Build your own vaccine

 RNA technology has reshaped the vaccine landscape. The biggest advantage is the flexibility. Every single virus, bacteria, fungus or parasite that has ever infected a human uses the same code set to make its proteins. It’s basically universal science Lego. This means we can build vaccines against any pathogen using the same tool kit.

In theory.

The first generation of RNA vaccines were highly effective during the pandemic, they significantly reduced the burden of disease and helped protect many people from hospitalisation and death. But more can be done to improve them for future pandemics or more routine use. One area that requires attention is how the RNA is delivered to cells.

RNA is a large, charged and fragile molecule. If injected into the body on its own, it gets chewed up; this is a natural defence system from the body to protect it against viruses, bacteria, fungi and parasites (which as mentioned above, all use RNA too, and would like nothing more than our cells to make their proteins). We therefore need to package it up to deliver the vaccine payload.

As part of a collaboration with Professor Cameron Alexander (University of Nottingham) and a Cambridge biotech company called Aqdot, we explored a new approach to package RNA. This was recently published in the journal Advanced Materials. The licensed RNA vaccines (from Pfizer and Moderna) use a delivery technology called LNPs. These are Lipid NanoParticles – they use fat particles to encapsulate the RNA, a bit like salad dressing (but not really). There is an alternative approach, that exploits the charge of the RNA molecule. It is long and negatively charged, this means it can then be combined with long positively charged molecules. However, long positive molecules can be toxic to cells – because they bind other stuff, disrupting normal function. The breakthrough in our published study: ‘Modular Supramolecular Polycations Enable Efficient Delivery of Diverse RNA Therapeutics and Vaccines’ was a novel way of packaging RNA, developed by the chemistry team at Nottingham.

Instead of using long molecules, they used much shorter ones, but used a linker to join it all together. The linker was developed by Aqdot – it is from a family of molecules named cucurbiturils (so called because they vaguely look like cucumbers). The cucurbituril molecules are barrel shaped and the hole in the middle is able to accommodate other polymers. This can then link everything together in a web – including the RNA. Using the new platform, we were able to show that you can deliver RNA vaccines and protect against infection with influenza.

There are three major challenges with RNA vaccines – the reactogenicity (side effects after immunisation), the longevity (how long the response lasts) and the stability (the need to keep them in -80). Much of this relates to how the RNA is packaged, the ability to deliver RNA with different formulations opens up opportunities for the further development of the RNA vaccine platform.

24 hours later: using systems vaccinology to understand responses to self-amplifying RNA

 Vaccination works by training the immune system to recognise parts of infectious micro-organisms, so that when we get infected with them we can better fight them off. This builds upon the immune system’s ability to remember what it has seen before. Training this memory takes some time from the initial encounter with the vaccine. Importantly, the events immediately after vaccination can shape the quality and quantity of the response.

These early responses to immunisation have been explored across a range of different vaccines. In our recent study ‘Systems vaccinology analysis of saRNA immunization identifies an acute innate immune signature correlated with adaptive immunity’ we measured the immune response at 24 hours to an RNA vaccine to explore how this might predict responses later on.

The vaccination part of the study took place as part of early trials of a self-amplifying RNA vaccine during the COVID-19 pandemic. This vaccine approach has potential advantages over the mRNA vaccines because the ability to amplify itself means the dose use can be much smaller. The study was performed towards the end of 2020 when the licensed vaccines were beginning to be offered widely to people. This reduced the number of people eligible for the study – because we wanted to evaluate responses in people who had not been vaccinated before. But we were able to enrol a small number of volunteers and undertook the study.

We used an approach called systems vaccinology. Which is a sciencey way of saying we measured lots of stuff and then looked for associations in the data. Blood was collected 24 hours after the initial injection. And three categories of immune markers were measured: RNA, protein and cells. The RNA gives a global picture of everything that has changed, the proteins focuses in on the way that immune cells communicate with one another and the cells gives a snapshot of what immune cells are moving from one place to another. These measurements because they are taken very soon after immunisation tell us about the short-term reaction to the vaccination, but the important thing is then to link them to the outcomes of vaccination. To do this we took a second blood sample several weeks later to measure whether the vaccine had indeed trained a memory response that could help prevent future infections.

We observed that immunisation led to significant changes in immune signalling in the blood. When the changes in RNA were profiled, there was a significant increase in genes linked to the immune response. Many of these fell in a family called the type I interferon pathway. This was not entirely surprising, RNA vaccines are made of RNA (#spoiler), so are many viruses; the type I interferon pathway is a programmed package of genes that fight viruses. We also saw an increase in genes that encode signalling molecules – particularly those that instruct immune cells to move from one place to another. These are called chemokines.

Having looked at the broad RNA picture, we focused down on proteins (as a quick reminder, RNA encodes proteins, proteins do the functional stuff). We saw similar increases in signalling molecules at the protein level in the blood, with increases in a chemokine called CCL2. One of the important functions of the immune system is to direct white blood cells (immune cells) to where the infection is occurring. CCL2 encourages the movement of a family of cells called monocytes, which have the ability to carry vaccine from one place to another and then engage with the cells that form the immune memory (called lymphocytes). Matching the increase in CCL2, we also saw an increase in the cells they recruit (monocytes) in the blood.

This demonstrated that the vaccine was somehow engaging the immune response, but did it mean anything with regards the training response? Well of course it did – otherwise I wouldn’t be writing about it. When we compared the magnitude of the CCL2 response in the blood with the magnitude of the antibodies targeting SARS-CoV-2 spike protein (the vaccine payload). People who had more CCL2 had more antibody.

So why is this important? It allows us to make predictions of how well a vaccine will work as soon as 24 hours after immunisation. It also enables us to design vaccines – if we can learn how to increase the induction of CCL2, we might improve the strength of the response. Of course this work was a massive team effort – and we owe a great debt to the volunteers, who during a time of considerable uncertainty undertook a study to help move us closer to having a working COVID vaccine.

All models are wrong, some are useful

 Cross referencing different systems can compensate for issues in individual systems and provide novel insight.

The saying about the limitations of models is attributed to a British Statistician, George Box. It speaks to how we use models to understand the world around us. This reflects is a deeper philosophical discussion as to what extent we can ever reach ‘truth’. But for now the point is that the models and observations we make in medicine and biology are often flawed. And we need to be aware of these flaws in order to better utilise the models.

In our recently published paper Comparative cross-species transcriptomics during RSV infection identifies targets to treat RSV disease we combined three different approaches to understand infection with Respiratory Syncytial Virus (RSV), a significant cause of illness in babies less than six months old.

Although natural RSV infection in children is the disease of interest, it is challenging to study directly in babies. It is very difficult to collect samples and it is often unknown when the child was first infected. An alternative is to use human infection challenge studies – where volunteers are deliberately infected with RSV. This has the advantage of being in the same species (the human) but the studies use young, healthy adults not babies. There is an additional challenge that all adults in these studies will have previously been infected with RSV at some time in their lives (probably several times) which will affect the immune response to any subsequent infection. The volunteers typically experience mild-to-moderate disease and so don’t fully recapitulate the disease seen in babies. Another alternative is to use mouse models. As well as the possibility to perform biological repeats in genetically identical individuals, following infection of a mouse, you can access all tissues and you can manipulate the response experimentally. However, there are interspecies differences in physiology, behaviour, viral tropism and genetics which can limit interpretation. All experimental approaches are ethically assessed, but there is a sliding scale – more can be done in adults than babies, more in mice than humans. And so by combining these different approaches we can compensate for limitations inherent to each.

In the published study, we compared the immune response in the blood, the lungs and the nose. As with different models, sampling different sites compensates for limitations of sampling an individual site. Blood is easily accessible and it is possible to collect large volumes of material, multiple times. But for a respiratory virus like RSV, it isn’t the actual site of infection (even though the lungs are highly perfused with blood). Blood can reflect cells moving into or out of the lungs, but not the complete picture. The lungs, as the site of infection tell us what is happening where the virus is and therefore can give us much more information. However, they are much harder to sample – it takes a medical procedure called a bronchoscopy to collect the tissue. Repeat sampling over time is not possible and these kind of samples cannot be collected from babies. The nose represents a good compromise. It is easily accessible and is the entry site for infection. Newer sampling methods have enabled the collection of good quality material from the nose without causing discomfort.

Having collected material from infected individuals, we then measured changes in gene expression. Cells, when they are infected or when they are responding to a local infection produce RNA that encodes the proteins they will use to fight off the virus. Profiling the changes in the RNA in a particular sample gives us a snapshot of how the immune system is working. We used an approach called RNA-Seq which captures all of the RNA in a sample and measures how many copies of each gene have been expressed. When two samples are compared side by side, the relative amounts of RNA can be evaluated; this is then presented as differentially expressed genes (DEG for short). The idea being that genes that change in amount are the ones that are important for the response to the infection.

Overall we evaluated 209 samples. When we pooled the data, we saw a clear increase in immune system genes following infection. This was not unexpected, the question we wanted to address was whether they were beneficial. The immune system has a dual role in disease, it is vital to protect us against viruses, but sometimes it overshoots and being in in excess of that required to clear the virus it can damage the lungs and cause us to feel sick. Within the data from this study, we observed increases in genes from the interleukin 17 family (IL-17). This is a type of signalling molecule called a cytokine which shapes the flavour of the immune response. It triggers a cascade of other genes, one of which is called S100A. Both IL-17 and S100A have been shown to cause enhanced disease following other viral infections, but their role in RSV is not known. Returning to the mouse model, we were able to block the action of S100A and show that when inhibited, there is less disease. Overall the study showed that integrating different data sets provides new insight and may ultimately lead to new treatments.