Rodent data is necessary but insufficient. That is a phrase you hear a lot in CNS drug development, and it took on sharper meaning for us the first time we saw our own NHP biodistribution results sitting next to our mouse data from six months earlier. The two datasets were consistent in some regions and divergent in others in ways that would have been very difficult to predict computationally. This post is an account of what we observed, why the differences matter for a longevity-focused CNS mRNA program, and how we are incorporating the lessons into our next formulation iteration.
We are not reporting this as a completed study. It is a first pass — a single NHP cohort using a leading LNP candidate, designed to generate biodistribution data rather than to answer efficacy questions. The framing is deliberate: at this stage, our highest-priority question is whether we can get mRNA payload into CNS tissue at all, not whether it has the downstream biological effect we want. That is a later question that depends entirely on getting delivery right first.
Study Design: What We Were Trying to Learn
The study used a cynomolgus macaque model with intrathecal administration as the primary delivery route, with a parallel intravenous group to characterize peripheral biodistribution and CNS penetration from systemic exposure. We chose intrathecal as the primary route because it bypasses the blood-brain barrier issue almost entirely, depositing LNP directly into the cerebrospinal fluid where it can distribute through the leptomeningeal space and contact CNS parenchyma. Intravenous was included to answer the question that matters most for a scalable therapeutic: whether systemic delivery can achieve meaningful CNS expression at doses that are tolerable peripherally.
The payload was a luciferase reporter mRNA, not our therapeutic candidate. This is standard practice — before you spend the time and resources validating a therapeutic payload in NHP, you want to know where the LNP is going. Luciferase gives you a quantitative, spatially resolved readout of where translation is occurring, which is more informative than simply measuring tissue LNP concentration because it tells you about functional delivery: LNPs that reach a cell, enter an endosome, escape, and successfully deliver mRNA to the cytoplasm.
LNP formulation used an ionizable lipid with a pKa of approximately 6.5, a phospholipid helper, cholesterol, and a PEGylated lipid at ratios informed by our prior mouse optimization work. Particle diameter was in the 80-110 nm range by DLS, with a polydispersity index below 0.15. Encapsulation efficiency was above 90% by the Ribogreen assay.
Intrathecal Delivery: Signal Distribution Was Not Uniform
The intrathecal results contained the largest surprise of the study. We observed robust luciferase signal in meningeal tissue and in cervical spinal cord segments close to the injection site, which was expected. What we did not anticipate was the sharp attenuation of signal in rostral structures — the cerebellum and cortical regions showed substantially weaker signal than the spinal segments, despite the fact that CSF flow carries LNP particles in the rostral direction continuously.
Our current interpretation, which we are still working through, is that LNP aggregation or nonspecific retention in the spinal subarachnoid space is competing with rostral transport. The leptomeningeal surface area in the spinal compartment is large, and positively charged LNP fractions — even in a formulation designed to be neutral at physiological pH — may adsorb preferentially to the negatively charged proteoglycan-rich CSF matrix before they can reach the cerebral subarachnoid space. This is a hypothesis, not a confirmed mechanism, but it is consistent with what we observe and with published data from several groups working on intrathecal AAV delivery who have reported similar rostral-caudal gradients.
The practical implication is that intrathecal administration alone, at least with this formulation, is unlikely to achieve the pan-CNS distribution we need for a Klotho or FOXO3 upregulation program targeting neurodegeneration broadly. It may be adequate for spinal programs. For cortical targets, we need a different approach — either reformulation to reduce nonspecific retention, ICV administration, or a systemic delivery route that crosses the BBB through a different mechanism.
Intravenous Group: What Actually Crossed the BBB
The intravenous data was more encouraging than we expected, but still requires honest interpretation. We did observe luciferase signal in CNS tissue — specifically, low-level but reproducible signal in cortical tissue, with higher signal in areas near the choroid plexus and in regions with higher vascular density. The signal was lower than in liver by approximately two orders of magnitude, which is consistent with published reports from other groups working on LNP CNS delivery and reflects the fundamental challenge of the BBB as a delivery barrier.
What this tells us: our current formulation does cross the BBB to some degree, and the payload that crosses is functionally delivered — it is not just LNP debris detected by lipid assay, it is mRNA that has been translated into protein. That is meaningful. Whether it is sufficient for therapeutic intent is a separate question, and the honest answer is that we do not yet know what expression level of Klotho or FOXO3 is required to produce measurable biological effect in CNS tissue. That question is one we are actively working on, but it is a research-stage question without a clear answer in the published literature either.
We are not claiming that our current IV formulation is ready for CNS therapeutic use. What we are saying is that the principle of LNP-mediated CNS delivery from systemic administration is demonstrated in our system, and we have a quantitative baseline to improve from.
Peripheral Biodistribution and Tolerability Observations
The peripheral results were consistent with what is established for ionizable LNP formulations. The liver dominated signal in the IV group, as expected — hepatocyte uptake of LNPs after IV administration is the basis for the entire approved LNP-mRNA therapeutics field. Spleen showed secondary signal, consistent with LNP uptake by splenic macrophages and marginal zone B cells. Lung showed modest signal in the IV group, likely reflecting endothelial uptake in the pulmonary circulation.
Tolerability at the doses used in this study was acceptable. Liver enzyme elevations were transient and within a range we would consider manageable, consistent with published benchmarks for similar formulations. We observed cytokine elevation in the IV group at the 6-hour timepoint, with return toward baseline by 24 hours. This pattern is familiar from the COVID-19 mRNA vaccine literature and from earlier LNP therapeutic programs — the innate immune response to mRNA and LNP components is a known variable that chemical modification of the mRNA backbone (specifically N1-methylpseudouridine substitution, which we use) substantially mitigates but does not eliminate.
We are not reporting specific numeric values here for tolerability endpoints because the study was not powered for statistical conclusions in that domain and because the animal numbers are small. The qualitative observation — manageable, transient, consistent with class — is the appropriate level of claim.
What We Are Changing in the Next Formulation Iteration
Based on the NHP data, we are pursuing two parallel formulation changes. The first addresses the intrathecal distribution problem: we are testing a modification of the helper lipid component to reduce nonspecific charge interactions with proteoglycans, with the goal of improving rostral transport efficiency. This is informed by published work on the role of surface charge in intrathecal LNP distribution and is a relatively focused formulation change that should be testable in rodents before returning to NHP.
The second addresses CNS penetration from systemic administration. We are evaluating a surface modification strategy — specifically, the incorporation of transferrin receptor-targeting ligands into the LNP shell — to leverage receptor-mediated transcytosis across brain endothelial cells. This is a longer-term development path. Receptor-targeted LNPs introduce manufacturing complexity and potential immunogenicity questions that need to be worked through carefully, but the potential gain in CNS delivery efficiency is large enough to justify the development investment.
There is also a route-of-administration question we have not fully resolved: whether the correct approach for a longevity CNS program is to accept partial BBB penetration from systemic delivery and design the therapeutic to be effective at that expression level, or to pursue intracerebroventricular delivery as a route that bypasses the BBB problem entirely but introduces a different set of tolerability and practical considerations. This is not a purely scientific question — it is also a clinical development and patient acceptability question that we think is worth engaging with now rather than later in development.
What NHP Data Actually Changes in a Preclinical Program
We want to be direct about what this study changes and what it does not. It does not change our scientific hypothesis — that Klotho and FOXO3 upregulation in CNS tissue could produce meaningful protective effects in neurodegeneration. It does not invalidate our prior rodent data. What it does is sharpen the delivery problem with a specificity that rodent models cannot provide, because the macaque brain is anatomically closer to the human CNS than the mouse brain is, and because CSF dynamics and BBB composition differ between species in ways that matter for LNP delivery.
The value of NHP biodistribution data at this stage of development is not that it gives you answers — it gives you the right questions to ask going into IND-enabling studies. We now know which regions of the formulation space are under-delivering and why, which is more useful than a rodent dataset that might have given us false confidence going into a more expensive and higher-stakes regulatory package.
The IND-enabling work we need to do remains substantial. But we have a much clearer picture of what "good enough" CNS delivery looks like in a primate model, and a formulation direction to pursue toward it. That is the real output of a first NHP study — not a validation, but a calibration.
