=Paper= {{Paper |id=Vol-2783/paper12 |storemode=property |title=Computer Simulation of Communication in Bacterial Populations under External Impact of Signal-Degrading Enzymes |pdfUrl=https://ceur-ws.org/Vol-2783/paper12.pdf |volume=Vol-2783 |authors=Christina Kuttler,Anna Maslovskaya |dblpUrl=https://dblp.org/rec/conf/birthday/KuttlerM20 }} ==Computer Simulation of Communication in Bacterial Populations under External Impact of Signal-Degrading Enzymes== https://ceur-ws.org/Vol-2783/paper12.pdf

Computer Simulation of Communication
in Bacterial Populations under External Impact
of Signal-Degrading Enzymes ? ??

Christina Kuttler1[0000−0002−5521−4159] and
Anna Maslovskaya2[0000−0002−5628−3519]
1
Technical University of Munich, Munich, Germany
kuttler@ma.tum.de
2
Amur State University, Blagoveshchensk, Russia
maslovskaya@mail.ru

Abstract. Time-dependent reaction-diffusion models of cell-to-cell bac-
terial communication in Gram-negative bacteria are developed with a
focus on the application of numerical methods and computer simulation
techniques. To model bacterial communication process in the special case
of quorum sensing, we consider an initial-boundary value problem for a
semilinear reaction-diffusion system. To solve the problem numerically,
we propose a computational algorithm based on a finite difference ap-
proach and Monte-Carlo simulations of bacterial population dynamics.
A special application software is designed in Matlab to perform com-
puter simulations of time-dependent characteristics of bacterial coopera-
tive behavior. The computer-simulated characteristics providing quorum
sensing are presented in the context of their changes under the external
addition of the defined signal and enzyme. The results of the computa-
tional experiments are discussed in view of the importance of controlling
the bacterial quorum sensing by means of the artificial impact of natural
enzymes.

Keywords: Bacterial communication · Quorum sensing · Reaction-dif-
fusion model · Finite difference method · Iterative algorithm · Monte-
Carlo simulation · Computer-animated signal substances.

1 Introduction

At the present stage, the design and development of mathematical models de-
scribing reaction-diffusion systems in biology and medicine have both, fundamen-
tal scientific and practical interest. Among the most relevant applied problems,
one can single out a class of models of microbiological communities, in particular,
?
Supported by the German Academic Exchange Service (DAAD), funding programm
number 57507437
??
Copyright c 2020 for this paper by its authors. Use permitted under Creative Com-
mons License Attribution 4.0 International (CC BY 4.0).
164 C. Kuttler, A. Maslovskaya

the model of the cell-to-cell communication of bacterial colonies due to quorum
sensing. In biology, quorum sensing describes ability of bacteria to detect and
respond to the (local) cell population density by a process of signaling and gene
regulation [16, 17]. This phenomenon is attributed to producing diffusible sig-
naling compounds by bacterial cells. The quorum sensing results in variability
and (often) mutual communication of bacteria. A bacterial colony with a high
local density can activate a variety of cellular processes including cooperative
activities, resistance to antibiotics and virulence factors [18].
The application of mathematical modelling and computer simulation tech-
niques to study bacterial cooperative behaviour is very useful for predicting and
controlling responses of bacterial communities to external exposures. It is less
costly and less time-consuming than laboratory experiments and allows even the
analysis of situations which may be not accessible by wet experiments, in case
of the system is already well-known. The mechanism of the bacterial quorum
sensing has been previously formalized mathematically as a dynamic system
described by ordinary differential equations [2, 11]. Numerous studies based on
a variety of model modifications have been reported elsewhere, taking into ac-
count bacterial population growth, feedback processes in formation of signal
compounds, a delay effect in a biological system, killing bacteria by antibiotics,
etc. [11, 13, 14]. In addition, the description of heterogeneous space distributions
of bacterial quorum sensing signaling molecules has been proposed in the form of
a system of reaction-diffusion parabolic partial differential equations (PDEs) [7].
This model is most suitable for specific bacterial strains which can produce spe-
cial enzymes (in particular, Lactonase), degrading signal substances produced
by bacteria. Such an inhibition of quorum sensing is also called ”quorum quench-
ing” [1, 4].
The strength of intercellular communication can be potentially reduced by
chemical agents. In these terms specific enzymes are promising candidates used
for degrading signaling molecules. Disconnected bacteria create biofilms with
essential difficulties, making them more vulnerable to both the immune system
and antimicrobial drugs. Biological studies have indicated that natural enzymes
can be used as inhibitors to prevent the development of infections caused by
pathogens [1]. Moreover, the idea of creating alternative antibiotics related to
the concept of natural enzymes application is under investigations [12].
In addition, the process of bacterial cell-to-cell communications under non-
equilibrium conditions by supplying external signal substances has been the
focus of intense research efforts, e.g., [3, 9]. In contrast to the previous case,
the addition of signalling compounds can lead to an amplification of the signal
and increasing bacterial community activities due to the positive feedback and
the other nonlinear interactions. But it may also lead to a faster decrease of
the signal, e.g., in the case of P. putida which produces the signal-degrading
Lactonase in a quorum sensing controlled way. As for some pathogenic bacterial
species their pathogeneity is controlled by quorum sensing, quorum quenching
processes are under discussion as possible treatments [4]. Therefore, it would be
of special interest to model the behaviour of time-dependent characteristics of
Computer simulation of communication in bacterial populations 165

quorum sensing taking into account the external addition of Lactonase or other
similar enzymes.
The current study aims at the further development of the dynamic semi-
linear reaction-diffusion model of bacterial quorum sensing with a focus on the
application of numerical methods and computer simulation techniques. The con-
tributions of this work are presented as follows: the computational approach for
modelling bacterial quorum sensing based on an iterative finite difference scheme
and a Monte-Carlo simulation of bacterial population growth, and the dynamic
characteristics of the biological system estimated by computational experiments
under the conditions of artificial additions of signal and/or enzymes.

2 Problem Formulation
2.1 Quorum Sensing: A Concise Biological Setup
In order to introduce the main characteristics of the state of a dynamic system,
let us present a brief formalized description of the bacterial communication pro-
cess due to quorum sensing, without pretending to be a complete exposition of
the biological foundations of this multifaceted and complex phenomenon. To be
definite, we will assume that the class of Gram-negative bacteria is under con-
sideration. More concretely, we survey a class of Pseudomonas bacterial species,
namely, Pseudomonas putida IsoF (P. putida). This bacterial strain is one of
the most studied from the point of view of the observation of joint dynamic
processes: the quorum sensing and quorum quenching by an enzymatic activity.
The functional regulation network of the quorum sensing mechanism in P.
putida is shown in Fig. 1.

AHL autoinducer AHL autoinducer AHL autoinducer

Lactonase
Target genes Target genes

PpuR AHL synthase PpuR-AHL AHL synthase PpuR-AHL AHL synthase
transcription (PpuI) complex (PpuI) complex (PpuI)
regulator

Fig. 1. Sketch of the sequential stages of the quorum sensing regulatory network in
P. putida (Ppu).

Bacterial quorum-sensing is realized by the generation and distribution of
special signaling molecules or autoinducers (N-acyl Homoserine Lactones – the
short notation AHL will be used hereinafter), which allows the bacterial colony
to control its behaviour by gene regulation dependent on the local population
density and to form a response to external influences. The other elements of the
regulation system are the follows: The AHL synthases are enzymes responsible
166 C. Kuttler, A. Maslovskaya

for the synthesis of autoinducers (proteins of the LuxI family), and transcrip-
tional regulators, which play important role in coordinating the expression of a
variety of genes (proteins of the LuxR family). The AHL can penetrate through
cell membranes (out and in), then diffuses. Inside the cells, it stimulates regu-
latory proteins. This process results in increasing the AHL concentration as a
function of the cell density due to a positive feedback in the biological regulation
system.
For special bacterial strains like P. putida IsoF, specific self-produced Lac-
tonase enzymes can degrade AHL-autoinducers. The generation of Lactonase can
be associated with the appearance of negative feedback in a dynamic system.
Also, during the evolution of the bacterial population, the natural degradation
of the AHL and Lactonase occur.
The systematical approach allows us to arrive at a mathematical model of
bacterial quorum sensing, which can be expressed by a system of time-dependent
reaction-diffusion equations [7].

2.2 Dynamic Reaction-Diffusion Model of Quorum Sensing
The system of the following time-varying semilinear reaction-diffusion PDEs
describes the dynamic biological system in the spatially one-dimensional case:
∂U ∂2U
= DU − γU U − γL→U LU + F1 (x, U ), (1)
∂t ∂x2
∂L ∂2L
= DL 2 − γL L + F2 (x, U ), 0 < x < l, 0 < t ≤ T, (2)
∂t ∂x
where U (x, t) is the AHL concentration and L(x, t) the Lactonase concentration
produced by bacteria in mol/l; l is the linear size of the domain solution in µm;
T is the observation time in h.
The model parameters will be specified in the following section devoted to
computational experiments. The governing equations (1), (2) define the dynam-
ics and diffusion of AHL and Lactonase concentrations, the natural degradation
of AHL and Lactonase, the production of AHL (positive feedback), the pro-
duction of the Lactonase and the degradation of AHL by Lactonase (negative
feedback).
We consider V discrete bacterial colonies. The source functions F1 (x,U ) and
F2 (x,U ) are specified using the normal distribution of bacterial cells density and
the general form of the Hill function:
V
!
2
X (x − xvc )
Fm (x, U ) = fm exp − , m = 1, 2, (3)
v=1
σ2

Un Un
f1 (U ) = αU + βU n , f2 (U ) = βL n , (4)
((Uth ) + U n ) ((Uth + ε) + U n )

where xvc is the position of the bacterial colony with v = 1, . . . , V ; σ, ε, αU , βU ,
βL , Uth , n are model parameters which also should be specified.
Computer simulation of communication in bacterial populations 167

The equations (1), (2) are completed with an appropriate set of initial and
boundary conditions:
U (x, 0) = 0, L(x, 0) = 0, 0 ≤ x ≤ l, (5)
U (0, t) = 0, U (l, t) = 0, L(0, t) = 0, L(l, t) = 0, 0 ≤ t ≤ T. (6)
We can stress that the basic model of quorum sensing (1)–(6) can be extended
by introducing the two-dimensional analogue. In this case we use a 2D solution
domain and define the dependent variables as functions U (x,y,t) and L(x,y,t) for
the AHL concentration and Lactonase concentrations. Here the diffusion terms
are written in the form Uxx + Uyy and Lxx + Lyy , respectively. Corresponding
transformations of the functions (3), (4), initial and boundary conditions (5),
(6), should be performed related to the two-dimensional case.
Thus, the mathematical model is formalized by the initial-boundary value
problem for system of semilinear reaction-diffusion PDEs (1)–(6). Some aspects
concerning existence and uniqueness of solutions can be found in [7]. The current
study focuses on the application of numerical methods, namely finite difference
methods combined with a Monte-Carlo simulation, to find an adequate solution
of the problem (1)–(6).

3 Numerical Aspects of the Computational Algorithm
3.1 Finite Difference Iterative Scheme for One-dimensional
Problem
The initial-boundary value problem (1)–(6) was solved numerically by an implicit
three-layer difference method combined with an iterative procedure [15]. Let us
introduce a rectangular space-time grid covering a solution domain:
Ωhτ = {xi = (i − 1)h, i = 1, 2, ..., N + 1, tk = (k − 1)τ, k = 1, 2, ..., K + 1},
where N and K are the positive integers.
The model equations (1) and (2) can be rewritten in the following general
view, that can be suitably used for modeling both AHL as well as Lactonase
concentrations:
∂v ∂2v
= D 2 − wv + F, (7)
∂t ∂x
where w = γU + γL→U L for the equation (1) and w = γL for the equation (2).
(s)
So far as we have a nonlinear term in (7), the iterative sequence {vi } for
k+1
s = 1, 2, ... for each time layer k + 1 is derived, converging to vi . As an
initial approximation for the iterative sequence, we take for each k + 1 time
(1)
layer vi = vik . Here we use the Crank–Nicolson scheme to find a solution at
the second time layer. After some transformations, we get for i = 2, 3, ..., N ,
s = 1, 2, ..., taking into account the initial conditions:

Dτ (s+1) Dτ τ 2 (s+1) Dτ (s+1) τ e
− 2 vi−1 + 1 + 2 + wi vi + − 2 vi+1 = F + F (xi , 0) ,
2h h 2 2h 2
(8)
168 C. Kuttler, A. Maslovskaya

(s)
where Fe = F xi , Ui is the source function for the equation (1) and Fe =
F (xi , 0) is the source function for the equation (2).
For others nodes we apply the three-layer implicit finite-difference scheme [15].
This yields that for i = 2, 3, ..., N , k = 2, 3, ..., K, s = 1, 2, 3... we arrive at

2Dτ (s+1) 4Dτ (s+1) 2Dτ (s+1)
− 2 vi−1 + 3 + 2 + 2τ wik+1 vi + − 2 vi+1 =
h h h

= 4vik − vik−1 + 2τ Fe, (9)
where wik+1 = γL for
the equation
(2) and wik+1 = γU + γL→U Lk+1i for the
(s) k

equation (1), F = F xi , U
e
i for the equation (1) and F = F xi , U for the
e
i
equation (2).
A standard procedure of approximation analysis for the computational scheme
(8), (9) results in the second order accuracy with respect to space and time vari-
ables, O h2 + τ 2 . However, to estimate the Lactonase concentration, we use of
the value of the AHL concentration from the previous time layer k. This can lead
to reducing the accuracy of the general scheme. To avoid this problem numeri-
cally, we apply a procedure based on a predictor-corrector approach. Actually,
first, we solve the equation for the Lactonase concentration, then we calculate
the AHL concentration, and finally using the specified value for the AHL concen-
tration on the k + 1 time layer, we estimate the refined value for the Lactonase
concentration applying the same schemes (8), (9) with Fe = F xi , Uik+1 .

The initial and boundary conditions are also satisfied: v1k = 0, vNk
+1 = 0 for
1
k = 1, 2, ..., K + 1, vi = 0 for i = 1, 2, ..., N + 1. In many practical situations,
the use of iterative procedures in solving semilinear reaction-diffusion equations
allows one to keep the level of accuracy corresponding to the order of approx-
imation of the applied numerical method [5, 6, 8]. Also, the stability analysis
by the linear approximation leads to absolute stability as well as unconditional
monotony of the numerical scheme (9). The system of linear algebraic equations
is solved efficiently by the high accuracy sweep method on each time layer.

3.2 Finite Difference Iterative Scheme for the Two-Dimensional
Problem

Similarly, the two-dimensional model equations can be attributed to the general
form of 2
∂2v

∂v ∂ v
=D + 2 − wv + F. (10)
∂t ∂x2 ∂y
In this case we assume Ωhτ 1 ,h2 to be a space-time grid covering a solution
domain: Ωhτ 1 ,h2 = {xi = (i − 1)h1 , i = 1, 2, ..., N + 1, yj = (j − 1)h2 , j =
1, 2, ..., M +1, tk = (k −1)τ, k = 1, 2, ..., K +1}. To solve the equation (10), here
we apply a splitting finite-difference method, namely the alternating direction
method [15]. At the first semi-step k + 1/2 for i = 2, 3, ..., N , j = 1, 2, ..., M ,
Computer simulation of communication in bacterial populations 169

k = 1, 2, ..., K, s = 1, 2, ... we can define:
k+1/2
" #
τ wi,j

Dτ (s+1) Dτ (s+1) Dτ (s+1)
− 2 vi−1,j + 1 + 2 + vi,j + − 2 vi+1,j =
2h1 h1 2 h1

k Dτ k k k
τ
= vi,j + 2 vi,j−1 − 2vi,j + vi,j+1 + Fe, (11)
2h2 2
(s)
where for the k + 1/2 time layer the iterative sequence {vi,j } converges to the
k+1/2 (1) k k+1/2
value vi,j , starting with vi,j = vi,j ; wi,j = γL for the equation (2) and
k+1/2 k+1/2
wi,j = γU + γL→U Li,j for the equation (1).
k

We define Fe = F xi , yj , Ui,j for the equation (2) (to estimate the Lactonase

(s)
concentration), then we use Fe = F xi , yj , Ui,j for the equation (1) for the AHL
concentration,
and then, again calculate the Lactonase concentration using F =
e
k+1/2
F xi , yj , Ui,j . Further we derive the scheme for the second time semistep
k + 1:
(s+1)
" #
τ wi,j

Dτ (s+1) Dτ (s+1) Dτ (s+1)
− 2 vi,j−1 + 1 + 2 + vi,j + − 2 vi,j+1 =
2h2 h2 2 h2

k+1/2 Dτ h k+1/2 k+1/2 k+1/2
i τ
= vi,j + v − 2v + v + Fe, (12)
2h21 i−1,j i,j i+1,j
2
(s) k+1 (1)
where the iterative sequence {vi,j } converges to the vi,j , starting with vi,j =
k+1/2 k+1 k+1
vi,j , wi,j = γL for the equation (2) and wi,j = γU + γL→U Lk+1 i,j for

k+1/2
the equation (1); F = F xi , yj , Ui,j
e is defined for the equation (2), F =
e

(s)
F xi , yj , Ui,j – for the equation (1).
Here, we apply the similar approach to estimate source functions. The equa-
tions (11) and (12) are also complemented by the discrete initial and bound-
1 k
ary conditions: vi,j = 0 for i = 1, 2, ..., N + 1, j = 1, 2, ..., M + 1, v1,j = 0,
k k k
vN +1,j = 0 for j = 1, 2, ..., M + 1, k = 1, 2, ..., K + 1, and vi,1 = 0, vi,M +1 = 0
for i = 1, 2, ..., N + 1, k = 1, 2, ..., K + 1 .
The derived scheme has the second order of approximation with respect to
space and time variables O h21 + h22 + τ 2 . Likewise to the one-dimensional case,
we can apply the sweep method to solve the obtained system on each time
semistep.

3.3 Stochastic Algorithm for the Bacterial Population Dynamics

Evidently, the model with static positions and sizes of bacterial colonies does
not suitable describe a realistic behaviour of microbe populations and, as a con-
sequence, spatial distributions of signal compounds. In this case we combine the
170 C. Kuttler, A. Maslovskaya

quorum sensing model with the time-dependent simulation of bacterial nucle-
ation and growth. During this process, a “mother” bacterial cell enlarges and
divides into two new “daughter” cells and these new bacterial cells can also
nucleate later. In general, the growth of bacterial population is a complex pro-
cess, which has geometrical, exponential or logistical character. Here we use the
ideas of stochastic principles of bacterial nucleation and the logistic mechanism
of bacterial growth. Thus, we combine the deterministic approach for the AHL
and Lactonase concentrations with a stochastic procedure to describe changes
in positions of bacterial colonies.
The stochastic algorithm includes the following steps. First, up to three bac-
terial colonies with different probabilities start to grow simultaneously from ran-
domly chosen points (for example, xvc for the one-dimensional case) as shown in
Fig. 2. All bacterial colonies are assumed to grow with a similar velocity. The
value of the linear size Rv (t) of each bacterial colony is determined by the logistic
law of growth. According to the Monte-Carlo method, a new bacterial “nucleus”
can appear at a fixed random time moment, but with a small probability on each
time layer. Then a new colony begins to grow in a self-similar way also according
to the logistic law. If the cells are overlapping, the source functions F1 and F2
are determined by a superposition of colonies contributions.

Fig. 2. The example of a state of bacterial population during growth phase.

The general computational schemes both for 1D model and 2D model were
implemented in Matlab. The designed application software can be used to per-
form computer simulations of time-dependent characteristics of bacterial coop-
erative behaviour at given parameters.

4 Computer Simulations Results and Discussion
In order to perform computer simulation we need to specify the conditions of
model implementation and initialize model parameters. For instance, let us take
a set of parameters estimated for the bacterial strain P. putida IsoF (see Table 1),
which have been reported previously in [7].
We aim to observe the simulation process during the defined time period
varied over the range 10-30 h. For the 1D-model we consider the solution domain
for 0 ≤ x ≤ 100 µm, and in the same way for 2D-model we take a square domain
limited by 0 ≤ x ≤ 100 µm, and 0 ≤ y ≤ 100 µm.
Computer simulation of communication in bacterial populations 171

Table 1. Parameter values estimated for Bacterium P. putida

Name Meaning of parameter Value Unit
DU Diffusion rate of AHL 100 µm2 /h
DL Diffusion rate of Lactonase 1 µm2 /h
γU Abiotic degradation rate of AHL 0.005545 1/h
γL Abiotic degradation rate of Lactonase 0.5 1/h
γL→U Degradation rate of AHL by Lactonase 0.65 · 109 l/(mol·h)
αU Low production rate of AHL 1.058 · 10−7 mol/(l·h)
βU Increased production rate of AHL 1.058 · 10−6 mol/(l·h)
βL Production rate of Lactonase 1.38 · 10−6 mol/(l·h)
Uth Threshold of AHL concentration between low 7 · 10−8 mol/l
and increased activity
ε Threshold shift for Lactonase production 5 · 10−9 mol/l
n Degree of polymerization 2.5 –
l Linear size of the object 100 µm

All computations we performed for N = M = 100 numbers of nodes over
space variables and 100 numbers of nodes over a time variable. This was enough
to get a sufficiently accurate solution. To examine the adequacy of the algo-
rithms, we solved test problems with known analytical solutions.

4.1 Dynamics of Components of the Quorum Sensing Model:
Numerical Analysis
In the first computational experiment we perform simulations of time-space dis-
tributions of the AHL and Lactonase concentrations for 1D model. In this case
three bacterial colonies are located at symmetrical positions of the solution do-
main: x1c = 0.25l, x2c = 0.5l, x3c = 0.75l. The observation time is set after 15
hours. To visualize the dynamics of simulated characteristics we use a time-
dependent profile at the central position of x2c = 0.5l.
Before considering the results of direct calculations of the AHL and Lactonase
concentrations, let us start with the numerical analysis of the contributions of
the different parts of the equations, which create the dynamical response of the
biological system.
To estimate these contributions, we calculated U (xi , tk ), L(xi , tk ) and then
evaluated each term in the right sides of PDEs (1), (2) multiplied by the time step
τ = 0.1 h. Here we apply the following notations for the equation (1): P1 = F1 τ
for the source function, P2 = DU τ Uxx for the diffusion term, P3 = γU U τ for the
AHL degradation term, P4 = γL→U LU τ for the term specified the degradation
of AHL by Lactonase. In the same way for the equation (2) we have: Q1 = F2 τ
for the source function, Q2 = DL τ Lxx for the diffusion term, Q3 = γL U τ for
the Lactonase degradation term.
Figure 3 shows the contributions of the different parts of the equation (1)
defined dynamics of the AHL concentration as time-dependent functions. The
dynamics of the model components for the Lactonase concentration is illustrated
in Fig. 4.
172 C. Kuttler, A. Maslovskaya

Fig. 3. The time-dependent parts of right side corresponding to the equation (1) defined
the AHL concentration dynamics: P1 (t) – 1, P2 (t) – 2, P3 (t) – 3, P4 (t) – 4.

Fig. 4. The time-dependent parts of the right hand side corresponding to the equation
(2) defined the Lactonase concentration dynamics: Q1 (t) – 1, Q2 (t) – 2, Q3 (t) – 3.

These graphs allow us to see which processes a priority affect the distributions
of the AHL and Lactonase concentrations for given parameter values. Here we
can conclude, that the AHL own degradation P3 (t) specified by the term of γU U τ
has a negligible influence on the resulting AHL distribution. Indeed, this degra-
Computer simulation of communication in bacterial populations 173

dation is very small under realistic conditions. At the initial time moments, the
dynamics of the AHL concentration is determined mainly by the production term
and diffusion process followed by the influence of the Lactonase. The changes
in the Lactonase concentration are determined by production, diffusion and the
own degradation simultaneously during the whole time range. It is obvious that
after 6-7 hours the state of the biological system tends to relaxation in general.
These effects are observed under natural conditions with self-produced AHL and
Lactonase. Hence we may assume that a variation of external exposure can re-
sult in changes in a state of the biological system. In particular, the artificial
supply can lead to relevant changes in the AHL and Lactonase concentrations.
This potentially means that we can artificially influence the quorum sensing of
a bacterial population.

4.2 Simulation of Quorum Sensing Characteristics under Changes
of External Conditions
In this Section we present the results of numerical simulation of bacterial quo-
rum sensing under variation of external exposure. Specifically, we consider the
artificial addition of the AHL and Lactonase at certain time point t∗ . Here we
also consider the implementation of the one-dimensional model for three bacte-
rial colonies located at fixed positions. The observation time is set to be 15 hour.
The simulated characteristics are visualized at the central position x2c = 0.5l.
Figures 5 and 6 show the time-dependent profiles of the AHL and Lactonase
concentrations under variations of external conditions.
Here we consider four modes: without any influence, under the addition of
the external AHL concentration 9.9 nmol/l equal to the value of the own AHL
concentration at the time moment t∗ = 6 h, under the addition of the external
Lactonase concentration 8.03 nmol/l equal to the current value of Lactonase at
the time moment t∗ = 6 h, and under the addition both of them. These data
suggest that the addition of the external AHL concentration leads to a rapid rise
not only in the AHL but also in the Lactonase concentration. This is followed by
the AHL concentration (with respect to natural conditions) due to the presence
of the negative feedback. These basic findings are consistent with research [9]
showing that slight fluctuations of signaling compounds arise under supplying
the external AHL, that can be visualized with the use of inverted microscopy.
At the same time the external addition of the Lactonase causes a significant
decrease in the concentration of signaling molecules. Furthermore, the addition
both of substances gives rise to an increase in the AHL concentration before
falling to certain value. The example of time-spatial visualization of the com-
puted AHL and Lactonase concentrations are shown in Fig. 7 and Fig. 8. The
simulation time equals 30 hours. During the period from hour 6 to hour 12, ev-
ery two hours the external concentration of Lactonase enzymes 32 nmol/l was
added. This yields that the AHL concentration drops dramatically to approxi-
mately 1.5 nmol/l.
Since the AHL concentration is referred to as the quorum sensing “level”,
we can make bacteria “fall silent” during the period of action of the external
174 C. Kuttler, A. Maslovskaya

Fig. 5. The time-dependent profiles of the AHL concentration at t∗ = 6 h under the
conditions: 1 – without impact, 2 – at the addition of the external AHL, 3 – at the
addition of the Lactonase, 4 – at the addition both of AHL and Lactonase.

Fig. 6. The time-dependent profiles of the Lactonase concentration at t∗ = 6 h under
the conditions: 1 – without impact, 2 – at the addition of the external AHL, 3 – at the
addition of the Lactonase, 4 – at the addition both of AHL and Lactonase.

Lactonase concentration. This implies that this period can be used to apply
antibiotic treatment for some pathogen bacterial strains, which demonstrate high
Computer simulation of communication in bacterial populations 175

Fig. 7. The time-spatial distribution of the AHL concentration under the external
addition of the Lactonase enzymes.

Fig. 8. The time-spatial distribution of the Lactonase concentration under the external
addition of Lactonase.

tolerance to antibiotics due to quorum sensing. For example P. putida strains
exhibit resistance to a large number of antibiotics [10]. The developed reaction-
diffusion model can potentially help to estimate the dose of external substances
176 C. Kuttler, A. Maslovskaya

needed to be added to control bacterial quorum sensing and the optimal time
points for these additions.

4.3 2D Simulation of Quorum Sensing Characteristics

The computer simulation based on the 2D-model allows us to visualize the quo-
rum sensing characteristics in more realistic manner. In this case we will present
the simulation results for the 2D reaction-diffusion model combined with the pro-
cedure of the Monte-Carlo simulation of bacterial population growth. According
to above algorithm, one, two or three bacterial colonies with a circle shape can
start to grow at initial time moment. The radius of each bacterial colony is set
to be 1 µm. Simulation results are presented in Fig. 9.

(a) (b)

Fig. 9. The dynamics of AHL space distribution at the fixed time moments: (a) t = 3 h;
(b) t = 7.25 h; (c) t = 7.75 h; (d) t = 25 h.

The total simulation time corresponds to 25 hours. During this period bac-
terial colonies are growing. If the total size of all bacterial colonies reaches the
value of R0 = 20 µm, the growth finishes due to the underlying model assumption
while new small colonies still can appear. One limitation of our implementation
is that only the stage of bacterial growth is under consideration (in reality the
degradation phase is followed by the growth phase). Here we assume that at the
Computer simulation of communication in bacterial populations 177

time point t∗ = 7.5 h, when the total linear size reaches R0 , the external Lac-
tonase concentration of 0.5·10−8 mol/l is added artificially. There are eighteen
bacterial colonies simulated with logistic-stochastic procedure during this com-
putational experiment. Figure 9 shows four frames with spatial distributions of
the AHL concentration calculated at the time moments: t = 3 h corresponding
the stage of bacterial growth under natural conditions (Fig. 9 (a)), t = 7.25 h
– before 0.25 h the addition of the external Lactonase enzymes (Fig. 9 (b)),
t = 7.75 h – after 0.25 h from the addition of Lactonase (Fig. 9 (c)) and t = 25 h
– at the end of the simulation process (Fig. 9 (d)).
The graph in Fig. 10 demonstrates the dynamics of changes in the maximum
value of the Lactonase concentration of the whole biological system. As it can
be expected that even a single injection of the external Lactonase leads to a
significant decrease of the AHL concentration. The maximum value of the AHL
concentration falls to the level of 5 nmol/l and then starts again to grow due to
mechanisms which underline the mathematical model.

Fig. 10. The time-dependent profile of maximum values of the Lactonase concentration
calculated during the simulation.

These findings suggest that artificial supply to bacterial populations by an
additional enzyme permits to influence the process of bacterial communication
significantly. For example, the external addition of Lactonase allows one to per-
form an antibacterial treatment in the sense of reducing, e.g., their pathogenic-
ity. Thus, the reaction-diffusion model of bacterial quorum sensing can be po-
tentially used as a scientific support for processes to control the population-
communication of bacteria.
178 C. Kuttler, A. Maslovskaya

5 Conclusion

In summary, this paper suggests a modification of the reaction-diffusion model
of bacterial quorum sensing based on combination of the deterministic approach
with logistic-stochastic simulation procedure of bacterial population dynamics.
For the numerical computer implementation of 1D as well as 2D models
we proposed computational schemes based on a joint application of finite dif-
ference methods and iterative procedures. This allows to provide the accuracy
corresponding to the order of approximation of applied numerical methods for
solving semilinear reaction-diffusion equations. The computational schemes have
second order accuracy with respect to space and time variables and are abso-
lutely stable. To describe the dynamics of bacterial colonies we also proposed the
algorithm based on Monte-Carlo simulation. The algorithm includes the scheme
of logistic growth of bacterial colonies.
The general computational schemes both for 1D model and 2D model were
implemented using Matlab programming. A designed application software was
applied for computer simulations of time-dependent characteristics of bacterial
cooperative behaviour.
We conducted a series of computational experiments specifically for Pseu-
domonas putida bacterial strains, for which the quorum sensing phenomenon
and producing a special enzyme, the Lactonase, are well studied in biological
experiments and corresponding mathematical models.
The computer simulated time-space distributions of the AHL and Lactonase
concentrations are presented with a focus on their changes under the external
addition of substances. Our findings indicate that the repeated addition of Lac-
tonase enzymes can significantly reduce the AHL concentration. As far as the
AHL concentration is associated with the quorum sensing “level”, we can make
bacteria “fall silent” during the period of action of the external Lactonase concen-
tration. This effect allows to make the application of antibiotic treatment more
effective for some pathogenic bacterial strains. Therefore, the developed model of
bacterial quorum sensing might be addressed in future studies of predicting and
controlling population-communication of bacteria, where spatial heterogeneity
plays a relevant role.

References

1. Ananda, M., Rusmana, I., Akhdiya, A.: Quorum quenching of Bacillus cereus
INT1c against Pseudomonas syringae. IOP Conf. Series: Journal of Physics: Conf.
Series 1277, 012010(9) (2019)
2. Dockery, J.D., Keener, J.P.: A mathematical model for quorum sensing in Pseu-
domonas aeruginosa. Bull. Math. Biol. 63, 95–116 (2001)
3. Fekete, A., Kuttler, C., Rothballer, M., Hense, B.A., Fischer, D., Buddrus-
Schiemann, K., Lucio, M., Müller, J., Schmitt-Koppling, P., Hartmann, A.: Dy-
namic regulation of N-acyl-homoserine lactone production and degradation in
Pseudomonas putida IsoF. FEMS Microbiol. Ecol. 72, 22–34 (2010)
Computer simulation of communication in bacterial populations 179

4. Kalia, V.C. (ed): Quorum sensing vs quorum quenching: a battle with no end in
sight. Springer, India (2015)
5. Kovtanyuk, A.E., Chebotarev, A.Yu., Botkin, N.D., Turova, V.L., Sidorenko, I.N.,
Lampe, R.: Continuum model of oxygen transport in brain. J. Math. Anal. Appl.
474, 1352–1363 (2019)
6. Kovtanyuk, A.E., Chebotarev, A.Yu., Botkin, N.D., Turova, V.L., Sidorenko, I.N.,
Lampe, R.: Nonstationary model of oxygen transport in brain tissue. Comp. Math.
Meth. Med. 2020, 4861654 (2020)
7. Kuttler, C.: Reaction-diffusion equations and their application on bacterial com-
munication. Handbook of statistics (Chapter 4), 55–91 (2017)
8. Maslovskaya, A.G., Moroz, L.I., Chebotarev, A.Yu., Kovtanyuk, A.E.: Theoretical
and numerical analysis of the Landau–Khalatnikov model of ferroelectric hysteresis.
Commun. Nonlin. Sci. Num. Simulat. 93, 105524 (2020)
9. Meyer, A., Megerle, J.A., Kuttler C., Müller, J., Aguilar, C., Eberl, L., Hense,
B.A., Rädler, J.O.: Dynamic of AHL mediated quorum sensing under flow and
non-flow conditions. Phys. Biol. 9, 026007(10) (2012)
10. Molina, L., Udaondo, Z., Duque, E., Fernández, M., Molina-Santiago, C., Roca, A.,
Porcel, M., de la Torre, J., Segura, A., Plesiat, P., Jeannot, K., Ramos, J.L.: An-
tibiotic resistance determinants in a Pseudomonas putida strain isolated from a
hospital. PloS one 9(1), e81604 (2014)
11. Müller, J., Kuttler, C., Hense, B.A., Rothballer, M., Hartmann, A.: Cell-cell com-
munication by quorum sensing and dimension-reduction. J. Math. Biol. 53, 672–
702 (2006)
12. Pang, Z., Raudonis, R., Glick, B.R., Lin, T.-J., Cheng, Z.: Antibiotic resistance
in Pseudomonas aeruginosa: mechanisms and alternative therapeutic strategies.
Biotechnology Advances 37(1), 177–192 (2019)
13. Perez-Velazquez, J., Gölgeli, M., Garcia-Contreras, R.: Mathematical modelling of
bacterial quorum sensing: a review. Bull. Math. Biol. 76, 1585–1639 (2016)
14. Roberts, P.A., Huebinger, R.M., Keen, E., Krachler, A.-M., Jabbari, S.: Mathe-
matical model predicts anti-adhesion antibiotic debridement combination therapies
can clear an antibiotic resistant infection. PLoS Comput. Biol. 15(7), e1007211(39)
(2019)
15. Samarskii, A.A.: The theory of difference schemes. Marcel Dekker Inc., New York
(2001).
16. Whitehead, N.A., Barnard, A.M.L., Slater, H., Simpson, N.J.L., Salmond, G.P.C.:
Quorum sensing in Gram-negative bacteria. FEMS Microbiol. Rev. 25, 365–404
(2001)
17. Williams, P., Winzer, K., Chan, W.C., Camara, M.: Look who’s talking: commu-
nication and quorum sensing in the bacterial world. Phil. Trans. R. Soc. B. 362,
1119–1134 (2007)
18. Zhao, X., Yu, Z., Ding, T.: Quorum-sensing regulation of antimicrobial resistance
in bacteria. Microorganisms 8(3), 425–446 (2020)