Overcoming the liver barrier in nucleic acid therapeutics – an interview with Shawn Davis

Background

Introducing therapeutic RNA or DNA into the body alone is insufficient. These molecules are rapidly cleared without exerting any therapeutic effect. For nucleic acids to be effective, they must be delivered directly into the specific cells they are intended to target – such as T cells (a type of white blood cell). Once inside, RNA initiates translation, producing proteins encoded by its sequence. These proteins drive the therapeutic mechanism – for example, chimeric antigen receptors (CARs), which equip T cells to recognise and attack cancer cells more effectively. As RNA does not integrate into the genome, this expression is temporary.

DNA therapies go a step further, requiring entry into the cell nucleus. They are typically paired with gene-editing tools such as Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) to correct faulty genes by silencing, removing or replacing them – the underlying cause of many diseases. Because this approach directly reprogrammes the genome, it results in stable, long-term expression (for example continued production of cancer fighting CAR-T cells) and is often referred to as a ‘one-and-done’ therapy. However, this permanence also carries greater risks in the event of any errors in the genetic re-coding.

To achieve this cellular delivery, RNA and DNA must be packaged into a vehicle known as a vector. Vectors fall into two broad categories: viral and non-viral. Viruses are naturally adept at navigating the body and infiltrating target cells, making them ideal candidates for repurposing. Scientists have engineered viruses – stripping them of their harmful components – to serve as delivery systems for therapeutic nucleic acids. This first generation of viral vectors, including adeno-associated virus (AAV) and lentivirus (LV), have led to some of the most successful advanced therapy medicinal products (ATMPs) currently available.

Despite their promise, viral vectors face significant limitations. Most notably, the body has a remarkable ability to mount an immune response against viruses – a vital defence mechanism that complicates repeated or long-term therapeutic use. To address this, researchers are developing modified viral vectors with enhanced specificity and reduced immunogenicity, improving their ability to selectively target desired cells while avoiding others.

In parallel, attention has turned to non-viral vectors, particularly lipid nanoparticles (LNPs). These synthetic carriers encapsulate nucleic acids for delivery into target cells. LNPs gained prominence during the COVID-19 pandemic, when mRNA-LNP vaccines achieved global clinical and commercial success. This breakthrough laid the groundwork for broader therapeutic applications of LNP technology.

One such application is the use of LNPs for personalised nucleic acid-based therapies. Such as mRNA-LNP therapies where the mRNA is designed to target the >30 neoantigens identified on a specific individual’s tumour. However, as most therapy manufacturing is set-up for producing large batches, manufacturing therapies in small, personalised batches is vastly more expensive and adds logistical complexity. We helped our client LEON develop a GMP instrument that overcomes this challenge, as it is purposefully designed to produce small batches of lipid nanoencapsulated nucleic acid therapies. This user-friendly pharmaceutical aseptic processing system, the LEON NANOme®, enabled significantly shorter batch run and product changeover times, as well as lower operational cost, in comparison to conventional systems. A step closer to mass accessibility for these novel modalities.

However, LNPs are not without challenges – their most common formulations tend to accumulate in the liver. As a result, the first wave of LNP-based therapies (excluding vaccines) are focussed on liver diseases. To overcome the hepatic barrier, researchers are exploring several strategies. One approach involves modifying surface ligands to reduce liver uptake and enhance targeting specificity. Another avenue is the development of alternative synthetic nanoparticles, such as polymer-based systems, which may avoid liver accumulation altogether. While these newer platforms show great promise, they remain less established than LNPs and lack the commercial momentum gained during the pandemic.

Beyond biochemical modifications to improve targeting, another complementary strategy is localised administration – delivering therapies directly into the affected tissue rather than systemically. This approach can enhance precision, reduce off-target effects, enable higher doses yet lower volumes, and bypass biological barriers that limit systemic delivery.

We sat down with Dr. Shawn Davis, CEO of Liberate Bio, who has spent over two decades innovating how medicines reach their target. Shawn discusses the development of the Raptor™ platform, the first scalable system to screen lipid nanoparticles directly in non-human primates (NHPs), powered by AI, to turn data into meaningful LNP design.

Liberate Bio’s Raptor™ platform enables extrahepatic delivery at industrial speed. Their lead candidate generates chimeric antigen receptor monocytes (CAR-Ms) in vivo. This is a far more efficient and cost-effective approach than the traditional ex vivo CAR-T cell therapies, which require the T cells to be removed from the patient’s body, isolated, modified with the CAR, expanded and reintroduced to the patients. The shift from T-cells to monocytes offers a unique opportunity to activate the immune system while potentially lowering the probability of cytokine release syndrome, and the primary safety concern of T-cell therapies. While Liberate Bio’s initial applications are focused on oncology and autoimmune diseases, their technology can also be utilised to develop treatments for many applications beyond these.

Shawn discusses the challenges and opportunities for developing platforms like Raptor™, that are reshaping complex drug development and delivery.

What were the challenges of therapeutic development?

The first challenge was actually more or less resolved post-COVID – the supply of cargo. At that time, getting even research-grade messenger RNA in small quantities was extraordinarily expensive, so it was hard to see how the modality could be scaled to a large population with a chronic disease. We saw smaller, rare indications, as real possibilities, but how can the cost of goods make sense post-COVID? After COVID, not only was the promise of mRNA clear to a global population, but the industrialisation of both cargo and delivery vehicles was well underway. Multiple high-quality CDMOs were all competing to see who could deliver the best quality material the fastest, at the lowest prices. It effectively commoditised the cargoes to a large degree, in my opinion.

Another problem was that everything was going to the hepatocytes (liver cells). Even with a relatively simple system of a four-component lipid nanoparticle, the design space was very large because of the combinatorial chemistry issue. My team was working on a review article and it was clear that the ionisable lipid makes a huge difference in terms of where things go, how well it internalises and operates – but so do the other components, the size of the particle, the delivery route and the formulation. Everything was making a difference. So, we had an enormous design space with a lot of complexity and no way to manage it.

How did you overcome the complex design space?

I had seen some of James Dahlman’s first publications using a DNA barcode to help track nanoparticles once they encapsulated mRNA and the barcode. I had a large design space with limited information being generated by the empirical methods available and no rational way to design a solution. Past efforts using receptor-mediated uptake had more or less failed to produce real results.

Then came a surprising outreach. Nessan Bermingham (a partner at Khosla Ventures at the time), the founding CEO of Intellia, and Mike Mitchell, a luminary in the lipid nanoparticle world, sent me an idea for a company that resonated with my findings. The need was clear – we have to deliver somewhere other than the liver. They thought we might have a way to do so using RNA barcodes instead of DNA barcodes. If we could multiplex the analysis by delivering and measuring the biodistribution of many particles in a single injection, we could generate a lot of data very efficiently. If the efficiency was high enough, we could test directly in non-human primates and understand the real-world distribution of different nanoparticle designs empirically in the most relevant species. Then we could use artificial intelligence to manage that complexity and to extract insights regarding how to design a solution.

And this idea is what led you to help establish Liberate Bio?

Yes, and I think we’ve accomplished exactly that. It took the team several years of hard work, but we now have a system that allows us to screen directly, shortcutting all of the translation issues in the traditional preclinical/clinical pathway. Providing a large, robust and reliable dataset on which to train the machine learning (ML) algorithms.

How has Liberate Bio generated this large screening dataset?

To start, putting messenger RNA barcodes into lipid nanoparticles. What I mean by that is the messenger RNA still encodes for something interesting, like a reporter protein, luciferase, as an example, and in the non-coding region, we have a unique molecular identifier that we can not only identify, but also quantify using deep sequencing techniques. So, over the two and a half years of establishing the Raptor™ or Rapid Particle Optimizer platform, we proved that not only does this work, but we could deliver up to 100 nanoparticles to a single animal, collect 10 organs from that animal and understand and quantify the biodistribution of those different nanoparticles throughout the 10 organs.

That taught us quickly what was and wasn’t working well in a given organ. It also generated large data sets and so, for the first time, we had the right training data to be able to use ML to extract those insights regarding what mechanisms were driving this approach. We could then use generative AI to help us produce new ideas, effectively automating the hypothesis generation portion of the whole Design-Make-Test-Analyse (DMTA) cycle.

What has this system unlocked?

To develop the system we have:

• Designed more than 3,000,000 lipids
• Tested several hundred of them directly in non-human primates for bioaccumulation
• Confirmed expression in unpooled experiments in rodents
• Confirmed that the translation from rodents to NHPs was, in fact, poor, at well below 10%.

And so,  even when you have a good hit in rodents, there’s more than a 90% chance that it’s not going to translate into primates, which is a huge gap for the industry as a whole.

The development has provided a solution to enable representative screening in a humane and cost-effective way.

The Raptor™ platform is a fantastic example of innovation from a start-up. Have you faced any challenges in developing this solution?

As a start-up with a small staff and limited financing, we knew we needed to move from platform development to generating a first asset quickly and efficiently. We found promising results in the bone marrow of NHPs during our first round of screening. These nanoparticles were accumulating in the bone marrow at much higher rates than typical, while simultaneously accumulating at a much lower rate in the liver than something like MC3 (a well-known ionizable lipid used in Onpattro) would. When we saw how effectively the nucleic acids reached the target tissue or ‘right area code’, while de-targeting the ‘bad area code’, we knew we had a strong foundation.

We have now proven that a single injection can transfect nearly 97% of all circulating monocytes, enabling them to express reporter genes and therapeutically relevant agents like chimeric antigen receptors (CARs). This means we can generate in vivo CAR monocytes (CAR-Ms) for the treatment of autoimmune diseases and some cancers. Our early success demonstrating B-cell depletion in primates has positioned us to move towards the clinic with those opportunities.

Can you expand on the specific challenges the liver presents when delivering LNPs and how Raptor™ is designed to overcome them?

When lipid nanoparticles are administered via intravenous (IV) administration, once they’re in the serum they rapidly interact with a collection of proteins binding to the surface of these nanoparticles. Apolipoprotein (ApoE) in particular creates a ‘protein corona’ around the lipid nanoparticle, which drives uptake by liver cells like hepatocytes that have receptors for ApoE. Within a couple of minutes of IV administration, 80 or 90% of nanoparticles could already be taken up in the liver. This makes it difficult for targeting strategies, like adding receptor-specific moieties, to work effectively, as the particles may never redirect to other tissues before being absorbed.

This is why Liberate Bio prioritises de-targeting the liver as the first step. Our goal is to minimise complexity, so we have developed a four-component nanoparticle that’s very easy to manufacture, scale and achieve optimal delivery results.

Investors and partners often ask about adding targeting moieties, but if we’re already transfecting 97% of target cells, the added complexity may not be worth it. For cell types with lower uptake (30 or 40%), targeted delivery could make sense if well-characterised receptors exist, but we still have to deal with the added complexity and cost of production.

Our primary goal is delivering value to patients globally, and so we focus on scalable, high-quality solutions. De-targeting the liver is the first step in that process and we’re exploring mechanisms like modulating the protein corona or increasing the circulating half-life in the body to give time to improve delivery to other tissues.

Once you've identified high-performing nanoparticle designs using AI/ML, have you been able to then identify the mechanism behind their uptake?

That is our goal, but we’re not quite there yet. We don’t fully understand exactly what’s driving the uptake in those cell types. Our goal is to analyse the common features among nanoparticles that achieve the desired results to uncover those mechanisms. It’s closer to directed evolution than rational design.

Have you observed precedent for that?

Yes, in a related but separate area, at Voyager Therapeutics. They’ve used viral vectors for delivery to the CNS, particularly to neurons, and identified variants that had good uptake. After identifying the subset of vectors that preferentially bound to the neurons, they then identified the receptor mediating uptake common to this subset. It was previously unknown. With it identified, they could then design improved vectors directed to that receptor and optimise their delivery.

So, Liberate Bio’s focus has been guided by delivery success rather than pre-defined targets?

I think that’s a fair characterisation. We identified promising delivery sites in the bone marrow, spleen and muscle, through high-throughput screening of hundreds of lipids. In a limited resource environment, the team had to prioritise what was working best and let data guide development.

We found that if you solve the delivery problem, even for a single cell type, like cardiomyocytes, it isn’t a narrow solution; it opens up several well-validated targets. Each target might underline a unique pathology requiring delivery of a different nucleic acid cargo, yet a single delivery vehicle could now support an entire therapeutics pipeline. So, there is this balance of directed energy and prioritisation, balanced with opportunistic reality. You prove that the platform works to produce the revenue that allows you to follow up on other hits.

Today, despite the success we’ve had delivering to immune cells, sitting in the back of all our minds is:

• What about the hits in other cell types and tissues we deprioritised?
• What about the hits in hemopoietic stem cells we haven’t followed up on?
• What about the hits in the heart that might confirm delivery into cardiomyocytes?
• Could we improve it by administering it intra-arterially rather than IV?
• What if we delivered these nanoparticles with intrathecal administration?

Across the industry, delivery to the CNS remains a tremendous challenge. What would the results look like if we bypassed the blood-brain barrier directly and explored hundreds of nanoparticles? Is there one that would do a great job in neurons or some other cell type?

They say strategy is not what you do, it’s what you choose not to do. We’ve chosen not to do all those things. But they also say the things you regret most in life are the ones you didn’t do. So, I don’t know what that balance looks like over a long-term horizon, but I’m trying to find it. Today, we’re focused on delivery in a single space to prove the potential of a new modality, CAR-M.

But even this ‘single space’ is bigger than you might imagine– if you can create CAR-Ms that effectively deplete malignant cells, whether with BCMA or with CD19 (different target antigens), even that one nanoparticle with these two targets could treat 50 indications in the autoimmune space, or solid tumours, or multiple myeloma. You can address so many unmet needs when you have effective delivery to a particular cell type. We all want to chase answers as scientists, but it’s hard to stay focused. That’s why it’s a difficult business in many cases. We’re very focused, but we also think we should follow the data and see where it takes us.

Unlike CAR-M, we know CAR-T works, and we believe in vivo CAR-T works. We don’t know much about CAR-M, we don’t know how that’s going to play out in terms of efficacy and safety. We have some hypotheses, we have some good supporting evidence, but until we get evidence in humans and understand what those profiles look like, it’s hard for me to tell you. Is multiple myeloma the best possible application because I can clear malignant cells both in the bone marrow as well as in circulation? I mean, that would be quite a good thing. Or is it possible that CAR-M has a much more friendly safety profile from cytokine release point of view that you could extend these therapies to earlier, less severe indications in the autoimmune space?

We think those things are possible, but until we prove them, we can’t be sure. What’s clear is that even with this sort of tight focus, there’s just so much opportunity.

When you’re working to solve fundamental delivery challenges of an emerging class of therapeutic modalities, the potential is enormous. We can afford to be selective. If the only thing you do in your entire career is solve sickle cell disease with an in vivo administration, you’ve had quite a good career. So, it is an exciting position to be in and it goes back to the founding of the company. Ultimately, Liberate Bio was created to solve this fundamental delivery challenge. A problem big enough, and valuable enough to dedicate ourselves to fully: create scalable, safe and effective delivery for genetic medicines.

How do you see physical targeting, beyond IV administration, fitting in Liberate Bio’s ‘jigsaw puzzle’ at the moment?

If I’ve learned anything in 20 years of drug delivery, it’s that if you want something to go somewhere – start by poking a hole in it. For years, people thought that injectable medicines would never compete with oral medications, but look at the rise of biologics and the introduction of more convenient injectables.

Alternative delivery routes like intra-arterial, subcutaneous or even intrathecal can shift biodistribution and improve access to hard-to-reach tissues. Delivery of nucleic acids as genetic medicines has tremendous potential as a third wave of medications after small molecules and biologics. In some cases, a ‘one-and-done approach’, such as gene editing, could justify more invasive delivery approaches, like physically drilling a hole. It’s all about choosing the right tool whether a ‘tiny screwdriver’ or a ‘sledgehammer’ to address the clinical need.

What are you most excited about and what do you see as the biggest remaining challenges in the space?

I think the excitement for me at the fundamental level is that we’re on the cusp of being in a world in which we can effectively cure monogenic diseases by delivering to the right cell type, with editing enzymes to make these changes. In the slightly more distant future, I think applications for polygenic diseases, which are the biggest, highest mortality diseases in the world, will also be possible.

The use of new tools like AI and ML to manage complexity present new opportunities to not only understand the fundamental biology that needs to be addressed but help us create the solutions to do that. I’ve mentioned testing in non-human primates. In a perfect world, we wouldn’t need to do any animal testing; we would go all the way back to the proteome to know what the result in the clinic would be, and then engineer something to solve that problem. That would be a great world. I’m hopeful that we could eventually overcome the challenge of generating data in the right systems to train these algorithms. We could remove that layer of abstraction and go back to first principles.

Want to read more on advanced drug delivery

This article was taken from our latest edition of Insight magazine. Sign up for your copy here.