Longevity Briefs: Unlocking Gene Therapy’s Full Potential

Longevity briefs provides a short summary of novel research in biology, medicine, or biotechnology that caught the attention of our researchers in Oxford, due to its potential to improve our health, wellbeing, and longevity.

The problem:

Gene therapy is an extremely powerful technique, allowing us to speak to cells in the language of life – the genetic code – and make them produce (or stop producing) exactly the proteins we want. Gene therapy involves enclosing genetic material inside a vector, like a virus or lipid nanoparticle, which then delivers that material to the interior of your cells. Yet there’s still a major hurdle for many gene therapies – it’s hard to get these vectors to go only where they are needed within the body. If you are delivering genes that promote neuron growth in order to treat dementia, for example, then you would want most of those genes to be active inside neurons, not other cell types where they might have unwanted effects.

In this study, researchers demonstrate how this problem could be overcome, by reverse-engineering a system that already exists within our cells. CREs (cis-regulatory elements) are sections of DNA that control the activity of nearby genes and can be highly specific to certain cell types. They don’t code for any proteins – instead, CREs bind to proteins called transcription factors, which regulate gene activity. 

We can make use of this system by identifying a CRE that binds to transcription factors that are specific to one type of cell. By putting that CRE in a gene therapy, we can ensure that it is only activated in the desired cell type that has those transcription factors. However, multiple transcription factors may bind to the same CRE, and there’s no guarantee that an optimal CRE (that is to say, one that is tissue-specific and binds strongly to its transcription factor) exists for every application we would need. But what if we could engineer CREs better than those that already exist?

The discovery:

Researchers started by developing a machine learning algorithm by feeding it CRE sequence data from bone marrow cells, liver cells, and nerve cancer cells. When given over 60,000 naturally occurring CRE sequences that were not in its training data, this model (named Malinois) was then able to accurately predict how those CREs would regulate gene activity in different cell types.

Next, the researchers developed another algorithm called Computational Optimization of DNA Activity (CODA). Rather than predicting the activity of existing CREs, CODA was designed to do the reverse – generate new CRE sequences to be as specific as possible to a given cell type. They found that the synthetic sequences it produced were far more likely to be cell-specific than the naturally occurring sequences. Even under their strictest conditions for cell-specificity, over half of the synthetic sequences qualified, compared to around 1 in 5 naturally occurring sequences identified by machine learning. Perhaps most importantly, these synthetic sequences actually worked: when researchers injected mice and zebrafish with gene therapies containing these CREs, the gene therapies were only expressed in the desired cell types.

The implications:

The ability to ensure gene therapies are only active in target cell types could open the door for far more effective treatments. Even when a gene therapy reaches the desired tissue, its effectiveness is often limited by off-target expression in other tissues causing unwanted effects that limit the safe dose. This new strategy gives us a potential way around this problem, and this is going to be essential for the reversal of ageing and age-related diseases, since different tissues and organs age in different ways and at different rates.