Team:USP-EEL-Brazil/Model


A serious problem that society has been dealing with for the past centuries is snakebit accidents. It is a tropical disease neglected by WHO (World Health Organization) and it affects a lot of people and can cause grave injuries and even be fatal. In the Southeast of Brazil, the viper that causes more accidents is the Bothrops jararaca, so Team USP-EEL Brazil decided to elaborate a new way to inhibit one of the most dangerous effects of the snake venom , the myonecrosis. Snake venoms are considered by many scientists the most complex mixture of nature, it has a lot of enzymes that include metalloproteases, hemoraginis, and of course the Phospholipase. The latter is responsible for the destruction of cellular membranes, exposing the organism to infectious agents. We know for a fact that snakes are immune to its own venom, that means it has some mechanism that prevents the venom from destroying the cells. That mechanism is the Phospholipase inhibitor enzyme (PLI), it is found in snake blood serum and we aimed to synthesize that protein by synthetic biology. There are plenty of types of Phospholipase and phospholipase inhibitors enzymes, but in this project we decided to focus on the PLA2 and the γPLI interaction since the latter is able to inhibit a wide range of PLA2.

To better predict the behavior of this enzymatic activity, we did two main types of modelling, a kinetic and a structural one.



The work we did at the modelling section began with the following questions: How can we figure out the kinetic behavior of our enzyme? How to inhibit it? Can we figure the 3D structures by a program? All those questions led us to do some research about the Phospholipase A2 and enzymatic kinetics. We learned that a molecule called 4-Nitro-3-(octanoyloxy)benzoic acid can absorb light at a certain wavelength, curiously, that acid is similar to some components of the substrate of the PLA2, so we used it to measure the effectiveness of this enzyme and used its properties to quantitatively determine the concentration over time. We did it by taking samples from our bioreactor and running it on the spectrophotometer at a specific wavelength so that the 4-Nitro-3-(octanoyloxy)benzoic acid could absorb the light and show us the concentration from each time.

Some collaborators helped us with the elucidation of the mechanisms of this specific enzymatic reaction telling us it appears to be a competitive inhibition, so we assumed a michaellian kinetic and integrated the equation to find a curve that could show the behavior of the protein.

To apply the michaellian equation for competitive inhibition, we assumed that the reaction is reversible, all the reactions steps are elementary and the complex ES and EI does not react with other substances present in the environment of the reaction,

Rewriting the expression above as a differential equation, we get:

Separating the variables and integrating both sides of the equality from the start to the end of the reaction, we get the following formula:

Graphing this equation we know the behavior of the enzyme, it starts with the initial concentration and then goes to zero as the time goes to infinity.

We found the missing constants by researching similar proteins (PLA2 and γPLI) from other animals using softwares. We assumed that if the similarity is high enough, then the kinetics should be the same because the active site is alike. Although the myonecrotic effect is not always present in different species snakebites, plenty of those animas has PLA2 is some concentration in its venom.

We chose the batch system because it is relatively simple and would better represent a snakebite accident. We did a few experiments using our substrate (4-Nitro-3-(octanoyloxy)benzoic acid), our enzyme (PLA2) and with and without the PLI.

The control showed a fast consumption of the acid, just as expected, mirroring a snakebite without proper treatment. While the other experiments used different concentrations of PLI, altogether with PLA2 and the substrate.

We did so we could figure out the ideal concentration and effectiveness of the inhibitor. Each experiment was analysed by a spectrophotometer, giving us the concentration in each situation. The tendency of the substrate going towards zero makes sense since we chose to operate in a discontinuous system.

The increase in the concentration of the PLI showed us that the activity of the PLA2 would decrease, giving evidence to sustain the assumption it is a competitive inhibitor in respect to the PLI.



Since the protein crystal of B. jararaca γPLI is not currently available, we chose to try to make a model starting with similar proteins from other species, from homology.

Some proteins evolve from a common ancestor. Even with the evolutionary mechanism of gene duplication and mutations, some divergences lead to the formation of protein families that may present some differences in their amino acid sequences, but which have a high degree of structural similarity. These proteins are known as homologous amino acid sequence and between two of these can be identical, similar or different. In the evolutionary process, the structural conformation is more conserved than the amino acid sequences, being proven that small changes in the sequence usually do not generate a great impact on the three-dimensional structure. For this reason, the 3D structure is extremely important in determining the specific function of the protein.

Protein homology modeling is nothing more than a process done on the computer through some softwares capable of predicting the structure of protein unknown based on the amino acid sequence and 3D structure of homologous proteins. For this, it is interesting that the range of sequential identity between the template protein and the target protein is more than 30%, depending on the length of the sequence and the number of residues. In general, if two proteins share 50% sequential identity, they may have overlapping central Cα with an RMDS value ≤ 1.0 Å.

In our project, we obtained by homology modeling the structure of γPLI from Bothrops jararaca from the known sequence of our γPLI,

and at the final, reaching a model with a 78% of TM-Score.

In homology modeling, there are some steps:

  1. Search for homologous proteins
  2. Alignment of sequences
  3. Construction and optimization of models
  4. Validation of models


Databases of three-dimensional protein structures are growing rapidly, despite the complexity and cost of experiments, e.g. crystallization, to discover these structures. An alternative to this is structural modeling by homology, as already explained. Many times when looking for a template to model a protein one finds several templates with similarity and undesirable alignment, then the modeling with multiple templates can be a great solution, since it will have more templates to guide the modeling. You can even use fragments of these templates to model other fragments of the protein to be modeled. In our case, we found templates with alignments and similarities that were not very well matched, but the TM score was lower than desired. So we performed the modeling using 10 different templates, achieving a model with high reliability.



One of the first steps to predict a protein structure is to analyse the alignment of the γPLI, that is the inhibitor that we want to model with the templates.

With the alignment of the amino acids it is possible to match the position and the amino acids of the templates with known structures, by comparing their sequence.

The figure 1, show us the alignment of our γPLI at the line Seq, with the other templates, that are, respectively: 3bt1U, 7bpsA, 6iomA, 7bpr, 7bpr, 6iomA, 7bps, 2i9bE, 6iomA, 7bpsA. The Alignment done in our project, showed us how each of these proteins match with the different regions of our sequence and contribute to the modelling. The differential here, that we can highlight is that, using 10 templates for this homology modelling we can predict a more trustable and precise model, due to the fact that our structure will be out of 10 references, being each one responsible for a different part of our protein model.



The best Model was selected by the best C-Score of 0.02 and a TM-Score of 72%. The C-Score shows the quality of the predicted model, based on the significance of the templates alignment and the convergence parameters of the structure.

Also the TM-Score is the score based on the structural similarity between the predicted model, compared to the used templates. Both scores show us that the predicted model achieved has a good quality, and accuracy, and can be used for further investigations, as we still lack the γPLI crystallography model.



  • Protein Data Bank: It is a 3D protein and nucleic acid database.
  • UCSF Chimera: It is a free program used for interactive visualization and analysis of molecular structures and related data, including density maps, trajectories and sequence alignments.
  • Python: It is a free object-oriented programming language.
  • i-Tasser(webServ): It is a bioinformatics multi web server used to predict three-dimensional structure models of protein molecules from amino acid sequences.
  • Lomets: It is a structural bioinformatics web server, that works with i-Tasser, dedicated to the homology modeling of 3D protein.

J Yang, Y Zhang. I-TASSER server: new development for protein structure and function predictions, Nucleic Acids Research, 43: W174-W181, 2015.

C Zhang, PL Freddolino, Y Zhang. COFACTOR: improved protein function prediction by combining structure, sequence and protein–protein interaction information. Nucleic Acids Research, 45: W291-W299, 2017.

UCSF Chimera--a visualization system for exploratory research and analysis. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, Ferrin TE. J Comput Chem. 2004 Oct;25(13):1605-12.