How to survive a phenotypic antiviral screening

Antiviral screening is a powerful tool to identify new compounds and therapeutic targets against viruses. Traditionally, an antiviral screening targets a viral protein critical for replication and uses high-throughput in vitro assays to identify inhibitors of that protein. This approach is advantageous because the drug target is known, which helps with lead optimisation. Also, specifically targeting viral proteins greatly limits the risk of side effects. Disadvantages of this traditional antiviral screening approach include a high chance of non-specific hits (e.g., candidate drugs that cause protein aggregation or precipitation) and hits that might not work in a physiologic context.

More recently, cell-based phenotypic screenings have become popular. These screening approaches can explore the effect of each molecule across the entire viral replication cycle in a hypothesis-free manner and physiological context. This often reveals novel mechanisms of host-pathogen interaction and new antiviral targets. Because of this, phenotypic screenings are best at identifying first in class molecules (i.e., molecules with a previously unencountered mechanism of action). Hits identified by phenotypic screening are able to penetrate the cell and often have low toxicity. Also, phenotypic screenings can identify both viral and cellular targets, therefore expanding the range of antiviral candidates by targeting both viral proteins and those cellular components involved in virus replication. Although the risk of side effect is higher when targeting the host, virus resistance to cellular protein inhibitors is less likely. Also, cell-based phenotypic screenings often identify compounds that display antiviral activity against multiple pathogens (known as broad-spectrum antivirals).

Given the unbiased nature of phenotypic screenings, some form of target deconvolution (i.e., identification of the molecular target of each hit) is usually required to make sense of the data and to move forward in the drug discovery process.

Based on our experiences in antiviral screenings, we have put together a few suggestions that we think it’s worth keeping in mind when embarking on a phenotypic screen. We hope you will find them useful!

1. Choose your assay according to your needs

If you are interested in a particular stage of the virus life cycle, try to design a high-throughput assay that will narrow your hits to that specific stage.

For example, if you are interested in viral entry, consider stopping the assay at an early time-point or using a viral system that gives a signal just after viral entry or fusion. A good example of this is the use of viral particles containing the beta-lactamase enzyme, which emits a fluorescent signal only after being released into the cytosol following viral fusion. Or perhaps you are interested in viral replication. In this case, you might want to stop the assay before the secondary rounds of replication begin.

Although using the real virus is often desirable, you may want to consider alternative systems (e.g., virus-like particles or replicons) that can only complete specific stages of viral infection.

2. Be ruthless with your hits

Depending on the type of screening, your hit rate should be around 1–3%. This is still a lot of hits, especially if you start from a big library.

Ask yourself, how many hits do you feel comfortable carrying forward? And remember, the process of deconvolution is likely to be much less high-throughput than the screening.

We highly recommend that you focus on your most potent compounds. Generally, we see little value in carrying forward hits that exhibit less than 60% activity at the concentration you run your assay (likely to be around 10 μM).

It is also a good idea to discard those hits that seem toxic, even mildly. Your screening assay might give you an indication of toxicity. For example, when we are performing microscopy-based assays, we like to compare the nuclei count of virus infected plus drug samples versus infected without the drug control. Nevertheless, a proper toxicity assay is usually worth the trouble.

We often perform 96-well plate MTT assays alongside our antiviral assay and then focus our attention of those hits with the highest selectivity index (SI).

3. Understand your hits

Once you have reduced your list of hits to a more manageable number of high performers, it’s time to think about their mode of action (MOA). Several techniques can be used when attempting to understand what your hits are targeting.

Here is a brief overview of some of the most traditional approaches:

• Affinity-based techniques involve immobilisation of hit compounds onto a solid support and mass spectrometry identification of the molecules that bind to the hits. The immobilisation occurs by conjugating tiny tags to the compound of interest in a way that does not affect its properties. As many interactions are likely to be transient or not strong enough to be preserved throughout the entire procedure, some chemical modifications can be designed to stabilise the bound conformation. It is worth remembering that not all of the interactions are likely to be preserved in this less physiological context and that the interaction with the therapeutic target can be missed, or hard to tear apart from more non-specific interactions. Some companies provide this type of service, often relying on their proprietary techniques.
• Expression cloning aims to identify the molecular target of a compound by expressing libraries of proteins in vitro (e.g., yeast two-hybrid, phage display) or, more recently, using genotypic screening. In genotypic screenings, haploid cells are used to knock-out (KO) single genes, one at a time. Reversion of the phenotype induced by the compound in the wild-type cells is a strong indication that the KO protein is involved in the antiviral activity of the drug.
• Knowledge-based approaches look for chemical similarities with reference bioactive compounds or exploit computational (in silico) approaches to identify potential docking surfaces and pockets where compounds might bind. These hypothesis-generating methods require experimental validation.
• Some companies provide a variety of panels of cellular targets (e.g., cellular kinases) or promoter signature profiling kits (e.g., luciferase reporter-based constructs) to test the activity of the selected hits. This approach is suitable if you have some knowledge about your compounds of choice or if there is some reason to target particular families of proteins.

Before embarking on the above approaches, remember that your virus can provide a very useful deconvolution tool. Defining the stage of the virus life cycle inhibited by a compound is much easier than identifying the specific target and helps with the drug discovery process. Assays can be designed to specifically probe entry, replication, and release of infectious particles, as well as other more virus-specific stages of the viral lifecycle. It is relatively straightforward, for instance, to determine whether the endocytic uptake of a virus is inhibited and whether such inhibition is restricted to the virus or also extends to other endocytic cargoes (e.g., transferrin).

4. But do I really need a mechanism?

Knowing the mechanism of action (MOA) of a compound is never a bad thing. This makes optimisation easier, provides insight into safety and toxicity, and might be required for intellectual property (IP) and regulatory issues.

However, there are several examples of compounds that are used in the clinic without any clear indication of how they work. Cyclosporine, used against Hepatitis C virus, is a classic example. Niclosamide, an effective anti-helminthic, also works through a mechanism that is not entirely clear.

If safe and effective, an antiviral can often get away without a thorough understanding of the MOA. So, while we recommend attempting to identify the stage(s) of viral replication that your compound blocks, the identification of a specific molecular target might not be strictly necessary. Phew!

5. How to move on

The chances are that the molecules identified in your screening won’t be the ideal antivirals just as they are, but will require medicinal chemistry and optimisation. Unless you are testing FDA approved sets, you should be looking for libraries where follow-up subsets of compounds are available. These subsets should include a fairly comprehensive panel of compounds with a similar structure to the original hit but carrying different modifications. These are helpful for structure-activity related (SAR) studies and the identification of the most effective compounds.

We also recommend establishing a strong collaboration with medicinal chemists who can help with further compound optimisation, including improvements in its biological activity and pharmacokinetics (PK). This stage cannot be overlooked, as this is what allows a generic hit to become a drug.

6. PK and toxicity in vivo

Once you have excluded the least effective, most toxic, and most non-specific molecules, as well as those with irremediably poor PK properties in vitro, it’s time to test how your best performers behave in vivo.

It is important that the in vivo PK of the compound is assessed first, as this will indicate a suitable compound dosing for virus challenge experiments. Toxicity is where most of the bad surprises arise. It is not uncommon to work with compounds that showed no signs of cellular toxicity in multiple cell types and over several days treatment, but that turn out to be highly toxic in vivo. Unless there is a clear and addressable PK issue (e.g., in its current form the molecule is metabolised into a toxic product, but this can be changed), toxic hits might have to be dropped.

We recommended testing multiple routes of administration and measuring the compound’s concentration in the blood at different time points, and in those organs in which the virus is known to accumulate. The goal is to surpass the in vitro EC50 of your compound.

These types of studies are required in any drug development process but are also important for simple proof-of-concept studies. The findings of these types of study help us to understand how to administer the drug and how to interpret negative in vivo challenge results. Did the compound fail because it just doesn’t work in vivo (for whatever reason), because it wasn’t present at a high enough concentration, doesn’t reach the required site, or doesn’t stay around long enough?

Final remarks

Moving from a high-throughput antiviral screening to a compound or a panel of compounds worth in vivo testing is not a linear process. It can be tackled in various ways for different viruses, and multiple approaches can be used simultaneously.

Nevertheless, it is important that every step in the pipeline is carefully planned before the screening begins. Care should be taken when designing the assay to ensure background noise, false positives, and toxicity are minimised. The library must be manageable yet sufficiently large so that you can discard poor hits without much regret.

Ideally, compound subsets should be available for subsequent SAR studies. We cannot overstress the importance of working closely with medicinal chemists to ensure a smooth reiterative process of modification and retesting.

Finally, pay attention to the PK, first in vitro and then in vivo: it’s hardly the most exciting part of the work, but the information it generates is worth every penny!

Good luck!