Model

Modelling was a critical part of our igem project. Our main goals were as follows:

• To understand and model the complex interactions in the skin microbiome that lead to dysbiosis
• To understand and model the structure, stability, and binding characteristics of an aptamer

We achieved our first goal by developing an agent-based model on MATLAB. While the second goal was realized by performing molecular dynamics simulations of an aptamer on GROMACS and AmberTools.

MatLab Modelling of Skin Microbial Interaction during Dysbiosis

The microorganisms present on the skin can be divided into two categories, commensals and pathogens. Although there are several microorganisms living in our skin In our model we only focused on three microbes: C. acnes, S.epidermidis and S.aureus.

S. epidermidis and C. acnes are the two most fundamental bacteria in the skin and are considered as commensal bacteria because they are harmless in healthy conditions. The second type are harmful and pathogenic microbes that colonise the skin and lead to infections and inflammation. S.aureus is one of the most predominant pathogenic bacteria in the skin. (Claudel et al., 2019 ; Fourniere et al., 2020)

In healthy skin microbiota S. epidermidis and C. acnes interact among each other to protect against and prevent pathogens, as well as participate in skin equilibrium with the secretion of beneficial metabolites. Thus they keep each other's population in balance and prevent growth of S.aureus. (Fourniere et al., 2020)

The various interactions can be summarised as follows

1. C. acnes inhibits proliferation of S. epidermidis through hydrolyzing sebum triglycerides and release of propionic acid
2. C. acnes releases coproporhyrin III promoting S. aureus biofilm
3. S. epidermidis inhibits proliferation of C. acnes through succinic acid (Claudel et al., 2019)
4. S. epidermidis can inhibit S. aureus biofilm formation with production of the serine protease glutamyl endopeptidase (Esp). Moreover, when Esp-expressing S. epidermidis induces keratinocytes to produce antimicrobial peptides via immune cell signalling, S. aureus is effectively killed. (Byrd et al., 2018)

An imbalance in skin microbiota distribution, also known as dysbiosis, leads to several skin problems. Sudden over-colonization of the pilosebaceous unit in the skin by C. acne can cause a loss of diversification, loss of metabolic balance and potentially causing acne. Research has also shown that a loss of microbial diversity and loss of balance between C. acnes phylotypes may lead to dysbiosis (Fourniere et al., 2020).

The interaction among the skin microbiome on either balanced skin or skin with dysbiosis has been depicted using a simple MATLAB Model.

Figure 1 shows the percentage bacterial distribution and level of metabolite concentration in the normal skin microbiome. We considered 1 to represent the reference value of metabolite concentration in the y-axis that indicates a normal level. Values lower than 1 represents a decrease in concentration relative to the normal level. While values higher than 1 indicate a rise in concentration relative to the normal level. An agent based model of microbial distribution is modeled on MATLAB as shown in figure 2.

Figure 1: Percentage of bacterial distribution and metabolite concentration in normal skin

Figure 2: Agent based model depicting balanced skin microbiome

Similarly, for the case of Dysbiosis, figure 3 and figure 4 depict the bacterial percentage distribution, variation in the levels of various metabolites, and microbial distribution the resolution imbalance in microbial distribution as seen in the animation.

Figure 3: Percentage of bacterial distribution and metabolite concentration in dysbiosis

Figure 4: Agent based model depicting dysbiosis skin microbiome

Assumptions of the Model

The crucial assumptions made in the MATLAB model are as follows:

• The skin is modelled as a two dimensional grid on MATLAB, representing a part of the skin also known as the face & scalp sebaceous unit.

• Under balanced skin conditions the distribution of skin microbiota is assumed to be 70% C.acnes, 20% S.epidermidis, 1% S.aureus and 9% of other bacteria. (Byrd et al., 2019)

• The number of square units in the grid signifies the population density of the different bacteria where each colour is associated with one bacteria. The population density depends on the growth and death rate of each type of bacteria.

• The growth and death rates are assumed to be proportional to the concentration of a certain metabolite secreted in the skin. The constants of proportionality have been assumed arbitrarily to fit the model and do not represent actual values.

• The proportionality relationships used are

  • Growth rate of C.acnes and S.epidermidis is constant

  • Death rate of C.acnes is proportional to the concentration of Succinic acid

  • Death rate of S.epidermidis is proportional to the concentration of Propionic acid

  • Growth rate of S.aureus is proportional to the concentration of coproporhyrin III

  • Death rate of S.aureus is proportional to the concentration of glutamyl endopeptidase (Esp)

• During dysbiosis there is a change in the values of proportionality constants for the growth and death rates leading to imbalance in bacterial population density.

AmberTools

Although DNA aptamers offer better stability over RNA aptamers in their laboratory use, there are no simulation tools aimed at DNA aptamer conformational and binding properties. As part of out project the aims were to generate a computational structure of an aptamer with the sequence:

5' ATACCAGCTTATTCAATTAGCAACATGAGGGGGATAGAGGGGGTGGGTTCTCTCGGCT 3'

Firstly, the structure was generated in Avogadro as a .pdb file, (assumed to be a right handed B DNA model by Arnott et al.) where a manually guided energy minimisation was performed based on previous work by Stoltenburg et al. This minimised structure showed binding similar to an expected secondary shape, however the universal force field (UFF) used in Avogadro is not ideal for DNA/RNA or single strand DNA energy minimisations.

A second version of the sequence was generated in AmberTools21. A tool called VMD was used to generate a double stranded version of the aptamer and the complementary strand was deleted in a tool called DSV. To avoid errors in xLeap with loading the PDB file, hydrogens on the terminal OH groups on both 5' and 3' ends were deleted as well. The single strand DNA structure was loaded to AmberTools and adapted for further processing in pdb4amber, which corrected the deleted terminal H atoms.

Preparation for molecular dynamics was done by adding solvent, ions and force field and running energy minimisation, heating and production.

The force field and solvent types used were: implicit solvent model IGB1 and force field ff99; and explicit solvent TIP3P and force field bsc1. The benefits of implicit solvent is that they reduce the computational resources necessary to perform a simulation, however they lack solvent viscosity for realistic simulation of electric potential.

The simulation was run in vacuo to establish initial values and fast simulation times, and was aimed to be repeated with both implicit and explicit solvent models as well as including and excluding counter ions.

The energies were extracted from the simulation to show the progress of the energy minimisation and showed local potential energy minima.

FIGURE

References

• Claudel, J. P., Auffret, N., Leccia, M. T., Poli, F., Corvec, S., & Dréno, B. (2019). Staphylococcus epidermidis: a potential new player in the physiopathology of acne?. Dermatology, 235(4), 287-294.
• Byrd, A. L., Belkaid, Y., & Segre, J. A. (2018). The human skin microbiome. Nature Reviews Microbiology, 16(3), 143-155.
• Fourniere, M., Latire, T., Souak, D., Feuilloley, M. G., & Bedoux, G. (2020). Staphylococcus epidermidis and Cutibacterium acnes: two major sentinels of skin microbiota and the influence of cosmetics. Microorganisms, 8(11), 1752.‘