LNPs are the most clinically consequential nanoparticle platform in the world right now. The gap between how far the science has come and how poorly we document it is wider than it has ever been.
Lipid nanoparticles have become the dominant drug delivery platform for nucleic acid therapeutics. From the mRNA COVID-19 vaccines to the first approved siRNA treatment to a growing pipeline of mRNA therapeutics for rare disease, cardiovascular conditions, and cancer, LNPs are the platform technology that is translating the nucleic acid revolution into clinical reality. Yet the documentation practices used to describe LNP designs remain inadequate for the complexity of the systems being designed. This article examines LNP biology and design in depth — the four-component lipid system, the role of each component, the structure-activity relationships that determine delivery efficiency and tolerability, and the formulation parameters that govern particle assembly. It then examines the data implications: what needs to be captured to describe an LNP design completely, why current documentation practices fail to capture it, and how SBOL-based structured design data addresses the gap. It closes with a forward look at where the LNP platform is going and what the data infrastructure needs to look like to get it there.
Keywords
1. Why LNPs Matter Now
The earlier articles in this series established the core argument: biology has discovery infrastructure but lacks design infrastructure, and the consequences are visible in manufacturing failures and regulatory friction. This article takes that argument and applies it to the most important nanoparticle platform in clinical use today — lipid nanoparticles — examining the biology and design in depth before returning to the data implications.
There is a reasonable argument that lipid nanoparticles are the most important drug delivery platform in the world right now. That is not hyperbole — it is a statement about clinical and commercial reality.
The two mRNA COVID-19 vaccines that have been administered to over five billion people are both LNP-based. The first approved siRNA therapeutic, Onpattro (patisiran, Alnylam), which treats hereditary transthyretin amyloidosis by silencing a disease-causing gene in the liver, is LNP-based. The first approved mRNA therapeutic outside of vaccines, mResvia (mRNA-1345, Moderna), which treats RSV in adults over 60, is LNP-based. And the pipeline behind these approved products — spanning rare genetic diseases, cardiovascular conditions, metabolic disorders, and personalised cancer vaccines — is dominated by LNP-delivered nucleic acid therapies.
The platform works. The scientific question is no longer whether LNPs can deliver nucleic acids to cells in vivo. It is which cells, at what dose, with what durability, and with what tolerability — and how to design LNP systems that hit the target for each specific application. Those are hard questions, but they are being answered, programme by programme, through a combination of rational design and empirical optimisation.
The infrastructure question — how we document, transfer, and regulate the designs being developed — has not received proportionate attention. The science of LNP design is advancing faster than the data practices that support it. That gap has consequences. This article examines both what we know about LNP design and what we need to do about how we document it.
2. The Four-Component Architecture
2.1 Ionisable Lipid
The ionisable lipid is the functional core of an LNP delivery system. Its job is threefold: to associate with the negatively charged nucleic acid cargo during formulation, to protect the cargo during transit in the biological environment, and to facilitate endosomal escape — the step at which the LNP releases its cargo into the cytoplasm of the target cell rather than being degraded in the endolysosomal pathway.
Ionisable lipids are so named because they carry a positive charge at low pH (the conditions of formulation, where organic and aqueous streams are mixed at approximately pH 4) and are neutral at physiological pH (approximately 7.4). This pH-dependent ionisation is the key to their biological behaviour: the positive charge at low pH drives electrostatic association with negatively charged nucleic acid, while the neutral charge at physiological pH reduces non-specific interactions with serum proteins and cell membranes during systemic circulation.
The pKa of the ionisable lipid — the pH at which it transitions between charged and neutral states — is a critical design parameter. For LNPs intended for hepatic delivery, the optimal pKa range is approximately 6.2 to 6.8, based on extensive empirical characterisation of structure-activity relationships. For LNPs intended for extrahepatic delivery — to the lung, the spleen, or specific immune cell populations — the optimal pKa range differs, and the relationship between pKa and delivery efficiency is less well characterised.
2.2 Phospholipid
The phospholipid provides structural stability to the LNP and contributes to membrane fusogenicity — the ability of the LNP to interact productively with endosomal membranes during escape. DSPC (1,2-distearoyl-sn-glycero-3-phosphocholine) is the phospholipid used in the Pfizer-BioNTech and Moderna COVID vaccines. It produces stable, well-characterised particles suitable for storage and administration, but its saturated acyl chains make it less fusogenic than unsaturated alternatives.
DOPE (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine) is a commonly used alternative phospholipid that forms a hexagonal phase at low pH and enhances endosomal escape, at the cost of reduced formulation stability. The choice of phospholipid is a design decision that involves trade-offs between stability, fusogenicity, and tolerability that are specific to each application.
2.3 Cholesterol
Cholesterol modulates membrane fluidity and contributes to particle stability. It is typically incorporated at 30–40 mole percent of total lipid content. Modifications to the cholesterol component — including the use of cholesterol analogues such as beta-sitosterol or 20-alpha-hydroxycholesterol — have been shown to affect hepatic versus extrahepatic biodistribution in ways that are not yet fully mechanistically understood. This is an active area of research with direct implications for the design of LNPs intended for extrahepatic targets.
2.4 PEG-Lipid
The polyethylene glycol-lipid conjugate serves two functions. During formulation, PEG on the particle surface stabilises the particles against aggregation by providing steric repulsion. In the biological environment, PEG reduces opsonisation — the coating of particles by serum proteins that targets them for clearance by the mononuclear phagocyte system.
PEGylation density and PEG chain length are both critical design parameters. Higher PEGylation density reduces immune recognition but also reduces cellular uptake. Longer PEG chains provide greater steric protection but may inhibit endosomal escape. The typical PEG-lipid content in clinical LNP formulations is 1.0 to 2.5 mole percent — a range that reflects the balance between stealth and activity that different applications require.
The LNP is not a passive vehicle. It is an active participant in the delivery process, and every component is a design decision with consequences that extend from formulation to the endosomal membrane of the target cell.
3. Structure-Activity Relationships and Their Data Implications
3.1 What We Know
The past decade has produced an increasingly detailed understanding of the structure-activity relationships that govern LNP delivery efficiency and tolerability. The work of the Cullis group at UBC, the Anderson group at MIT, the Siegwart group at UT Southwestern, and others has established that ionisable lipid geometry — specifically the relationship between headgroup area, tail length, and degree of unsaturation — is a primary determinant of delivery efficiency and endosomal escape.
The Lipid Nanoparticle Structure-Activity Relationship (LNP-SAR) landscape is now partially mappable using computational tools. Machine learning models trained on ionisable lipid structure-activity data can make reasonable predictions about the delivery efficiency of novel ionisable lipid candidates, and these predictions are being used to accelerate the design of next-generation ionisable lipids for challenging delivery targets including the lung, the brain, and specific immune cell populations.
3.2 What the Data Requires
The existence of computational SAR models creates a new data requirement that current documentation practices are not designed to meet. A machine learning model trained on LNP structure-activity data is only as good as the data it is trained on. If that data is documented in inconsistent formats across laboratories — with particle size measured by different instruments, encapsulation efficiency measured by different assays, and delivery efficiency measured in different cell lines under different conditions — the model will learn noise as well as signal.
Structured design data helps address this problem. Within the Nanomedicine Design Stack, L2 Nanomaterial Architecture can be linked to L1 Manufacturing & Specification and to structured characterisation assertions containing not only measured values but also methods and conditions. SBOL-encoded design objects can then be aggregated across laboratories and time points more reliably than prose- or PDF-based records. The data infrastructure required to accelerate LNP design through machine learning is therefore closely related to the infrastructure required for reproducibility and regulatory review.
4. The Delivery Landscape: Where LNPs Are Going
4.1 Extrahepatic Delivery
The dominant approved LNP products — Onpattro, the COVID vaccines, mResvia — are all liver-targeted or administered intramuscularly for local and systemic immune response. The next frontier for the LNP platform is extrahepatic delivery: nucleic acid therapeutics delivered to the lung for respiratory conditions, the central nervous system for neurological diseases, muscle for muscular dystrophies, and specific immune cell populations for cancer immunotherapy.
Extrahepatic delivery is harder than hepatic delivery. The liver has evolved to capture and metabolise foreign particles from the circulation. Most intravenously administered LNPs end up in the liver whether or not that is the intended target. Directing LNPs to other tissues requires either local administration (intratracheal for lung, intrathecal for CNS) or the addition of targeting ligands — antibodies, peptides, aptamers — that direct the particle to cells expressing a specific receptor.
4.2 Personalised Cancer Vaccines
The most ambitious near-term application of the mRNA-LNP platform is the personalised cancer vaccine — a therapeutic vaccine encoding the neoantigens specific to an individual patient’s tumour, manufactured on a timescale of weeks from tumour sequencing to first dose. BioNTech and Moderna both have personalised cancer vaccine programmes in late-stage clinical development in partnership with Merck.
The data challenges of personalised cancer vaccines are an order of magnitude greater than those of standard biopharmaceuticals. Each vaccine is a unique product, manufactured once for a single patient. The manufacturing process must be rapid, highly controlled, and extensively characterised within a very short timeframe. The regulatory framework for these products is still developing. The documentation infrastructure that will support this framework does not yet exist in mature form.
For personalised cancer vaccines, the Stack would allow each patient-specific product to be represented across the same five design levels: therapeutic intent and patient-specific objective; relevant biological interactions; delivery architecture; the LNP and mRNA nanomaterial architecture; and the manufacturing and specification record for the individual lot. The supporting SBOL3 representation could encode the mRNA sequence, formulation parameters, characterisation data and provenance links to sequencing data, design algorithms and manufacturing records.
5. The Data Infrastructure Required
The LNP platform is advancing toward applications that place increasing demands on design documentation — more complex formulations, more sensitive cargo, more demanding delivery targets, and in the case of personalised medicines, more patient-specific variation. The documentation practices that were adequate for the first generation of LNP therapeutics are not adequate for the second.
What is required is not an entirely new documentation technology, but a more coherent application of structured biological design. The Nanomedicine Design Stack provides the architecture: five levels connecting therapeutic intent to manufacturing and specification, with SBOL3-oriented implementation dimensions for representing the resulting design data. Existing tools such as Benchling and SynBioHub can contribute to that implementation layer, while regulatory structured-data initiatives provide a direction of travel.
What is required is the decision, by the organisations developing the next generation of LNP therapeutics, to document their designs as structured data from the outset — not as an afterthought for regulatory submission, but as the primary record from which all other documentation is generated. That decision would compound in value with every design iteration, every manufacturing transfer, and every regulatory submission that draws on the design data.
The LNP platform is too important, and too complex, to continue documenting in formats designed for simpler systems. The field has the tools. The question, as always, is whether it will use them before the next avoidable failure makes the argument for us.
The final article in this series steps back from the technical and examines the strategic question: why the pharmaceutical industry’s instinct to treat design data as proprietary is costing it more than it gains, and why open standards are not a threat to competitive advantage but the foundation on which it is built.
References
- Hou, X. et al. (2021). Lipid nanoparticles for mRNA delivery. Nature Reviews Materials, 6, 1078–1094.
- Kulkarni, J. et al. (2021). The current landscape of nucleic acid therapeutics. Nature Nanotechnology, 16, 630–643.
- Sabnis, S. et al. (2018). A novel amino lipid series for mRNA delivery. Molecular Therapy, 26(6), 1509–1519.
- Hassett, K. et al. (2019). Optimization of lipid nanoparticles for intramuscular administration of mRNA vaccines. Molecular Therapy Nucleic Acids, 15, 1–11.
- Kauffman, K. et al. (2015). Optimization of lipid nanoparticle formulations for mRNA delivery in vivo. Nano Letters, 15(11), 7300–7306.
- Billingsley, M. et al. (2020). Ionizable lipid nanoparticle-mediated mRNA delivery for human CAR T cell engineering. Nano Letters, 20(3), 1578–1589.
- Cheng, Q. et al. (2020). Selective organ targeting (SORT) nanoparticles for tissue-specific mRNA delivery and CRISPR-Cas gene editing. Nature Nanotechnology, 15, 313–320.
- Semple, S. et al. (2010). Rational design of cationic lipids for siRNA delivery. Nature Biotechnology, 28, 172–176.