Ask most LNP groups what they measure to characterize a new formulation, and you'll hear the same answer: dynamic light scattering for size and PDI, RiboGreen-based encapsulation efficiency assay, and if the group is thorough, cryo-TEM for morphology and zeta potential by electrophoretic light scattering. This panel — let's call it the standard panel — is fine for what it was designed to do: confirm that a particle formed correctly, that it has cargo inside, and that it isn't aggregating.
For hepatic delivery, the standard panel is adequate as a first pass. If your LNP is headed to the liver, the downstream biology is relatively forgiving: ApoE corona formation drives receptor-mediated uptake, and the endosomal escape step in hepatocytes is reasonably well-characterized for common ionizable lipid chemistries. Aberrant formulations are caught by size and PDI. The biology fills in the gaps.
For CNS delivery, the standard panel misses most of what matters. This article describes what it misses, why it misses it, and what an expanded characterization panel for CNS-targeted LNPs should include.
What the Standard Panel Measures — and What It Doesn't
Dynamic light scattering (DLS)
DLS measures the hydrodynamic diameter of particles in solution by analyzing time-dependent fluctuations in scattered laser light due to Brownian motion. The z-average diameter (intensity-weighted mean) and PDI (a measure of size distribution width) are the primary outputs. DLS is fast, non-destructive, and well-suited for batch QC.
What DLS does not tell you: it cannot distinguish between unilamellar spheres, multilamellar structures, empty particles, and loaded particles of the same size. A population of aggregated small particles and a population of large unilamellar particles can produce identical DLS readouts. PDI below 0.20 is a necessary but not sufficient indicator of formulation quality. DLS also provides no information about surface chemistry, targeting ligand presentation, or cargo integrity. For surface-engineered CNS LNPs specifically, DLS provides no indication of whether targeting ligands are present, correctly oriented, or accessible — it measures hydrodynamic size only.
An additional limitation for CNS formulations is that DLS is performed in well-defined buffer, not in the physiological milieu the particle will encounter in vivo. LNP size in serum differs from DLS measurements in PBS due to protein corona adsorption — the effective hydrodynamic diameter after corona formation can increase by 15–40 nm depending on serum concentration and LNP surface chemistry. This matters for transcytosis because particle size affects the geometry of receptor engagement and endosomal processing.
RiboGreen assay for encapsulation efficiency
The RiboGreen fluorescence-based assay quantifies free (unencapsulated) nucleic acid by comparing fluorescence signal with and without detergent-mediated disruption of the LNP membrane. Encapsulation efficiency (EE%) is calculated as: EE% = 1 − (fluorescence without detergent / fluorescence with detergent) × 100.
What EE does not tell you: it measures how much nucleic acid is topologically inside a lipid membrane. It does not measure whether the nucleic acid is in a form that will be released and active after endosomal escape. Degraded RNA encapsulated inside an LNP produces EE% identical to intact RNA. A formulation can have 85% EE and deliver zero functional cargo if the RNA was fragmented during the encapsulation process — due to ribonuclease contamination in the aqueous buffer phase, thermal degradation during microfluidic mixing, or buffer pH excursions during tangential flow filtration.
For CNS applications in particular, where the cargo may be a large base editor mRNA (4.5–5.5 kb) or a prime editor construct with both editor mRNA and pegRNA, integrity of the full-length cargo matters more than in small siRNA applications (where siRNA duplexes are short enough to remain functional even with partial degradation). An EE of 82% tells you nothing about whether those base editor transcripts are full-length and functional.
Cryo-TEM morphology
Cryo-transmission electron microscopy provides direct imaging of LNP morphology at nanometer resolution in a near-native hydrated state. It reveals particle shape, internal structure (electron-dense cargo presence), multilamellar features, and fusion events. For CNS LNPs with surface modifications, cryo-TEM is useful for confirming that the surface engineering step — conjugation of targeting ligands to PEG-lipid, for example — did not disrupt particle morphology or cause aggregation.
What cryo-TEM does not tell you: it is a sampling technique — a well-prepared cryo-TEM grid images thousands of particles, but this is a small fraction of the total population. Rare aggregates or aberrant particles are undersampled. It provides no functional information — a visually perfect LNP at cryo-TEM may have non-functional targeting ligands, degraded cargo, or suboptimal ionizable lipid packing geometry that compromises endosomal escape. The inverse is also true: an LNP with some visible heterogeneity by cryo-TEM (bimodal population, some ellipsoidal particles) may still transduce neurons efficiently. Structural appearance and functional performance are not the same measurement.
Zeta potential
Zeta potential by electrophoretic light scattering measures the electrokinetic potential at the particle surface (more precisely, at the slipping plane of the electrical double layer). For ionizable lipid LNPs, this measurement is performed at physiological pH (pH 7.4) to confirm that the ionizable lipid is in its near-neutral charged state — a critical safety and specificity parameter. Cationic LNPs at physiological pH trigger non-specific cell uptake, complement activation, and immune responses. A near-neutral or slightly negative zeta potential (−5 to +5 mV at pH 7.4) is the target for CNS-targeted LNPs, for the same reasons it is the target for liver-optimized LNPs.
What zeta potential does not tell you: it does not confirm that the ionizable lipid will become sufficiently cationic in the low-pH environment of neural cell endosomes to destabilize the endosomal membrane and release cargo. pKa characterization — typically by TNS (6-(p-toluidino)-2-naphthalenesulfonic acid) fluorescence assay across a pH range — is a separate measurement that reports on the pH-dependent ionization behavior of the lipid. Zeta potential at pH 7.4 confirms the near-neutral character needed for circulation; pKa value predicts endosomal escape potential.
The Gap: What the Standard Panel Misses for CNS Applications
1. BBB model uptake and transcytosis efficiency
The most important thing the standard panel omits is any measurement of whether the LNP can actually cross the BBB. For a CNS delivery formulation, this is the primary performance metric — yet it is absent from virtually every published LNP characterization panel in the literature. Transwell hCMEC/D3 assays — measuring the appearance of fluorescently labeled LNPs on the basolateral side of a tight junction-forming endothelial cell monolayer — provide direct evidence of transcytosis competence. The assay is not an in vivo BBB, but it discriminates between formulations with active transcytosis capability and those without, and it has a 72–96 hour throughput that is compatible with formulation screening workflows.
A CNS-targeted LNP that is not evaluated in a BBB cell model before in vivo work is not characterized for its intended application. TEER measurement of the monolayer should accompany each transcytosis assay to confirm barrier integrity; TEER below 50 Ω·cm² in hCMEC/D3 cultures indicates a compromised monolayer that will overestimate transcytosis for all formulations equally — and invalidate the experiment. Lucifer yellow permeability as a paracellular integrity marker should run in parallel to each experiment as an orthogonal barrier integrity check.
2. Endosomal escape efficiency in neural cells
Endosomal escape is the step after cellular uptake where LNP cargo must escape from the endosomal compartment into the cytoplasm to exert its function. The galectin-8 (Gal8) recruitment assay provides a quantitative measure of endosomal disruption events per cell — Gal8 is a cytoplasmic lectin recruited to ruptured endosomal membranes, visualizable by immunofluorescence or via a Gal8-GFP reporter system. The assay is cell-type specific, which is exactly what is needed: endosomal escape in HeLa or HEK293 cells does not predict endosomal escape in post-mitotic cortical neurons, where endosomal maturation kinetics and late endosomal pH differ.
For ionizable lipids with pKa values in different ranges, the relationship between pKa and endosomal escape efficiency is measurably different in hepatocyte-lineage cells versus neuronal cells. This is the practical reason why liver-optimized ionizable lipids (pKa 6.2–6.5) underperform in CNS cell models: the pKa optimized for hepatocyte late endosomal pH is not optimal for the endosomal acidification profile in post-mitotic neurons. The Gal8 assay, run in primary cortical neurons (DIV7–10) or SH-SY5Y differentiated cells, captures this difference directly and provides a quantitative basis for pKa candidate selection for CNS applications.
In a representative ionizable lipid screening experiment using a panel of nine lipid candidates spanning pKa 5.7 to 6.6, the Gal8 puncta count per SH-SY5Y cell peaked at pKa 6.0–6.1 — which is shifted downward by approximately 0.3–0.4 units relative to the pKa optimum we observe for the same lipid panel in Hep3B cells. This is consistent with the hypothesis that neuronal late endosomes acidify to a lower terminal pH than hepatocyte late endosomes, requiring a lower ionizable lipid pKa to achieve maximal membrane disruption at the relevant pH.
3. Neuronal transduction efficiency
Functional delivery — actually getting cargo into the nucleus and achieving a biological effect — is the terminal readout that all upstream characterization is working toward. For CRISPR cargo, this means measuring indel frequency (for Cas9 nuclease delivery by ICE analysis or amplicon deep sequencing), base conversion efficiency at the target adenine or cytosine (for ABE8e or CBE4max delivery), or reporter expression (for mRNA delivery). In CNS LNP development, the relevant cell types are primary cortical neurons (DIV7–10), differentiated SH-SY5Y cells as a more tractable surrogate, and patient-derived iPSC-derived neurons for disease-relevant target evaluation.
Neuronal transduction efficiency is lower than in proliferating cell lines for any non-viral delivery system — this is an expected consequence of post-mitotic cell biology and is not evidence of formulation failure by itself. Primary cortical neurons at DIV8 transfected with the same LNP formulation that achieves 60% editing in HEK293 cells may show 8–15% indel frequency. That is the correct benchmark comparison to use: optimized CNS formulation versus non-optimized CNS formulation, not CNS neuron versus HEK293 cell. An LNP formulation optimized through the full expanded CNS panel should show measurably higher neuronal transduction efficiency than a standard LNP from the same base ionizable lipid chemistry.
4. RNA cargo integrity after encapsulation
mRNA and sgRNA are susceptible to nuclease degradation during the encapsulation process and to thermal or pH-induced hydrolysis during buffer exchange steps. The standard RiboGreen assay measures quantity but not quality. Bioanalyzer or TapeStation RNA Integrity Number (RIN) analysis on extracted cargo provides a direct measure of cargo integrity. For mRNA cargo, a RIN equivalent ≥ 7 and a single dominant high-molecular-weight band corresponding to the full-length transcript on a RNA nano chip are the minimum quality indicators. For sgRNA, a sharp single-band profile on a small RNA chip with no low-molecular-weight degradation products is the equivalent check. For prime editing pegRNAs — which are longer and more complex than standard sgRNAs — full-length profile confirmation is especially important since truncated pegRNAs lose spacer or 3'-RT template function depending on where truncation occurs.
Running cargo integrity analysis on extracted RNA from each batch is not optional for CNS formulations. The path from formulation to functional editing in primary neurons is long enough — BBB crossing, endosomal uptake, endosomal escape, nuclear delivery, editing activity — that degraded cargo will produce a clean-looking characterization panel followed by zero editing activity, with no intermediate signal to indicate where the failure occurred.
5. Apparent permeability (Papp) in transwell BBB models
Papp, the apparent permeability coefficient, is a quantitative measure of flux across the BBB model monolayer under defined conditions. It is calculated from the rate of appearance of fluorescent tracer on the basolateral side over time, normalized to membrane surface area and the apical starting concentration: Papp = (dQ/dt) / (A × C₀), where dQ/dt is the rate of accumulation on the basolateral side, A is the membrane area, and C₀ is the initial apical concentration. For CNS LNPs, Papp provides a single-number comparator across formulations that complements the percentage transcytosis efficiency readout. Reference compounds — lucifer yellow for paracellular integrity assessment, and Rhodamine 123 (a P-gp substrate) for P-gp efflux activity validation — should run in parallel in each experiment to confirm that the model is functioning as expected.
6. Protein corona characterization
After incubation in serum or whole blood, LNPs adsorb a protein corona that substantially modifies their surface chemistry within seconds. Proteomic analysis of the hard corona (proteins remaining after stringent washing) by LC-MS/MS after LNP isolation from serum incubate reveals which serum proteins are enriched and which are depleted relative to serum composition. For a CNS-targeted LNP, the critical question is whether the targeting ligand — TfR1-binding peptide or LRP1-targeting ApoE-mimetic — remains sterically accessible after corona formation, or is buried under enriched serum proteins.
This can be assessed by comparing receptor-binding activity of LNPs pre- and post-corona formation in a competitive binding assay, or by including anti-targeting-ligand antibody blocking experiments in the transcytosis assay after corona formation under serum conditions. A CNS LNP that achieves 60% transcytosis in serum-free conditions but loses activity completely in 10% serum has a surface engineering problem that no amount of ionizable lipid optimization will fix. Corona characterization is resource-intensive and is not practical for every batch QC, but it is essential during formulation development.
Building the Expanded Panel in Practice
The expanded characterization panel is not a replacement for the standard panel — it builds on it. Our current panel for CNS-LNP characterization runs in this sequence:
- Batch QC (every batch): DLS (z-average, PDI), zeta potential at pH 7.4, pKa by TNS assay, EE% by RiboGreen, morphology by cryo-TEM, RNA cargo integrity by Bioanalyzer/TapeStation
- CNS functional panel (new formulations and periodic batch testing): hCMEC/D3 transwell transcytosis assay with TEER and lucifer yellow validation; Gal8 endosomal escape assay in SH-SY5Y cells; neuronal transduction efficiency in primary cortical neurons (indel frequency or base conversion readout)
- Formulation development (during candidate screening): Papp quantification in hCMEC/D3 transwell with Rhodamine 123 P-gp control; protein corona characterization by LC-MS/MS; receptor binding competition assays pre- and post-corona formation
This panel produces formulations where the quality assessment is actually predictive of CNS delivery performance — not just predictive of particle formation. The standard panel tells you that a particle formed. The expanded panel tells you whether it will reach a neuron.
A Practical Note on Resource Constraints
Primary cortical neurons are time-consuming to isolate and maintain — rat E18 cortical dissociates require specific dissection technique and plating conditions, and are not usable for editing assays until DIV7–10. hCMEC/D3 transwell assays require cell culture expertise and take 72–96 hours from seeding to transcytosis readout; cell passage number must be controlled tightly (passages 25–35 are optimal for TEER reliability in this line). Gal8 assays require fluorescence microscopy capability and image analysis software for puncta quantification. Not every LNP group has immediate access to all of these, particularly early in a program.
The practical recommendation is to tier the panel by program stage. At the formulation screening stage — evaluating a library of ionizable lipid candidates — the Gal8 endosomal escape assay in SH-SY5Y cells provides the most information per unit of effort. It is faster than primary neuron work, requires less specialized cell culture expertise, and discriminates between ionizable lipid candidates on the CNS-relevant functional metric. Transwell transcytosis assays follow once the ionizable lipid is selected. Primary neuron editing data is the final confirmation step for candidates selected by both prior screens.
We are not suggesting that these assays are beyond reach for early-stage CNS LNP programs. We are suggesting that building the expanded panel capability alongside formulation development — rather than attempting to retrofit it after a failed in vivo experiment — is the more efficient path. The specific characterization protocols we use are summarized on the LNP Technology platform page.
