Investigation of Binding Interactions at the Pae MurC (PDB: 6X9N) Active Site

[Yuhang-CADD]

Dear all!

I am Yuhang and I am a PhD student in Prof. Matthew Todd's group. I have been recently trying to design and synthesise some pyrazolopyrimidines based on the properties of Pseudomonas aeruginosa MurC and visualise its interactions for the GitHub Wiki. However, I am now confronting several problems when examining the structure. As I am still a beginner in the bioinformatics area, I would really appreciate any insight from the OSA team @drc007. The original paper for this experiment was Pyrazolopyrimidines Establish MurC as a Vulnerable Target in Pseudomonas aeruginosa and Escherichia coli For full details please see my ELN page at this link ELN_Yuhang Many thanks!

Aim:

1. To understand how compound AZ5595 was interacting with the Pseudomonas aeruginosa (Pae) MurC (6X9N)

2. To figure out how the actual binding interactions between AZ5595 and MurC (in the 6X9N) at ATP binding site deviated from the predictions of the original paper DOI

3. To design and synthesise an amine derivative of AZ5595 based on the structural properties of 6X9N.

Methods:

PLIP and PyMOL were used for the inspection of interactions (3D). Discovery Studio (academic) was used for generating the 2D diagram of interactions. Chain A of 6X9N was selected as a representative to illustrate binding interactions at ATP site. 3UAG (E. Coli MurD) and 1P3D (H. Influenzae MurC) were aligned with 6X9N at ATP site for validation of Mur ligases' conserved topology.

Results:

1. Interactions at the ATP binding site of 6X9N:

PLIP Result (as a reference)

H-bonding Interactions: ASN 190 had an amide hydrogen to donate to the Nitrogen atom in the pyrimidine ring. ASN 292 interacted with an NH from the Pyrazole moiety and an NH from bridging amine via 2 H-bonds respectively. THR 189 acted as a H-bond Acceptor (HBA) by interacting with the NH group of the pyrazole moiety. TYR 246 donated its hydrogen to the NH group of the pyrazole moiety which connected to the pyrimidine.

Hydrophobic Interactions: HIS 124, HIS 288, LEU 287 and LEU 291 were involved in the formation of hydrophobic pockets

Pi-cation Interactions: HIS 288 and the pyrazole group had a Pi-cation interaction.

Fig 1. Binding interactions between AZ5595 (light magenta) and ATP site residues (coloured in purple and labelled in orange). Water molecules were represented by red balls (licorice style). Polar bonding interactions were coloured in yellow.

Fig 2. 2D map of 6X9N binding interactions at ATP site with AZ5595.

2. Alignment of 3UAG and 6X9N:

RMSD = 7.0 (1231 atoms) shows that the ATP binding site of Pae MurC is topologically similar to the E. Coli MurD.

Fig 3. Alignment of ATP binding sites from E. Coli MurD (3UAG, white/grey) and P. aeruginosa MurC (6X9N, purple). ADP (yellow, from 3UAG) and AZ5595 (light magenta, from 6X9N) overlapped.

3. Alignment of 1P3D and 6X9N:

RMSD = 2.5 (1734 atoms) shows that the ATP binding site of MurC is highly conserved among Hin and Pae.

Fig 4. Alignment of ATP binding sites from H. Influenzae MurC (1P3D, cyan) and P. Aeruginosa MurC (6X9N, purple). ANP (yellow, from 1P3D) and AZ5595 (light magenta, from 6X9N) overlapped.

Conclusion:

1. PLIP has found most of the interactions correctly, but still, in order to explore all proper interactions, PyMOL and visual inspections are necessary!

2. The Pae MurC homology model from the original paper DOI shows a certain degree of accuracy in binding predictions. However, it still missed some important interactions as compared to the binding information from the actual crystal structure 6X9N. Furthermore, the orientation of AZ5595 in 6X9N was different from what was predicted in the homology model from the original paper DOI.

3. MurC ATP site was highly conserved among H. Influenzae and P. Aeruginosa.

4. P. Aeruginosa MurC and E. Coli MurD have a certain degree of similarity at the ATP binding site.

Questions:

1. The alignment of Pae MurC (6X9N chain A) & E. Coli MurD (3UAG), ADP sites overlap and it shows that the MurC seems to lack the C-terminus part? Was it being manually deleted or that was just natural? I also checked another crystal structure of Pae MurC, 5VVW, it showed the same topology.

2. If that (lacking the C-terminus) was its natural topology, then the OH group of the AZ5595 was facing towards the solvent, (just double-check) am I right about this?

3. When dealing with crystal structures, the water molecules around the active site, keep or remove? Not so sure, but I kept water molecules within 3 Angstroms to AZ5595 (named UYD in pdb structure).

4. For the breaking loops in the interaction diagram (such as the following pic), do we need to do homology modelling to fix it before posting it or maybe in the future for publication purposes? But what I have been concerned about is that if we do homology modelling, and especially the breaking loops are in the binding centre of 6X9N, it might change the topology of the binding pocket and may influence the docking experiment and the illustration of binding interactions When having 8 slightly different binding poses in a multi-chain structure, shall we just present one of them for drug design or should we consider all of them? Fig 5. The detailed structural overlap between H. Influenzae MurC (1P3D, cyan) and P. Aeruginosa MurC (6X9N, purple) at ATP binding sites with ANP (yellow) and AZ5595 (light magenta) involved.

5. In the "Results and Discussion" section, "Proposed Binding Mode of Pyrazolopyrimidine" sub-section, "the 3rd sentence" of the original paper DOI, the author mentioned that they minimised the homology model before docking. I am confused by this information as the minimisation would change the shape of the protein and have a chance of increasing the error level. So, I want to know is that a necessary step before docking and how to do it right (e.g. minimise the structure with binding substrate in the binding site).

[mattodd]

Hi @WieselHenson great! These analyses are key to understanding design of amine analogues that was kicked off by @dayang-us in e.g. #31 . There are lots of things here, but let's focus first on the binding that is revealed by the MurC crystal structure with AZ5595. I think it's really crucial that the compound's alcohol is pointing towards solvent. My question is: how different is the crystal structure from the pose proposed in the original Hameed et al paper? Might you be able to take a snapshot of that image (if there is one)? @eyermanncj may know the proposed binding mode in detail and how it differs, so I'm tagging him in case.

[Yuhang-CADD]

@mattodd As far as I know (also mentioned in my ELN), the original paper didn't have a crystal structure of Pae MurC, so they used Hin MurC as a template to build a homology model of it. This was based on the fact that Hin MurC has a certain degree of similarity to the Pae MurC, especially at the ATP binding site (conserved >85%). After that, they docked AZ5595 into the ATP binding site of modelled Pae MurC and gave the predicted pose. That was how they came up with the following two pictures (The paper is copyrighted by the ACS but I am using the pictures as Fair Use).

Fig 6. The predicted binding interactions of pyrazolopyrimidines at the ATP binding site of the Pae MurC homology model.

Fig 7. The predicted binding interactions of pyrazolopyrimidine (10) at the ATP binding site of the Pae MurC homology model. Pyrazolopyrimidine (10) is now known as AZ5595.

**In my opinion, the difference may originate from the process of homology modelling and molecular docking. What I have done was based on the published crystal structure 6X9N from Seatle group. So that may explain why the difference occurred.**

[dayang-us]

Re: ATP binding site differences between structure determined from X-ray crystallography vs. homology modelling

Hi, I've done some work on binding interactions for my project and thought it would be useful to add some of my personal observations on the differences in the binding site here (adapted from my meeting slides):

My observations:

• Interactions in 6x9n is ‘different’ from Hameed et al’s putative binding site from the homology model – it seems like it’s flipped, but the residues interacting with the scaffold are the same (highlighted by the red boxes)
• Thr128 of the “phosphate binding site” in model interacts with phenyl glycinol moiety, but in the crystal structure it’s nearer to the tert-butyl/pyrazole ring
• Tyr246 and Cys226 interacts with the pyrazolopyrimidine ring instead of the pyrazole ring

In the paper they have also mentioned that:

• AZ5595-resistant strains were found to develop a single-point mutation on the MurC gene resulting in a Y246C amino acid substitution - which I believe is a strong indication that Tyr246, found in both the homology model and 6x9n, is important for biological inhibition.

Hope this helps!

[Yuhang-CADD]

Reply to Question 4 and 5

Based on suggestions from Dr Chris Swain, Mr Chenggong Hui (@huichenggong), and PhD Girinath Pillai, I have had some understandings to solve Question 4 & 5, here is the conclusion:

1. Normally we don't do energy minimisation for the entire protein as we have the crystal structure (normally with a decent resolution), especially when Electron Density Map fits perfectly to the atom coordinates. Modifying hydrogen bonding networks and optimising steric clashes or torsions among residues are necessary steps and sometimes need to be done manually. These protocols are unlikely to change the entire protein structure. So, what could happen is we do constrained energy minimisation of the protein structure with all of the heavy atoms frozen and only minimise the hydrogens.

2. When you have the crystal structure of a protein and it has some residues or loops missing, we do homology modelling to fix it, followed by constrained energy minimisation. When you don't have a crystal structure of a protein, homology modelling can also work by fitting the sequence information into a structurally similar protein template and generate the modelled protein structure. Constrained energy minimisation should also be applied to such model as the supplementary material of the original paper showed (DOI).

I would really appreciate it if Chris @drc007 could double-check if my conclusions are reasonable! Many thanks for this!

[Yuhang-CADD]

Reply to Question 2

The crystal structure 6X9N and 5VVW seemed to have the C-terminus of P. aeruginosa MurC missing. Not sure why, but this certainly made me concerned about the possibility that the OH group of AZ5595 might not be pointing towards the solvent as it seemed to be. So, in order to find where the C-terminus should be positioned, I did an alignment of 3UAG and 4HV4 (Fig 8). This was based on the paper (DOI) which found the Yersinia pestis MurC was 61% similar to the P. aeruginosa MurC protein in sequence.

Fig 8. The structural overlap between Yersinia pestis MurC (4HV4, blue) and P. Aeruginosa MurC (6X9N, purple) at ATP binding sites with AZ5595 (light magenta) sitting in the pocket.

From this image, we can reasonably hypothesize that the missing C-terminus of 6X9N should be positioned around where the one of 4HV4 was sitting. At the same time, we can see that the OH group of the compound AZ5595 was nowhere near the C-terminus (see the orange circle which showed the position of 4HV4 C-terminus, also a hypothetic position of 6X9N C-terminus) and was pointing towards the solvent (see the green arrow which indicated the orientation of OH).

[eyermanncj]

Yuhang,

Nice analysis. My understanding from the Seattle group is that the C-terminus was engineered out of the Pae murC protein in order to get a crystal structure.

Joe

[Yuhang-CADD]

Yuhang,

Nice analysis. My understanding from the Seattle group is that the C-terminus was engineered out of the Pae murC protein in order to get a crystal structure.

Joe

Many thanks! Joe! That is really helpful!

Best,

Yuhang