV
(experimental)Valeric acid (predicted)Valeric acid
(experimental)
Valericacid pulse
Fig. 3. Detail of the experimental and simulated concentrations
of valeric acid and the monomers 3-hydroxybutyrate (3HB) and
3-hydroxyvalerate (3HV) just before and after the firstpulse in
Fig. 2. Time zero denotes the start of the mixotrophic phase.
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
864
-
coexistence of both the autotrophic and heterotrophic pathways
duringbacterial growth and PHA accumulation. For the energy
housekeepingof C. necator in heterotophic conditions, it is
beneficial to repress the ccbgenes involved in CO2 fixation by the
energy intensive CBB cycle.However, partial derepression occurs on
some substrates (e.g. fructose)to convert CO2 from oxidative
decarboxylation into 3HB monomers(Shimizu et al., 2015), in order
to circumvent adverse effects of CO2 oncell functioning. In our
study, we assume that the supplied CO2 was
converted to 3HB, rather than CO2 from decabroxylation of
valeric acidintermediates, because the presence of CO2
thermodynamically ham-pers such a decarboxylation.
The repression and derepression of the ccb genes is
achievedthrough the CbbR protein in C. necator (CbbRRE), which is
modulatedby metabolites signaling the nutritional state of the cell
to the cbbsystem (Esparza et al., 2015). Above the critical valeric
acid con-centration, CO2 fixation seems to be entirely repressed,
while at
0 5 10 15 20 25 30
Time (h)
0
5
10
15
20
25
30
35
Con
cent
ratio
n (g
.L-1
)
0
10
20
30
40
50
60
70
80
Per
cent
age
(%)
Dry cell weight
PHBV
Residual biomass
3HB
3HV
PHBV content
Fig. 4. Concentrations of dry cell weight, residual biomass,
poly(3-hydroxybutyrate-co-3-hydroxyvalerate) (PHBV)
3-hydroxybutyrate (3HB) and 3-hydroxyvalerate (3HV) during
themixotrophic phase of the experiment with pH-stat addition of
valeric acid under constant sparging of CO2, H2 and O2. Time zero
denotes the start of the mixotrophic phase.
Fig. 5. Model validation with independentdata for the prediction
of 3-hydroxybutyrate,3-hydroxyvalerate and
poly(3-hydro-xybutyrate-co-3-hydroxyvalerate) and the va-leric acid
concentrations. Time zero denotesthe start of the mixotrophic
phase.
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
865
-
limiting valeric acid levels, it is only partially repressed.
Given thispartial repressing, we suggest that a metabolite of the
valeric acid-to-3HV pathway signals the cell to mainly utilize
valeric acid, while CO2fixation to 3HB remains partly derepressed
at low levels of this valericacid-derived metabolite.
3.2. Semi-continuous valeric acid feeding in mixotrophic
conditions
3.2.1. Process performanceThe semi-continuous valeric acid
feeding rate was 0.003 L h−1 with
a valeric acid concentration of 0.75 ± 0.47 g L−1. Online
monitoringof the pH and cumulative addition of valeric acid and
NaOH over timeis displayed in the Supplementary Information. The
evolution of the drycell weight and monomers is represented in Fig.
4. In this experiment,the total 3HB concentration reached a maximal
level of 7.0 g L−1, while3HV amounted to 17.7 g L−1 within a 28 h
accumulation phase (Fig. 4).So, a total PHBV concentration of 24.7
g L−1 was achieved. The si-multaneous production of both monomers
in Fig. 4 is indicative for theproduction of random PHBV. The
measured residual biomass con-centration was initially 13.9 g L−1,
but decreased to 6.8 g L−1, possiblydue to larger valeric acid
additions during the pH-stat, causing in-hibitory effects and cell
lysis (Khanna and Srivastava, 2007). The PHBVcontent at the end of
the experiment was 78%
+− −(24.66 g PHBV L /31.49 g(PHBV X) L )1 1 , with a 3HV
fraction in PHBVof 0.72 − −(17.7 g 3HV L /24.66 g PHBV L )1 1 . The
obtained experimental3HV yield over valeric acid was 0.69 − −(17.7
g 3HV L /24.45 g Val L )1 1 ,resulting in a value of 0.70 for the
experimentally measured fraction ffrom the maximal 3HV yield over
valeric acid. This is close to the 0.65previously determined,
confirming the annotated stoichiometry in Eq.(11).
Garcia-Gonzalez et al. (2015) used identical autotrophic
conditions(H2:O2:CO2=84:2.8:13.2 vol%) to produce pure PHB and
reached apolymer concentration of 28 g L−1, which is comparable to
the24.7 g L−1 obtained in this work. Nevertheless, the productivity
fromGarcia-Gonzalez et al. (2015) was 0.17 − −g L h1 1 and
significantly lowerthan the productivity of 0.87 − −g L h1 1
obtained in this study. This isattributed to the higher
accumulation rate of heterotrophic valeric acid.In the study of
Park et al. (2014), both valeric acid and a gas mixture ofH2, O2
and CO2 (H2:O2:CO2= 77.78:11.11:11.11 vol%) were usedduring a six
day accumulation phase. The final PHBV concentration was1.07 −g L
1. The corresponding productivity of 0.007 − −g L h1 1 was
sig-nificantly lower than the 0.87 − −g L h1 1 obtained in this
experiment. This
may be due to a lower biomass concentration after the biomass
growthphase, although that concentration was not reported.
The 3HV synthesis rate was 0.63 − −g L h1 1, while the 3HB
synthesisrate was 0.25 − −g L h1 1. This contradicts with a
reported 39% slowerreactivity of the polymerase enzyme phaC towards
R-3HV-CoA (relativeto R-3HB-CoA), which are the last metabolites in
the pathway to PHBVmonomers (Zhang et al., Jul 2001). However,
macroscopic and micro-scopic phenomena may explain this
discrepancy. On a macroscopiclevel, valeric acid already is in the
liquid phase of the medium, whilethe gaseous substrate experiences
an additional resistance for masstransfer to the liquid phase. This
favors a faster conversion of valericacid into R-3HV-CoA and
subsequent addition of a 3HV monomer to thegrowing PHBV
copolyester. On a microscopic level, CO2 has to undergomuch more
enzymatic conversion steps in the CBB cycle compared tovaleric
acid, motivating a faster 3HV synthesis. Indeed, the reactivity
ofphaC only relates to R-3HV-CoA and R-3HB-CoA, not taking into
ac-count upstream processes leading to these metabolites.
Compared to Park et al. (2014), this work achieved a 525
timeshigher 3HV synthesis rate and a 26 times higher 3HB synthesis
rate. Thesignificantly higher 3HV synthesis rate can be attributed
to the constantpresence of valeric acid by the pH-stat approach.
Indeed, Park et al.(2014) only provided valeric acid once or twice
at the beginning of theaccumulation phase. The obtained higher 3HB
synthesis rate in thiswork may indicate that process conditions
from Garcia-Gonzalez et al.(2015) are more optimal than those of
Park et al. (2014).
3.2.2. Model validationFor a series of substrate, monomer and
polymer concentrations, the
validation errors (absolute deviations) between the new
experimentaldata and the model outputs are represented in Fig. 5.
The predicted3HB, 3HV and PHBV concentrations are close to the
experimental data(Fig. 5). The 3HB error/measurement ratio was on
average 10%, with amaximum of 31% for the fourth data point, while
the 3HV error/measurement ratio was on average 21%, with a maximum
of 51% forthe third data point. For PHBV, the PHBV
error/measurement ratio was14% on average, with a maximum of 37%
for the third data point. Forthe valeric acid concentration, the
validation errors were low at thebeginning of phase two, but larger
at the end. The valeric acid error/measurement ratio was on average
134%, with a maximum of 321% forthe fourth data point. These larger
biases are caused by larger varia-tions in valeric acid
concentrations, probably due to sub-optimal pH-stat control
(Supplementary Information) and the observed interferenceof cell
lysis prompted by temporary higher valeric acid concentrationson
HPLC analysis. However, as the goal of the model was to predict
theproduction of PHBV and its monomers, rather than the valeric
acidconcentration, the calibrated model is valid for its purpose
and wasused to set up an operating diagram for tailored synthesis
of PHBV.
3.2.3. Operating diagram for targeted PHBV productionThe
calibrated model was applied to predict the 3HV fraction and
PHBV content evolution over time for various valeric acid inflow
ratesand a residual biomass concentration of 13.9 g L−1. The
results weresummarized in an operating diagram, displayed in Fig.
6. When appliedto the second experiment (valeric acid inflow rate
of 0.003 g L−1 for28 h mixotrophic fermentation), a 3HV fraction of
=F 0.703HV waspredicted, matching the experimentally measured value
of 0.72 veryaccurately. The operating diagram also indicated a PHBV
content
=F 0.613HV of (61%), while the experimentally obtained one
was78.30%. The difference is attributed to the decrease of residual
biomassover time (from 13.9 g L−1 to 6.8 g L−1 in Fig. 4), while
simulationswere conducted assuming a constant concentration of
initial residualbiomass. The PHBV content calculated with the final
residual biomassconcentration was 60%, which is close to its
prediction.
The operating diagram from Fig. 6 can thus be applied to
producePHBV copolyesters with a predefined composition. It
demonstrates thatlow 3HV fractions (
-
(
-
Schwartz, E., Voigt, B., Zühlke, D., Pohlmann, A., Lenz, O.,
Albrecht, D., Schwarze, A.,Kohlmann, Y., Krause, C., Hecker, M.,
Friedrich, B., 2009. A proteomic view of thefacultatively
chemolithoautotrophic lifestyle of Ralstonia eutropha H16.
Proteomics9 (22), 5132–5142.
Shimizu, R., Dempo, Y., Nakayama, Y., Nakamura, S., Bamba, T.,
Fukusaki, E., Fukui, T.,2015. New insight into the role of the
calvin cycle: Reutilization of co2 emittedthrough sugar
degradation. Sci. Rep. 5 11617 EP –.
Sp̆oljarić, I.V., Lopar, M., Koller, M., Muhr, A., Salerno, A.,
Reiterer, A., Malli, K., Angerer,H., Strohmeier, K., Schober, S.,
Mittelbach, M., Horvat, P., 2013. Mathematicalmodeling of
poly[(R)-3-hydroxyalkanoate] synthesis by Cupriavidus necator
DSM545 on substrates stemming from biodiesel production. Bioresour.
Technol. 133,482–494.
Tanaka, K., Ishizaki, A., Kanamaru, T., Kawano, T., 1995.
Production of poly(D-3-hy-droxybutyrate) from CO2, H2, and O2 by
high cell density autotrophic cultivation ofAlcaligenes eutrophus.
Biotechnol. Bioeng. 45 (3), 268–275.
Volova, T.G., Kalacheva, G.S., 2005. The synthesis of
hydroxybutyrate and hydro-xyvalerate copolymers by the bacterium
Ralstonia eutropha. Microbiology 74 (1),54–59.
Volova, T.G., Kalacheva, G.S., Steinbüchel, A., 2008.
Biosynthesis of multi-componentpolyhydroxyalkanoates by the
bacterium Wautersia eutropha. Macromol. Symp. 269(1), 1–7.
Volova, T.G., Kiselev, E.G., Shishatskaya, E.I., Zhila, N.O.,
Boyandin, A.N., Syrvacheva,D.A., Vinogradova, O.N., Kalacheva,
G.S., Vasiliev, A.D., Peterson, I.V., 2013. Cellgrowth and
accumulation of polyhydroxyalkanoates from CO2 and H2 of a
hydrogen-oxidizing bacterium, Cupriavidus eutrophus B-10646.
Bioresour. Technol. 146,215–222.
Wang, Y., Yamada, S., Asakawa, N., Yamane, T., Yoshie, N.,
Inoue, Y., 2001. Comonomercompositional distribution and thermal
and morphological characteristics of
bacterialpoly(3-hydroxybutyrate-co-3-hydroxyvalerate)s with high
3-hydroxyvalerate con-tent. Biomacromolecules 2 (4), 1315–1323
pMID: 1177740.
Z̆agar, E., Krzăn, A., Adamus, G., Kowalczuk, M., 2006.
Sequence distribution in
microbialpoly(3-hydroxybutyrate-co-3-hydroxyvalerate) co-polyesters
determined by NMRand MS. Biomacromolecules 7 (7), 2210–2216 pMID:
1682758.
Zhang, S., Kamachi, M., Takagi, Y., Lenz, R., Goodwin, S., Jul
2001. Comparative study ofthe relationship between monomer
structure and reactivity for two poly-hydroxyalkanoate synthases.
Appl. Microbiol. Biotechnol. 56 (1), 131–136.
S. Ghysels et al. Bioresource Technology 249 (2018) 858–868
868
http://refhub.elsevier.com/S0960-8524(17)31910-7/h0145http://refhub.elsevier.com/S0960-8524(17)31910-7/h0145http://refhub.elsevier.com/S0960-8524(17)31910-7/h0145http://refhub.elsevier.com/S0960-8524(17)31910-7/h0145http://refhub.elsevier.com/S0960-8524(17)31910-7/h0150http://refhub.elsevier.com/S0960-8524(17)31910-7/h0150http://refhub.elsevier.com/S0960-8524(17)31910-7/h0150http://refhub.elsevier.com/S0960-8524(17)31910-7/h0155http://refhub.elsevier.com/S0960-8524(17)31910-7/h0155http://refhub.elsevier.com/S0960-8524(17)31910-7/h0155http://refhub.elsevier.com/S0960-8524(17)31910-7/h0155http://refhub.elsevier.com/S0960-8524(17)31910-7/h0155http://refhub.elsevier.com/S0960-8524(17)31910-7/h0160http://refhub.elsevier.com/S0960-8524(17)31910-7/h0160http://refhub.elsevier.com/S0960-8524(17)31910-7/h0160http://refhub.elsevier.com/S0960-8524(17)31910-7/h0165http://refhub.elsevier.com/S0960-8524(17)31910-7/h0165http://refhub.elsevier.com/S0960-8524(17)31910-7/h0165http://refhub.elsevier.com/S0960-8524(17)31910-7/h0170http://refhub.elsevier.com/S0960-8524(17)31910-7/h0170http://refhub.elsevier.com/S0960-8524(17)31910-7/h0170http://refhub.elsevier.com/S0960-8524(17)31910-7/h0175http://refhub.elsevier.com/S0960-8524(17)31910-7/h0175http://refhub.elsevier.com/S0960-8524(17)31910-7/h0175http://refhub.elsevier.com/S0960-8524(17)31910-7/h0175http://refhub.elsevier.com/S0960-8524(17)31910-7/h0175http://refhub.elsevier.com/S0960-8524(17)31910-7/h0180http://refhub.elsevier.com/S0960-8524(17)31910-7/h0180http://refhub.elsevier.com/S0960-8524(17)31910-7/h0180http://refhub.elsevier.com/S0960-8524(17)31910-7/h0180http://refhub.elsevier.com/S0960-8524(17)31910-7/h0185http://refhub.elsevier.com/S0960-8524(17)31910-7/h0185http://refhub.elsevier.com/S0960-8524(17)31910-7/h0185http://refhub.elsevier.com/S0960-8524(17)31910-7/h0190http://refhub.elsevier.com/S0960-8524(17)31910-7/h0190http://refhub.elsevier.com/S0960-8524(17)31910-7/h0190
Targeted poly(3-hydroxybutyrate-co-3-hydroxyvalerate) bioplastic
production from carbon dioxideIntroductionMaterials and
methodsExperimental set-upOrganism and inoculumBioreactor set-up
and controlOperating conditions of the heterotrophic phaseOperating
conditions of the mixotrophic phaseAnalytical procedures
PHBV production modelBioconversion stoichiometry and
kineticsMass balancesModel calibration and validationCalculation of
process performance parameters
Results and discussionPulse-feeding valeric acid in mixotrophic
conditions3HV conversion stoichiometryModel calibrationSubstrate
interaction between CO2 and valeric acid
Semi-continuous valeric acid feeding in mixotrophic
conditionsProcess performanceModel validationOperating diagram for
targeted PHBV productionNMR spectroscopic validation of targeted
PHBV production
Implications, limitations and future perspectives
ConclusionsAcknowledgementsSupplementary dataReferences