Abstract
Messenger RNA (mRNA) therapeutics have advanced rapidly, yet sequence design must now satisfy not only biological potency but also scalable manufacturing, product consistency, and immune compatibility. The 3′poly(A) tail is increasingly recognized as a pivotal factor linking these competing demands. While long homopolymeric tails can enhance translation and prolong RNA persistence, their repetitive nature can also compromise DNA template stability, increase in vitro transcriptional slippage and 3′-end heterogeneity, and promote the formation of immunostimulatory byproducts. We define this design trade-off as the Production–Expression paradox, in which sequence features that favor expression may simultaneously undermine manufacturability and immune safety. In this review, we summarize the molecular functions of the poly(A) tail in regulating translation and mRNA decay, and discuss emerging engineering strategies that move beyond conventional linear homopolymers, including segmented designs, chemical modifications, and experimental topological engineering through structured 3′modules, along with their potential impacts on protein yield (often termed a “translational tax”). We further connect upstream production-associated defects to downstream innate immune activation and outline a proposed Quality-by-Design framework for poly(A) optimization, linking critical quality attributes with fit-for-purpose analytical methods ranging from routine release assays to nucleotide-resolution profiling. Reframing poly(A) architecture as an engineerable design parameter, rather than a fixed default element, may improve product comparability, reduce immunogenic burden, and support the more robust translation of mRNA medicines into clinical applications.
Graphical Abstract
Introduction
Messenger RNA (mRNA) therapeutics have revolutionized modern medicine, a journey that began with the first proof-of-concept of in vivo expressed mRNA and culminated in the rapid deployment of COVID-19 vaccines (Wolff et al., 1990; ; Polack et al., 2020; Sahin et al., 2014). Beyond infectious diseases, this modality is now expanding into protein replacement therapies and cancer immunotherapies (Sahin et al., 2014; ). A major determinant of the pharmacokinetic profile and therapeutic efficacy of these molecules is the 3′polyadenosine (poly(A)) tail (To and Cho, 2021), a homopolymeric sequence typically ranging from 100 to 250 nucleotides in mammalian transcripts (). While historically viewed as a monotonous, passive appendage protecting the 3′end, the poly(A) tail is now increasingly recognized as a versatile engineering module subject to precise optimization (Sahin et al., 2014). Mechanistically, it functions as a scaffold for Poly(A)-Binding Proteins (PABP) to facilitate the “closed-loop” conformation with the 5′cap, thereby governing the delicate kinetic balance between translation initiation and mRNA decay—a process often described as a “molecular timer” for intracellular protein output (Passmore and Coller, 2022; ; ; Svitkin et al., 2009; Kahvejian et al., 2001).
However, the engineering of therapeutic poly(A) tails faces what we describe here as a fundamental “Production-Expression Paradox.” Biologically, longer tails (typically >100 nt) are generally required to maximize translational potency and extend half-life in mammalian cells (Passmore and Coller, 2022; ). Conversely, in the context of manufacturing, long homopolymeric tracts can induce instability in bacterial plasmid templates, leading to recombination, deletion, and polymerase slippage (Macdonald et al., 1993; Molodtsov et al., 2014; ). This conflict forces developers to navigate a narrow design space where biological optimality may clash with manufacturing robustness. Consequently, the poly(A) tail is emerging as a focal point for Design for Manufacturing (DFM), potentially determining not only the drug’s efficacy but also the consistency and safety of the final drug substance (Sahin et al., 2014; To and Cho, 2021).
In this Review, we treat the poly(A) tail as a primary engineering variable in therapeutic mRNA development. Rather than discussing tail properties only in descriptive terms, we organize the field around the logistical and biological decisions developers face: which tail architecture to choose, which risk each design is intended to mitigate, which new liabilities it may introduce, and which analytical controls are required to verify performance. We first summarize the well-established mechanistic basis by which poly(A) tails regulate translation and decay. We then examine emerging design strategies that move beyond conventional linear homopolymers, evaluating how these architectures reshape the trade-off between potency and production robustness. Next, we discuss how template- and IVT-derived failure modes propagate into impurity formation and innate immune sensing, and how process controls and purification strategies can reduce these risks. We then place poly(A) characterization within a proposed Quality by Design (QbD) framework, emphasizing the transition from approximate sizing methods to release-compatible and sequence-resolved analytical workflows (; Kato et al., 2006; Pichlmair et al., 2006; Kawai and Akira, 2010). Finally, we extend this discussion to non-canonical RNA platforms, including circular RNA, where initial poly(A)-mimetic or poly(A)-inspired strategies are beginning to serve analogous translational functions.
Scope and literature selection
This review is intended as a narrative review rather than a systematic review. The literature discussed here was selected to support three interconnected aims: (i) to summarize the mechanistic basis by which poly(A) tails influence translation and mRNA decay; (ii) to examine how poly(A)-related design choices affect manufacturability, impurity formation, and analytical control in therapeutic mRNA development; and (iii) to highlight emerging architectures and design concepts that may extend or reconfigure poly(A)-associated functions. Priority was given to studies that were mechanistically informative, experimentally well characterized, or directly relevant to development-stage decision-making in RNA therapeutics and CMC. Foundational studies were included where necessary to establish biological principles, whereas more recent reports were used to illustrate evolving engineering strategies, analytical methods, and translational considerations. Because the field is developing rapidly and includes areas with uneven levels of validation, some sections are intended to be illustrative rather than exhaustive, particularly those addressing emerging topological architectures, AI-assisted design, and forward-looking translational frameworks.
Biological basis of poly(A) tail regulation
Poly(A) tail engineering can be framed as a stoichiometry-and-kinetics problem: PABP occupancy is a major driver of translation, while deadenylation kinetics largely determine the time-to-decay (Figure 1).
FIGURE 1
The PABP footprint and closed-loop formation
Although in vivo mRNA conformations appear highly dynamic rather than rigidly “circular,” (; Miao and Hu, 2025; Tarun and Sachs, 1996) the closed-loop model remains a highly useful framework for therapeutic mRNA. Cytoplasmic poly(A)-binding protein 1 (PABPC1) is the key mediator of this cap–tail synergy (Nicholson-Shaw et al., 2022). PABP binds the tail with an ∼27 nt footprint (), and recruits eIF4G to stabilize cap-bound initiation complexes, which in turn is thought to enhance productive translation (; Nicholson-Shaw et al., 2022; Re et al., 2014; Vicens et al., 2018; ). Consequently, a minimal tail of ∼20–30 nt is considered the threshold to support the first functional PABP binding event (To and Cho, 2021; Xiang and Bartel, 2021). For therapeutic use, tails commonly exceed ∼100 nt to enable cooperative PABP occupancy, improve ribosome recycling, and sterically protect the 3′end from nucleolytic access (; Wells et al., 1998; Munroe and Jacobson, 1990; Safaee et al., 2012).
Deadenylation kinetics and mRNA stability
mRNA durability is extensively regulated by deadenylation, executed in a biphasic manner by PAN2–PAN3 and CCR4–NOT (Yamashita et al., 2005; ; Wahle and Winkler, 2013). When tail length falls below ∼25 nt, PABP dissociates, exposing the transcript to decapping (DCP2) and 3′→5′degradation (exosome), thereby committing the mRNA to rapid clearance (Passmore and Coller, 2022; ). Importantly, CCR4–NOT is highly processive on uninterrupted poly(A); non-A residues can stall its catalytic core. While spontaneous 3′uridylation (e.g., via TUT4/7) accelerates decay (Lim et al., 2014), engineered G/C interruptions have been shown to impede CCR4–NOT processivity, offering a potential strategy to extend the functional window without simply “buffering” by length (; Lee et al., 2024; Lim et al., 2018).
The coupling between translation and decay
Translation and decay exist in a competitive relationship through shared access to the poly(A) tail (Webster et al., 2018; ). High PABP occupancy promotes initiation while sterically limiting deadenylase engagement, coupling active translation to increased stability (; ; ). Conversely, deadenylation erodes the PABP–eIF4G bridge, reducing protein output and licensing decay (Preiss et al., 1998; Torkzaban et al., 2025). These mechanistic principles motivate architectures that preserve sufficient PABP occupancy while slowing CCR4–NOT—design constraints that intersect directly with manufacturability.
Poly(A) tail engineering: from linear sequences to structured designs
Harnessing the biological principles outlined above, the poly(A) tail can be reconceptualized. Rather than being viewed merely as a static structural appendage, it can serve as a potential “programmable timer” that helps orchestrate the kinetic equilibrium between translational output and mRNA decay (Passmore and Coller, 2022; Nicholson and Pasquinelli, 2019). While endogenous tails are dynamic substrates subject to constant enzymatic remodeling (; Nicholson and Pasquinelli, 2019), therapeutic mRNA typically demands engineered stability to withstand the rigors of manufacturing and sustain in vivo potency (Rosa et al., 2021). Current strategies have evolved from simple length optimization to segmentation, chemical “armoring,” and proposed 3D topological engineering (Figure 2).
FIGURE 2
The length paradox: production stability vs expression potency
Therapeutic mRNA design faces what has been termed the “Production-Expression Paradox”: while high-level expression in many mammalian systems generally requires long tails, E. coli—the standard production host for plasmid templates—often exhibits intrinsic instability toward long homopolymeric tracts (Trepotec et al., 2019).
The Biological Demand: In eukaryotic cells, poly(A) tail length often correlates positively with translational output, driven by the cooperative binding of PABP (; Wells et al., 1998). Recent high-resolution analyses in specific cell-free systems suggest that while mRNA stability increases linearly with tail length, translational initiation efficiency actually peaks at approximately 75 nt and may decline beyond 100 nt; studies suggest that this is possibly because longer tails provide diminishing translational benefit and may impose structural constraints (; Peng and Schoenberg, 2005).
The Manufacturing Ceiling: Conversely, bacterial manufacturing imposes strict physical and thermodynamic constraints. Long poly (dA:dT) sequences (>100 bp) in plasmids are widely regarded as hotspots for polymerase slippage and homologous recombination (). These repetitive tracts have been shown to induce the formation of non-B DNA structures (e.g., triplexes or bent DNA), which stall the replication fork and trigger extensive deletions, leading to heterogeneity in the final mRNA drug substance (Trepotec et al., 2019; ).
Practical Compromise Length: Across platforms, many developers target an initial tail length on the order of ∼80–120 nt as a starting point (), then tune length in the context of the full construct (5′/3′UTRs, ORF, nucleoside chemistry, cap structure) and the intended dose and route (To and Cho, 2021; Linares-Fernández et al., 2020; ; Kim et al., 2024). In this range, PABP occupancy is thought to support closed-loop translation, while limiting some of the known manufacturing liabilities. Importantly, the optimal length appears to be context-dependent and is typically justified using potency, durability, and innate immune readouts together with lot-to-lot distribution control (Linares-Fernández et al., 2020).
Segmented design for improved template stability
To address the ∼100 nt plasmid bottleneck and achieve more durable expression, segmented poly(A) tails have emerged as a practical engineering strategy. By inserting short non-adenosine linkers, these designs disrupt the homopolymeric continuity of the DNA template (Trepotec et al., 2019). This modification can substantially reduce bacterial recombination associated with long homopolymeric tracts without compromising the biological function of the RNA (To and Cho, 2021).
The Spacer Evolution: Early successful designs, such as Pfizer/BioNTech’s BNT162b2 (A30-L10-A70), utilized a 10-nt linker to stabilize the template (). Trepotec et al. (2019) demonstrated that splitting a 120 nt tail into two 60-nt segments (A60-L-A60) drastically reduced plasmid recombination loss from ∼50% to negligible levels (Trepotec et al., 2019).
Spacer Composition and Design Trade-offs: The sequence of the linker also matters. While guanosine (G) spacers are superior for stabilizing DNA templates compared to C or T (Trepotec et al., 2019), they introduce a trade-off: G-rich RNA sequences can form G-quadruplexes or trigger innate immune sensors, potentially compromising cell viability (Li C. Y. et al., 2022). Therefore, the choice of spacer should be viewed as a balance between upstream plasmid stability (favoring G) and downstream immunogenicity (favoring C- or A-rich linkers).
Hyper-segmentation: Recent studies have extended this concept further. Spiewla et al. (2026) introduced a hyper-segmented design using repeating CA15 motifs (Spiewla et al., 2026). This architecture may provide dual benefits: the repetitive insertion of cytidines (C) disrupts homopolymeric continuity, thereby improving plasmid stability, and may also slow deadenylation by introducing distributed interruptions within the tail. This design stabilized tails exceeding 200 nt and reportedly provided a substantial increase in protein yield compared to standard A90 tails (Spiewla et al., 2026). These findings suggest that segmented architectures may enable the use of tail lengths that are otherwise difficult to manufacture reliably.
Tail hardening strategies to resist deadenylases
Since deadenylation by the CCR4–NOT complex is the rate-limiting step of decay, protecting the 3′terminus is a potentially effective strategy for extending mRNA half-life (Webster et al., 2018; ).
Sequence Doping: Leveraging the substrate specificity of deadenylases, researchers have incorporated non-adenosine residues—specifically cytidines (C)—at the 3′end. These C residues may act as local barriers to nuclease progression, delaying the onset of decay (Li C. Y. et al., 2022).
Chemical Modification: A more robust approach involves chemical alteration of the backbone. Introducing phosphorothioate (PS) linkages at the 3′end provides increased resistance to exonucleases while maintaining translational activity (Strzelecka et al., 2020). This strategy is conceptually related to stabilization approaches long used in antisense oligonucleotides (ASOs) (). However, directly transferring ASO-type modifications to much larger therapeutic mRNAs (∼2,000 nucleotides) creates two major development challenges. These span both formulation and safety: on the formulation side, hydrophobic PS linkages may alter the tail’s charge and flexibility, disrupting mRNA–lipid interactions and yielding suboptimal LNP morphology (; ); on the safety side, extensive PS incorporation has been associated with liver injury, platelet activation/thrombocytopenia, and off-target protein binding, making non-human primate (NHP) validation, beyond standard murine models, essential for robust translational assessment (; Valenzuela et al., 2023).
Topological engineering of structured 3′modules: benefits and trade-offs
Poly(A) engineering is extending from linear sequence design to 3′topological architectures that resist decay, but these gains must be weighed against added CMC and in vivo delivery complexity.
Structural Shielding and the Manufacturing Constraints: Structured 3′modules, including looped termini (Oh et al., 2025) and branched poly(A) tails (), can reduce exonucleolytic access and increase local PABP density, but they usually require post-transcriptional assembly, lowering GMP yield, increasing batch variability, and complicating characterization. Their altered size and rigidity may also affect LNP encapsulation and endosomal escape, while structured 3′ends may increase the risk of unintended dsRNA-like motifs that engage RIG-I/MDA5 (Schlee et al., 2009). Consequently, robust implementation requires rigorous quality control (QC) gates, such as advanced biophysical characterization (e.g., SEC-MALS) combined with highly sensitive innate immune profiling.
Dynamic Repair and Pharmacokinetic Variability: Biomimetic motifs such as A7S can recruit TENT4A/B to restore tail length in the cytoplasm, prolonging mRNA persistence (Lim et al., 2018; Kim et al., 2020; ). However, this mechanism depends on host TENT4 abundance, creating cell type–and tissue state–dependent pharmacokinetics (). Sustained competition for TENT4 may also perturb endogenous RNA homeostasis under repeated dosing (Lim et al., 2018; ). To adequately manage these risks, analytical strategies must shift toward cell-line-specific potency and half-life assays tailored to the intended clinical target tissue.
Regulatory Switches and Co-delivery Hurdles: Tail-linked control systems, including ASO-recruited poly(A) mimetics (Torkzaban et al., 2025) and 5′UTR pA regulators (Luo et al., 2024), enable inducible translation but impose demanding co-delivery requirements. Precise co-delivery of mRNA and a second regulator is difficult in vivo and often results in discordant biodistribution and reduced efficacy (; Li S. et al., 2022; ). Validating such multi-component systems therefore necessitates advanced QC gates, leveraging dual-fluorophore tracking in NHP or complex animal models to accurately map in vivo co-delivery kinetics.
A comparative framework for poly(A) design in therapeutic development
The transition from wild-type poly(A) homopolymers to engineered topologies reframes tail selection as a multi-parameter optimization problem, since maximizing in vitro stability yields diminishing returns if it inadvertently exacerbates in vivo immunogenicity or undermines GMP feasibility.
To support development decisions, we constructed a benchmarking matrix (Table 1) comparing poly(A) architectures across plasmid stability, expression kinetics, GMP scalability, immunogenicity risk, and the Intellectual Property (IP)/Freedom to Operate (FTO) constraints.
TABLE 1
| Poly(A) architecture | Design logic | Template stability/plasmid compatibility | Reported expression/persistence effect | Manufacturing feasibility | Immune-related concerns | IP/FTO considerations | Evidence basis |
|---|---|---|---|---|---|---|---|
| Conventional linear WT tails (>100 nt) | Long uninterrupted poly(A) tail | Often limited | Strong in some settings; length-dependent | Simple concept; upstream burden at long length | Generally low; impurity risk indirect | Relatively open | Well supported |
| Segmented tails | Non-A spacers interrupt homopolymer | Often improved | Maintained or improved; system-dependent | High practicality | Spacer-dependent | Increasingly crowded | Supported by multiple studies |
| 3′end chemical hardening | Terminal protection against deadenylation | Limited upstream benefit | Potentially improved persistence | Moderate complexity | Mixed; added chemistry-related liabilities | Chemistry-dependent | Context-dependent |
| Structured/topological 3′modules | Looped, branched, or structured tails | Not primarily template-focused | Potentially improved persistence | Often limited | Possible dsRNA-like features | Design-specific | Evidence emerging |
| Dynamic repair motifs (e.g., A7S/TENT4-related) | Recruit host machinery to maintain tail function | Compatible if sequence is stable | Potentially prolonged persistence | Biologically variable | PK and homeostasis concerns | Mechanism-specific | Promising but early |
| Tail-linked regulatory/co-delivery systems | Translation control by co-delivered regulator | Depends on encoded backbone | Inducible, but delivery-dependent | Low to moderate practicality | Multi-component liability | Potentially constrained | Exploratory |
Interpretive comparison of representative poly(A) engineering strategies for therapeutic mRNA development.
This table provides an interpretive comparison of representative poly(A) engineering strategies intended to support conceptual evaluation in therapeutic mRNA, development. The categories summarize general design logic, commonly discussed advantages, and frequently noted liabilities in the literature, but they should not be interpreted as a standardized performance ranking or meta-analytic conclusion. The “Evidence basis” column is included to distinguish more mature areas from emerging or exploratory concepts.
As the matrix highlights, linear WT tails >100 nt are associated with substantial manufacturing liabilities, with plasmid recombination rates reaching ∼50% in standard cloning hosts. Segmented tails (e.g., A60–L–A60 or hyper-segmented CA repeats) currently offer a favorable balance of low recombination, strong translational performance, and GMP-compatible manufacturability, although their IP landscape is becoming increasingly crowded (Trepotec et al., 2019; Spiewla et al., 2026). By contrast, branched or looped architectures provide stronger biological shielding but face major industrial constraints: they rely on post-transcriptional chemical coupling or complex folding, which caps process yield and raises CMC characterization and regulatory burden (Oh et al., 2025; ). Ultimately, the optimal poly(A) design is modality-dependent: acute indications (e.g., vaccines) favor robust manufacturability (segmented tails), whereas chronic replacement therapies may justify higher CMC complexity for self-regenerating or topologically hardened modules.
Manufacturing robustness: overcoming poly(A) hurdles
Transitioning poly(A) design from the bench to industrial scale demands a robust Design for Manufacturing (DFM) framework. The core challenge lies in the intrinsic thermodynamic instability of long homopolymers, which manifests differently across distinct production phases—from plasmid propagation to enzymatic synthesis. To ensure batch-to-batch consistency and regulatory compliance, developers should deploy a multi-tiered strategy encompassing host engineering, strict process control, and purification logic (Table 2).
TABLE 2
| Process stage | Representative failure mode | Likely root cause | Potential impact | Representative mitigation | Suggested detection/control | Evidence basis |
|---|---|---|---|---|---|---|
| Plasmid propagation | Tail deletion or contraction | Replication slippage; recombination | Template heterogeneity; altered tail length | Segmented tails; strain selection | Plasmid sequencing; encoded-tail verification | Well supported |
| Plasmid maintenance/fermentation | Plasmid multimerization | Recombination; segregational instability | Lower effective yield; higher COGs | Circular architectures; multimer-control systems | Topology assessment; yield monitoring | Supported |
| Template architecture selection | Poor manufacturability of specialized vectors | Linear or low-productivity vector formats | Reduced scalability | Favor simpler encoded designs when possible | Template qualification; COG comparison | Context-dependent |
| IVT reaction | Tail-length heterogeneity | T7 slippage on homopolymers | Broad tail distribution; batch inconsistency | Reduce repetitiveness; optimize IVT conditions | Tail-distribution analysis | Well supported |
| IVT reaction | dsRNA byproduct formation | Self-templating; antisense synthesis; IVT kinetics | Innate immune activation; lower tolerability | Process tuning; downstream dsRNA removal | dsRNA assays; functional immune testing | Well supported |
| Polymerase/reagent sourcing | Supply-chain fragility | Reliance on specialized enzymes | Cost and GMP risk | Prefer broadly available reagents when possible | Vendor qualification; comparability testing | Development-oriented |
| Downstream purification | Residual truncated/tailless transcripts | Abortive transcription; incomplete separation | Heterogeneity; immunostimulatory species | Oligo-dT enrichment; orthogonal cleanup | Transcript sizing; impurity profiling | Supported |
| Downstream purification | Residual dsRNA contamination | Incomplete impurity clearance | RIG-I/MDA5 activation risk | Cellulose purification; RP-HPLC | dsRNA-specific assays | Well supported |
| Tail installation strategy | Poor comparability with enzymatic tailing | Broad stochastic tail-length distribution | Weaker lot control | Prefer template-encoded tails for controlled products | CGE; sequence-resolved methods | Supported |
| Sequence design/linker composition | Unintended immune liability | U-rich motifs; GC-rich structures; dsRNA-like pairing | Increased innate sensing risk | Lower-risk linker design; construct-level validation | Structure review; immune testing | Evidence emerging |
| Integrated control strategy | Over-reliance on one assay | Multi-stage defect sources | Incomplete QC; weak root-cause analysis | Tiered control strategy | CGE + sequencing + functional assays | Strong practical support |
Representative failure modes and control strategies for poly(A)-related mRNA manufacturing.
This table summarizes representative poly(A)-related manufacturing risks and corresponding control strategies across different process stages. It is designed as a practical, fit-for-purpose framework rather than a complete inventory of all failure modes, mitigations, or analytical controls. The examples shown may vary in relevance depending on construct architecture, manufacturing platform, purification workflow, and intended clinical application.
Upstream controls: stabilizing the DNA template
The primary source of manufacturing failure occurs before transcription begins: the instability of the plasmid DNA template. While segmented tails (discussed previously) are the primary DFM solution to disrupt homopolymer continuity (Trepotec et al., 2019; Li C. Y. et al., 2022), residual recombination risks in long poly (dA:dT) tracts persist, necessitating further intervention (Molodtsov et al., 2014; Triana-Alonso et al., 1995; ).
Host Engineering: Standard cloning strains often fail to maintain long repetitive sequences. Consequently, the use of specialized Escherichia coli strains (e.g., Stbl3 or recA-deficient derivatives) is often advantageous. Crucially, simple recA deficiency is often insufficient, as long homopolymers remain susceptible to RecA-independent recombination instability mechanisms (e.g., replication slippage) (; ). Therefore, strains should be specifically validated to suppress these residual instability mechanisms, preserving the integrity of full-length templates during bacterial propagation.
Combating Segregational Instability: A more subtle issue is the formation of plasmid multimers (). Poly(A) tracts can accelerate homologous recombination between plasmid copies, driving the formation of dimers, trimers, and higher-order multimers (; ). This reduces the number of independent segregation units and can lead to uneven partitioning during cell division, thereby increasing plasmid loss during fermentation (; Viguera et al., 2001). To counteract this, modern therapeutic vectors may incorporate active resolution systems, such as cer sites (which utilize the host XerCD recombinase to resolve multimers back into monomers) or active partitioning loci (e.g., parABS), to support stable inheritance (; ).
Vector Architecture and the Cost-of-goods Trade-offs: To propagate extreme tail lengths (∼500 nt) that exceed the practical limits of circular vectors, linear vectors such as pEVL (derived from pJazz) use telomere-capped ends to bypass recombination entirely (). However, this approach introduces substantial manufacturing bottlenecks. Poor E. coli fermentation titers increase cost of goods (COGs) relative to standard high-copy circular vectors, while the altered physiology of the required host strains can complicate downstream alkaline lysis and primary recovery. Consequently, linear plasmids may remain too costly for broad deployment, favoring scalable segmented tails within standard circular architectures in many manufacturing settings (Trepotec et al., 2019; ).
Midstream controls: IVT fidelity and impurity mitigation
Even with a stable DNA template, the in vitro transcription (IVT) reaction introduces its own fidelity issues. T7 RNA polymerase is prone to transcriptional slippage on homopolymers, in which the enzyme loses register on the template, resulting in tail-length heterogeneity (±5%–10%) and premature termination (Macdonald et al., 1993; Molodtsov et al., 2014; Triana-Alonso et al., 1995).
Polymerase Engineering and Supply Chain Constraints: Emerging strategies focus on engineering T7 polymerase variants with improved register fidelity. Analogous to the development of high-fidelity polymerases for next-generation sequencing (NGS) (; Leconte et al., 2010), these variants could in principle reduce indel errors and slippage. However, their industrial implementation currently faces major supply-chain barriers (Kuzmine and Martin, 2001; Wu et al., 2021; ). In contrast, wild-type (WT) T7 is already a widely available, industry-standard GMP reagent. Switching to custom variants may create dependence on a limited supplier base and substantially increase costs. Until multiple vendors can reliably provide these enzymes, WT T7 is likely to remain the more practical option for many manufacturing workflows.
Suppressing dsRNA: At the same time, reliance on WT T7 requires manufacturers to manage its known byproducts. One of the most important safety concerns at this stage is the formation of double-stranded RNA (dsRNA). The poly(A) tail may serve as a hotspot for RNA-dependent RNA polymerase-like activity, in which T7 polymerase engages the 3′end of the nascent transcript to synthesize a complementary antisense strand (Figure 3) (; Mu et al., 2018). Process tuning can help suppress this mechanism (i) Thermodynamic Control: High-temperature IVT (e.g., 42 °C): destabilizes the transient RNA:DNA hybrids that promote slippage (Wu et al., 2020). (ii) Kinetic Control: A “low steady-state UTP feeding” strategy creates a kinetic bottleneck that stalls backward-slipping polymerases, effectively reducing dsRNA byproducts by 60%–70% (Ziegenhals et al., 2023). (iii) Additives: The inclusion of chaotropic agents can further disrupt tail self-pairing, maintaining the RNA in a single-stranded state (Piao et al., 2022).
FIGURE 3
Downstream controls: purification and strategy selection
The final quality gate involves removing truncated species and making strategic choices regarding tail-addition methods.
Purification Strategy: For the mRNA product, oligo-dT affinity chromatography remains the industry standard for enriching full-length tails and separating them from abortive transcripts (; ). However, affinity purification alone is insufficient to address all safety concerns. Removal of dsRNA contaminants—often by RP-HPLC or cellulose purification—remains an important step in limiting innate immune activation, consistent with the principle established by Karikó et al. (; Karikó et al., 2011)
Template-encoded versus Enzymatic Tailing: Finally, manufacturers must choose between encoding the tail in the template and adding it enzymatically. While enzymatic tailing offers flexibility during early-stage research, it produces a stochastic distribution of tail lengths. In a Quality by Design (QbD) context, template-encoded polyadenylation is strongly preferred for GMP manufacturing because it provides better reproducibility and tighter length control. Despite the upstream cloning challenges, this approach is better aligned with the need to maintain tail length as a controlled quality attribute across production batches (Rosa et al., 2021; Pardi and Krammer, 2024).
Mitigation strategies: aligning sequence design and process control
To reduce innate immune recognition that can compromise therapeutic efficacy, manufacturers should employ an integrated strategy. This approach combines process purification (to remove immunostimulatory byproducts) with structural design (to minimize immunogenic features at the sequence level) (Table 2).
Purification: beyond IVT optimization
While upstream optimizations of the in vitro transcription (IVT) reaction—such as high-temperature synthesis (Wu et al., 2020) and low-UTP feeding (Ziegenhals et al., 2023)—significantly reduce the baseline formation of double-stranded RNA (dsRNA), they rarely eliminate it entirely. Consequently, downstream purification remains an important phase for improving product safety.
dsRNA Depletion: The removal of dsRNA contaminants is a high priority, as even trace amounts can trigger RIG-I and MDA5 sensors (; Kato et al., 2006; Pichlmair et al., 2006; Kawai and Akira, 2010; Mu et al., 2018; ). Cellulose chromatography has emerged as a scalable and robust method for selectively depleting these byproducts (). High-capacity affinity chromatography has also been reported to enable scalable removal of dsRNA impurities from IVT mRNA (). Alternatively, RP-HPLC remains a widely used high-resolution method for separating dsRNA based on hydrophobicity, building on purification principles established in earlier work by Karikó et al. (Karikó et al., 2011)
Enzymatic Treatment: Some processes employ RNase III, an enzyme specific to double-stranded RNA, to digest helical regions in the crude product. However, this approach requires rigorous control to prevent unwanted cleavage of therapeutically relevant RNA structures ().
Integrity Control: Finally, oligo-dT affinity purification acts as an important quality-control step (; ). By selecting for full-length poly(A) tails, this step also reduces the abundance of tailless abortive transcripts. These truncated species may be particularly immunostimulatory because their exposed 5′-triphosphates can serve as ligands for RIG-I activation (; Pichlmair et al., 2006; Schlee et al., 2009; ).
Sequence design: structural considerations and linker constraints
A complementary strategy is to reduce immunogenicity through sequence design itself. Beyond widely used chemical modifications (e.g., N1-methylpseudouridine) (; Karikó et al., 2005), the structural properties of the poly(A) region may also influence innate immune recognition.
Homopolymeric versus Structured Tails: Pure homopolymeric poly(A) sequences are often favored because they tend to remain flexible and predominantly single stranded, enabling efficient PABP binding. In contrast, more complex or GC-rich tail sequences may form intramolecular hairpins that resemble dsRNA-like structures and could increase innate immune sensing (Mu et al., 2018).
Linker composition in segmented tails: When segmented tails are used to improve plasmid stability, linker composition becomes an important design variable. G/C-containing spacers may be preferable to U-rich linkers in some contexts (Trepotec et al., 2019). U-rich regions may introduce motifs recognized by TLR8 or facilitate non-specific antisense hybridization, thereby promoting dsRNA formation (Kawai and Akira, 2010; Li C. Y. et al., 2022; ; ).
Poly(A) integrity and immune exposure
Taken together, these observations suggest that the poly(A) tail contributes not only to mRNA stability and translation, but also to the extent to which the transcript remains exposed to innate immune surveillance.
In a properly engineered therapeutic mRNA, a full-length tail recruits PABP, supports closed-loop formation, and promotes active translation. This PABP-associated state may also reduce immune exposure by limiting access of some sensors to the 3′end and by favoring engagement of the transcript with the translation machinery rather than with cytosolic surveillance pathways (; Luo et al., 2024).
Conversely, tail truncation can erode PABP protection, reduce translational efficiency, and increase the likelihood that transcripts enter stress-associated RNA states. Under these conditions, the RNA may become more exposed to pathways involving Protein Kinase R (PKR) and 2′-5′-Oligoadenylate Synthetase/RNase L, with downstream consequences including eIF2α phosphorylation and translational arrest (Mu and Hur, 2023). In this sense, maintaining poly(A) integrity is not only a manufacturing objective, but also a biologically relevant determinant of intracellular RNA behavior.
Integrated QC strategy and regulatory context
In practice, a one-size-fits-all analytical strategy is insufficient. Instead, a tiered quality-control (QC) framework is more appropriate, distinguishing between high-throughput methods for routine manufacturing and high-content methods for deeper biological and structural characterization (Figure 4) (Rosa et al., 2021).
FIGURE 4
Routine lot release: pragmatic assays for distribution control
For routine GMP manufacturing, capillary gel electrophoresis (CGE) serves as a practical quantitative assay (; Wu et al., 2024; Webb et al., 2025). Its adoption is driven by its speed, precision, and operational similarity to established protein QC methods (e.g., CE-SDS), making it accessible to many QC laboratories. CGE provides a rapid profile of tail-length distribution and can support batch-level consistency assessment when aligned with predefined specifications (Table 3) (United States Pharmacopeia USP, 2023).
TABLE 3
| Analytical technique | Typical output | Resolution | Turnaround time | Main limitations | Best use context | Evidence basis |
|---|---|---|---|---|---|---|
| CGE | Tail-length distribution | High | Hours | No direct sequence information; sensitive to sample design | Routine QC; lot comparison; release-supporting analysis | Well supported |
| IP-RP HPLC | Bulk size/purity profile | Low to moderate | Hours | Limited precision for tail-length assignment | Orthogonal purity check; in-process control | Established, but limited for precise tail sizing |
| 3′AIM-seq/related short-read NGS | Tail composition; non-A residues; sequence-level variation | High | Days | Less robust for very long homopolymers; higher workflow complexity | Development; method validation; root-cause analysis | Strong in development settings |
| Direct RNA sequencing (ONT) | Full-length RNA context; isoforms; approximate tail length | Moderate | Hours to days | Limited precision for exact homopolymer length | Exploratory characterization; full-length transcript analysis | Useful for context, less suitable for release quantification |
| RT-ddPCR | Targeted abundance or threshold readout | Not length-resolving | Hours | Does not measure tail-length distribution directly | Rapid process check; targeted monitoring | Useful as a supportive assay |
Fit-for-purpose comparison of representative methods for poly(A)-tail analysis.
This table compares representative analytical methods that may be used to assess poly(A)-tail attributes in different settings, including development, investigation, comparability, and routine control. The entries are intended to highlight typical strengths, limitations, and use contexts, rather than to imply universal equivalence or a fixed regulatory hierarchy. Method suitability remains context-dependent and should be justified according to the specific product, question, and stage of development.
Deep characterization: sequence-resolved methods for development and comparability
However, physical sizing alone cannot resolve sequence-level defects. During process characterization, method development, or root-cause investigation, next-generation sequencing (NGS)-based methods—such as 3′AIM-seq—or long-read platforms such as Oxford Nanopore Technologies (ONT) can provide higher-resolution information on tail composition and sequence fidelity. These methods may help identify untemplated additions, non-A residues, or other sequence features that are not readily captured by electrophoretic sizing alone (Table 3) (Workman et al., 2019; Subtelny et al., 2014; Krause et al., 2019; ; Legnini et al., 2019).
Regulatory interpretation and QbD alignment
From a Quality by Design (QbD) perspective, it is increasingly reasonable to treat poly(A)-tail attributes as quality-relevant analytical features when they are linked to product performance, consistency, or comparability. Draft and emerging guidance documents, including those cited here, support closer attention to RNA structural and sequence-related attributes, but the extent to which poly(A)-tail properties are formally designated or assessed may vary across regulatory context, product class, and stage of development.
Accordingly, manufacturers should be prepared to justify whether tail length, composition, and distribution should be monitored as part of an overall control strategy, particularly when these features may influence potency, stability, or immunogenicity. In comparability settings, it may also be important to show that poly(A)-related attributes remain acceptably consistent with the clinical reference material used in pivotal studies (; ; U.S. Food and Drug Administration FDA, 2022; ).
Consequently, the use of at least one validated quantitative method for poly(A)-tail distribution analysis may be justified within a release or comparability framework, while higher-resolution assays may be reserved for development, investigation, or extended characterization (Table 3). With robust metrology established, the next step is to integrate these measurements into a broader QbD framework for process understanding and control.
Poly(A)-inspired design in emerging RNA architectures
As RNA therapeutics evolve beyond linear formats, the functions traditionally associated with the poly(A) tail are increasingly being re-examined in other architectural contexts. Rather than viewing the poly(A) tail solely as a terminal appendage, recent studies suggest that some of its key roles—particularly PABP recruitment and translational support—may, in certain settings, be partially recapitulated in alternative topologies. By functionally separating PABP recruitment from the physical position of a 3′end, these approaches expand current design strategies for RNA stability and translation.
Circular RNA and internal poly(A)-based translational support
Circular RNAs (circRNAs) present an important design contrast to linear mRNAs: their covalently closed structure confers strong resistance to exonucleases (), whereas the absence of a 5′cap and 3′tail has historically limited translational efficiency relative to linear mRNA (Legnini et al., 2017). To improve their translational performance, recent biomimetic strategies have attempted to recreate some features of cap–tail functional coupling within a circular scaffold ().
Mechanistic Basis: Early studies showed that eukaryotic ribosomes could engage circular templates containing internal poly(A) tracts (). More recent engineering efforts have extended this concept by introducing a precisely spaced internal poly(A) segment (typically ∼12–20 nt) to recruit PABP within the circular RNA (). When combined with an internal ribosome entry site (IRES) and an aptamer capable of recruiting eIF4G, this configuration may help functionally bridge translation-supporting elements that, in linear mRNA, are normally coordinated through cap–tail communication (; Tong et al., 2025; Shen et al., 2025; Khan et al., 2023). In some reported systems, such architectures have approached the protein output of linear mRNAs (; Tong et al., 2025; Shen et al., 2025; Khan et al., 2023).
Design Constraints in Circular Architectures: However, these architectures also introduce important design constraints. Internal poly(A) elements require careful placement, because ribosomal read-through into inappropriate sequence contexts may promote poly-lysine-associated stalling or frameshifting (Juszkiewicz and Hegde, 2017). In addition, the cyclization process can leave structured junctional sequences that may alter translation behavior or increase the likelihood of dsRNA-like features (Pichlmair et al., 2006; Schlee et al., 2009). If such junctions are translated aberrantly, they could in principle generate unintended peptide products, which may be relevant when considering repeated dosing or chronic applications. Analytically, these considerations support the use of junction-resolved RNA characterization to confirm the integrity of the internal PABP-binding region after ligation (Ma et al., 2019). From a translational perspective, the immunogenicity and in vivo tolerability of these engineered junctions would likely require careful evaluation in advanced preclinical models.
Structural analogues and topological hybrids
Beyond circular RNAs, other emerging architectures use structured elements to substitute for, mimic, or stabilize functions typically associated with the poly(A) tail.
Viral-derived Structural Motifs in SaRNA: Alphavirus-derived self-amplifying RNAs (saRNAs) contain 3′conserved sequence elements (CSEs) that fold into defined secondary structures such as pseudoknots. These elements may slow tail remodeling or degradation by limiting deadenylase access, thereby contributing to transcript persistence in some vaccine-associated settings (Khan et al., 2023; ).
Dumbbell RNAs: A related concept is represented by dumbbell RNA architectures (). By sealing both termini with tight stem-loops, these molecules achieve strong resistance to exonuclease attack while retaining a single-stranded poly(A)-containing region for PABP interaction. Conceptually, this design combines enhanced structural stability with retention of a translation-supportive element, although the extent to which these advantages translate across delivery and therapeutic settings remains to be determined ().
Formulation and Translational Constraints: A major open question for topological hybrids such as dumbbell RNAs is whether their structural advantages can be maintained through clinically relevant delivery workflows. From a biophysical modeling perspective, it is hypothesized that their rigid double-stranded termini may increase the hydrodynamic radius of the payload relative to flexible linear transcripts. This theoretical alteration in geometry could affect encapsulation behavior in LNP systems originally optimized for linear mRNAs (Li S. et al., 2022; Schober et al., 2024). This altered geometry may also influence internal lipid organization and, potentially, endosomal escape, although the extent of these effects is likely to be formulation-dependent (). In addition, encouraging durability observed in mice may not translate directly across species, given known differences in innate immune sensitivity and RNA sensing thresholds (Schlee et al., 2009; Lemdani et al., 2024; ; ). Accordingly, the comparative translational value of these architectures relative to optimized segmented tails remains unresolved and will require further validation in more predictive preclinical models.
Discussion
Architectural evolution and translational CMC
Resolving the Production–Expression paradox will likely require moving beyond simple homopolymeric tails toward segmented and, in some cases, more structurally complex architectures (Trepotec et al., 2019). However, any such shift must be evaluated in the context of manufacturability, analytical controllability, and therapeutic use case. To improve translational feasibility, these design changes may need to be accompanied by a more adaptable Chemistry, Manufacturing, and Controls (CMC) framework (Qin et al., 2022; Pardi et al., 2018). Upstream, this includes layered control of plasmid template integrity before transcription. Downstream, it supports a tiered analytical strategy in which capillary gel electrophoresis (CGE) is used for routine distribution analysis and next-generation sequencing (NGS) or related high-resolution methods are applied for deeper structural characterization when needed. Rather than serving as an absolute guarantee of transcript integrity, this framework is better viewed as a practical basis for process understanding, comparability, and control within a QbD-oriented workflow ().
In silico modeling and AI-assisted design
Given the growing complexity of tail architectures and the associated CMC constraints, empirical optimization alone may become increasingly inefficient. In this context, artificial intelligence (AI) and computational biophysics are emerging not as absolute replacements for experimental validation, but as powerful complementary tools for rational design. Beyond optimizing coding sequences for translation and RNA structure (Zhang et al., 2023; Wayment-Steele et al., 2021), future predictive frameworks may also need to model the 3′UTR and poly(A) region as functionally coupled design variables, particularly when sequence context influences stability, immunogenicity, or translational performance (Wayment-Steele et al., 2022; ; Mo et al., 2025). Likewise, molecular simulations may help evaluate how RNA topology influences LNP formulation behavior before experimental testing, although such approaches are still developing and should be interpreted alongside empirical validation ().
A fit-for-purpose framework for clinical deployment
Ultimately, CMC strategy and computational design should support clinical performance rather than operate as stand-alone optimization goals. From this perspective, it is unlikely that a single poly(A) architecture will be optimal across all therapeutic settings (Qin et al., 2022). A more useful framework may be to align tail design with the intended pharmacodynamic window and treatment context (Pardi et al., 2018; Verbeke et al., 2022). For example, prophylactic vaccines—which typically necessitate acute antigen bursts—may benefit from architectures (such as segmented tails) that balance strong short-term expression with highly scalable manufacturability, whereas other applications may prioritize duration, controllability, or reduced immunostimulatory burden (Pardi et al., 2018). In precision genetic interventions such as CRISPR- or CAR-T-related workflows, prolonged RNA persistence may introduce additional safety or off-target considerations, although the magnitude of such risk will depend on construct design, dosing, and delivery context (; ). By contrast, chronic systemic therapies may, in some cases, justify the additional complexity of more durable or regenerative tail architectures if their functional benefits clearly outweigh manufacturing and formulation costs ().
Concluding perspective
Taken together, current evidence supports the view that poly(A) design should be treated as an adjustable engineering parameter rather than a fixed default feature. Integrating CMC considerations, higher-resolution analytics, emerging computational design tools, and therapeutic context may improve how poly(A) architectures are selected and evaluated across mRNA platforms. In that sense, the poly(A) tail is best understood not as a passive appendage, but as a highly engineerable control module that fundamentally bridges expression kinetics, manufacturing robustness, and therapeutic safety in next-generation RNA medicines (; Pardi et al., 2018; ).
Statements
Author contributions
Y-YZ: Writing – review and editing, Writing – original draft, Visualization. J-PZ: Writing – review and editing, Supervision, Funding acquisition, Conceptualization. X-BZ: Writing – review and editing, Conceptualization, Funding acquisition, Supervision.
Funding
The author(s) declared that financial support was received for this work and/or its publication. This work was supported by the National Key Research and Development Program of China (Grant Nos. 2019YFA0110802 and 2019YFA0110803), the National Natural Science Foundation of China (Grant Nos. 82570286, 82402188, 81870149, 82070115, 81890990, and 81730006).
Acknowledgments
We thank members of the XBlab for discussions and feedback during manuscript preparation.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1
AbeN.MatsumotoK.NishiharaM.NakanoY.ShibataA.MaruyamaH.et al (2015). Rolling circle translation of circular RNA in living human cells. Sci. Rep.5, 16435. 10.1038/srep16435
2
AbrahamJ. (2010). “International conference on harmonisation of technical requirements for registration of Pharmaceuticals for human use,” in Handbook of transnational economic governance regimes. Editors TietjeC.BrouderA. (Leiden, Netherlands, Boston, MA: Brill | Nijhoff), 1041–1053.
3
Al-AllafF. A.TolmachovO. E.ZambettiL. P.TchetchelnitskiV.MehmetH. (2013). Remarkable stability of an instability-prone lentiviral vector plasmid in Escherichia coli Stbl3. 3 Biotech.3 (1), 61–70. 10.1007/s13205-012-0070-8
4
AlexopoulouL.HoltA. C.MedzhitovR.FlavellR. A. (2001). Recognition of double-stranded RNA and activation of NF-κB by Toll-like receptor 3. Nature413, 732–738. 10.1038/35099560
5
AnK.KurekD.MahadeoM.ZhangY.ThewaltJ. L.CullisP. R.et al (2022). On the influence of nucleic acid backbone modifications on lipid nanoparticle morphology. Langmuir38, 14036–14043. 10.1021/acs.langmuir.2c01492
6
ArslanS.GarciaF. J.GuoM.KellingerM. W.KruglyakS.LeVieuxJ. A.et al (2024). Sequencing by avidity enables high accuracy with low reagent consumption. Nat. Biotechnol.42, 132–138. 10.1038/s41587-023-01750-7
7
BadenL. R.El SahlyH. M.EssinkB.KotloffK.FreyS.NovakR.et al (2021). Efficacy and safety of the mRNA-1273 SARS-CoV-2 vaccine. N. Engl. J. Med.384, 403–416. 10.1056/nejmoa2035389
8
BaerB. W.KornbergR. D. (1980). Repeating structure of cytoplasmic poly(A)-ribonucleoprotein. Proc. Natl. Acad. Sci. U. S. A.77, 1890–1892. 10.1073/pnas.77.4.1890
9
BaiersdörferM.BorosG.MuramatsuH.MahinyA.VlatkovicI.SahinU.et al (2019). A facile method for the removal of dsRNA contaminant from in vitro-Transcribed mRNA. Mol. Ther. Nucleic Acids15, 26–35. 10.1016/j.omtn.2019.02.018
10
BartelD. P. (2018). Metazoan MicroRNAs. Cell.173, 20–51. 10.1016/j.cell.2018.03.006
11
BegikO.DiensthuberG.LiuH.Delgado-TejedorA.KonturC.NiaziA. M.et al (2023). Nano3P-seq: transcriptome-wide analysis of gene expression and tail dynamics using end-capture nanopore cDNA sequencing. Nat. Methods20, 75–85. 10.1038/s41592-022-01714-w
12
BiziaevN.ShuvalovA.SalmanA.EgorovaT.ShuvalovaE.AlkalaevaE. (2024). The impact of mRNA poly(A) tail length on eukaryotic translation stages. Nucleic Acids Res.52, 7792–7808. 10.1093/nar/gkae510
13
BzymekM.LovettS. T. (2001). Instability of repetitive DNA sequences: the role of replication in multiple mechanisms. Proc. Natl. Acad. Sci. U. S. A.98, 8319–8325. 10.1073/pnas.111008398
14
CaiX.LiJ. J.LiuT.BrianO.LiJ. (2021). Infectious disease mRNA vaccines and a review on epitope prediction for vaccine design. Brief. Funct. Genomics20, 289–303. 10.1093/bfgp/elab027
15
CarrascoM. J.AlishettyS.AlamehM.-G.SaidH.WrightL.PaigeM.et al (2021). Ionization and structural properties of mRNA lipid nanoparticles influence expression in intramuscular and intravascular administration. Commun. Biol.4, 956. 10.1038/s42003-021-02441-2
16
Castillo-HairS.FedakS.WangB.LinderJ.HavensK.CertoM.et al (2024). Optimizing 5’UTRs for mRNA-delivered gene editing using deep learning. Nat. Commun.15 (1), 5284. 10.1038/s41467-024-49508-2
17
ChangH.LimJ.HaM.KimV. N. (2014). TAIL-seq: genome-wide determination of poly(A) tail length and 3′ end modifications. Mol. Cell.53, 1044–1052. 10.1016/j.molcel.2014.02.007
18
ChenC. Y.SarnowP. (1995). Initiation of protein synthesis by the eukaryotic translational apparatus on circular RNAs. Science268, 415–417. 10.1126/science.7536344
19
ChenR.WangS. K.BelkJ. A.AmayaL.LiZ.CardenasA.et al (2023). Engineering circular RNA for enhanced protein production. Nat. Biotechnol.41, 262–272. 10.1038/s41587-022-01393-0
20
ChenH.LiuD.GuoJ.AdithamA.ZhouY.TianJ.et al (2025). Branched chemically modified poly(A) tails enhance the translation capacity of mRNA. Nat. Biotechnol.43, 194–203. 10.1038/s41587-024-02174-7
21
ClarkN. E.KozarskiM.AsciS. D.HeuvelJ.V. d.SchrautM. R.WintersR. A.et al (2025). Removal of dsRNA byproducts using affinity chromatography. Mol. Ther. Nucleic Acids36, 102549. 10.1016/j.omtn.2025.102549
22
CollartM. A. (2016). The Ccr4-Not complex is a key regulator of eukaryotic gene expression. Wiley Interdiscip. Rev. RNA7, 438–454. 10.1002/wrna.1332
23
CollartM. A.PanasenkoO. O. (2012). The ccr4–not complex. Gene492, 42–53. 10.1016/j.gene.2011.09.033
24
CollerJ.ParkerR. (2005). General translational repression by activators of mRNA decapping. Cell.122, 875–886. 10.1016/j.cell.2005.07.012
25
CorbettK. S.NasonM. C.FlachB.GagneM.O’ConnellS.JohnstonT. S.et al (2021). Immune correlates of protection by mRNA-1273 vaccine against SARS-CoV-2 in nonhuman primates. Science373, eabj0299. 10.1126/science.abj0299
26
CrookeS. T.BakerB. F.CrookeR. M.LiangX.-H. (2021). Antisense technology: an overview and prospectus. Nat. Rev. Drug Discov.20, 427–453. 10.1038/s41573-021-00162-z
27
CullisP. R.HopeM. J. (2017). Lipid nanoparticle systems for enabling gene therapies. Mol. Ther.25, 1467–1475. 10.1016/j.ymthe.2017.03.013
28
DeverT. E.GreenR. (2012). The elongation, termination, and recycling phases of translation in eukaryotes. Cold Spring Harb. Perspect. Biol.4, a013706. 10.1101/cshperspect.a013706
29
Di GrandiD.DayehD. M.KaurK.ChenY.HendersonS.MoonY.et al (2023). A single-nucleotide resolution capillary gel electrophoresis workflow for poly(A) tail characterization in the development of mRNA therapeutics and vaccines. J. Pharm. Biomed. Anal.236, 115692. 10.1016/j.jpba.2023.115692
30
DieboldS. S.KaishoT.HemmiH.AkiraS.Reis E SousaC. (2004). Innate antiviral responses by means of TLR7-mediated recognition of single-stranded RNA. Science303, 1529–1531. 10.1126/science.1093616
31
EdmondsM.VaughanM. H.NakazatoH. (1971). Polyadenylic acid sequences in the heterogeneous nuclear RNA and rapidly-labeled polyribosomal RNA of HeLa cells: possible evidence for a precursor relationship. Proc. Natl. Acad. Sci. U. S. A.68, 1336–1340. 10.1073/pnas.68.6.1336
32
EnukaY.LauriolaM.FeldmanM. E.Sas-ChenA.UlitskyI.YardenY. (2016). Circular RNAs are long-lived and display only minimal early alterations in response to a growth factor. Nucleic Acids Res.44, 1370–1383. 10.1093/nar/gkv1367
33
ErasmusJ. H.KhandharA. P.O’ConnorM. A.WallsA. C.HemannE. A.MurapaP.et al (2020). An Alphavirus-derived replicon RNA vaccine induces SARS-CoV-2 neutralizing antibody and T cell responses in mice and nonhuman primates. Sci. Transl. Med.12, eabc9396. 10.1126/scitranslmed.abc9396
34
European Medicines Agency (2025). Guideline on the quality aspects of mRNA vaccines (Draft). Available online at: https://www.ema.europa.eu/en/documents/scientific-guideline/draft-guideline-quality-aspects-mrna-vaccines_en.pdf (Accessed February 27, 2026).
35
European Medicines Agency (EMA) (2021). Comirnaty: EPAR - Public assessment report. EMA/707383/2020 Corr.2.
36
EwaishaR.AndersonK. S. (2023). Immunogenicity of CRISPR therapeutics—critical considerations for clinical translation. Front. Bioeng. Biotechnol.11, 1138596. 10.3389/fbioe.2023.1138596
37
FieldC. M.SummersD. K. (2011). Multicopy plasmid stability: revisiting the dimer catastrophe. J. Theor. Biol.291, 119–127. 10.1016/j.jtbi.2011.09.006
38
FrazierK. S. (2015). Antisense oligonucleotide therapies: the promise and the challenges from a toxicologic pathologist’s perspective. Toxicol. Pathol.43, 78–89. 10.1177/0192623314551840
39
FriehsK. (2004). “Plasmid copy number and plasmid stability,”. New trends and developments in biochemical engineering. Editor ScheperT., 86, 47–82. 10.1007/b12440
40
GallieD. R. (1991). The cap and poly(A) tail function synergistically to regulate mRNA translational efficiency. Genes Dev.5, 2108–2116. 10.1101/gad.5.11.2108
41
GarneauN. L.WiluszJ.WiluszC. J. (2007). The highways and byways of mRNA decay. Nat. Rev. Mol. Cell. Biol.8, 113–126. 10.1038/nrm2104
42
GearyR. S.NorrisD.YuR.BennettC. F. (2015). Pharmacokinetics, biodistribution and cell uptake of antisense oligonucleotides. Adv. Drug Deliv. Rev.87, 46–51. 10.1016/j.addr.2015.01.008
43
GholamalipourY.Karunanayake MudiyanselageA.MartinC. T. (2018). 3′ end additions by T7 RNA polymerase are RNA self-templated, distributive and diverse in character—RNA-Seq analyses. Nucleic Acids Res.46, 9253–9263. 10.1093/nar/gky796
44
GrierA. E.BurleighS.SahniJ.CloughC. A.CardotV.ChoeD. C.et al (2016). pEVL: a linear plasmid for generating mRNA IVT templates with extended encoded poly(A) sequences. Mol. Ther. Nucleic Acids5, e306. 10.1038/mtna.2016.21
45
GTEx Consortium (2020). The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science369, 1318–1330. 10.1126/science.aaz1776
46
GuoC.MaX.GaoF.GuoY. (2023). Off-target effects in CRISPR/Cas9 gene editing. Front. Bioeng. Biotechnol.11, 1143157. 10.3389/fbioe.2023.1143157
47
HeilF.HemmiH.HochreinH.AmpenbergerF.KirschningC.AkiraS.et al (2004). Species-specific recognition of single-stranded RNA via toll-like receptor 7 and 8. Science303, 1526–1529. 10.1126/science.1093620
48
HellgrenF.RosdahlA.CerveiraR. A.LenartK.OlsS.GwonY.-D.et al (2024). Modulation of innate immune response to mRNA vaccination after SARS-CoV-2 infection or sequential vaccination in humans. JCI Insight9, e175401. 10.1172/jci.insight.175401
49
HoltkampS.KreiterS.SelmiA.SimonP.KoslowskiM.HuberC.et al (2006). Modification of antigen-encoding RNA increases stability, translational efficacy, and T-cell stimulatory capacity of dendritic cells. Blood108, 4009–4017. 10.1182/blood-2006-04-015024
50
HornungV.EllegastJ.KimS.BrzózkaK.JungA.KatoH.et al (2006). 5′-Triphosphate RNA is the ligand for RIG-I. Science314, 994–997. 10.1126/science.1132505
51
HouX.ZaksT.LangerR.DongY. (2021). Lipid nanoparticles for mRNA delivery. Nat. Rev. Mater6, 1078–1094. 10.1038/s41578-021-00358-0
52
HyrinaA.JonesC.ChenD.ClarksonS.CochranN.FeuchtP.et al (2019). A genome-wide CRISPR Screen identifies ZCCHC14 as a host factor required for hepatitis B surface antigen production. Cell. Rep.29, 2970–2978.e6. 10.1016/j.celrep.2019.10.113
53
International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) (2009). Pharmaceutical development Q8(R2) (current step 4 version, august 2009). ICH. Available online at: https://database.ich.org/sites/default/files/Q8_R2_Guideline.pdf (Accessed May 8, 2026).
54
JacksonR. J.HellenC. U. T.PestovaT. V. (2010). The mechanism of eukaryotic translation initiation and principles of its regulation. Nat. Rev. Mol. Cell. Biol.11, 113–127. 10.1038/nrm2838
55
JiangM.RongM.MartinC.McAllisterW. T. (2001). Interrupting the template strand of the T7 promoter facilitates translocation of the DNA during initiation, reducing transcript slippage and the release of abortive products. J. Mol. Biol.310, 509–522. 10.1006/jmbi.2001.4793
56
JungS.-J.SeoJ. J.LeeS.HyunS.-I.LeeJ.-E.LeeS.et al (2025). RNA stability enhancers for durable base-modified mRNA therapeutics. Nat. Biotechnol.10.1038/s41587-025-02891-7
57
JuszkiewiczS.HegdeR. S. (2017). Initiation of quality control during poly(A) translation requires site-specific ribosome ubiquitination. Mol. Cell.65, 743–750.e4. 10.1016/j.molcel.2016.11.039
58
KahvejianA.RoyG.SonenbergN. (2001). The mRNA closed-loop model: the function of PABP and PABP-interacting proteins in mRNA translation. Cold Spring Harb. Symp. Quant. Biol.66, 293–300. 10.1101/sqb.2001.66.293
59
KarikóK.BucksteinM.NiH.WeissmanD. (2005). Suppression of RNA recognition by Toll-like receptors: the impact of nucleoside modification and the evolutionary origin of RNA. Immunity23, 165–175. 10.1016/j.immuni.2005.06.008
60
KarikóK.MuramatsuH.LudwigJ.WeissmanD. (2011). Generating the optimal mRNA for therapy: HPLC purification eliminates immune activation and improves translation of nucleoside-modified, protein-encoding mRNA. Nucleic Acids Res.39, e142. 10.1093/nar/gkr695
61
KatoH.TakeuchiO.SatoS.YoneyamaM.YamamotoM.MatsuiK.et al (2006). Differential roles of MDA5 and RIG-I helicases in the recognition of RNA viruses. Nature441, 101–105. 10.1038/nature04734
62
KawaiT.AkiraS. (2010). The role of pattern-recognition receptors in innate immunity: update on Toll-like receptors. Nat. Immunol.11, 373–384. 10.1038/ni.1863
63
KhanD.TerenziF.LiuG.GhoshP. K.YeF.NguyenK.et al (2023). A viral pan-end RNA element and host complex define a SARS-CoV-2 regulon. Nat. Commun.14, 3385. 10.1038/s41467-023-39091-3
64
KimD.LeeY.JungS.-J.YeoJ.SeoJ. J.LeeY.-Y.et al (2020). Viral hijacking of the TENT4–ZCCHC14 complex protects viral RNAs via mixed tailing. Nat. Struct. Mol. Biol.27, 581–588. 10.1038/s41594-020-0427-3
65
KimY.-A.MousaviK.YazdiA.ZwierzynaM.CardinaliM.FoxD.et al (2024). Computational design of mRNA vaccines. Vaccine42, 1831–1840. 10.1016/j.vaccine.2023.07.024
66
KrauseM.NiaziA. M.LabunK.Torres CleurenY. N.MüllerF. S.ValenE. (2019). tailfindr: alignment-free poly(A) length measurement for Oxford Nanopore RNA and DNA sequencing. RNA25, 1229–1241. 10.1261/rna.071332.119
67
KuzmineI.MartinC. T. (2001). Structure in nascent RNA leads to termination of slippage transcription by T7 RNA polymerase. Nucleic Acids Res.29, 2601–2606. 10.1093/nar/29.12.2601
68
LeconteA. M.PatelM. P.SassL. E.McInerneyP.JaroszM.KungL.et al (2010). Directed evolution of DNA polymerases for next-generation sequencing. Angew. Chem. Int. Ed. Engl.49, 5921–5924. 10.1002/anie.201001607
69
LeeY.LevdanskyY.JungY.KimV. N.ValkovE. (2024). Deadenylation kinetics of mixed poly(A) tails at single-nucleotide resolution. Nat. Struct. Mol. Biol.31, 826–834. 10.1038/s41594-023-01187-1
70
LegniniI.Di TimoteoG.RossiF.MorlandoM.BrigantiF.SthandierO.et al (2017). Circ-ZNF609 is a circular RNA that can Be translated and functions in myogenesis. Mol. Cell.66, 22–37.e9. 10.1016/j.molcel.2017.02.017
71
LegniniI.AllesJ.KaraiskosN.AyoubS.RajewskyN. (2019). FLAM-seq: full-length mRNA sequencing reveals principles of poly(A) tail length control. Nat. Methods16, 879–886. 10.1038/s41592-019-0503-y
72
LemdaniK.MarlinR.MayetC.PerkovV.PascalQ.RipollM.et al (2024). Distinct dynamics of mRNA LNPs in mice and nonhuman primates revealed by in vivo imaging. NPJ Vaccines9, 113. 10.1038/s41541-024-00900-5
73
LiC. Y.LiangZ.HuY.ZhangH.SetiasabdaK. D.LiJ.et al (2022a). Cytidine-containing tails robustly enhance and prolong protein production of synthetic mRNA in cell and in vivo. Mol. Ther. Nucleic Acids30, 300–310. 10.1016/j.omtn.2022.10.003
74
LiS.HuY.LiA.LinJ.HsiehK.SchneidermanZ.et al (2022b). Payload distribution and capacity of mRNA lipid nanoparticles. Nat. Commun.13, 5561. 10.1038/s41467-022-33157-4
75
LimJ.HaM.ChangH.KwonS. C.SimanshuD. K.PatelD. J.et al (2014). Uridylation by TUT4 and TUT7 marks mRNA for degradation. Cell.159, 1365–1376. 10.1016/j.cell.2014.10.055
76
LimJ.KimD.LeeY.-S.HaM.LeeM.YeoJ.et al (2018). Mixed tailing by TENT4A and TENT4B shields mRNA from rapid deadenylation. Science361, 701–704. 10.1126/science.aam5794
77
Linares-FernándezS.LacroixC.ExpositoJ.-Y.VerrierB. (2020). Tailoring mRNA vaccine to balance innate/adaptive immune response. Trends Mol. Med.26, 311–323. 10.1016/j.molmed.2019.10.002
78
LuoL.JeaJ.D.-Y.WangY.ChaoP.-W.YenL. (2024). Control of mammalian gene expression by modulation of polyA signal cleavage at 5′ UTR. Nat. Biotechnol.42, 1454–1466. 10.1038/s41587-023-01989-0
79
MaX.-K.WangM.-R.LiuC.-X.DongR.CarmichaelG. G.ChenL.-L.et al (2019). circexplorer3: a clear pipeline for direct comparison of circular and linear rna expression. Genomics Proteomics Bioinforma.17, 511–521. 10.1016/j.gpb.2019.11.004
80
MacdonaldL. E.ZhouY.McAllisterW. T. (1993). Termination and slippage by bacteriophage T7 RNA polymerase. J. Mol. Biol.232, 1030–1047. 10.1006/jmbi.1993.1458
81
MiaoX.HuG. (2025). Poly(A) tail regulation in stem cells and early development. Cell. Regen.14, 48. 10.1186/s13619-025-00250-0
82
MoO.ZhangZ.ChengX.ZhuL.ZhangK.ZhangN.et al (2025). mRNAdesigner: an integrated web server for optimizing mRNA design and protein translation in eukaryotes. Nucleic Acids Res.53, W415–W426. 10.1093/nar/gkaf410
83
MolodtsovV.AnikinM.McAllisterW. T. (2014). The presence of an RNA:DNA hybrid that is prone to slippage promotes termination by T7 RNA polymerase. J. Mol. Biol.426, 3095–3107. 10.1016/j.jmb.2014.06.012
84
MuX.HurS. (2023). Immunogenicity of in vitro-transcribed RNA. Acc. Chem. Res.54 (21), 4012–4023. 10.1021/acs.accounts.1c00521
85
MuX.GreenwaldE.AhmadS.HurS. (2018). An origin of the immunogenicity of in vitro transcribed RNA. Nucleic Acids Res.46, 5239–5249. 10.1093/nar/gky177
86
MunroeD.JacobsonA. (1990). mRNA poly(A) tail, a 3′ enhancer of translational initiation. Mol. Cell. Biol.10, 3441–3455. 10.1128/mcb.10.7.3441
87
NicholsonA. L.PasquinelliA. E. (2019). Tales of detailed poly(A) tails. Trends Cell. Biol.29, 191–200. 10.1016/j.tcb.2018.11.002
88
Nicholson-ShawA. L.KofmanE. R.YeoG. W.PasquinelliA. E. (2022). Nuclear and cytoplasmic poly(A) binding proteins (PABPs) favor distinct transcripts and isoforms. Nucleic Acids Res.50, 4685–4702. 10.1093/nar/gkac263
89
OhA.LeeS.ParkH.-J.YoonS.HaD.LeeJ.et al (2025). Loop structure in poly(A) tail of mRNA vaccine enhances antigen translation efficiency and mRNA stability. NPJ Vaccines10, 234. 10.1038/s41541-025-01287-7
90
PardiN.KrammerF. (2024). mRNA vaccines for infectious diseases—advances, challenges and opportunities. Nat. Rev. Drug Discov.23, 838–861. 10.1038/s41573-024-01042-y
91
PardiN.HoganM. J.PorterF. W.WeissmanD. (2018). mRNA vaccines — a new era in vaccinology. Nat. Rev. Drug Discov.17, 261–279. 10.1038/nrd.2017.243
92
PassmoreL. A.CollerJ. (2022). Roles of mRNA poly(A) tails in regulation of eukaryotic gene expression. Nat. Rev. Mol. Cell. Biol.23, 93–106. 10.1038/s41580-021-00417-y
93
PengJ.SchoenbergD. R. (2005). mRNA with a <20-nt poly(A) tail imparted by the poly(A)-limiting element is translated as efficiently in vivo as long poly(A) mRNA. RNA11, 1131–1140. 10.1261/rna.2470905
94
PiaoX.YadavV.WangE.ChangW.TauL.LindenmuthB. E.et al (2022). Double-stranded RNA reduction by chaotropic agents during in vitro transcription of messenger RNA. Mol. Ther. Nucleic Acids29, 618–624. 10.1016/j.omtn.2022.08.001
95
PichlmairA.SchulzO.TanC. P.NäslundT. I.LiljeströmP.WeberF.et al (2006). RIG-I-mediated antiviral responses to single-stranded RNA bearing 5′-phosphates. Science314, 997–1001. 10.1126/science.1132998
96
PolackF. P.ThomasS. J.KitchinN.AbsalonJ.GurtmanA.LockhartS.et al (2020). Safety and efficacy of the BNT162b2 mRNA covid-19 vaccine. N. Engl. J. Med.383, 2603–2615. 10.1056/NEJMoa2034577
97
PreissT.MuckenthalerM.HentzeM. W. (1998). Poly(A)-tail-promoted translation in yeast: implications for translational control. RNA4, 1321–1331. 10.1017/s1355838298980669
98
QinS.TangX.ChenY.ChenK.FanN.XiaoW.et al (2022). mRNA-based therapeutics: powerful and versatile tools to combat diseases. Signal Transduct. Target Ther.7, 166. 10.1038/s41392-022-01007-w
99
ReA.JoshiT.KulberkyteE.MorrisQ.WorkmanC. T. (2014). RNA-protein interactions: an overview. Methods Mol. Biol.1097, 491–521. 10.1007/978-1-62703-709-9_23
100
RosaS. S.PrazeresD. M. F.AzevedoA. M.MarquesM. P. C. (2021). mRNA vaccines manufacturing: challenges and bottlenecks. Vaccine39, 2190–2200. 10.1016/j.vaccine.2021.03.038
101
SafaeeN.KozlovG.NoronhaA. M.XieJ.WildsC. J.GehringK. (2012). Interdomain allostery promotes assembly of the poly(A) mRNA complex with PABP and eIF4G. Mol. Cell.48, 375–386. 10.1016/j.molcel.2012.09.001
102
SahinU.KarikóK.TüreciÖ. (2014). mRNA-based therapeutics—developing a new class of drugs. Nat. Rev. Drug Discov.13, 759–780. 10.1038/nrd4278
103
SchleeM.RothA.HornungV.HagmannC. A.WimmenauerV.BarchetW.et al (2009). Recognition of 5′ triphosphate by RIG-I helicase requires short blunt double-stranded RNA as contained in panhandle of Negative-strand virus. Immunity31, 25–34. 10.1016/j.immuni.2009.05.008
104
SchoberG. B.StoryS.AryaD. P. (2024). A careful look at lipid nanoparticle characterization: analysis of benchmark formulations for encapsulation of RNA cargo size gradient. Sci. Rep.14 (1), 2403. 10.1038/s41598-024-52685-1
105
ShenY.LiB.DongL.TangW.RenJ.ChenF.et al (2025). Self-splicing RNA circularization facilitated by intact group I and II introns. Nat. Commun.16, 7376. 10.1038/s41467-025-62607-y
106
SpiewlaT.CzubakK.PilchZ.BaranowskiM. R.KrawczykP. S.AffekK.et al (2026). PolyA tail segmentation improves the stability of the template DNA and increases the translatability of in vitro transcribed mRNA. Nucleic Acids Res.54, gkaf1412. 10.1093/nar/gkaf1412
107
StrzeleckaD.SmietanskiM.SikorskiP. J.WarminskiM.KowalskaJ.JemielityJ. (2020). Phosphodiester modifications in mRNA poly(A) tail prevent deadenylation without compromising protein expression. RNA26, 1815–1837. 10.1261/rna.077099.120
108
SubtelnyA. O.EichhornS. W.ChenG. R.SiveH.BartelD. P. (2014). Poly(A)-tail profiling reveals an embryonic switch in translational control. Nature508, 66–71. 10.1038/nature13007
109
SvitkinY. V.EvdokimovaV. M.BraseyA.PestovaT. V.FantusD.YanagiyaA.et al (2009). General RNA-binding proteins have a function in poly(A)-binding protein-dependent translation. EMBO J.28, 58–68. 10.1038/emboj.2008.259
110
TarunS. Z.SachsA. B. (1996). Association of the yeast poly(A) tail binding protein with translation initiation factor eIF-4G. EMBO J.15, 7168–7177. 10.1002/j.1460-2075.1996.tb01108.x
111
ToK. K. W.ChoW. C. S. (2021). An overview of rational design of mRNA-based therapeutics and vaccines. Expert Opin. Drug Discov.16, 1307–1317. 10.1080/17460441.2021.1935859
112
TongM.PalmerN.DailamyA.KumarA.KhaliqH.HanS.et al (2025). Robust genome and cell engineering via in vitro and in situ circularized RNAs. Nat. Biomed. Eng.9, 109–126. 10.1038/s41551-024-01245-z
113
TorkzabanB.ZhuY.LopezC.AlexanderJ. M.MaJ.SunY.et al (2025). Use of polyadenosine tail mimetics to enhance mRNA expression from genes associated with haploinsufficiency disorders. Mol. Ther. Nucleic Acids36, 102453. 10.1016/j.omtn.2025.102453
114
TrepotecZ.GeigerJ.PlankC.AnejaM. K.RudolphC. (2019). Segmented poly(A) tails significantly reduce recombination of plasmid DNA without affecting mRNA translation efficiency or half-life. RNA25, 507–518. 10.1261/rna.069286.118
115
Triana-AlonsoF. J.DabrowskiM.WadzackJ.NierhausK. H. (1995). Self-coded 3′-extension of run-off transcripts produces aberrant products during in vitro transcription with T7 RNA polymerase. J. Biol. Chem.270, 6298–6307. 10.1074/jbc.270.11.6298
116
United States Pharmacopeia (USP) (2023). General chapter 〈1239〉 vaccines for human use—viral vaccines. Rockville, MD: United States Pharmacopeia.
117
U.S. Food and Drug Administration (FDA) (2022). Comp. Protoc. Postapproval Changes Chem. Manuf. Controls Inf. NDA, ANDA, or BLA Guid. Industry.
118
ValenzuelaA.AyusoM.BuyssensL.BarsC.Van GinnekenC.TessierY.et al (2023). Platelet activation by antisense oligonucleotides (ASOs) in the göttingen minipig, including an evaluation of glycoprotein VI (GPVI) and platelet factor 4 (PF4) ontogeny. Pharmaceutics15, 1112. 10.3390/pharmaceutics15041112
119
VerbekeR.HoganM. J.LoréK.PardiN. (2022). Innate immune mechanisms of mRNA vaccines. Immunity55, 1993–2005. 10.1016/j.immuni.2022.10.014
120
VicensQ.KieftJ. S.RisslandO. S. (2018). Revisiting the closed-loop model and the nature of mRNA 5′-3′ communication. Mol. Cell.72, 805–812. 10.1016/j.molcel.2018.10.047
121
VigueraE.CanceillD.EhrlichS. D. (2001). Replication slippage involves DNA polymerase pausing and dissociation. EMBO J.20, 2587–2595. 10.1093/emboj/20.10.2587
122
WahleE.WinklerG. S. (2013). RNA decay machines: deadenylation by the Ccr4–Not and Pan2–Pan3 complexes. Biochim. Biophys. Acta1829, 561–570. 10.1016/j.bbagrm.2013.01.003
123
Wayment-SteeleH. K.KimD. S.ChoeC. A.NicolJ. J.Wellington-OguriR.WatkinsA. M.et al (2021). Theoretical basis for stabilizing messenger RNA through secondary structure design. Nucleic Acids Res.49, 10604–10617. 10.1093/nar/gkab764
124
Wayment-SteeleH. K.KladwangW.WatkinsA. M.KimD. S.TunguzB.ReadeW.et al (2022). Deep learning models for predicting RNA degradation via dual crowdsourcing. Nat. Mach. Intell.4, 1174–1184. 10.1038/s42256-022-00571-8
125
WebbA. L. J.WelbourneE. N.EvansC. A.DickmanM. J. (2025). Characterisation and analysis of mRNA critical quality attributes using liquid chromatography based methods. J. Chromatogr. A1745, 465724. 10.1016/j.chroma.2025.465724
126
WebsterM. W.ChenY.-H.StowellJ. A. W.AlhusainiN.SweetT.GraveleyB. R.et al (2018). mRNA deadenylation is coupled to translation rates by the differential activities of ccr4-not nucleases. Mol. Cell.70, 1089–1100.e8. 10.1016/j.molcel.2018.05.033
127
WellsS. E.HillnerP. E.ValeR. D.SachsA. B. (1998). Circularization of mRNA by eukaryotic translation initiation factors. Mol. Cell.2, 135–140. 10.1016/s1097-2765(00)80122-7
128
WolffJ. A.MaloneR. W.WilliamsP.ChongW.AcsadiG.JaniA.et al (1990). Direct gene transfer into mouse muscle in vivo. Science247, 1465–1468. 10.1126/science.1690918
129
WorkmanR. E.TangA. D.TangP. S.JainM.TysonJ. R.RazaghiR.et al (2019). Nanopore native RNA sequencing of a human poly(A) transcriptome. Nat. Methods16, 1297–1305. 10.1038/s41592-019-0617-2
130
WuM. Z.AsaharaH.TzertzinisG.RoyB. (2020). Synthesis of low immunogenicity RNA with high-temperature in vitro transcription. RNA26, 345–360. 10.1261/rna.073858.119
131
WuH.WeiT.YuB.ChengR.HuangF.LuX.et al (2021). A single mutation attenuates both the transcription termination and RNA-dependent RNA polymerase activity of T7 RNA polymerase. RNA Biol.18 (1), 451–466. 10.1080/15476286.2021.1954808
132
WuK.XuF.DaiY.JinS.ZhengA.ZhangN.et al (2024). Characterization of mRNA-LNP structural features and mechanisms for enhanced mRNA vaccine immunogenicity. J. Control Release376, 1288–1299. 10.1016/j.jconrel.2024.11.007
133
XiangK.BartelD. P. (2021). The molecular basis of coupling between poly(A)-tail length and translational efficiency. eLife10, e66493. 10.7554/eLife.66493
134
YamashitaA.ChangT.-C.YamashitaY.ZhuW.ZhongZ.ChenC.-Y. A.et al (2005). Concerted action of poly(A) nucleases and decapping enzyme in mammalian mRNA turnover. Nat. Struct. Mol. Biol.12, 1054–1063. 10.1038/nsmb1016
135
ZhangH.ZhangL.LinA.XuC.LiZ.LiuK.et al (2023). Algorithm for optimized mRNA design improves stability and immunogenicity. Nature621, 396–403. 10.1038/s41586-023-06127-z
136
ZiegenhalsT.FrielingR.WolfP.GöbelK.KochS.LohmannM.et al (2023). Formation of dsRNA by-products during in vitro transcription can be reduced by using low steady-state levels of UTP. Front. Mol. Biosci.10, 1291045. 10.3389/fmolb.2023.1291045
Summary
Keywords
dsRNA impurities, in vitro transcription, innate immunity, manufacturability, mRNA therapeutics, poly(A) tail, quality by design, sequence engineering
Citation
Zhang Y-Y, Zhang J-P and Zhang X-B (2026) Engineering the poly(A) tail for therapeutic mRNA: from expression control to manufacturing robustness. Front. Bioeng. Biotechnol. 14:1838589. doi: 10.3389/fbioe.2026.1838589
Received
25 March 2026
Revised
19 April 2026
Accepted
23 April 2026
Published
13 May 2026
Volume
14 - 2026
Edited by
Andrea Caporali, University of Edinburgh, United Kingdom
Reviewed by
Sarbjeet Makkar, University of Michigan, MI, United States
Dinglingge Cao, Icahn School of Medicine at Mount Sinai, United States
Updates
Copyright
© 2026 Zhang, Zhang and Zhang.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Jian-Ping Zhang, zhangjianping@ihcams.ac.cn; Xiao-Bing Zhang, zhangxbhk@gmail.com
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.