# Chapter 107. Extracellular and Cell-Surface Host RNA, Circulating RNA, Vesicles, Biomarkers, and Intercellular Communication

## Scope Note

This chapter treats host RNA outside the cytosol and nucleus as a set of physical states with different topologies, carriers, destinations, and evidentiary requirements. It covers RNA exposed on the outer surface of living cells, including covalent glycoRNA and noncovalently presented RNA-RBP assemblies; RNA enclosed in or associated with extracellular vesicles; nonvesicular ribonucleoprotein and lipoprotein carriers; circulating RNA biomarkers; and defined donor-to-recipient RNA transfer. The chemistry, attachment site, glycan-linkage evidence, and biosynthetic machinery of glycoRNA belong to [Chapter 49](chapter1045.md). Innate and adaptive immune-sensing pathways belong to [Chapter 108](chapter1103.md) and [Chapter 109](chapter1104.md), protocol-level analyte measurement belongs to [Chapter 132](chapter1120.md), and therapeutic delivery belongs to Chapters [153](chapter1137.md)-[157](chapter1140.md). Environmental and mixed-community transcriptomics belong to [Chapter 114](chapter1108.md).

## Executive Summary

Extracellular RNA is not one organelle, carrier, or pathway. An extracellular RNA molecule may be enclosed within a membrane-bounded vesicle, adsorbed to a vesicle surface, bound to Argonaute or another RNA-binding protein (RBP), associated with a lipoprotein, embedded in cell-death material, exposed on the outer face of an intact cell, or briefly free before extracellular ribonucleases degrade it. These states are not interchangeable. Each creates different access to nucleases, receptors, recipient cells, purification methods, and biological interpretations.

Cell-surface RNA has expanded the classical picture of the glycocalyx, the carbohydrate-rich molecular layer outside the plasma membrane. Some surface RNAs are **glycoRNAs**, RNA molecules bearing covalently attached glycans. Other surface RNAs appear to be held noncovalently by RBPs, heparan-sulfate proteoglycans, or related surface assemblies. Flynn et al. (2021) established that conserved small RNAs can carry sialylated N-glycans and occur on living-cell surfaces. Subsequent studies connected particular surface-RNA states to neutrophil adhesion and recruitment, RNA-RBP nanodomains used by a cell-penetrating peptide, and heparan-sulfate-RNA-RBP complexes capable of recruiting an immune receptor. These results establish that surface RNA can be functional in defined models. They do not establish that all surface RNA is glycosylated, that all surface RNA uses one anchoring mechanism, or that every receptor found near surface RNA is activated by it.

Topology is the central reasoning problem. Detecting RNA in a membrane fraction does not show that the RNA is exposed outside the cell. Nuclease sensitivity can support external accessibility only if the cells remain intact and the treatment does not cause membrane leakage or secondary stress. Metabolic sugar labeling can identify a glycan-associated species, but co-purifying glycoproteins and other glycoconjugates can confound some workflows. Conversely, stringent chemistry and mass-spectrometric evidence can establish a covalent RNA-glycan linkage without determining how the molecule is retained on a particular cell surface. A strong surface-RNA claim combines intact-cell topology, molecular identity, physical association, orthogonal perturbations, and a function-specific causal test.

Extracellular vesicles and nonvesicular particles create a parallel assignment problem. Many small RNAs in plasma reside in RNPs rather than vesicles, and vesicle-enriched preparations can contain lipoproteins, protein aggregates, virions, and cell debris. The MISEV2023 framework therefore emphasizes operational particle descriptions, orthogonal characterization, and transparent limits on biogenesis claims (Welsh et al., 2024). RNase resistance or sedimentation in a high-speed pellet alone does not prove vesicle encapsulation.

Circulating RNA can still be a useful biomarker even when its carrier, source tissue, or biological function is uncertain. Clinical utility, however, requires a locked specimen and assay definition, appropriate normalization, calibration, and independent validation. Blood collection tube, clotting, hemolysis, platelet activation, residual cells, processing delay, extraction chemistry, library construction, and site-specific batch effects can each create or erase an apparent disease signature. The exRNAQC Consortium (2025) provides direct evidence that collection and purification choices alter transcriptome recovery and should be selected for the intended RNA classes.

Functional RNA communication is a stricter claim than extracellular detection. The defensible causal chain is donor production, release or surface presentation, protected transit, recipient exposure, access to the relevant membrane or intracellular compartment, molecular engagement, and a sequence- or mechanism-specific consequence that is lost and rescued under controlled perturbation. Copy number, carrier heterogeneity, surface retention, endosomal sequestration, and effector abundance often make this chain quantitatively restrictive. This chapter keeps observation, carrier assignment, topology, receptor proximity, delivery, and function as separate claims.

## Concept Inventory

- **Extracellular RNA:** RNA detected outside intact cells. The term specifies location, not carrier, origin, stability, or function.
- **Cell-surface RNA:** RNA physically accessible at the extracellular face of an intact cell. Cell-surface RNA can be covalently modified, noncovalently tethered, or transiently adsorbed.
- **GlycoRNA:** RNA carrying a covalently attached glycan. GlycoRNA is a chemical class; surface exposure is a localization state, so neither term is a synonym for the other.
- **Glycocalyx:** The carbohydrate-rich layer at the extracellular face of a cell, conventionally built from glycoproteins, glycolipids, and proteoglycans and now understood to include some RNA-containing assemblies.
- **Cell-surface RBP:** An RNA-binding protein detected at the outer cell surface. A cell-surface RBP may tether, cluster, protect, or present RNA without being a transmembrane protein.
- **Heparan sulfate:** A sulfated glycosaminoglycan attached to proteoglycans at the cell surface and in extracellular matrices. Heparan sulfate can bind many basic proteins and, in defined systems, participates in ternary complexes with RNA and RBPs.
- **Surface topology:** The physical orientation and accessibility of a molecule relative to the plasma membrane. Outer-surface exposure must be distinguished from luminal, cytosolic, endosomal, or membrane-fragment association.
- **Extracellular carrier:** A particle or macromolecular association that protects RNA or changes its extracellular behavior, such as an EV, RNP, lipoprotein, virion, apoptotic body, or matrix complex.
- **Circulating RNA biomarker:** An RNA or RNA profile measured in blood or another biofluid for diagnosis, prognosis, classification, exposure assessment, or treatment monitoring.
- **Functional intercellular RNA transfer:** Movement of RNA from a donor to a recipient followed by molecular engagement and a causal recipient-cell effect.
- **Evidence ladder:** An ordered set of increasingly specific claims. Detection, external exposure, carrier association, molecular interaction, receptor recruitment, and biological function occupy different rungs.

## What to Know Before Reading This Chapter

The first prerequisite is membrane topology. The plasma membrane separates the cytosol from extracellular space. An RNA recovered with a membrane fraction might be cytosolic RNA attached to the inner leaflet, RNA trapped in an endosome, RNA inside a vesicle, RNA on the outer surface, or RNA released from damaged cells during fractionation. A surface claim therefore begins with intact, nonpermeabilized cells and a membrane-integrity test. An externally added nuclease cannot normally cross an intact plasma membrane; loss of an RNA signal after extracellular nuclease treatment supports outside accessibility only when viability and leakage controls remain acceptable.

The second prerequisite is the difference between a covalent modification and a molecular assembly. A glycoRNA has a glycan chemically attached to RNA. An RNA-RBP-heparan-sulfate complex may place RNA in the glycocalyx through noncovalent binding. Both may appear in a surface-capture assay, but only the first is RNA glycosylation. [Chapter 49](chapter1045.md) explains the attachment chemistry, acp3U-linked N-glycans, glycan biosynthesis, and chemical-assay inference. This chapter asks where these molecules are, what holds them at the surface, and what functions have been demonstrated.

The third prerequisite is claim-specific evidence. Sequencing identifies a sequence in a sample, not its physical carrier. Colocalization places two signals near one another at the resolution of the assay, not necessarily in direct contact. Proximity labeling identifies a local molecular neighborhood, not a stable stoichiometric complex. Receptor recruitment does not automatically establish receptor activation, and a phenotype after RNase treatment does not prove that one named RNA caused the effect. The sections below repeatedly move from an observation to the strongest claim that the experiment supports, then state the next control needed for a stronger inference.

## 107.1. Extracellular RNA carriers, vesicles, lipoproteins, and RNP particles

RNA enters extracellular space through secretion, membrane shedding, cell-cell contacts, extrusion of organelles or particles, apoptosis, necrosis, tissue injury, and sample-processing damage. The resulting material is a population of physical states rather than a uniform pool. The most useful first question is therefore, “What protects or presents this RNA?” before asking what the RNA does.

Extracellular vesicles are membrane-bounded particles released by cells. Small EV is an operational size-enriched category that can include endosome-derived exosomes and small plasma-membrane-derived vesicles. Larger ectosomes or microvesicles bud from the plasma membrane, whereas apoptotic bodies arise during programmed cell fragmentation and can contain RNA, DNA, proteins, and organelle material. The terms overlap when biogenesis is not directly observed. MISEV2023 recommends reporting source, isolation, size, density, abundance, and marker evidence rather than converting a size fraction into an unproven biogenesis label (Welsh et al., 2024).

![Figure 107.1. Carrier Taxonomy for Extracellular RNA](../assets/figures/chapter1102_figure1.png)

**Figure 107.1. Carrier Taxonomy for Extracellular RNA.** Extracellular RNA is a physical-state category. Orthogonal fractionation, topology perturbation, and marker evidence are required before a sequence detected outside cells can be assigned to a carrier.

RNA associated with an EV preparation may be intraluminal, embedded in the membrane, bound to its exterior, or carried by a contaminant that co-fractionates with the vesicle. These topologies predict different results. Intraluminal RNA should resist external RNase until detergent disrupts the membrane. Exterior RNA may be RNase sensitive without detergent. An RBP-shielded RNA may require protease before nuclease access. An aggregate can sediment with EVs but lack appropriate membrane markers and detergent behavior. No single treatment is decisive, because detergent, protease, and nuclease can alter particle integrity and expose secondary substrates. The pattern across orthogonal separations is more informative than any one assay.

EV cargo includes microRNAs, tRNA fragments, Y-RNA fragments, rRNA fragments, mRNA fragments, long noncoding RNA fragments, circular RNAs, mitochondrial RNAs, and viral RNAs. Cargo composition reflects some combination of intracellular abundance, RNA stability, RNA-binding proteins, sequence or modification-dependent sorting, stress, cell type, and degradation. Garcia-Martin et al. (2022) showed that defined microRNA sequence features can bias cellular retention versus small-EV release. This establishes selective sorting for that experimental system, not a universal export code for all EV RNAs.

Sharma et al. (2025) reported glyco-modified small noncoding RNAs as intraluminal exosomal cargo and observed transfer to recipient cells. This result raises an important topological boundary: luminal vesicle glycoRNA and outer cell-surface glycoRNA need not share the same retention mechanism, extracellular exposure, or function. It also illustrates why glycoRNA workflows require independent chemical and contamination controls. Evidence that a metabolic sugar label and RNA signal co-purify with an EV fraction is strengthened by nuclease topology, detergent dependence, particle purification, glycoprotein depletion, molecular linkage evidence, and recipient-cell tracking. The chemistry of the glycan-RNA bond remains the responsibility of [Chapter 49](chapter1045.md).

Nonvesicular carriers are abundant and biologically consequential. Much circulating microRNA is found with Argonaute or other RNPs. Lipoproteins can carry small RNAs and can change biodistribution through lipoprotein-receptor interactions. Ribosomal material, nucleoprotein complexes, extracellular protein aggregates, chromatin fragments, virions, and damaged-cell debris can overlap with EVs in size or density. Apoptotic bodies and other large particles can be depleted by low-speed centrifugation yet fragment during handling. A study that labels every high-speed pellet “exosomal RNA” collapses distinct carriers and topologies into one unsupported mechanism.

Carrier state also changes clearance and access. A soluble Argonaute complex can protect a small RNA from nuclease while remaining too large or chemically unsuitable for spontaneous membrane passage. A lipoprotein-associated RNA encounters receptors and clearance pathways that differ from those of an EV. An apoptotic body can be engulfed by a professional phagocyte, exposing its RNA to endosomal receptors after uptake. A surface-tethered RNA may act without leaving the donor cell at all. Thus, nuclease stability is not a general proxy for delivery competence: protection, biodistribution, cellular encounter, and productive compartment access are separate properties.

Source attribution is similarly conditional. Platelet activation can produce EVs and RNPs during phlebotomy; tissue injury can release both actively secreted particles and passive debris; infection can add virions or microbial particles; and tumors can change systemic blood-cell activation. Cell-type markers, paired tissue data, genetic tracing, and controlled release experiments can narrow the source, but an RNA sequence shared among tissues rarely identifies its origin by itself. A carrier profile therefore combines RNA identity with particle markers, proteins, lipids, and specimen context rather than relying on sequence alone.

Carrier assignment is strongest when the RNA co-fractionates with the proposed particle across at least two independent physical principles. Size-exclusion chromatography separates by hydrodynamic size; density gradients separate by buoyant density; immunocapture enriches a marker-positive subset; asymmetric-flow or other fractionation approaches add further resolution. Particle imaging, protein and lipid markers, negative markers, detergent sensitivity, nuclease accessibility, and targeted depletion then test whether the RNA behaves as predicted. Immunocapture does not recover marker-negative EVs, and a density peak can still contain different particles. The conclusion should name the preparation actually measured—for example, “small-EV-enriched fraction”—unless the biogenesis and topology are directly demonstrated.

> **Box 107.1. Carrier Assignment Is an Evidence Claim**
>
> An extracellular RNA sequence arrives without a carrier label. RNase resistance can arise from a membrane, protein shield, lipoprotein, virion, aggregate, or matrix. A high-speed pellet can contain several of these states. Immunocapture enriches the marker-positive subset rather than all vesicles. Assign a carrier only when the RNA follows the proposed particle across orthogonal separations, expected positive and negative markers behave correctly, and targeted disruption changes the RNA signal as predicted. When those conditions are not met, report the preparation that was measured—for example, “small-EV-enriched fraction”—rather than the biogenesis that was assumed.

Carrier abundance and cargo abundance must also be separated. A population can contain a high concentration of particles but very few copies of a particular RNA per particle. A bulk measurement can be consistent with a rare loaded subpopulation, many particles carrying fractional population-average occupancy, or an external RNP contaminant. Single-particle assays, limiting-dilution logic, independent copy-number measurements, and perturbation of the carrier help resolve these alternatives. The quantitative consequences become especially important when a carrier is proposed to deliver regulatory RNA to another cell.

## 107.2. Cell-surface RNA, glycoRNA-RBP assemblies, topology, and candidate functions

Cell-surface RNA is RNA accessible on the extracellular face of a living cell. This location differs from RNA inside a secreted vesicle and from cytosolic RNA near the inner plasma-membrane leaflet. It is also broader than glycoRNA. Surface RNA can include covalently glycosylated small RNAs, RNA held by extracellular RBPs, RNA associated with heparan sulfate or other glycocalyx components, and RNA temporarily adsorbed from the medium. The scientific task is to identify which state is present in a particular cell type and then test its function without using one state as a proxy for all others.

### Discovery changed the molecular boundary of the cell surface

Earlier work used surface-selective sequencing, intact-cell labeling, fluorescence in situ hybridization, and antisense probes to identify nuclear-encoded RNA exposed on mammalian cell surfaces. Huang et al. (2020) called these molecules membrane-associated extracellular RNAs and linked candidate surface RNAs to monocyte-endothelial adhesion. These experiments did not require every detected surface RNA to be glycosylated. They established the broader proposition that RNA can be a stable, cell-type-dependent component of the outer molecular surface.

Flynn et al. (2021) then provided evidence that conserved small noncoding RNAs can bear sialylated N-glycans. The study combined metabolic glycan labeling, RNA purification, enzymatic perturbations, genetic manipulation of glycan biosynthesis, mass spectrometry, and intact-cell measurements. GlycoRNAs occurred across cell types and mammalian species, and a large fraction of the detected glycoRNA signal was accessible at the cell surface. The work also reported interactions with Siglec-family lectins. These results introduced a third covalently glycosylated scaffold alongside proteins and lipids and connected RNA chemistry to the glycocalyx.

Chemical identity was subsequently strengthened by identification of 3-(3-amino-3-carboxypropyl)uridine, abbreviated acp3U, as an N-glycan attachment site (Xie et al., 2024). Native sialoglycoRNA enrichment, enzyme-dependent release, isotope behavior, and matching to synthetic acp3U and acp3U-GlcNAc standards supplied compound-level linkage evidence. The release workflow does not preserve the identity of the complete RNA molecule or measure the fraction of a named RNA species carrying the linkage, and acp3U has not been established as the only possible N-glycan attachment route. The result therefore anchors covalent chemistry without specifying how a glycoRNA travels through the secretory system, remains at the plasma membrane, presents its glycan outward in a particular cell type, or affects a receptor. Those are separate carrier, occupancy, trafficking, topology, and functional questions. [Chapter 49](chapter1045.md) owns the linkage chemistry and biosynthetic detail.

### Surface retention uses more than one physical model

The simplest model imagines a glycoRNA as a membrane-anchored glycoprotein, but RNA does not necessarily contain a transmembrane segment or lipid anchor. Surface retention can instead emerge from multivalent extracellular assemblies. An RBP can bind RNA and a second surface component. A basic RBP can associate with negatively charged heparan sulfate. A glycan on RNA can engage a lectin. Multiple weak interactions can concentrate RNA in nanodomains even when no single interaction permanently anchors it. Shedding, endocytosis, extracellular nuclease activity, proteolysis, and competition by soluble ligands can make these assemblies dynamic.

Perr et al. (2025) detected a group of RBPs on living-cell surfaces and found that surface RBPs organized into nanoclusters enriched for multiple RBPs and glycoRNAs. Extracellular RNase disrupted cluster organization. The clusters served as interaction sites for the HIV trans-activator of transcription cell-penetrating peptide, commonly called TAT; removal of surface RNA or loss of TAT RNA-binding activity reduced internalization. This study supports a functional surface RNA-RBP nanodomain in a defined entry process. The result does not imply that every cell-penetrating peptide uses the same domain, that every RBP in the neighborhood directly contacts the same RNA, or that the domain is a phase-separated condensate. “Nanocluster” describes spatial organization measured at the cell surface; it should not be upgraded to a thermodynamic phase-separation claim without the relevant physical tests.

Li et al. (2025) established a complementary presentation model. The investigators used Toll-like receptor 7 as an engineered surface-RNA probe, performed a genome-wide knockout screen, and identified heparan sulfate as important for cell-surface RNA presentation. Surface proximity labeling placed heparan-sulfate-associated RNAs near RBPs, and spatioselective RNA-protein crosslinking supported an RNA-RBP-heparan-sulfate ternary complex. The surface RNAs could recruit the immune receptor KIR2DL5 in human cell-line experiments. This work supports presentation and receptor recruitment. It does not yet establish a general endogenous signaling response, receptor activation in primary tissues, or organism-level immune outcome. TLR7 in this experiment was a probe; deep TLR7 sensing biology remains in [Chapter 108](chapter1103.md).

These models are compatible rather than mutually exclusive. Some glycoRNAs may participate in RBP-rich nanodomains. Some nonglycosylated RNAs may be presented by RBPs and heparan sulfate. An individual RNA might move between a vesicular lumen, a cell surface, a soluble RNP, and an endosome at different stages. The correct ontology therefore records at least four axes: chemical state, carrier or binding partners, topology, and biological context.

### Neutrophil recruitment supplies a causal adhesion example

Zhang et al. (2024) studied surface RNA on murine neutrophils. Removing accessible surface RNA impaired recruitment to inflammatory sites in vivo and reduced adhesion to and migration through endothelial monolayers. Neutrophil glycoRNA was predominantly surface exposed, P-selectin recognized the glycoRNA-associated surface state, and knockdown of mammalian SIDT-family RNA transporter homologs reduced glycoRNA expression and phenocopied aspects of surface-RNA loss. The causal chain joins topology, a defined cell type, an adhesion context, and an in vivo phenotype.

The experiment also teaches how to read perturbations. Extracellular RNase removes many accessible RNAs rather than one sequence and can alter an RNP network. SIDT perturbation may change RNA trafficking or other cell properties in addition to a named surface RNA. P-selectin binding places a receptor-ligand interaction in the mechanism, but adhesion occurs within a larger selectin, integrin, chemokine, and shear-force system. The strongest conclusion is that a surface-RNA state contributes to neutrophil-endothelial interactions in the tested models. Sequence-resolved sufficiency, exact glycan determinants, and generality across vascular beds remain narrower questions.

### Immune recognition can be enabled, recruited, or suppressed

Surface RNA sits in a location where immune receptors, antibodies, complement components, and extracellular nucleases can encounter it. That location does not predict a single immune outcome. A glycan may create a lectin ligand; an RNA-RBP-heparan-sulfate complex may concentrate a receptor; an exposed RNA may become immunostimulatory after uptake; or the glycan may mask an RNA feature that would otherwise activate an endosomal sensor.

Graziano et al. (2025) provided direct evidence for the masking model. Enzymatic removal of N-glycans from cell-culture-derived and circulating glycoRNA elicited inflammatory responses involving Toll-like receptor 3 and Toll-like receptor 7, whereas intact N-glycans concealed the immunostimulatory acp3U-containing RNA state. During efferocytosis—the uptake of apoptotic cells by phagocytes—glycosylated surface RNA helped prevent inappropriate inflammatory sensing of self RNA. Genetic deletion of DTWD2, an enzyme required for acp3U formation, and synthetic acp3U-containing RNAs connected the modified nucleotide to the observed sensing response. Within this chapter, the result establishes a surface and extracellular function: glycosylation can suppress exposure of an immunostimulatory RNA feature during dead-cell clearance. The downstream receptor signaling, cell-type-specific sensor network, and broader self/nonself logic belong to [Chapter 108](chapter1103.md) and [Chapter 109](chapter1104.md).

Receptor proximity and receptor activation must remain distinct. Flynn et al. showed Siglec interactions; Zhang et al. supported P-selectin recognition; Li et al. showed KIR2DL5 recruitment; Graziano et al. demonstrated inflammatory consequences after deglycosylation in defined sensor-dependent models. These studies support several different biological verbs—bind, recognize, recruit, mask, and signal. They should not be compressed into the statement that “glycoRNA is an immune receptor.” RNA is the ligand-bearing or assembly-forming molecule; the receptor and the measured response must be named.

### Topology and molecular identity require orthogonal tests

An intact-cell surface experiment should show that the plasma membrane remains impermeable during labeling or nuclease treatment. Viability dyes, cytosolic-enzyme release, intracellular-RNA preservation, and morphology provide complementary checks. Surface-selective biotinylation or enzymatic labeling should be tested for membrane impermeability. Imaging should include nonpermeabilized conditions, optical-section or super-resolution controls appropriate to the claimed scale, and colocalization analysis that does not infer direct binding from diffraction-limited overlap. Cell-surface capture should include negative intracellular markers and a known positive surface control.

Nuclease accessibility is powerful but not self-interpreting. RNase can remove surface RNA, disrupt RNA-mediated assemblies, expose or release proteins, activate stress responses, and damage membranes at excessive dose or duration. A useful design includes multiple nucleases or sequence-specific oligonucleotide perturbations, catalytically inactive controls where possible, washout or rescue, time courses shorter than transcriptional remodeling, and assays for cell integrity. Protease and heparinase perturbations can test RBP and heparan-sulfate contributions, but each changes many surface interactions. Convergent effects across independent perturbations support the model more strongly than a single enzyme treatment.

GlycoRNA identity adds a second control layer because each analytical family observes a different part of the conjugate. Metabolic sugar reporters can enter several glycoconjugate pathways, and reporter uptake and metabolic conversion make signal intensity difficult to compare quantitatively across cell types. Native periodate labeling such as RNA-optimized periodate oxidation and aldehyde labeling (rPAL) avoids precursor uptake, but it preferentially reports sialylated glycans and requires reaction conditions that limit background oxidation of terminal RNA ribose. Lectins and glycan-binding aptamers recognize selected motifs on proteins and lipids as well as RNA. Dual-recognition imaging can require both a glycan probe and an RNA-sequence probe, yet the resulting proximity remains resolution-limited rather than proof of a covalent bond. Released-glycan mass spectrometry can define carbohydrate structures while losing RNA identity; released-glyconucleoside analysis can define a linker while losing the complete RNA carrier and its fractional occupancy (Xie et al., 2024; Sun et al., 2026).

Purification creates a separate ambiguity. Phase-separation and silica-based RNA workflows can carry proteins or other glycoconjugates into an RNA fraction. Kim et al. (2025) and Kegel et al. (2025) showed that purification-dependent, RNase-insensitive glycoconjugates or glycoproteins can appear in preparations used for glycoRNA analysis. These findings do not erase orthogonal chemical evidence for covalent glycoRNA; they show that a labeled glycan signal in an RNA preparation is insufficient by itself. No single current readout simultaneously establishes RNA sequence, glycan structure, covalent linkage, intact surface topology, carrier identity, and fractional occupancy. A rigorous workflow therefore uses denaturing proteolysis, orthogonal RNA purification, nuclease and glycosidase controls, molecular linkage evidence, matched process blanks, and an assay that requires simultaneous recognition of the RNA and glycan components, with separate measurements for topology and abundance.

### Candidate functions should be graded by the endpoint measured

Surface RNA can plausibly organize adhesion, receptor presentation, barrier properties, particle uptake, cell-penetrating cargo entry, immune tolerance, and intercellular recognition. Evidence is already causal for some context-specific endpoints: neutrophil recruitment, TAT internalization, and glycan-dependent suppression of inflammatory sensing during efferocytosis. Other endpoints remain candidate functions supported by association, receptor proximity, or perturbations that affect many molecules. A chapter-level synthesis should therefore report the cell type, RNA state, perturbation, receptor or partner, endpoint, and level of causal resolution.

The distribution across cell types is also unresolved. Monocytes, neutrophils, cultured tumor cells, epithelial cells, and apoptotic cells have yielded different surface-RNA profiles or phenotypes. Surface composition can change with activation, differentiation, stress, cell cycle, extracellular nuclease exposure, and culture conditions. A surface state established in one transformed cell line should not be treated as a constitutive feature of every primary cell. Tissue imaging, primary-cell validation, in vivo perturbation, and sequence-resolved measurements are the next evidence steps.

## 107.3. Circulating RNA biomarkers and preanalytical confounders

Circulating RNA is RNA measured in blood or another moving biofluid. It can derive from regulated secretion, blood cells, tissue turnover, cell death, platelets, tumors, pregnancy, exercise, infection, or sample handling. A biomarker can be a single RNA, a ratio, a multi-RNA classifier, a fragment pattern, a modification signal, or a carrier-enriched signature. The biomarker need not cause disease. It must, however, be measured reproducibly in the intended population and add clinically useful information.

Intended use determines the required signal. An early-detection test must distinguish small disease-associated changes from age, inflammation, comorbidity, and organ injury in a low-prevalence population, where false positives can dominate. A treatment-monitoring marker can instead be evaluated within the same patient and may succeed by tracking change from a personal baseline. A prognostic marker must add information beyond stage, pathology, and other established variables. A pharmacodynamic marker may report that a pathway changed without predicting clinical benefit. These tasks cannot share one generic “accuracy” threshold.

Plasma and serum are different specimen systems. Serum is collected after clotting, during which platelets and leukocytes can release RNA and particles. Plasma avoids clotting but retains sensitivity to anticoagulant, centrifugation, processing time, and residual platelets. Hemolysis releases abundant erythrocyte-associated RNAs and can produce a false disease signature. Leukocyte lysis, platelet activation during phlebotomy, storage temperature, freeze-thaw cycles, tube additives, exercise, time of day, and medications can further change the measured pool. Max et al. (2018) showed that plasma and serum possess distinctive extracellular small-RNA profiles rather than interchangeable backgrounds.

**Table 107.1. Preanalytical Variables in Circulating RNA Studies.** Collection tube, processing delay, hemolysis, platelet activation, centrifugation, storage, and extraction can alter circulating-RNA composition before measurement; documentation and controls should connect each variable to its expected artifact.

| Variable | Mechanistic effect | Typical misleading result | Detection or documentation | Mitigation |
| --- | --- | --- | --- | --- |
| **Plasma versus serum** | Clotting activates platelets and can release leukocyte and platelet RNA | A specimen-type signature interpreted as disease biology | Record matrix and tube; compare matched specimens when changing matrix | Do not merge plasma and serum without explicit bridging validation |
| **Hemolysis** | Erythrocytes release abundant RNA and other molecules | False circulating microRNA or fragment signature | Spectral hemolysis index and erythrocyte-associated RNA ratios | Standardize phlebotomy and exclude or model affected samples using prespecified rules |
| **Residual platelets** | Platelets contribute RNA and EVs and respond to handling | Apparent carrier or activation difference | Platelet counts or markers after centrifugation | Use defined centrifugation and document residual platelet burden |
| **Processing delay** | Ongoing cell release, lysis, and nuclease action alter composition | Site or cohort difference | Timestamp collection, first processing, and freezing | Set and validate a maximum processing interval |
| **Collection tube** | Anticoagulant, preservative, and inhibitor chemistry alter recovery | Transcript-class or reverse-transcription bias | Report manufacturer and lot; use process controls | Select tube for intended analyte and lock before validation |
| **Freeze-thaw and storage** | Particle disruption, RNA loss, or fragment redistribution | Batch-associated abundance shift | Track cycles, duration, and temperature | Aliquot and use matched storage histories |
| **RNA purification** | Class-specific recovery and inhibitor carryover | Method-dependent biomarker panel | Spike-ins across sizes and chemistries; recovery comparison | Validate one workflow for the intended RNA classes |
| **Library construction** | Ligation, reverse-transcription, and terminal-chemistry bias | Fragment or isomiR distortion | Technical replicates and orthogonal assays | Keep platform fixed and validate key features independently |

The exRNAQC Consortium (2025) compared blood collection tubes and RNA-purification strategies for extracellular transcriptome profiling. The practical conclusion is not that one workflow is universally best. Different combinations recover different transcript classes and differ in stability, inhibition, and contamination. A study should select the specimen and extraction workflow for its intended analyte, lock that workflow before validation, and avoid merging cohorts collected under incompatible conditions. Van der Schueren et al. (2025) further documented incomplete reporting of preanalytical variables in plasma RNA studies, which makes replication and meta-analysis harder even when the underlying assay is sound.

Analytical choices add class-specific bias. Extraction kits differ in recovery of short RNAs, long fragments, circular RNAs, protein-bound RNA, and inhibitor carryover. Reverse transcriptases and ligases are sensitive to terminal chemistry and internal modifications. Short reads can map ambiguously among tRNAs, rRNAs, repetitive elements, isomiRs, microbes, and synthetic controls. Spike-ins measure selected process steps but do not define a biological denominator. Library-size normalization can fail when a few abundant fragments change sharply, whereas volume normalization, carrier normalization, or stable endogenous references answer different questions.

Feature identity needs molecular validation. A short read assigned to a tRNA fragment can also match another tRNA gene, a pseudogene, or a degradation product with a different end. A circular-RNA junction requires alignment and assay controls that exclude template switching or genomic rearrangement. A methylation- or glycosylation-associated enrichment signal can reflect altered chemistry, altered abundance, or carryover of another molecule. High-throughput discovery should therefore be followed by an orthogonal assay designed around the proposed molecular feature rather than merely repeating the same library chemistry.

Machine-learning classifiers amplify both signal and leakage. If all cases were collected at one hospital and all controls at another, a model can learn site, tube lot, storage time, or processing batch. Feature selection performed before train-test separation leaks outcome information. Repeated samples from one person can leak personal baselines across folds. A defensible analysis partitions patients and collection sites appropriately, freezes preprocessing and thresholds, reports calibration and decision curves where relevant, and tests performance in the setting in which the assay will actually be used.

A credible discovery study prespecifies specimen handling, extraction, quality-control exclusions, covariates, normalization, and model evaluation. A credible validation study locks the feature set and threshold, uses an independent cohort or prospective design, measures calibration as well as discrimination, and compares the RNA test with existing clinical information. Cross-validation within one batch is not independent clinical validation. Conversely, incomplete knowledge of tissue source or mechanism does not invalidate a biomarker if analytical performance and intended use are strong.

Carrier enrichment can improve a biomarker by reducing background or increasing tissue relevance, but it also adds another assay. An EV-enriched signature is reproducible only if particle isolation, recovery, purity, and RNA topology are stable across sites. A cell-surface-RNA marker requires a viable-cell or validated capture assay rather than total plasma sequencing. The physical state named in the biomarker must therefore be part of the locked test definition. Protocol-level detection and analytical-validation depth belongs to [Chapter 132](chapter1120.md).

## 107.4. Intercellular and cross-organism RNA transfer: uptake, stoichiometry, and functional evidence

Intercellular RNA transfer means that RNA produced in one cell reaches another cell. Functional communication is the narrower claim that the transferred RNA changes a defined recipient process. EVs, RNPs, lipoproteins, apoptotic material, tunneling structures, direct contacts, and exposed cell-surface assemblies can all participate. The route determines what must be measured: a surface-presented RNA may alter adhesion without entering the recipient, whereas a microRNA proposed to repress a cytosolic target must escape into the recipient cytosol and engage Argonaute.

Recipient-cell association is not equivalent to internalization. Surface-bound particles, trapped extracellular material, endosomal retention, lysosomal degradation, and transfer of a fluorescent lipid or dye without intact RNA can mimic delivery. A strong localization design therefore combines sequence-specific detection, intact-cargo tracking, microscopy, biochemical fractionation, and an effector or target assay. Orthogonal labeling methods can connect EV RNA to recipient-cell interactors, as illustrated by Zhang et al. (2025), but labels can change stability and trafficking and must be paired with unlabeled functional controls.

![Figure 107.3. Evidence Ladder for Intercellular and Cross-Organism RNA Communication](../assets/figures/chapter1102_figure3.png)

**Figure 107.3. Evidence Ladder for Intercellular and Cross-Organism RNA Communication.** Functional RNA communication is a chain, not a single detection event. The required recipient compartment follows from the proposed mechanism, but every route needs quantitative exposure, molecular engagement, and causal perturbation.

The causal chain has seven links. First, the donor produces the RNA. Second, a defined release or presentation pathway places it outside the donor. Third, the RNA survives in a carrier or assembly. Fourth, the relevant recipient encounters it at a plausible dose. Fifth, the molecule reaches the receptor, membrane domain, cytosol, nucleus, or endosome required by the proposed mechanism. Sixth, it engages a defined molecular partner or target. Seventh, perturbing the donor RNA changes the recipient outcome and a sequence- or mechanism-appropriate rescue restores it. Co-delivered proteins, lipids, metabolites, cytokines, virions, and cell debris are alternative causes at every step.

Each link has a different experiment. Donor-side genetics or sequence editing tests production. Release-pathway perturbation tests dependence on a vesicle, RNP, or surface-presentation system, but broad secretion inhibitors often affect many cargos. Fractionation and protection assays test transit state. Physiological dosing and tissue localization test encounter. Effector capture, target-site reporters, receptor competition, or direct binding test engagement. Recipient-cell genetics tests whether the proposed pathway is necessary. Rescue with a defined RNA distinguishes sequence or structure from the many other molecules changed by donor perturbation.

> **Box 107.3. Detection Is Not Functional RNA Communication**
>
> Ask seven questions. Did the donor produce the RNA? Did a defined route release or present it? Did the RNA survive in a protected state? Did the relevant recipient encounter a plausible dose? Did the intact molecule reach the membrane or intracellular compartment required by the mechanism? Did it engage a receptor, RBP, Argonaute, or molecular target? Did loss of the RNA remove the outcome and a mechanism-matched rescue restore it? A sequence can be detected after surface binding, contamination, label transfer, or endosomal degradation without answering the final questions.

Stoichiometry connects a detectable signal to biological plausibility. Suppose a preparation contains one target microRNA copy per one hundred particles on average, only one particle in ten reaches the relevant cell, one percent of internalized cargo escapes endosomes, and the recipient contains thousands of target transcripts. Detection of donor RNA after a large experimental dose could coexist with negligible repression at physiological exposure. The exact numbers vary, but the multiplication of inefficiencies is the key. Useful studies report RNA copies per volume and particle population, particle dose per recipient, internalized fraction, compartment-access fraction, target abundance, and dose-response. A population-average occupancy must not be rewritten as one RNA molecule in every vesicle.

Not all communication requires entry. The cell-surface examples in [Section 107.2](chapter1102.md) show that RNA can participate in adhesion, molecular recruitment, or peptide uptake at the plasma membrane. In such cases, receptor occupancy, surface density, shear conditions, and lateral organization can matter more than cytosolic copy number. The evidence standard changes with the mechanism, but it does not become weaker: surface retention, physical interaction, perturbation specificity, and causal endpoint still require independent tests.

For a microRNA-like intracellular mechanism, the strongest recipient evidence is loading of the transferred sequence into recipient Argonaute together with repression of a sequence-matched target and loss of repression after target-site mutation. Total recipient-cell RNA, fluorescence, or a generic stress response is weaker because none shows entry into the RNA-interference pathway. For an endosomal-sensor mechanism, cytosolic escape is unnecessary but delivery to the sensor-positive endosome, receptor dependence, and appropriate downstream signaling are required. For adhesion, the relevant test may instead be binding under physiological shear and receptor-specific blockade. Mechanism determines the endpoint.

Cross-organism systems provide stringent sequence-defined tests of transfer. Plant-pathogen, host-parasite, and engineered RNA-interference systems can connect a donor sequence to a complementary target and an Argonaute-dependent effect. Yet environmental carryover, dietary contamination, mapping ambiguity, and high experimental exposure can also create donor-like reads. [Chapter 113](chapter1107.md) owns plant immunity and agricultural RNA applications, [Chapter 111](chapter1106.md) owns organism-pair host-pathogen biology, and [Chapter 114](chapter1108.md) owns community inference. The bridge retained here is the causal transfer standard: identify the donor, protected route, recipient compartment, plausible dose, molecular target, loss, and rescue.

Broad claims that dietary microRNAs act as endocrine regulators in mammals remain high-threshold. Very low read counts can arise from index leakage, reagent contamination, sequence homology, or background unrelated to diet. Failure to observe uptake in one workflow does not prove that uptake is impossible, but detection after feeding does not establish physiological regulation. Dose scaling, tissue localization, molecular engagement, sequence-specific perturbation, and replication across laboratories are needed.

Soma-to-germline RNA communication illustrates another specialized route. Defined model-organism experiments can trace small RNAs or RNA-dependent information from somatic tissues to germ cells and then ask whether inherited molecular or phenotypic changes depend on the RNA pathway (Conine and Rando, 2022). Such evidence does not imply that any RNA found in adult blood can enter gametes or transmit an acquired state. The anatomical route, developmental timing, germline compartment, and persistence across generations must be measured.

## Experimental Foundations and Evidence

The same RNA sequence can support different biological conclusions depending on when the sample was collected, which physical compartment was isolated, and which controls survived the workflow. The experimental foundation is therefore organized by claim rather than by instrument: establish specimen integrity, assign a carrier or surface topology, validate molecular identity, select an abundance denominator, and add causal tests only after the earlier physical claims hold.

## 107.5. Biofluid sampling, carrier contamination, normalization, and evidence thresholds

Extracellular and surface RNA studies begin with a sampling model. Plasma, serum, urine, saliva, cerebrospinal fluid, milk, airway fluid, and tissue interstitial fluid differ in cell content, nuclease activity, protein concentration, inhibitors, collection volume, and normal physiological variability. Surface-RNA experiments add cell detachment, washing, temperature, culture medium, cell density, activation state, and viability. Every step can change the measured RNA before the intended assay begins.

### Build controls around the physical claim

For extracellular detection, document removal of intact cells and quantify residual-cell markers. For a carrier claim, combine fractionation with positive and negative particle markers and topology perturbations. For a surface claim, work on intact nonpermeabilized cells and show membrane integrity. For glycoRNA, demonstrate that the readout requires both RNA and glycan rather than a copurifying glycoprotein. For delivery, measure the relevant recipient compartment. For function, perturb the proposed molecule by at least two independent routes and test rescue. This claim-matched architecture is more informative than applying one generic control panel to every experiment.

![Figure 107.4. Control Architecture for Low-Biomass Extracellular and Surface RNA Studies](../assets/figures/chapter1102_figure4.png)

**Figure 107.4. Control Architecture for Low-Biomass Extracellular and Surface RNA Studies.** Controls become interpretable when tied to a physical failure mode. A blank detects contamination, an integrity assay protects topology inference, and rescue tests a causal mechanism; none substitutes for the others.

Extracellular RNA degradation can create stable fragments rather than simply removing signal. tRNA, Y-RNA, rRNA, and structured-RNA fragments can persist or be preferentially cloned. Some are regulated biological products; others arise after collection or during extraction. Conventional RNA-integrity numbers developed for cellular messenger RNA do not summarize these small-fragment pools. Process timing, defined spike-ins, fragment-size profiles, target-specific assays, and matched degradation controls are more useful.

Contamination occurs at multiple scales. Cellular carryover introduces erythrocyte, platelet, leukocyte, epithelial, sperm, or tissue RNA. Carrier co-isolation introduces lipoproteins, soluble RNPs, virions, protein aggregates, and membrane fragments. Molecular purification can retain glycoproteins or other glycoconjugates. Laboratory contamination includes synthetic oligonucleotides, extraction-column nucleic acids, aerosols, water, skin, and prior libraries. Sequencing contamination includes index hopping and barcode leakage. Collection blanks, extraction blanks, carrier-negative fractions, process controls, randomized batches, and replicate libraries address different sources and should be analyzed rather than merely reported.

Control abundance should bracket the sample regime. A blank with zero reads does not prove absence of low-level contamination if it was prepared or sequenced differently from the samples. A spike-in at a concentration orders of magnitude above endogenous RNA can track gross recovery while missing nonlinear loss near the detection limit. Positive controls should resemble the target in length, structure, chemistry, and carrier state closely enough to test the vulnerable step. Negative controls should pass through the same tubes, extraction batches, and library pools as biological specimens.

### Match topology assays to alternative models

An RNase-protection matrix is most informative when RNA is measured after RNase alone, protease plus RNase, detergent plus RNase, and matched untreated controls. RNase loss without detergent supports external exposure or lack of protection. Resistance until detergent is compatible with membrane enclosure. Sensitization by protease is compatible with protein shielding. These interpretations remain conditional because treatments can disrupt particles or cells. Fractionation and markers must agree with the proposed state.

For cell-surface RNA, use intact-cell nuclease accessibility alongside nonpermeant surface capture, imaging under nonpermeabilized conditions, and exclusion of cytosolic markers. Measure viability and leakage immediately, not only hours later. A rescue can be physical, such as restoring a defined surface RNA or RNP, or genetic, such as re-expressing a trafficking factor. A sequence-specific antisense reagent can provide resolution beyond broad RNase treatment, but extracellular hybridization may itself sterically alter adhesion or receptor binding. The interpretation should state what the reagent changed physically.

For glycoRNA, chemical specificity is indispensable. A useful chain combines glycan metabolic or enzymatic labeling, stringent denaturing RNA isolation, protease and nuclease controls, glycosidase sensitivity, RNA sequence identification, and direct or proximity-constrained linkage evidence. Dual-recognition assays such as RNA-sequence plus glycan proximity can improve specificity but inherit the spatial resolution and off-target properties of both probes. Deep protocol choices and analytical platforms belong to [Chapter 132](chapter1120.md), while linkage chemistry belongs to [Chapter 49](chapter1045.md).

Replication should cross the artifact boundary. Repeating the same extraction and sequencing library confirms technical repeatability but may reproduce the same bias. A stronger replication changes the vulnerable physical principle: surface capture plus intact-cell imaging, size separation plus density separation, metabolic glycan labeling plus direct linkage chemistry, sequencing plus digital PCR, or broad RNase perturbation plus a sequence-specific loss and rescue. Agreement across such methods narrows the set of shared artifacts.

### Choose a denominator before interpreting abundance

Normalization should reflect the question. Input-fluid volume estimates concentration per specimen volume. Particle normalization estimates RNA relative to a defined particle population. Total-RNA or library-size normalization estimates composition within recovered RNA. Cell normalization estimates release or surface abundance per source cell. Spike-ins estimate recovery through selected technical steps. None is universally correct, and each can move in the opposite direction when total RNA, particle abundance, or cell count changes.

Absolute measurements can clarify disagreements but require calibrated standards, recovery estimates, and a definition of the sampled compartment. Digital PCR can estimate target copies in an extract; it does not identify how copies are distributed across particles or surface domains. Particle counting can be biased by size and refractive index. Surface fluorescence is affected by probe accessibility and local density. Reporting raw quantities and all denominators allows readers to test alternative interpretations.

### Advance claims one rung at a time

The evidence ladder begins with detection. The next rungs are external location, carrier or surface assignment, molecular interaction, cell-type or tissue specificity, receptor or effector engagement, and causal function. A biomarker ladder differs: detection, analytical precision, association, locked classification, independent validation, clinical utility. A transfer ladder adds donor identity, route, recipient compartment, target engagement, loss, and rescue. A study can be rigorous while stopping on an early rung if the conclusion stops there too.

**Table 107.2. Evidence Thresholds by Extracellular and Surface RNA Claim.** Claims about extracellular location, carrier, surface topology, biomarker value, delivery, or functional communication require progressively different controls; detection or association alone cannot support uptake or biological effect.

| Claim | Minimal evidence | Stronger evidence | Common false inference | Best next test |
| --- | --- | --- | --- | --- |
| **RNA is extracellular** | Detection after documented removal of intact cells with blanks | Independent recovery, cellular-contamination markers, and fraction assignment | Any RNA in a biofluid tube is a stable extracellular signal | Repeat with matched cell-depleted fractions and lysis markers |
| **RNA is EV associated** | Cofractionation with characterized EV markers plus topology-compatible nuclease behavior | Orthogonal separation, contaminant depletion, detergent dependence, and carrier perturbation | Pelleting or RNase resistance proves exosome encapsulation | Compare size, density, immunocapture, protease, detergent, and RNase behavior |
| **RNA is on an intact cell surface** | Nonpermeant capture or extracellular nuclease accessibility with intact cells | Orthogonal imaging, surface sequencing, membrane-integrity controls, and rescue | Membrane-fraction recovery proves outer-surface exposure | Pair intact-cell topology with cytosolic-marker exclusion and a second physical method |
| **RNA is glycoRNA** | RNA-associated glycan signal with nuclease and glycosidase dependence | Direct linkage evidence, denaturing proteolysis, orthogonal purification, and dual RNA-glycan recognition; intact-transcript and occupancy measurements are additional rungs | Glycan signal in an RNA preparation proves covalent linkage, complete carrier identity, and uniform occupancy | Test copurifying glycoproteins, require molecular evidence for both components, and state which identity or abundance information the assay loses |
| **RNA recruits a receptor** | Receptor proximity changes with RNA perturbation | Direct interaction or competition, primary-cell validation, and dose response | Recruitment proves activation | Measure receptor-proximal biochemistry and downstream response |
| **RNA is a circulating biomarker** | Prespecified association under documented handling | Locked assay, independent cohort, calibration, and added clinical value | Discovery significance establishes clinical utility | Prospective blinded validation against the clinical comparator |
| **RNA is delivered** | Donor RNA reaches recipient above contamination background | Intact-cargo localization, compartment access, molecular engagement, dose response, loss, and rescue | Recipient association proves functional delivery | Measure the mechanism-required compartment and effector |

The surface-RNA comparison figure and table consolidate this claim-matched logic across the new molecular states. They return to the examples in [Section 107.2](chapter1102.md) while keeping chemical identity, topology, partner association, receptor recruitment, and function as independently testable propositions.

![Figure 107.5. Molecular States and Evidence Boundaries of Cell-Surface RNA](../assets/figures/chapter1102_figure5.png)

**Figure 107.5. Molecular States and Evidence Boundaries of Cell-Surface RNA.** Cell-surface RNA is a localization class containing several chemical and physical states. Covalent glycosylation, RBP or heparan-sulfate presentation, nanoscale clustering, receptor recruitment, and function are separate claims that require separate evidence.

**Table 107.3. Cell-Surface RNA Models, Representative Evidence, and Remaining Boundaries.** Cell-surface RNA may be membrane-tethered, glycan-associated, receptor-bound, vesicular, or adsorbed, and each model supports different verbs; orthogonal topology and contamination controls are needed before assigning chemistry or signaling.

| Surface-RNA state or model | Representative evidence | Strongest supported verb | What is not yet implied | Orthogonal control |
| --- | --- | --- | --- | --- |
| **Covalent sialylated glycoRNA** | Metabolic labeling, glycan-pathway genetics, enzymatic tests, intact-cell exposure, and acp3U linkage evidence | Exists and can be surface displayed | One resolved linker identifies every intact RNA carrier, establishes uniform occupancy, or explains all surface retention | Direct linkage plus intact-transcript identity, occupancy, intact-cell topology, and glycoprotein-depletion controls |
| **Broad nonglycoRNA-specific surface RNA** | Surface-selective sequencing, nonpermeant probing, imaging, and sequence-specific perturbation | Is externally accessible and cell-type dependent | Every recovered sequence is stably anchored or functional | Membrane integrity, independent probe, wash resistance, and rescue |
| **GlycoRNA-csRBP nanodomain** | Surface proteomics, nanoscale imaging, RNase disruption, and peptide-entry perturbation | Clusters and supports TAT entry in a defined system | Liquid phase separation or universal cell-penetrating-peptide entry | Pairwise binding tests, dynamics, primary-cell validation, and matched peptide controls |
| **HS-RNA-RBP ternary presentation** | Genome-wide screen, HS perturbation, proximity labeling, and spatioselective crosslinking | Presents RNA and recruits KIR2DL5 in tested cells | KIR2DL5 activation or a general immune outcome | Primary-cell signaling, competition, stoichiometry, and in vivo perturbation |
| **Neutrophil surface glycoRNA state** | Intact-cell RNA removal, adhesion and transmigration assays, in vivo recruitment, P-selectin recognition, and SIDT perturbation | Contributes to neutrophil recruitment | One identified RNA or glycan is sufficient in every vascular bed | Sequence-resolved perturbation, physiological-flow testing, and rescue |
| **Immune-masked glycoRNA on apoptotic cells** | Deglycosylation, sensor dependence, DTWD2 genetics, synthetic acp3U RNA, and efferocytosis assays | Prevents inflammatory self-RNA sensing in defined models | All glycoRNA is immunosuppressive in every context | Tissue-specific loss and rescue with sensor- and dose-resolved outcomes |
| **Transiently adsorbed surface RNA** | Wash-sensitive or medium-dependent external RNA association | Is present at the assay time | Regulated export, stable retention, or endogenous function | Medium exchange, kinetic washout, source tracing, and surface-partner perturbation |

Negative results also require calibration. Failure to detect a surface RNA may reflect probe inaccessibility, nuclease degradation, low abundance, cell-state dependence, or an assay optimized for a different RNA class. Failure to observe EV delivery at physiological dose can constrain a proposed mechanism even when delivery occurs after concentrated dosing. Disagreement between biochemical purification and intact-cell imaging may reveal different molecular states rather than simple technical error. The productive response is to name the tested topology and sensitivity limit.

> **Box 107.4. Five Verbs That Should Not Be Collapsed**
>
> - **Bind:** purified or cell-associated molecules interact under defined conditions.
> - **Recognize:** a receptor or probe discriminates an RNA-containing state from controls.
> - **Recruit:** a receptor becomes enriched near a surface RNA or domain.
> - **Mask:** a modification prevents receptor access to an otherwise stimulatory feature.
> - **Signal:** receptor-proximal biochemistry and downstream cellular responses change.
>
> Binding can occur without recruitment, recruitment without activation, and masking without any receptor interaction at the untreated surface. A complete statement names the cell type, RNA state, partner, perturbation, and endpoint.

## Biological Contexts Across Systems

In blood, extracellular RNA reflects a rapidly changing mixture of vascular cells, blood cells, platelets, tissue turnover, renal and hepatic clearance, immune activation, and clinical handling. Neutrophils provide a mechanistic surface-RNA example because adhesion and transendothelial migration can be measured under flow and in vivo. Monocytes provide another because cell-type-resolved surface probing has connected candidate RNAs to endothelial adhesion. These cases should not be generalized to erythrocytes, lymphocytes, endothelium, or tumor cells without direct topology data.

Apoptotic cells provide a distinct context. Their membranes initially preserve an ordered surface while exposing “eat-me” signals, and phagocytes clear them through efferocytosis. Glycosylated surface RNA can coexist with this recognition system and, in defined models, prevent self RNA from provoking endosomal sensors after uptake. Late apoptosis and secondary necrosis change membrane permeability, release intracellular RNA, and transform the sampling model. Experiments must stage cell death and separate intact apoptotic bodies from lysed material.

Tumors and inflamed tissues can alter EV release, glycosylation, RBP localization, cell death, and vascular permeability at the same time. A circulating signature may therefore report tumor burden, immune composition, tissue injury, or treatment response rather than tumor-specific secretion. Cell-line surface nanodomains can identify a mechanism worth testing, but primary tumors add extracellular matrix, stromal cells, hypoxia, proteases, and heterogeneous vascular access.

In epithelial barriers, airway secretions, saliva, urine, and milk, surface and extracellular RNA encounter mucus, antimicrobial proteins, high nuclease activity, microbes, and mechanical flow. An RNA may function locally at the surface without entering circulation. Abledu et al. (2025), for example, linked perturbation of alveolar epithelial surface glycoRNA to barrier-associated phenotypes in a defined system; broader respiratory mechanisms require replication and separation of direct RNA effects from global surface remodeling.

Plants and host-pathogen interfaces contain different walls, extracellular matrices, vesicles, and uptake routes. Comparative examples can test protection, transfer, and sequence-specific effects, but their cellular architecture should not be mapped directly onto mammalian glycocalyx models. Detailed plant and organism-pair mechanisms hand off to [Chapter 113](chapter1107.md) and [Chapter 111](chapter1106.md).

## Technology, Computational, Clinical, and Engineering Links

Surface RNA creates engineering opportunities and liabilities. A cell-penetrating peptide may use an RNA-RBP domain as an entry site, suggesting that surface RNA composition can influence delivery. Heparan-sulfate-associated RNA assemblies could affect how basic proteins, peptides, or nanoparticles bind. These insights may inform targeting, but endogenous association does not guarantee safe or efficient drug delivery. Therapeutic dosing, endosomal escape, biodistribution, immunogenicity, manufacturing, and release testing belong to Chapters [153](chapter1137.md)-[157](chapter1140.md).

EVs are being engineered as RNA carriers because membranes can protect cargo and some particles display cell-interaction molecules. The translational barriers are cargo loading, particle heterogeneity, purity, potency, scalable manufacturing, storage, biodistribution, and unwanted immune or coagulation effects. An engineered EV is a drug product, not merely a concentrated endogenous secretion. Its RNA topology, particle dose, contaminants, and functional assay must be defined.

Computational analyses of extracellular RNA must model short-fragment ambiguity, multiple origins, batch structure, and nonrandom missingness. A small read can match an isomiR, tRNA fragment, repeat, microbial sequence, or synthetic oligonucleotide. A surface-enrichment dataset adds contamination from lysed cells and abundance-dependent probe accessibility. Useful pipelines preserve multi-mapping information, report annotation versions, include negative controls in the statistical model, and validate classification across collection sites.

Clinical translation requires an intended-use statement. A test for early cancer detection, treatment response, organ injury, or inflammatory state has a different acceptable threshold, comparator, and prevalence. Carrier- or surface-defined biomarkers require an assay that reproduces that physical state, not merely the same RNA sequence in total plasma. Mechanistic uncertainty is acceptable for a validated marker; analytical ambiguity is not.

## Recent Consensus

Extracellular host RNA is an established biological and analytical category, but carrier and topology must be demonstrated rather than assumed. EVs carry RNA, yet RNPs, lipoproteins, virions, aggregates, and debris are common alternative or co-isolated states. Operational nomenclature and orthogonal characterization are the current standard.

Cell-surface RNA is now supported by multiple intact-cell, biochemical, imaging, genetic, and functional approaches. Covalently glycosylated RNA is one surface-RNA class, and acp3U-linked N-glycan chemistry provides a defined molecular anchor. Surface RNA also participates in noncovalent RBP- and heparan-sulfate-containing assemblies. The field has moved beyond asking whether surface RNA exists to asking which molecular states occur in each cell type and what each state does.

Functional evidence is context-specific. Neutrophil recruitment, TAT internalization through surface RNA-RBP domains, and glycan-dependent suppression of inflammatory sensing during efferocytosis have causal support in defined models. Receptor binding or recruitment findings provide strong mechanistic leads but should not be generalized to universal signaling functions. GlycoRNA purification can be confounded by copurifying glycoproteins and other glycoconjugates; direct linkage and dual-component controls are therefore essential.

Circulating RNA remains promising for biomarkers, but collection, processing, extraction, normalization, and independent validation determine whether a signature is portable. Functional intercellular transfer is credible in defined systems, yet donor-like sequence detection alone is insufficient. The consensus evidence chain includes physical state, dose, recipient compartment, molecular engagement, and perturbation-rescue logic.

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

Open questions:

- Which RNA sequences and modification states occupy the surface of each primary cell type under homeostasis, activation, stress, differentiation, and disease?
- What fraction of surface RNA is covalent glycoRNA, RBP-tethered RNA, heparan-sulfate-associated RNA, or transiently adsorbed extracellular RNA?
- How do glycoRNAs and nonglycosylated RNAs traffic to, remain at, and leave the cell surface?
- Are RNA-RBP surface nanodomains stable complexes, rapidly exchanging networks, or several physically distinct assemblies?
- Which receptor-recruitment events produce signaling, adhesion, internalization, tolerance, or no measurable downstream response in primary tissues?
- How general are the neutrophil-recruitment and efferocytosis mechanisms across species, vascular beds, inflammatory states, and disease?
- How many RNA molecules occupy each carrier, surface domain, or recipient cell, and which subpopulation supplies the functional dose?
- Which circulating RNA signatures remain calibrated across centers when collection and purification are standardized prospectively?

Controversies:

- Current evidence supports covalent glycoRNA, but some widely used enrichment workflows also recover RNase-insensitive glycoproteins or other glycoconjugates. The dispute concerns assay specificity and the identity of each recovered species, not a license to treat every glycan signal as RNA or to dismiss orthogonal linkage evidence.
- Receptor binding, recognition, recruitment, and signaling are sometimes merged in summaries of surface RNA. These are different endpoints. Each receptor-context claim should retain the exact experimental verb.
- High experimental concentrations can demonstrate that EV RNA or surface RNA is capable of an effect without showing that the same mechanism operates at physiological abundance.

Deprecated or weakened claims:

- “Extracellular RNA is mainly exosomal microRNA” is too narrow. Extracellular RNA spans many RNA classes and vesicular, protein-bound, lipoprotein-associated, surface-bound, and debris-associated states.
- “RNA is absent from the outer surface of intact mammalian cells” is no longer tenable. Multiple independent approaches now support surface-exposed RNA, although its composition and anchoring differ by context.

Common misconceptions:

- “All cell-surface RNA is glycoRNA.” GlycoRNA is a covalently modified chemical class; nonglycosylated RNA can also be held at the surface by RBPs, heparan sulfate, or other interactions.
- “Any glycan detected in an RNA purification proves covalent glycoRNA.” Glycoproteins and other glycoconjugates can co-purify, so direct linkage, denaturing proteolysis, nuclease controls, and orthogonal purification are required.
- “RNase sensitivity proves that a phenotype is caused by one named surface RNA.” RNase can remove many RNAs and reorganize RNP assemblies; sequence-resolved perturbation and rescue are needed for a single-RNA claim.
- “Colocalization or proximity labeling proves direct binding.” These assays define neighborhoods at a method-dependent resolution; crosslinking, reconstitution, competition, or structural evidence is needed for direct interaction.
- “Receptor recruitment proves receptor activation.” Recruitment can increase proximity without producing signaling. A downstream biochemical or cellular endpoint must be measured.
- “RNase resistance proves vesicle encapsulation.” Protein, lipoprotein, virion, aggregate, or matrix association can also protect RNA.
- “A significant plasma RNA difference is automatically a clinical biomarker.” Clinical use requires a locked assay, appropriate comparator, calibration, independent validation, and evidence that the result improves a decision.
- “Detecting donor RNA in a recipient proves functional transfer.” Surface binding, contamination, endosomal sequestration, and label transfer can mimic delivery; molecular engagement, dose, loss, and rescue complete the mechanism.