Nisin Modelling

Mathematical modelling, docking and molecular dynamics of nisin


The production of nisin by E. coli has not yet been thoroughly studied and characterized. To put it in context of our application when grown in a co-culture, this increases the complexity of the system. It is for this reason that we modelled the production of nisin, in order to have an understanding of the production of Nisn and how it would be affected when put under the constraints of a co-culture.

Also, for the ideal use of Cellucoat, nisin would have to come in contact with BC and PHB. To characterize and study the interactions and behaviour of nisin, PHB and BC we conducted docking simulations for revealed binding dynamics of these molecules. Molecular dynamics simulations were also conducted and analyzed to study the conformational changes nisin undergoes when put under environmental conditions during productions and use of Cellucoat.

Nisin Production

To model nisin, we followed the same approach used for the co-culture model and that includes deriving a system of differential equations that describe the dynamics of what goes on in the system. The equations used in nisin production modelling were obtained from Calleja et. al 2016 that modeled protein production in fed-batch E. coli cultures (1).This was an ideal model to use as it implemented our chassis, E. coli, and also considered the fed-batch strategy which is synonymous to the intermittent feeding strategy used in wet lab experiments.

The Calleja model has three components which includes a biomass growth model which consists of the biomass concentration and the total volume. The dynamics of the inducer (IPTG) uptake model was based on the biomass growth model. Lastly, the protein production model is a function of the results of the changes in the concentration of the inducer IPTG and its workings with the repressor (1).

Figure 1. Blocks and Variables of the Model by Calleja et. al (1)

The model takes into account two timestamps; before induction and after induction. During the non-induced stage, only the biomass model is taken into account while after induction, the other models come into play with modifications on the biomass model to account for induction. However, we made modifications to the model to be more reflective of our experimental conditions. One of such is the change in the volume equation. The model takes into account a different feeding flow rate equation which the volume is dependent on. So we changed that to reflect our periodic feeding. This was then reflected in all equations that utilize the feeding flow rate. Hence, we present the following equations that describe our nisin production for a monoculture of E. coli:

We then used the solve_ivp function in the SciPy package of Python to solve the system of equations in order to plot and help visualize the changes in the modelled parameters. The python files for protein production in a monoculture and co-culture can be found here here . These resulted in the following simulation plots:

Figure 2. Change in E. coli biomass during the simulation run

Figure 3. Change in Substrate Consumption during the simulation run

Figure 4. Protein production during the simulation run

Non-induced period of the simulation is shown in blue and induced periods of the simulaiton is shown in red.

Applying the Model to the Co-culture

Although initially produced in a monoculture, E. coli would be in a co-culture with K. xylinus. Therefore, it was important that we understood how this could affect the production of nisin. So we added changes to the model for a co-culture and plotted the model results. This resulted in the following:

Figure 5. Change in Substrate Consumption forthe Co-culture during the simulation run

Figure 6. Change in Bacterial Cellulose in a Co-culture (note that bacterial cellulose amount is used to quantify the amount of K. xylinus)

Figure 7. Change in E. coli biomass in a co-culture

Figure 8. Change in Acetic acid production by K. xylinus in a co-culture

Figure 9. Change in Gluconic acid production by K. xylinus in a co-culture

Figure 10. Protein production in the co-culture during the simulation run

Non-induced period of the simulation is shown in blue and induced periods of the simulaiton is shown in red.

Comparing both the monoculture and the simulation runs to the co-culture results, an evident difference is the decrease in protein production in the co-culture. This can be attributed to the introduction of K. xylinus in the culture. Since K. xylinus utilizes the same substrate as E. coli, it reduces the amount available to E. coli to utilize in the production of nisin. However, this change is not too significantly different as at the 30 hr mark the production unit in th E. coli monoculture is 12000mg protein / L while that for the co-culture is 10000mg protein / L, which is a 16.7% difference. However, when scaled up, this difference can be more. So in order to ensure that nisin is optimally produced and incorporated into our BC we would need to ensure that sufficient substrate and nutrients are available for the co-culture.

Another result from this model is the difference from the Co-culture model. In the co-culture model the simulation resulted in the substrate being consumed at a faster rate. Thus, it resulted in a shorter period of growth for both K. xylinus and E. coli and reduction in BC production. However, the model showed BC production for a longer period even though it is the same amount of substrate concentration used in both models. In the wet lab the reality is that if K. xylinus does grow for a longer period than established by the co-culture model, then this model is more representative of the actual experiments. Hence, utilizing this in place of the co-culture model in future iterations would provide more accurate predictions helping to understand the dynamics of the co-culture.

Molecular Docking

Docking is a bioinformatics modeling method that can be used to give predictions to the structure of a ligand binding to a protein. This is mainly done by going through possible conformations of the ligand-receptor complex and ranking this based on some scoring function. When implemented in Cellucoat, nisin comes into contact with bacterial cellulose. Since there is no immobilizing agent that attaches nisin onto bacterial cellulose, the placement of nisin onto BC is solely dependent on the present interactions they have between one another. To look into these interactions and confirm that they are suitable we performed docking simulations.

Nisin is not the only incorporated molecule in our bacterial cellulose. For strength purposes we are including Polyhydroxybutyrate (PHB) in our material. Due to the fact that PHB and nisin are in the same environment they would have interactions with one another. Hence, we also conducted docking on these two to have an understanding as to how they might bind to one another and how this binding could have an effect on nisin’s functioning or its binding to BC.


There exists a variety of software which utilize several algorithms, scoring functions and even perform several types of docking ranging from protein-protein, to protein-ligand. In our case, our two docking simulations were with a ligand which was BC or PHB and protein which is nisin, while understanding that BC and PHB are not ligands but have larger sizes compared to nisin. However, to utilize the software we made the assumption that these molecules were small in size. Hence, we made use of a small chain of both polymers and nisin in our docking simulations.

The software implemented was a web server CB-DOCK 2 (2) that is used for protein-ligand docking. Given the three-dimensional structure of a protein and ligand, it integrates cavity detection, a docking algorithm and homologous template fitting to predict the sites and affinity that a ligand has to a protein (2).


After conducting docking simulations, the software gave rise into five possible binding conformations with varying Vina scores, which is the metric used in measuring binding affinity. The score works in such a way that the more negative it is the less energy is needed to initiate binding and the better the binding affinity. So we took into account the most negative Vina scores and based our conclusions on it.

Protein + Ligand Vina Score Cavity Volume Center Docking Size
nisin + BC -5.4 83 13, 16, -12 33, 33, 33
nisin + PHB -4.1 83 13, 16, -12 27, 27, 27

Table 1. Docking Results of nisin and BC, and nisin and PHB

Figure 11. 3D Visualization of nisin binding unto a chain of BC (Contact residues of nisin to BC shown in red and other parts in green. BC chain shown in red and white)

Figure 12. 3D Visualization of nisin binding unto a chain of PHB (Contact residues of nisin to PHB shown in red and other parts in green. PHB chains shown in yellow and red)

After conducting docking, we analysed the contact residues of PHB and BC with nisin. We discovered that the best binding conformation for both is at the same amino acid residues. This implies that during PHB’s and nisin’s integration to BC, PHB and BC would tend to cling to the same or similar contact point on Nsin which could have a potential impact on its immobilization unto BC. However, BC (-5.4) has a lower vina score than PHB (-4.1) as seen through the docking simulations meaning that it has a higher binding affinity. So empirically nisin would favourably bind more to BC than PHB. Thus, allowing for immobilization of Nisn into BC.

Informing the Project

The wetlab team’s work targeted making and developing experiments with nisin, PHB, and BC that were based on macroscale outcomes. Through molecular dynamics and docking simulations, we can optimize the interactions occurring in different systems and understand what is occurring on a nanoscale. Docking allowed us to predict how BC, nisin, and PHB would interact as a single unit, whether they would bind, how they would bind, what affinity they would bind with, and if nisin’s antimicrobial binding sites would be covered.

A significant result from molecular dynamics showed that the nisin-PHB-BC binding sites do not hinder nisin’s antimicrobial capacity (3). In nisin’s antimicrobial mechanism, when nisin reaches the cell membrane surface, rings A and B at the N-terminus of nisin bind to the pyrophosphate group of lipid II (3). The C-terminal region inserts into the membrane to form a pore using a flexible mechanism. The aforementioned regions and rings do not overlap with the binding of PHB and BC (3).

Molecular Dynamics

When it came to the development of Cellucoat, one important factor that we had to ensure was that the conformation of nisin was not affected in any way that would affect its function from the processes of production to use. Looking at the production-to-use pipeline, there existed several varying environmental conditions and we hypothesized it could have an effect on nisin. nisin’s functionality under these conditions could not be validated experimentally due to the time and resource constraints. So we turned over to conducting molecular dynamics simulations to validate this.

Molecular dynamics is a modelling method, often computational, that involves simulating the brownian motion of atoms in a protein or molecule over a period of time. By stimulating their motion, it can enable one to understand the thermo/pH stability of the system, look at system instabilities and also identify conformational changes the molecule undergoes under certain conditions. Hence, we used Molecular dynamics to understand nisin’s dynamics under several conditions. We conducted these simulations using GROMACS on a supercomputer provided by our University (4). The process of conducting the simulation involves feeding into the software a “.pdb” file that contains the atomic structure of the protein/molecule. Then simulation parameters are specified in other files some of which includes temperature, pressure, simulation time. These where the parameters we modified to simulate our desired conditions.

One of such extreme conditions we considered was nisin being placed under extreme temperature and pressure conditions of the autoclave or other sterilizing machines after the production of BC so as to kill off any bacteria present and lyse open E. coli to release nisin as found on the Co-culture page. On the other end of the spectrum, we understand that when fruits are in use and in storage, they are often placed in cold conditions in fridges and refrigerators. So we used molecular dynamics to try and simulate these conditions and understand the stability, conformational changes of nisin in these conditions.


Before conducting molecular dynamics on any extreme condition, we needed a control to represent optimal and normal conditions. To do so, we performed a molecular dynamics simulation on just nisin by setting the simulation temperature and pressure to match room temperature and atmospheric pressure conditions.

When it comes to simulating autoclave conditions, these environments mimic sterilization by implementing high temperature and pressure conditions. So we changed the .mdp files in the simulation run which contains parameters that control the pressure and temperature of the simulation. The temperature and pressure was changed to 121 degrees Celsius and 15 psi respectively which is representative of the standard autoclave used in the lab (5). The same approach was applied to that of the fridge conditions. We used a temperature of 0 degrees Celsius which was the average for storing fruit conditions (6). For pressure, we realized several refrigerants can lead to several pressure conditions. So for our simulation, we assumed the use of a R134 refrigerant which is a common refrigerant implemented (7). This resulted in 28psi at a temperature of 0 degrees Celsius (8).

Simulation Simulation Temperature Simulation Pressure
Normal Conditions (control) 27 ° C or 300K 14.5 psi or 1 bar
Autoclave Conditions 0 ° or 273K 15 psi or 1.03421 bar
Fridge Conditions 0 ° or 273K 28 psi or 1.93053 bar

Table 2. Temperature and Pressure values of the several molecular dynamics simulations on nisin


Figure 13. 3D Visualization of the molecular dynamics simulation of nisin under normal conditions

Figure 14. 3D Visualization of the molecular dynamics simulation of nisin under autoclave conditions

Figure 15. 3D Visualization of the molecular dynamics simulation of nisin under fridge conditions

After conducting the simulations we wanted to analyse how the simulation condition affects nisin’s conformation. So to do this we expressed the change in conformation terms of how the distance between key amino acids are changing over the period of the simulation. Two amino acids we studied were cysteine 7 and serine 33 which are the first and last amino acid residues in nisin that are in contact with BC as resulted from the docking simulation conducted.

So we used a software called VMD to generate the distance between the two amino acids over the frames of the simulation. These were extracted as a file of the distance over frames and also a histogram of the distance.

The results are as follows:

Figure 16. Distance between Cysteine 7 and Serine 33 residues in nisin under normal conditions

Figure 17. Distance between Cysteine 7 and Serine 33 residues in nisin under fridge conditions

Figure 18. Distance between Cysteine 7 and Serine 33 residues in nisin under fridge conditions

Figure 19. Distance distribution of nisin under normal, fridge and autoclave conditions between Cysteine 7 and Serine 33 residues

Again it was also important to validate that nisin is still able to bind unto BC so that it can be adequately integrated in Cellucoat. As we had done before, we conducted docking simulations with the nisin structure that had undergone the molecular dynamics simulation. We removed all the water molecules and ions and got the protein structure of nisin and conducted docking with it to BC. These resulted in the following:

Simulation Condition Vina Score Cavity Volume Center Docking Size
Autoclave (nisin + BC) -5.3 76 40, 40, 8 33, 33, 33
Fridge (nisin + BC) -4.9 42 30, 25, 15 33, 33, 33

Table 3. Docking Results of nisin and BC under autoclave and fridge conditions

From the docking simulations, the reductions in binding efficacy of nisin and BC were below 5%. Thus, we can confirm through the model that nisin would still be able to bind unto BC and still have similar conformations to itself when under normal conditions while in the conditions of the autoclave or fridge.

Changing pH and its effect on nisin

Another environmental condition that had the potential to affect the functioning of nisin was the potential acidic conditions of fruits that it comes in contact with. Since Cellocoat is designed to be a replacement for clamshell packaging, we realized that the most commonly packaged fruits in clamshells are berries and grapes. We found that these could have an average pH of 3.46 (9). So we used PROTEINPREPARE from the PlayMolecule website to change the pronation state of nisin, giving it a pH of 3.46 and then conducted molecular docking simulation on it.

SProtein + Ligand Vina Score Cavity Volume Center Docking Size
Autoclave (nisin + BC) -5.2 76 9, 22, -4 33, 33, 33

Table 4. Docking Results of protonated nisin and BC to simulate low pH conditions of fruits

From the docking simulations, the reductions in binding efficacy of nisin and BC were below 5%. Thus, we can confirm through the model that nisin would still be able to bind unto BC while under the pH conditions of the fruits that it would be used for in Cellocoat.

The protein structure files, docking complexes, and molecular dynamics output files of nisin, PHB and BC can be found here

Modelling Collaboration

This year we collaborated with iGEM 2 IISER Pune inorder to have a better understanding on molecular dynamics work and share molecular docking strategies used. More information on the work we conducted for them can be found on the collaborations page. However, when it came to conducting docking simulations, iGEM 2 IISER Pune had utilised several docking software to confirm their binding dynamics which they were trying to understand. So we wanted to implement this. However, what we realised was that there was a limitation in variety of web servers we could use for our docking, so we asked if they could help us utilize another software for docking - Auto dock Vina. After running protein-ligand docking for us on nisin and BC and nisin and PHB this resulted in the following:

Figure 20. Docking Results of nisin and BC from iGEM 2 IISER Pune

Figure 21. Docking Results of nisin and PHB from iGEM 2 IISER Pune

From the docking results we took into account the modes that had the lowest scores as this ranks as the best affinity. Although implementing different algorithms in docking, both our software CB-Dock 2 and Auto Dock Vina utilize the same scoring functions we could compare these scores. For docking of nisin and PHB, Auto dock Vina showed that the confirmation that had the most affinity had a vina score of -3.9 (kcal/mol) while CB-Dock showed a vina score of -4.1 (kcal/mol). These are similar in size having a percent difference of 4.9%. While when we look at the best conformations for Nisn and Cellulose, CB-Dock-2 gave -5.4 kcal/mol and Auto dock vina resulted in -5.2 kcal/mol. This also has a percent difference of 3.7%.

So we can validate through the results from iGEM 2 IISER Pune that our docking results are consistent and hence suitable to make the conclusions we made when performing our original docking.


  1. Calleja D, Kavanagh J, de Mas C, López-Santín J. Simulation and prediction of protein production in fed-batch E. coli cultures: An engineering approach: Simulation of E. coli Protein Production. Biotechnol Bioeng [Internet]. 2016 Apr [cited 2022 Oct 10];113(4):772–82. Available from:
  2. Liu Y, Yang X, Gan J, Chen S, Xiao ZX, Cao Y. CB-Dock2: improved protein-ligand blind docking by integrating cavity detection, docking and homologous template fitting. Nucleic Acids Res. 2022 May 24;gkac394.
  3. Rachel MindellRachel Mindell is a Special Projects Editor at Submittable. She also writes and teaches poetry. Connect with her on LinkedIn., Mindell R. 7 best practices for creating an impactful CSR strategy [Internet]. Submittable Blog. 2022 [cited 2022Oct8]. Available from:,What%20is%20CSR%20strategy%3F,communication%20approaches%2C%20and%20evaluation%20procedures.
  4. Abraham MJ, Murtola T, Schulz R, Páll S, Smith JC, Hess B, et al. GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX [Internet]. 2015 Sep 1 [cited 2022 Oct 11];1–2:19–25. Available from:
  5. Autoclave overview [Internet]. [cited 2022 Oct 10]. Available from:
  6. Fruits and vegetables - optimal storage conditions [Internet]. [cited 2022 Oct 10]. Available from:
  7. Freon™ refrigerants for residential refrigeration [Internet]. [cited 2022 Oct 10]. Available from:
  8. Refrigerant pressures, states, and conditions | achr news [Internet]. [cited 2022 Oct 10]. Available from:
  9. Food and foodstuff - ph values [Internet]. [cited 2022 Oct 10]. Available from: