# Chapter 52. Biological Dynamics and Cross-Mark Synthesis of RNA Modifications

## Scope Note

This chapter synthesizes biological dynamics across RNA-modification chemistries, RNA classes, RNP states, cell states, and organisms. It owns installation/removal as biological regulation, stoichiometry, cross-mark coupling, development, stress, immunity, disease, and controversies. Comparative catalytic mechanisms, enzyme recognition, kinetics, specificity, evolution, and inhibitors belong to [Chapter 47](chapter1163.md); glycoRNA chemical identity and biosynthetic evidence to [Chapter 49](chapter1045.md); cell-surface RNA topology and extracellular assemblies to [Chapter 107](chapter1102.md); evidence adjudication to [Chapter 46](chapter1043.md); and assay workflows to [Chapter 132](chapter1120.md).

## Executive Summary

RNA modification dynamics describe changes in the chemical state of RNA molecules across time, RNA maturation, cellular condition, organismal development, stress, infection, and disease. The word "dynamic" should be used carefully. A dynamic signal can mean active installation of a mark on newly synthesized RNA, enzymatic removal from existing RNA, selective decay of modified or unmodified molecules, altered transcription or processing of the RNA substrate, redistribution of cell types in a mixed sample, or changes in protein binding that affect detection. Only some cases demonstrate true reversible chemistry on the same RNA molecule. Reviews of dynamic RNA modifications emphasize that the strongest examples combine modification mapping, enzyme perturbation, chemical validation, and biological readout rather than relying on differential enrichment alone.

Stoichiometry is the fraction of molecules carrying a defined modification at a defined site in a defined RNA population. Stoichiometry is different from peak height, read count, writer expression, or total modified nucleoside abundance. It is the denominator-aware quantity that tells the reader whether a site marks nearly every molecule, a substantial subset, or a rare state. This distinction matters because high-occupancy tRNA or rRNA modifications can define mature RNP function, whereas low-occupancy mRNA marks may affect specialized RNA subpopulations, transient life-cycle stages, or only a small part of the transcript pool.

RNA modifications rarely act alone. A modification changes the physical and biological state of an RNA only in an RNA-protein complex, cellular compartment, and temporal window. A methylated adenosine can alter local RNA structure, recruit a reader protein, affect splice-site choice, change mRNA decay, or become irrelevant if the RNA is already bound by another protein or rapidly degraded. A tRNA anticodon-loop modification can interact with neighboring tRNA modifications and with codon usage. A viral RNA modification can interact with viral replication strategy and host immune sensing. Emerging molecule-resolved and RNP-resolved observations expose co-occurrence and protein-assembly context, but the field still lacks comprehensive maps that connect chemical identity, stoichiometry, isoform, RNP composition, and RNA fate.

Disease associations are abundant but uneven in evidence strength. RNA modification enzymes and reader proteins are often altered in cancer, immune activation, infection, fibrosis, neurological disease, and developmental disorders. Some studies connect a specific enzyme, RNA substrate, modified site, and phenotype, as in NAT10-mediated ac4C modification of ITGB5 mRNA in pancreatic ductal adenocarcinoma and NSUN2-mediated m5C modification of hepatitis B virus RNA. Many other reports remain correlative because writer expression or global modification abundance does not establish that a particular substrate-level event causes disease. Clinical interpretation requires causal mechanism, patient-context validation, and a safety framework before a pathway becomes a biomarker or target.

The main biological controversies concern generalization across marks. Not every dynamic signal reflects reversible chemistry, not every low-occupancy site is biologically negligible, not every writer or reader has one substrate class, and not every mark has a context-independent effect. Cross-mark claims are strongest when they distinguish co-occurrence on the same molecule from co-variation across cell populations and when they connect stoichiometry to a defined RNP transition or fate. General evidence adjudication belongs to [Chapter 46](chapter1043.md), and method-specific failure modes to [Chapter 132](chapter1120.md).

## Concept Inventory

- **RNA modification:** a covalent chemical feature of an RNA nucleotide or RNA end that differs from the canonical unmodified ribonucleotide state expected for that RNA. Examples include N6-methyladenosine, abbreviated m6A; 5-methylcytidine, abbreviated m5C; pseudouridine; N4-acetylcytidine, abbreviated ac4C; inosine; 2′-O-methylated ribose residues; and complex tRNA modifications. [Chapter 49](chapter1045.md) treats individual chemistries in more detail. This chapter asks how the amount, location, context, and interpretation of such marks change.
- **Modification installation:** the enzymatic addition or generation of a mark on RNA. The enzyme or enzyme complex is often called a writer. A writer may act co-transcriptionally on nascent RNA, during precursor processing, during RNP assembly, or on mature RNA. Installation depends on enzyme abundance, catalytic state, cofactors, RNA sequence, RNA structure, subcellular localization, competing RNA-binding proteins, and the lifetime of the RNA substrate.
- **Modification removal:** direct chemical reversal or excision of a mark from RNA. The enzyme is often called an eraser. The evidence standard for an eraser is high because a population-level decrease in a mark can result from RNA decay, dilution by newly made unmodified RNA, altered cell composition, or loss of the modified RNA substrate. True reversible demethylation is best established for some m6A-related pathways; many other RNA modifications are better interpreted as installed marks whose apparent turnover occurs through RNA processing and decay unless direct removal has been demonstrated.
- **Modification turnover:** the change in modification signal over time. Turnover can include installation, removal, RNA decay, RNA maturation, RNA export, RNP remodeling, selective translation, or changes in cell population composition. A turnover experiment should therefore define whether the measured quantity is modified nucleoside abundance, site occupancy, modified RNA abundance, reader occupancy, or downstream phenotype.
- **Stoichiometry:** the fraction of eligible RNA molecules that carry a specified mark at a specified site under a specified condition. A statement such as "this transcript is m6A modified" is incomplete without stoichiometry or a statement that stoichiometry is unknown. A site with 10 percent occupancy and a site with 90 percent occupancy can have different biological interpretations even if both are detected reproducibly.
- **Cell-state dependence:** that modification installation, stoichiometry, reader engagement, or consequence differs among cell types, differentiation stages, activation states, stress states, disease states, or infection stages. In mixed tissue, a modification change can reflect a change inside each cell, a change in the fraction of cell types, or both. Single-cell transcriptomics can annotate cell states, but most single-cell RNA-seq workflows do not directly quantify RNA modifications.
- **Modification crosstalk:** coupling among modifications or between modifications and RNP context. Crosstalk can be direct, such as two nearby marks affecting the same RNA structure or reader-binding surface. It can be enzymatic, such as one modification changing access for another writer. It can be life-cycle based, such as modification interacting with splicing, 3′ end formation, translation, or decay. It can also be apparent rather than direct, when two marks change together because the RNA substrate or cell state changes.
- **Cross-mark co-occurrence:** two or more modifications present on the same RNA molecule or molecular complex in a defined state. Co-occurrence is stronger than bulk co-variation, which can arise because RNA abundance, cell composition, or developmental state changes both marks without direct molecular coupling.

## What to Know Before Reading This Chapter

The reader should know that an RNA molecule has a sequence, a chemical backbone, a folded structure, and a set of binding partners. A modification is not a free-floating regulatory label. It is a chemical feature embedded in that physical RNA molecule. The same mark can have different consequences depending on whether the RNA is a tRNA anticodon stem-loop, an rRNA decoding-center residue, an mRNA stop-codon-proximal region, a viral RNA genome, or a synthetic therapeutic RNA.

The reader should also know the difference between abundance and occupancy. RNA abundance is the number of molecules of an RNA species. Modification occupancy is the fraction of those molecules carrying the mark at a site. If a stress response doubles the abundance of an mRNA and leaves the number of modified molecules unchanged, an enrichment experiment can appear to change because the denominator changed. If the RNA abundance falls but the modified subpopulation is stable, the apparent modification fraction can increase. Interpreting dynamics without modeling RNA abundance can therefore produce wrong mechanisms.

Finally, the reader should separate association from causation. A disease sample can show higher writer expression, higher global modification levels, or stronger enrichment peaks. Those observations may be useful biomarkers, but they do not by themselves show that the modification causes disease. A causal claim needs a defined RNA substrate, a defined modification event, perturbation and rescue when possible, downstream molecular consequences, disease-relevant models, and controls for pleiotropic enzyme effects.

## 52.1. Modification installation, removal, and turnover as biological regulation

Modification installation begins when an enzyme or enzyme complex encounters an RNA substrate in a compatible chemical and cellular context. A writer enzyme must recognize enough information to choose the right RNA and nucleotide. That information can include sequence motif, RNA secondary structure, position relative to a stop codon or splice junction, transcript region, subcellular compartment, transcriptional timing, RNA-binding protein partners, and RNP assembly state. For example, m6A installation in many mammalian mRNAs depends on a methyltransferase complex that reads sequence and transcript context rather than modifying every adenosine. In contrast, many tRNA and rRNA modifications are installed at stereotyped positions during RNA maturation and RNP assembly. The word "writer" covers both cases, but the biological logic differs.

The causal steps for installation are useful to state explicitly. First, the RNA substrate must be produced and accessible. Second, the writer or writer complex must localize to the same compartment or nascent RNA context. Third, the enzyme must bind or be recruited to the RNA. Fourth, the active site must place the target atom and donor molecule in the correct geometry. Fifth, the modified RNA must survive long enough for the mark to be detected or to affect function. A failure at any step can reduce the observed modification signal. A change in modification after stress can therefore arise from altered writer expression, altered writer localization, altered RNA structure, altered transcription rate, altered RNA stability, or altered competition by RNA-binding proteins.

![Figure 52.1. Biological Routes to a Dynamic Modification State](../assets/figures/chapter1048_figure1.png)

**Figure 52.1. Biological Routes to a Dynamic Modification State.** A flow diagram separates altered installation, direct removal where demonstrated, selective RNA decay, synthesis and dilution, isoform switching, and changing cell-state composition. The visual emphasizes that several biological routes can yield the same observed increase or decrease in a marked population.

Removal is more restrictive than loss of signal. Direct removal requires an enzyme that chemically reverses the mark or otherwise restores an unmodified nucleotide state. The m6A field made the writer-reader-eraser vocabulary popular because demethylase activities offered a model for reversible methylation. Even there, the appropriate interpretation is enzyme- and context-specific. For other modifications, such as many tRNA hypermodifications, rRNA modifications, pseudouridine, and ac4C, a decrease in cellular signal often reflects reduced installation, altered RNA maturation, or degradation of modified RNA rather than active removal from a mature RNA molecule. Reviews of m5C and dynamic modifications warn against treating every differential signal as evidence of a reversible epitranscriptomic switch.

> **Box 52.1. Dynamic Does Not Necessarily Mean Reversible**
>
> Use m6A, m5C, pseudouridine, and stable-RNA examples to compare direct removal with loss through RNA turnover or dilution. The teaching goal is to make the biological route explicit before assigning an eraser.

Turnover combines chemistry with RNA life history. A newly transcribed pre-mRNA can be modified co-transcriptionally, spliced, exported, translated, stored, localized, or degraded. If modified molecules decay faster, the steady-state occupancy can remain low even when installation is frequent. If unmodified molecules are preferentially degraded, occupancy can rise without an increase in writer activity. If a modification promotes reader binding that recruits decay factors, installation and removal cannot be interpreted without RNA decay kinetics. [Chapter 35](chapter1033.md) treats mRNA decay pathways, and [Chapter 36](chapter1034.md) treats codon optimality-linked stability; those pathways provide the denominator for modification turnover.

Nascent RNA illustrates why timing matters. Xu and colleagues reported dynamic control of chromatin-associated m6A methylation linked to nascent RNA synthesis, showing that modification biology can be coupled to transcription rather than acting only on finished cytoplasmic mRNAs. A nascent mark can influence processing or RNP recruitment before the RNA is counted in steady-state poly(A) RNA. Conversely, a steady-state mRNA map can miss short-lived modification events that occurred during transcription or nuclear processing. Time-resolved designs, metabolic labeling, chromatin-associated RNA fractionation, and matched input measurements are therefore essential for understanding installation timing.

RNA class determines the expected turnover regime. Transfer RNA and rRNA modifications are often coupled to maturation and quality control, and mature molecules can be long-lived. A defect in modification can block biogenesis, create unstable intermediates, or impair translation. Messenger RNA modifications occur on shorter-lived and more heterogeneous substrates. Viral RNA modifications occur during infection, often under changing host-defense and replication conditions. Plant pseudouridine biology adds organism-specific developmental and stress contexts, including chloroplast and nuclear RNA pathways that should not be assumed to match mammalian mRNA rules.

The key boundary case is apparent dynamics caused by cell composition. A tissue sample can show increased m6A, m5C, pseudouridine, or ac4C signal because a cell type with high modification occupancy became more abundant. Microglia studies and cancer organoid studies show how strongly tissue and tumor samples can vary by cell state and intercellular context, even when they are not themselves direct modification maps. A modification dynamics study in tissue should therefore ask whether the same cells changed their modification state, whether the proportions of cells changed, or whether both occurred.

## 52.2. Biological stoichiometry and cell-state dependence

Stoichiometry is the central quantitative variable for this chapter. At a single site, stoichiometry can be written as the number of molecules carrying the modified residue divided by the number of molecules that contain that residue and are eligible for modification. The denominator is not trivial. It may be a transcript isoform, a mature RNA class, a cellular compartment, a viral genome population, or a purified tRNA species. If the denominator changes, the apparent stoichiometry can change even when the number of modified molecules does not.

![Figure 52.2. Stoichiometry Is a Denominator-Aware Fraction](../assets/figures/chapter1048_figure2.png)

**Figure 52.2. Stoichiometry Is a Denominator-Aware Fraction.** A numerator-over-denominator diagram compares a nearly saturated stable-RNA site, a low-occupancy mRNA isoform, a viral RNA population that changes with infection stage, and a rare cell state. Identical percentages can carry different biological meanings because molecule lifetime, RNP fate, and eligible populations differ.

Consider an mRNA site with 20 percent m6A occupancy in proliferating cells and 40 percent occupancy after differentiation. That twofold increase could mean each molecule has become more likely to be methylated. It could also mean the methylated isoform became more abundant, an unmodified isoform was degraded, or a cell subpopulation expanded. The biological interpretation therefore requires RNA abundance, isoform usage, site occupancy, and cell-state composition to be considered together.

**Table 52.1. What a Dynamic Modification State Can Mean.** A changing RNA-modification signal can reflect altered writing, erasing, decay, cell composition, compartment, or RNA age; time, occupancy, and matched population measurements are needed to distinguish these models.

| Biological observation | Compatible biological models | Distinguishing variable | Unsafe inference |
| --- | --- | --- | --- |
| **Marked fraction rises after stress** | Faster installation, slower decay of marked RNA, isoform shift, or state-composition change | Time-resolved occupancy within a defined RNA and cell-state denominator | Stress activates reversible marking |
| **Marked fraction falls during differentiation** | Reduced installation, selective loss of marked molecules, or lineage-composition shift | RNA birth, maturation, decay, and lineage-resolved state | An eraser removed the mark |
| **Reader-bound fraction rises** | More marked substrate, changed reader abundance, or altered RNP assembly | Modification dependence and matched RNP state | Reader binding proves more modification |
| **Two marks rise together** | Direct crosstalk, shared upstream regulation, or abundance/composition change | Same-molecule coexistence and causal transition order | The marks cooperate directly |

Quantitative studies across m6A, pseudouridine, and tRNA modification show that occupancy must be assigned to a defined site and RNA population. Their shared biological lesson is not that all RNA classes should have the same occupancy distribution. Mature tRNAs and rRNAs often carry high-occupancy modifications coupled to biogenesis, whereas an mRNA site can mark a transient subset of one isoform. Assay mechanisms and quantitative benchmarking are treated in [Chapter 132](chapter1120.md).

**Table 52.2. Stoichiometry Across RNA Classes and Cell States.** Modification occupancy uses different denominators across messenger RNA, transfer RNA, ribosomal RNA, and mixed cell states, so identical percentages can represent different biological populations and should not be compared without denominator control.

| RNA population | Relevant denominator | Biological interpretation of occupancy | Boundary case |
| --- | --- | --- | --- |
| **mRNA** | One transcript isoform in a defined compartment and state | May identify a transient or fate-biased subpopulation | Isoform abundance can change the denominator |
| **tRNA** | One mature tRNA species or isodecoder | Often reports maturation and decoding competence | Pooling isodecoders hides distinct states |
| **rRNA** | One rRNA position in assembled or assembling ribosomes | Often reports ribosome biogenesis and specialized subpopulations | Precursors and mature ribosomes are different denominators |
| **Viral RNA** | Genome or transcript class at a defined infection stage | Can couple to replication, translation, or immune recognition | Host contamination and stage shifts alter the pool |
| **Rare cell state** | Molecules within the specified cell state | Can reveal a restricted regulatory population | Bulk averages can erase or simulate the state |

Cell-state dependence enters because cells do not share a single transcriptome or modification state. Immune activation, differentiation, nutrient availability, stress, developmental stage, cell cycle, infection, and tissue microenvironment can alter writers, erasers, readers, substrates, and RNA decay pathways. Single-cell RNA sequencing can identify cell states and trajectories, but standard single-cell RNA-seq usually measures cDNA abundance rather than native RNA modifications. It can therefore explain a denominator problem without directly solving the modification measurement problem.

Mixed samples need extra care. A tumor biopsy may contain malignant cells, immune cells, stromal cells, endothelial cells, dying cells, and extracellular RNA. A bulk modification increase could come from the tumor cells, from infiltrating immune cells, from necrotic RNA damage, or from changes in the abundance of highly modified stable RNAs. Organoid and tissue models can capture heterogeneity better than uniform cell lines, but they also complicate interpretation because cell states, intercellular exchange, and stress gradients become part of the measurement.

Stoichiometry also changes the expected phenotype. A near-stoichiometric rRNA modification can affect most ribosomes produced under a condition. A low-occupancy mRNA modification may affect only a small RNA subpopulation unless the modified molecules are selectively translated, localized, bound by a potent reader, or routed into decay. A low global abundance can still be biologically important if the mark controls a rate-limiting RNA subset, but that argument must be demonstrated with molecule-level or functional evidence. The opposite mistake is also common: a high global modification abundance does not identify the functional substrate.

The chapter's recommended wording is denominator-aware. Instead of "stress increases pseudouridine," write "stress increases pseudouridine signal in poly(A)-selected RNA, with site-level occupancy unknown" or "stress increases occupancy of a mapped pseudouridine site in this transcript after correction for transcript abundance." Instead of "NSUN2 methylates viral RNA," write "NSUN2-mediated m5C modification of hepatitis B virus RNA was reported to positively regulate HBV replication in the studied infection system" when using the Feng study. That wording preserves organism, enzyme, RNA substrate, mark, and context.

## 52.3. Crosstalk among modifications and RNP context

Modification crosstalk means that one RNA feature changes the installation, recognition, or consequence of another feature. The simplest model is chemical co-occurrence on the same molecule. For example, two nearby modifications can alter local structure or protein binding together. But most biological crosstalk is broader. A mark can influence RNA folding, which changes access for a second enzyme. A writer can be recruited by a processing factor. A reader can recruit decay machinery. A modification can shift the set of RNA-binding proteins assembled on the RNA. A change in transcription elongation can alter both modification timing and splice-site exposure. Therefore crosstalk is a network concept, not merely a list of co-modified nucleotides.

![Figure 52.3. Cross-Mark States in RNP Context](../assets/figures/chapter1048_figure3.png)

**Figure 52.3. Cross-Mark States in RNP Context.** One RNA moves among unmodified, mark-A, mark-B, and double-marked states. Structure, RNA-binding proteins, processing, translation, and decay alter transition rates and consequences. A parallel panel distinguishes same-molecule co-occurrence from bulk co-variation caused by cell-state mixtures.

RNP context is the reason crosstalk matters. An RNA-protein complex, or RNP, is the molecular form in which most cellular RNAs function. Proteins can expose or shield modification sites, bind preferentially to modified RNA, recruit writers, block erasers, remodel RNA structure, or alter decay. Ducoli and colleagues used irCLIP-RNP and Re-CLIP approaches to reveal dynamic protein assemblies on RNA, illustrating that protein occupancy is not static around RNA regulatory features. Szeto and colleagues studied dynamic RNA binding and unfolding by the nonsense-mediated mRNA decay factor UPF2, reinforcing that RNA fate factors can remodel RNA rather than passively reading a fixed sequence. These studies are not complete modification maps, but they provide the RNP logic needed to interpret modification effects.

Co-transcriptional and processing context are major sources of crosstalk. A nascent RNA can be modified while it is still associated with chromatin, polymerase, spliceosome components, cleavage factors, and export adaptors. Xu and colleagues linked chromatin-associated m6A methylation to nascent RNA synthesis. The broader principle is that RNA modification, transcriptional timing, and RNA processing should not be treated as independent layers when they occur on the same nascent transcript.

Multi-feature observations in native human transcriptomes motivate a biological model in which modification state, poly(A) tail length, isoform structure, RNA editing, and RNA fate can coexist on the same molecule. The key distinction is between same-molecule coexistence and population-level co-variation. Even a double-marked molecule establishes coexistence rather than functional cooperation unless one state changes the transition into or consequence of the other.

tRNA modification biology illustrates direct crosstalk especially well. Many tRNAs contain multiple modifications that stabilize structure, tune decoding, and protect the RNA from surveillance. A missing modification can affect the installation or interpretation of another mark. The lesson for mRNA is not that every mRNA has a tRNA-like code; it is that modification effects are embedded in mature RNP architecture.

Viral RNA provides another crosstalk setting. A viral RNA can be a genome, mRNA, replication template, pathogen-associated molecular pattern, and RNP scaffold in the same infection cycle. NSUN2-mediated m5C modification of hepatitis B virus RNA was reported to promote HBV replication. Innate immune recognition of viral RNA depends on RNA structure, 5′ end chemistry, double-stranded RNA features, nucleoside modifications, and host sensors. A modification that helps viral replication may do so by changing RNA stability, translation, immune visibility, RNP assembly, or polymerase access. Distinguishing those mechanisms requires experiments that separate replication, RNA abundance, immune signaling, and modification status.

Cancer studies often use "crosstalk" for interactions among chromatin regulators, RNA modification pathways, microRNAs, and signaling pathways. Yi and colleagues reported that EZH2 crosstalk with RNA methylation promotes prostate cancer progression through modulation of an m6A autoregulation pathway. Such examples are valuable because disease phenotypes often involve chromatin, transcription, RNA processing, translation, and decay together. They also require caution: a pathway interaction in cancer cells does not automatically prove that two modifications physically co-occur on the same RNA molecule.

The practical crosstalk test is to ask four questions. Are the features on the same RNA molecule or merely in the same sample? Is the coupling direct, through enzyme access or reader binding, or indirect, through RNA abundance or cell state? Does perturbing one feature change the other at a defined site after controlling for RNA abundance? Does the combined state change an RNA fate, such as splicing, export, translation, localization, decay, immune sensing, or replication? Without these answers, crosstalk should be phrased as co-variation or pathway association rather than mechanism.

## 52.4. Development, stress, immunity, and disease associations

Developmental regulation of RNA modifications is expected because development changes transcription programs, RNA processing, translation demand, cell-cycle state, metabolic cofactors, and RNP composition. A differentiating cell does not merely turn genes on and off; it changes the entire environment in which RNA molecules mature and function. Some modification changes may instruct developmental decisions, while others may mark the altered RNA composition of a new cell state. Distinguishing instruction from consequence is the central developmental problem.

Plants provide a useful reminder that modification dynamics are not mammal-specific. Reviews of plant pseudouridine emphasize that pseudouridine modification participates in plant RNA biology across nuclear and organellar contexts and may be linked to development and stress responses. Plant cells have chloroplasts and plant-specific developmental programs, so mammalian mRNA assumptions cannot be imported wholesale. [Chapter 104](chapter1099.md) treats plant and non-animal RNA biology in broader context.

Stress changes modification biology in at least three ways. First, stress can change writer, eraser, and reader expression or localization. Second, stress can change RNA substrate abundance through transcription, processing, and decay. Third, stress can remodel RNPs, translation, compartments, and RNA lifetimes. Heat shock, oxidative stress, viral infection, or nutrient shifts can therefore change both modification-state transitions and the fates of the marked molecules.

Innate immunity is one of the clearest biological contexts in which RNA chemistry matters. Host sensors distinguish self and non-self RNA using features such as double-stranded RNA length, 5′ end chemistry, cap status, sequence composition, localization, and nucleoside modifications. Han and Xu review RNA modification roles in the immune system, and Marques and colleagues place antiviral RNA sensing in evolutionary perspective. The key principle is that immune sensing reads combinations of features. A modified nucleoside can reduce or alter sensing in one context, while an improperly capped, double-stranded, or mislocalized RNA can remain immunostimulatory.

Therapeutic and viral RNAs sharpen the distinction between endogenous regulation and engineered chemistry. Synthetic mRNAs can use modified nucleosides to tune innate immune sensing, translation, and stability, whereas viral RNAs can acquire or exploit host-installed marks during infection. An engineered, nearly uniform nucleoside substitution and a low-occupancy endogenous site are therefore different biological states even when they contain the same modified nucleoside. Therapeutic design is treated in [Chapter 156](chapter1139.md), and delivery and vaccine immunobiology in Chapters [157](chapter1140.md) and [160](chapter1143.md).

Cancer associations dominate much of the disease literature. RNA modification writers, erasers, and readers are often differentially expressed in tumors, and some studies link specific modification events to invasion, proliferation, immune escape, or therapy response. Yu and Ueda review RNA modifications in cancer and their detection. Huang and colleagues reported that NAT10-mediated ac4C modification of ITGB5 mRNA promotes perineural invasion in pancreatic ductal adenocarcinoma. Yi and colleagues linked EZH2 and RNA methylation pathways in prostate cancer progression. These studies show the value of substrate-level mechanisms, but the broader field still contains many weaker association claims based only on enzyme expression or global modification abundance.

Fibrosis and cardiovascular disease illustrate another disease frame. Wu and colleagues review RNA modification advances in myocardial fibrosis. Fibrosis involves cell-state changes, extracellular matrix remodeling, inflammation, injury responses, and tissue composition shifts. A modification change in a fibrotic sample may occur in cardiomyocytes, fibroblasts, immune cells, endothelial cells, or damaged RNA pools. Causal interpretation should therefore connect the modified RNA event to a cell type, molecular pathway, and fibrotic phenotype rather than stopping at global modification differences.

Neurological and immune diseases raise similar denominator problems. Microglia can adopt complex states across development, injury, aging, and neurodegeneration. If a bulk brain sample shows altered modification signal, cell-state annotation is needed before attributing the signal to a specific disease mechanism. The same logic applies to immune tissues, tumors, and infected organs. Cell-state methods are not substitutes for modification measurements, but they prevent misassignment of bulk signals to the wrong cells.

Viral infection can convert modification biology into replication biology. Feng and colleagues reported that NSUN2-mediated m5C modification of HBV RNA positively regulates HBV replication. This is a clear example of why the RNA substrate matters: the substrate is viral RNA in an infection context, and the outcome is viral replication. A host enzyme can influence viral RNA fate without implying that the same modification has the same function on host mRNAs. The boundary between host regulation and pathogen exploitation must be stated explicitly.

The clinical translation rule is conservative. A disease-associated modification pathway becomes clinically meaningful only after several questions are answered. Is the mark reproducibly measured in clinically relevant material? Is the change independent of cell composition, RNA abundance, and batch effects? Does the modification event cause a disease-relevant phenotype in suitable models? Can the pathway be targeted without unacceptable effects on essential tRNA, rRNA, or mRNA processes? Does the proposed biomarker outperform existing markers? Until then, "associated with disease" should not be rewritten as "driver" or "therapeutic target."

> **Box 52.2. Reading Disease Associations at the Correct Level**
>
> Separate pathway expression, modification state, substrate-level mechanism, clinical association, biomarker performance, and therapeutic opportunity. Each is a distinct endpoint, and success at one level does not automatically establish the next.

## 52.5. Biological controversies and deprecated cross-mark claims

**Table 52.3. Deprecated and Preferred Cross-Mark Claims.** Preferred cross-mark statements preserve RNA class, site, stoichiometry, enzyme, and cellular context; broad claims of epitranscriptomic crosstalk are weakened when those distinctions or causal tests are missing.

| Deprecated claim | Preferred biological statement | Missing distinction |
| --- | --- | --- |
| **Dynamic means reversible** | The marked population changes; the responsible biological route remains to be identified | Direct removal versus installation, decay, dilution, or composition |
| **Every mark has a writer, reader, and eraser** | Regulatory roles differ among chemistries and RNA classes | Demonstrated role versus imported vocabulary |
| **Two marks co-vary, so they crosstalk** | The marks co-vary in this population; direct coupling is unproven | Same-molecule state and causal route |
| **Writer expression measures modification** | Writer abundance is one determinant of a substrate-specific state | Enzyme activity, localization, substrate access, and RNA denominator |
| **Disease association defines a target** | The pathway is associated with disease; causality and tractability are separate questions | Substrate mechanism, selectivity, pharmacology, and safety |

The central biological controversy is whether useful principles can be generalized across marks without erasing chemistry, RNA class, and RNP context. The writer-reader-eraser vocabulary is helpful for some systems, but it can imply a shared regulatory grammar that does not exist universally. Some marks are installed nearly quantitatively during stable-RNA maturation; others occupy a minority of mRNA molecules; some are chemically removable; others turn over mainly when the RNA turns over. Cross-mark synthesis should compare rate structure and biological consequence without forcing every chemistry into one model.

One deprecated habit is treating "dynamic" as synonymous with "actively reversible." Many modifications are dynamic at the population level because RNA molecules are synthesized, processed, localized, translated, and degraded. Some m6A-related marks have direct demethylase pathways, but many modifications have no demonstrated eraser under physiological conditions. For m5C and other marks, reviews emphasize emerging dynamic roles while also noting unresolved questions about direct removal and detection. A careful chapter should say "modification signal changes" unless removal from the same RNA molecule has been shown.

Another deprecated habit is treating writer or reader expression as a proxy for modification state. A writer can be abundant but inactive, mislocalized, lacking cofactors, blocked by RNA structure, or acting on different substrates. A reader can be expressed but unable to bind because the relevant sites are absent or occupied by other proteins. Conversely, a small change in writer activity can strongly affect a small set of high-value substrates. Expression data are useful hypotheses, not direct modification measurements.

Another deprecated claim is that two marks that change together must interact directly. Bulk co-variation can be produced by a common writer program, altered RNA abundance, a shared maturation stage, or a shift in cell composition. Direct crosstalk requires a defined route: one mark changes installation of another, two marks jointly alter structure or binding, or both marks coexist on the same molecules that enter a distinct fate. Single-molecule co-occurrence is especially valuable, but co-occurrence itself still does not prove functional cooperation.

The mRNA-centric idea of a dynamic epitranscriptomic switch should not be projected onto every RNA class. Many tRNA and rRNA marks are maturation checkpoints with high occupancy and strong consequences for RNP biogenesis. A small decrease can reflect accumulation of immature particles rather than a regulated switch on mature RNA. Conversely, a low-occupancy mRNA mark may identify a transient, localized, or specialized subpopulation. The same fraction has different meaning when the denominator is a stable ribosome, a tRNA isodecoder, a short-lived mRNA isoform, or a viral genome.

Universal reader claims are also weakened when reader proteins have modification-independent RNA binding, multiple domains, or context-dependent partners. A reader may prefer a modified motif in vitro yet be excluded by RNA structure or another RBP in cells. A mark may act by changing structure without a dedicated reader. The biological unit is therefore the modified RNP transition, not a free mark-reader pair abstracted from time and compartment.

Disease-wide phrases such as "m6A drives cancer" or "epitranscriptomic dysregulation causes inflammation" collapse heterogeneous substrates and opposing effects. The same enzyme can modify essential stable RNAs, selected mRNAs, viral RNAs, or noncoding RNAs, and its consequence can reverse across cell types or stages. A defensible synthesis names the substrate, site or region, stoichiometry, RNP effect, disease-relevant cell, and temporal position in pathogenesis.

GlycoRNA illustrates a frontier cross-mark claim whose layers now have explicit owners. The appropriate synthesis distinguishes chemical identity and linkage, biosynthetic dependence, surface topology, binding partners, and physiological function rather than treating them as one settled mechanism. [Chapter 49](chapter1045.md) owns the chemical and biosynthetic evidence; [Chapter 107](chapter1102.md) owns cell-surface presentation and extracellular RNA-protein assemblies; general evidence adjudication belongs to [Chapter 46](chapter1043.md) and protocol-level validation to [Chapter 132](chapter1120.md). This chapter retains only the cross-mark lesson that support for one layer does not automatically establish the others.

> **Box 52.3. Cross-Mark Frontier: From Co-Variation to Biological Coupling**
>
> Use glycoRNA and multi-feature native-RNA observations to distinguish chemical identity, same-molecule coexistence, topology, RNP partners, causal ordering, and organismal function. The box ends with biological models that would discriminate cooperation, antagonism, and shared upstream response.

The cross-mark synthesis rule is therefore biological rather than methodological: compare marks only after defining RNA class, eligible site population, occupancy, life-cycle stage, RNP state, cell state, and consequence. Method-specific detection limits and controls remain essential, but they are taught in [Chapter 46](chapter1043.md) and [Chapter 132](chapter1120.md) rather than repeated here.

## 52.6. Cross-mark quantitative models and open questions

Quantitative modeling starts with a minimal kinetic picture. For a single site on a defined RNA population, unmodified molecules can be produced, modified, demodified if an eraser exists, and degraded. Modified molecules can also be degraded. At steady state, occupancy depends on installation rate, removal rate, production rate, and decay rates of both modified and unmodified molecules. The important lesson is not the exact formula; it is that modification fraction is not determined by writer activity alone. RNA synthesis and decay are part of the modification model.

A simple model can be written in words. Newly made RNA enters an unmodified pool. A writer moves some molecules into a modified pool. An eraser, if present, moves modified molecules back to the unmodified pool. Decay removes molecules from both pools, possibly at different rates. Translation, localization, or RNP assembly can be modeled as additional states. A stress response can change any rate. If the modified pool decays faster than the unmodified pool, stronger installation may be needed to maintain the same occupancy. If the unmodified pool decays faster, occupancy can rise without stronger installation.

Cross-mark models require more than one modified pool. For two candidate marks, molecules can occupy unmodified, A-only, B-only, or A-plus-B states, with transitions determined by writer access, order of installation, removal if it exists, RNP remodeling, and state-specific decay. If mark A exposes the site for mark B, the transition into the double-marked state should depend on A. If both marks merely respond to the same cell state, their rates covary without a direct transition. This distinction formalizes direct crosstalk versus shared regulation.

Molecule-level and cell-state models answer different questions. A double-marked fraction asks whether marks coexist on individual molecules. A cell-state mixture model asks whether one subpopulation is enriched for mark A and another for mark B. Bulk averages can make these cases look identical. The biological model should therefore include isoform, compartment, maturation state, and cell identity before assigning cooperation or antagonism between marks. Measurement error must also be represented, but assay-specific calibration and benchmarking are treated in [Chapter 132](chapter1120.md).

RNP dynamics require models beyond site occupancy. Liebau and colleagues' work on quantitative dynamics in the eukaryotic RNA exosome complex illustrates how 4D structural biology can treat RNP machines as dynamic ensembles rather than fixed structures. For modification biology, this matters because the effect of a mark often depends on whether a reader, helicase, decay factor, spliceosome component, or ribosome encounters the modified RNA at the right time. A static site map cannot explain a time-dependent RNP process without kinetic and structural context.

A useful quantitative framework for disease studies separates four layers. The first layer is cell composition: which cell types and states are present? The second layer is RNA abundance and isoform usage: which substrates exist? The third layer is site occupancy and modification co-occurrence: which molecules are chemically marked? The fourth layer is functional response: what changes in splicing, translation, decay, immune sensing, replication, invasion, or fibrosis? Many disease studies measure only one or two layers and infer the rest. Future studies should combine them directly.

Open questions remain substantial. How many low-occupancy sites control specialized subpopulations? Which apparent dynamics reflect active enzymatic control rather than RNA turnover or cell composition? Which modifications have physiological erasers? How often do multiple marks co-occur on the same RNA molecule, and when does co-occurrence alter RNP fate? Which cross-mark relationships are conserved across RNA classes or organisms? Which disease associations survive substrate- and cell-specific causal tests?

Plant RNA biology highlights another open frontier. Plant modification dynamics raise specific problems in organellar RNA processing, stress adaptation, development, environmental response, and lineage-specific enzymes. The broader lesson is that one organism's modification grammar should not become the universal template for all life.

The strongest future studies will be denominator-aware, time-resolved, perturbation-resolved, and cross-mark-aware. They will measure RNA abundance and modification fraction together, define cell states, distinguish co-occurrence from co-variation, model synthesis and decay, and connect enzyme, substrate, site, stoichiometry, RNP state, and biological consequence. Method selection and benchmarking follow [Chapter 132](chapter1120.md); this chapter's endpoint is a biological model that predicts how the marked RNA population changes state.

## Technology, Computational, and Clinical Links

Technology enables cross-mark biology when it can connect chemical state to isoform, molecule, compartment, cell state, and RNP fate. Assay mechanisms, workflows, spike-ins, caller calibration, and benchmark design belong to [Chapter 132](chapter1120.md). The biological requirement here is that the selected measurements distinguish occupancy from abundance and same-molecule co-occurrence from population-level co-variation.

Computational models should integrate RNA synthesis, isoform abundance, modification-state transitions, RNP binding, decay, cell-state composition, and functional output. A model trained on one mark, RNA class, or organism cannot be assumed to generalize to another because rate structure and denominators differ. [Chapter 146](chapter1149.md) extends these ideas into systems RNA biology.

Clinically, RNA modification pathways are attractive because enzymes and reader proteins can be druggable or biomarker-like. They are also risky because many modification enzymes act on essential RNAs. Inhibiting a writer that modifies a tumor-promoting mRNA may also perturb tRNA, rRNA, immune-cell function, or normal tissue renewal. Clinical translation therefore requires selectivity, context, dosing, pharmacology, and safety evidence. Disease reviews are useful for mapping associations, but the chapter's standard for intervention is mechanistic and clinical validation rather than enthusiasm.

## Recent Consensus

The current consensus is that RNA modifications are chemically real, biologically important, and often context-dependent, but that many transcriptome-wide claims require sharper quantification and validation. The strongest claims specify chemical identity, RNA class, site or region, stoichiometry, context, method, and functional consequence. The weakest claims use global modification levels, enzyme expression, or enrichment peaks as if they directly identified functional sites.

There is broad agreement that mRNA modification biology must be interpreted alongside RNA abundance, isoform choice, processing, translation, decay, and RNP assembly. There is also broad agreement that immune and disease contexts are important but confounded by cell-state changes. The field is moving from maps to models: from "where is the mark?" to "what fraction of which molecules carry the mark, when, in which RNP state, and with what consequence?"

Consensus is more limited for some emerging or controversial areas. Direct erasure is not established for every dynamic signal. Single-molecule co-occurrence is not comprehensively mapped. Many disease associations remain correlative. GlycoRNA biology and other frontier claims require careful chemical and spatial validation. These caveats should be treated as a path to stronger science, not as a reason to ignore RNA modification biology.

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

Open questions:

- Which low-occupancy marks define specialized RNA or cell subpopulations with distinct fates?
- When do two marks coexist on the same molecule and directly cooperate, compete, or occur independently?
- How do modification-state transitions couple to RNP assembly during development, stress, infection, and recovery?
- Which writer, eraser, and reader functions are mark-dependent versus modification-independent protein functions?
- Which cross-mark relationships are conserved across mRNA, stable RNAs, organellar RNAs, and viral RNAs?
- When does a disease-associated change initiate pathology, amplify it, or merely report a changed cell-state mixture?

Common misconceptions:

- "Every dynamic mark is reversible." Dynamics can come from installation, removal, RNA turnover, altered substrate abundance, altered cell composition, or altered detection.
- "Every detected modification is functional." Function requires a measurable molecular or biological consequence, and causality should be tested with substrate, site, enzyme, reader, or rescue experiments.
- "Low stoichiometry means no biological relevance." Low occupancy can matter if the modified subpopulation has a specialized fate, but that fate must be demonstrated.
- "Writer expression measures modification state." Writer abundance is only one variable among enzyme activity, localization, cofactors, substrate access, RNA abundance, and competing RNPs.
- "Disease association implies therapeutic opportunity." Therapeutic opportunity requires causality, selectivity, delivery, pharmacology, safety, and patient-context validation.