Homology model of PfηCarbonic Anhydrase

[drc007]

The recent brilliant results have identified OSM-S-106 as a potent inhibitor of PfηCA (Ki of 2.2 nM ) with useful selectivity over human CAI (Ki 345 nM) and CAII (Ki 53 nM).

It would be great if we had a crystal structure of OSM-S-106 bound to PfηCA but this is a major undertaking especially since the enzyme has not been previously crystallised. It may be possible to get a structure of OSM-S-106 bound to the human enzyme but that will also take time and there is little sequence homology between the human and Pf enzymes. Perhaps an example of convergent evolution.

The sequence of the Pf enzyme is available in Uniprot

tr|Q8IHW5|Q8IHW5_PLAF7 Carbonic anhydrase OS=Plasmodium falciparum (isolate 3D7) OX=36329 GN=PF3D7_1140000 PE=4 SV=2
MKLLYLLYPILLFYNVNVFINYKKSRLMLEMIDKYNTHFVQTTKPYYEFNVTNLTNSKKK
KKKKKRENHLIGSGENMQKKDEKNIKDFHINDYEIDGKTIHNKENKDSFKMNKNKLNDNE
ELFYMDNILSYKPNKKKLFTYSFSENEGNSEKEETLYNFKNMKNINSVQNNINKTFLYNK
LKNVDYYEHGYNWDIGQCKTGKYQSPVDLPMKDLKERELKNISDVYLNLFDDDNYAWNNY
NKPWMKGDFFYYYEYFIKKIVINRQNNIFQIKAARDGIIPFGVLFTTEQPAMFYADQIHF
HAPSEHTFQGSGNRREIEMQIFHSTNYFYDIQDDKSKYKKKYGLHIYNNLKKNSKETSKK
DSSRYHSYLMSFLMNSLSNEQLQNKYNKKKRIKKMKNQYEVISITFTSAEINASTINAFK
KLPSEKFLRTIINVSSAVHVGSDPTLVELKDALNLDALMMMLNIEDMQFLSYQGSSTLPL
CDENVSWKVAKQPLPVSTETILNFYYLLKKHTPNYSGSDNDNYRSLQNVEDNTRHYRKFS
LVQVFPIQVLISSAISNIEDKKVINIIKDISPKNMSFTYYSKWDIYFILFIFYNIVLFLF

A comparison with sequences in the protein database identified PDB code 1Y7W.A Halotolerant alpha-type carbonic anhydrase (DCA II) as the closest match (<20%). Taking this as the starting point, I "mutated" each of the residues in the active site of the enzyme in the crystal structure to the corresponding amino-acid found in the Pf enzyme, checking at each stage that there were no steric clashes or unfavourable torsions. In some cases I also compared with the side-chain alignment of other crystal structures. In the main I've left the loops unchanged since modelling loops is really, really hard, and hopefully they won't impact too much on the topology of the active site. So what I have is a 3D structure of a chimera with the active site of Pf and the loops of the bacterial carbonic anhydrase.

As you can see the binding site has a Zinc atom co-ordinating to three histidine residues, and I let the residues minimise around the Zinc. I then tried docking OSM-S-106 into the active site.

This gave a plausible pose with the sulphonamide binding to the Zinc as expected.

I'm always rather cautious about homology models but:-

• It is likely to be a long while (if ever) before we get a crystal structure of a ligand bound to the Pf enzyme
• Looking at docked poses can stimulate ideas and generate questions that might be asked with molecular design
• We can feed back any SAR results back and refine the model

But please remember a few notes of caution:-

• People see the 3D structure and think it is 100% correct, it is certainly not.
• It is tempting to over interpret differences in binding scores inappropriately, the scoring functions are imperfect even in the best case scenarios.
• Proteins are not static objects, there is considerable flexibility and movement

Please keep those caveats in mind. I'll post the structure tomorrow together with a Jupyter notebook that will allow you to dock structures into the active site. It will require installing a number of command line tools and I'll provide links to instructions.

[holeung]

Thank you, Chris! This is a great start.

[drc007]

Whilst there are number of Open Source computational toolkits and command-line tools they often present a step learning curve for new users. In an effort to provide a simpler environment to access these tools this post will highlight a series of Jupyter notebooks that users can use to run key computational studies that might be undertaken in a drug discovery project.

On the Open Source Antibiotics project we have created a couple of notebooks that help with specific tasks that might be undertaken in a Drug Discovery Project. In particular there is a Jupyter Notebook to aid docking. I've modified this notebook to use the chimeric protein highlighted above.

To use these notebooks you will need to have Jupyter and a number of Python libraries installed. The easiest way to do this is to use Anaconda. Anaconda is a modern package manager and seems to be becoming the preferred source of scientific software. To install all the software libraries and toolkits follow the instructions on this page. Faustina bravely volunteered to go through the process as a complete beginner and the discussion on this thread is quite instructive, particularly if you have not used the command line before, and was used to improve the documentation. The instructions should work for any Unix based machine (Mac OSX, Linux), I've not used a Windows machine for nearly 20 years so if someone wants to write a guide that would be great.

The notebook uses SMINA (https://sourceforge.net/projects/smina/) a fork of AutoDock Vina for docking, but if anyone wants to use a different docking engine then is would be great if they could upload a modified Jupyter notebook.

The protein with and without ligand, the ligand alone and random set of 100 sulphonamides to try docking can be downloaded here.

[Skame161]

Perhaps a mute point given the recent developments with the tRNA aynthetase inhibition; however, alphafold now has a model for pf-CA if anyone would like to do a comparison with the homology model. Unfortunately, I don't have enough computing power to be able to do this myself.

https://alphafold.ebi.ac.uk/entry/Q8IHW5