Epitope Modeling
Our project primarily functions based on being able to use the functional epitopes of the tyrosine-rich oocyst wall protein (TrOWP2) derived from Toxoplasma Gondii to produce an immunological response from the cat, thus preventing future toxoplasmosis and reducing shedding of the parasite into the environment. The peptides that comprise linear B-cell epitopes can be used in place of antigens when producing antibody responses via immunization. As such, we can attach an epitope peptide sequence to our phage-based vaccine in order to effectively immunize the cat. To better understand and determine the locations of these epitopes, we used several web-based software in order to predict, as well as visualize the location of the antigen’s epitope region.
Three web based computational epitope prediction tools were used. Bepriped, then ABCpred and SVMTriP software’s were used to determine, and then additionally validate the sequence of the epitope from the given wall protein antigen sequence we found. Bepried was used in the initial determination of the epitope sequence. As protein structures contain approximately the same amount of hydrophilic and hydrophobic residues, as well as comparing scales for surface accessibility, flexibility, and β-propensity, multi algorithmic programs such as those run by the IEDB’s Bepried can be used to determine the potential epitope region based on the hydrogen bond driven secondary structure characteristics of the amino acid from the sequence. We were able to get an initial prediction of 18 potential peptides in varying locations from the antigen sequence:
In order to become more specific in determining the epitope region, as well as to validate which of the potential peptide regions were most likely to be the functional epitope, we turned to two additional web-based software’s in ABCpred and SVMtrip. These models, primarily developed on the basis of machine learning algorithms, consist of artificial neural networks utilizing positive data training models, though for ABCpred specifically random peptides were used as negative data. Such ML base systems have been found to, in certain instances, produce greater accuracy than propensity scales, while others show little improvement, however the difference in training data does allow this as an effective method to additional validate the results. We determined a 15 amino acid long peptide sequence that shared high confidence intervals and properties from all three servers of Bepried, ABCpred, and SVMtrip, with confidence intervals stated at .555, .94, and 1 from each server respectively.
Using the newfound sequence FKCAEGTTETIDGDCKRLKQFPP, Alpha Fold via the collabfold web server was used to produce 3D protein models that we could then use for additional analysis of the epitope’s properties based on the structure. Below is the predicted 3D structure for the epitope from the amino acid sequence:
Using the 3D structure and PDB file, we were able to utilize the web-based computational mapping server FTMap in order to determine binding hotspots that could additionally be used for binding studies to help further validate the effectiveness of the epitope. FTMap has greater accuracy than classical mapping methods, being developed as computational analogues of X-ray crystallography experiments, using 16 probes to determine binding hotspots. This is particularly relevant as epitope mapping in experimental settings can often be carried out via x ray crystallography. Once regions that are capable of binding several probed (Consensus sites, CSs) are determined, being a fundamental property of the protein structure, sequences that are easily susceptible to bound and unbound interactions can be later used to help determine likelihood of antigen bonding. Below are the locations shown to be hot spots for binding in the protein structure, as well as locations in the sequences for unbound structures.
Phage Production Modeling
The efficiency of our toxoplasma vaccination system is highly reliant on the mechanisms of infection of phage M13 in the cat intestine. We aim to model the production of phage M13 through infection of E. coli bacteria over time through a system of ordinary differential equations (ODEs).
M13, as a filamentous phage, follows a unique life cycle that is not observed in most other bacteriophages. Unlike the lytic and lysogenic models of infection, the host cell - most commonly E. Coli - produces M13 phages that bud out from the cell membrane without lysing and/or killing the cell. In this manner, M13 causes a chronic infection of the host cell, in which viral progeny are continually secreted upon infection.
We created a system of ODEs describing the populations of susceptible cells, infected cells, and free phages, in order to model the chronic infection of E. coli by M13. Here, let S be the concentration of susceptible E. coli cells, I be the concentration of infected E. coli cells, and V be the concentration of free M13 phages.
As E. coli follows a logistical growth model, the increase of the susceptible cell population depends on parameters a (growth rate of E. Coli) and Kc (intestinal carrying capacity of E. coli cells). The population of susceptible cells decreases as they encounter free phages at rate b (rate of phage infection) and become infected cells.
As infected cells are not lysed, but are simply exploited for viral production, their population is also prone to increase as infected cells divide. However, viral infection and utilization of the host cell’s cellular machinery causes the growth to be impacted by a penalty factor < 1, here represented as the parameter ψ. The population of infected cells increases as susceptible cells encounter phages and become infected.
M13 is a phage that buds, not lyses, therefore our model does not employ latency time or burst size parameters. However, we modified the traditional lytic cell model to reflect the continuous phage production rate h. Phage growth is also logistic and limited by intestinal carrying capacity Kp. The free phage population decreases due to both adsorption into cells as well as natural phage clearance or decay, accounted for by the rate δ.
Reasonable estimates for all parameters, determined both in vivo and in vitro, were obtained by conducting a thorough scientific literature search. However, certain parameters, such as the true viral and cell carrying capacity of a cat’s intestine, may need to be verified through further experimentation or applications of resource availability and the Monod equation.
In order to solve and graph the system of ODEs, we assumed starting conditions S0, I0, and V0. S0, the number of F+ E. Coli in the cat’s intestine, was estimated through our research into the cat intestinal microbiome. I0, the number of infected CFUs in the ingestible vaccine, was estimated through our research into probiotics, however will have to be accordingly varied depending on the predictions of our model.
The below graph depicts the progression of the M13 infection over time:
From this, we are able to obtain the maximum concentration of the free phage at equilibrium, which is notably lower than the assumed carrying capacity.
As we adjust various parameters to produce a more accurate model of the production of M13 by E. coli in the cat intestine, we aim to use this production data in conjunction with our transcytosis model to determine extent the impact of the antigen-conjugated M13 on the cat’s immune response. With this, we will work in reverse to determine a healthy and effective dosage of the vaccine to administer to cats.
Diffusion of Phages across Epithelial Cells
In the pdf listed below, we have a detailed explanation of all the steps that we followed to model the diffusion of the phages across epithelial cells.