A P system model of swarming and aggregation in a Myxobacterial colony
Abstract
Bacterial communities provide an interesting subject for the study of emergence and complexity as the consequence of many local interactions. In particular, the soildwelling social bacterium Myxobacteria demonstrates two distinct types of motility, social motility via the sensing of bacterial slime deposits and adventurous motility. Both modes of motility are governed by local interactions. Using P systems, a membrane computing methodology based on compartmental rewrite rules for modelling computational processes; this work demonstrates how minimal set of rules can model swarming and aggregating behaviour in Myxobacteria bacterial populations. Our model uses a multienvironment P system similar a 2D cellular automaton to represent the substrate environment whilst stochastic rule selection dictates Myxobacterial motion according to behaviour observed in vitro. The rules account for both mechanisms of motility, the deposit and detection of slime, a change in direction due to Csignal induction and the mixing of population numbers. Simulations demonstrate an extensible computational framework for the modelling of bacterial behaviour, with the potential for extension into additional emergent behaviours.
Keywords
P systems Multienvironment P systems Myxobacteria Membrane computing Nature inspired computing Emergence1 Introduction
Bacterial populations provide an interesting subject for the study of emergence and complexity phenomena. Bacteria react to the environment in complex ways and use shortdistance local communication to start coordinated behaviour that can involve hundreds of thousands of cells. Tissues in multicellular organisms, on the other hand, are often controlled centrally via hormones and other signals. The Gramnegative soildwelling bacterium, Myxobacteria, best known for population level rippling and fruiting body stages whilst under starving conditions, provides an interesting and extensively studied vehicle for understanding cellular communication and spatial behaviour in bacteria [1, 2, 3, 4, 5, 6, 7]. In particular, phenomena such as swarming, aggregation and rippling in the Myxobacteria Myxococcus xanthus continue to elude researchers in their understanding of the mechanisms by which Myxobacteria communicate and organise themselves to predate prey bacteria [8].
Myxobacteria are able to move over a substrate using slow gliding motions driven by two modes of regulated transport, adventurous (Amotility) and social (Smotility) [7, 9, 10]. Bacteria at the colony edge or under isolation can undergo an adventurous mode of motility through the secretion of slime to propel the organism forward. Through celltocell contactsignal (Csignal) transduction, bacterium undergoes reversal of the intracellular motility mechanism [11]. After a number of hours, the accumulation of Csignals across a population will cause an oscillatory motion within the population, giving a visual impression of a diffusing wave. Although it appears as though colliding waves pass through one another, the headtohead impact of bacterium causes the wave to reverse resulting in very little net displacement. Under Smotility the cell relies on the extension and retraction of type IV pili: these extend from the leading cell pole and grip the polysaccharide cell surface or solid substrate surface. Both type IV pilibased motility and slime extrusion based motility require the Frz pathway to induce forward movement in the direction of the long axis of the cell.
P systems are distributed parallel computability models based on cellular membrane compartmentalisation and diffusing chemical signals within and across membranes [12]. A system is defined within a unique outer membrane known as the skin that can hold any number of membrane structures, similar to the organelles inside a living cell. Each membrane has a cellular region which can hold a multiset of objects defined by some alphabet and a set of rewrite rules that enables the multiset of objects to evolve under the correct conditions. Objects can diffuse across membranes analogous to cellular signals, and membranes can dissolve passing their chemical contents onto the parent membrane region. Rules can be assigned a priority rating, giving the set of rules for a single membrane a specific order of execution similar to the downstream transduction of receptor activation. If a highpriority rule cannot execute under current conditions, then any additional rules are tested. The priority of rules is typically denoted as \(r_{1} > r_{2}\), to indicate that the first rule has a higher execution priority over the second rule. Additionally, the prioritising of rules can be assigned using a stochastic constant [13]. These stochastic P systems use the Gillespie algorithm to determine which rule to execute and the order.
The evolution of a P system is performed in parallel across all cellular regions similar to the parallel operation of the organelles within the cell and the chemical operations within the main cytoplasmic space itself. Operations are performed either according to the order of each rule, a stochastic selection process, or a mixture of the two. P systems can generally be divided into two categories: the first, noncooperative, where only single objects can evolve; and the second, cooperative, where multiple objects can evolve in parallel [14]. If every membrane is incapable of evolving over its multiset, the system is said to halt. Upon halting, values can be read from a specific membrane region as defined in the P system formal definition. Under such conditions, the P system is said to be computationally complete, yet unlike the model proposed in this study, these systems would be used to generate a value or a decision rather than spatial behaviour.
Due to inherent compartmentalisation of a P systems model and their encapsulation of chemical diffusion across membranes, they have proved very apt in modelling Quorum sensing [15, 16, 17, 18, 19]. The spatial dynamics of diffusing signals can be captured in a gridlike P system, known as a multienvironment, similar to that of a cellular automaton [20]. Further works of modelling biological systems using membrane computing can be seen in Cardona et al. [21, 22], which look at the dynamics of ecosystems. To address the inherent randomness and uncertainty in large biological systems, multienvironment P systems have been combined with probabilistic terms [23, 24].
We explore the use of the P system formalism to increase our understanding of spatial phenomena in Myxobacteria. Recent extensions to the P system formalism allow simulating the models with a stochastic engine, which is crucial to capture realistic biochemical scenarios [25]. While the spatial behaviour of Myxobacteria is controlled primarily by a nondiffusible contactbased signal, we explore commonalities between this form of communication and chemical quorum sensing which underlies other modelling work using P systems.
2 Methods
The methods proposed in this study begin with a brief description of the formal definition for a P system; the reader is advised to refer to the work of Păun et al. [26]. The P system construct is then described in light of an extension to incorporate a twodimensional spatial construct. The execution of such a multienvironment P system is covered with a brief description of the Gillespie algorithm. Finally, the twodimensional model and Myxobacterial rewrite rules are reviewed.
2.1 P system formal definition
P systems can be graphically represented as Venn diagrams and trees. Venn diagrams help illustrate the membrane structure in respects to the physiology of a real biological cell, in addition to showing the multiset of objects and membrane rules.
A P system containing \(n\) cellular regions delimited by membranes is defined formally as

\({{\varSigma }}\) is an alphabet of objects that populate a multiset;

\(\mu\) defines the membrane structures including the skin. The labels for the membranes are onetoone with the number of defined regions. For example, \(\left[ {\left[ {} \right]_{2} \left[ {\left[ {} \right]_{4} } \right]_{3} } \right]_{1}\) defines a P system \({{ \Phi}}_{4}\);

\(w_{i}\) defines the initial multiset content of each membrane, as indicated by the respective label. The multiset for each region is defined as the set \(w_{i} \in {{\varSigma }}^{*}\);

\(\left( {R_{i} ,p_{i} } \right)\) refers to the rules and rule priority, where \(0 \le p_{i } \le 1\), for the membrane region i. If there is only one rule for a given membrane region the priority is set to 0. Multiple rules for the same region require that each rule be given a priority;

\(i_{0}\) defines the output region which should be read only when the system halts.
2.2 Multienvironment P system

\({{\varSigma }} = \left( {{\text{slime}}, \ldots } \right)\) is an alphabet used to construct the contents of a multienvironment region. The alphabet contains chemical signals such as slime and lysing antibiotic chemicals;

\({\text{H}}\) is a finite set of labels used to distinguish each multienvironment region. We use the Cartesian ordered pair \(X \times Y = \left\{ {\left( {i,j} \right):i \in X \,\,{\text{and }}\,\,j \in Y} \right\}\) to layout each multienvironment region similar to the coordinates of a twodimensional cellular automata;

\({\text{G}} = \left( {V,S} \right)\) is a graph with nodes \(V = \left( {\left( {1,1} \right), \ldots ,(x,y)} \right)\) represent the multienvironment regions and the edges S define the links between regions. Mycobacteria P systems are able to traverse the multienvironment P system by the link. This is an alternative notation to the P system \(\mu\), which is used to outline a membrane structure;

\(E_{i,j} = \left( {w_{i,j} ,R_{i,j} ,\varphi_{n} } \right)\), for each environment region \(\left( {1,1} \right) \le \left( {i,j} \right) \le \left( {x,y} \right)\), \(E_{i,j}\) is the initial configuration of the region. \(w_{i,j} \in {{\varSigma }}^{*}\) is a multiset of characters over the alphabet \({{\varSigma }}\) and \(R_{i,j}\) defines a set of rules and \(\left\{ {\varphi_{n} \in {{ \Phi}}:1 \le n \le 8} \right\}\) defines a number of bacterial P systems within that multienvironment P system cell, where each bacterial P system can adopt up to one of the eight possible directions according to the underlying Moore neighbourhood arrangement of the multienvironment grid structure;

\(R_{i,j}\) is a finite set of rule for each multienvironment regions. Rule priority is defined on a stochastic basis, using the likelihood that a rule will be executed given certain environmental constraints;

\({{ \Phi}}\) is the inclusion of n Myxobacterial P systems within this particular multienvironment region. A single Myxobacterial P system is denoted by one of eight directions defined by a Moore neighbourhood. The sum of Myxobacteria present in a multienvironment P system is defined by the sum of k (see below) across the eight possible Myxobacterial P systems present at any one time.
2.3 Simulating P system models
The chemical master equation gives the probability of obtaining a state S at time t from an initial set of conditions [25]. Yet, even in short simulations, this form of stochastic modelling can be very computationally expensive [27]. An alternative means of modelling stochastic processes introduced by Gillespie [28] updates the state of a system using probability to determine the order and time for rule execution. The collections of trajectories are updated after the execution of each rule to reflect the change to any of the system variables. In this work, the Gillespie algorithm is modified to calculate the likelihood of each rule being executed in parallel on a constant time step (a unitless value \(\Delta t = 1\)). If a region of the multienvironment P system is not populated with a Myxobacterial P system, the stochastic values for that region are set to zero. The algorithm provides a simple yet elegant means of using environmental properties, such as the volume of a biological space and the probability of chemical collision.
2.4 Myxobacterial P system
The structure of a Myxobacterial P system can be defined as

\({{\varGamma }}\) is an alphabet of multiset elements within the P system region;

\({{\mu }}\) represents the structure of a bacterium P system, that is, \(\left[ {\quad} \right]_{l}\) where \(l \in L\);

\(L\) is the set \(\left\{ {\text{myxo}} \right\}\) of labels used on the membrane;

\(E\) represents the initial configuration of the Myxobacterial P system. Each Myxobacterial P system contains a direction, a means of motility and a bacteria population, defined as n, m and k for the purpose of rule presented in this work.
The configuration of a single Myxobacterial P system is defined by a tuple:

\(N\), where \(1 \le N \le M\), defines the quantity of Myxobacteria in a P system and M denoted the maximum number of Myxobacterial cells;

\(D = \left\{ {{\text{N,}}\;{\text{NE,}}\;{\text{E,}}\;{\text{SE,}}\;{\text{S,}}\;{\text{SW,}}\;{\text{W,}}\;{\text{NW}}} \right\}\) specifies the direction in which a P system of bacteria faces;

\(M = \left\{ {A_{\text{m}} ,\;S_{\text{m}} } \right\}\) specifies the motility engine of the cell.
As with the structure of cellular compartmentalisation of a regular P system, the use of multienvironment model in this study provide cellular space for the inclusion of a Myxobacterial P system and the connection between multienvironment regions allows the diffusion of Myxobacterial P systems between multienvironment regions. For example,
3 Results and discussion
We begin by describing the behaviour of Myxobacteria as a set of multicomponent P system rewrite rules. These rules are then explored further via a demonstration of typical motility, complete Smotility, and complete Amotility from the initial conditions of a bacteria colony. Finally, we see in closer detail the behaviour of Myxobacteria in the presence of slime upon the substrate.
3.1 Stochastic rules
The list of multicomponent P system rules
Rule ID  Rule  Description 

Social motility  
r _{1}  \(\left[ {\left[ {n,{\text{E}},S_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j} \left[ {\quad} \right]_{i + 1,j} \mathop \to \limits^{c1}\left[ {\quad} \right]_{i,j}^{\prime }  \left[ {\left[ {n,{\text{E}},S_{\text{m}} } \right]_{\text{myxo}} } \right]_{i + 1,j}^{\prime }\)  Smotility 
Adventurous motility  
r _{2}  \(\left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j} \left[ {\quad} \right]_{i + 1,j} \mathop \to \limits^{c2} \left[ {\text{slime}} \right]_{i,j}^{\prime }  \left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i + 1,j}^{\prime }\)  Amotility leave slime behind 
Adventurous motility into slime  
r _{3a}  \(\left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {\text{slime}} \right]_{i + 1,j} \mathop \to \limits^{c3a} \left[ {\text{slime}} \right]_{i,j}^{\prime }  \left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i + 1,j}^{\prime }\)  Amotility into slime 
r _{3b}  \(\left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {\text{slime}} \right]_{i + 1,j  1} \mathop \to \limits^{c3b} \left[ {\text{slime}} \right]_{i,j}^{\prime }  \left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i + 1,j  1}^{\prime }\)  Amotility into slime 
r _{3c}  \(\left[ {\left[ {n,E,Am} \right]_{myxo} } \right]_{i,j}  \left[ {slime} \right]_{i + 1,j + 1} \mathop \to \limits^{c3c} \left[ {slime} \right]_{i,j}^{'}  \left[ {\left[ {n,E,Am} \right]_{myxo} } \right]_{i + 1,j + 1}^{'}\)  Amotility into slime 
r _{4a}  \(\left[ {\left[ {n,{\text{NE}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {\text{slime}} \right]_{i,j  1} \mathop \to \limits^{c4a} \left[ {\text{slime}} \right]_{i,j}^{\prime }  \left[ {\left[ {n,{\text{NE}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j  1}^{\prime }\)  Amotility into and with slime 
r _{4b}  \(\left[ {\left[ {n,{\text{NE}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {\text{slime}} \right]_{i + 1,j  1} \mathop \to \limits^{c4b} \left[ {\text{slime}} \right]_{i,j}^{\prime }  \left[ {\left[ {n,{\text{NE}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i + 1,j  1}^{\prime }\)  Amotility into and with slime 
r _{4c}  \(\left[ {\left[ {n,{\text{NE}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {\text{slime}} \right]_{i + 1,j} \mathop \to \limits^{c4c} \left[ {\text{slime}} \right]_{i,j}^{\prime }  \left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i + 1,j}^{\prime }\)  Amotility into and with slime 
Bacteria mixing  
r _{5}  \(\left[ {\left[ {n,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {\left[ {m,{\text{W}},  } \right]_{\text{myxo}} } \right]_{i + 1,j} \mathop \to \limits^{c5} \left[ {\left[ {\left( {n  k} \right),{\text{E,AM}}} \right]^{\prime }_{\text{myxo}} } \right]_{i,j}^{\prime }  \left[ {\left[ {k,{\text{E}},A_{\text{m}} } \right]_{\text{myxo}} \left[ {m,{\text{W}},  } \right]_{\text{myxo}}^{\prime } } \right]_{i + 1,j}^{\prime }\)  Amotility mixing into neighbour 
r _{6a}  \(\left[ {\left[ {n,{\text{E}},S_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {m,{\text{E}},S_{\text{m}} } \right]_{i + 1,j} \mathop \to \limits^{c6a} \left[ {\left[ {\left( {n  k} \right),{\text{E}},S_{\text{m}} } \right]_{\text{myxo}}^{\prime } } \right]_{i,j}^{\prime }  \left[ {\left[ {\left( {m + k} \right),{\text{E}},S_{\text{m}} } \right]_{\text{myxo}}^{\prime } } \right]_{i + 1,j}^{\prime }\)  Smotility into neighbour 
r _{6b}  \(\left[ {\left[ {n,{\text{E}},S_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {m,{\text{NE}},S_{\text{m}} } \right]_{i + 1,j  1} \mathop \to \limits^{c6b} \left[ {\left[ {\left( {n  k} \right),{\text{SE}},S_{\text{m}} } \right]_{\text{myxo}}^{\prime } } \right]_{i,j}^{\prime }  \left[ {\left[ {\left( {m + k} \right),{\text{NE,}}} \right]_{\text{myxo}}^{\prime } } \right]_{i + 1,j  1}^{\prime }\)  Smotility into neighbour 
r _{6c}  \(\left[ {\left[ {n,{\text{E}},S_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j}  \left[ {m,{\text{NE}},S_{\text{m}} } \right]_{i + 1,j + 1} \mathop \to \limits^{c6c} \left[ {\left[ {\left( {n  k} \right),{\text{SE}},S_{\text{m}} } \right]_{myxo}^{\prime } } \right]_{i,j}^{\prime }  \left[ {\left[ {\left( {m + k} \right),{\text{NE}},  } \right]_{\text{myxo}}^{\prime } } \right]_{i + 1,j + 1}^{\prime }\)  Smotility into neighbour 
Direction and motility switching  
r _{7}  \(\left[ {[n,{\text{E}},  ]_{\text{myxo}} } \right]_{i,j}  \left[ {[m,{\text{W}},  ]_{\text{myxo}} } \right]_{i + 1,j} \mathop \to \limits^{c7} \left[ {[n,{\text{W}},  ]_{\text{myxo}}^{\prime } } \right]_{i,j}  \left[ {\left[ {m,{\text{E}},  } \right]_{\text{myxo}}^{\prime } } \right]_{i + 1,j}\)  Csignal reversal 
r _{8}  \(\left[ {\left[ {n,  ,A_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j} \mathop \to \limits^{c8} \left[ {\left[ {n,  ,S_{\text{m}} } \right]_{\text{myxo}}^{\prime } } \right]_{i,j}\)  Amotility to Smotility switch 
r _{9}  \(\left[ {\left[ {n,  ,S_{\text{m}} } \right]_{\text{myxo}} } \right]_{i,j} \mathop \to \limits^{c9} \left[ {\left[ {n,  ,A_{\text{m}} } \right]_{\text{myxo}}^{\prime } } \right]_{i,j}\)  Smotility to Amotility switch 
r _{10}  \(\left[ {\left[ {n,{\text{E}},  } \right]_{\text{myxo}} } \right]_{i,j} \mathop \to \limits^{c10} \left[ {\left[ {n,{\text{W}},  } \right]_{\text{myxo}}^{\prime } } \right]_{i,j}\)  Spontaneous reversal 
The movement of the Myxobacteria is presented in rules \(r_{1}\) through to \(r_{6}\): \(r_{1}\) uses social motility to move a population of Myxobacteria into an empty neighbouring cell; \(r_{2}\) applies to Myxobacteria using adventurous motility to propel itself forward into an empty neighbour whilst depositing slime; \(r_{{ 3 {\text{a}}}}\), \(r_{{ 3 {\text{b}}}}\) and \(r_{{ 3 {\text{c}}}}\) allows the population of Myxobacteria to detect slime perpendicular to its long axis, followed by movement into the slime whilst itself depositing slime; \(r_{{ 4 {\text{a}}}}\), \(r_{{ 4 {\text{b}}}}\) and \(r_{{ 4 {\text{c}}}}\) allows slime detection on an angle; and \(r_{5}\) allows the mixing of a population of Myxobacteria from one neighbouring cell into another, in either motility modes.
The movement of Myxobacteria from one region into another populated region is determined by social motility movement and cell number mixing, perpendicular or on an angle to the moving P system in question, is controlled using \(r_{{ 6 {\text{a}}}}\), \(r_{{ 6 {\text{b}}}}\) and \(r_{{ 6 {\text{c}}}}\).
Collision between populations of Myxobacteria can result in a switch in motility mechanism, from adventurous into social as per Csignal induced reversal (\(r_{7}\)).
As the quantity of adventurous Myxobacteria occupy a single region increases there is an evergreater likelihood of becoming socially mobile, and therefore joining any social Myxobacteria within that region (as per \(r_{8}\)). The probability of switching motility takes on a rate similar in form to a ligand associating with a substrate as per the Michaelis–Menten equation, and in which a surplus of Myxobacterial asymptotes the probability of switching. On the other hand, if the region is sparely populated with social Myxobacteria, they can switch back into adventurous motility, as seen in \(r_{9}\). Myxobacterium can undergo spontaneous reversal via a switch to the internal cellular motility mechanism (\(r_{10}\)). This causes the Myxobacteria to spend a lot of its solitary, swarming and aggregating life moving along its long axis.
3.2 Simulation examples
Default values for the stochastic selection of rules. Stochastic values are retrieved for valid rules (cellular conditions are correct), arranged descending by size and transformed onto the range (0, 1) for random selection
Rule identifier  Stochastic value  Rule identifier  Stochastic value 

r _{1}  0.05  r _{5}  0.8 
r _{2}  5  r _{6a}  0.05 
r _{ 3a}  5  r _{6b}  0.05 
r _{3b}  5  r _{ 6c}  0.05 
r _{3c}  5  r _{7}  0.05 
r _{4a}  5  r _{8}  0.05 
r _{4b}  5  r _{9}  0.02 
r _{4c}  5  r _{10}  0.05 
3.2.1 Simulation examples of a Myxobacterial colony
During the early simulation steps of a Myxobacteria colony capable of social and adventurous exploration (Fig. 2), the Myxobacterial P system populations at the colony periphery orientated out towards the empty adjacent multicomponent P system regions. Those Myxobacterial P systems within the heart of the colony would have potentially undergone a number of Csignal reversals from headtohead collisions with immediate neighbours (\(r_{7}\)), spontaneous reversals (\(r_{10}\)) or exchange of Myxobacterial subpopulations between neighbouring Myxobacterial P systems (\(r_{{ 6 {\text{a}}}}\), \(r_{{ 6 {\text{b}}}}\) and \(r_{{ 6 {\text{c}}}}\)). Driven by the adventurous motility (\(r_{2}\)) of the periphery Myxobacterial P systems, immediate neighbours were able to detect local deposits of slime encouraging movement away from the colony. As the initial colony structure reduces, some of those adventurous and freeexploring Myxobacteria switched to Smotility (\(r_{9}\)), enabling neighbouring Smotility cells to begin mixing and moving amongst themselves (\(r_{{ 6 {\text{a}}}}\), \(r_{{ 6 {\text{b}}}}\) and \(r_{{ 6 {\text{c}}}}\)). Before too long, the periphery Myxobacterial P systems having broken away from the main colony to setup smaller colonies can switch back into Amotility (\(r_{8}\)) and once again venture out to form additional colonies.
By being able to turn towards the direction of slime deposits (\(r_{6a}\), \(r_{{ 6 {\text{b}}}}\) and \(r_{{ 6 {\text{c}}}}\)), switching direction randomly (\(r_{10}\)), and interacting with neighbouring adventurous P systems (\(r_{5}\)), there is a high degree of variability compared to the initial simulation parameters and an increased ability to form further subcolonies. It is interesting to note, by decreasing the probability of a population switching from Amotility to Smotility, the colony immediately breaks up and begins randomly exploring the environment (data not shown). This is seen when the genes responsible for the Myxobacteria’s Smotility are knocked out, preventing the Myxobacteria from gripping onto neighbouring Myxobacteria [29].
A simulation of a Myxobacterial colony using only the Amotility (rules \(r_{2}\), \(r_{{ 3 {\text{a}}}} ,\; r_{{ 3 {\text{b}}}} ,\; r_{{ 3 {\text{c}}}}\), \(r_{{ 4 {\text{a}}}} ,\; r_{{ 4 {\text{b}}}} , \;r_{{ 4 {\text{c}}}}\)) rules for movement and removing the ability for the P systems to switch motility was performed for 250 time steps (Fig. 3). Compared to the normalmotility simulation, during the initial simulation steps the centre of the colony remains intact whilst showing only minor indication of the periphery Myxobacterial P system orientating towards the empty substrate to forward (\(r_{2}\)) into the empty substrate. Without Smotility, the Myxobacterial P systems are unable to maintain a bacterial aggregate, rather they move out into explore the empty substrate (\(r_{2}\)) or follow the slime deposits of Myxobacterial P system directly in front (\(r_{{ 3 {\text{a}}}}\)), or at an angle (\(r_{{ 3 {\text{b}}}}\), \(r_{{ 3 {\text{c}}}}\)). The eight orientations available can be clearly seen from step 100 to step 250, with a less dense covering of Myxobacteria as populations detected neighbouring slime on the angle off from their long axis.
Similarly, as with Amotility, when the movement of the population has been reduced to only Smotility (\(r_{1}\), \(r_{{ 6 {\text{a}}}}\), \(r_{{ 6 {\text{b}}}}\), and \(r_{{ 6 {\text{c}}}}\)) (Fig. 4), the eight possible orientations can also be discerned across the Myxobacterial P system population. Without adventurous motility, aggregates of P system Myxobacteria migrate along these orientations with little in the way of breakaway Myxobacteria, where each colony having formed due to collisions (\(r_{7}\)) and spontaneous reversal (\(r_{10}\)).
4 Conclusion
Using a multicomponent P system substrate and a model P system to represent dynamic quantities of Myxobacteria, this work presented simulations of a swarming and aggregating Myxobacterial population. The results demonstrate how both modes of bacterial motility, social and adventurous were necessary to replicate both swarming of bacteria from the colony edge, exploration of the empty substrate and intermittent subcolony formation. A simulation using only social behaviour explored the substrate by forming uniformed subcolonies, yet with no means of acting as solitary agents, the distribution of populations was along the eight orientations. Whereas, adventurous motility only simulations results in Myxobacterial P systems at the colony periphery venturing into the empty substrate and in turn leads to neighbouring P systems breaking from the colony. Yet, with no mixing of P system populations (from social behaviour), had the simulation been left indefinitely, it is quite possible that the complete population would have been distributed equally over the substrate space.
With all languages being Turing equivalent, it is possible to implement a P System model using any language as mathematically, all programming languages are Turing equivalent. However, Java was selected given its native support for encapsulation and inheritance. Objectorientated languages such as Java and C++ lend themselves very well to the paradigm of membrane computing. Although of interest are the efforts to encapsulate P system rules in a formal language [31], and very recently using in vitro techniques to produce a real P system [32].
Despite the swarming and aggregating dynamics observed using only a small set of P system rewrite rules, there are a number of limitations; orientation is limited to the eight adjacent spaces in the multienvironment P system substrate; the time step is a simple count and the execution of rules are determined stochastically with no consideration of temporal dynamics; and, slime remains indefinitely, reducing the Amotility to a random walk as the simulation progresses. Compared with recent particle continuum [33] and continuous [34] based modelling, conventional models of Myxobacteria continue to dominate, chiefly through their ability to model structural properties of bacterium. Finally, there is a degree of abstraction from the traditional cellline P system paradigm in the way in which the mixing of bacteria is treated. Traditionally, rewrite rules and the diffusion of cell content would be required to combine Myxobacterial populations. However, given the framework presented in this study, further exploration into locally induced global phenomena of bacterial colonies using P systems is entirely possible.
Notes
References
 1.Shimkets, L. J., & Kaiser, D. (1982). Induction of coordinated movement of Myxococcus xanthus cells. Journal of Bacteriology, 152, 45–461.Google Scholar
 2.Kaiser, D., & Crosby, C. (2005). Cell movement and its coordination in swarms of Myxococcus xanthus. Cytoskeleton, 3, 227–245.Google Scholar
 3.Igoshin, O., Goldbeter, A., Kaiser, D., & Oster, G. (2004). A biochemical oscillator explains several aspects of Myxococcus xanthus behavior during development. Proceedings of the National Academy of Sciences USA, 101, 15760–15765.CrossRefGoogle Scholar
 4.Sliusarenko, O., Neu, J., Zusman, D. R., & Oster, G. (2006). Accordion waves in Myxococcus xanthus. Proceedings of the National Academy of Sciences USA, 103, 1534–1539.CrossRefGoogle Scholar
 5.Wu, Y., Chen, N., Rissler, M., Jiang, Y., Kaiser, D., & Alber, M. (2006). CA models of Myxobacteria swarming. In S. E. Yacoubi, B. Chopard, & S. Bandini (Eds.), Cellular automata. ACRI 2006. Lecture notes in computer science (pp. 192–203). Berlin, Heidelberg: Springer.Google Scholar
 6.Welch, R., & Kaiser, D. (2001). Cell behavior in traveling wave patterns of Myxobacteria. Proceedings of the National Academy of Sciences USA, 98, 14907–14912.CrossRefGoogle Scholar
 7.Wolgemuth, C., Hoiczyk, E., Kaiser, D., & Oster, G. (2002). How myxobacteria glide. Current Biology, 12, 369–377.CrossRefGoogle Scholar
 8.Berleman, J. E., Chumley, T., Cheung, P., & Kirby, J. R. (2006). Rippling is a predatory behavior in Myxococcus xanthus. Journal of Bacteriology, 188, 5888–5895.CrossRefGoogle Scholar
 9.Mauriello, E. M., Yang, T., & Zusman, D. R. (2010). Gliding motility revisited: How do the Myxobacteria move without flagella? Microbiology and Molecular Biology Reviews, 74, 229–249.CrossRefGoogle Scholar
 10.MuñozDorado, J., MarcosTorres, F. J., GarcíaBravo, E., MoraledaMuñoz, A., & Pérez, J. (2016). Myxobacteria: Moving, killing, feeding, and surviving together. Frontiers in Microbiology, 7, 781.CrossRefGoogle Scholar
 11.Mignot, T., Merlie, J. P., Jr., & Zusman, D. R. (2005). Regulated poletopole oscillations of a bacterial gliding motility protein. Science, 310, 855–857.CrossRefGoogle Scholar
 12.Păun, G. (2000). Computing with membranes. Journal of Computer and System Sciences, 61, 108–143.MathSciNetCrossRefzbMATHGoogle Scholar
 13.Pu, Y., Yu, Y., Dong, X. (2006). Simulation of biomolecular processes by using stochastic p systems. In: Proc Workshop on High Performance Computing in the Life Sciences. Ouro Preto.Google Scholar
 14.Sburlan, D. (2006). Noncooperative P systems with Priorities Characterize PsET0L. In: Freund R, Lojka G, Oswald M, Păun G, editors. PreProc. of sixth international workshop on membrane computing (WMC6), Vienna, pp. 530–539.Google Scholar
 15.Cámara, M. (2006). Quorum sensing: A cellcell signalling mechanism used to coordinate behavioral changes in bacterial populations. In H. J. Hoogeboom, G. Păun, G. Rozenberg, & A. Salomaa (Eds.), Membrane Computing. WMC 2006. Lecture notes in computer science (pp. 42–48). Berlin, Heidelberg: Springer.Google Scholar
 16.Bianco, L., Pescini, D., Siepmann, P., Krasnogor, N., RomeroCampero, F. J., & Gheorghe, M. (2006). Towards a P systems Pseudomonas quorum sensing model. In H. J. Hoogeboom, G. Păun, G. Rozenberg, & A. Salomaa (Eds.), Membrane computing. WMC 2006. Lecture notes in computer science (pp. 197–214). Berlin, Heidelberg: Springer.Google Scholar
 17.Bernardini, F., Gheorghe, M., & Krasnogor, N. (2007). Quorum sensing P systems. Theoretical Computer Science, 371, 20–33.MathSciNetCrossRefzbMATHGoogle Scholar
 18.Terrazas, G., Krasnogor, N., Gheorghe, M., Bernardini, F., Diggle, S., & Cámara, M. (2005). An environment aware Psystem model of quorum sensing. In S. B. Cooper, B. Löwe, & L. Torenvliet (Eds.), New Computational Paradigms. CiE 2005. Lecture Notes in Computer Science (pp. 479–485). Berlin, Heidelberg: Springer.Google Scholar
 19.RomeroCampero, F. J., & PérezJiménez, M. J. (2008). A model of the quorum sensing system in Vibrio fischeri using P systems. Artificial Life, 14, 95–109.CrossRefGoogle Scholar
 20.Frisco, P., Corne, D.W. (2007). Modeling the dynamics of HIV infection with conformonp systems and cellular automata. In: Eleftherakis G, Kefalas P, Păun G, Rozenberg G, Salomaa A, editors. Membrane Computing. WMC 2007. Lecture notes in computer science, Berlin, Heidelberg. pp. 21–31.Google Scholar
 21.Cardona, M., Colomer, M. A., Margalida, A., PérezHurtado, I., PérezJiménez, M. J., & Sanuy, D. (2009). A P system based model of an ecosystem of some scavenger birds. In G. Păun, M. J. PérezJiménez, A. RiscosNúñez, G. Rozenberg, & A. Salomaa (Eds.), Membrane computing. WMC 2009. Lecture notes in computer science (pp. 182–195). Berlin, Heidelberg: Springer.Google Scholar
 22.Cardona, M., Colomer, M. A., PérezJiménez, M. J., Sanuy, D., & Margalida, A. (2008). Modeling ecosystems using p systems: The bearded vulture, a case study. In D. W. Corne, P. Frisco, G. Păun, G. Rozenberg, & A. Salomaa (Eds.), Membrane Computing. WMC 2008. Lecture Notes in Computer Science (pp. 137–156). Berlin, Heidelberg: Springer.Google Scholar
 23.Colomer, M. A., PérezHurtado, I., PérezJiménez, M. J., & RiscosNúñez, A. (2011). Comparing simulation algorithms for multienvironment probabilistic P systems over a standard virtual ecosystem. Natural Computing, 11, 369–379.MathSciNetCrossRefzbMATHGoogle Scholar
 24.GarcíaQuismondo, M., & MartínezdelAmor, M. A. (2014). Probabilistic guarded P systems, a new formal modelling framework. International Conference on Membrane Computing., 2014, 193–214.zbMATHGoogle Scholar
 25.Gillespie, D. T. (1992). A rigorous derivation of the chemical master equation. Physica A: Statistical Mechanics and its Applications, 188, 404–425.CrossRefGoogle Scholar
 26.Păun, G., Rozenberg, G., Salomaa, A., editors. (2010). The oxford handbook of membrane computing, Oxford University Press, Oxford.Google Scholar
 27.Lu, T., Volfson, D., Tsimring, L., & Hasty, J. (2004). Cellular growth and division in the Gillespie algorithm. Systems Biology, 1, 121–128.CrossRefGoogle Scholar
 28.Gillespie, D. T. (1997). Exact stochastic simulation of coupled chemical reactions. The Journal of Physical Chemistry B, 81, 2340–2361.CrossRefGoogle Scholar
 29.Shimkets, L. J. (1990). Social and developmental biology of myxobacteria. Microbiological Reviews, 54, 473–501.Google Scholar
 30.Diggle, P. J. (2003). Statistical analysis of spatial point patterns. New York: Academic Press.zbMATHGoogle Scholar
 31.DíazPernil, D., PérezHurtado, I., PérezJiménez, M.J., RiscosNúñez, A. (2008). A Plingua programming environment for membrane computing. In: D. W. Corne, P. Frisco, G. Păun, G. Rozenberg, A. Salomaa (Eds.), Membrane computing. WMC 2008. Lecture notes in computer science, vol. 5391. Berlin, Heidelberg, pp. 187–203.Google Scholar
 32.Mayne, R., Phillips, N., Adamatzky, A. (2018). Towards experimental Psystems using multivesicular liposomes. arXiv, 2018;1812.05476.Google Scholar
 33.Degond, P., Manhart, A., & Yu, H. (2018). An agestructured continuum model for myxobacteria. Mathematical Models and Methods in Applied Sciences, 29, 1737–1770.MathSciNetCrossRefzbMATHGoogle Scholar
 34.Hendrata, M., Yang, Z., Lux, R., & Shi, W. (2011). Experimentally guided computational model discovers important elements for social behavior in myxobacteria. PLoS One., 2011, 6.Google Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.