# Chapter 159. RNA Synthesis, In Vitro Transcription, Purification, Formulation, Stability, and Analytical Release Testing

## Scope Note

This chapter explains how research, diagnostic, and therapeutic RNA molecules are manufactured and released for use, from chemical oligonucleotide synthesis and enzymatic in vitro transcription through finishing reactions, impurity control, purification, formulation, stability management, scale-up, good manufacturing practice, and analytical release testing. The chapter emphasizes mechanistic sources of product heterogeneity because RNA quality is not a single property: sequence identity, length, end structure, modification pattern, residual template, residual enzyme, double-stranded RNA, nuclease burden, formulation state, and storage history can each change biological activity and safety.

## Executive Summary

RNA manufacturing is the controlled production of a chemically fragile, highly charged, sequence-defined polymer whose biological behavior depends on both intended design features and unintended process-derived species. Short RNAs are commonly made by solid-phase oligonucleotide synthesis, in which protected nucleotide building blocks are coupled stepwise on a support and then deprotected, cleaved, purified, and characterized. Longer RNAs, especially messenger RNA and self-amplifying RNA, are usually made by in vitro transcription, in which a DNA template bearing a phage promoter directs RNA polymerase to synthesize many RNA copies in solution. These two production modes share the need to verify sequence, length, purity, and absence of harmful contaminants, but their major impurity mechanisms differ.

In solid-phase synthesis, incomplete coupling, depurination, deprotection side reactions, protecting-group residues, truncated products, deletion products, and failure sequences dominate process design. In in vitro transcription, template design, promoter sequence, nucleotide concentrations, magnesium, pyrophosphate, enzyme loading, reaction time, temperature, and modified nucleotide choice influence yield, full-length fraction, abortive products, run-off heterogeneity, misincorporation, residual DNA, residual enzyme, and double-stranded RNA. Double-stranded RNA is especially important for therapeutic mRNA because it can activate innate immune receptors and reduce translation; its abundance is controlled by template design, reaction conditions, modified nucleotides, and purification.

Product finishing reactions define many RNA modalities. Synthetic mRNA usually needs a 5′ cap and a poly(A) tail; circular RNA requires ligation or autocatalytic splicing; guide RNAs may need defined chemical ends; self-amplifying RNA must preserve a long replicon architecture. Capping can be performed during transcription with cap analogs or after transcription with capping enzymes, and each approach has different failure modes. Poly(A) sequences can be encoded in the DNA template or added enzymatically; encoded tails provide greater definition, whereas enzymatic tailing can produce length distributions.

Purification is both a concentration step and an impurity-specification step. Chromatographic and filtration methods separate product from short transcripts, template DNA, enzymes, nucleotides, salts, endotoxin, organic residues, host-cell impurities from enzyme production, and double-stranded RNA. Analytical release testing is the evidence system that links the manufacturing process to product claims. A release panel for a therapeutic RNA product typically includes identity, concentration, integrity, purity, residual DNA, residual protein or enzyme, residual solvent or reagent where relevant, endotoxin, bioburden or sterility, pH, osmolality, particle attributes for formulated products, encapsulation efficiency for lipid nanoparticles, potency, and stability-indicating assays.

Formulation converts purified RNA into a usable product. Naked RNA is vulnerable to hydrolysis and nucleases and is usually unsuitable for systemic delivery, while lipid nanoparticles protect RNA and mediate cellular uptake but introduce additional quality attributes such as particle size, polydispersity, lipid identity, lipid impurities, encapsulation, pH sensitivity, aggregation, and freeze-thaw behavior. Stability is a property of the RNA sequence, chemical modifications, buffer, excipients, container, headspace, temperature, light exposure, freeze-thaw history, shear, and formulation. Cold-chain storage, lyophilization, and container-closure choices are therefore part of the manufacturing design, not downstream packaging details.

## Concept Inventory

- **In vitro transcription:** enzymatic RNA synthesis outside cells from a defined DNA template. The term usually refers to phage single-subunit RNA polymerases such as T7, SP6, or T3 RNA polymerase, although other polymerases can be used for specialized applications. A promoter is the DNA sequence that recruits and positions the polymerase; in therapeutic and research manufacturing, promoter design is not merely a cloning detail because promoter-proximal bases influence initiation, abortive products, and yield.
- **Solid-phase oligonucleotide synthesis:** chemical stepwise assembly of an oligonucleotide attached to an insoluble support. For RNA, the reactive 2′ hydroxyl must be protected during synthesis and later deprotected without damaging the product. A failure sequence is a molecule that lacks one or more intended residues because a coupling or deprotection step did not proceed to completion. The longer the oligonucleotide, the more cumulative yield depends on small per-cycle losses.
- **Cap:** a specialized 5′ end structure that supports translation, stability, and innate immune discrimination in many eukaryotic and therapeutic mRNAs. Cap0 contains an inverted 7-methylguanosine connected by a 5′ to 5′ triphosphate bridge. Cap1 additionally contains 2′-O-methylation on the first transcribed nucleotide. Poly(A) tail refers to an adenosine-rich 3′ tract that can support translation and stability of mRNA; an encoded poly(A) tail is copied from a DNA template, whereas enzymatic tailing is added after transcription by a poly(A) polymerase.
- **Double-stranded RNA:** RNA in which complementary regions form duplexes. In an in vitro transcription product, dsRNA can include full-length antisense duplexes, shorter complementary fragments, self-complementary structures, or RNA generated by polymerase-dependent side reactions. The category is functional as well as structural because dsRNA-like impurities can activate cellular sensors such as TLR3, MDA5, RIG-I-like pathways, PKR, and OAS/RNase L, depending on length, end chemistry, sequence context, formulation, and cell type.
- **Good manufacturing practice:** the regulated quality system for making clinical and commercial products. GMP is not a purification method. It is a system of controlled materials, validated or qualified processes, documented deviations, trained personnel, calibrated equipment, defined specifications, stability programs, and batch records that allow a product lot to be released or rejected based on pre-specified evidence.

## What to Know Before Reading This Chapter

The reader should know that RNA is a polymer of ribonucleotides connected by phosphodiester bonds, that RNA has directional ends conventionally described as 5′ and 3′, and that the ribose 2′ hydroxyl makes RNA more chemically labile than DNA under many conditions. The reader should also know that therapeutic RNA can act by several modalities: a small RNA can guide RNA interference or gene editing, an antisense oligonucleotide can bind a target RNA, and an mRNA can encode a protein. Manufacturing details differ because each modality asks the molecule to survive different stresses and perform a different biological task.

The central manufacturing distinction is chemical versus enzymatic synthesis. Chemical synthesis builds short chains residue by residue and is powerful for oligonucleotides that need nonnatural sugars, backbones, or conjugates. Enzymatic transcription copies a DNA template and is efficient for long RNAs, but it introduces polymerase-specific heterogeneity and template-derived impurities. A 21-nucleotide siRNA, a 100-nucleotide guide RNA, a 2,000-nucleotide mRNA vaccine antigen, and a 9,000-nucleotide self-amplifying RNA replicon are all "RNA products," but no single process or release assay adequately describes all four.

Three running examples will recur. First, a chemically synthesized guide RNA illustrates short-RNA production, where sequence definition and modified residues matter more than bulk yield. Second, a therapeutic mRNA encoding a secreted protein illustrates in vitro transcription, capping, purification, lipid nanoparticle formulation, and potency testing. Third, a circular RNA candidate illustrates why end structure, ligation, residual linear precursor, and innate immune impurities must be considered together. These examples are not protocols; they are manufacturing cases that expose the causal logic of quality attributes.

## 159.1. Solid-Phase Oligonucleotide Synthesis

Solid-phase oligonucleotide synthesis is the dominant production method for many short therapeutic and experimental RNAs, including small interfering RNA strands, antisense-like RNA segments, chemically modified guide RNA components, aptamers, primer-like RNA standards, and hybrid RNA-DNA oligonucleotides. The word "solid-phase" means that the growing chain is tethered to an insoluble support, usually through its future 3′ end, while soluble reagents flow over the support. After each reaction cycle, excess reagents and byproducts can be washed away without losing the immobilized product. This simple physical separation is what makes repetitive chemistry feasible.

The standard conceptual cycle contains four steps. First, a protecting group is removed from the terminal hydroxyl of the support-bound chain to expose a reactive site. Second, an activated nucleotide phosphoramidite couples to the growing chain. Third, unreacted chains are capped so they cannot continue as deletion products in later cycles. Fourth, the new internucleotide linkage is oxidized or sulfurized to the intended phosphate or phosphorothioate state. The chain then enters the next cycle. Because RNA contains a reactive 2′ hydroxyl, RNA phosphoramidites carry additional protecting groups compared with DNA phosphoramidites. These protecting groups must be stable during chain assembly but removable after synthesis under conditions that do not degrade the RNA.

![Figure 159.1. Solid-phase RNA synthesis cycle and cumulative yield logic](../assets/figures/chapter1142_figure1.png)

**Figure 159.1. Solid-phase RNA synthesis cycle and cumulative yield logic.** Solid-phase RNA synthesis builds a protected chain one residue at a time. The same cycle that works efficiently for short RNAs creates cumulative losses and sequence-related impurities as oligonucleotide length increases.

The main mechanistic lesson is that a high-yielding individual cycle can still produce a poor full-length yield when repeated many times. If each coupling cycle is 99 percent efficient, a 20-mer can still contain a substantial minority of truncated or capped molecules, and a 100-mer becomes much more difficult. Therefore, chemical synthesis is best suited to short RNAs and heavily modified oligonucleotides, not to conventional kilobase-scale mRNAs. The boundary is not absolute; improvements in chemistry, supports, and purification can push lengths upward, and ligation of shorter fragments can produce longer RNAs. Still, each added nucleotide increases the burden on coupling efficiency, side-reaction control, and purification resolution.

**Table 159.1. Chemical synthesis impurity classes and control levers.** Solid-phase RNA impurity classes arise from specific chemical steps and require assays that can distinguish sequence, mass, charge, and modification state.

| Impurity class | Likely source in the synthesis cycle | Useful detection or readout | Control lever or caveat |
| --- | --- | --- | --- |
| **N-minus and other failure sequences** | Incomplete coupling, incomplete capping, or loss of full-length material across repeated cycles | Ion-exchange or reverse-phase chromatography, capillary electrophoresis, mass spectrometry | Per-cycle efficiency compounds with length; high single-cycle yield can still leave a substantial truncated population. |
| **N-plus, deletion, and branch-like side products** | Carry-through after failed capping, side reactions during coupling, or unintended elongation of defective chains | Orthogonal chromatography plus mass analysis or mapping assays | A single main peak may conceal near-isobaric or co-eluting sequence-related variants. |
| **Depurination, deamination, and base-damaged products** | Acidic or otherwise harsh synthesis, cleavage, deprotection, or hold conditions | Mass spectrometry, nucleoside analysis, chromatographic shoulders | Gentler conditions reduce damage but must still remove protecting groups completely. |
| **Incompletely deprotected or residual protecting-group species** | Incomplete post-synthesis deprotection or cleavage from support | Mass shifts, reverse-phase behavior, ion-exchange behavior | More aggressive deprotection can improve removal while increasing RNA damage risk. |
| **Phosphorothioate linkage heterogeneity** | Sulfurization at selected internucleotide linkages, especially when stereochemistry is not controlled | Specialized chromatography, mass-based characterization, functional comparability assays | A stereochemical mixture may be part of the intended product, but it must be specified and controlled. |
| **Salts, counterions, and small-molecule residues** | Activators, oxidizers or sulfurizers, cleavage reagents, protecting-group byproducts, desalting carryover | Conductivity, residual reagent assays, residual solvent assays, chromatography | These are not sequence impurities, but they can affect annealing, formulation, stability, and biological assays. |
| **Duplex assembly impurities for siRNA products** | Strand ratio imbalance, incomplete annealing, mismatched duplexes, residual single strands | Native electrophoresis, chromatography, melting behavior, strand-specific assays | Each strand may pass synthesis tests before the duplex fails its own process controls. |

For a chemically synthesized guide RNA segment, the product specification may include exact sequence, identity and placement of modified nucleotides, stereochemical status if phosphorothioate linkages are used, residual protecting groups, water content, salt form, counterions, endotoxin, bioburden, and purity by chromatography or electrophoresis. The relevant impurity profile includes n-minus sequences, n-plus sequences, depurination-derived fragments, deamination or base-modified products, incompletely deprotected species, branch-like side products, and residual small molecules from synthesis and cleavage. A single mass peak can conceal isobaric or near-isobaric variants, so orthogonal assays are often required when clinical quality or mechanistic interpretation depends on a specific modification pattern.

Purification after solid-phase synthesis commonly starts with crude deprotected material containing the full-length product plus truncated sequences and small-molecule contaminants. Desalting removes salts and low-molecular-weight reagents but does not reliably remove sequence-related impurities. Ion-exchange chromatography separates molecules by charge and length, reverse-phase chromatography separates by hydrophobicity and protecting-group or tag status, and hydrophilic interaction or size-based methods may be used for specialized cases. For duplex siRNA, each strand is usually synthesized, purified, and characterized before annealing; duplex formation itself becomes a process step with its own controls for strand ratio, residual single strand, mismatched duplexes, and aggregate-like species.

Evidence for product quality in solid-phase synthesis is mostly analytical rather than biological. A chromatogram, mass spectrum, capillary electrophoresis trace, or nuclease mapping assay does not prove that an oligonucleotide will work in cells, but it can establish that the material being tested is close to the intended chemical entity. This distinction is important. If a guide RNA editing experiment fails, poor biological delivery, target chromatin state, nuclease degradation, sequence design, or protein compatibility may be responsible; but if the guide RNA itself contains substantial deletion products or wrong modifications, downstream biological conclusions become uninterpretable.

Boundary cases matter. Some therapeutic oligonucleotides are not ordinary RNA: they may contain 2′-O-methyl, 2′-fluoro, locked nucleic acid, phosphorothioate, morpholino, peptide nucleic acid, or other noncanonical chemistry, many of which are treated in more detail in [Chapter 149](chapter1133.md) and [Chapter 150](chapter1134.md). This chapter focuses on RNA synthesis as it intersects with RNA manufacturing. The same plant may synthesize DNA, RNA, and modified oligonucleotide drugs, but the quality risks differ because ribose chemistry, nuclease susceptibility, immune recognition, and formulation behavior differ.

Add review and primary anchors for RNA protecting groups, coupling chemistry, impurity mechanisms, purification choices, and oligonucleotide release specifications.

## 159.2. In Vitro Transcription Enzymes, Templates, Promoters, Yields, Abortive Products, and Impurities

In vitro transcription, abbreviated IVT, is the main route for producing long synthetic RNA. In a typical IVT reaction, a linear DNA template contains a promoter recognized by an RNA polymerase, followed by the sequence to be transcribed. The polymerase binds the promoter, initiates RNA synthesis at a defined start site, elongates by adding ribonucleoside triphosphates, and runs off the end of the template. The product is a population of RNA molecules that share the intended sequence but may differ in length, ends, structure, and chemical composition.

The most common IVT enzymes for manufacturing are bacteriophage-derived single-subunit RNA polymerases, especially T7 RNA polymerase, with SP6 and T3 RNA polymerases used in some contexts. These enzymes are compact, processive, promoter-specific, and capable of high RNA output in simple biochemical mixtures. Their usefulness comes from separation from cellular regulation: the reaction can be driven with purified enzyme, nucleotide triphosphates, magnesium, buffer, template DNA, and additives. Their limitations also come from this simplicity. A purified polymerase does not perform all the capping, processing, surveillance, folding assistance, and quality control that cellular transcription systems use.

Template design starts with the promoter. T7 RNA polymerase, for example, has strong sequence preferences near the transcription start site, and many high-yield templates begin with guanosine-rich starts. This creates a design tension for mRNA therapeutics and guide RNAs because the optimal promoter-proximal bases for polymerase initiation may not be the desired first bases of the product. Extra 5′ bases can affect translation, innate immune recognition, folding, guide activity, or downstream processing. Some workflows accept a short designed leader; others use ribozymes, promoter variants, or post-transcriptional processing to generate precise ends. The promoter is therefore a manufacturing control element rather than an invisible upstream sequence.

![Figure 159.2. IVT reaction architecture and major impurity origins](../assets/figures/chapter1142_figure2.png)

**Figure 159.2. IVT reaction architecture and major impurity origins.** In vitro transcription quality is shaped by template architecture, polymerase initiation, elongation, run-off termination, reaction chemistry, and post-reaction handling.

The DNA template may be a linearized plasmid, polymerase chain reaction product, synthetic DNA fragment, or other defined DNA substrate. Template purity affects RNA purity. Residual nicked plasmid, incomplete linearization, heterogeneous template ends, incorrect sequence, residual primers, and contaminating nucleases can create run-on products, shorter products, wrong products, or degraded products. For mRNA, an encoded poly(A) tail is often part of the template. Long homopolymeric poly(A) tracts can be unstable in bacterial plasmids or difficult to sequence completely, so template verification may require methods that read through repetitive regions or infer tail length by orthogonal assays.

The chemistry of IVT is nucleotide addition. The polymerase uses ribonucleoside triphosphates as substrates and releases pyrophosphate. Magnesium supports catalysis but also influences RNA folding, pyrophosphate precipitation, and hydrolysis risk. Nucleotide concentrations influence yield, incorporation balance, misincorporation, and final reaction composition. High concentrations can drive output but may also favor side reactions or complicate downstream purification. Reaction time and temperature are not neutral variables: longer incubations can increase yield while also increasing degradation, template-independent extension, dsRNA formation, or accumulation of byproducts. Process development therefore seeks a window, not simply the maximum transcript mass.

Abortive initiation is a defining impurity mechanism for phage polymerase IVT. During early transcription, the polymerase can repeatedly synthesize and release short RNA products before escaping the promoter into stable elongation. These abortive products are usually short, but they matter because they consume nucleotides, complicate purification, and can contribute to small-RNA impurities. Promoter escape is influenced by promoter sequence, initiating nucleotides, temperature, enzyme properties, template context, and reaction composition. Abortive products are conceptually similar to promoter-proximal abortive transcripts in bacterial transcription, but in manufacturing they are treated as process-related impurities.

Run-off transcription creates another end-definition problem. A perfect run-off transcript ends at the last templated base, but real reactions can contain non-templated additions, prematurely terminated transcripts, read-through products if the template is not properly terminated, and products arising from damaged or truncated templates. For long RNAs, even a small frequency of premature termination can produce a complex smear of shorter products. The distinction between a short abortive product and a long prematurely terminated product matters analytically because the former may be removed by size-based methods, while the latter may co-purify with the target and affect translation or immunogenicity.

**Table 159.2. IVT product-related and process-related impurities.** IVT impurity control requires separating RNA variants from template, enzyme, nucleotide, salt, endotoxin, and degradation-related impurities.

| Impurity class | Product-related or process-related origin | Useful analytical approaches | Release relevance or caveat |
| --- | --- | --- | --- |
| **Abortive transcripts** | Product-related short RNAs from repeated initiation before promoter escape | Small-RNA chromatography, capillary electrophoresis, size-sensitive assays | Short products consume substrates and complicate purification even when full-length RNA is abundant. |
| **Truncated and prematurely terminated RNA** | Product-related early termination, pausing, RNase cleavage, or chemical fragmentation | Electrophoresis, capillary electrophoresis, chromatography, fragment analysis | Total RNA mass can overstate the amount of full-length functional product. |
| **Extended RNA, wrong ends, and tail heterogeneity** | Product-related run-on transcription, heterogeneous template ends, or finishing variation | End mapping, tail-length assays, sequencing-compatible methods, chromatography | End structure can affect translation, immune recognition, circularization, and comparability. |
| **dsRNA, antisense RNA, and copy-back products** | Product-related duplex-like species from complementary transcripts, self-complementary regions, or polymerase side reactions | Antibody assays, nuclease assays, chromatography, cell-based immune or translation assays | dsRNA-like species are safety- and potency-relevant, but no single assay defines the full burden. |
| **Misincorporated or modification-biased products** | Product-related polymerase errors, nucleotide imbalance, or altered incorporation of modified substrates | Sequencing where feasible, nucleoside composition analysis, mass spectrometry | Modified nucleotide use can change both biology and analytical behavior. |
| **Residual DNA template** | Process-related carryover after IVT and DNase treatment | qPCR, ddPCR, DNA-sensitive fluorescence, sequencing-informed assays | Primer choice matters; a low qPCR result can miss fragments outside the assay target. |
| **Residual enzymes and proteins** | Process-related polymerase, DNase, pyrophosphatase, RNase inhibitor, or supplier-derived protein | Protein assays, immunoassays, enzyme activity assays | Enzyme cleanup can solve one impurity problem while adding another process impurity. |
| **Nucleotides, salts, magnesium, pyrophosphate, and small molecules** | Process-related reaction components and byproducts | Conductivity, chromatography, elemental or small-molecule assays | Residual small molecules can affect stability, downstream purification, and formulation. |
| **Endotoxin and bioburden** | Process-related raw material, water, equipment, or handling contamination | Endotoxin assays, bioburden tests, sterility testing where required | Low endotoxin does not rule out immune activation from RNA ends, dsRNA, formulation, or cell context. |

Impurities in IVT reactions fall into several mechanistic classes. Product-related RNA impurities include short abortive transcripts, truncated transcripts, extended transcripts, antisense or complementary RNA, dsRNA, misincorporated products, uncapped or improperly capped mRNA, and molecules with wrong tail length or end chemistry. Process-related impurities include template DNA, polymerase, pyrophosphatase if used, ribonuclease inhibitors, unincorporated nucleotides, salts, magnesium, pyrophosphate, organic reagents, endotoxin, and bioburden. Degradation products include RNase-generated fragments and chemically hydrolyzed RNA. Analytical release testing must decide which impurities are safety critical, which are potency critical, which are process consistency indicators, and which are controlled by upstream raw material specifications.

Yields are reported in ways that can mislead. Total RNA mass does not equal full-length functional RNA. A high-yield reaction with abundant truncated transcripts or dsRNA may be worse than a lower-yield reaction with clean product. For mRNA, the meaningful denominator may be full-length capped polyadenylated RNA capable of expressing protein after formulation. For circular RNA, the meaningful denominator may be correctly circularized RNA free of linear precursor and nicked products. For a guide RNA, the meaningful denominator may be full-length chemically and sequence-correct guide. Manufacturing development should therefore track yield and quality together.

> **Box 159.1. Full-Length Yield Is Not Product Quality**
>
> A reported RNA yield should always be read with a definition of the product in the numerator. Ask whether the value was measured before or after DNase treatment, purification, capping, tailing, circularization, buffer exchange, and formulation. Ask whether the assay counts abortive transcripts, truncated products, uncapped molecules, wrong-tail species, dsRNA, degraded fragments, or unencapsulated RNA. A high absorbance value can reflect abundant RNA while overstating the amount of full-length, correctly finished, low-impurity material. For mRNA, a process that lowers total RNA mass but increases capped, polyadenylated, full-length RNA may be better. For circular RNA, the useful yield excludes linear precursor, nicked circles, and concatemers. For guide RNAs or siRNAs, full-length sequence and correct modification placement matter more than bulk mass. Report yield as a chain: crude RNA, purified full-length recovery, finishing efficiency, formulation recovery, and potency-relevant material.

Evidence for IVT process behavior comes from biochemical reconstitution, analytical method comparison, sequencing of templates and products, electrophoretic sizing, chromatographic impurity separation, mass spectrometry of nucleoside composition or ends, immunostimulatory assays, and potency assays in cells or animals. Each evidence class has limitations. Gel electrophoresis can show gross integrity but not precise sequence. qPCR can quantify residual DNA but not necessarily its physical association with RNA or particles. Antibody-based dsRNA assays can be useful screens but may have epitope and length biases. Cell-based cytokine assays are biologically relevant but can conflate dsRNA, RNA ends, formulation, endotoxin, and cell-state effects.

## 159.3. Capping, Tailing, Circularization, Ligation, and Enzymatic Finishing

Many RNA products need finishing reactions after or during synthesis because the biologically active molecule is defined by more than its internal sequence. A eukaryotic mRNA needs a 5′ cap that recruits cap-binding proteins, supports translation initiation, protects against exonucleolytic decay, and helps distinguish self-like mRNA from immunostimulatory RNA. A poly(A) tail supports translation and stability through poly(A)-binding proteins and mRNP architecture. A circular RNA needs covalent continuity between its 3′ and 5′ ends. A guide RNA may require chemically stabilized termini or a precise 5′ end for Cas protein loading. The finishing step is therefore part of the product identity.

Capping can be co-transcriptional or post-transcriptional. In co-transcriptional capping, a cap analog is included in the IVT reaction so that initiation can occur with the cap structure. Modern anti-reverse cap analogs and related designs improve the fraction of cap incorporated in the correct orientation compared with older symmetric analogs. Co-transcriptional capping is operationally simple because it avoids a separate enzymatic reaction, but it can reduce yield, create mixtures of capped and uncapped RNA, and impose constraints on initiating nucleotide composition. The cap analog is also an expensive raw material, so its stoichiometry and recovery matter at scale.

Post-transcriptional enzymatic capping uses capping enzymes after RNA synthesis. A canonical enzymatic workflow converts a 5′ triphosphate RNA end to a capped structure through triphosphatase, guanylyltransferase, and methyltransferase activities, sometimes followed by 2′-O-methyltransferase to generate Cap1. Enzymatic capping can produce high cap quality when the 5′ end is accessible and reaction conditions are well controlled. It adds process complexity, enzymes, cofactors, donor methyl groups, and additional purification needs. The major failure modes include incomplete capping, incomplete methylation, degradation during handling, residual enzyme, residual cofactors, and difficulty capping structured or heterogeneous 5′ ends.

![Figure 159.3. Capping, tailing, ligation, and circularization as product-defining finishing reactions](../assets/figures/chapter1142_figure3.png)

**Figure 159.3. Capping, tailing, ligation, and circularization as product-defining finishing reactions.** RNA finishing reactions define biologically meaningful ends and topologies. Each finishing strategy creates distinct incomplete-product and residual-reagent risks.

The difference between Cap0 and Cap1 is not cosmetic. Cap0 has the inverted methylguanosine cap but lacks 2′-O-methylation on the first transcribed nucleotide, whereas Cap1 contains that methylation. In mammalian cells, cap methylation status can influence innate immune discrimination and translation behavior. Therapeutic mRNA manufacturing therefore often specifies cap structure, not merely "capped RNA." Assays for cap status include enzymatic digestion followed by liquid chromatography or mass spectrometry, cap-specific electrophoretic or chromatographic methods, immunoassays in limited contexts, and indirect translation or immune assays. No single assay perfectly substitutes for direct structural evidence when cap identity is critical.

Poly(A) tailing can be encoded or enzymatic. An encoded tail is copied directly from the DNA template, which helps define tail length and reduces post-transcriptional processing. The challenge is that long adenosine tracts can be unstable during plasmid propagation and difficult to verify by short-read sequencing. Enzymatic tailing uses poly(A) polymerase to add adenosines after transcription. This can rescue templates without a long encoded tail and can tune average tail length, but it usually produces a distribution rather than a single tail length. For some products this distribution is acceptable; for others it complicates potency and comparability.

Circularization creates a different product class. Circular RNA can be generated by splint-mediated ligation, ribozyme-assisted ligation, group I intron or permuted intron-exon systems, chemical ligation, or engineered autocatalytic systems. In all cases, the intended product lacks free 5′ and 3′ ends, which can improve resistance to exonucleases and alter translation or immune recognition. The manufacturing challenge is that linear precursor, nicked circles, concatemeric circles, unligated splints, residual enzymes, and incorrectly joined products can be difficult to separate from the desired circle. Circularity also changes analytical behavior: a circular molecule may migrate differently from a linear molecule of the same length, and some sequencing or reverse transcription methods can create artifacts at junctions.

Ligation is also useful outside circular RNA. Chemically synthesized RNA fragments can be ligated to make longer modified RNAs that would be difficult to synthesize directly. Splinted ligation uses a complementary DNA or RNA splint to align donor and acceptor ends for enzymatic joining. The key requirements are compatible end chemistry, correct annealing, adequate ligase specificity, and removal of splint and unreacted fragments. A ligated product can contain single-nucleotide junction errors, missing phosphate groups, residual splints, or positional isomers if multiple ends can react. Therefore, junction verification is often as important as gross length verification.

Enzymatic finishing also includes dephosphorylation, phosphorylation, removal of template DNA, RNA repair, tail trimming, ribozyme cleavage, and nuclease-based cleanup. DNase treatment removes template DNA after IVT, but the DNase itself becomes a process impurity and must be removed or controlled. Nucleases can trim unwanted sequences or digest linear RNA after circularization, but they can also damage the intended product if specificity or timing is poor. A process step that improves one quality attribute can threaten another; DNase reduces DNA but may introduce protein impurity, nuclease cleanup improves circular RNA purity but may nick the product, and enzymatic capping improves cap identity but can lower recovery.

The evidence basis for finishing reactions is a chain of orthogonal measurements. Capping is supported by direct end analysis, cap-specific digestion, and translation or immune-response assays. Tailing is supported by tail-length assays, sequencing strategies that handle homopolymers, and functional expression data. Circularization is supported by RNase R resistance only cautiously, because resistance is not unique to true circles and susceptibility can reflect structure as well as topology; stronger evidence uses junction sequencing, exonuclease challenge, electrophoretic behavior, and direct mapping. Ligation is supported by junction-specific assays and mass or sequence confirmation. The shared caution is that endpoint biological activity cannot identify which finishing defect is present.

## 159.4. Modified Nucleotides and Impurity Profiles Including Double-Stranded RNA

Modified nucleotides are intentionally altered ribonucleotides incorporated into RNA to change stability, translation, immune recognition, structure, or protein interaction. In therapeutic mRNA, modified nucleotides such as pseudouridine or N1-methylpseudouridine can reduce innate immune activation and improve translation in many systems. In guide RNAs and siRNAs, modifications such as 2′-O-methyl, 2′-fluoro, phosphorothioate linkages, and terminal stabilizing groups can improve nuclease resistance, binding behavior, or pharmacokinetics. The same word "modification" therefore covers distinct manufacturing situations: substrate replacement during IVT, protected monomer choice during chemical synthesis, post-synthetic enzymatic modification, or conjugation chemistry.

A modified nucleotide changes the reaction environment. An RNA polymerase that efficiently incorporates canonical uridine triphosphate may incorporate a modified uridine analog with different kinetics, fidelity, sequence bias, and termination behavior. A modification that improves innate immune tolerance can also alter RNA folding, duplex stability, chromatographic behavior, and susceptibility to nucleases or analytical enzymes. In solid-phase synthesis, modified phosphoramidites can have different coupling efficiencies and protecting-group requirements. Therefore, modifications cannot be treated as biological decorations added after manufacturing is solved; they alter the process design and impurity profile.

![Figure 159.4. Modified nucleotide effects on process, structure, immune sensing, and analytics](../assets/figures/chapter1142_figure4.png)

**Figure 159.4. Modified nucleotide effects on process, structure, immune sensing, and analytics.** Modified nucleotides are process variables as well as biological design features. Their effects propagate through synthesis, purification, analysis, and product performance.

Impurity profiles for modified RNAs include both ordinary impurities and modification-specific impurities. Ordinary impurities include truncated transcripts, residual DNA, residual proteins, and degradation fragments. Modification-specific impurities include incomplete substitution, wrong positional incorporation, modified nucleotide degradation products, unmodified carryover, altered misincorporation spectra, and nucleoside analog contaminants. For chemically synthesized oligonucleotides, stereochemical mixtures at phosphorothioate linkages can be intended or controlled depending on the product; for IVT mRNA, the product is usually not a stereochemical mixture at the backbone, but it may be a mixture in cap status, tail length, and modification incorporation.

Double-stranded RNA is the impurity class that most clearly links manufacturing chemistry to immunobiology. dsRNA can arise when an RNA product anneals to a complementary transcript, when polymerase creates antisense or copy-back products, when self-complementary regions form long duplexes, or when partially complementary fragments co-purify. The structural boundary between a designed RNA secondary structure and an unwanted dsRNA impurity is not always sharp. A stem-loop inside an mRNA is not automatically a process impurity, but long duplex contaminants or duplexed antisense products can strongly stimulate innate immune receptors. The manufacturing question is not "does any base pairing exist?" but "what dsRNA-like species are present at what abundance, length, context, and biological activity?"

**Table 159.3. dsRNA assay classes and interpretation limits.** dsRNA assays are not interchangeable because each method detects a different structural or biological property of duplex-like RNA.

| Assay class | Primary readout | Best use in a control strategy | Interpretation limit |
| --- | --- | --- | --- |
| **dsRNA antibody dot blot or immunoassay** | Binding signal from antibody-recognized duplex-like epitopes | Rapid screening, lot comparison, and purification-process monitoring | Signal depends on duplex length, sequence, modification, conformation, and assay matrix. |
| **dsRNA- or ssRNA-selective nuclease sensitivity** | Loss or persistence of RNA after nuclease challenge | Mechanistic support for duplex-like or single-stranded character | Structured single-stranded RNA and modified RNA can confound specificity; nuclease exposure can damage product. |
| **Chromatographic or cellulose-based separation** | Enrichment or depletion of duplex-like species under selected conditions | Process development and impurity removal for the actual RNA product | Co-elution and product loss are possible; performance may not transfer across length, sequence, or modification. |
| **Electrophoresis or capillary sizing** | Mobility, size distribution, and gross integrity | Detecting large contaminants, degradation, or major duplex-associated shifts | Mobility does not prove dsRNA identity, cap status, tail status, or absence of co-migrating species. |
| **Sequencing or strand-origin analysis** | Read patterns suggesting antisense, copy-back, junction, or fragment origin | Identifying sequence origins of suspicious RNA species | Library preparation, reverse transcription, modifications, and low abundance can bias recovery. |
| **Cell-based cytokine or translation assay** | Biological response after RNA exposure or delivery | Testing whether remaining material has immune or potency consequences | Responses can arise from dsRNA, RNA ends, endotoxin, lipid impurities, residual solvent, dose, or cell state. |
| **Orthogonal assay panel** | Concordance across structural, analytical, process, and biological readouts | Release justification and comparability assessment | Discordant results are expected when assays measure different properties; they require investigation rather than averaging. |

Innate immune sensing creates a high consequence for low-abundance contaminants. Endosomal TLR3 can respond to dsRNA, TLR7 and TLR8 can respond to single-stranded RNA features, cytosolic RIG-I-like receptors can respond to certain RNA ends and duplexes, PKR can respond to dsRNA and inhibit translation, and OAS enzymes can activate RNase L after dsRNA recognition. These pathways are treated mechanistically in [Chapter 108](chapter1103.md). For manufacturing, the important point is that an impurity too small to dominate an electropherogram may still influence cytokine release, reactogenicity, or translation shutdown in sensitive cells. Conversely, not every immune response in a formulated RNA product proves that dsRNA is the cause; endotoxin, lipid impurities, residual solvent, particle size, RNA ends, and target-cell state can also contribute.

Analytical detection of dsRNA is difficult because "dsRNA" is a family of structures. Antibody-based dot blots or enzyme-linked assays can detect some dsRNA epitopes but may under- or over-report depending on length, sequence, modification, and conformation. Cell-based assays are sensitive to biological activity but not specific. Chromatographic methods can remove or estimate duplex-like species if the method resolves them from the target, but co-elution is possible. Nuclease digestion with dsRNA-specific or ssRNA-specific enzymes can provide mechanistic clues, but structured single-stranded RNA and modified RNA can confound interpretation. A robust control strategy usually combines process understanding, purification, analytical assays, and biological readouts.

> **Box 159.2. Reading a dsRNA Signal**
>
> Treat a dsRNA result as a question to investigate, not as a finished diagnosis. First ask what the method detects: an antibody epitope, nuclease sensitivity pattern, chromatographic behavior, sequencing-derived complementarity, or a cellular response. Second, ask whether the signal could come from the intended RNA structure, antisense or copy-back transcripts, short complementary fragments, aggregates, formulation components, or assay matrix effects. Third, test whether purification or process changes reduce the signal without damaging the intended RNA. Fourth, ask whether biological readouts such as cytokine release, translation inhibition, or potency track with the analytical result. Discordant dsRNA assays are common because each assay measures a different property of duplex-like RNA. The safest interpretation is orthogonal: combine process knowledge, structural or analytical measurements, impurity removal data, and product-relevant biological assays before assigning causality.

Modified nucleotides can reduce dsRNA formation or reduce immune consequences of dsRNA-like structures in some IVT systems, but this is context-dependent. A modified substrate may change polymerase side reactions, product folding, or sensor recognition. It may also change downstream purification behavior. Therefore, a product made with N1-methylpseudouridine is not automatically free of immunostimulatory impurities, and a product made with canonical nucleotides is not automatically unacceptable. The claim must be tied to a specific sequence, process, purification method, formulation, dose, route, and assay.

The evidence basis for modification incorporation differs by product class. For IVT mRNA, nucleoside composition after enzymatic digestion can show global incorporation but not always positional distribution. Sequencing can identify sequence and some misincorporation events but may not faithfully read all modifications. Mass spectrometry can provide strong evidence for specific nucleosides or oligonucleotide fragments, but complete mapping of a long RNA is technically demanding. For chemically synthesized oligonucleotides, mass spectrometry and chromatographic purity are often more direct because the product is shorter, but positional isomers and stereochemical details can still require specialized methods.

"modified RNA is non-immunogenic." A more accurate statement is that certain modifications, in certain sequence and formulation contexts, can reduce activation of particular innate immune sensors and improve expression or tolerability. Immunogenicity is a property of the whole product and host context, including impurities, delivery material, dose, route, repeat dosing, and tissue. This distinction is essential for interpreting both therapeutic claims and manufacturing specifications.

## 159.5. Purification, Chromatography, Nuclease Control, and Double-Stranded RNA Removal

Purification is the set of operations that turns a synthesis reaction into a defined drug substance, research reagent, or intermediate. Purification is not simply "cleaning up" RNA. It establishes which molecular species remain in the product and at what levels. For clinical products, purification must be reproducible, scalable, documented, compatible with product stability, and linked to specifications. For research products, purification still matters because many RNA experiments fail or mislead when crude reaction mixtures are used as if they were pure molecules.

A typical IVT reaction after transcription contains the intended RNA, template DNA, polymerase, nucleotides, salts, magnesium, pyrophosphate, short transcripts, truncated transcripts, dsRNA, possible endotoxin from raw materials, and process enzymes such as DNase. Early purification often includes DNase digestion, clarification, concentration, and buffer exchange. Tangential-flow filtration or ultrafiltration-diafiltration can remove small molecules and exchange buffer while retaining high-molecular-weight RNA, but these methods do not reliably separate full-length RNA from similarly sized RNA impurities. Chromatography or selective precipitation may be needed for sharper impurity removal.

Chromatography separates molecules by a chosen property. Anion-exchange chromatography exploits the negative charge of nucleic acids; longer RNAs generally bind more strongly, but structure, salt, modification, and counterions affect behavior. Reverse-phase and ion-pair reverse-phase chromatography exploit hydrophobic interactions and ion pairing; these methods can resolve oligonucleotides and some RNA species but must be evaluated for solvent compatibility, scale, and residue control. Size-exclusion chromatography separates by hydrodynamic size but has limited resolving power for closely related long RNA species. Affinity approaches can exploit poly(A) tails, tags, hybridization, or binding proteins, but they introduce ligand leaching, capacity, and specificity questions.

![Figure 159.5. Purification decision tree for IVT RNA](../assets/figures/chapter1142_figure5.png)

**Figure 159.5. Purification decision tree for IVT RNA.** Purification combines impurity removal, concentration, buffer exchange, and product protection. The appropriate path depends on RNA length, structure, modification, and intended use.

dsRNA removal has become a central purification problem for IVT mRNA. Strategies include chromatographic separation under conditions that distinguish duplex-like species, cellulose-based approaches that preferentially bind dsRNA under selected conditions, selective nuclease treatments, process conditions that reduce formation, and upstream template or reaction design. Each strategy has tradeoffs. A method that removes dsRNA may also remove structured full-length RNA or reduce recovery. A nuclease that digests dsRNA may nick the product or leave enzyme residue. A cellulose or chromatography method may perform differently for different RNA lengths, modifications, and formulations. Process development must demonstrate performance for the actual product, not only for a model RNA.

Nuclease control is both an upstream and downstream requirement. RNases are ubiquitous, stable, and sometimes difficult to inactivate. They can enter through raw materials, water, personnel handling, equipment, enzymes, or contaminated containers. A small amount of RNase can produce fragments that appear as lowered integrity, reduced potency, increased innate immune activity, or inconsistent chromatography. Control measures include qualified raw materials, RNase-free water and consumables, clean equipment, closed processing where appropriate, environmental monitoring, enzyme supplier controls, in-process integrity testing, and rapid cold handling. RNase inhibitors can be useful in research but are not a universal manufacturing solution because they add another component that may need removal or qualification.

For chemically synthesized oligonucleotides, purification decisions are different but the logic is the same. A full-length 21-mer siRNA strand can be separated from n-minus species by ion-exchange or reverse-phase chromatography more readily than a full-length 4,000-nucleotide mRNA can be separated from a 3,900-nucleotide truncated transcript. However, highly modified oligonucleotides can have unusual chromatographic behavior, and stereochemical mixtures can complicate peak interpretation. Duplex oligonucleotide products add an annealing and purification problem: the final duplex must be distinguished from residual single strands and mismatched or aggregate-like duplexes.

**Table 159.4. Release-testing panel for representative RNA products.** Release testing expands as product complexity increases from a purified oligonucleotide to a formulated RNA nanoparticle.

| Product or intermediate | Core identity and integrity tests | Impurity and safety tests | Special quality attributes or caveats |
| --- | --- | --- | --- |
| **Chemically synthesized siRNA strand** | Sequence or mass confirmation, length or purity by chromatography or electrophoresis, modification placement where relevant | N-minus and n-plus species, incomplete deprotection, residual protecting groups, salts, counterions, endotoxin, bioburden as appropriate | Passing a single-strand assay does not prove correct duplex formation after annealing. |
| **IVT mRNA drug substance** | Sequence identity where feasible, full-length integrity, concentration, cap structure and capped fraction, poly(A) tail characterization | dsRNA, residual DNA, residual polymerase or DNase, nucleotides, salts, pyrophosphate, endotoxin, bioburden | The release-relevant yield is full-length functional RNA, not total RNA mass. |
| **Circular RNA intermediate** | Junction verification, topology-sensitive electrophoresis or chromatography, exonuclease challenge interpreted with controls | Linear precursor, nicked circles, concatemers, residual splint, ligase, nucleases, dsRNA-like byproducts | Exonuclease resistance alone does not prove homogeneous circular topology. |
| **Formulated mRNA-LNP drug product** | RNA integrity after formulation, RNA content, encapsulation efficiency, particle size, polydispersity, lipid identity and ratio | Residual ethanol or solvent, unencapsulated RNA, lipid impurities, endotoxin, sterility or bioburden, visible and subvisible particles | Product quality includes both RNA attributes and particle attributes; freeze-thaw and storage history can change release-relevant properties. |

Purification also has to protect the product. RNA can hydrolyze under alkaline conditions, fragment under heat or metal-catalyzed stress, adsorb to surfaces, shear during pumping, precipitate with salts or solvents, or aggregate with proteins and lipids. A purification method that gives excellent impurity removal but damages the RNA is not acceptable. Therefore, process development monitors recovery, integrity, impurity clearance, buffer composition, temperature, hold time, shear exposure, and compatibility with the next step. For long RNA, apparently minor hold steps can become important stability risks.

Analytical limitations should be explicit. A chromatographic peak labeled "main peak" is not automatically the correct RNA. It may contain co-eluting truncated molecules, dsRNA, aggregates, or conformers. A gel band at the expected size is not proof of cap status, tail length, sequence identity, or absence of template DNA. A low residual DNA qPCR result does not prove that DNA is absent if the assay target misses fragments or modified templates. A low endotoxin result does not rule out all innate immune stimulation. Good purification science uses orthogonal tests because each assay observes a different property.

Cross-chapter handoff: [Chapter 132](chapter1120.md) treats chromatography and mass spectrometry for RNA modifications in more analytical detail; [Chapter 156](chapter1139.md) treats formulated lipid nanoparticles after purified RNA is combined with lipids; [Chapter 158](chapter1141.md) treats regulatory toxicology and specification logic. This section provides the manufacturing bridge among those topics.

## 159.6. Formulation, Stability, Cold Chain, Lyophilization, Container Closure, Scale-Up, GMP, Release Specifications, and Analytical Methods

Formulation is the process of placing RNA into a physical and chemical environment that allows storage, administration, delivery, and biological activity. A purified RNA solution may be appropriate for an in vitro assay, but many therapeutic RNAs require a formulation that protects the molecule, controls biodistribution, and enables cellular uptake. The formulation can be as simple as a buffered saline solution for local administration of a chemically stabilized oligonucleotide or as complex as a lipid nanoparticle containing ionizable lipid, helper phospholipid, cholesterol, polyethylene glycol lipid, RNA, buffer, cryoprotectant, and controlled residual solvents.

The main stability problem is that RNA is chemically and biologically fragile. Chemical hydrolysis can occur through the 2′ hydroxyl attacking the phosphodiester backbone, especially under alkaline conditions or in the presence of catalytic metal ions. RNases can cleave RNA even when the bulk buffer is chemically benign. Oxidation, depurination, deamination, and modification-specific degradation can contribute under some conditions. Physical instability can include adsorption to surfaces, aggregation, precipitation, shear damage, concentration-dependent self-association, and, in nanoparticles, particle growth, fusion, leakage, lipid oxidation, and loss of encapsulation. Stability is therefore measured by a panel, not a single number.

![Figure 159.6. Stability stress map from purified RNA to filled drug product](../assets/figures/chapter1142_figure6.png)

**Figure 159.6. Stability stress map from purified RNA to filled drug product.** RNA stability is a product-system property. Sequence, chemistry, buffer, formulation, container, and storage history jointly determine shelf life and potency.

For lipid nanoparticle mRNA products, formulation is commonly performed by rapid mixing of an RNA-containing aqueous phase with a lipid-containing organic phase, often ethanol-based. Ionizable lipids are designed to be positively charged under acidic mixing conditions so they complex RNA, but less charged near physiological pH to reduce toxicity. The final particles are then buffer exchanged, concentrated, sterile filtered if feasible for the particle size and process, filled into containers, and stored under defined conditions. [Chapter 156](chapter1139.md) covers lipid nanoparticle composition and trafficking in depth. The manufacturing point here is that particle attributes become product quality attributes alongside RNA attributes.

Critical formulation attributes include RNA concentration, encapsulation efficiency, particle size, polydispersity, lipid composition, lipid impurities, residual ethanol or solvent, pH, osmolality, buffer species, visible and subvisible particles, sterility or bioburden, endotoxin, and potency. Encapsulation efficiency is not merely a yield measurement; unencapsulated RNA may degrade faster, stimulate immune sensors differently, or alter dose calculations. Particle size can influence biodistribution, filtration, reactogenicity, and storage behavior. Lipid oxidation or hydrolysis can change potency or tolerability. A release strategy for formulated RNA must therefore test both the RNA drug substance and the formulated drug product.

Cold-chain storage slows many degradation processes but does not solve all stability problems. Freezing can protect against chemical reactions but introduce ice interfaces, freeze-concentration of solutes, pH shifts, mechanical stress, and particle aggregation. Refrigerated storage can avoid freeze-thaw stress but may allow slower hydrolysis or lipid degradation. Ultra-cold storage can improve stability for some products but complicates distribution, administration sites, and global access. Freeze-thaw history is a process variable because repeated cycles can change particle size, encapsulation, and RNA integrity. Stability claims must specify the formulation, container, temperature, duration, and allowable excursions.

Lyophilization, or freeze-drying, removes water under low temperature and reduced pressure to create a dry or partially dry product that may be more stable and easier to distribute. For RNA products, lyophilization is attractive but challenging. The freezing step can stress RNA nanoparticles; drying can alter particle arrangement; residual moisture must be controlled; reconstitution must restore the intended particle size and potency; excipients such as sucrose, trehalose, buffers, and bulking agents must be compatible with RNA and lipids. Lyophilization does not automatically produce room-temperature stability, and an apparently elegant cake can hide loss of encapsulation or potency.

Container closure is the set of vial, stopper, syringe, cartridge, seal, and contact materials that protect the product until use. RNA products can adsorb to glass, interact with elastomers, experience pH changes from leachables, or be affected by silicone oil, tungsten residues, oxygen, moisture, or extractables. For lipid nanoparticles, container surfaces and headspace can influence aggregation and oxidation. Container closure integrity also protects sterility. A stability program should test the actual container and closure system, not only the buffer in a convenient research tube.

Scale-up changes the physics of manufacturing. A reaction that works in a microliter tube may not mix, cool, filter, or purify the same way in liter scale. IVT scale-up must control heat, mixing, pyrophosphate accumulation, viscosity, enzyme distribution, template quality, and hold times. Chromatography scale-up must control column loading, flow, pressure, resin lifetime, cleaning, carryover, and fraction collection. LNP scale-up must preserve mixing time and solvent exchange behavior because particle assembly occurs rapidly. The correct scale-up question is not "can the same recipe be made larger?" but "which dimensionless or mechanistic variables must be preserved to maintain the same product quality?"

GMP places all these operations inside a quality system. Raw materials must be specified and qualified, including nucleotides, modified nucleotides, enzymes, DNA templates, lipids, solvents, buffers, and containers. Equipment must be cleaned, calibrated, and appropriate for contact with product. Operators must be trained. Deviations must be investigated. In-process controls must detect failures early. Batch records must document what happened. Specifications must define which test results allow release. Method validation or qualification must match development stage and regulatory expectations. GMP does not guarantee that a product is biologically effective; it supports confidence that the product lot is what it claims to be and was made under controlled conditions.

Analytical release testing is the formal evidence package for a batch. Identity testing can include sequencing, enzymatic mapping, mass spectrometry, or hybridization-based methods. Integrity and size can be assessed by capillary electrophoresis, agarose gels, chromatography, or fragment analysis. Purity can be assessed by chromatography, electrophoresis, residual impurity assays, and dsRNA assays. Content and concentration can be measured by absorbance, fluorescence, chromatography, or phosphorus-based methods, each with biases. Cap and tail assays may be needed for mRNA. Residual DNA can be measured by qPCR or ddPCR. Residual proteins or enzymes can be measured by immunoassay or activity assay. Endotoxin, bioburden, sterility, mycoplasma where relevant, pH, osmolality, particulate matter, residual solvent, and container closure integrity belong to the product control strategy.

> **Box 159.3. From Quality Attribute to Release Test**
>
> A release test should answer a product-specific decision question. Start by naming the critical quality attribute: identity, integrity, cap status, tail status, modification composition, dsRNA burden, residual DNA, residual enzyme, endotoxin, sterility, particle size, encapsulation, potency, or another attribute. Then choose a method that actually measures that attribute in the relevant matrix and state its limitation. A cap assay does not measure tail length; an endotoxin assay does not rule out RNA-triggered innate immune activation; a particle-size assay does not prove RNA integrity. Next define an acceptance criterion justified by product risk, stage of development, process capability, stability data, and regulatory expectations. Finally, specify the batch decision: release, reject, reprocess if allowed, investigate, or place on stability hold. A strong release panel is not the longest possible list of assays; it is the smallest defensible set that controls the product's clinically or experimentally relevant failure modes.

Potency is often the hardest assay to design. For mRNA, potency may be measured by protein expression in cells, enzymatic activity of the encoded protein, antigen expression, or functional immune readout. For siRNA, potency may be target knockdown in a cell model. For a guide RNA, potency may be editing activity with a Cas protein. For circular RNA, potency may be expression, stability, translation, or another platform-specific function. A potency assay should be stability-indicating and related to the intended mechanism, but it also must be robust enough for release testing. Highly mechanistic assays can be too variable; highly simple assays can miss relevant failure modes.

Regulatory boundaries should be kept clear. This chapter explains scientific and manufacturing logic; it does not define a universal regulatory checklist. Release specifications depend on product class, route, dose, clinical stage, jurisdiction, risk assessment, platform history, and regulatory agreement. A research-grade RNA standard, an early phase personalized mRNA cancer vaccine, and a commercial prophylactic mRNA vaccine do not have identical specifications. Still, the common principle is stable: specifications should be justified by critical quality attributes, process knowledge, analytical capability, and clinical or nonclinical relevance.

"If the RNA sequence is correct, the product is correct." Sequence identity is necessary but not sufficient. The product also has an end structure, modification pattern, length distribution, impurity profile, formulation state, stability history, and biological potency. Another misconception is that formulation problems can be solved after synthesis. In practice, formulation and synthesis interact: buffer salts, RNA concentration, residual impurities, dsRNA, cap status, tail length, and RNA length can all influence particle formation, stability, and potency.

## Recent Consensus

The current practical consensus is that RNA manufacturing must be product-specific but mechanism-guided. Short chemically modified oligonucleotides, long IVT mRNAs, self-amplifying RNAs, circular RNAs, and formulated nanoparticles share some analytical vocabulary but not identical critical quality attributes. A process that is adequate for a research guide RNA can be inadequate for a clinical mRNA. A purification method that works for one mRNA sequence may need re-optimization for another because length, structure, modification, and formulation change impurity behavior.

Another consensus is that orthogonal analytics are required. RNA products are too heterogeneous for a single chromatogram, gel, or cell assay to define quality. Sequence, length, cap, tail, modification, dsRNA, residual DNA, residual protein, endotoxin, formulation attributes, and potency each need appropriate evidence. The exact panel can vary, but the logic cannot be replaced by a single "purity percent." The best release programs combine direct structural assays, impurity assays, process controls, biological potency, and stability-indicating methods.

The current consensus also treats finishing and formulation as product-defining operations rather than optional polishing. A therapeutic mRNA is not fully described by its coding sequence until cap structure, cap orientation, Cap0 or Cap1 status, poly(A) architecture, modified-nucleotide composition, dsRNA burden, residual template, residual enzyme, purity, potency, and stability profile are measured in a product-relevant way. For lipid nanoparticle products, RNA quality and particle quality must be controlled together because residual salts, dsRNA, RNA length distribution, cap status, lipid composition, particle size, encapsulation, and storage history can interact. For GMP work, release specifications should be justified by critical quality attributes and method capability rather than copied from another RNA modality.

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

Open questions:

- How should quantitative dsRNA standards be improved for long and modified RNA products?
- Which assays best distinguish problematic dsRNA-like species in modified RNA?
- How can complete end and tail characterization be scaled for long RNA?
- How can stability models for lipid nanoparticle RNA products become more predictive?
- How can purification methods maintain high recovery for very long RNA while removing process- and product-related impurities?
- How should analytical attributes be linked to clinical tolerability?
- How should platform manufacturing changes be compared across products?
- How should circular RNA and self-amplifying RNA products be analyzed for topology, replicon integrity, long-RNA potency, innate immune activation, and analytical resolution?

Controversies:

- The most persistent controversies are not about whether RNA can be manufactured reproducibly, but about how much mechanistic evidence is enough for a given use. A dsRNA antibody signal, a nuclease-resistance result, a chromatography shoulder, or a cell-based cytokine readout may each point to duplex-like impurities, but none is a universal measurement of all immunostimulatory RNA species.
- Modified nucleotides usually reduce some innate immune signals and improve translation for many mRNA designs, but the statement is context-dependent because sequence, purification, cap status, tail design, formulation, dose, and cell type can dominate the biological readout.
- Lyophilization, cold storage, and LNP encapsulation are not generic stability solutions; each must be qualified against the actual RNA, lipid composition, container, reconstitution procedure, and intended shelf life.

Common misconceptions:

- "Sequence-correct means product-correct." The molecule also has ends, modifications, topology, impurity profile, formulation state, stability history, and biological potency.
- "Modified RNA means non-immunogenic RNA." DsRNA impurities, 5′ triphosphates, contaminants, formulation, dose, and cell context still matter.
- "GMP grade means universally safe or effective." GMP grade means the lot was made and tested under a defined quality system against agreed specifications.
- "Platform claims transfer automatically across sequences, cap chemistries, ionizable lipids, purification methods, routes, doses, and formulations." Platform knowledge can support inference, but the usual unit of evidence is the product and process.

Deprecated or weakened claims:

- IVT products should no longer be described only by total RNA yield; a high-yield reaction can produce unacceptable truncated RNA, abortive products, dsRNA, residual template, residual enzyme, or poorly capped material.
- A main peak or expected-size gel band should not be treated as proof of identity, purity, cap status, tail length, circularity, or potency.
- Exonuclease resistance alone should not be treated as proof of homogeneous circular topology; residual linear precursor, nicked circles, concatemers, and dsRNA-like byproducts can survive or confound simplified assays.