Things to consider when designing the experiment

When a new compound, antibody or test article is being tested for antiviral activity, there are several aspects to consider when designing the experiment. The choice of virus, cells and readout is important, but so are the concentration range, cytotoxicity, controls, and number of replicates.

A good first antiviral experiment should provide a broad and reliable picture of what the candidate does, while generating enough information to distinguish a genuine antiviral effect from cytotoxicity, assay variability, or other unexpected effects. These are also the parameters regulators expect to see characterised: the FDA’s guidance on conducting and submitting virology studies sets out what cell culture antiviral activity data should establish, including the relationship between antiviral activity and cytotoxicity.

The first experiment cannot answer every question, but it needs to provide enough good-quality information to understand the result and decide what to investigate next.

Here we describe some considerations to keep in mind.

Look at the whole concentration range

A first antiviral experiment should give a reasonable sense of the concentration-response relationship, so that it is possible to establish not only within which range a compound is active, but also whether the concentration-dependent effect is gradual or whether there is a sharp transition from inactive to active. This may provide some indication of the mode of action of the compound, although it will not, on its own, identify one mechanism over another.

Observing the full curve can also reveal other anomalies. For example, a compound may be antiviral at all concentrations tested without showing a clear concentration-response. In this case, further testing at a different concentration range or using a larger dilution step may be appropriate. Alternatively, antiviral activity may plateau at less than 100% inhibition, with increasing concentrations failing to shift the plateau. This could indicate that the mechanism targeted by the compound is not the only mechanism involved in infectivity.

Cytotoxicity is not an afterthought

Cytotoxicity should not be treated as an afterthought. A concentration-response cytotoxicity curve should help define the concentrations at which the compound starts to affect the cells, as determining antiviral activity within a cytotoxic range makes it very difficult to establish whether the antiviral effect is due to a direct activity on the target or to the fact that the cells are, to some extent, compromised. Equally, determining how far the concentration of a compound can be increased before it becomes cytotoxic can be just as informative, as it allows the compound dosage to be increased while remaining within a safe range.

The controls are part of the experiment

As in any assay, controls are paramount, not only to ensure that the assay performs as expected, but also to provide reference points. As a minimum, untreated uninfected and untreated infected cells must be included. The untreated infected cells provide the maximum level of infection against which the antiviral-treated wells can be compared; the untreated uninfected cells provide the assay baseline, allowing any assay noise to be identified and measured by comparison.

Without these two controls, no assay can be properly interpreted. Because of this, it is worth including them in multiple replicates, so that if some wells have to be excluded, there is still enough information to make robust conclusions. Multiple replicates also provide important information on intra-plate variability, coefficient of variation and Z’-scores.

When available, a known or previously validated positive control is particularly useful because it indicates whether the assay behaved as expected. If the known antiviral does not produce the expected response, a negative result with the test candidate becomes much harder to interpret.

In certain assays, a negative control may also be important. For example, in a neutralisation assay, a sample-matched isotype control can help determine how specific the neutralising properties of the test articles are, or whether a high concentration of antibody, or even the dilution buffer in which it is stored, can affect infectivity.

Replicates matter

Multiple replicates in the two reference controls, untreated infected and untreated uninfected cells, are critical both for interpreting the assay and for statistical analysis. The number of replicates for the test samples themselves also needs to be considered.

A good balance generally needs to be established during the validation stage, and the NIH Assay Guidance Manual remains a useful reference on the statistical validation of assay performance. Two or three replicates are typically sufficient for a robust assay, but three are recommended to give a better sense of variability and to identify outliers with more confidence.

A well-validated assay can rely on duplicates when screening many compounds, and this approach is more defensible in the context of a concentration-response study, where the serial dilution itself can help determine the trend. However, triplicate is always recommended when possible, and particularly in virology, where the variability introduced by the stochastic elements of a viral infection combines with the intrinsic variability of a live-cell-based assay.

It is tempting to think that the more replicates, the better, and in a way this is true, particularly when more complex statistics are involved. However, the number of replicates should be commensurate with the variability of the assay. Four replicates may be warranted in systems where high variability is observed, for instance with an insensitive readout of a spreading infection. But limiting the number of replicates to what is necessary to guarantee a robust assay has advantages in terms of cost and resources, and can sometimes make it possible to run more samples and controls on the same plate, allowing better comparisons for certain applications.

What about independent repeats? The typical N=3 required by academic publications and by some regulatory routes is often considered the gold standard. However, a good indication of the accuracy of the measurement is provided during the validation stage. During this stage, inter-experimental variability is assessed and an error associated with the assay (in an antiviral test normally relative to the positive control) can be established, allowing the uncertainty of subsequent measurements to be estimated.

Having said that, if time and budget allow, three independent repeats remain good practice, as the error will be specific to the compound being tested.

As an early antiviral study is unlikely to be an end in itself, it can be even more valuable to re-test the same compound in different experiments. This might mean re-introducing it in subsequent rounds of lead optimisation as a reference, allowing the uncertainty of the assay for that type of compound to be determined, or including it in follow-up studies aimed at determining other parameters, such as mode of action or resistance passaging. In fact, the most robust results are not necessarily obtained by repeating exactly the same thing multiple times, but by obtaining coherent results under different conditions.

Choose a readout that measures what actually matters

There are many ways of measuring viral infection or replication, but they do not all provide the same information.

For an antiviral study where nothing is known about the mode of action, it can be useful to keep the assay as broad as possible, by selecting a readout that measures the entire virus replication cycle. Conversely, when only entry events are of interest, it may be cleaner to choose a readout that focuses on these specific events.

For example, immunofluorescence-based readouts can be used to look at viral replication after multiple rounds, or at expression of the incoming virus without waiting for further replication cycles. A CPE assay, by contrast, will generally reflect events occurring over multiple replication rounds.

q(RT)PCR provides information on genome copies, not necessarily infectious virus, although increasing copy numbers over time would generally be consistent with viral replication. It is also worth remembering that even after washing, some of the readout will still come from the input virus, even if it may no longer be infectious at the time of the assay, so the input background will need to be taken into account.

q(RT)PCR can also be complementary to an infectivity readout to tease apart downstream modes of action. If there is an increasing abundance of genome copies but not of infected cells, this may suggest that something downstream of replication, such as maturation or egress, is affected.

Pay attention to things that don’t fit the expected result

One of the most useful aspects of a broad first experiment is that it can reveal more than a simple “active” or “inactive” result.

For example, an unexpected concentration-response relationship, a separation between antiviral activity and cytotoxicity, activity that appears only at particular stages of infection, or a pattern that differs substantially between assays or virus strains may all provide useful information. These observations may provide clues about the mechanism.

A candidate that behaves differently from an established antiviral may be worth investigating for different reasons than one that simply reproduces a familiar concentration-response curve.

Similarly, an unusual result may suggest a useful follow-up experiment rather than being dismissed as experimental noise.

So what should the first antiviral test look like?

In broad terms, a useful first-pass antiviral study should aim to:

Cover a broad concentration range. Enough concentrations should be included to see the shape of the response rather than simply asking whether one dose works.

Include the relevant controls. Cytotoxicity, untreated controls and an appropriate positive control are essential for interpreting the result.

Use sufficient replication. A first indication needs to be reproducible enough to support a decision about what to do next.

Use a robust assay. The assay needs to be capable of producing consistent results across the concentration range while accommodating the controls and replicates.

Measure infectious virus where that is the relevant question. This gives a direct measure of whether infectious virus production has been reduced, rather than relying solely on the presence of viral genomes.

Look beyond the headline result. The shape of the response, separation from cytotoxicity and unexpected patterns may all provide useful information for subsequent experiments.

The aim is not to squeeze the maximum amount of information out of a single experiment. It is to obtain enough good-quality information to know what question to ask next.

Once a candidate shows a convincing antiviral effect, the experimental strategy can become more focused. Potency can be characterised more precisely, the effect can be confirmed using complementary endpoints, the mechanism can be investigated, and additional virus strains or experimental conditions can be introduced where appropriate.

How VRS can help

Designing a first antiviral experiment well is often the difference between a result you can act on and one that has to be repeated. Our team designs and runs these studies routinely, across antiviral drug screening, neutralisation assays, and resistance passage studies, using validated assays with established controls and defined error.

Get in touch to talk through the right design for your compound.