# Chapter 157. Comparative Non-LNP RNA Delivery: Conjugates, Polymers, Peptides, Vectors, Extracellular Vesicles, and Tissue Barriers

## Scope Note

This chapter compares RNA delivery modalities that are not primarily conventional lipid nanoparticles: receptor-targeted conjugates, ligand-decorated systems, cationic and degradable polymers, dendrimers, inorganic carriers, hybrid materials, viral vectors, virus-like particles, extracellular vesicles, and extracellular-vesicle-inspired systems. [Chapter 156](chapter1139.md) owns lipid nanoparticle chemistry, ionizable-lipid behavior, endosomal trafficking and escape, and lipid-driven biodistribution. The therapeutic modality chapters, including [Chapter 150](chapter1134.md), [Chapter 151](chapter1135.md), [Chapter 153](chapter1137.md), and [Chapter 154](chapter1138.md), own product pharmacology for antisense, small interfering RNA, messenger RNA, and programmable products. The present chapter owns the comparative question: given a defined RNA cargo, tissue, route, intracellular destination, repeat-dose requirement, and manufacturing and safety boundary, which non-LNP delivery strategy is mechanistically plausible and what evidence would justify choosing it? [Chapter 158](chapter1141.md) treats integrated pharmacokinetics, pharmacodynamics, immunogenicity, toxicology, and regulatory science.

## Executive Summary

RNA delivery is the art of moving a charged, nuclease-sensitive, often immunostimulatory macromolecule from a dose container to the cellular compartment where the RNA can act. Delivery is not a single step. A useful delivery system must survive formulation and storage, remain chemically intact in biological fluid, avoid premature renal filtration or phagocytic clearance when systemic exposure is needed, cross vascular and tissue barriers, bind or enter target cells, escape or bypass degradative compartments, release an active RNA species, and do all of this with a safety margin wide enough for the disease context. Each delivery modality solves only some of these problems.

Ligand-conjugated oligonucleotides are the clearest success case for receptor-targeted RNA delivery. The best-established example is triantennary N-acetylgalactosamine, usually abbreviated GalNAc, conjugation to antisense or siRNA drugs for hepatocyte uptake through the asialoglycoprotein receptor. This strategy works because the liver sinusoid is fenestrated, the receptor is abundant on hepatocytes, the ligand-receptor interaction is efficient, and chemically stabilized oligonucleotides can act after productive endosomal release. The same logic is harder outside the liver: many extrahepatic receptors have lower accessible expression, less favorable trafficking, tighter endothelial barriers, stronger uptake by macrophages, or poorer endosomal escape.

Antibodies, peptides, aptamers, lipids, vitamins, sugars, and small molecules can all be used as targeting ligands. A ligand may increase cell association without increasing functional delivery, because binding and internalization are not equivalent to cytosolic or nuclear access. The design question is therefore not "does the vehicle bind the target?" but "does the complete route from dose to active RNA improve enough to justify added complexity?" Assays that measure total tissue radioactivity, fluorescent particle uptake, or bulk RNA abundance often overstate targeting if they do not distinguish surface-bound material, trapped endosomal cargo, degraded RNA fragments, and productive pharmacology.

Polymers, dendrimers, inorganic particles, and hybrid systems expand the design space beyond lipid nanoparticles. Cationic polymers can condense RNA into polyplexes; degradable polymers can tune release; dendrimers offer controlled branching and multivalency; inorganic cores can support imaging, photothermal effects, magnetic guidance, or high surface area. These platforms are powerful experimentally, but their clinical translation is constrained by toxicity, complement activation, batch heterogeneity, slow biodegradation, scale-up reproducibility, and sometimes limited endosomal escape.

Viral vectors, virus-like particles, exosomes, and extracellular vesicle-inspired systems exploit biological entry pathways. Viral vectors can be exceptionally efficient for gene transfer or RNA expression cassettes, but payload size, pre-existing immunity, integration risk for some vector classes, redosing limitations, manufacturing complexity, and long persistence may make them mismatched for transient RNA pharmacology. Virus-like particles and engineered capsids can package RNA or RNA-protein complexes without full viral replication. Exosomes and extracellular vesicles are attractive because cells naturally secrete membrane-bound carriers containing RNA and protein, but therapeutic use requires careful distinction between biological plausibility, measurable cargo loading, and reproducible functional delivery.

Tissue targeting is not a label attached to a particle; it is the net result of anatomy, vascular permeability, extracellular matrix, receptor accessibility, immune surveillance, disease state, route of administration, cargo chemistry, and dose. Liver is the easiest systemic target for several RNA formats. Eye and central nervous system can be reached by local dosing in selected compartments. Lung is accessible by inhalation but faces mucus, surfactant, macrophages, and epithelial barriers. Muscle is large and poorly fenestrated for many systemic particles. Tumors are heterogeneous and cannot be assumed to share one enhanced permeability and retention rule. Placenta and fetal-interface delivery raise distinctive safety and biodistribution thresholds.

The practical decision among delivery modalities depends on cargo size and chemistry, tissue and target cell, route, dominant anatomical barrier, required uptake and release path, duration of action, repeat-dosing need, acceptable immune stimulation, manufacturing scale, analytical control, and clinical risk tolerance. These variables form a modality-by-tissue decision matrix, not a platform ranking. A delivery system that looks elegant in a cell-culture uptake assay may be inappropriate for a chronic systemic drug, whereas a complex but efficient system may be reasonable for a severe disease treated once, locally, or ex vivo. Delivery should therefore be evaluated as a complete pharmacological system rather than as a decorative targeting component.

## Concept Inventory

- **Receptor-targeted conjugate:** an RNA or oligonucleotide covalently attached to a ligand that binds a cell-surface receptor and promotes uptake. The ligand may be a sugar, peptide, antibody fragment, aptamer, lipid, vitamin, or small molecule. The key boundary is that receptor engagement must produce functional intracellular delivery, not merely binding.
- **GalNAc conjugate:** an oligonucleotide conjugate bearing N-acetylgalactosamine residues, often in a multivalent arrangement, used to bind the asialoglycoprotein receptor on hepatocytes. GalNAc conjugation is a leading example of a clinically validated receptor-targeted RNA delivery strategy, especially for siRNA and antisense formats (Springer and Dowdy, 2018; Nair et al., 2014; Roberts et al., 2020).
- **Polyplex:** a complex formed when a cationic polymer electrostatically condenses an anionic nucleic acid. Polyplex behavior depends on polymer charge density, molecular weight, degradability, hydrophobicity, buffering capacity, particle size, and serum interaction.
- **Dendrimer:** a branched macromolecule with controlled generations and surface groups. Dendrimers can display many positive charges or ligands, but multivalency can also amplify toxicity and nonspecific protein adsorption.
- **Viral vector:** a viral delivery system engineered to carry a therapeutic genetic payload while removing or disabling pathogenic replication functions. Viral vectors can deliver RNA indirectly by carrying DNA expression cassettes, or directly in specialized RNA virus or virus-like systems.
- **Virus-like particle:** a particle assembled from viral structural proteins that resembles a virus in architecture but lacks a complete infectious viral genome. Virus-like particles can be engineered to package RNA, messenger RNA, guide RNA-protein complexes, or other nucleic acid cargo.
- **Exosome and extracellular vesicle:** membrane-bound vesicles released by cells. Exosomes are often defined operationally as small extracellular vesicles associated with endosomal multivesicular body origin, but routine preparations often contain mixed vesicle populations. This chapter uses "extracellular vesicle" when biogenesis is not proven.
- **Tissue targeting:** preferential functional delivery to a tissue or cell type relative to non-target tissues. Functional targeting requires evidence of active RNA effect, not just particle accumulation.
- **Endosomal escape:** movement of internalized RNA cargo from endosomes into the cytosol or another productive compartment before degradation or recycling. Endosomal escape is frequently the limiting step after receptor-mediated uptake.

## What to Know Before Reading This Chapter

RNA therapeutics differ from many small-molecule drugs because the active agent is large, charged, and usually unable to passively diffuse across cell membranes. Oligonucleotides such as antisense oligonucleotides and siRNAs can be chemically stabilized and can sometimes act at low intracellular copy numbers. Messenger RNA and self-amplifying RNA are much larger and must reach the cytosol intact enough to be translated. Guide RNAs for editing or CRISPR systems may need to arrive with, or be expressed alongside, a protein enzyme.

Readers should distinguish four measurements that are often confused. Biodistribution measures where administered material goes. Cell uptake measures association with or internalization into cells. Endosomal escape measures release from vesicular compartments into a productive compartment. Pharmacodynamic effect measures the intended RNA-driven biological change, such as target mRNA knockdown, splice correction, protein expression, or editing. A delivery platform can score well on the first two measurements and still fail on the last two.

This chapter assumes familiarity with lipid nanoparticle mechanisms from [Chapter 156](chapter1139.md), especially ionizable lipid charge switching, apolipoprotein-mediated liver tropism, and endosomal trafficking. It also assumes basic understanding of antisense, siRNA, mRNA, and editing modalities from Chapters [149](chapter1133.md) to [154](chapter1138.md). The present chapter uses those modalities as cargo examples rather than reintroducing all therapeutic mechanisms.

## 157.1. GalNAc and other receptor-directed oligonucleotide conjugates

A receptor-targeted conjugate is the simplest conceptual form of active delivery: attach a ligand to an RNA drug, let the ligand bind a cell-surface receptor, and rely on receptor-mediated endocytosis to bring the RNA into the cell. A ligand is a molecule recognized by a receptor. A receptor is a cellular molecule, usually a membrane protein, that binds a ligand and triggers uptake, signaling, adhesion, or another cellular response. For RNA therapeutics, the useful receptor is not merely abundant; the receptor must be accessible from the dosing route, internalize efficiently, recycle or traffic in a way that leaves some cargo able to escape degradation, and be expressed in the desired cells more than in dose-limiting off-target cells.

GalNAc conjugation is the benchmark for receptor-targeted oligonucleotide delivery. N-acetylgalactosamine is a sugar recognized by the asialoglycoprotein receptor, a hepatocyte-enriched receptor that clears desialylated glycoproteins from blood. GalNAc conjugates usually use multivalent GalNAc display, often triantennary architecture, because clustered ligands bind the receptor far more avidly than a single sugar. The RNA cargo is typically a chemically stabilized siRNA duplex or antisense oligonucleotide. After subcutaneous injection, the conjugate reaches the liver through the circulation, encounters fenestrated liver sinusoidal endothelium, binds hepatocyte asialoglycoprotein receptor, enters by clathrin-associated endocytosis and related uptake routes, and then faces the major bottleneck: only a small fraction of internalized oligonucleotide reaches the cytosol or nucleus where RNA interference or RNase H recruitment can occur (Springer and Dowdy, 2018; Nair et al., 2014; Willoughby et al., 2018; Roberts et al., 2020).

The original multivalent GalNAc-siRNA study illustrates the causal evidence behind this model while also defining its limits. In primary mouse hepatocytes, uptake was reduced by excess free GalNAc ligand, calcium chelation, and loss of the ASGR2 receptor subunit; triantennary GalNAc outperformed a biantennary design in that assay. In mice, liver uptake and transthyretin silencing were greater after subcutaneous than intravenous administration of the tested conjugate, consistent with slower absorption extending receptor exposure, and closely spaced doses showed that receptor recycling could support repeated uptake. Those experiments link ligand architecture, receptor dependence, route, tissue exposure, and RNA interference in one chain, but they are preclinical, design-specific results rather than universal dose or saturation rules for all GalNAc drugs (Nair et al., 2014).

Figure 157.1 summarizes this route as a chain of conditional steps rather than a single targeting event.

![Figure 157.1. Conditional route of a receptor-targeted RNA conjugate](../assets/figures/chapter1140_figure1.png)

**Figure 157.1. Conditional route of a receptor-targeted RNA conjugate.** A receptor-targeted RNA conjugate is not delivered when it binds a receptor; it is delivered only when a sufficient fraction of intact cargo reaches the productive compartment. GalNAc-ASGPR hepatocyte delivery is shown as the reference example, with extrahepatic decision points marked where other receptor systems often fail.

The strength of GalNAc delivery is not only ligand affinity. The whole biological context is favorable. Hepatocytes face blood across a discontinuous, fenestrated endothelium. The target receptor is highly expressed and naturally specialized for ligand internalization. Many liver-expressed therapeutic targets are accessible to oligonucleotide mechanisms. Chemically stabilized siRNAs can sustain RNA-induced silencing complex activity after limited productive release, and antisense gapmers can act catalytically through RNase H once they reach a productive nuclear or cytosolic compartment. This combination explains why GalNAc should not be treated as a generic proof that any receptor ligand can solve extrahepatic delivery.

Other receptor-targeted conjugates try to extend the same principle to different tissues. Examples include transferrin receptor ligands for endothelial, muscle, or central nervous system-related delivery; integrin ligands for angiogenic vasculature or tumors; folate receptor ligands for selected tumor or inflammatory cells; mannose and other carbohydrate ligands for antigen-presenting cells; and peptides or small molecules that bind receptors enriched in muscle, kidney, or immune-cell subsets. The mechanistic questions are the same for each system: Is the receptor exposed to the drug in vivo? Is the receptor on the desired cell type rather than only on nearby endothelial cells, macrophages, or stromal cells? Does receptor internalization send the cargo to a compartment compatible with escape? Does ligand attachment alter RNA stability, protein binding, renal clearance, or immune recognition?

Table 157.1 should be read as a design map rather than a ranked list of "good" receptors.

**Table 157.1. Receptor-targeted conjugate examples and evidence requirements.** Receptor-targeted conjugates differ in accessible tissue, ligand, RNA cargo, trafficking route, and functional endpoint; receptor binding or uptake is insufficient without productive intracellular activity and artifact controls.

| Receptor or target class | Ligand example | Intended tissue or cell type | Compatible RNA cargo | Required functional endpoint | Major artifact or caveat | Citation status |
| --- | --- | --- | --- | --- | --- | --- |
| **Asialoglycoprotein receptor** | Clustered GalNAc, commonly triantennary | Hepatocytes reached through fenestrated liver sinusoids | Chemically stabilized siRNA or antisense oligonucleotide | Receptor-dependent hepatocyte uptake plus target mRNA knockdown, RNase H activity, or durable RNAi effect | Liver accumulation can reflect non-hepatocyte uptake, and only a small internalized fraction escapes endosomes; route and saturation results are construct-specific | Direct anchors: Nair et al. 2014; Springer and Dowdy 2018; Willoughby et al. 2018; Roberts et al. 2020 |
| **Transferrin receptor and related transcytosis targets** | Antibody fragment, peptide, or transferrin-mimetic ligand | Brain endothelium, muscle, or other receptor-accessible barriers | Stabilized oligonucleotide or ligand-decorated particle cargo | Cell-type-resolved RNA effect beyond endothelial binding or lysosomal trapping | Receptor-mediated uptake can stop in endothelium and may perturb iron-transport physiology | Curate transferrin-receptor delivery and extrahepatic conjugate sources |
| **Folate receptor** | Folate or folate-decorated carrier | Folate-receptor-positive tumors or activated inflammatory cells | siRNA, antisense oligonucleotide, or particle-loaded RNA | Target knockdown or splice/protein change in sorted malignant or immune target cells | Folate receptor expression is heterogeneous and can mark inflammatory or stromal cells rather than malignant cells | Curate folate-targeted RNA delivery and tumor-cell specificity sources |
| **Mannose receptor and other lectin targets** | Mannose-rich glycan or clustered carbohydrate ligand | Macrophages, dendritic cells, and selected lymphoid or inflamed tissues | siRNA, mRNA, or antisense cargo in conjugates or particles | Antigen-presenting-cell-specific RNA effect or intended immune modulation | Phagocytic uptake may be desired for vaccines but harmful for tolerogenic or gene-correction uses | Curate mannose-receptor and carbohydrate-targeted delivery sources |
| **Integrin target class** | RGD-like or other integrin-binding peptide | Angiogenic endothelium, tumors, fibrotic tissue, or inflamed tissue | Oligonucleotide conjugate or ligand-decorated nanoparticle | Functional RNA effect in the target cell population rather than vascular association alone | Integrins are shared by endothelium, stroma, immune cells, and tumor cells, and disease state changes accessibility | Curate integrin-targeted RNA delivery and tumor-vascular targeting sources |
| **Receptor-specific small-molecule or transporter target** | Vitamin, metabolite, enzyme-substrate mimic, or tissue-homing small molecule | Receptor- or transporter-enriched parenchymal or immune-cell subset | Stabilized oligonucleotide or compact carrier cargo | Dose-dependent pharmacodynamic gain over a matched non-targeted construct | Species differences, receptor saturation, and off-target transporter expression can erase apparent selectivity | Curate receptor-specific small-molecule and transporter-targeting sources |

The main evidence basis for receptor targeting includes binding assays, receptor competition experiments, receptor knockout or knockdown studies, tissue and cell-type biodistribution, microscopy with surface-stripping controls, pharmacodynamic readouts, and dose-response comparison against non-targeted controls. A receptor-targeted claim is strongest when excess free ligand blocks uptake, genetic loss of the receptor reduces both uptake and functional effect, the target cell type shows the expected RNA mechanism, and a nonbinding conjugate fails under the same conditions. A claim is weaker when it rests on fluorescent puncta in cultured cells, because many fluorescently labeled conjugates accumulate in endosomes or lysosomes and the dye may not report intact RNA.

Several artifacts recur. A high-affinity ligand can become a sink that traps conjugate in a nonproductive cell population. Receptor expression measured by bulk RNA sequencing may not reflect receptor protein abundance at the accessible plasma membrane. Ligand density can change particle size, serum protein corona, complement activation, and macrophage uptake. Receptor binding can saturate, so a low dose may look targeted while a high dose spills into nonspecific uptake pathways. Some receptors are expressed in disease tissue but not in healthy tissue, or the reverse; this matters for translation from animal models to human patients.

The boundary between conjugate delivery and chemistry is also important. For short oligonucleotides, chemical modifications such as phosphorothioate linkages, 2′ sugar modifications, locked nucleic acids, or morpholino backbones can strongly influence plasma protein binding, tissue distribution, nuclease resistance, and intracellular trafficking. A ligand may improve delivery only in the context of a specific oligonucleotide chemistry. Conversely, a poorly stabilized RNA may be cleared or degraded before receptor targeting can matter. [Chapter 149](chapter1133.md) treats the chemistry; this chapter treats the delivery consequences.

The consensus is that GalNAc-ASGPR delivery is established for hepatocyte-targeted oligonucleotides, while most other receptor-targeted conjugate strategies remain context-dependent or investigational. The open question is not whether receptors can be used; they clearly can. The open question is which receptor-ligand-cargo combinations provide enough functional intracellular delivery, tissue selectivity, repeat-dose tolerability, and manufacturable consistency to become general extrahepatic platforms (Roberts et al., 2020; Paunovska et al., 2022; Anand et al., 2025).

## 157.2. Antibody, peptide, aptamer, lipid, and small-molecule targeting ligands

Ligands differ in size, valency, specificity, immunological profile, manufacturing route, and the way they change the physical behavior of the RNA carrier. Antibodies and antibody fragments offer high-affinity binding to cell-surface proteins and can exploit decades of biologics engineering. Peptides can be short, modular, and synthetically accessible. Aptamers are folded nucleic acids selected to bind targets through shape and chemistry. Lipid ligands can promote membrane association or albumin binding. Vitamins, metabolites, sugars, hormones, and other small molecules can hijack natural transport pathways. Each class is a tool for changing one or more steps in the delivery route, but none automatically solves the endosomal escape problem.

Antibody-based RNA delivery includes several architectures. An antibody can be covalently linked to an oligonucleotide, fused or assembled with an RNA-binding domain, attached to a nanoparticle surface, or used as a bispecific bridge between a cell receptor and a nucleic acid carrier. The appeal is cell-type specificity: a monoclonal antibody can recognize an antigen enriched on a tumor cell, immune subset, muscle cell, or endothelial cell. The challenge is that many antibody-antigen complexes internalize into endosomes and lysosomes, where RNA may be degraded unless the conjugate is engineered for release. Antibody size can also reduce tissue penetration, increase exposure in antigen-expressing off-target cells, and complicate stoichiometric control of RNA loading (Malecova et al., 2023; Roberts et al., 2020).

Peptide ligands occupy a broad design space. A targeting peptide binds a receptor or tissue component. A cell-penetrating peptide, sometimes called a CPP, is designed to promote cellular entry through electrostatic interaction, membrane perturbation, or endocytic uptake. An endosomolytic peptide is designed to disrupt endosomal membranes, often by changing conformation or charge state at acidic pH. These functions can be combined but should not be conflated. A peptide that increases cell-associated fluorescence may be acting as a sticky cationic surface rather than as a true transporter. A peptide that disrupts membranes may also damage plasma membranes, activate innate immunity, or cause hemolysis at high concentration.

Figure 157.2 compares common ligand-display architectures and the failure points that separate binding from pharmacology.

![Figure 157.2. Ligand-display architectures for RNA delivery](../assets/figures/chapter1140_figure2.png)

**Figure 157.2. Ligand-display architectures for RNA delivery.** Targeting ligands differ in size, valency, manufacturability, pharmacokinetics, and intracellular fate. Valency is not a one-directional targeting dial: it can shift injection-site retention, circulation, organ exposure, and self-association. The figure separates ligand binding from internalization, release, and RNA activity so that active targeting is not confused with surface association.

Aptamers are especially conceptually attractive for RNA delivery because the targeting ligand is itself a nucleic acid. An aptamer can be selected in vitro to bind a protein, small molecule, or cell surface, then linked to an siRNA, antisense region, ribozyme, or other RNA module. Aptamer-siRNA chimeras show the idea clearly: one region binds a cell-surface target, while the cargo region must remain accessible for Dicer processing when the design requires it and for loading into the RNA-induced silencing complex. Receptor-mediated endocytosis does not itself explain how the chimera escapes the endosome, and the escape mechanism remains poorly resolved for many aptamer systems. The main boundaries are therefore nuclease stability, chemical-modification compatibility with aptamer folding and cargo processing, target accessibility in vivo, and the difference between local or intratumoral efficacy and systemic delivery. Aptamers can also bind unintended proteins, and some apparent cell-specific aptamers selected on cultured cells lose specificity in serum or in animal tissues (Sivakumar et al., 2019).

Lipid and hydrophobic ligands are not the same as lipid nanoparticles. A cholesterol-conjugated siRNA, for example, can bind serum lipoproteins and albumin, changing circulation time and tissue uptake. Fatty acids, tocopherol, and other hydrophobic moieties can tune association with plasma proteins and membranes. This can improve exposure in some tissues but may also increase uptake by liver, spleen, and phagocytic cells. Hydrophobic conjugation is often useful for small oligonucleotides because it changes pharmacokinetics without requiring a large carrier. It is harder to generalize to mRNA because mRNA size, fragility, and innate immune activation usually require more protective formulation.

Ligand valency is not a one-directional potency dial. In a mouse study of mono-, di-, and trimeric fatty-acid-siRNA conjugates, increasing valency changed hydrophobicity and self-association, but it also changed absorption and organ exposure in different directions: trivalent constructs were retained mainly at the subcutaneous injection site, monovalent constructs entered the circulation rapidly and accumulated predominantly in kidney, and divalent constructs showed intermediate behavior with preferential liver accumulation and functional distribution to selected extrahepatic tissues. Chemical structure predicted productive extrahepatic silencing better than bulk hydrophobicity alone. The transferable lesson is that valency, linker, and ligand chemistry must be evaluated through pharmacokinetics and cell-specific pharmacodynamics; the particular organ pattern is construct-, dose-, route-, species-, and target-dependent (Biscans et al., 2019).

Small-molecule ligands include vitamins, receptor ligands, enzyme substrates, and tissue-homing molecules. Folate-targeted systems exploit folate receptor expression in selected cancers and activated macrophages. Mannose-targeted systems aim at mannose receptors on macrophages and dendritic cells. Other examples include integrin-binding motifs, ligands for low-density lipoprotein receptor family members, and molecules that bind transporters. The danger is that a receptor name can create a false sense of precision. Many receptors have broad expression, change during inflammation, or localize to inaccessible surfaces. A ligand may direct a carrier to endothelial cells lining a tissue rather than to parenchymal cells within that tissue.

Table 157.2 places ligand classes alongside the cargo and evidence issues they most often raise.

**Table 157.2. Ligand classes for RNA delivery.** Antibodies, peptides, aptamers, lipids, sugars, vitamins, and small molecules offer different targeting, trafficking, immunogenicity, and manufacturing tradeoffs; confirmatory assays must measure the relevant productive-delivery step.

| Ligand class | Usual attachment format | Main advantage | Main delivery bottleneck | Safety or immunogenicity issue | Manufacturing issue | Best confirmatory assay |
| --- | --- | --- | --- | --- | --- | --- |
| **Full antibody** | Covalent antibody-oligonucleotide conjugate or antibody displayed on a particle surface | High-affinity recognition and mature biologics engineering | Internalized antibody cargo often traffics to endosomes and lysosomes before RNA release | Anti-drug antibodies, Fc-mediated uptake, and antigen sinks can alter exposure | Conjugation stoichiometry, aggregation, linker stability, and batch comparability | Matched binding-defective control plus target-cell pharmacodynamic endpoint |
| **Antibody fragment or bispecific binder** | Fragment-oligonucleotide conjugate, bispecific bridge, or RNA carrier adaptor | Smaller size and tunable valency relative to whole antibody | Shorter half-life or weaker internalization can reduce productive exposure | Immunogenicity and off-target antigen binding remain target-specific risks | Expression, purification, conjugation-site control, and fragment stability | Receptor competition or knockout with intact-cargo detection and RNA mechanism readout |
| **Cell-penetrating peptide** | Direct peptide-oligonucleotide conjugate or peptide-polyplex | Compact synthetic module that can increase cell association | Uptake can be nonspecific and strongly serum-dependent | Polycationic membrane injury, hemolysis, complement activation, or inflammation | Sequence purity, conjugation ratio, and proteolytic stability | Uptake, viability, and pharmacodynamic effect compared with scrambled or inactive peptide |
| **Endosomolytic peptide** | pH-responsive or fusogenic peptide linked to cargo or carrier | Addresses the post-uptake escape bottleneck directly | Productive escape must occur without damaging the plasma membrane | Membrane disruption can cause cytotoxicity and inflammatory danger signals | Activation pH, linker stability, and coformulation reproducibility | Quantitative endosomal-release assay tied to knockdown, splice correction, or protein expression |
| **Aptamer** | Aptamer-siRNA chimera or aptamer-linked oligonucleotide | Targeting domain and therapeutic RNA can be encoded in related nucleic acid chemistry | Folding, serum stability, endosomal escape, and in vivo target accessibility can fail outside selection conditions | Modified or double-stranded regions may trigger innate sensing or bind unintended proteins | Chemical modification and junction design must preserve aptamer folding and any required cargo processing | Binding in biological fluid plus route-matched target-cell RNAi, antisense, or editing endpoint |
| **Cholesterol or other hydrophobic ligand** | Cholesterol, fatty acid, tocopherol, or lipid conjugated to a small oligonucleotide | Alters albumin, lipoprotein, membrane association, and absorption without a large carrier | Valency and structure can shift retention toward the injection site or exposure toward kidney, liver, spleen, or phagocytic cells instead of the intended tissue | Hydrophobic aggregation and protein-corona effects can increase clearance or immune recognition | Control of valency, hydrophobic impurities, self-association, linker chemistry, and construct identity | Intact-cargo pharmacokinetics and cell-resolved pharmacodynamics across matched valencies and an unconjugated control |
| **Folate, vitamin, or small-molecule ligand** | Compact ligand attached directly or displayed on a carrier | Chemically defined targeting element with low added mass | Receptor heterogeneity and endosomal trapping limit functional delivery | Off-target receptor expression in inflammatory or normal tissues can narrow safety margin | Ligand stoichiometry, linker stability, and preservation of receptor affinity | Receptor competition with sorted-cell or spatial pharmacodynamic readout |
| **Carbohydrate ligand** | GalNAc, mannose, or other clustered sugar display | Exploits lectin and scavenger-receptor uptake pathways | Requires accessible receptor, suitable trafficking, and productive release | Macrophage, Kupffer-cell, or dendritic-cell uptake may dominate unexpectedly | Multivalent architecture, stereochemistry, and batch consistency | Receptor-loss or competition experiment plus cell-type-resolved functional delivery |

Good ligand experiments separate three questions. First, does the ligand bind the intended target with suitable affinity and selectivity in the biological fluid and species of interest? Second, does binding change cell or tissue exposure of intact cargo? Third, does the cargo produce the expected molecular effect at lower dose, higher selectivity, or improved safety compared with a matched non-targeted system? These questions require different assays. Surface plasmon resonance, flow cytometry, and cell binding studies address binding. Quantitative tissue distribution, cell sorting, in situ hybridization, and mass spectrometry address exposure. Target knockdown, splice correction, protein expression, editing, or immunological function address pharmacology.

> **Box 157.1. Binding Is Not Delivery**
>
> When evaluating a targeted RNA vehicle, separate the route into checkpoints. First, the receptor protein must be present on an accessible surface in the relevant species, disease state, and dosing route. Second, the intact product must bind that receptor under biological conditions such as serum proteins, extracellular matrix, and tissue flow. Third, the complex must enter the intended cell population rather than nearby endothelium, macrophages, or stromal cells. Fourth, active RNA must escape to the cytosol or nucleus before degradation. Fifth, the cargo must produce the expected mechanism: RNAi, RNase H cleavage, splice correction, protein expression, editing, or another defined endpoint. A bright uptake image or high tissue count can satisfy only part of this chain. The strongest design includes a binding-defective or non-targeted control that loses both uptake and functional RNA activity at matched dose and toxicity.

The misconception to avoid is that active targeting always dominates passive biodistribution. In vivo, circulation time, particle size, charge, plasma protein corona, renal filtration, mononuclear phagocyte uptake, endothelial permeability, and local injection route can dominate receptor binding. A highly specific ligand may add little if the vehicle never reaches the receptor. Conversely, a modest-affinity ligand may work well if the anatomy and trafficking are favorable. Ligand selection therefore belongs late enough in design that the route and tissue barrier are defined, but early enough that chemistry, particle architecture, and assays can be built around the intended receptor path.

For RNA drugs, the ligand also has to respect the cargo's subcellular destination. A splice-switching oligonucleotide or RNase H gapmer may need nuclear access; an siRNA must reach the cytosol and enter Argonaute loading; an mRNA must reach cytosol as a translatable molecule; a guide RNA may need to assemble with a nuclease or editor. Ligand-mediated endocytosis is common, but endosomes are not the destination for most RNA mechanisms. Targeting ligands should therefore be paired with release chemistry, endosomolytic domains, degradable linkers, or carrier designs that improve productive escape without unacceptable toxicity.

The current consensus is conservative. Ligand targeting is a powerful optimization layer and has clear successes in selected contexts, especially hepatocyte-targeted conjugates. Outside those contexts, ligand systems should be judged by functional pharmacology and safety, not by elegant binding diagrams (Roberts et al., 2020; Anand et al., 2025; Malecova et al., 2023; Sivakumar et al., 2019; Osborn et al., 2019; Gait et al., 2019).

## 157.3. Polymers, polyplexes, cell-penetrating peptides, and inorganic carriers

Polymers are macromolecules composed of repeating units. In RNA delivery, many polymers are cationic, meaning positively charged, because RNA has a negatively charged phosphate backbone. Mixing a cationic polymer with RNA can condense the RNA into a nanoscale complex called a polyplex. Condensation can protect RNA from nucleases, reduce hydrodynamic size, and promote cellular uptake. The same positive charge that helps bind RNA can also bind serum proteins, cell membranes, complement components, and extracellular matrix, producing toxicity or nonspecific uptake. Polymer delivery is therefore a balance between binding RNA tightly enough to protect it and releasing RNA soon enough to permit function.

Cell-penetrating peptides sit at the boundary between molecular conjugates and material carriers. A peptide covalently attached to a short oligonucleotide is a defined conjugate, whereas a peptide that electrostatically condenses RNA, decorates a polyplex, or supplies a pH-responsive membrane-disruptive domain behaves as part of a carrier. Arginine-rich or otherwise cationic peptides can increase association with cell surfaces, but productive delivery still requires intact cargo, intracellular release, and access to the cytosol or nucleus. Peptide-mediated phosphorodiamidate morpholino oligomer delivery to muscle illustrates the opportunity and the boundary: increased myofiber exposure can improve splice correction, yet renal exposure, membrane toxicity, species-specific uptake, and chronic repeat-dose tolerability must be measured rather than inferred from cultured-cell penetration.

Classic cationic polymers include polyethyleneimine, poly-L-lysine, chitosan derivatives, poly(beta-amino ester)s, and many synthetic degradable polycations. Polyethyleneimine is often discussed because its protonatable amines can buffer over the endosomal pH range, a property proposed to drive chloride influx, osmotic swelling, and membrane rupture. Direct measurements with compartment-localized pH sensors did not detect a PEI-dependent rise in lysosomal pH across the tested concentrations and time points, and estimates of lysosomal PEI abundance raised doubt that this simple proton-sponge sequence is the dominant escape mechanism. Those experiments constrain one version of the model; they do not prove that protonation, ion flux, or osmotic effects never contribute in another polymer, formulation, cell type, or endosomal stage. Endosomal escape can instead depend on membrane interaction, fusogenic chemistry, polymer architecture, cargo release, and cell-type-specific trafficking. Indeed, a polymer-design comparison summarized in a recent review found that adding a buffering imidazole group did not improve escape, whereas a fusogenic modification did, reinforcing the need to measure escape rather than infer it from buffering capacity (Benjaminsen et al., 2013; Friesen and Blakney, 2024).

Degradable polymers address one major limitation of stable cationic materials. Hydrolyzable, reducible, enzymatically cleavable, or pH-sensitive linkages can lower persistence and permit cargo release after uptake, but degradability is beneficial only when cleavage occurs on a useful timescale and produces tolerable products. Poly(beta-amino ester)s are a prominent example of a large design family in which monomers, end groups, branching, hydrophobicity, and degradability can be varied. Chitosan-based systems illustrate the effect of natural-polymer properties: molecular weight, deacetylation, solubility, and charge state strongly influence RNA binding and mucosal delivery. Across these systems, complexation and decomplexation are coupled: weak binding can expose RNA before uptake, whereas excessive binding can prevent release into the productive compartment. Polymers can also be formulated with lipids, peptides, targeting ligands, or inorganic cores to build hybrid particles (Friesen and Blakney, 2024; Byun et al., 2022).

Dendrimers are branched polymers built in generations from a central core. A higher-generation dendrimer has many terminal groups that can bind RNA or display ligands. Poly(amidoamine), often abbreviated PAMAM, is a commonly cited dendrimer family. The benefit is multivalency and structural control. The risk is dense surface charge, which can disrupt membranes, activate complement, damage blood cells, and accumulate in tissues depending on size and degradability. Dendrimer design often modifies terminal amines with neutral, degradable, hydrophobic, or targeting groups to reduce toxicity and improve release.

Figure 157.3 organizes polymer, dendrimer, inorganic, and hybrid carriers by the physical jobs they perform: condensation, protection, circulation control, targeting, endosomal escape, cargo release, imaging, and biodegradation.

![Figure 157.3. Material functions in polymer, dendrimer, inorganic, and hybrid RNA carriers](../assets/figures/chapter1140_figure3.png)

**Figure 157.3. Material functions in polymer, dendrimer, inorganic, and hybrid RNA carriers.** Polymers, dendrimers, inorganic particles, and hybrid systems solve different physical problems in RNA delivery. Protection and release are coupled: weak complexation can expose RNA prematurely, whereas excessive complexation can prevent productive decomplexation. The same properties that support condensation or uptake, especially cationic charge and multivalency, can also drive toxicity and nonspecific clearance.

Inorganic carriers include gold nanoparticles, silica nanoparticles, calcium phosphate, iron oxide, carbon-based materials, layered double hydroxides, and other mineral or metallic systems. They can provide rigid cores, tunable surface chemistry, optical or magnetic properties, and high loading surfaces. Calcium phosphate and other mineral carriers can dissolve under acidic conditions, potentially linking endosomal acidification to cargo release. Gold particles support dense oligonucleotide functionalization and imaging. Iron oxide can support magnetic resonance imaging or magnetic handling. These advantages must be weighed against persistence, unclear degradation products, organ accumulation, inflammatory responses, and scale-up challenges.

Hybrid systems combine material classes to solve multiple problems. A polymer-lipid hybrid may use a polymer core for RNA condensation and a lipid shell for biocompatibility. An inorganic core may be coated with polymer, lipid, polyethylene glycol, peptide, or antibody to improve stability and targeting. A peptide-polymer conjugate may combine endosomal disruption with RNA binding. Hybridization increases design flexibility but also increases analytical complexity. Every component adds a source of batch variation, impurity, degradation, and immune recognition.

The assays for material carriers must account for physical heterogeneity. Particle size distributions measured by dynamic light scattering can hide small populations of large aggregates. Zeta potential measured in simple buffer may not predict charge in serum after protein adsorption. Gel retardation can show RNA binding but not release. Fluorescently labeled RNA may dissociate from the carrier or be quenched. Reporter protein expression from delivered mRNA can reflect a small number of highly transfected cells rather than broad tissue delivery. A complete evaluation needs orthogonal measurements: particle characterization, RNA integrity, serum stability, cell viability, intracellular trafficking, functional RNA effect, and histological evidence of tissue injury.

Polymer and inorganic systems are often strongest where local delivery is possible. Intratumoral injection, topical ocular dosing, inhaled lung delivery, mucosal application, wound delivery, and ex vivo cell manipulation reduce some systemic barriers. Local dosing does not eliminate toxicity; it changes which toxicity matters. A polymer acceptable in a single intratumoral injection may be unacceptable for chronic systemic administration. A material suitable for ex vivo cell engineering may not be suitable for direct intravenous dosing.

The manufacturing boundary is central. Clinical translation requires defined raw materials, reproducible synthesis, controlled molecular-weight distribution and dispersity, verified end groups and substitution, a specified polymer-to-RNA charge ratio, low residual monomer, endotoxin, and solvent, scalable mixing, stable storage, and release assays that predict function. Hydrodynamic size, polydispersity, morphology, loading, intact-RNA protection, decomplexation, and activity should be measured in biologically relevant matrices because a size or zeta-potential value in simple buffer can change after serum or mucus exposure. Some academic formulations are made by hand mixing under conditions that are hard to reproduce at scale. Others rely on polymers with broad molecular-weight distributions or poorly defined substitution patterns. Such systems can teach biological principles but may be difficult drug products (Friesen and Blakney, 2024; Byun et al., 2022).

The consensus is that polymers, dendrimers, inorganic carriers, and hybrids are scientifically important and can be powerful in selected applications, but few have yet matched the clinical maturity of hepatocyte-targeted conjugates or lipid nanoparticle platforms. The most promising paths emphasize degradability, defined chemistry, local or ex vivo use cases, and assays that connect physical properties to functional RNA delivery (Yan et al., 2022; Byun et al., 2022; Friesen and Blakney, 2024).

## 157.4. Viral vectors, virus-like particles, and extracellular-vesicle delivery

Biological delivery systems use entry programs that evolution has already optimized. Viruses enter cells, traffic through barriers, release genomes, and sometimes target particular tissues. Cells secrete extracellular vesicles that carry proteins, lipids, and RNAs to other cells. Therapeutic engineering can borrow from these systems, but borrowing biological machinery also brings biological constraints: immunity, tropism, replication safeguards, cargo limits, manufacturing variability, and difficulty controlling dose composition.

Viral vectors are engineered viruses or viral particles used to deliver genetic material. Adeno-associated virus, lentiviral vectors, adenoviral vectors, retroviral vectors, and RNA virus-derived vectors differ in genome type, capacity, persistence, integration behavior, tissue tropism, and immunogenicity. Many viral vectors deliver DNA expression cassettes rather than RNA drug molecules directly, but they are relevant to RNA therapeutics when the expressed product is a therapeutic RNA, guide RNA, short hairpin RNA, RNA editing guide, or mRNA-like transcript. Some RNA virus-derived systems and replicon particles deliver RNA more directly.

The advantage of viral vectors is efficiency. Viral capsids and envelopes can mediate attachment, uptake, endosomal escape or membrane fusion, nuclear entry for some vectors, and expression. Tissue tropism can be altered by capsid engineering, envelope pseudotyping, promoter selection, microRNA target sites that detarget expression, and route of administration. The disadvantage is that the same efficiency is hard to make transient, repeatable, and immunologically silent. Pre-existing antibodies can block vector activity. Innate sensing can cause inflammation. Cytotoxic T cells can eliminate transduced cells. Some vectors persist for months to years, which is desirable for genetic replacement but mismatched for short-lived RNA pharmacology. Integrating vectors raise insertional mutagenesis concerns when integration is part of the biology.

Virus-like particles, or VLPs, occupy an intermediate position. They are assembled from viral structural components but lack a complete infectious genome. VLPs can be used as vaccines, gene-delivery particles, or RNA carriers. Engineered VLPs can package mRNA, guide RNA, Cas ribonucleoproteins, base editor components, or RNA-programmed enzymes. The key design problem is packaging specificity: the particle must load the intended cargo efficiently without carrying unwanted producer-cell RNA, DNA, proteins, or replication-competent contaminants. VLPs also need a release route that puts the cargo in the correct compartment, not simply inside an endosome.

Figure 157.4 compares viral vectors, VLPs, exosomes, and engineered extracellular vesicles in terms of biogenesis, cargo loading, entry, release, and control points.

![Figure 157.4. Biological carriers: viral vectors, VLPs, and extracellular vesicles](../assets/figures/chapter1140_figure4.png)

**Figure 157.4. Biological carriers: viral vectors, VLPs, and extracellular vesicles.** Viral vectors, virus-like particles, and extracellular vesicles all exploit biological entry systems, but they differ in cargo identity, persistence, immune recognition, and controllability. Therapeutic development depends on proving what is packaged, how it is released, and which contaminants are absent.

Exosomes are small extracellular vesicles often associated with multivesicular body biogenesis, while microvesicles bud from the plasma membrane and apoptotic bodies arise during cell death. In practice, many therapeutic preparations are mixed extracellular vesicle populations unless biogenesis markers, size, density, and isolation methods are carefully controlled. Extracellular vesicles can carry RNA species, proteins, lipids, and metabolites. Their natural role in cell communication makes them attractive as RNA carriers, especially for tissues where synthetic particles perform poorly. However, natural vesicle cargo is complex and can be biologically active on its own.

Therapeutic extracellular vesicle engineering has several routes. Producer cells can be loaded with RNA so that some RNA is packaged during vesicle biogenesis. Vesicles can be loaded after isolation by electroporation, sonication, extrusion, detergent treatment, or chemical transfection. Surface proteins can be engineered to display targeting ligands. Donor cell type can be chosen for tissue tropism or immunological compatibility. Each approach has caveats. Electroporation can aggregate RNA. Loading efficiency can be low or hard to quantify. Surface engineering can alter vesicle identity. Producer-cell impurities, residual DNA, adventitious agents, and culture conditions can affect safety.

The evidence standard for extracellular vesicle RNA delivery must be high because the field is artifact-prone. A vesicle preparation can co-isolate protein aggregates, lipoproteins, ribonucleoprotein particles, or free RNA bound to contaminants. Bulk small-RNA sequencing of a vesicle preparation does not prove that a specific RNA is inside a vesicle lumen or delivered to recipient-cell cytosol. Functional effects attributed to vesicle RNA may be caused by vesicle proteins, lipids, cytokines, or contaminating medium components. Strong experiments use density gradients, protease and nuclease protection assays, detergent controls, single-particle characterization, donor-cell negative controls, recipient-cell mechanism assays, and dose normalization by particle number, protein, lipid, and cargo copy number.

> **Box 157.2. Controls for Extracellular Vesicle RNA Delivery**
>
> An extracellular vesicle RNA-delivery claim should clear four controls before it is interpreted mechanistically. Purity controls should show that the preparation is not mainly lipoproteins, protein aggregates, free ribonucleoproteins, or residual transfection reagent. Cargo-location controls should use nuclease protection, protease treatment when relevant, and detergent disruption to distinguish RNA inside vesicles from RNA stuck to the outside. Dose controls should report particle number, protein or lipid amount, and cargo copy number, not only input RNA or enrichment by quantitative PCR. Mechanism controls should show that the recipient-cell effect follows the expected RNA pathway and is lost with a mutated cargo, donor-cell cargo knockout, recipient target control, or nonloaded vesicle preparation. Mock-conditioned medium and donor-cell negative controls are essential because vesicle proteins, lipids, cytokines, or culture contaminants can produce effects that look like RNA delivery.

Viral and vesicle systems also differ in manufacturability. Viral vectors have established industrial processes but require cell substrates, helper functions, purification of full from empty capsids for some vectors, replication-competent virus testing, potency assays, and careful control of host-cell DNA and protein impurities. Extracellular vesicle products require donor-cell control, culture media control, isolation scalability, identity markers, potency assays, sterility, and reproducible cargo loading. VLPs sit between these categories, with viral-protein assembly quality and cargo loading as central release attributes.

Table 157.3 should emphasize tradeoffs rather than implying that biological systems are inherently safer or more precise than synthetic carriers.

**Table 157.3. Biological RNA carrier tradeoffs.** Viral vectors, virus-like particles, exosomes, microvesicles, and extracellular-vesicle-inspired carriers differ in cargo strategy, duration, redosing, immune risk, and analytical control; carrier identity alone does not establish cargo delivery.

| Carrier class | Typical cargo strategy | Duration | Repeat-dosing concern | Main analytical control | Major safety issue | Suitable use cases |
| --- | --- | --- | --- | --- | --- | --- |
| **AAV-like vector** | DNA cassette expressing shRNA, guide RNA, RNA editing guide, or therapeutic transcript | Long-lasting expression, often months to years | Neutralizing anti-capsid antibodies and cellular anti-transgene responses can block redosing | Vector genome titer, full-empty capsid ratio, potency, and residual host-cell impurities | Persistent expression, tissue inflammation, dose-related organ toxicity, and limited reversibility | Durable genetic or local indications where long expression is desired |
| **Adenoviral vector** | DNA expression cassette delivered by nonintegrating viral particle | Transient to intermediate, usually shorter than AAV persistence | Pre-existing and induced anti-adenovirus immunity can be strong | Replication-competent virus testing, vector genome identity, potency, and impurity profile | Acute innate inflammation and cytokine induction | Vaccines, cancer immunotherapy, and local applications where immunogenicity can be acceptable |
| **Lentiviral or retroviral vector** | Integrating cassette for RNA expression or engineered-cell programming | Durable or permanent in transduced cells | Redosing is usually less central ex vivo, but vector and product immunity still matter | Integration-site profile, vector copy number, replication-competent vector testing, and potency | Insertional mutagenesis and persistent off-target expression | Ex vivo cell engineering and selected one-time genetic interventions |
| **RNA-virus-derived vector or replicon particle** | Packaged self-amplifying RNA, replicon RNA, or RNA expression system | Transient but amplified, with duration set by replicon design and immunity | Anti-vector immunity and innate sensing can reduce repeat activity | RNA integrity, replicon competency limits, particle identity, and envelope or capsid quality | Excess innate immune activation or unwanted persistence | Vaccination and transient high-level expression where amplification is useful |
| **Virus-like particle** | Packaged mRNA, guide RNA, ribonucleoprotein, or editor component without a complete infectious genome | Transient cargo delivery | Anti-capsid or anti-envelope responses can limit repeat use | Cargo identity, unwanted host nucleic acid, particle potency, and replication-competent contaminants | Off-target tropism, inflammatory recognition, and unintended packaged cargo | Transient editing, protein expression, or RNA-programmed delivery needing biological entry |
| **Engineered extracellular vesicle** | Producer-cell loading or post-isolation loading of RNA cargo | Transient and often formulation-dependent | Donor-cell antigens or vesicle surface proteins may create repeat-dose immunity | Vesicle identity, purity, cargo copy number, potency, and co-isolated contaminant testing | Mixed vesicles, lipoproteins, free RNPs, cytokines, or unwanted donor-cell cargo | Local or targeted delivery programs where synthetic carriers perform poorly and analytics are strong |
| **EV-inspired synthetic hybrid** | Synthetic vesicle or nanoparticle decorated with selected EV-like lipids or proteins | Transient and design-dependent | Surface proteins, PEG, or lipid components can induce adaptive or innate responses | Defined composition, cargo loading, residual biological impurities, and functional potency | Unclear biological identity, immune recognition, and manufacturing drift | Modular biological mimicry with more control than native EV preparations |

The misconception to avoid is that "natural" means non-immunogenic or tissue-specific. Natural vesicles and viral capsids are recognized by immune systems. Natural tropism may favor liver, spleen, lung, or phagocytic clearance rather than the intended tissue. Natural cargo may be unwanted. Engineered viral vectors can be safer than wild-type viruses, but they still require safety logic appropriate to their persistence and tissue distribution. Engineered extracellular vesicles may be less inflammatory than some synthetic particles in a specific assay, yet still present immunological and manufacturing risks.

The consensus is that viral vectors are mature for several gene-transfer applications, VLPs are a rapidly developing platform for transient genetic and RNA-programmed delivery, and extracellular vesicle systems remain biologically promising but analytically demanding. For RNA delivery, the decision depends on whether the desired therapeutic action should be transient or persistent, local or systemic, repeatable or one-time, and whether the cargo can tolerate the constraints of biological packaging (Ronzitti et al., 2020; Ikwuagwu and Tullman-Ercek, 2022; Welsh et al., 2024; Muskan et al., 2024; Mizenko et al., 2024).

## 157.5. Tissue barriers, cell entry, intracellular release, biodistribution, and repeat dosing

Tissue targeting begins with anatomy. A systemically administered RNA carrier leaves the injection site or bloodstream through vascular beds that differ in fenestration, flow, endothelial junctions, basement membrane, extracellular matrix, resident macrophages, lymphatic drainage, and disease remodeling. A locally administered carrier faces different barriers: mucus, vitreous humor, cerebrospinal fluid flow, tissue pressure, epithelial tight junctions, or injection-related spread. A delivery platform should therefore be matched to tissue physiology before receptor or material optimization begins.

The liver is the reference tissue for systemic RNA delivery because blood flow is high, sinusoidal endothelium is fenestrated, hepatocytes express useful receptors, and Kupffer cells and liver sinusoidal endothelial cells actively clear particles and macromolecules. These features make the liver easy to reach but not always easy to target specifically. A carrier accumulating in liver may be in hepatocytes, Kupffer cells, endothelial cells, stellate cells, or biliary cells. For hepatocyte-targeted siRNA, GalNAc conjugation provides a strong solution. For mRNA expression, lipid nanoparticles often transfect hepatocytes efficiently. For macrophage or stellate-cell targets, different uptake routes and safety issues dominate (Springer and Dowdy, 2018; Juliano, 2016; Paunovska et al., 2022).

Central nervous system delivery faces the blood-brain barrier, blood-cerebrospinal fluid barrier, extracellular matrix, perivascular clearance, and sensitive neuroinflammatory thresholds. Systemic carriers usually have poor access to brain parenchyma unless they exploit receptor-mediated transcytosis, disease-disrupted barriers, focused ultrasound, or specialized routes. Intrathecal, intracerebroventricular, and intraparenchymal administration can bypass some vascular barriers but impose distribution limits and procedural risk. Antisense oligonucleotides can distribute through cerebrospinal fluid after intrathecal dosing, while large particles and mRNA systems often face stronger spread and uptake limitations. Receptor targeting through transferrin receptor or insulin receptor pathways is plausible but must avoid trapping in endothelial lysosomes or altering receptor physiology.

Skeletal and cardiac muscle are attractive for genetic disease, vaccines, and protein-expression therapies, but muscle is not uniformly permissive. Intramuscular injection can produce local expression and immune activation useful for vaccination. Systemic delivery to widespread skeletal muscle requires crossing continuous endothelium, basement membrane, extracellular matrix, and the sarcolemma of large multinucleated fibers. Conjugates that bind transferrin receptor or other muscle-enriched targets, cell-penetrating peptides for phosphorodiamidate morpholino oligomers, and engineered viral capsids have all been explored. Cardiac muscle adds safety constraints because inflammation, conduction disturbance, or off-target editing can have severe consequences.

The eye illustrates why local anatomy can make difficult molecules clinically practical. Intravitreal, subretinal, suprachoroidal, and topical routes expose different ocular compartments. The eye is small, relatively compartmentalized, and accessible for local injection, so oligonucleotides, viral vectors, and particles can be used at lower total doses than systemic therapies. The barriers include vitreous diffusion, inner limiting membrane, retinal cell uptake, immune privilege that is partial rather than absolute, and long-term safety in nonrenewable retinal cells. A delivery success in eye should not be generalized to systemic tissues without accounting for dose, volume, and route.

The lung is accessible through inhalation, but inhaled delivery is not simply "direct." Particles encounter device aerosol physics, airway branching, mucociliary clearance, mucus mesh, surfactant, alveolar macrophages, epithelial tight junctions, cough, inflammation, and disease-state remodeling. Cystic fibrosis, asthma, chronic obstructive pulmonary disease, viral infection, and pulmonary fibrosis create different mucus and immune environments. Local delivery can reduce systemic exposure, but the lung is highly immunologically surveilled. Polymer, lipid, peptide, and viral systems must be evaluated for airway irritation, complement activation, macrophage uptake, and regional deposition.

Immune-cell delivery is valuable for vaccines, cancer immunotherapy, inflammatory disease, and ex vivo cell engineering. Dendritic cells, macrophages, T cells, B cells, natural killer cells, and hematopoietic stem cells differ in uptake routes and sensitivity to innate immune activation. In vivo delivery to lymphoid organs may exploit subcutaneous drainage, lymph node-resident phagocytes, mannose receptors, Fc receptors, complement receptors, or lipid uptake pathways. Ex vivo delivery can use electroporation, viral vectors, nanoparticles, or transfection reagents under controlled conditions, followed by cell selection and quality testing. The key distinction is that immune activation may be desired for vaccination but harmful for gene correction or tolerogenic therapy.

Tumor delivery is often described through the enhanced permeability and retention effect, the idea that abnormal tumor vasculature and poor lymphatic drainage can permit nanoparticle accumulation. This effect is real in some animal models and tumor regions but varies within one lesion, among lesions in one patient, and among patients. Vascular permeability alone is insufficient: perfusion, collapsed vessels, interstitial fluid pressure, extracellular-matrix density, stromal and macrophage uptake, necrosis, antigen heterogeneity, and prior therapy all shape access and penetration. Consequently, a particle can reach a tumor yet remain perivascular or accumulate in tumor-associated macrophages rather than malignant cells. Imaging or other companion measures of vascular function and nanoparticle deposition may help stratify EPR-dependent approaches, but a proxy nanoparticle is not automatically predictive of a chemically different RNA product. Intratumoral or regional delivery can bypass some access problems for accessible lesions; metastatic disease still requires product-specific systemic exposure and cell-resolved pharmacology (Golombek et al., 2018).

Placental and fetal-interface delivery is the most ethically and biologically constrained tissue-targeting problem in this chapter. The placenta is not a passive filter; it is a dynamic organ with trophoblast layers, immune cells, transporters, endocrine function, and anatomy that changes across gestation and species. Human maternal and fetal circulations are separated at the exchange surface by a different trophoblast organization than in common rodent models, and trophoblast invasion and endocrine control also differ. Delivery could target maternal disease, placental dysfunction, fetal disease, or deliberately avoid fetal exposure during maternal treatment; each goal therefore needs separate maternal, placental, and fetal measurements. Evidence that some nontherapeutic nanoparticles reach or cross placental models does not establish RNA delivery, and reported transfer depends on composition, size, charge, surface chemistry, protein corona, route, developmental stage, and assay. Rodent studies should be paired where feasible with human-relevant systems such as ex vivo perfused placental cotyledons, trophoblast models, or placenta-on-chip systems, while recognizing that no isolated model reproduces a whole pregnancy. Placenta-targeted RNA delivery therefore requires exceptional mechanistic and developmental-safety evidence before clinical translation (Adams and Stapleton, 2023; Aengenheister et al., 2021).

Repeat dosing changes the tissue comparison. A GalNAc oligonucleotide program for hepatocytes can often be designed around periodic subcutaneous dosing, whereas an intrathecal CNS or intravitreal eye program adds procedural burden and compartment-specific inflammation to every dose. Repeated inhalation can accumulate airway irritation or macrophage activation even when systemic exposure is low. Chronic muscle therapy may require broad distribution at every administration, making modest inefficiency a cumulative dose problem. Viral capsids, antibody carriers, peptide conjugates, and extracellular-vesicle products create different adaptive immune memories, so successful first-dose delivery does not establish later-dose exposure or safety. Repeat-dose studies should therefore measure both pharmacodynamic durability and whether the carrier's tissue distribution, immune recognition, and productive release change across administrations.

Figure 157.5 places these tissue barriers on one comparative map. The point is not to rank tissues from easy to hard but to show which barrier dominates for each route and modality.

![Figure 157.5. Tissue barriers, cell entry, intracellular release, and repeat dosing](../assets/figures/chapter1140_figure5.png)

**Figure 157.5. Tissue barriers, cell entry, intracellular release, and repeat dosing.** Tissue targeting is the combined outcome of route, anatomy, vascular permeability, extracellular barriers, cell-type uptake, immune surveillance, intracellular release, and treatment schedule. A delivery signal in a tissue homogenate can come from cells that are not the therapeutic target, and a successful first dose does not prove that a platform can be redosed safely.

> **Box 157.3. What Counts as Tissue Targeting?**
>
> Treat tissue targeting as a hierarchy of evidence. Whole-organ biodistribution is useful for screening where administered material accumulates, but it cannot identify the therapeutic cell. Cell sorting, spatial imaging, in situ hybridization, or reporter systems can show which cell populations contain intact cargo. The decisive step is a mechanism readout in those cells: target knockdown, splice correction, translated protein, editing, or another cargo-specific endpoint. Compartment matters as well, because surface-bound, endothelial, macrophage, endosomal, or lysosomal signal is not the same as productive cytosolic or nuclear delivery. A liver signal without hepatocyte pharmacology, a brain signal restricted to endothelium, or a tumor signal dominated by macrophages is biodistribution rather than target-cell delivery. The meaningful comparison is functional gain over a matched untargeted design at an acceptable dose and safety margin.

The evidence basis for tissue targeting should include cell-type-resolved readouts. Bulk tissue RNA or fluorescence is rarely enough. A liver homogenate cannot distinguish hepatocyte from Kupffer-cell delivery. A tumor homogenate cannot distinguish cancer cells from macrophages or endothelium. A brain homogenate cannot distinguish endothelial trapping from neuronal or glial delivery. Useful methods include single-cell RNA sequencing when the pharmacodynamic effect is transcriptionally visible, fluorescence-activated cell sorting with intact-cargo detection, in situ hybridization, immunohistochemistry for protein expression, laser capture microdissection, reporter animals, cell-type-specific target knockdown, and spatial methods. Each method has artifacts, so the strongest claims combine several.

The consensus is that tissue barriers, not ligand imagination, set the practical boundary of RNA delivery. Liver and local compartments such as eye and cerebrospinal fluid are currently more tractable than broad systemic delivery to brain, muscle, tumors, or placenta. Muscle, eye, lung, central nervous system, tumor, and immune-cell programs cannot be placed on one universal difficulty scale because local administration, cell identity, cargo mechanism, and acceptable treatment frequency differ. Extrahepatic targeting is advancing, but progress should be measured by functional delivery at safe doses in relevant species and disease states (Anand et al., 2025; Holm et al., 2022; Man et al., 2023; Zhang et al., 2024; Golombek et al., 2018; Adams and Stapleton, 2023; Berdecka et al., 2024).

## 157.6. Modality-by-tissue selection, safety, manufacturing, and decision criteria

Delivery systems fail clinically for reasons that may be invisible in a proof-of-concept experiment. A modality-by-tissue decision therefore begins with a tuple rather than a platform name: RNA cargo and required intracellular destination; tissue and target cell; route and dominant barrier; uptake and release mechanism; desired duration and repeat-dose schedule; and manufacturing and safety boundary. Safety includes acute infusion reactions, complement activation, cytokine induction, tissue inflammation, coagulation changes, off-target organ accumulation, genotoxicity for integrating systems, reproductive and developmental toxicity when relevant, and toxicity from impurities or degradation products. Immunogenicity includes innate immune sensing of RNA, carrier materials, viral proteins, vesicle proteins, antibodies, and adaptive immune responses that block redosing or injure transduced cells. Manufacturability includes the ability to make the same product repeatedly with controlled identity, purity, potency, and stability.

RNA cargo and carrier safety are inseparable. An unmodified or double-stranded RNA contaminant can trigger Toll-like receptors, RIG-I-like receptors, protein kinase R, or OAS/RNase L pathways. A cationic carrier can damage membranes or activate complement. A targeting ligand can bind an unintended tissue. A viral vector can persist longer than desired. An extracellular vesicle preparation can carry donor-cell proteins or nucleic acids. A polymer can degrade into toxic products or fail to degrade. The safety question is not "is the RNA safe?" or "is the carrier safe?" but whether the assembled product is safe under the intended dose, route, schedule, and patient population.

Figure 157.6 shows how cargo and intracellular destination, tissue and target cell, route and barrier, productive uptake and release, repeat-dose need, and manufacturing and safety constraints converge on a product-specific decision.

![Figure 157.6. Modality-by-tissue decision framework for non-LNP RNA delivery](../assets/figures/chapter1140_figure6.png)

**Figure 157.6. Modality-by-tissue decision framework for non-LNP RNA delivery.** A non-LNP delivery modality becomes credible only when the cargo and intracellular destination, tissue and target cell, route and dominant barrier, productive uptake and release path, repeat-dose requirement, and manufacturing and safety controls align for the intended product. No platform is universally preferred across tissues or RNA cargos.

Manufacturability is often the decisive boundary between a clever platform and a drug product. Conjugates are attractive because covalent chemistry can yield defined molecular species, although stereochemistry, impurity profiles, and conjugation site control still matter. Polymer and hybrid nanoparticles may require careful control of mixing, molecular weight, substitution, charge ratio, particle size, residual monomer, and aggregation. Viral vectors need producer-cell systems, helper components, capsid or envelope quality, potency assays, and tests for replication-competent contaminants. Extracellular vesicles need donor-cell banks, culture control, isolation methods, cargo loading quantification, identity markers, and potency assays. A product that cannot be characterized cannot be reliably dosed.

Scalability depends on both material and indication. A rare-disease intrathecal oligonucleotide may require modest annual production but high purity and long-term safety. A seasonal vaccine or pandemic-response RNA product may require enormous scale, rapid release testing, cold-chain management, and robust raw-material supply. An autologous ex vivo cell therapy can tolerate individualized manufacturing but must control chain of identity and sterility. A chronic cardiometabolic drug must support repeat dosing, low cost of goods, and minimal cumulative toxicity. The "best" delivery modality changes when the dose, population size, and treatment frequency change.

Decision criteria should begin with the cargo. A short chemically stabilized siRNA can use conjugates or carriers that would not protect an mRNA. An mRNA requires cytosolic delivery of a long, translation-competent molecule and often benefits from particulate protection. A splice-switching oligonucleotide may need nuclear access and can sometimes be dosed locally. An RNA editing guide may need co-delivery with a protein or expression system. A self-amplifying RNA may reduce dose but increase innate immune and persistence considerations. A circular RNA may change stability and translation requirements but still needs cytosolic entry.

Table 157.4 makes the comparison concrete across GalNAc and other ligand conjugates, peptide-oligonucleotide conjugates, polymer and inorganic carriers, viral vectors, virus-like particles, extracellular vesicles, and locally administered stabilized oligonucleotides. Each row specifies compatible cargo, the tissue or compartment in which the approach is plausible, the dominant access and release bottlenecks, repeat-dose compatibility, and manufacturing and safety liabilities. The table includes liver, muscle, eye, lung, central nervous system, tumors, immune cells, and the maternal-fetal interface, but it deliberately does not assign a universal winner: the same platform can be a rational local or ex vivo choice and an unjustified systemic choice.

**Table 157.4. Modality × RNA cargo × tissue/barrier decision matrix for non-LNP delivery.** A non-LNP delivery choice is a product-specific combination of modality, RNA cargo, tissue barrier, uptake and release route, repeat-dose need, manufacturability, and safety rather than a universally superior carrier.

| Non-LNP modality | Compatible RNA cargo | Plausible tissue or use context and dominant barrier | Uptake and productive-release requirement | Repeat-dose fit | Manufacturing and safety boundary |
| --- | --- | --- | --- | --- | --- |
| **GalNAc-ASGPR conjugate** | Chemically stabilized siRNA or antisense oligonucleotide | Liver hepatocytes; favorable sinusoidal access and abundant receptor, but whole-liver signal can include Kupffer and endothelial cells | ASGPR-mediated uptake should be demonstrated by competition or receptor-loss controls, followed by rare productive endosomal release to cytosol for RNAi or nucleus/cytosol for RNase H | Generally compatible with repeat subcutaneous dosing when cargo and conjugate toxicology permit; receptor-recycling and route data remain construct-specific | Chemically defined conjugate with controllable stoichiometry; monitor liver injury, receptor saturation, impurity profile, and off-target oligonucleotide effects |
| **Antibody-oligonucleotide conjugate or antibody-targeted carrier** | Stabilized siRNA, antisense oligonucleotide, splice-switching oligonucleotide, or compact RNA complex | Muscle, selected tumors, immune-cell subsets, or receptor-accessible endothelium; continuous endothelium, antigen heterogeneity, and antigen sinks are common barriers | Target-antigen binding must be followed by internalization, linker or cargo release, and access to the cargo-specific cytosolic or nuclear compartment | Potentially repeatable, but anti-drug antibodies, target-mediated clearance, and cumulative off-target uptake are product-specific | Biologic expression and purification plus conjugation-site and drug-to-antibody-ratio control; assess Fc effects, immunogenicity, aggregation, and off-tissue antigen expression |
| **Peptide-oligonucleotide conjugate, including peptide-PMO strategies** | Antisense, splice-switching, or other short stabilized oligonucleotide | Skeletal or cardiac muscle and selected local compartments; vascular access, extracellular matrix, and sarcolemma entry dominate | Peptide-mediated cell association must lead to cytosolic or nuclear delivery without nonspecific membrane injury | Repeat dosing can be plausible but may be limited by peptide immunity, renal exposure, hemolysis, or cumulative tissue toxicity | Sequence-defined synthesis is feasible, but conjugation purity, proteolysis, aggregation, membrane toxicity, and species-specific uptake require control |
| **Hydrophobic or small-molecule oligonucleotide conjugate** | siRNA or antisense oligonucleotide | Liver and selected extrahepatic tissues reached through albumin, lipoprotein, receptor, or transporter pathways; injection-site retention, renal exposure, systemic protein binding, and phagocytic uptake can dominate | Carrier-protein association or receptor uptake must still yield intact cargo and productive intracellular release; valency and bulk hydrophobicity alone do not predict efficacy | Often designed for repeat dosing, with schedule constrained by tissue persistence and cumulative chemistry toxicity | Compact synthetic product, but ligand structure, valency, self-association, hydrophobic impurities, linker stability, protein corona, and unintended organ exposure matter |
| **Degradable polymer, polyplex, or peptide-polymer carrier** | mRNA, siRNA, guide RNA, self-amplifying RNA, or oligonucleotide | Inhaled lung, local tumor, mucosal sites, or ex vivo immune-cell engineering are more plausible than universal intravenous targeting; mucus, serum destabilization, and phagocyte uptake are key barriers | Condensation and protection must be balanced against timely decomplexation and measured endosomal escape without membrane toxicity; buffering capacity alone is insufficient evidence | Repeat use is possible only when polymer clearance, innate sensing, complement activation, and cumulative degradation products are acceptable | Molecular-weight distribution and dispersity, end-group or substitution identity, charge ratio, residual monomer, mixing, size and polydispersity in relevant matrices, stability, biodegradation, and potency require control |
| **Inorganic or synthetic hybrid carrier** | siRNA, mRNA, or imaging-coupled oligonucleotide cargo | Local tumor, eye, lung, skin, or other locally accessible research contexts; extracellular diffusion and long-term material clearance dominate | Surface-bound cargo must remain intact, enter target cells, detach or be released, and reach the correct intracellular compartment | Repeat-dose fit is often uncertain because persistence and organ accumulation may outlast RNA pharmacology | Rigid-core composition and loading can be measurable, but aggregation, metal or mineral impurities, biodegradation, chronic inflammation, and scale-up reproducibility are major boundaries |
| **Viral vector carrying an RNA-expression cassette** | DNA cassette expressing shRNA, guide RNA, editing guide, regulatory RNA, or therapeutic transcript | Eye, CNS, muscle, liver, and ex vivo engineered cells when durable expression is desired; capsid tropism and tissue immunity dominate | Cell transduction, nuclear delivery, transcription, and correct RNA processing replace direct cytosolic RNA release as the productive route | Often poor for routine redosing because neutralizing antibodies and cellular immunity can be durable | Mature but complex producer-cell manufacture; control capsid identity, genome integrity, empty particles, replication-competent contaminants, insertion risk where relevant, and persistent expression |
| **Virus-like particle or engineered capsid for direct cargo** | mRNA, guide RNA, ribonucleoprotein, editor protein-RNA complex, or transient RNA expression system | Muscle, liver, immune cells, tumors, or local compartments depending on engineered tropism; claims are capsid- and route-specific | Packaged cargo must survive entry, escape or fuse productively, and dissociate in an active form without unwanted nucleic-acid transfer | Repeat dosing may be limited by anti-capsid or anti-envelope immunity and should not be inferred from a first dose | Control assembly, packaged-cargo identity, host-cell nucleic acids, potency, particle heterogeneity, replication competence, and immune recognition |
| **Engineered extracellular vesicle or EV-inspired carrier** | siRNA, mRNA, guide RNA, protein-RNA complex, or regulatory RNA | Local CNS, lung, tumor, or immune-cell applications are investigational; tissue tropism, reticuloendothelial clearance, and heterogeneous cargo loading dominate | Prove that cargo is intravesicular, reaches the intended cell, exits the vesicular pathway, and causes the expected RNA mechanism rather than an effect of vesicle protein or lipid | Repeat-dose behavior is uncertain and depends on donor-cell antigens, vesicle surface composition, innate sensing, and product consistency | Donor-cell banks, culture conditions, isolation, purity, identity, cargo copies per particle, co-isolated lipoproteins or RNPs, sterility, and predictive potency are central controls |
| **Locally dosed stabilized oligonucleotide without a large carrier** | Antisense or splice-switching oligonucleotide and selected siRNA formats | CNS by intrathecal or intracerebroventricular dosing and eye by intravitreal or subretinal dosing; fluid distribution, extracellular spread, and target-cell uptake dominate | Chemical stability and local exposure must yield cellular uptake and nuclear or cytosolic access in the intended neurons, glia, retinal cells, or retinal pigment epithelium | Repeat dosing can be feasible but procedural burden, compartment-specific toxicity, and long tissue residence set the schedule | Chemically defined manufacturing is favorable; control sequence-dependent toxicity, metabolites, aggregation, sterility, local inflammation, and procedural risk |
| **No default platform for placenta or fetal interface** | Any RNA cargo requires indication-specific justification | Maternal tissue, placenta, or fetus are distinct targets; trophoblast organization, developmental timing, transporters, endocrine function, and species-specific placentation prevent simple extrapolation | Compartment-resolved evidence must distinguish maternal, placental, and fetal intact cargo and mechanism; pair animal studies with human-relevant perfusion, trophoblast, or chip models where feasible | Repeat dosing cannot be presumed safe; developmental stage and cumulative fetal exposure change the risk calculation | Reproductive and developmental toxicology, long follow-up, ethical justification, and the limits of every model dominate; nontherapeutic nanoparticle transfer is not evidence of RNA delivery |

The second criterion is duration. If a transient effect is desired, a persistent viral vector may be excessive. If lifelong correction is desired, repeated dosing with a weak nonviral carrier may be impractical. If redosing is likely, adaptive immunity to viral capsids, antibodies, or carrier components becomes central. If the therapy must be stopped quickly after toxicity, long-lived expression systems are risky. Duration should be matched to disease biology rather than platform fashion.

The third criterion is route. Subcutaneous delivery favors stable conjugates and depot-like absorption. Intravenous delivery exposes a product to complement, serum proteins, liver, spleen, kidney, and vascular endothelium. Intrathecal delivery changes distribution and safety thresholds. Intravitreal and subretinal delivery use local ocular compartments. Inhaled delivery demands aerosol stability and lung tolerability. Ex vivo delivery allows aggressive physical methods that would be unacceptable in vivo.

The fourth criterion is the relevant safety margin. For a life-threatening disease with no alternatives, a one-time vector with serious but manageable risks may be justified. For a preventive therapy, chronic metabolic drug, or pregnancy-related indication, the tolerated risk is much lower. Tissue-targeted delivery does not remove the need for systemic toxicology because small off-target fractions can matter when doses are high or tissues are sensitive.

The matrix is a screening tool, not a substitute for product-specific evidence. GalNAc-siRNA performance in hepatocytes does not validate GalNAc for mRNA, an antibody-oligonucleotide result in muscle does not establish tumor-cell delivery, an inhaled polymer result does not establish systemic lung selectivity, and local ocular or intrathecal success does not predict distribution from intravenous dosing. A platform claim becomes defensible only when the cargo, route, species, target cell, intracellular mechanism, dose, schedule, and assay endpoint match the intended product.

## Recent Consensus

The recent consensus across non-LNP delivery modalities is pragmatic rather than platform-driven. GalNAc-ASGPR conjugates are established for hepatocyte-directed oligonucleotide delivery because the ligand, receptor, liver sinusoidal anatomy, oligonucleotide chemistry, subcutaneous pharmacology, and intracellular mechanism align. This success is real, but it is not a general proof that ligand attachment can redirect any RNA cargo to any tissue. Most extrahepatic conjugates, including transferrin receptor, folate receptor, integrin, mannose receptor, antibody, peptide, aptamer, lipid, vitamin, and small-molecule systems, remain context-dependent and should be judged by intact-cargo exposure, cell-type-resolved pharmacodynamics, and safety rather than by binding alone.

Material carriers have a similarly conditional consensus. Polymers, dendrimers, inorganic particles, and hybrid systems can condense RNA, protect cargo, alter tissue exposure, and support local or ex vivo use cases, but cationic charge, multivalency, persistence, protein corona formation, complement activation, and batch heterogeneity remain recurring translational limits. Degradable and chemically defined systems are favored because they give developers more control over release, toxicity, and manufacturing. Viral vectors are powerful when durable expression, efficient transduction, or ex vivo cell engineering is desired; they are less attractive when the intended RNA pharmacology is transient, repeatedly dosed, or must be rapidly stopped after toxicity. Virus-like particles are an active bridge between biological entry and transient cargo delivery. Exosomes and extracellular vesicles are biologically plausible delivery vehicles, but the field increasingly requires stricter identity, purity, cargo-loading, potency, and contaminant controls before therapeutic claims are persuasive.

For tissue targeting, the consensus is that anatomy and route usually dominate ligand imagination. Liver and selected local compartments, especially eye and cerebrospinal fluid spaces, are more tractable than broad systemic delivery to brain parenchyma, widespread skeletal or cardiac muscle, solid tumors, or the placenta. Lung, immune-cell, tumor, and placental delivery can be feasible in defined settings, but each has a different safety boundary: airway irritation and macrophage uptake in lung, desired versus harmful immune activation in lymphoid tissues, malignant-cell versus stromal or macrophage uptake in tumors, and developmental risk at the maternal-fetal interface. Across modalities, credible delivery evidence connects formulation attributes to intact cargo, productive intracellular release, cell-type-specific RNA mechanism, clinically relevant dose, repeat-dosing feasibility, manufacturable quality, and a safety margin appropriate for the indication (Paunovska et al., 2022; Roberts et al., 2020; Nguyen et al., 2020; FDA CBER, 2020; FDA CBER, 2023).

## Open Questions, Controversies, Deprecated Models, and Common Misconceptions

Open questions:

- How can researchers make extrahepatic systemic RNA delivery broadly reliable? Candidate answers include new receptor-ligand pairs, engineered capsids, tuned polymer or hybrid carriers, endosomolytic chemistries, VLPs, extracellular vesicle-inspired particles, local administration routes, and disease-state targeting. None is yet a universal solution. The hardest unsolved problems are quantitative endosomal escape in vivo, prediction of human tissue tropism from animal and organoid models, safe redosing after adaptive immune responses, reproducible manufacture of complex biological or hybrid carriers, and potency assays that predict clinical performance instead of only in vitro uptake.

Common misconceptions:

- "Active targeting is not proven by ligand binding." A receptor expressed in bulk RNA-seq data may not be present as accessible protein on the relevant cell surface. A biodistribution signal does not prove intact, active RNA. Fluorescent puncta or total tissue radioactivity can reflect surface-bound material, endosomal sequestration, lysosomal degradation, macrophage uptake, endothelial trapping, or free label rather than productive delivery. Polymer transfection potency in cultured cells does not predict in vivo tolerability. Tumor accumulation does not prove malignant-cell delivery. Lung exposure after inhalation does not prove epithelial or immune-cell delivery, because mucus, surfactant, macrophages, and regional deposition can dominate. Placental transfer in one animal model does not establish human fetal-interface safety.

Deprecated or weakened claims:

- Several models should be treated as simplified or deprecated when they are used too broadly. The enhanced permeability and retention model can describe some tumor regions and animal models, but it does not justify assuming uniform malignant-cell delivery in human cancers. The proton-sponge model for polymers is useful as one possible contributor to endosomal disruption, but it is not a complete explanation for polyplex escape, toxicity, or release. The idea that exosomes are intrinsically safe because they are natural is an overgeneralization; extracellular vesicle preparations can contain mixed vesicles, lipoproteins, free ribonucleoproteins, donor-cell proteins, unwanted nucleic acids, and immunologically active contaminants. The assumption that viral efficiency makes viral vectors the best RNA delivery system ignores duration, redosing, immune memory, integration or persistence risks, payload limits, and manufacturing burden. GalNAc success should not be generalized from hepatocytes to all receptor-targeted delivery.
- Boundary cases matter for modality choice. A complex antibody, VLP, or viral vector may be justified for a severe disease, one-time intervention, local administration, or ex vivo engineered cell product, but inappropriate for a chronic preventive or cardiometabolic drug requiring inexpensive repeat dosing. A polymer or inorganic system may be useful for topical, intratumoral, inhaled, ocular, or ex vivo delivery even if systemic intravenous use is limited by toxicity or clearance. Immune activation can be beneficial for vaccines and cancer immunotherapy but unacceptable for gene correction, tolerogenic therapy, pregnancy-related indications, or sensitive tissues such as CNS and heart. The most defensible delivery claims therefore state the cargo, route, tissue, target cell, endpoint, species, dose, schedule, safety context, and manufacturing control rather than naming a platform as generally targeted.