Team:OUC-China/Model

Team OUC-China: Main

Overview

The aim of our project is to develop a better post-transcriptional regulation strategy and use it in monocistron and polycistron. Here we built models to design and predict our work.

miniToe —— a better transcriptional regulate strategy

To achieve a better post-transcriptional regulation strategy, we designed a system which is composed of an RNA endoribonuclease (Csy4) and an RNA module named miniToe. We modeled to describe the dynamics of the miniToe system and found a way to achieve different regulation level. The ODEs and molecular dynamics were two main tools to explore it. We used the ODEs to describe the reaction curve and the molecular dynamics in order to give some explanations to experimental data.

Below you can follow the several questions we point out to have a better understanding of model work and the miniToe system. We will discuss some structures of Csy4 in different stages (Q1), some structures of miniToe system in different stages (Q2), the reaction order and some key points of miniToe system (Q3), the simulation of ODEs model (Q4), some significant symbols in molecular dynamics (Q5) and the ways to different regulation levels (Q6).

Q1 : What is the structure of Csy4?

click to see more

click to see less

The endoribonuclease Csy4 from CRISPR family is the main role of the miniToe system. Csy4 (Cas6f) is a 21.4 kDa protein which can recognize and cleave a specific 22nt RNA hairpin which consists of an N-terminal ferredoxin-like domain and a C-terminal domain. The later domain constitutes most of the recognition interactions with the RNA. The RNA adopts a stem-loop structure (the specific 22nt RNA hairpin) with five base pairs in A-form helical stem capped by GUAUA loop containing a sheared G11-A15 base pair and a bulged nucleotide U14. In the binding structure of Csy4-RNA complex, the RNA stem-loop straddles the β-hairpin formed by strands β6-7 of Csy4. The Fig.1 and Fig.2 shows two structure of Csy4: with and without the hairpin bound.

Fig.1 The structure of Csy4 without the hairpin bound. (PDB ID: 4AL5, resolution 2.0 A)

Fig.2 The structure of Csy4 with the hairpin bound. (PDB ID: 4AL5, resolution 2.0 A)

Q2 : What is the structure of miniToe?

click to see more

click to see less

Based on the function of Csy4, we designed a new cis-regulatory RNA element named miniToe, which can be recognized by Csy4. The whole system works as a translational activator including three modular parts:

1. A cis-repressive RNA (crRNA) serves as a translation suppressor by pairing with RBS, and therefore constitutes the critical part of the miniToe structure.
2. A Csy4 site serves as a linker between cis-repressive RNA and RBS, which can be specifically cleaved by Csy4 enzyme.
3. Csy4 enzyme --- A CRISPR endoribonuclease.

Fig.3 The structure of miniToe.

Fig.4 and Fig.5 blow show the two complex of miniToe structures: with and without specific site of hairpin cleaved, which is called the precursor complex and product complex respectively.

Fig.4 The precursor complex of wild-type Csy4.

Fig.5 The product complex of wild-type Csy4.

Q3 : What is the reaction process and key points of miniToe system?

click to see more

click to see less

Fig.6 The working process of miniToe system.

All the reaction happened in our miniToe system can be described chronologically by following five main steps in Fig.6:

(1) The miniToe is produced and accumulated.
(2) Csy4 is produced after induced by IPTG.
(3) Csy4 binds to the miniToe structure and forms the Csy4-miniToe complex.
(4) Csy4 cleaves the special site and divides the miniToe structure into two parts: the Csy4-crRNA complex and the mRNA of sfGFP.
(5) sfGFP is produced.

From the description above, we can get four key problems in our system to make sure whether our system can work successfully:

(1) Can Csy4 dock correctly with the miniToe structure (hairpin)?
(2) How about the ability of binding between the Csy4 and miniToe structure (hairpin)?
(3) How about the ability of cleavage between the Csy4 and miniToe structure (hairpin)?
(4) Can cis-repressive RNA be released from the RBS successfully?

Q4 : How about the simulation results of the ODEs model?

click to see more

click to see less

According to the work process, we built an ODEs model and simulated our miniToe system for 30h, the result can be seen in the Fig.7.

Fig.7 The dynamics of sfGFP by model prediction

We compared the experimental data to the simulation result, find it fit perfectly as Fig.7 shows.

Fig.8 The comparison between experimental data and simulation data.

To see more details

Q5 : How about simulation result of the molecular dynamics?

click to see more

click to see less

The main work our molecular dynamics is to give an explanation for the experimental data at atom level and four key points we discuss before: the experiment has proved that the miniToe structure is working well which means the four key points work well too. We will present you the result of molecular dynamics following the four key points.

For the first key point, we have the interaction matrix to describe the molecular docking, and the heatmap of the matrix can be seen in Fig.9.

Fig.9 The heatmap of interaction matrix for wild-type Csy4.

For the second problem, we calculated the free binding energy of Csy4/RNA complex. The result of binding free energy for wild-type Csy4 is

△ G_{b i n d i n g} = - 59154.9251 \overset{}{} k j / mol

.

For the third key point, we check the distance of Ser151(OG)-G20(N2’), which is a key interaction in the active site of Csy4 to describe the ability of cleavage. The distance curve of Ser151(OG)-G20(N2’) for wild-type Csy4 can be seem in Fig.10.

Fig.10. The distance of Ser151(OG)-G20(N2’) in wild-type Csy4

For the last key points, we use the RMSD of product to describe the release of crRNA. The result can be seen in the Fig.11.

Fig.11. The distance of Ser151(OG)-G20(N2’) in wild-type Csy4

The RMSD is unstable which give an explain to experiment that crRNA is release from RBS.

To see more details

Q6: How to achieve the goal of different regulate level?

click to see more

click to see less

When we make the sensitivity analysis for our ODEs model, we find that while the cleavage rate of Csy4 will influence the expression in GFP, which indicated if we change the wild-type Csy4 to some mutation then we can achieve the different expression of GFP. The Fig.12 is showing that what will happen in the GFP expression curve if we change the cleavage rate of Csy4. It can achieve the goal of different regulate level.

Fig.12 The curve of sfGFP with the changing cleavage rate

miniToe Family —— The way to fifferent regulate level

In the miniToe family, the protein and hairpin are mutated to meet the goal of the different regulation level. In this part, the model helps to design mutants. For the own feature of Csy4 and hairpin, different strategies are used to design: molecular dynamics plays an important role in designing protein mutants while the bioinformatics and machine learning support us to choose hairpin mutants.

We are going to discuss the method to design Csy4 mutants (Q7)，how the method work in design and the result (Q8), the problem different from Csy4 designing when design the hairpin mutants and how to solve it (Q9) and the result of the mutants designing (10).

Q7: How to design the Csy4 mutants?

click to see more

click to see less

The wet lab members give us four important sites, Gln104, Tyr176, Phe155, His29, which play important roles in binding and cleavage in protein Csy4 which can be seen in Fig.13. Considering 20 kinds of amino acids, we have 80 mutants to explore and choose if we only have one site mutated.

Fig.13 The four important site in Csy4

In Q3, we point out four key points which will directly influence the work of our miniToe system. And in Q5, according to the molecular dynamics, we have four significant symbols to describe the four key points.

Now we are going to construct a logic line to show you how to use the three main information above to designing the Csy4 mutants:

What we know and proved by the experiment is that the wild-type Csy4 with the miniToe system is working well, which means that all the important key points we discussion did not exist in the wild-type Csy4. The wild-type Csy4 can dock correctly with the miniToe structure and the Csy4 have a good ability to bind and cleave the miniToe structure, finally the crRNA release from the RBS. So we choose the wild-type Csy4 as a standard, and all the Csy4 mutant can check the four key points by comparing to wild-type Csy4.

Now for the four key points in Q3 we have something in mathematical forms to describe it in Q5. The most important thing is that how to make a comparison between mutant and wild-type Csy4.

Q8: How does the design methods work?

click to see more

click to see less

In Q7, we have the full logic lines in designing the Csy4 mutnats. Then here we will give the comparison method for the four key points in miniToe system between mutants and Csy4 wild-type.

Now we have four mathematical forms including two curves, a numerical value, and a matrix. Four things can be divided into two kinds of data: the matrix and the numerical value. The interaction matrix and the curve can be regarded as a matrix because the curve is discrete, and the binding free energy is just a numerical value.

For the matrix we can use Euclidean distance to describe the difference between two matric:

D (p, q_{W T}) = \sqrt{\sum_{i}^{m} \sum_{j}^{n} {(p_{i, j} - q^{W T}_{i, j})}^{2}}

For the free bind ing energy, we used the formula below to calculate the difference between the wild type and mutants:

\ln (K_{d r e l}) = \ln (\frac{K_{d W T}}{K_{d M U T}}) = △ △ G_{b i n d i n g}

According to description above, we define four value used to compare four key points between mutant and wild-type:

D_{1} (int t e r a c t i o n \overset{}{} m a t r i x)

,

\ln (K_{d r e l})

,

D_{3} (S e r 151 - G 20 \overset{}{} c u r v e)

,

D_{4} (R M S D)

.

By using the four values, five Csy4 mutants is designed in the following table.

Csy4	$D_{1}$	$\ln (K_{d r e l})$	$D_{3}$	$D_{4}$
WT	0	0	0	0
Q104A	0.483	2483	9.48	30.82
Y176F	0.592	-382	11.61	40.62
F155A	0.233	-1627	13.41	35.71
H29A	0.173	833	15.29	316.22

Q9: How to design the hairpin mutant?

click to see more

click to see less

The design of hairpin mutant is qiute different from the Csy4 mutant due to the large library. Except for the two cleaved sites, G20 and C21, we can have 4²⁰ mutants.

Combining the bioinformatics and machine learning, we present an algorithm to pre-processing our big mutation library. Fig.14 is the flow chart of the pre-processing algorithm.

Fig.14 the flow chart of the pre-processing algorithm

The SVM model is training well and the result can be seen in the Fig.15.

Fig.15 The training result

After training the SVM model, we use it to evaluate the hairpin mutants. We choose the hairpin mutants which has high ranks to check the four key points. Finally, we choose the five hairpin mutants. The following chart shows the DR-Score which is the evaluated result of the SVM model for them.

Hairpin-Mutant	$D R - S core$
miniToe1	76.6306
miniToe2	65.6278
miniToe3	66.7160
miniToe4	62.5537
miniToe5	52.9794

Q10: How about the result of mutant designinig result?

click to see more

click to see less

After designing the protein mutant and hairpin mutants, the wet lab members test the all the Csy4 mutants and hairpin mutants. The result can see in the Fig.16.

Fig.16 The experimental result of mutants

And we try to give a comparison between the special value we used before for evaluating the mutant and experimental result to check our model.

For the protein mutants, we give a comparison between D₃ and experimental result. Fig.17 is the result.

Fig.17 The comparison between model and experiment for protein mutant

As we can see in the Fig.3-2, we can find the inner relationship between D₃ and experiment result: the D₃ value describe the difference in the ability of cleavage between the wild-type and mutant. The higher D₃ value means that it will have an big weaker than the wild-type Csy4 in it.

For the hairpin mutants, we give a comparison between DR-Score and experimental result. Fig.18 is the result.

Fig.18 The comparison between model and experiment for hairpin mutant

As we can see in the Fig.18 we can also can find the inner relationship between DR-Score and experiment result except for the miniToe 1. It is reasonable because the machine learning is quite sensitive to the data amounts and the R² is not 1 in our training result of SVM model.

After all, our wet lab member test 30 combinations of our Csy4 and hairpin. Fig.19 is the heatmap result of it.

Fig.19 The heatmap result of 30 combination