COVID-19 mRNA Vaccines: From New Technology to Medical Use

First-generation COVID-19 mRNA vaccines showed outstanding early performance against the ancestral virus and early variants. The pivotal BNT162b2 and mRNA-1273 trials established high efficacy against symptomatic COVID-19 in the pre-Omicron setting (1, 2), and subsequent observational studies during the Alpha and Delta periods showed high protection against symptomatic disease, hospitalization, and death, supporting mRNA vaccination as a central public-health intervention during the acute pandemic phase (12, 16–19).

That performance did not remain static. With the emergence of Omicron and its sublineages, vaccine effectiveness against infection and mild disease declined substantially because of both immune escape and waning antibody titers. However, protection against hospitalization and death remained substantially more durable, especially after booster doses and updated formulations targeting Omicron-descended lineages such as XBB.1.5 and JN.1 (9, 12–15, 20–24). This distinction between protection against infection and protection against severe outcomes should be stated explicitly, because treating these endpoints as interchangeable has contributed to confusion in parts of the literature.

Moreover, effectiveness estimates should be interpreted with caution because they are shaped by prior infection, hybrid immunity, testing behavior, age, risk profile, and study design. Observational estimates can therefore vary meaningfully across populations and time periods even when the biologic performance of the vaccine platform is broadly consistent (25–30). To avoid treating VE as a single fixed number, Table 1 summarizes representative effectiveness estimates by variant period, vaccine product or regimen, clinical endpoint, and main interpretation.

Post-authorization surveillance supports a favorable overall safety profile for COVID-19 mRNA vaccines. Most adverse events are mild and self-limited, including injection-site pain, fatigue, headache, myalgia, and transient fever, and these reactions are consistent with short-lived innate immune activation after vaccination (31–34). Rare serious adverse events, particularly myocarditis and pericarditis, have been reported after mRNA vaccination and require risk-stratified communication and continued follow-up; population-level studies indicate that their overall prognosis is generally better than that of infection-associated cardiac inflammation (33–38).

From a public-health perspective, the benefits were substantial. Modeling studies and registry-based analyses indicate that COVID-19 vaccination prevented large numbers of deaths globally and remains most cost-effective in older adults and other high-risk groups, supporting a risk-stratified booster strategy rather than a one-size-fits-all approach for all age groups (39–42).

The main limitations revealed by clinical use are now clear: protection against infection wanes faster than protection against severe disease; variant evolution can outpace the breadth of neutralizing antibodies elicited by prior formulations; repeated boosting is operationally difficult at population scale; and intramuscular administration does not reliably establish durable sterilizing immunity in the upper airway (9–13). These limitations set the agenda for the next phase of RNA-vaccine development and provide the rationale for the following sections, which examine the comparative platform advantages of mRNA during the pandemic and how RNA engineering, delivery innovation, immune shaping, and next-generation RNA formats may address the remaining gaps.

Although no single immune marker fully predicts protection across all COVID-19 endpoints, neutralizing antibody levels remain the best-supported correlate or surrogate marker for short-term protection against infection and symptomatic disease at the population level (9, 47–51). The coordinated immune mechanisms underlying both early protection and persistent severe-disease protection are summarized in Figure 1. After primary immunization or boosting, antibody titers rise quickly and correlate with short-term protection against symptomatic infection, but they also decline over time and are strongly affected by spike mutations in antigenically distant variants (9, 47–50). This is why breakthrough infection became increasingly common after Omicron despite continued vaccination.

However, reduced protection against infection does not imply immunologic failure. Repeated vaccination and/or breakthrough infection continue to refine germinal-center and memory-B-cell responses, increasing clonal continuity, affinity maturation, and recognition of conserved epitopes across variants (52–55). When neutralization is incomplete after variant challenge, conserved-epitope T-cell responses and Fc-mediated non-neutralizing antibody functions can still contribute to infected-cell clearance and disease attenuation (56–60). Together, these mechanisms help explain why protection against hospitalization and death has remained more stable than protection against infection.

Thus, one of the main lessons from COVID-19 is that correlates of protection are endpoint-specific: neutralizing antibodies are most informative for infection risk, whereas memory B cells, T cells, and non-neutralizing effector mechanisms are central to sustained protection against severe disease. This framework is well supported, even though the quantitative thresholds needed for different endpoints remain under active study (9, 47, 59).

Immune imprinting has emerged as an important consideration for repeated SARS-CoV-2 vaccination, but its implications are more nuanced than a simple loss-of-function model. Historically, the related concept of original antigenic sin was introduced by Francis in 1960 to describe the tendency of prior antigenic exposure to shape responses to later variant antigens (61, 62). In the COVID-19 setting, prior infection or vaccination can bias recall toward previously seen epitopes, especially within conserved regions of spike, which may reduce the de novo recruitment of responses to newly emerged variant-specific epitopes (61, 63–65).

At the same time, imprinting is not uniformly detrimental. Cross-reactive memory can also serve as a useful scaffold for rapid recall, affinity maturation, and the generation of broader antibodies after boosting or breakthrough infection (66, 67). The key unresolved question is therefore not whether imprinting exists—it clearly does—but under what antigenic distances, dosing intervals, and exposure histories it becomes mainly constraining rather than beneficial. Variant-updated boosters can partly redirect the response, but they do not erase immunologic history (52, 65, 67).

Repeated exposure to spike antigen through mRNA vaccination has been associated in some cohorts with an increase in spike-specific IgG4. This observation is reproducible at the serological level, especially after repeated antigen exposure, but its clinical meaning remains unsettled (68–70). It would therefore be premature to frame IgG4 switching as either clearly harmful or clearly beneficial.

On one hand, IgG4 has weaker canonical pro-inflammatory Fc functions than IgG1, including reduced Fcγ receptor and complement-mediated effector activation (71), which has led to concern that high IgG4 levels might reduce certain effector functions important for viral clearance. On the other hand, the consequences of subclass redistribution depend on the broader antibody repertoire, total antibody titer, epitope specificity, Fc-receptor context, and the infecting variant. Current evidence supports treating IgG4 as an informative immunologic signature of repeated exposure rather than as an established mechanism of reduced vaccine effectiveness (68–70).

Intramuscular COVID-19 mRNA vaccines are highly effective at inducing systemic immunity but comparatively limited at generating durable upper-airway mucosal immunity. This mismatch helps explain why post-booster protection against severe disease can remain strong while infection and onward transmission are only incompletely prevented (10, 11, 72).

Intranasal and inhaled approaches are being explored to address this gap. Animal studies and early clinical data suggest that mucosal boosting can induce secretory IgA, lung- or airway-resident memory T cells, and more localized antiviral responses, especially in heterologous prime-boost settings (73–85). Nevertheless, this field is still evolving. It is not yet established that currently available mucosal strategies can reproducibly provide durable sterilizing immunity in humans, and formulation, delivery, and safety trade-offs remain major translational challenges (10, 75, 80–82). Overall, repeated COVID-19 mRNA vaccination has clarified that protection is endpoint-specific: neutralizing antibodies are central to short-term protection against infection, whereas memory B cells, T cells, and non-neutralizing effector mechanisms help sustain protection against severe disease.

The success of first-generation COVID-19 mRNA vaccines rests on precise engineering of both the RNA molecule and the encoded antigen, as illustrated in Figure 2. At the RNA level, nucleoside modification, cap optimization, untranslated-region design, and poly(A)-tail engineering improve translation efficiency, limit undesirable innate sensing, and increase stability. The foundational study by Karikó and colleagues showed that modified nucleosides such as pseudouridine suppress Toll-like receptor-mediated recognition of exogenous RNA, a finding that helped enable modern nucleoside-modified mRNA vaccines (86, 87). Subsequent work has further refined cap, UTR, and tail design, and has shown that host factors such as TENT5A can enhance vaccine performance by re-adenylating vaccine mRNA after delivery (4, 88).

At the antigen-design level, structural biology rapidly translated into vaccine engineering. The introduction of the prefusion-stabilizing 2P mutation locked spike in an immunogenically favorable conformation, and the later HexaPro/6P design added further proline substitutions to improve stability and expression, as shown schematically in Figure 2 (89, 90). Newer work suggests that in some contexts non-stabilized spike antigens may broaden epitope targeting and promote antibodies to conserved regions, emphasizing that maximally stable antigen design is not always synonymous with maximally broad immunity (91).

LNPs remain the clinically dominant delivery system for mRNA vaccines because they solve the two central delivery problems of RNA therapeutics: nuclease instability and inefficient cytosolic entry. Their four-component architecture—ionizable lipid, phospholipid, cholesterol, and PEG-lipid—supports RNA encapsulation, tissue delivery, and endosomal escape (92–100). Figure 3 depicts this delivery process from LNP composition and cellular uptake to acidic endosomal escape, cytosolic mRNA release, antigen translation, and immune activation. Ionizable lipids exhibit pH-responsive behavior: they remain largely neutral at physiological pH but become protonated in acidic endosomes, thereby promoting endosomal membrane destabilization and cytosolic release of mRNA (95–98).

Current work is focused on improving the balance between potency, biodistribution, tolerability, and manufacturability. Representative strategies include biodegradable ionizable lipids, AI-guided lipid design, PEG alternatives, and LNP structural tuning to favor more localized antigen expression (96–101). At the same time, next-generation delivery formats such as hydrogels, microneedle patches, and nanoemulsions are being explored to extend antigen exposure, reduce cold-chain dependence, simplify administration, or better target antigen-presenting tissues (102–109). These approaches are not intended simply as replacements for LNPs; rather, as summarized in Table 2 and illustrated in the right panel of Figure 3, they address specific delivery bottlenecks such as sustained antigen exposure, skin/APC targeting, administration convenience, and mucosal or immune-programming potential.

Although mRNA-LNP formulations already possess intrinsic adjuvant activity, next-generation design is moving from generic self-adjuvancy toward more deliberate immune shaping. Co-delivery of cytokine-encoding RNAs such as IL-12 or IL-7 can enhance cytotoxic T-cell responses or support longer-lasting humoral responses, while integration of innate agonists such as STING-activating materials may improve lymphatic delivery and T-cell priming (110–112). The conceptual shift is important: the aim is no longer only to generate a strong response, but to generate the right response for the desired clinical endpoint.

Clinical translation of this concept is already visible in local tumor immunotherapy. Intratumoral MEDI1191, an mRNA encoding IL-12, has been evaluated with durvalumab in patients with advanced solid tumors (113), and mRNA-2752 delivers mRNAs encoding OX40L, IL-23, and IL-36γ for intratumoral administration either alone or in combination with durvalumab (114). More recently, STX-001 extended this approach to an LNP-encapsulated self-replicating mRNA encoding IL-12 for intratumoral administration (115). These examples illustrate that immune-shaping RNA is not limited to vaccine antigen delivery, but can also be used to remodel the tumor microenvironment through transient local cytokine expression.

Next-generation RNA platforms should be discussed according to developmental maturity. Figure 4 and Table 3 together summarize the structural and translational differences among conventional mRNA, saRNA, and circRNA platforms. Self-amplifying RNA (saRNA) is currently the more clinically advanced modality. By encoding both the target immunogen and a viral replicase complex, saRNA amplifies intracellular RNA copies and can achieve strong immunogenicity at substantially lower doses than conventional mRNA. Clinical studies of ARCT-154 showed favorable safety, immunogenicity, and efficacy, and follow-up analyses reported durable immune responses up to 12 months in comparison with BNT162b2 (116–118). Importantly, ARCT-154/Kostaive has already received regulatory approval in Japan and the European Union, making saRNA a clinically realized platform rather than an early-stage concept (119, 120).

Additional examples reinforce this point. GEMCOVAC-OM, an Omicron-specific thermostable self-amplifying mRNA booster delivered intradermally, received emergency authorization in India and showed favorable phase 2/3 immunogenicity and safety data (121). Other programs, such as HDT-301, illustrate that platform diversity extends beyond a single product lineage and includes clinical-stage repRNA candidates with distinct formulation and backbone designs (122). Clinical phase 1 studies of optimized self-replicating RNA platforms have also shown that very low-dose or single-dose vaccination can be immunogenic and that durability can be comparable to or better than a licensed comparator vaccine in another infectious-disease setting (123, 124). The diversity of saRNA backbones is also expanding. Comparative and mechanistic studies indicate that alphaviral backbone choice can shape expression, innate sensing, and immunogenicity, supporting a move away from a one-backbone-fits-all view of the platform (125, 126). This point is reinforced by a recent EEEV-derived self-replicating RNA platform developed for combination cancer immunotherapies, which demonstrated the importance of empirically selecting fit-for-purpose srRNA vectors for different therapeutic applications (127).

By contrast, circular RNA (circRNA) remains largely preclinical in the vaccine field. Its closed-loop structure is attractive because it resists exonuclease degradation and may support longer protein expression, but clinical evidence for COVID-19 circRNA vaccines remains limited compared with saRNA (128). Thus, the key distinction is not only architectural but also translational: as reflected in Figure 4 and Table 3, circRNA should be described as a promising emergent architecture, not as a platform currently matched to saRNA in clinical maturity.

Beyond changes in RNA format, next-generation COVID-19 vaccine design also includes multivalent and pan-sarbecovirus strategies. These include updated bivalent or monovalent variant-matched vaccines, mosaic nanoparticle displays, and immunogens targeting conserved regions such as the S2 subunit and fusion machinery (129–134). Such approaches seek to reduce the need for frequent strain chasing and to provide broader resilience against future coronavirus evolution. Collectively, next-generation RNA-vaccine development is shifting from maximizing antigen expression alone toward optimizing the match among RNA architecture, antigen breadth, delivery site, innate immune tuning, and clinical objective.

Immunocompromised populations remain among the clearest examples of why vaccination policy must be individualized. Solid-organ transplant recipients, patients receiving B-cell-depleting therapies, and some individuals with hematologic malignancies often show markedly blunted humoral responses to COVID-19 mRNA vaccines, although T-cell responses may be partially retained in some subgroups (135–141). Updated boosters and additional doses can improve responses in part of this population, but heterogeneity remains large, making serologic monitoring and risk-stratified scheduling reasonable in selected high-risk settings (136, 141–144).

Evidence to date supports COVID-19 mRNA vaccination during pregnancy and lactation. Large cohort and meta-analytic data have not shown increases in major adverse birth outcomes attributable to vaccination (145, 146). Modeling work further highlights the potential health costs of excluding pregnant participants from vaccine trials, supporting more deliberate inclusion where ethically appropriate (147). Additional population-level, immunologic, and guidance evidence indicates that vaccination during pregnancy and lactation is compatible with favorable early-childhood outcomes and can transfer protective antibodies through the placenta and breast milk (148–152).

In older adults, immunosenescence reduces the magnitude and speed of vaccine responses, but booster vaccination remains clinically valuable. Updated boosters improve neutralizing titers and reduce hospitalization and death, making older age groups the strongest candidates for continued risk-based seasonal or periodic boosting (14, 15, 153–156).

In children and adolescents, the risk-benefit profile differs from that in older adults but still favors vaccination in appropriate settings. Available evidence supports good safety and benefits that extend beyond prevention of acute disease to reduction in post-COVID condition risk in some studies, although policy decisions should remain age- and risk-sensitive (157–161).