Delivering Base Editors: Why Cargo Size Changes Everything

Base editing is not the same as standard CRISPR-Cas9 — and the delivery problem reflects that difference

Base editing and prime editing are often grouped with standard CRISPR-Cas9 as members of the "CRISPR-based gene editing" toolbox. For delivery purposes, this grouping is misleading. The molecular cargo that must reach the cell nucleus for base editing is physically larger, biochemically more complex, and more sensitive to delivery conditions than the standard SpCas9 + sgRNA combination. Understanding why cargo size changes the delivery optimization problem — and what that means for LNP formulation — requires examining the base editor architecture in detail.

What base editors are and what they deliver

Base editors perform single-nucleotide conversions in the genome without introducing a double-strand DNA break (DSB). The two major base editor classes are:

Cytosine base editors (CBEs) convert cytosine (C) to thymine (T) at target positions. The core CBE architecture fuses a catalytically impaired Cas9 (nCas9, which makes a single-strand nick rather than a DSB) to a cytidine deaminase enzyme (APOBEC1 in early CBE designs; evolved TadA or APOBEC3A in more recent CBE3max and related variants). The cytidine deaminase converts C to U in the R-loop generated by the nCas9-sgRNA complex; U is subsequently converted to T through base excision repair or DNA replication.

Adenine base editors (ABEs) convert adenine (A) to guanine (G). ABEs fuse nCas9 to an engineered adenine deaminase (the TadA domain, evolved through multiple directed evolution cycles in the ABE7.10 → ABE8e lineage). ABE8e achieves A-to-G editing efficiencies of 50–95% at target positions within the editing window (positions 4–8 of the protospacer, counted from the PAM-distal end), comparable to or exceeding SpCas9 indel rates for standard cleavage.

Prime editors (PEs) perform any transversion or transition and can insert or delete short sequences without a DSB. PEs fuse nCas9 to an engineered reverse transcriptase (RT) domain (derived from MMLV reverse transcriptase in the PE2 design). The pegRNA (prime editing guide RNA) carries both the target-binding sequence and the RT template encoding the desired edit. PE3+ designs add a second sgRNA to nick the non-edited strand and improve editing efficiency.

The functional difference between these editor classes and standard SpCas9 for delivery is the protein cargo size and the guide RNA complexity:

Why cargo size matters for LNP formulation

LNP-based delivery of base editors and prime editors uses mRNA (encoding the editor protein) co-encapsulated with, or separately co-administered alongside, the guide RNA (sgRNA or pegRNA). For LNP delivery, the packaging constraint is not the ~4.7 kb AAV cap — LNPs accommodate mRNAs of any length. ABE8e mRNA of ~5.5 kb coding region (plus 5'/3' UTR, poly-A tail, making the full-length transcript approximately 6.5–7.0 kb) is routinely encapsulated in LNPs at encapsulation efficiencies comparable to shorter mRNA species.

However, larger mRNA cargo does change the LNP formation and encapsulation physics in ways that matter for formulation optimization.

Encapsulation efficiency as a function of mRNA length. Longer mRNA molecules have more phosphate groups per molecule and require more ionizable lipid contacts per RNA strand for complete encapsulation at a given N/P ratio. The encapsulation efficiency of long mRNA (5–7 kb) is more sensitive to N/P ratio than short mRNA (<2 kb), because incomplete charge saturation at a given N/P ratio is more likely to leave exposed RNA segments when the molecule is longer. Formulations optimized for encapsulating short siRNA or mRNA of 2–3 kb must be re-validated for larger mRNA cargo rather than assumed to transfer directly.

Particle size distribution. Longer mRNA molecules introduce more hydrodynamic volume into the LNP interior during assembly. At fixed lipid concentration and N/P ratio, LNPs encapsulating 6.5 kb mRNA tend to form slightly larger particles than LNPs encapsulating 2 kb mRNA — because the mRNA-ionizable lipid condensate that constitutes the aqueous core has greater volume. Particle size shifts of 10–20 nm (from, say, 90 nm to 105 nm) are within the acceptable range for hepatic delivery (hepatic sinusoidal fenestrae 100–180 nm) but require re-characterization. Formulations targeting the tight 80–100 nm range may drift into the 95–120 nm range when the mRNA cargo is doubled in length.

RNA integrity during encapsulation. Longer mRNA strands are more vulnerable to mechanical shear during microfluidic synthesis. The ethanol injection process subjects the RNA solution to high fluid shear in the mixing channel. Full-length mRNA of 6.5 kb has a higher probability of strand fragmentation under high-shear conditions than a 2 kb mRNA — simply because the longer molecule spans more fluid gradient. Optimizing the microfluidic synthesis for minimum shear (lower flow rate, lower ethanol concentration, optimized channel geometry) is more critical for base editor mRNA than for standard Cas9 mRNA.

The guide RNA coformulation problem

For base editors and prime editors, the guide RNA is not just a ~100 nt sgRNA. Prime editing pegRNAs are 150–250 nt, encoding both the spacer sequence and the RT template. PE3+ strategies require co-delivery of both the pegRNA and a second nicking sgRNA — two distinct RNA species that must reach the same cell simultaneously.

Co-encapsulation of two distinct RNA species in a single LNP particle is possible but requires formulation optimization for the specific RNA combination. The ratio of mRNA to guide RNA in the particle, their relative encapsulation efficiencies, and the intracellular stability of each species once released from the endosome all affect the editing outcome. Excess sgRNA relative to Cas9 protein (after mRNA translation) maximizes Cas9 occupancy with guide RNA and minimizes unguided Cas9 activity. Excess Cas9 mRNA relative to sgRNA maximizes the pool of translated Cas9 available for RNP complex formation. The optimal ratio depends on the relative translation efficiency and intracellular stability of the mRNA versus the guide RNA in the specific target cell type.

For base editor programs, the guide RNA chemistry is also relevant. Chemically modified sgRNAs (2'-O-methyl and phosphorothioate backbone modifications at the 3' end) are substantially more stable in the cytoplasm than unmodified sgRNA and are standard for clinical CRISPR programs. The LNP encapsulation efficiency for chemically modified sgRNA can differ from unmodified sgRNA — the 2'-O-methyl modification reduces the RNA's negative charge density, changing the electrostatic interaction with ionizable lipid during encapsulation. This is another dimension of formulation optimization specific to base editor delivery that does not apply to standard Cas9 delivery programs.

How LNP formulation optimization changes for base editor programs

In a standard SpCas9 LNP delivery program, the formulation optimization target is well-defined: maximize hepatocyte transfection efficiency (luciferase mRNA reporter in HepG2 or primary hepatocytes), subject to particle size, PDI, and EE constraints. The formulation that maximizes luciferase expression reliably predicts high SpCas9 mRNA delivery.

For base editor programs, the same transfection screen is necessary but not sufficient. Additional considerations:

• RNA integrity of the specific mRNA cargo matters more. ABE8e mRNA at 6.5 kb is larger than a luciferase reporter mRNA (typically 1.8 kb). A formulation optimized for luciferase reporter delivery must be re-validated for ABE8e mRNA delivery, with specific attention to whether the EE and particle size results transfer to the longer cargo.
• Co-encapsulation ratio optimization. For programs using mRNA + chemically synthesized sgRNA co-encapsulation (versus separate co-administration), the formulation must be optimized for both the mRNA:sgRNA encapsulation ratio and the total RNA:lipid ratio.
• Editing efficiency assay, not just transfection efficiency. Luciferase reporter assay confirms LNP delivery of mRNA and translation. Actual base editing efficiency (measured by next-generation sequencing of the target locus — EditR analysis or Sanger sequencing with EditR deconvolution) is needed to confirm that translated base editor protein is functional and that the guide RNA was co-delivered and is active. A formulation that delivers lots of ABE8e mRNA but fails to co-deliver functional sgRNA at the right intracellular ratio will show high luciferase (from a reporter mRNA) but low editing efficiency at the target locus.

The Gendelivr computational screening pipeline was designed primarily for standard ionizable lipid LNP optimization targeting hepatocyte transfection efficiency. For base editor programs, the in silico pre-screen logic applies for the same formulation parameter space (ionizable lipid pKa, helper lipid composition, PEG density, particle size targets) — but the bench validation protocol must include cargo-specific characterization steps that go beyond the standard luciferase transfection endpoint.

The 2025 base editor delivery landscape

As of mid-2025, LNP-delivered base editing programs have advanced to clinical stage for several hepatic targets. These programs have confirmed that LNP delivery of ABE8e and CBE mRNA in conjunction with chemically synthesized sgRNA is technically feasible, safe, and achieves durable hepatic gene correction in human subjects at doses compatible with the therapeutic window. The clinical data from hepatic transthyretin and other programs has validated the in vivo pharmacology of LNP-delivered base editors.

The formulation development challenge for new base editor programs is not whether LNP delivery of 5.5 kb base editor mRNA is possible — it is demonstrated. The challenge is achieving consistent high editing efficiency across the intended patient population, with an LNP formulation that maintains drug product quality through the manufacturing and clinical administration process. This is the formulation optimization problem that computational pre-screening is designed to accelerate: not whether the approach works, but which specific formulation composition maximizes efficiency and quality within the window that clinical development requires.