With the onset of the coronavirus disease 2019 (COVID-19) pandemic, caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), it came to light that there were no specific and highly effective preventive or therapeutic agents against this virus. The result has been an enormous amount of studies reporting the efficacy or otherwise of novel drug candidates or repurposed drugs against this infection.
A new study, which was released on the bioRxiv* preprint server, describes the use of a combination of virtual and experimental screening protocols to identify potential inhibitors of the virus, which could lead to rapid drug development.
Priming by TMPRSS2
The virus, like other coronaviruses, is dependent on host enzymes to affect its entry into the host cell. Attachment to the latter is mediated by the spike protein protruding from the virus surface, which recognizes and binds to the corresponding host receptor, the angiotensin-converting enzyme 2 (ACE2). However, further steps demand that the spike be primed by a human transmembrane serine protease, called TMPRSS2.
While ACE2 is important in catalyzing many physiological reactions required for cardiovascular health, specifically by counteracting the ACE2 enzyme, TMPRSS2 is not known to have any biological function as of now.
TMPRSS2 is known to be involved in priming other pathogenic coronaviruses, including the Middle East respiratory syndrome and the earlier SARS-CoV, and the influenza virus.
TMPRSS2 knockout mice do not show any obvious differences compared to the wildtype mice, but are resistant to viral infections, indicating that this protein is not an essential part of physiological functions. TMPRSS2 transcription inhibitors also reduce the infectivity of the virus in human lung cells.
Because of this observation, TMPRSS2 appears to be a promising drug target for this and future coronavirus infections.
Already, Camostat, Nafamostat and gabexate are known to target this enzyme, but the first two are non-specific, inhibiting multiple serine proteases. Secondly, these bond with the central ester and thus with the active site serine of the serine protease, which leads to degradation.
Both of these, as well as the metabolite of camostat called FOY251, also bond covalently, whereas less reactive structures would be better suited for therapeutic inhibition.
In order to identify novel TMPRSS2 inhibitors, the researchers adopted a combination of both computational modeling and experimental screening. All known TMPRSS2 inhibitors were tested first, and found to bind covalently to the enzyme.
Expression of purified TMPRSS2
The researchers used E. coli to produce purified TMPRSS2, which was then purified from the cell lysate as insoluble aggregates. These were then treated as required to obtain a correctly refolded protein with enzymatic activity.
The researchers found that the known inhibitors of TMPRSS2 form covalent bonds with the enzyme, which accounts, at least partly, for the very low nanomolar concentrations required to inhibit the enzyme at 50% (50% inhibitory concentration, IC50).
Camostat has a half-life of less than a minute within the cell, being broken down to the protease inhibitor FOY251. While the latter has similar inhibitory potency as camostat, it has a somewhat longer half-life. It is broken down into an inactive compound. This indicates the need to find newer inhibitory compounds with other structures that can provide a longer duration of activity.
The study's gains are thus two-fold: not only did the investigators identify non-covalent hits that can be used to prevent or treat SARS-CoV-2 infections, but these can be modified for use as advanced TMPRSS2 inhibitors.
Creating a TMPRSS2 homology model
The virtual screening was based on a molecular dynamics/simulated annealing-based docking protocol, and the receptor side chains, being flexible, reflect small alterations in the energetics and conformational profile of the enzyme-ligand substrate. These allowed a subset of ligands to be identified. Subsequent biochemical testing showed them to be active inhibitors of the enzyme.
The model for docking protocols had to be synthesized for the active soluble domain, since none were available based on crystallographic or nuclear magnetic resonance structures. The generated structure, using a suitable web interface, was highly homologous with the TMPRSS2 peptidase domain. It also contained a ligand that provided a template for the docking of possible ligands via pharmacophores.
The researchers introduced novelty into the mechanisms used to position the small molecules for the docking study. Firstly, they studied docking with the rigid receptor. This included placing the small molecule inhibitors in different positions via pharmacophores based on ligands bound to other serine protease complexes. Alternatively, they explored those generated by a random arrangement of molecular conformations and random positioning in the binding pocket.
The second phase of the docking study used a new fastdock framework, where pharmacophores were superposed onto bound compounds in experimentally solved structures. Flexible ligands combined with flexible receptors were simultaneously run through a flexible docking search algorithm, combining the different methods mentioned above with a search protocol that maximized the sampling of the different conformations of the receptor.
Again, they developed a new method to score the ligand binding and the enzyme-based on conformational clustering of the ligand with the flexible side chains.
What were the results?
This virtual screening methodology succeeded in identifying conformationally active residues near the active site that can bind various ligands, especially glutamine and lysine, at positions 438 and 342, which seem to take part in stabilizing the bound ligands.
It also successfully identified the three known TMPRSS2 inhibitors mentioned earlier as among the top five hits. Moreover, all three were positioned such as to display the mechanism by which the reaction of the catalytic serine residues with the inhibitors occurs. Not only the serine, but the guanidium functionality, are both positioned to achieve the right interactions with the active site at aspartate-435.
Covalent inhibition also involves histidine and serine at positions 296 and 441, respectively. This led to repeated docking experiments exploring these three residues and the changes in their charges. This showed interesting changes in position and intermolecular distances to allow the enzyme and ligand to react with each other.
The screening protocol also threw up several approved drugs among the top hits, such as pentamidine, propamidine, and debrisoquine. All have a guanidium functional group, and all show a similar docking position, with the positive charge directed towards the Asp-435. These were all inhibitors of TMPRSS2, albeit with varying potencies.
Debrisoquine is interesting in that it is more of a binding fragment than a small molecule. However, it shows some selectivity for TMPRSS2 over trypsin that hints at the possibility of modifying such fragments to obtain more selective serine proteases. This could be by docking with the Lys-342 that is found only in TMPRSS2.
What are the implications?
The researchers thus suggest a solution, in this preprint, to the lack of an experimentally verified structure for docking studies, by substituting a homology model developed from other serine protease domains. With the TMPRSS2, the protease domain has been refractory to purification and refolding in the E. coli system.
This has hindered high-throughput screening efforts so far. This study overcomes this through a combination of modeling and experimental techniques. It also allows the efficient use of TMPRSS2, which is notoriously hard to purify in active form, by first triaging large numbers of compounds before actually assaying the small number of promising hits.
In the future, “having identified promising scaffolds with high ligand efficiencies, future work will be dedicated towards improving potency and selectivity of these inhibitors.” Other hits will be biochemically characterized as well.
bioRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.