1 s2.0 S135951132400237X Main

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Process Biochemistry 146 (2024) 304–315

Contents lists available at ScienceDirect

Process Biochemistry
journal homepage: www.elsevier.com/locate/procbio

A mathematical model-based evaluation of yeast extract’s effects on


microbial growth and substrate consumption for lactic acid production by
Bacillus coagulans
Agata Olszewska-Widdrat a, * , Majharulislam Babor a , Marina M.-C. Höhne a, b , Maria Alexandri c ,
Jose Pablo López-Gómez a, d , Joachim Venus a
a
Leibniz Institute for Agricultural Engineering and Bioeconomy, Max-Eyth-Allee 100, Potsdam 14469, Germany
b
Department of Computer Science, University of Potsdam, An der Bahn 2, Potsdam 14476, Germany
c
Ionian University, Department of Food Science and Technology, Vergoti Ave. Argostoli 28100 Kefalonia, Greece
d
AINIA, Parque tecnológico de Valencia, c/ Benjamin Franklin, 5-1, E46980 Paterna, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: The microbial production of lactic acid, a crucial platform chemical for the development of biobased products,
Bacillus coagulans aims to replace fossil-based materials, although current production costs remain relatively high. Fermentation
Lactic Acid requires the addition of yeast extract (YE) to exclude nutritional lag, which increases the cost of production;
Kinetic model
therefore, optimizing its amount is essential. This study explores batch fermentation with varying concentrations
Fermentation
of YE (5 %, 10 %, 15 % and 20 %) in a medium using the strain Bacillus coagulans A166 to produce lactic acid
from glucose. The results reveal that the level of YE significantly influences the product yield, substrate con-
sumption rate, and process productivity. The kinetic model applied to experimental data indicates that higher YE
concentrations lead to increased glucose uptake and microbial growth rates, while concentrations exceeding
10 % result in reduced performance. It was determined that a concentration of 10 % YE in the medium delivered
the highest maximum specific rates for glucose consumption and microbial growth, leading to a 0.99 g/g yield of
glucose lactic acid. Furthermore, a real-time monitoring strategy for glucose consumption and lactic acid for-
mation was established using a mathematical model based on the rate of alkaline consumption throughout
fermentation. The cross-validation results demonstrate the reliability of this approach, with mean absolute errors
(expressed as a percentage of the concentration range) for the prediction of glucose ranging from 2.8 % to 5.0 %
and for the prediction of lactic acid ranging from 3.4 % to 5.4 %.

1. Introduction friendly alternative to the conventional petrochemical route, producing


a high purity product [19]. Notably, lactic acid serves as a fundamental
The manufacture of lactic acid (LA) plays a key role as a platform building block for the production of poly-lactic acid (PLA), making it a
chemical in the shift from fossil-based production to white biotech- substantial bulk chemical [16,20,21]. In 2021, the lactic acid market
nology [1,2]. In this microbial process of production of lactic acid, was valued at USD 2.9 billion, with an anticipated Compound Annual
additional supplementation with nutrients, especially in the form of a Growth Rate (CAGR) of 18.7 % [22].
nitrogen source, is essential to support the optimal growth of lactic As the demand for lactic acid continues to surge, sustainable sup-
acid-producing bacteria and meet their metabolic demands [3-10]. plementation becomes a fundamental strategy for cost reduction, with a
Lactic acid, characterized as an organic acid and a C3 monomer, boasts particular focus on yeast extract (YE). In typical batch fermentation
diverse applications in the food, chemical and pharmaceutical sectors processes, YE is incorporated to provide the bacterial strain with the
[11-17]. Although lactic acid has traditionally been synthesized by optimal nutrients required for growth and product formation. Due to its
chemical means, there is a growing trend toward its biotechnological high nitrogen content, YE is recognized for enhancing lactic acid pro-
production [18]. Fermentative production offers an environmentally duction [23], but this advantage comes at the expense of increased

* Corresponding author.
E-mail addresses: aolszewska-widdrat@atb-potsdam.de (A. Olszewska-Widdrat), malexandri@ionio.gr (M. Alexandri), plopez@ainia.es (J.P. López-Gómez).

https://doi.org/10.1016/j.procbio.2024.07.017
Received 17 April 2024; Received in revised form 20 June 2024; Accepted 9 July 2024
Available online 11 July 2024
1359-5113/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

production costs for additional YE expenses. Table 1


This study used strain Bacillus coagulans A166 [24], recognized for its Nomenclature.
significant production of lactic acid, although it does not fall into the Measurement
Symbol Description
category of lactic acid bacteria (LAB) [3,25,26]. This facultative unit.
anaerobic strain demonstrates exceptional growth capabilities across a B Biomass concentration [g/L]
wide range of conditions. It can grow at extreme temperatures, such as G Glucose concentration [g/L]
61 ◦ C, and in highly alkaline environments with a pH range of 10.5–11.0, L Lactic acid concentration [g/L]
although its optimal temperature and pH ranges lie between 50 and N Free amino nitrogen concentration [mg/L]
A Alkaline (base) consumption [L]
55 ◦ C [27-30] and 4.0–7.0 [30,31], respectively. Due to its versatility, V Working volume of the bioreactor [L]
robustness, and efficient carbon utilization, B. coagulans represent an D Dilution rate [h− 1]
attractive resource for biotechnological applications, not only in the µB Specific growth rate for biomass [h− 1]
production of lactic acid, but also in other areas. In addition to its ap- µB,max Maximum specific growth rate for biomass [h− 1]
Specific glucose uptake rate [h− 1]
plications in the production of thermostable enzymes and antimicrobial ΩG
ΩG,max Maximum specific glucose uptake rate [h− 1]
peptides like coagulin [32-34], B. coagulans has the potential to serve as ΩN Specific amino nitrogen uptake rate [h− 1]
a dietary additive for probiotic and antibiotic purposes. ΩN,max Maximum specific amino nitrogen uptake rate [h− 1]
Understanding the interactions between microbial activity, chemical YBG Yield coefficient from glucose to biomass [g/g]
substrates, and productivity in biotechnological processes is challenging YBN Yield coefficient from amino nitrogen to biomass [g/mg]
YLG Yield coefficient from glucose to lactic acid [g/g]
due to their complex dynamics. Relying solely on experimental data for Yield coefficient from glucose to alkaline
various setups is demanding, requiring significant amounts of data and YAG [L/g]
consumption
time. Therefore, the implementation of mathematical models signifi- t Time [h]
cantly advances comprehension of complex biotechnological phenom- KG Saturation parameter for glucose [g/L]
Saturation parameter for free amino nitrogen-
ena and the interaction between various process parameters. KN [mg/L]
source
Mathematical models for fermentation illustrate how the dynamic I Inhibitory effect of lactic acid concentration [1]
behavior of microbial growth impacts the process variables, including Representation of the effect of glucose on initial lag
∝G [1]
substrate consumption and product formation. In such cases, mathe- phase
matical modeling is a widely employed tool, facilitating researchers and Representation of the effect of free amino nitrogen
∝N [1]
on initial lag phase
enterprises in simulating system behaviors and optimizing processes for
n Exponent to represent the toxic power of lactic acid [1]
enhanced productivity. The application of mathematical models has Representation for the physiological state of the
qG [1]
been extensively reported in various bioprocesses, including lactic acid inoculum, glucose
production [27,35-41], propionate production [42], lipid production qN
Representation for the physiological state of the
[1]
inoculum, free amino nitrogen
[43] and ethanol production [44-46]. In this context, mathematical
Lmax Maximum lactic acid concentration [g/L]
models play a crucial role, not only in simulating, monitoring and Bmax Maximum biomass concentration [g/L]
controlling the processes but also in advancing the comprehension of dB/dt Biomass formation rate [g/L]
microbial activity, enabling the optimization of key parameters for dG/dt Glucose accumulation rate [g/L]
enhanced efficiency. dN /dt Free amino nitrogen accumulation rate [g/L]
dL/dt Lactic acid formation rate [g/L]
In this study, we investigate the performance of the strain Bacillus
dA/dt Alkaline accumulation rate [g/L]
coagulans A166 in batch fermentation systems while considering varying dV/dt Working volume accumulation rate [g/L]
concentrations of a nitrogen-containing nutrient, yeast extract. The At Consumed alkaline at time t [L]
main contributions of this paper were as follows: We introduce a At− 1 Consumed alkaline at time t-1 [L]
Kinetic parameter to represent the rate of glucose
comprehensive mathematical model designed to simulate the existing rGA [g/L2]
consumption with alkaline consumption rate
batch fermentation system, enabling the identification and quantifica- Kinetic parameter to represent the rate of lactic acid
tion of key process parameters that illustrate microbial activity. These rLA [g/L2]
formation with alkaline consumption rate
parameters include the maximum glucose uptake rate, the amino ni- PG Productivity [g/L/h]
trogen uptake rate, and the microbial growth rate. The dynamics of Ltotal Total lactic acid produced [g]
Ct Fermentation process completion time [h]
microorganisms in the presence of varying concentrations of amino ni-
trogen provide information on the optimal conditions for the fermen-
tation process to be effective. The composition and consumption profile
of amino acids during the fermentation process was investigated.
Additionally, a real-time monitoring strategy for glucose consumption
and lactic acid formation was established through a mathematical
model-based approach, utilizing the rate of alkaline consumption across
the fermentation.Table 1

2. Materials and methods

2.1. Microorganism and preculture preparation

The B. coagulans strain A166, utilized in this study as shown in Fig. 1,


was initially isolated from hemp at the Leibniz Institute for Agricultural
Engineering and Bioeconomy (ATB), Potsdam, and is an integral part of
the institute’s own strain collection. Comprehensive characterization of
this strain was performed by the Leibniz Institute Deutsche Sammlung
von Mikroorganismen und Zellkulturen (DSMZ) GmbH. Furthermore,
matrix-assisted laser desorption/ ionization time-of-flight mass spec-
Fig. 1. : Transmission electron microscope (TEM) image of B. coagulans
trometry (MALDI-TOF MS) was utilized for the detailed characterization
A166 strain.
of the B. coagulans A166 strain. Inoculum preparation was performed in

305
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

shake flasks, using De Man Rogosa and Sharpe (MRS) medium (Merck 2.3.2. Sample preparation for cell dry weight (CDW) analysis
KGaA, Germany) with Everzit®Dol (0.5–2.5 mm) (Evers e.K., Germany) For CDW analysis, 5 mL of each sample was transferred to plastic test
as a buffering system. The precultures were incubated at 52 ◦ C while tubes (100 ×16 mm, VWR International, Germany). Samples were
agitating at 100 rpm for 17 hours. For every 30 mL of MRS medium, centrifuged at 4920 x g, at 6 ◦ C, for 15 minutes using the SIGMA 4–16KS
cells from a single agar slant were used. centrifuge (Sigma Laborzentrifugen GmbH, Germany). The supernatant
was decanted, and 2 mL of Milli-Q water was added to each tube to wash
2.2. Fermentation the cell pellets. After another round of centrifugation, cell pellets were
transferred to pre-dried and weighed ceramic crucibles (VWR Interna-
Laboratory-scale fermentations were performed using 5 L BIOSTAT tional, Germany). The cell mass was dried until a constant weight was
bioreactors (Sartorius AG, Goettingen, Germany) with a working vol- achieved, at 105 ◦ C, over a period of 24 hours. Subsequently, the dried
ume of 3 L. The fermentation conditions were as follows: a temperature samples were weighed using an ALJ 220 – 4 M balance (Kern & Sohn
of 52 ◦ C and agitation at 300 rpm. The pH was maintained consistently GmbH, Germany).
at 6 employing a 20 % (w/w) NaOH solution. A standard inoculation
volume of 6 % (v/v) was used for all laboratory-scale experiments. A 2.3.3. Amino acid analysis
synthetic medium containing 100 g/L of glucose and YE was utilized, Amino acid analysis was performed using a SYKAM Amino Acid
with variations in the concentration of YE. In batch experiments, YE Analyzer. Samples from batch experiments with yeast extract (YE)
concentrations of 5, 10, 15 and 20 g/L were tested. The conditions for concentrations of 5, 10, 15, and 20 g/L were analyzed for amino acid
this strain were based and selected according to [47,48]. content at the beginning of fermentation and after 3 hours of fermen-
tation, shortly before feeding.
2.3. Analytical assays
2.3.4. Transmission electron microscope analysis
The stored samples were thawed and subjected to centrifugation For electron microscopy, B. coagulans cells were concentrated by
(SIGMA 4–16KS, Sigma Laborzentrifugen GmbH, Germany) at 4920 × g centrifugation (5000 rpm, 5 minutes, 4 ◦ C). The cells were adsorbed
and 6 ◦ C for 15 minutes. The resulting supernatant was diluted 1:20 with onto copper grids and rinsed once with distilled water. The samples were
distilled water and then filtered through a 0.2 µm filter (Chromafil CA - imaged using a Zeiss 912 Omega transmission electron microscope at
20/25, Macherey - Nagel GmbH & Co. KG, Germany). For the mea- 120 kV.
surements, 1.5 mL glass vials (Macherey - Nagel GmbH & Co. KG, Ger-
many) were prepared, while backup samples were stored in 2 mL cryo- 2.4. Kinetic model for B. coagulans
tubes (Th. Geyer GmbH & Co. KG, Germany) at − 80 ◦ C.
The concentrations of glucose and lactic acid were determined using In practice, the growth of bacteria during a complete fermentation
high-performance liquid chromatography (HPLC) (UltiMate® 3000, exhibits distinct phases, which depend on multiple interconnected fac-
Thermo Fisher Scientific Dionex, USA). A mobile phase consisting of tors that can be represented as a system of mathematical equations.
0.01 N H2SO4 was pumped through the system at a flow rate of 0.8 mL/ Given specific initial conditions, such as initial biomass, glucose, and
min. The measurements were conducted at 35 ◦ C and a pressure of lactic acid concentrations, a mathematical model can simulate the sys-
1.5 MPa. A refractive index detector (RI-71) was employed for substance tem’s dynamics throughout the fermentation. The problem can be
detection. An auto-sampler (GINA 50 T, Fa. Dionex, USA) injected 10 µL approached in two stages. Firstly, a mathematical representation of the
of each sample into the system. To measure the L/D ratio, the same interdependencies and dynamics among process variables (e.g.,
HPLC system was utilized. Samples were diluted 1:100 using 1 mM biomass, glucose, lactic acid), which relies on understanding the
Cu2SO4 instead of water and were then analyzed using a CHIRALPAK® fermentation system. Secondly, optimization of the process parameters
MA (+) column (50 ×4.6 mm, 3 µm), connected to the HPLC system. (e.g., microbial growth rate) in the mathematical model to make it
The mobile phase (2 mM Cu2SO4 + 0.1 % ACN) was pumped at a flow representative of the studied fermentation. In this study, a mathematical
rate of 0.8 mL/min. The working conditions were adjusted to 30 ◦ C and model was applied that describes the time-dependent dynamics of
a pressure of 58 bar. A UV detector (236 nm) and auto-sampler (GINA biomass (B), glucose (G), and lactic acid (L) using experimental data
50 T, Fa. Dionex, USA) were employed, and 10 µL of each sample was derived from the fermentation of B. coagulans.
injected into the system. The structure of mathematical models presented in publications
The free amino nitrogen concentration (N) was determined using the varies, however, the range of models employed remains relatively
ninhydrin colorimetric method as outlined in the European Brewery consistent considering the diversity in microorganisms and their growth
Convention from 1987. The color reagents were dissolved in 1 L of in different combinations of substrates, essential nutrients and products
distilled water, maintaining the pH between 6.6 and 6.8 through the [38,49]. In the mathematical modeling of lactic acid production, several
cautious addition of KH2PO4. The mixture of dilution reagents was filled critical factors demand consideration, including inhibitory activities,
up to 1 L with distilled water. The sample preparation process involved nutrient limitations, and the dynamic interactions between microbial
the following steps: growth and metabolite production, among others. These key aspects
play an essential role in shaping the comprehensive model that captures
2.3.1. Sample preparation for free amino nitrogen analysis the complexities of fermentation. Glazer and Venus [37] conducted a
Samples were diluted 1:200, resulting in 2 mL in glass test tubes (160 study on the batch fermentations of B. coagulans using mixed substrate
×16 mm, VWR International, Germany). Four different concentrations medium, including glucose, xylose, and arabinose as carbon sources,
of 2 mL glycine standard solution (0.5, 1, 1.5, and 2 mg/L) were trans- with the aim of developing a mathematical model. The authors have
ferred to separate test tubes. Distilled water (2 mL) was used as a blank. explained the three distinct phases with additive Monod type of growth
Subsequently, 1 mL of the color reagent was added to each tube, and all considering product inhibitory effect. The B. coagulans strains that were
tubes were mixed on a shaker (VWR International, Germany) before studied appeared to consume no substrates and produce no lactic acid
sealing with aluminum foil. The sealed tubes were placed in a boiling during the stationary phase, despite the presence of substrates of lower
water bath (1003, GFL, Germany) for 16 minutes, followed by cooling concentrations. Furthermore, the consumption of free amino nitrogen
on ice for an additional 20 minutes. To each tube, 5 mL of the dilution was absent in the proposed model, given the consistent concentration of
reagent was added after shaking. The tubes were shaken once more, and yeast extract added in all fermentations (15 %) [36]. However, because
the absorbance at 570 nm was measured using a spectrophotometer (UV this research employed a different experimental setup with glucose as
1, Spectronic Unicam, UK). the only carbon source, their proposed model was not applicable.

306
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

Several studies described the growth of B. coagulans using the Ver- substrate concentration while accounting for the dilution rate (D) using
hulst model [50] or a logistic model and product formation employing Eq. 7. The dilution rate was influenced by the addition of an alkaline to
the Luedeking-Piret function to account for production when microbial control the pH level of the medium. Eq. 8 provided insights into the
growth is zero [27,43]. In this study, a mathematical model was applied temporal dynamics of glucose consumption. The simulation of amino
that accounts for the specific glucose uptake rate ΩG, specific amino nitrogen consumption by B. coagulans was performed using Eq. 9.
nitrogen uptake rate ΩN, and specific microbial growth rate µB to Furthermore, Eq. 10 outlined the lactic acid production, primarily
replicate the nutrient consumption during the stationary growth phase influenced by the glucose uptake rate and the conversion coefficient
of B. coagulans. To model the process of lactic acid production from from glucose to lactic acid (YLG), which was determined based on
glucose by using B. coagulans, the system of equations can be described experimental data. To express the rate at which alkaline is added to the
as follows. The specific growth rate (µB) was calculated by using only fermentation system for pH regulation, considering the variation in
current biomass concentration (B), as represented in Eq. 1. Lactic acid glucose uptake rate, the Eq. 11 with a coefficient denoted as YAG,
production inhibition was quantified using Eq. 2 at any fermentation quantifying the alkaline needed per unit of glucose uptake. The alkaline
stage with the degree of this effect characterized by the parameter n, concentration remained constant at 20 % for all fermentations and was
where with n > 1 the inhibition term (I) exhibits a hyperbolic behavior. omitted from the model for the sake of simplicity. Finally, Eq. 13 was
Eq. 3 integrated the role of the physiological state of the inoculum in utilized to update the dilution factor, accounting for the introduction of
adapting to a new environment with a parameter qG determining the the alkaline into the fermentation medium that was measured during the
quickness at which it shifts from the lag phase to the exponential growth experimental run.
phase [51]. The influence of the amount of free amino nitrogen on the
adaptation of the inoculum was described using Eq. 4.
( ) 2.5. Process parameters optimization
B
µB = µB,max 1− (1)
Bmax
To apply the proposed mathematical model, the system of differen-
( )n tial equations representing the dynamics of concentrations of process
1− L
I= (2) variables: biomass (B), glucose (G), lactic acid (L), free amino nitrogen
Lmax
(N), and alkaline consumption (A) had to be solved, which was done by
qG using the 4th order Runge-Kutta method. To implement the solution
∝G = (3) method and simulating the fermentation process, Python programming
t)
qG − e(− ΩG,max •
language (version 3.10) was used. However, despite having the experi-
qN mental data, eight essential process kinetic parameters, namely the
∝N = (4) maximum specific growth rate for biomass (µB,max), the maximum spe-
t)
qN − e(− ΩN,max •
cific rates of nutrient uptake for glucose and free amino nitrogen (ΩG,
max, ΩN,max), the parameters for nutrient saturation (KG, KN), the phys-

⎪ G
⎨ ΩG, max • ∝G • I • , ifG > 0 iological state parameters (qG, qN), and the lactic acid production inhi-
µG = G + KG (5)

⎩ bition exponent (n), remained initially unknown. These parameters
0 otherwise
serve to distinguish the fermentations from one another, effectively
⎧ reflecting the influence of varying YE levels.
N

⎨ ΩN, max • ∝N •I • , if G > 0 To illustrate the impact of YE on fermentation dynamics, process
µN = N + KN (6) parameters were optimized using experimental data on process vari-


0 otherwise ables measured at different fermentation stages. Fig. 2 shows that the
optimization procedure starts with initial values for the process vari-
dB ables and the process parameters to simulate a fermentation using a
= µB • B− D • B (7)
dt mathematical model. As the initial process parameters were randomly
set, The simulated data showed significant differences from the experi-
dG 1
= − ΩG •B• − D•G (8) mental data and the dissimilarity in the fitting was taken into account
dt YBG
during the error calculation. The optimization procedure was executed
with the speed-constrained particle swarm optimization algorithm [52].
dN 1
= − ΩN •B• − D•N (9) It is an iterative approach, proposing possibly a new set of process pa-
dt YBN
rameters to repeat the simulation. The optimal set of parameter values
dL dG corresponds to the model-based simulated data that best aligns with the
= • YLG − D • L (10) experimental data, signifying the best representation of a fermentation.
dt dt
Throughout the optimization process, the fitness quality was determined
dA dG based on the sum of the mean absolute error percentage in the con-
= • YAG (11)
dt dt centration range (SMAEP) for biomass, glucose, lactic acid, and free
amino nitrogen, with an objective of minimizing the difference between
dV dA the simulated and experimental data. The inclusion of a percentage in
= (12)
dt dt the concentration range ensures a balanced assessment, particularly for
variables with lower concentrations, such as biomass with a maximum
dA 1
D= • (13) concentration of <7 g/L and glucose with >90 g/L. The mean absolute
dt V
error (MAE), mean absolute error percentage in the concentration range
The specific glucose uptake rate (ΩG) and the specific amino nitrogen (MAEP) and sum of mean absolute error percentage in the concentration
uptake rate (ΩN) were calculated by considering the inhibitory effect of range (SMAEP) were calculated using the equations 14 – 16.
lactic acid concentration (I) using Eq. 2, the initial lag phase (∝G and ∝N )
E ⃒ ⃒
using Eqs. 3 – 4, and nutrient source limitations, which include glucose ∑ ⃒ exp
⃒yj − ysim

j ⃒
and amino nitrogen using the formulas G+K G
G
(Eq. 5) and N+K N
N
(Eq. 6), MAE =
j=1
(14)
respectively. A logistic model, proposed by Verhulst [48] was employed E
to describe biomass formation (B) with an absence of dependency on

307
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

Fig. 2. : Optimization procedure to determine the optimal set of maximum specific growth rates for biomass (µB,max), the maximum rates of nutrient uptake for
glucose and free amino nitrogen (ΩG,max, ΩN,max), the parameters for nutrient saturation for glucose and free amino nitrogen (KG, KN), the physiological state
representations (qG, qN), and the lactic acid production inhibition exponent (n). It involves three stages: a) Utilizing experimental data, b) Employing a system of
equations describing the dynamics of the process to simulate biomass formation (B), glucose consumption (G), free amino nitrogen consumption (N), lactic acid
formation (L), and alkaline consumption (A), and c) Implementing an optimizer (speed-constrained particle swarm optimization algorithm) to find the optimal set of
process parameters by minimizing the sum of mean absolute error percentage in the concentration range (SMAEP) between the model-simulated data and experi-
mental data.

100 real-time without needing to know the Bmax and Lmax values. Eq. 17 il-
MAEP = MAE • (15)
yexp
max − yexp
min
lustrates the calculation of alkaline consumption by taking into account
the difference in alkaline consumption over a short time interval (At −
M
∑ At− 1). The glucose consumption and lactic acid formation rates were
SMAEP = MAEPi (16) characterized using Eq. 18 and Eq. 19, respectively. In this context, the
i=1
kinetic parameters rGA and rLA were employed to signify the rates of
where j is the index for experimental measurements, E is the total glucose consumption and lactic acid formation, respectively, derived
number of experimental measurements from a process run, yexp is the j- from the real-time alkaline consumption rate. The alterations in volume
j
and dilution rate resulting from the addition of alkaline were calculated
th experimental value for a variable, ysim is the corresponding model-
j using Eq. 20 and Eq. 21, respectively. For the establishment of the real-
based simulated value, and yexp exp
max and ymin are the maximum and min- time monitoring of glucose and lactic acid concentration, experimental
imum experimental values for a variable, and i is the index of process data from three fermentations were used for training the mathematical
variables and M is the total number of process variables. model, and one fermentation was excluded for validation purposes. This
process was repeated four times with different combinations of fer-
2.6. Real-time alkaline consumption-based process monitoring mentations in the training set to ensure that each fermentation was left
out once for validation.
Kinetic growth models prove valuable in describing the dynamics of
ΔA = At − At− (17)
fermentation, aiding in the design, monitoring, and control of processes. 1

Given the interdependence of fermentation process variables, the


dG
mathematical representations facilitate the dynamics of one variable in = − rGA • ΔA − D • G (18)
dt
relation to others. This approach offers the advantage of estimating
multiple process variables based on the measurement of a single vari- dL
able, provided that a representative model is employed. The kinetic = − rLA • ΔA − D • L (19)
dt
model-based characterization of fermentation presented in Section 2.5
can be used to simulate the lactic acid production process. However, it dV
= ΔA (20)
cannot be applied for real-time monitoring of key process variables dt
because of the lack of real-time measurements. Additionally, offline
1
measurements are time-consuming and the maximum biomass (Bmax) D= ΔA • (21)
and maximum lactic acid (Lmax) values are unknown beforehand. The V
proposed kinetic model was employed only after all offline data for a Fig. 3 illustrates the process of training the parameters of the
specific fermentation had been obtained. To overcome this shortcoming, mathematical model and using cross-validation to monitor glucose and
real-time records of alkaline consumption were utilized to understand lactic acid levels in real-time. With alkaline consumption data from the
the system’s dynamics. This approach operates under the assumption training fermentations, both glucose consumption and lactic acid for-
that the rate of alkaline addition corresponds to the rate of lactic acid mation were calculated based on mathematical models (Eqs. 17 – 21).
formation in the medium, and consequently, the rate of glucose con- To simulate the concentrations of glucose and lactic acid during
sumption. Therefore, glucose and lactic acid levels can be predicted in fermentation time, they are employed with their initial concentrations,

308
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

Fig. 3. : Illustration of training and cross-validation of online process monitoring with glucose and lactic acid concentration predictions based on real-time alkaline
addition used to keep pH constant during batch fermentation of B. coagulans. The experimental data were split into training and validation processes. Real-time
alkaline addition (A), initial glucose (G0), and lactic acid (L0) concentrations from training fermentation processes were used to find optimized kinetic parame-
ters (rGA, rLA) in the mathematical model that can represent the dynamics between alkaline addition, glucose consumption and lactic acid formation by minimizing
the sum of the mean absolute error percentage in the concentration range (SMAEP) of the prediction of glucose and lactic acid from all training processes. The
optimized kinetic parameters were used to predict the glucose and lactic acid concentrations of the left-out (validation process) fermentation using only real-time
alkaline addition (A), initial glucose (G0), and lactic acid (L0) concentrations. The procedure was repeated until each fermentation process was validated once. Here
MAE is the mean absolute error of the prediction.

which are known. By assigning random initial values to the kinetic glutathione (GLU), valine (VAL), and leucine (LEU) are relatively
parameters—rGA and rLA—the predicted glucose and lactic acid values similar, fluctuating between 20 and 30 nM. Aspartic acid (ASP), threo-
were compared with the corresponding experimental data. This facili- nine (THR), glycine (GLY), isoleucine (ILEU), phenylalanine (PHE),
tated the evaluation of the fitness of the proposed rGA and rLA values. By lysine (LYS), and proline (PRO) exhibited concentrations ranging from
applying the optimization algorithm, PSO, rGA and rLA were optimized to 10 nM to 20 nM. The lowest concentrations, below 10 nM, were
ensure that the predicted values of the mathematical model presented
the closest fit with minimal prediction error compared to the experi-
mental data for glucose and lactic acid in all training fermentations. The
optimized values of rGA and rLA were validated by predicting glucose and
lactic acid concentrations using alkaline addition, initial glucose (G0)
and lactic acid (L0) concentration data from a fermentation, which was
excluded during training.

3. Results and discussions

The batch fermentation processes of B. coagulans with yeast extract


(YE) of 5 % (BFYE5), 10 % (BFYE10), 15 % (BFYE15), 20 % (BFYE20)
were performed to produce lactic acid from glucose. To achieve higher
microbial growth and increase productivity, previous studies have used
15 g/L [36,37], and 20 g/L [53] of YE. In addition to experimental data
on free amino nitrogen throughout the fermentation runtime, an anal-
ysis of the amino acid profile of the yeast extract used was conducted. In Fig. 4. : Composition of yeast extract utilized in the study. The yeast extract
the yeast extract utilized in this study, 16 amino acids were identified. solution consisted of 15 g/L of yeast extract dissolved in MilliQ-water. The
The yeast extract composition in terms of AA concentration is shown in analyzed amino acids include aspartic acid (ASP), threonine (THR), serine
Fig. 4. Alanine (ALA) stands out as the most abundant amino acid in the (SER), glutamic acid (GLU), glycine (GLY), alanine (ALA), valine (VAL),
yeast extract, surpassing 60 nM. The concentrations of serine (SER), methionine (MET), isoleucine (ILEU), leucine (LEU), tyrosine (TYR), phenyl-
alanine (PHE), histidine (HIS), lysine (LYS), arginine (ARG), and proline (PRO).

309
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

observed for methionine (MET), tyrosine (TYR), histidine (HIS), and was used. Although ALA is the most abundant amino acid in YE, it was
arginine (ARG). only consumed to the extent of 25 % during fermentation. The con-
During the initial 3 hours of fermentation, amino acid (AA) analysis centration of arginine (ARG) was relatively low in the YE and was fully
was conducted. The concentrations of various AAs were measured at consumed in all batch fermentations. The quantities of ASP, THR, and
both time 0 and after 3 hours of each batch fermentation, corresponding SER in the YE sample were significantly higher compared to ARG and
to the observed shift in free amino nitrogen (N) analysis. It allowed an were completely utilized during the fermentation process.
observation of which AAs are essential for the microorganism at the It has been established that the B. coagulans A166 strain exhibits
onset of fermentation. The percentage consumption of each AA is specific requirements for amino acids and vitamins, as reported by
illustrated in Fig. 5. Within the first three hours of fermentation, almost Campbell et al. [57] and Marshall et al. [58]. Alanine (ALA), as high-
100 % consumption was observed for aspartic acid (ASP), threonine lighted in a study [59], plays a crucial role in the growth and viability of
(THR), serine (SER), and arginine (ARG) in all fermentation processes. Bacillus subtilis. Various techniques, such as continuous cell recycling
However, methionine (MET) was fully depleted in BFYE5, while in all fermentation [60] and immobilized cell technology [61], have been
other cases, 50 % was consumed. There is a noticeable trend where the explored to optimize the usage of YE concentration. This study [24]
addition of more yeast extract led to increased amino acid consumption, compared batch and continuous fermentation methods to evaluate the
particularly evident in the case of alanine (ALA), tyrosine (TYR), efficient use of YE of B. coagulans. Additionally, alternative and more
phenylalanine (PHE), histidine (HIS), and glutamic acid (GLU). For cost-effective sources of YE were explored, [62,63], including corn steep
glycine (GLY), the fermentation with 5 g/L of YE, almost 75 % was powder in repeated-batch fermentation [64]. By applying different
consumed, whereas for rest of the batch fermentations, less than 25 % feeding strategies, the yield of LA in relation to the used YE increased, by
was consumed. Valine (VAL) consumption remained relatively stable, successfully reducing the YE by 36 %. Considering more complex media,
fluctuating around 25 % consumption after 3 hours. Lysine (LYS) and such as, municipal solid waste, the techno-economic analysis are
proline (PRO) consumption displayed specificity, indicating more PRO necessary to show, that pilot scale production of lactic acid utilizing
detected in the sample after 3 hours of fermentation. In the batch yeast extract as nitrogen source is competitive with technologies already
fermentation with 20 g/L of YE, almost 25 % of PRO was utilized. existing [65].
Studies describing the dynamics of free amino acid levels during Table 2 provides an overview of fermentation conditions and pro-
fermentation using Lactobacillus brevis species have shown a similar in- ductivity coefficients. Fermentations BFYE5 and BFYE10 were stopped
crease in PRO content during the initial 4 hours of fermentation [54]. after 50 hours (see Ct in Table 2 c), with 3.1 g / L and 4.5 g / L glucose
Other research has indicated that the number of prolines increases with remaining, respectively (see G in Table 2b). In contrast, BFYE15 and
increasing thermostabilities of oligo-1,6-glucosidase of B. coagulans, BFYE20 consumed all glucose from the medium in 28 and 24 hours,
suggesting that the substitution of proline is a crucial factor for the se- respectively, showing higher productivity (see PG in Table 2c). No
lection of thermostabilities in oligo-1,6-glucosidases [55]. Additional fermentation fully utilized the supplementary nitrogen source (N), with
studies, such as the one conducted by Tsuchiya et al. [56], have reported BFYE15 and BFYE20 retaining a huge surplus of nitrogen. Lactic acid
similar results, emphasizing that various strategies may be preferred for production was higher in fermentations with increased concentrations
thermostabilization. of yeast extract (BFYE15 and BFYE20), as outlined in the final lactic acid
The consumption of other amino acids exhibited fluctuations. In concentration (L) in Table 2b and the total lactic
particular, the four primary amino acids, namely ASP, THR, SER, and acid (Ltotal) in Table 2c. While there is a significant difference in
ARG, displayed consistent patterns. Within the initial 3 hours of all four fermentation completion time (see Ct in Table 2c) among the processes,
batch processes, 100 % consumption was observed for arginine (ARG), it is interesting that the total alkaline consumption (A) exhibits only
which was comparable to ASP, THR, and SER. Their consumption was marginal variations, ranging from 0.5 to 0.55 L across the different
notably the highest during the initial phase of the experiment. Methio- fermentations. This is exemplified by the coefficient YAG, indicating the
nine (MET) was consumed entirely during the batch experiment with amount of alkaline needed for each unit of glucose consumption, varying
5 g/L YE, while for the remaining fermentations, approximately 50 % from 0.0064 to 0.0066 L. Although BFYE10 had the lowest initial

Fig. 5. : Percentage consumption of amino acids during first 3 hours of batch fermentations using B. coagulans A166 strain to produce lactic acid from glucose with
varying yeast extract (YE) concentrations: 5 % (BFYE5), 10 % (BFYE10), 15 % (BFYE15), 20 % (BFYE20). The analyzed amino acids include aspartic acid (ASP),
threonine (THR), serine (SER), glutamic acid (GLU), glycine (GLY), alanine (ALA), valine (VAL), methionine (MET), isoleucine (ILEU), leucine (LEU), tyrosine (TYR),
phenylalanine (PHE), histidine (HIS), lysine (LYS), arginine (ARG), and proline (PRO).

310
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

Table 2 Furthermore, there was a significant increase in the yield from glucose to
Initial and final conditions, along with productivity coefficients for the batch lactic acid (YLG) in fermentations with 10 % yeast extract (BFYE10),
fermentation of B. coagulans employed in lactic acid production from glucose while an increasing concentration of yeast extract was observed to
with varying yeast extract (YE) concentrations: 5 % (BFYE5), 10% (BFYE10), diminish the yield in BFYE15 and BFYE20.
15 % (BFYE15), and 20 % (BFYE20). Sections a) and b) illustrate the initial and In a process to formulate the biomass formation, a comparison was
final conditions of the fermentations, section c) shows the calculated process
made by fitting multiple unstructured kinetic bacterial growth models
parameters using only experimental data.
with the experimental data. Kinetic growth models such as Monod [66],
BFYE5 BFYE10 BFYE15 BFYE20 Blackman [67], Haldane [68], Tesseir [69], Moser [70], Contois [71],
a) Initial conditions Logarithmic [72], Aiba-Edwards [73], Han and Levenspiel [74], Yano
B [g/L] 0.155 0.05 0.44 0.45 and Koga [75], Powell [76], Luong [77], and Webb [78] showed poor
G [g/L] 100.8 93.0 98.4 101.3
fitting with biomass formation (results not shown). However, the Ver-
N [mg/L] 180 301 671 857
L [g/L] 0.0 0.91 1.18 1.13 hulst model [50], which considers only biomass concentration to
A [L] 0.0 0.0 0.0 0.0 describe kinetic growth [79] and is known as a substrate-independent
b) Final conditions model, fit satisfactorily with the obtained experimental cell dry weight
B [g/L] 3.99 4.18 4.63 5.54 data for biomass formation.
G [g/L] 3.12 4.57 0.0 0.0
The experimental investigation involved varying the concentration
N [mg/L] 48 133 370 498
L [g/L] 74.68 73.95 82.31 83.73 of yeast extract from 5 % to 20 % in fermentations, directly impacting
A [L] 0.54 0.50 0.54 0.55 B. coagulans’ growth rate, glucose uptake rate, and lactic acid formation.
c) Productivity parameters obtained from experimental data In addition to experimental data, insights into the kinetic aspects of the
Ltotal [g] 251.4 246.9 279.4 287.4
fermentation process were extracted using a mathematical model with
Ct [h] 50 50 28 24
PG [g/L/h] 1.63 1.48 3.07 5.04
an aim to understand how variations in nutrient supplementation in-
YAG [L/g] 0.0065 0.0066 0.0065 0.0064 fluence microbial growth and substrate consumption dynamics. The
YBG [g/g] 0.033 0.055 0.050 0.060 comparison between kinetic model-based simulated and experimental
YBN [g/mg] 0.025 0.033 0.021 0.022 data for biomass formation, glucose consumption, and lactic acid pro-
YLG [g/g] 0.92 0.99 0.93 0.95
duction can be observed in Fig. 6. The experimental data indicate that
glucose consumption and lactic acid production persist even as the dry
biomass concentration, it exhibited a significant growth rate of biomass weight of the biomass reaches the stationary phase. In the early stages of
dry weight until the final stage, even with a slightly lower initial glucose growth, a brief lag phase is evident, reflected in both glucose con-
concentration (93 g/L) compared to other fermentations (>98 g/L). The sumption and lactic acid formation as well. Afterward, biomass un-
yield coefficient from glucose to biomass (YBG) indicates a slightly lower dergoes exponential growth until the 6th hour of the fermentation
value for BFYE10 (0.055 g/g) compared to BFYE20 (0.060 g/g). The process, followed by no significant change reflects the stationary phase.
conversion coefficient from free amino nitrogen to biomass (YBN) shows During batch fermentation, the maximum specific growth rate (µB,
− 1
that BFYE10 had the maximum conversion rate (0.033 g/mg). max) for B. coagulans achieved 0.74 h with 5 % yeast extract in the

Fig. 6. : Process dynamics of lactic acid production from glucose using batch fermentation (BF) of B. coagulans A166 strain with mediums containing different yeast
extract (YE) concentrations: 5 % (BFYE5), 10 % (BFYE10), 15 % (BFYE15), 20 % (BFYE20). Experimental data (e.g., Glucoseexpr) are displayed with markers and
modelbased simulated data (e.g., Glucosesim) with lines. recorded the highest at 0.19 h− 1. With higher yeast extract concentrations, BFYE15 and BFYE20 demon-
strated rates of 0.12 h− 1 and 0.16 h− 1, indicating a reduction in the maximum specific glucose consumption rate.

311
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

medium (FBYE5). The highest specific growth rate occurred in the


fermentation BFYE10 with 10 % yeast extract (1.2 h− 1), and subsequent
increases in yeast extract concentration to 15 % (0.73 h− 1) and 20 %
(0.69 h− 1) did not result in further enhancement (Table 3). A compa-
rable pattern was observed for the maximum specific glucose uptake
rate (ΩG,max) with BFYE5 showing the lowest value (0.09 h− 1), while
FBYE10 recorded the highest at 0.19 h− 1. With higher yeast extract
concentrations, BFYE15 and BFYE20 demonstrated rates of 0.12 h− 1 and
0.16 h− 1, indicating a reduction in the maximum specific glucose con-
sumption rate.
Fig. 7 presents the comparison between the observed and model-
based simulated data for free amino nitrogen (N) in batch fermentations.
Data on free amino nitrogen exhibit alignment with biomass formation,
demonstrating exponential consumption until 6th hour, followed by
slower consumption until the completion of the processes. The opti- Fig. 7. : Experimental and model-based simulated free amino nitrogen (N)
mized process parameters indicate an increasing trend in the maximum concentration for batch fermentation of B. coagulans A166 strain with mediums
specific uptake rate of free amino nitrogen (ΩN,max) as the concentration containing different yeast extract (YE) concentrations: 5 % (BFYE5), 10 %
of yeast extract in fermentation increases. In particular, BFYE20, char- (BFYE10), 15 % (BFYE15), 20 % (BFYE20) to produce lactic acid from glucose.
acterized by the highest concentration of yeast extract, demonstrated a Experimental data (e.g., BFYE5expr) are displayed with markers and model-
maximum specific uptake rate of free amino nitrogen (ΩN,max) at based simulated data with lines (e.g., BFYE5sim).
0.41 h− 1. This represents a significant augmentation compared to the
rates observed in other fermentations, which ranged from 0.10 to inhibition (ranging from 1.31 to 2.05) was reported [36], along with a
0.14 h− 1. This implies that as the concentration of yeast extract in- slightly increased lactic acid concentration (up to 94 g/L) in comparison
creases, the fermentation completion time may be shortened, potentially to the present investigation (<84 g/L).
improving productivity. However, this leads to an excess of sources of The alkaline consumption-based real-time monitoring model was
free amino nitrogen remaining in the medium after completion and its trained to predict glucose and lactic acid concentrations employing
consumption does not necessarily align with a similar trend in lactic acid system dynamics. To validate this approach, one fermentation was left
yield from glucose (YLG). out each time from the training set. By using only real-time alkaline
Table 3 shows the kinetic model parameters for the batch fermen- consumption rate, initial glucose, and lactic acid as input variables, the
tations obtained from the experimental data. The parameter for glucose trained model was used to predict glucose and lactic acid concentration.
saturation (KG) was highest (24 g/L) for BFYE10 while for free amino Fig. 8 illustrates the results of cross-validation in fermentations marked
nitrogen, it shows an increasing pattern with increasing YE in the me- by varying concentrations of yeast extract. The validation outcomes
dium. The physiological state parameter of the inoculum for glucose (qG) revealed alignment between the predicted values of glucose and lactic
exhibited a range of values, from 3.09 in FBYE10 to 7.98 in FBYE5. This acid formation and their actual experimental data.
agrees with the findings obtained by Glaser and Venus [36] in a study The training and prediction results for glucose consumption and
with 15 % yeast extract and hydrolysates of lignocellulosic feedstock lactic acid formation are outlined in Table 4. The kinetic parameter rGA,
that reported a comparable range of values (0.8–13.1) across five which signifies the rate of glucose consumption associated with alkaline
distinct strains of B. coagulans. Nevertheless, the physiological state consumption, shows slight variations, ranging from 175.4 to 178.5 [g/
parameters of the inoculum for free amino nitrogen, denoted as qN, L2] in various combinations of training sets. Similarly, the kinetic
exhibited a notable increase in FBYE20 (42.46) compared to the values parameter rLA, representing the rates of lactic acid formation in relation
obtained from other fermentations (<1.2). The parameter representing to alkaline consumption, demonstrates consistency in a range of
the extent of lactic acid production inhibition, denoted as n, signifies the 162.2–164.0 [g/L2], regardless of the different combinations of training
inhibitory effect due to the elevated concentration of lactic acid (L) in sets. This minimal variability in the model parameters indicates a
the medium, resulting in a decrease in the rate of substrate consumption persistent correlation between alkaline consumption and the dynamics
and lactic acid production, as evident from the experimental data ac- of glucose and lactic acid in the system. The highest accuracy in glucose
quired from the later stage of fermentations. The n values highlighted a prediction was noted with a mean absolute error (MAE) of 2.73 g/L for
minimal degree of inhibition in all fermentations, with no substantial fermentation BFYE5, whereas BFYE20 showed the least prediction ac-
variance observed (ranging from 1.01 to 1.17). In a study involving curacy with MAE of 5.12 g/L. In terms of lactic acid prediction, MAE was
lignocellulosic feedstock hydrolysates, a slightly higher degree of 2.98 g/L for BFYE5, 2.48 g/L for BFYE10, 3.67 g/L for BFYE15, and
4.48 g/L for BFYE20.
The research focus on model-based calibration in fermentation
Table 3 monitoring has been an attractive area for the past few decades, driven
Optimized process parameters derived from the kinetic models (Eqs. 1 – 13) by the challenges of acquiring extensive experimental data for training
using experimental data from the batch fermentation of B. coagulans employed machine learning models with sensor signals [35,44,80,81]. This
in lactic acid production from glucose with varying yeast extract (YE) concen-
approach eliminates the necessity of experimental data, which are
trations: 5 % (BFYE5), 10 % (BFYE10), 15 % (BFYE15), and 20 % (BFYE20). The
time-consuming, laborious, and prone to errors. Leveraging mod-
presented optimized process parameters represent the dymanics for each
fermentation described by kinetic model.
el-simulated data as target values in lieu of experimental data enables
the use of sensor signals from very short time intervals to train models
BFYE5 BFYE10 BFYE15 BFYE20
that learn system dynamics precisely. Alemneh et al. [35] employed a
mathematical model-based calibration approach with 2D fluorescence
− 1
µB,max [h ] 0.74 1.20 0.73 0.69
[h− 1] 0.09 0.19 0.12 0.16
ΩG,max
spectra and predicted the concentration of lactic acid with an error of
ΩN,max [h− 1] 0.10 0.11 0.14 0.41
KG [g/L] 5.9 24.0 3.8 1.1
8.0 % (root mean squared error percentage of the concentration range
KN [mg/L] 74.7 264.1 408.9 742.7 (RMSEP)). Similarly, Yousefi-Darani et al. [44] achieved <7 % RMSEP
qG [1] 7.98 3.09 6.57 3.96 prediction error for ethanol in batch cultivation of S. cerevisiae by
qN [1] 1.17 0.89 1.10 42.46 applying model-based calibration of gas sensor signals. Babor et al. [80]
n [1] 1.10 1.17 1.01 1.07
proposed an approach to train a chemometric model from 2D

312
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

Fig. 8. : One fermentation left out cross-validation for the prediction of glucose consumption and lactic acid production based on real-time alkaline consumption
using batch fermentation (BF) dynamics (Eqs. 17 – 21) of B. coagulans A166 strain with mediums containing different yeast extract (YE) concentrations: 5 % (BFYE5),
10 % (BFYE10), 15 % (BFYE15), 20 % (BFYE20). Experimental data (e.g., Glucoseexpr) are displayed with markers and model-based predicted data with lines (e.g.,
Glucosepred). The experimental data from three fermentations were used to find the best parameters (rGA, rLA), which were used to predict the glucose consumption
and lactic acid formation based on real-time alkaline consumption of the left out fermentation.

4. Conclusion
Table 4
Cross validation of the glucose and lactic acid prediction using alkaline con-
This study focused on the lactic acid production from glucose, where
sumption in mathematical model (Eqs. 17 – 21) for batch fermentation of
B. coagulans with varying yeast extract (YE) concentrations: 5 % (BFYE5), 10 % yeast extract serves as a vital yet cost-intensive nutrient for optimal
(BFYE10), 15 % (BFYE15), and 20 % (BFYE20). The prediction accuracy is growth of versatile Bacillus coagulans A166 strain in batch fermentation
shown using the mean absolute error (MAE) and the mean absolute error per- processes. With a mathematical model, the key kinetic parameters
centage in range of the concentration (MAEP). correspond to the dynamics of microorganisms in response to fluctuating
BFYE5 BFYE10 BFYE15 BFYE20
yeast extract were studied to provide valuable insights for optimizing
production processes. Moreover, a mathematical model-based real-time
a) Mathematical model parameters
monitoring of glucose consumption and lactic acid formation from
rGA [g/L2] 175.9 175.4 178.5 176.1
rLA [g/L2] 163.6 162.2 164.0 163.4 alkaline consumption rate was studied.
b) Glucose prediction The findings implied that an increase in yeast extract led to reduced
MAE [g/L] 2.73 4.06 4.01 5.12 fermentation completion time, enhancing productivity, however,
MAEP [%] 2.8 4.6 4.1 5.0
resulting in a significant surplus of free amino nitrogen. Remarkably,
c) Lactic acid prediction
MAE [g/L] 2.98 2.48 3.67 4.48 with 10 % yeast extract in the medium, the fermentation showed the
MAEP [%] 4.0 3.4 4.5 5.4 highest maximum specific growth rate for B. coagulans, maximum spe-
cific glucose consumption rate, and the optimal yield of lactic acid from
glucose. The mathematical model-based predictions highlighted a
fluorescence spectra with model-based data to predict biomass, while strong correlation between real-time alkaline consumption and process
glycerol predictions were derived by using predicted biomass as an input dynamics, with mean absolute errors (expressed as a percentage of the
in a mathematical model describing the dynamics of biomass and concentration range) for glucose and lactic acid predictions being less
glucose in the fermentation. This approach outperformed predictions than or equal to 5.05 % and 5.43 %, respectively. Further experiments
based on model-based calibration alone, achieving an RMSEP of 5.2 % can be explored using a range of yeast extract concentrations in the
for glycerol compared to 8.5 %. It implies that predicting substrates fermentation medium, specifically between 5 % and 15 %, incorpo-
using different sensors can be challenging since, in many instances, there rating replicates with an aim to identify optimal conditions for lactic
are no reasonable direct responses in the signals observed. Nevertheless, acid production from glucose, considering factors such as productivity,
in this particular research, solely a pH sensor was employed to maintain yield of lactic acid from glucose, and the excess amount of yeast extract.
optimal pH in the medium by adding alkaline, highlighting it as a
dependable metric for real-time monitoring of glucose and lactic acid Funding
concentration.
This work was funded from the European Union’s Horizon 2020

313
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

Research and innovation Program under Grant Agreement No. 95294 [14] M.I. Torino, G.F. de Valdez, F. Mozzi, Biopolymers from lactic acid bacteria. Novel
applications in foods and beverages, Front. Microbiol. 6 (2015) 1–16, https://doi.
(BIOMAC project).
org/10.3389/fmicb.2015.00834.
[15] L. Pei, M. Schmidt, W. Wei, Conversion of biomass into bioplastics and their
CRediT authorship contribution statement potential environmental impacts, Biotechnol. Biopolym. 364 (2011), https://doi.
org/10.5772/683.
[16] E. Castro-Aguirre, F. Iñiguez-Franco, H. Samsudin, X. Fang, R. Auras, Poly(lactic
Marina M.-C. Höhne: Methodology, Formal analysis. Majhar- acid)—Mass production, processing, industrial applications, and end of life, 107 (n.
ulislam Babor: Data curation, Conceptualization. Agata Olszewska- d.) 333–366. https://doi.org/10.1016/j.addr.2016.03.010.
[17] F.A. Castillo Martinez, E.M. Balciunas, J.M. Salgado, J.M. Domínguez González,
Widdrat: Writing – original draft, Supervision, Formal analysis, Data A. Converti, R.P. de S. Oliveira, Lactic acid properties, applications and production:
curation, Conceptualization. Joachim Venus: Supervision, Project a review, Trends Food Sci. Technol. 30 (2013) 70–83, https://doi.org/10.1016/j.
administration, Methodology, Funding acquisition. Jose Pablo López tifs.2012.11.007.
[18] Y. Wang, W. Deng, B. Wang, Q. Zhang, X. Wan, Z. Tang, Y. Wang, C. Zhu, Z. Cao, G.
Gómez: Investigation, Formal analysis, Conceptualization. Maria
Wang, H. Wan, Chemical synthesis of lactic acid from cellulose catalysed by lead
Alexandri: Data curation, Conceptualization. (II) ions in water, 4 (n.d.) 2141. https://doi.org/10.1038/ncomms3141.
[19] D. Yankov, Fermentative Lactic Acid Production From Lignocellulosic Feedstocks:
From Source to Purified Product, Front. Chem. 10 (2022), https://doi.org/
Declaration of Competing Interest 10.3389/fchem.2022.823005.
[20] R.P. Pawar, S.U. Tekale, S.U. Shisodia, J.T. Totre, A.J. Domb, Biomedical
Applications of Poly(Lactic Acid), 4 (n.d.) 40–51. 〈https://www.eurekaselect.
The authors declare that they have no known competing financial com/article/59945〉.
interests or personal relationships that could have appeared to influence [21] S. Inkinen, M. Hakkarainen, A.C. Albertsson, A. Södergård, From lactic acid to poly
(lactic acid) (PLA): characterization and analysis of PLA and Its precursors,
the work reported in this paper.
Biomacromolecules 12 (2011) 523–532, https://doi.org/10.1021/bm101302t.
[22] Global Industry Report 2030, Research Grand View. Global lactic acid market, (n.
Data availability d.). 〈https://www.grandviewresearch.com/industry-analysis/lactic-acid-and-pol
y-lactic-acid-market〉.
[23] V. Juturu, J.C. Wu, Production of high concentration of l-lactic acid from oil palm
Data will be made available on request. empty fruit bunch by thermophilic Bacillus coagulans JI12, Biotechnol. Appl.
Biochem. 65 (2018) 145–149, https://doi.org/10.1002/bab.1567.
[24] T. Payot, Z. Chemaly, M. Fick, Lactic acid production by Bacillus
Acknowledgement
coagulans—kinetic studies and optimization of culture medium for batch and
continuous fermentations, Enzym. Microb. Technol. 24 (1999) 191–199, https://
We would like to thank Maria Zoidi and Christopher Schieweck for doi.org/10.1016/S0141-0229(98)00098-2.
[25] L. Jaramillo, D. Santos, D. Guedes, D. Dias, E. Borges, N. Pereira, Production of
their valuable contribution in this study.
lactic acid enantiomers by lactobacillus strains under limited dissolved oxygen
conditions in the presence of a pentose fraction, Appl. Biochem. Microbiol. 57
References (2021) 452–457, https://doi.org/10.1134/S0003683821040050.
[26] X. Tian, H. Chen, H. Liu, J. Chen, Recent Advances in Lactic Acid Production by
Lactic Acid Bacteria, 193 (n.d.) 4151–4171. https://doi.org/10.1007/s12010-02
[1] M. Alexandri, R. Schneider, K. Mehlmann, J. Venus, Recent advances in d-lactic
1-03672-z.
acid production from renewable resources: case studies on agro-industrial waste
[27] F. Zhang, J. Liu, X. Han, C. Gao, C. Ma, F. Tao, P. Xu, Kinetic characteristics of long-
streams, Food Technol. Biotechnol. 57 (2019) 293–304, https://doi.org/10.17113/
term repeated fed-batch (LtRFb) l-lactic acid fermentation by a Bacillus coagulans
ftb.57.03.19.6023.
strain, 20 (n.d.) 562–570. https://doi.org/10.1002/elsc.202000043.
[2] R. Alves de Oliveira, A. Komesu, C.E. Vaz Rossell, R. Maciel Filho, Challenges and
[28] N. Adibpour, M. Hosseininezhad, A. Pahlevanlo, M.A. Hussain, A review on
opportunities in lactic acid bioprocess design—from economic to production
Bacillus coagulans as a Spore-Forming Probiotic, 6 (n.d.) 91–100. https://doi.
aspects, Biochem. Eng. J. 133 (2018) 219–239, https://doi.org/10.1016/j.
org/10.22037/afb.v6i2.23958.
bej.2018.03.003.
[29] M. Alexandri, A.-K. Neu, R. Schneider, J.P. López-Gómez, J. Venus, Evaluation of
[3] S.A.A. Rawoof, P.S. Kumar, D.V.N. Vo, K. Devaraj, Y. Mani, T. Devaraj,
various Bacillus coagulans isolates for the production of high purity L-lactic acid
S. Subramanian, Production of optically pure lactic acid by microbial fermentation:
using defatted rice bran hydrolysates, 54 (n.d.) 1321–1329. https://doi.org
a review, Environ. Chem. Lett. 19 (2021) 539–556, https://doi.org/10.1007/
/10.1111/ijfs.14086.
s10311-020-01083-w.
[30] M. Aulitto, S. Fusco, S. Bartolucci, C.J. Franzén, P. Contursi, Bacillus coagulans
[4] N. Nancib, A. Nancib, A. Boudjelal, C. Benslimane, F. Blanchard, J. Boudrant, The
MA-13: a promising thermophilic and cellulolytic strain for the production of lactic
effect of supplementation by different nitrogen sources on the production of lactic
acid from lignocellulosic hydrolysate, 10 (n.d.) 210. https://doi.org/10.1186/s130
acid from date juice by Lactobacillus casei subsp. rhamnosus, Bioresour. Technol.
68-017-0896-8.
78 (2001) 149–153, https://doi.org/10.1016/S0960-8524(01)00009-8.
[31] E. De Clerck, M. Rodriguez-Diaz, G. Forsyth, L. Lebbe, N.A. Logan, P. DeVos,
[5] K.Y. Ooi, J.C. Wu, Use of dry yeast cells as a cheap nitrogen source for lactic acid
Polyphasic characterization of Bacillus coagulans strains, illustrating heterogeneity
production by thermophilic Bacillus coagulans WCP10-4, Front. Chem. Sci. Eng. 9
within this species, and emended description of the species, 27 (n.d.) 50–60.
(2015) 381–385, https://doi.org/10.1007/s11705-015-1534-2.
https://doi.org/10.1078/0723-2020-00250.
[6] I. de la Torre, M. Ladero, V.E. Santos, Production of d-lactic acid by Lactobacillus
[32] J.R. Endres, A. Clewell, K.A. Jade, T. Farber, J. Hauswirth, A.G. Schauss, Safety
delbrueckii ssp. delbrueckii from orange peel waste: techno-economical assessment
assessment of a proprietary preparation of a novel Probiotic, Bacillus coagulans, as
of nitrogen sources, Appl. Microbiol. Biotechnol. 102 (2018) 10511–10521,
a food ingredient, 47 (n.d.) 1231–1238. https://doi.org/10.1016/j.fct.2009.02.01
https://doi.org/10.1007/s00253-018-9432-4.
8.
[7] M. Hujanen, Y.Y. Linko, Effect of temperature and various nitrogen sources on L (
[33] S.S. Kanwar, R.K. Kaushal, H. Sultana, S.S. Chimni, Purification of a moderate
+)-lactic acid production by Lactobacillus casei, Appl. Microbiol. Biotechnol. 45
thermotolerant Bacillus coagulans BTS1 lipase and its properties in a hydro-gel
(1996) 307–313, https://doi.org/10.1007/s002530050688.
system, Acta Microbiol. Et. Immunol. Hung. (2006), https://doi.org/10.1556/
[8] T. Gao, K.-P. Ho, L-lactic acid production by Bacillus subtilis MUR1 in continuous
AMicr.53.2006.1.5.
culture, 168 (n.d.) 646–651. https://doi.org/10.1016/j.jbiotec.2013.09.023.
[34] C. Le Marrec, B. Hyronimus, P. Bressollier, B. Verneuil, M.C. Urdaci, Biochemical
[9] D. Pleissner, F. Demichelis, S. Mariano, S. Fiore, I.M. Navarro Gutiérrez, R.
and genetic characterization of coagulin, a new antilisterial bacteriocin in the
Schneider, J. Venus, Direct production of lactic acid based on simultaneous
pediocin family of bacteriocins, produced by Bacillus coagulans I4, Appl. Environ.
saccharification and fermentation of mixed restaurant food waste, 143 (n.d.)
Microbiol. (2000), https://doi.org/10.1128/AEM.66.12.5213-5220.2000.
615–623. https://doi.org/10.1016/j.jclepro.2016.12.065.
[35] S.T. Alemneh, M. Babor, V. Zettel, A. von Wrochem, B. Hitzmann, Online
[10] D. Dietz, R. Schneider, F. Papendiek, J. Venus, Leguminose green juice as an
monitoring of the growth of probiotic bacteria and metabolites in the fermentation
efficient nutrient for L(+)-lactic acid production, J. Biotechnol. 236 (2016) 26–34,
of a teff substrate using model-based calibration of 2D fluorescence spectra,
https://doi.org/10.1016/j.jbiotec.2016.07.008.
Microorganisms 11 (2023), https://doi.org/10.3390/microorganisms11041032.
[11] D. Zhao, T. Zhu, J. Li, L. Cui, Z. Zhang, X. Zhuang, J. Ding, Poly(lactic-co-glycolic
[36] R. Glaser, J. Venus, Co-fermentation of the main sugar types from a beechwood
acid)-based composite bone-substitute materials, 6 (n.d.) 346–360. https://doi.
organosolv hydrolysate by several strains of Bacillus coagulans results in effective
org/10.1016/j.bioactmat.2020.08.016.
lactic acid production, Biotechnol. Rep. 18 (2018) 22–27, https://doi.org/
[12] A. Vlachopoulos, G. Karlioti, E. Balla, V. Daniilidis, T. Kalamas, M. Stefanidou, N.
10.1016/j.btre.2018.e00245.
D. Bikiaris, E. Christodoulou, I. Koumentakou, E. Karavas, D.N. Bikiaris, Poly
[37] R. Glaser, J. Venus, Model-based characterisation of growth performance and L-
(Lactic Acid)-based microparticles for drug delivery applications: an overview of
lactic acid production with high optical purity by thermophilic Bacillus coagulans
recent advances, Pharmaceutics 14 (2022), https://doi.org/10.3390/
in a lignin-supplemented mixed substrate medium, N. Biotechnol. 37 (2017)
pharmaceutics14020359.
180–193, https://doi.org/10.1016/j.nbt.2016.12.006.
[13] P. Cichońska, M. Ziarno, Legumes and legume-based beverages fermented with
lactic acid bacteria as a potential carrier of probiotics and prebiotics,
Microorganisms 10 (2022), https://doi.org/10.3390/microorganisms10010091.

314
A. Olszewska-Widdrat et al. Process Biochemistry 146 (2024) 304–315

[38] Y.L. Gordeeva, A.G. Borodkin, E.L. Gordeeva, E.G. Rudakovskaya, Mathematical [58] R. Marshall, R.J. Beers, Growth of Bacillus coagulans in chemically defined media,
Modeling of Continuous Fermentation in Lactic Acid Production, 53 (n.d.) J. Bacteriol. 94 (1967) 517–521.
501–508. https://doi.org/10.1134/S0040579519040183. [59] K.R. Sidiq, M.W. Chow, Z. Zhao, R.A. Daniel, Alanine metabolism in Bacillus
[39] S.P. Patel, A.M. Gujarathi, P.B. Vanzara, Multi-criteria analysis of cell-recycle subtilis, Mol. Microbiol. 115 (2021) 739–757, https://doi.org/10.1111/
based continuous lactic acid production process, 38 (n.d.) 1932–1941. https://doi. mmi.14640.
org/10.1080/10426914.2023.2190387. [60] R.K. Lee, H.W. Ryu, H. Oh, M. Kim, Y.J. Wee, Cell-recycle continuous fermentation
[40] R. Spann, A.E. Lantz, C. Roca, K.V. Gernaey, G. Sin, Model-based process of Enterococcus faecalis RKY1 for economical production of lactic acid by
development for a continuous lactic acid bacteria fermentation, in: Computer reduction of yeast extract supplementation, J. Microbiol. Biotechnol. 24 (2014)
Aided Chemical Engineering, Elsevier, n.d.: pp. 1601–1606. https://doi.org/10.10 661–666, https://doi.org/10.4014/jmb.1402.02017.
16/B978-0-444-64235-6.50279-5. [61] S. Norton, C. Lacroix, J.C. Vuillemard, Reduction of yeast extract supplementation
[41] Y. Wang, Dynamic simulation of continuous mixed sugar fermentation with in lactic acid fermentation of whey permeate by immobllized cell technology,
increasing cell retention time for lactic acid production using Enterococcus mundtii J. Dairy Sci. 77 (1994) 2494–2508, https://doi.org/10.3168/jds.S0022-0302(94)
QU 25, (n.d.). 77191-5.
[42] L. Van der Hauwaert, A. Regueira, L. Selder, A.P. Zeng, M. Mauricio-Iglesias, [62] L. Schroedter, R. Schneider, L. Remus, J. Venus, L-(+)-lactic acid from reed:
Optimising bioreactor processes with in-situ product removal using mathematical comparing various resources for the nutrient provision of B. coagulans, Resources 9
programming: a case study for propionate production, Comput. Chem. Eng. 168 (2020) 89, https://doi.org/10.3390/resources9070089.
(2022), https://doi.org/10.1016/j.compchemeng.2022.108059. [63] K. Ma, T. Maeda, H. You, Y. Shirai, Open fermentative production of l-lactic acid
[43] X.H. Xu, Z.X. Liu, X.Y. Shi, C. Miao, S. Sheng, Y. Xu, F.A. Wu, J. Wang, Fed-batch with high optical purity by thermophilic Bacillus coagulans using excess sludge as
fermentation of yarrowia lipolytica using defatted silkworm pupae hydrolysate: a nutrient, 151 (n.d.) 28–35. https://doi.org/10.1016/j.biortech.2013.10.022.
dynamic model-based approach for high yield of lipid production, Waste Biomass- [64] Y. Wang, J. Chang, D. Cai, Z. Wang, P. Qin, T. Tan, Repeated-batch fermentation of
.-. Valoriz. 9 (2018) 2399–2411, https://doi.org/10.1007/s12649-017-0180-y. L-lactic acid from acid hydrolysate of sweet sorghum juice using mixed
[44] A. Yousefi-Darani, M. Babor, O. Paquet-Durand, B. Hitzmann, Model-based neutralizing agent under unsterilized conditions, 92 (n.d.) 1848–1854. https://doi.
calibration of a gas sensor array for on-line monitoring of ethanol concentration in org/10.1002/jctb.5195.
Saccharomyces cerevisiae batch cultivation, Biosyst. Eng. 198 (2020) 198–209, [65] S.-M. Ioannidou, J.P. López-Gómez, J. Venus, M.A. Valera, V. Eßmann, I. Alegria-
https://doi.org/10.1016/j.biosystemseng.2020.08.004. Dallo, I.K. Kookos, A. Koutinas, D. Ladakis, Techno-economic evaluation and life
[45] A. Yousefi-Darani, O. Paquet-Durand, J. Hinrichs, B. Hitzmann, Parameter and cycle assessment for sustainable alternative biorefinery concepts using the organic
state estimation of backers yeast cultivation with a gas sensor array and unscented fraction of municipal solid waste, Green. Chem. 25 (2023) 4482–4500, https://doi.
Kalman filter, Eng. Life Sci. 21 (2021) 170–180, https://doi.org/10.1002/ org/10.1039/D3GC00244F.
elsc.202000058. [66] J. Monod, The growth of bacterial cultures, Annu. Rev. Microbiol. 3 (1949)
[46] R.-K. Lee, H.-W. Ryu, H. Oh, M. Kim, Y.-J. Wee, Cell-recycle continuous 371–394, https://doi.org/Doi: 10.1146/annurev.mi.03.100149.002103.
fermentation of Enterococcus faecalis RKY1 for economical production of lactic [67] F.F. Blackman, Optima and limiting factors, Ann. Bot. os-19 (1905) 281–296.
acid by reduction of yeast extract supplementation, 24 (n.d.) 661–666. https://doi. [68] J.B.S. Haldane, Enzymes, MIT Press, Cambridge, 1965.
org/10.4014/jmb.1402.02017. [69] G. Teissier, Growth of bacterial populations and the available substrate
[47] R. Alves De Oliveira, C.E. Vaz Rossell, B.H. Lunelli, P.O.M. Schichi, J. Venus, R. concentration, Rev. Sci. Instrum. (1942).
M. Filho, Different strategies to improve lactic acid productivity based on [70] H. Moser, Structure and dynamics of bacterial populations maintained in the
microorganism physiology and optimum operating conditions, Ind. Eng. Chem. chemostat, Cold Spring Harb. Symp. Quant. Biol. 22 (1957) 121–137, https://doi.
Res. 57 (2018) 10118–10125, https://doi.org/10.1021/acs.iecr.8b01655. org/10.1101/SQB.1957.022.01.015.
[48] E. Cubas-Cano, J. Venus, C. González-Fernández, E. Tomás-Pejó, Assessment of [71] D.E. CONTOIS, Kinetics of bacterial growth: relationship between population
different Bacillus coagulans strains for L-lactic acid production from defined media density and specific growth rate of continuous cultures, J. Gen. Microbiol. 21
and gardening hydrolysates: effect of lignocellulosic inhibitors, J. Biotechnol. 323 (1959) 40–50, https://doi.org/10.1099/00221287-21-1-40.
(2020) 9–16, https://doi.org/10.1016/j.jbiotec.2020.07.017. [72] K.J.H.Hans V. Westerhoff, Juke S. Lolkema, Roel Otto, Thermodynamics of growth
[49] L.S. Gordeev, A.V. Koznov, A.S. Skichko, Y.L. Gordeeva, Unstructured non-equilibrium thermodynamics of bacterial growth the phenomenological and
mathematical models of lactic acid biosynthesis kinetics: A review, Theor. Found. the mosaic approach, Biochim. Et. Biophys. Acta (BBA) - Rev. Bioenerg. 683 (1982)
Chem. Eng. 51 (2017) 175–190, https://doi.org/10.1134/S0040579517020026. 181–220, https://doi.org/10.1016/0304-4173(82)90001-5.
[50] M. Vogels, R. Zoccklcr, D.M. Stasiw, L.C. Ccrny+, P.F. Verhulst’s “Notice sur la loi [73] M.N.S. and, M. Aiba, Shoda, Kinetics of product inhibition in alcohol fermentation,
que la populations suit dans son accroissement” from Correspondence Biotechnol. Bioeng. 67 (2000) 671–690, https://doi.org/10.1002/(sici)1097-0290
Mathematique, n.d. (20000320)67:6<671::aid-bit6>3.0.co;2-w.
[51] J. Baranyi, T.A. Roberts, Review Paper A dynamic approach to predicting bacterial [74] K. Han, O. Levenspiel, Extended monod kinetics for substrate, product, and cell
growth in food, 1994. inhibition, Biotechnol. Bioeng. 32 (1988) 430–447, https://doi.org/10.1002/
[52] M. Babor, J. Senge, C.M. Rosell, D. Rodrigo, B. Hitzmann, Optimization of no-wait bit.260320404.
flowshop scheduling problem in bakery production with modified pso, neh and sa, [75] T. Yano, S. Koga, Dynamic behavior of the chemostat subject to substrate
Processes 9 (2021), https://doi.org/10.3390/pr9112044. inhibition, Biotechnol. Bioeng. 11 (1969) 139–153, https://doi.org/10.1002/
[53] X. Zhou, L. Ye, J.C. Wu, Efficient production of l-lactic acid by newly isolated bit.260110204.
thermophilic Bacillus coagulans WCP10-4 with high glucose tolerance, Appl. [76] E. Powell, Growth rate Microorg. a Funct. Substr. Conc. (1967).
Microbiol. Biotechnol. 97 (2013) 4309–4314, https://doi.org/10.1007/s00253- [77] J.H. Luong, Generalization of monod kinetics for analysis of growth data with
013-4710-7. substrate inhibition, Biotechnol. Bioeng. 29 (1987) 242–248, https://doi.org/
[54] C. Collar, A.F. Mascarós, J. Prieto, C.B. Barber, Changes in free amino acids during 10.1002/bit.260290215.
fermentation of wheat doughs started with pure culture of lactic acid bacteria, (n. [78] J.L. Webb, Enzyme and metabolic inhibitors, Academic Press, New York, 1963,
d.). 〈https://www.semanticscholar.org/paper/Changes-in-free-amino-acids-durin https://doi.org/10.5962/bhl.title.7320.
g-fermentation-of-Collar-Mascar〉ós/ [79] T.A. Omer, Anal. Bact. Popul. Growth Using Ext. Logist. Growth Model Distrib.
74149f3ca71e4c46dbf582fcc06fa4efcc2b4d5e. (2018), https://doi.org/10.48550/arXiv.1807.09108.
[55] K. Watanabe, K. Kitamura, Y. Suzuki, Analysis of the critical sites for protein [80] B. Babor, M. Paquet-Durand, O. Berg, C. Büchs, J. Hitzmann, Online process state
thermostabilization by proline substitution in oligo-1,6-glucosidase from Bacillus estimation for hansenula polymorpha cultivation with 2D fluorescence spectra-
coagulans ATCC 7050 and the evolutionary consideration of proline residues., 62 based chemometric model calibrated from a theoretical model in place of offline
(n.d.) 2066–2073. 〈https://www.ncbi.nlm.nih.gov/pmc/articles/PMC167984/〉. measurements, Fermentation 9 (2023), https://doi.org/10.3390/
[56] D. Tsuchiya, T. Sekiguchv, A. Takenaka’ J, Crystal structure of 3-isopropylmalate fermentation9020095.
dehydrogenase from the moderate facultative thermophile, bacillus coagulans: two [81] B.H.O. Paquet-Durand, T. Ladner, J. Bu¨chs, Calibration of a chemometric model by
strategies for thermostabilization of protein structures, J. Biochem 122 (1997) using a mathematical process model instead of offline measurements in case of a h.
1092–1104. 〈https://academic.oup.com/jb/article/122/6/1092/777615〉. polymorpha cultivation, Chemom. Intell. Lab. Syst. 171 (2017) 74–79, https://doi.
[57] L.L. Campbell, E.E. Sniff, Folic acid requirement of Bacillus coagulans, J. Bacteriol. org/10.1016/j.chemolab.2017.10.003.
78 (1959) 267–271, https://doi.org/10.1128/jb.78.2.267-271.1959.

315

You might also like