Description

Docking is done to test the binding affinity between our different parts, the receptor, and the ligand.

Since all our interacting parts are proteins and peptides, we used protein-protein and protein-peptide docking tools. We found several tools and settled on three; LightDock [1] v0.9.2 standalone on Ubuntu release 20.04, ClusPro2.0 ,[2], and GalaxyTongDock_A [3] web servers.

LightDock is a protein-protein and protein-peptide docking tool that uses the Ab initio protocol based on the Glowworm Swarm Optimization (GSO) algorithm. LightDock can do rigid body and flexible protein docking; it depends on the ProDy package to do flexible docking [4].

The benefit of LightDock is it is downloadable, so it will be easier to use for the automated docking of multiple proteins without human interference. We wrote a bash script for this job; it also offers an online web server.

GalaxyTongDock_A is an online web server that performs Ab-initio rigid-body docking to do symmetric and asymmetric docking of protein structures [2].
ClusPro 2.0 webserver is a highly benchmarked tool that generates different conformations and does rigid body docking followed by clustering based on RMSD (Root Mean Square Deviation) and selecting clusters of the lowest energy, which has a higher probability of existence; afterward is the structure refinement which based on energy minimization [3].

Since we are using three tools, each one has its own scoring function to rank the complexes; we had to find another scoring method that can be used to rank all models with the same parameters. We used PRODIGY for this purpose. It calculates the affinity based on intermolecular contacts and properties derived from non-interface surfaces and lists many parameters like delta free energy and interacting residues [5].

Plug-Sink system

I. HtrA1 model

HtrA1 exists in nature as a trimer. Modeling gave the structure file of the monomer. So, we had to do trimerization. GalaxyTongDock and ClusPro were used to do the trimerization step of HtrA1, and the best model was chosen based on Cβ-deviation and RMSD after alignment with the experimental model (RCSB: 3NZI) was 1.47 Å on RCSB Pairwise Structure Alignment tool.

Fig.1: HTRA1 model aligned to experimental model of HTRA1 (3NZI)

II. Inhibitors and HtrA1 binding peptide

After selecting the final HtrA1 trimer, the model was docked with H1A (HtrA1 binding peptide), WAP-four domain 13 inhibitor, and SPINK8 with their respective binding affinities, which were -32.8, -38.3, and -25 kcal/mol respectively.

Then the model of each inhibitor is aligned with the top docked model of H1A vs HTRA1 to determine their relative position on HtrA1 trimer, afterward the distances between their four terminals were calculated and subtracted from the distance between clamp terminals to determine the best length of the linkers connecting them to the clamp.

Fig.2: HTRA1 Trimer docked with inhibtor 1 aligned to HTRA1 H1A binding peptide

After modeling the complete composite part (inhibitor1/2-linker1-BP1-linker2-BP2-linker3-HtrA1 peptide), the best structures obtained after quality assessment are 3D aligned to the single and docked structure of inhibitor 1, inhibitor2, and H1A peptide to determine how their structures are changed upon putting them together in fused protein, the results were:

Table 1. RMSD (in Å) with docked structure of the following

Peptide name

inhibitor1

inhibitor2

H1A

switch 10

-

3.84

1.64

Switch 12

-

5.05

1.66

Switch 15

6.08

-

1.67

Switch 18

6.16

-

4.21

III. Tau and Aβ binding peptides

After filtration, the best models of tau binding peptides (TD28rev and WWW) and Aβ binding peptide (Aβ_37-42) were docked with three Tau models; paired helical filaments (PHF) of tau aggregates, PHF seed, and PHF^* seed, and Aβ₄₂ to measure its affinity towards their target and if they can bind to the other target as reported by paper[6, 7].

Fig.3: graph representing the binding affinity of binding peptides (TD28rev, WWW and seed) to targets (PHF, PHF*, tau, Amyloid-beta) using Galaxy, Cluspro and lightDock

The results showed that TD28REV peptide has the highest affinity towards PHF and PHF^* followed by WWW, then Aβ_37-42 in both Galaxy and ClusPro. In PHF the affinity of seed exceeds that of WWW in ClusPro this might be because WWW is designed especially for PHF^*. In both tau filaments and Aβ₄₂, the energies are measured twice, the energy of binding of the whole structure which considers the contacts between tau filaments, and the energy of binding due to contacts between tau filaments and the binding peptides only.

In all graphs, the trend of energies is almost the same regarding the tools: LightDock > Galaxy > ClusPro.

Although we estimated that the tau binding peptide binds to Aβ₄₂, we did not expect their affinity to be higher than the seed as in the Aβ42 graph. On the other hand, the seed showed good affinity towards PHF, PHF^*, and tau filaments even better than its affinity towards Aβ₄₂ and in some cases exceed TD28rev and WWW.

The three peptides were used to construct the clamps. TD28rev and WWW are used to construct the Tau clamp while the seed is repeated twice and used to construct the Aβ₄₂ clamp. Every two peptides are attached together using five different linkers followed by model construction and filtering to get the best models.

IV. The clamps

Regarding clamps We have done three dockings which are:
1- Tau binding peptide clamps With tau aggregates and its seeds.
2- Aβ₄₂ binding peptides clamp with Aβ₄₂.
3- Clamps with each other.

We docked them together to ensure that the binding affinity between clamps would not be high enough to cause the clamps to aggregate as we had a concern that the clamps may aggregate on each other and the affinity between the clamps and each other to be higher than the clamps with either tau of Aβ₄₂

The affinity of Aβ₄₂ binding peptide clamps to bind with themselves was very low which means that they are less likely to aggregate. However, tau clamps’ affinity with themselves is high. So, it may form aggregates.

Fig.4: graphical representation of clamps binding energy and compare them with their building binding peptides

According to the docking results of Galaxy, the affinity of tau clamps are slightly less than tau clamps towards PHF and PHF^*, but when tested against tau, they had better scores against TD28REV, while the score of WWW seems to be outlier.

However, ClusPro results show that their affinity is higher than tau binding peptides regarding all targets. The affinity of Aβ₄₂ binding peptides clamp with Aβ₄₂ is higher than that of seed alone on both Galaxy and ClusPro.

V. HtrA1 against the target protein:

The remaining part before constructing the whole system is testing the affinity of our free protease HtrA1 towards its targets: PHF, PHF^*, tau aggregates and amyloid beta on different docking tools.

Fig.5:Cartoon representation of HTRA1 docked to different targets (a-K) can be found in the table above. Green: HTRA1, Cyan: docking target

from the results it is obvious that HtrA1 has high affinity towards tau and Aβ42. However, these results are affected by the fact that docking receptor and docking ligands are both multi-chain models. And the calculated models affinity is high since the affinity of HtrA1 monomers to each other and the affinity of tau fibrils to each other are also counted.

VI. The whole system:

After modeling the (inhibitor-linker1-TBP1-linker2-TBP2-linker3-HtrA1 peptide) we docked it with HtrA1, and we expect it to act as the basic parts; the inhibitor binds at the catalytic domain, and the H1A binds to the PDZ domain leaving the clamp part exposed and accessible so that it can bind to our target protein (tau or Aβ₄₂)

We could not use Galaxy in the docking as it requires the proteins to be less than or equal 1000 aa and the HtrA1 trimer with the switch exceeded that limit. In addition, LightDock will take days of working on finishing these jobs. So, we continued with ClusPro solely.

In the basic models of the switches, the HtrA1 binding peptide is predicted to fold in a way that makes it embedded inside the protein. This gives a docked structure that deviates from our expectation. However, from the previous docking of basic HtrA1 binding peptide we expect that the affinity of binding will be high enough to expand the structure of the switch to make the fold as expected in our design. Unfortunately, ClusPro does rigid body docking so the change in the fold of the switch will not be predicted.

Fig.6: cartoon representation of the top predicted structure of the switches a)Switch 10 b)Switch 15 C)Switch 12 D)Switch 18. Blue: H1A peptide, Red: Binding peptide, Cyan: inhibitor and green: linkers

To get a hint of this assumption, we docked the switches basic parts with the basic interacting parts (H1A with HtrA1 PDZ domain, inhibitor with HtrA1 catalytic domain, and clamps with their targets). We did not use tau aggregates as it needs more computing resources and takes more time.

The docking proved that H1A binds to PDZ with high affinity and in case of switch 18 it exceeds the inhibitor, while the affinity of the clamp to PHF and PHF^* is low when compared to H1A and the inhibitor, which opposes our concept. This means that our parts need more optimization. In addition, the scores obtained by PRODIGY depends on the contacts between docking partners which is low in PHF and PHF^* as they are small peptides of 6 amino acid residues. It is expected to be higher when docked with the tau aggregates like the previous results.

Snitch system:

In the snitch system we tested the interaction of three parts; Trim21-linker-DocS, GST-Coh2-linker-TBP, and the target protein tau; PHF, mutated PHF and whole PHF aggregates.

1- Coh2 and DocS:

For the Wet-Lab work, we had to add GST tag to one of DocS or Coh2, to determine to which one GST should be added we searched the literature and found that when DocS domain is expressed alone it has low yield and stability in solution [8], so we have to express DocS fused to another protein to stabilize it. We found in previous studies GST has been fused to DocS and has given better yield [9]].

However, in the composite part DocS will be fused to truncated trim21 so we cannot add GST the fusion protein as it will generate a very high molecular weight protein, but a 6xHis tag can be added as it will not generate a high molecular weight protein.

We need both interacting parts to be tagged with different parts for the pull-down assay. Therefore, in the composite part we added GST to Coh2 [10].

We built both DocS and Coh2 basic models with GST and 6xHis tags once each alternatively and tested how GST can influence their interaction by docking.

’

Fig.7: Cartoon representation of docked structures a) GST-Coh2?with 6xHis-DocS on Galaxy, b)GST-Coh2 with 6xHis-DocS on Cluspro, c) GST-DocS with 6xHis-Coh2 on Galaxy d) GST-DocS with 6xHis-Coh2 on ClusPro

The docking results showed that the GST tag can interact with Coh2 or DocS when fused with the other part. However, experimentally GST will be used in the pull-down assay. So, we were more concerned about the binding affinity of models at which DocS and CoH2 bind to each other.

2- Trim-linker-DocS vs GST-Coh2-linker-TBP:

After determining to which part the GST tag is added, we had to test the interaction of the whole fusion proteins together, and how the binding affinity will be affected.

We could not use galaxy in this docking job since it accepts proteins with a distance of less than 240 Å between the two furthest atoms.

Fig.8:Cartoon representation of the docked structure of Trim21-(G4S)3-DocS with GST-Coh2-(G4S)3-TBP a) WWW TBP on LightDock, b) WWW on ClusPro, c) TD28rev on LighDock, d) TD28rev on ClusPro.

The affinity scores showed that the binding of Coh2 to DocS is better when they are fused than when they are free, this may be due to the contribution of truncated Trim21 in forming some bonds with the GST-Coh2 model.

For further tests of the docked model, we cannot depend on the affinity score only. Therefore, we uploaded the models on the SWISS assessment server to determine the clash score, QMEANDisCo, and other parameters discussed in the modelling section. In addition to measuring the RMSD with experimental Coh2-DocS complex (PDB id: 2CCL).

From the table above since model a cannot be uploaded, we did not use it for further tests. The remaining models to compare were models c and d. Model c has better clash score, QMEANDisCo, MolProbity Score, and RMSD with lower bad bonds and angles. So, we are to complete our work with this model.

3- Whole system vs target proteins:

The final docking step in the tTrim21 system is done to detect whether tau can bind to the whole system (Trim-linker-DocS --- GST-Coh2-linker-TBP) and at its binding peptides or not. At first the docking was done blindly, but the results showed no binding of tau models at the binding peptide site. So, the docking is repeated with defining the docking site at the binding peptide site to see whether it will bind or not then measuring the affinity.

Docking of tau aggregates could not be done on Galaxy due to the distance issue mentioned above, while on LightDock there was an error related to ProDy package as its atoms count is different from LightDock atoms count so we could not proceed with LightDock.

Fig.9:Cartoon representations of docked structures of Trim21-(G4S)3-DocS, GST-Coh2-TBP and tau seeds (PHF and PHF*). a-l) are found in the table above, red: TBP, cyan: docking target (PHF or PHF*), Orange: Coh2, Blue: DocS, grey: GST tag and Green: Trim21 and linkers