Model
AlphaFold 2, developed by DeepMind (a subsidiary of Alphabet, Google's parent company), is a revolutionary computational biology tool that has made significant strides in the field of protein structure prediction. The software uses artificial intelligence and deep learning to predict the 3D structures of proteins with remarkable accuracy. In the 14th Critical Assessment of Structure Prediction (CASP14) competition in 2020, AlphaFold 2 showcased its remarkable accuracy by outperforming other models in predicting protein structures, marking a significant milestone in the field of computational biology and bioinformatics [1]. We want to use AlphaFold2 to see if our T6SS_component-linker-drug model will disturb the original function of the peptide drug or the T6SS component.
In our project, we utilize a python notebook created by sokrypton on GitHub [2], which provides an easy way to use the AlphaFold model to predict our target protein. To run the prediction, we separately input the sequence of the drug and the complex, adjust parameters, and then compare their structure in PYMOL to see if the conformation of the drug or the T6SS component is altered by our engineered complex.
Results have displayed that the structure of T6SS component or drug will not affect each other, and the size of drug protein is within the edge of T6SS components(protein larger than the components is likely to block the secretion pathway). So our prediction is that they can be successfully secreted. But due to the low pLDDT, which indicates the confidence of prediction, of certain regions, we need to see if they can be successfully secreted in tests, and change the choice of peptide and our prediction accordingly. The predicted structure of three types of drugs, fusion protein, and their prediction confidence(pLDDT) are given as follows:
We noticed that there is a significant drop of pLDDT on the right side of the graphs, so we looked into it by analyzing the PAE (Domain position confidence), which indicates the pairwise confidence of each prediction, and we discovered that the drop is probably due to the flexibility of our linker (notice that we use an AAAGGG linker for Hcp1, and for VgrG3 and Tse6 we use their natural linker), so it leads to a result that the linker we use in our project and the natural linker of VgrG3 and Tse6 are flexible linkers. The results of PAE are given as follows:
In this section, we use ordinary differential equations to describe the behaviors of T6SS.
Firstly, we construct 2 ODEs to describe the transcription and translation of the drug delivery gene carried on our plasmid (notice that we generalize the drug delivery component to VgrG for a clearer view, other drug-carrying components like PAAR and HCP should work similarly):
$$ \frac{d[VgrG(mRNA)]}{dt}=k_{\text{transcribe}}n_{\text{plasmid}} - \gamma_{\text{mRNA}}[VgrG(mRNA)] $$ $$ \frac{d[VgrG]}{dt}=k_{\text{translate}}[VgrG(mRNA)]-\gamma_{\text{VgrG}}[VgrG] + k_{\text{disassemble}}[T6SS] - 3k_{\text{assemble}}[VgrG]^3 $$$n_{\text{plasmid}}$: The number of plasmids carrying the VgrG gene.
$k_{\text{transcribe}}$: Transcriptional velocity of VgrG gene, affected by the chosen promoter.
$k_{\text{translate}}$: Translational velocity of VgrG mRNA.
$\gamma_{\text{mRNA}} \& \gamma_{\text{VgrG}}$: Degradation rate of VgrG mRNA and VgrG.
$k_{\text{disassemble}}$: Disassemble rate of T6SS.
Assembly of T6SS:
The equation of T6SS assembly is given as this (notice that we don't take membrane complexes into consideration because they are reusable and don't partake in the main injection structure of T6SS):
$$ nHCP + 3VgrG + PAAR + nTssB + nTssC + TagA + TssA \rightleftharpoons T6SS $$To denote the assembly velocity of T6SS, the equation will be:
$$ v_{T6SS} = k_{\text{assemble}}[Hcp]^{n_{\text{Hcp}}}[VgrG]^3[PAAR][TssB]^{n_{\text{TssB}}}[TssC]^{n_{\text{TssC}}}[TagA][TssA] $$$k_{\text{assembly}}$: The rate constant of T6SS assembly.
$n_{\text{Hcp/TssB/TssC}}$: The number of Hcp/TssB/TssC, which is unsure in a T6SS structure, given that the back of T6SS should touch the other side of the cell, the number of them is determined by the size of the bacterium.
However, due to the self-regulating feature of T6SS genes, the concentration of these parts can be considered constant, so here we assume the velocity of assembling solely depends on VgrG to make it less intimidating. As a result, the equation should look like this:
$$ v_{T6SS} = k_{\text{assemble}}[VgrG]^3 $$$k_{\text{T6SS}}$: The rate constant of T6SS assembly considering only VgrG.
Therefore, the change rate of T6SS complex concentration can be denoted as this:
$$ \frac{d[T6SS]}{dt} = k_{\text{assemble}}[VgrG]^3 - (k_{\text{injection}} + k_{\text{disassemble}})[T6SS] $$$k_{\text{injection}}$: The rate constant of T6SS injection.
Finally, the change rate of the concentration of VgrG in prey cells can be denoted as this:
$$ \frac{d[VgrG\_prey]}{dt}=p_{\text{succ}}k_{\text{injection}}[T6SS]\frac{V_{bac}}{V_{prey}}-\gamma_{\text{vgrg\_prey}}[VgrG\_prey] $$$p_{\text{succ}}$: The percentage of successful injection of T6SS.
$V_{bac} \& V_{prey}$: Volume of bacterium and prey cell.
$\gamma_{\text{vgrg\_prey}}$: Degradation rate of VgrG in prey cell.
If we were to consider the concentration of VgrG in the environment the ODE would be:
$$ \frac{d[VgrG\_env]}{dt}=(1-p_{\text{succ}})k_{\text{injection}}[T6SS]\frac{V_{bac}}{V_{env}}-\gamma_{\text{vgrg\_env}}[VgrG\_env] $$$V_{bac} \& V_{env}$: Volume of bacterium and secretion environment.
$\gamma_{\text{vgrg\_env}}$: Degradation rate of VgrG in secretion environment.
Furtherly, under the stable state assumption, we can infer the relationship between the concentration of VgrG and VgrG_prey by this equation, which can also be utilized in parameter estimation:
$$ [VgrG\_prey]=\frac{p_{succ}V_{bac}}{\gamma_{VgrG\_prey}V_{prey}(1+\frac{k_{disassemble}}{k_{injection}})}[VgrG]^3 $$KLA is an active anti-tumor peptide drug that requires to be delivered into tumor cells to activate tumor cell apoptosis. V. cholerae is originally an intestine pathogen and is able to aggregate in an intestinal environment, which makes it a great microbe for colon cancer elimination. Colon cancer is often characterized by loss of the mucosal layer. Because the hapA gene in the strain we use (V. cholerae v52 rhh) is knocked out, the bacteria won't be able to move through the mucus layer, only to attach to the cancer cell and aggregate, injecting the drug into tumor cells. However, the mucosal layer on the surface of the intestine with other kinds of diseases can also lack the mucus layer (e.g., irritable bowel syndrome, Crohn's disease). KLA is also able to damage those cells if we activate T6SS there. Therefore, we decided to design a circuit to regulate the activity of T6SS. Inspired by the hypoxia biosensor provided by LZU-CHINA-2022 [3], an improved lactate biosensor provided by Wageningen-UR-2022 [4], and an efficient AND-gate [5], we were able to construct the following regulatory circuit:
Notice that V. cholerae is naturally able to sense a low-pH environment and activate colonization when faced with such an environment. Therefore, we can predict a truth table as follows:
| low pH | lactate | hypoxia |
|---|---|---|
| - | - | - |
| - | - | + |
| - | + | + |
| + | - | - |
| + | - | + |
| + | + | - |
| + | + | + |
| Colonization | Secretion |
|---|---|
| - | - |
| - | - |
| - | + |
| + | - |
| + | - |
| + | - |
| + | + |