A poorly reactogenic lipid nanoparticle-mRNA vaccine unveils an innate immune pathway for adverse reactions

Abstract

Lipid nanoparticle (LNP)-mRNA vaccines robustly activate immune responses, contributing to their high efficacy and frequent adverse reactions (ARs). Here, we identified an LNP-mRNA formulation with a more favorable balanced immunogenicity-reactogenicity profile. Immune profiling in a mouse model defined the reactogenic LNP-mRNA vaccine as a potent inducer of HMGB1 release, pro-inflammatory cytokine production, and concurrent neutrophil infiltration. HMGB1 induced TNF-α secretion from monocyte subsets, yet in vivo blockade studies revealed the contribution of multiple cytokines (TNF-α, IL-1, and IL-6) to reactogenicity. Among the reactogenic cytokines, IL-1 was identified as the key mediator of vaccine-induced ARs, but was dispensable for humoral immunity. The clinical relevance was confirmed in a well-controlled vaccine cohort where IL-1 pathway activation correlated with fever severity but not with neutralizing antibody titers. We dissected early innate pathways specifically linked to vaccine reactogenicity, providing a rationale for selectively reducing ARs in next-generation vaccines.

Introduction

Lipid nanoparticle (LNP)-mRNA vaccine platform, along with other technologies, has played a crucial role in achieving herd immunity and mitigating the pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) viruses^1–4. This vaccine platform demonstrated >94% efficacy after the second dose, eliciting robust humoral and cellular immune responses that exceed the levels induced by natural infection. However, vaccination also triggers strong pro-inflammatory responses that lead to adverse reactions (ARs), such as fever and fatigue, in > 50% of non-elderly individuals following the second dose^2,5–7. Therefore, reducing the reactogenicity of mRNA vaccines, while maintaining the immunogenicity, is crucial for promoting the continued use and broader application of this platform beyond SARS-CoV-2, particularly as the high incidence of ARs contributes vaccine hesitancy at some extents^8–10.

Intensive efforts have identified key immune correlates of protection, most notably neutralizing antibody titers, which now serve as a reliable standard for evaluating vaccine immunogenicity and antigenic mismatch between vaccine strains and emerging variants¹¹. These correlates have become key drivers in the development of next-generation vaccines with improved immunogenicity and breadth to variants and related viruses. In contrast, the immune mechanisms driving the reactogenic responses remain less appreciated, despite the identification of several potential biomarkers associated with mRNA vaccine reactogenicity.

System vaccinology approaches have suggested that mRNA vaccines uniquely and rapidly induce interferon (IFN)-γ-associated signatures, unlike other platforms^12,13. Furthermore, analyses of reactogenic LNP components, an ionizable lipid used in Moderna mRNA vaccine, suggested a role of the interleukin-1 (IL-1) signaling pathway in pro-inflammatory responses mediated by liposomes containing ionizable lipids¹⁴. However, the extent to which IL-1 signaling or other damage-associated molecular patterns (DAMPs), such as high mobility group box 1 (HMGB1), contribute to vaccine-induced ARs remains unclear. This knowledge gap is partly derived from the limitations of current mouse models in replicating human AR symptoms and of integrated analyses linking early local immune responses with systemic AR severity.

Ionizable lipids, essential for LNP-mRNA vaccine function, facilitate mRNA encapsulation, intracellular delivery, and cytoplasmic release by fusing with endosomal membranes in response to pH changes^15,16. While these lipids have been optimized primarily for efficient delivery and maximal protein expression^15,17–20, they also contribute to inflammatory responses^21,22. However, comprehensive assessments of their reactogenic potential remain limited.

Here, we developed a mouse model for in vivo assessment of vaccine reactogenicity, characterized by fever and reduced locomotor activity, which replicate clinical symptoms in humans following mRNA vaccination. We employed this model to compare LNPs formulated with diverse ionizable lipids and to probe the early immune pathways that distinguish reactogenicity from immunogenicity. Furthermore, we analyzed well-controlled human cohorts stratified by AR severity scores to determine whether similar immune responses apply in clinical settings. Collectively, this study was designed to define immunological processes specifically linked to systemic ARs, providing an immunological basis for rational design of next-generation mRNA vaccines with improved safety profiles.

Results

SM-102 LNP vaccination elevates body temperature and reduces locomotor activity

To investigate the contribution of inflammatory cytokines in systemic ARs, an animal experimental model was initially established to evaluate ARs following mRNA vaccination. To assess the severity of vaccine-related ARs while replicating clinical symptoms, a device was implanted in the abdominal cavity of the mice to monitor both body temperature and locomotor activity (Fig. 1a). Mice received two intramuscular doses of SARS-CoV-2 spike LNP-mRNA formulated with the ionizable lipid SM-102, equivalent in composition to Moderna’s Spikevax (hereafter referred to as “SM-102 LNP”), administered 21 days apart. Following the booster vaccination, a febrile response was observed, peaking at 4 h post-administration, along with a significant reduction in nocturnal locomotor activity compared with that in the saline control group (Fig. 1b, c). Given the nocturnal behavior of mice, the average body temperature during the light period following the booster dose and the ratio of locomotor activity during the dark period on the day of vaccination relative to that of the preceding night were calculated. These parameters served as quantitative indicators of LNP-mRNA vaccine-related ARs, enabling the assessment of vaccine reactogenicity in mice (Fig. 1d, e).

Fig. 1. SM-102 LNP vaccination elevates body temperature and reduces locomotor activity.

a Experimental design. Nano tags were implanted in C57BL/6 mice, which subsequently received intramuscular prime and boost immunizations with either saline or SM-102 LNP vaccine. Body temperature and locomotor activity were continuously monitored, and plasma was collected for RBD–ACE2 binding inhibition assays. b Longitudinal body temperature profiles after vaccination (n = 5 biologically independent mice per group). c Locomotor activity recorded before and after vaccination (n = 5 mice per group). d Mean body temperature during the light period following vaccination (n = 5 mice per group). e Ratio of mean locomotor activity during the dark period on the immunization day relative to the day before (n = 5 mice per group). f Body weight loss 1 day after immunization with saline or escalating doses of SM-102 LNP (n = 8 mice per group). g Mean body temperature during the light period after immunization (n = 8 mice per group). h Relative locomotor activity during the dark period after immunization (n = 8 mice per group). Each dot represents an individual mouse; horizontal bars indicate means. Data in (b, c) are shown as mean ± SD. In (d, e), each dot represents an individual mouse; horizontal bars indicate group means. Statistical significance was determined using two-way repeated-measures ANOVA with Greenhouse–Geisser correction (b, c), unpaired two-tailed Welch’s t-test (d, e), and one-way ANOVA with Dunnett’s multiple-comparison test (f–h). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To verify the sensitivity of these ARs parameters, changes in body temperature and spontaneous locomotor activity were compared with body weight loss, a commonly used indicator of reactogenicity and toxicity in murine models. The SM-102 LNPs were administered at doses ranging from 0.3 µg to 10 µg, and body weight loss, temperature changes, and spontaneous locomotor activity were monitored post-vaccination. Although significant weight loss was only observed in the groups receiving a 10 µg dose, a pronounced febrile response was noted in the 1 µg group, and a reduction in locomotor activity was evident at a dose exceeding 3 µg (Fig. 1f–h). These findings indicate that the parameters assessed in this study provide greater sensitivity compared with body weight loss alone in detecting reactogenicity.

Identification of a low-reactogenic ionizable lipid in LNP–mRNA vaccination

To identify ionizable lipids with reduced reactogenicity, LNP-mRNA candidates were formulated using 23 commercially available ionizable lipids (Fig. 2a and Supplementary Fig. 1a). The molar ratio of ionizable lipid: cholesterol: DSPC: PEG lipid was fixed at 50:38.5:10:1.5, maintaining the same composition as that of the Moderna Spikevax vaccine, except for the ionizable lipid. Twenty LNPs with > 80% mRNA encapsulation efficiency were selected for evaluation of reactogenicity in a mouse model (Supplementary Fig. S1b). The particle sizes of the selected LNPs ranged from 69 to 136 nm (Supplementary Fig. S1c). Following prime-boost vaccinations with 10 µg each, LNP9 (Lipid 5), LNP13 (DOTAP), LNP15 (DODAP), and LNP17 (DOTMA) significantly reduced both fever responses and attenuated declines in locomotor activity compared with the SM-102 LNP (Fig. 2b, c). At both the 10 µg and 0.3 µg doses, Lipid 5 LNP showed RBD–ACE2 binding inhibition activity comparable to that of SM-102 LNP, indicating robust antibody responses (Fig. 2d). The severity of the ARs induced by these LNPs did not correlate with particle size (Supplementary Fig. 1d, e), suggesting the involvement of other factors. Across the 20 LNP formulations tested in vivo, no correlation was observed between antibody responses and systemic ARs (Supplementary Fig. S1f, g).

Fig. 2. Identification of a low-reactogenic ionizable lipid in LNP–mRNA vaccination.

a Schematic of lipid nanoparticle (LNP) preparation incorporating different ionizable lipids with SARS-CoV-2 ancestral spike mRNA. Mean body temperature (b) during the light period and relative locomotor activity (c) during the dark period after secondary immunization with saline, 10 µg of SM-102 LNPs, or ionizable lipid–modified LNPs (n = 5 mice per group). d Relative inhibition of ancestral RBD–ACE2 binding on day 35 (n = 5 mice per group). e Plasma levels of anti-ancestral RBD IgG (n = 5 mice per group). f Frequencies of germinal center (GC) B cells and memory B cells (B_mem) specific to ancestral RBD among total B220⁺CD138⁻ B cells (n = 5 mice per group). g Frequencies of spike-specific non-resident T follicular helper (Tfh) cells among CD4⁺ T cells (n = 5 mice per group). h Frequencies of spike-specific effector CD4⁺ T cells among CD4⁺ T cells (left) and effector CD8⁺ T cells among CD8⁺ T cells (right) (n = 5 mice per group). i Pseudotyped virus neutralization titers (IC₅₀) plotted against mean body temperature (left) or relative locomotor activity (right) at indicated doses (n = 5 mice per group). In (b–h), each dot represents an individual mouse, and horizontal bars indicate means. In (i), data are shown as mean ± 95% CI. Statistical significance was assessed by one-way ANOVA followed by Dunnett’s multiple-comparison test (b–d, relative to SM-102 LNP) and unpaired two-tailed Welch’s t-test (e–h). ns not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To further characterize humoral immune responses, we next assessed germinal center (GC) and memory B (B_mem) cell responses, which represent a critical component of vaccine–induced humoral immunity^13,23. On day 28 (7 days after the secondary immunization), both LNP formulations at 1 µg dosage elicited comparable SARS-CoV-2 spike receptor-binding domain (RBD)-reactive IgG titers (Fig. 2e). Consistent with this, GC B cell responses against ancestral spike and RBD were similar between Lipid 5 and SM-102 LNPs (Fig. 2f and Supplementary Fig. 2a, b). Moreover, both LNPs induced B_mem responses to the ancestral spike, RBD, and variant RBDs at comparable levels (Fig. 2f), along with equivalent frequencies of T follicular helper (Tfh) cells (Fig. 2g and Supplementary Fig. 2c, d).

To assess cellular immunity beyond the Tfh subset, we further quantified spike peptide-specific effector CD4⁺ and CD8⁺ T cells in draining lymph nodes at day 28 using activation-induced marker (AIM) assays and intracellular cytokine staining. Lipid 5 LNP effectively induced spike-specific CD4⁺ T cell responses at levels equivalent to SM-102 LNP. Regarding spike-specific CD8⁺ effector T cells, Lipid 5 LNP achieved a significant 6.9-fold increase over the saline control group. While this induction was marginally lower (1.8-fold) than that of SM-102 LNP, the data confirms that Lipid 5 LNP maintains a high capacity for activating spike-specific CD8⁺ T cell immunity (Fig. 2h).

Finally, we compared neutralizing antibody titers induced by SM-102 and Lipid 5 LNPs. Both LNPs elicited pseudovirus neutralization titers in a dose-dependent manner. However, SM-102 LNP consistently induced higher neutralization titers at equivalent doses to Lipid 5 LNP for unknown reasons (Fig. 2h). Despite the inferiority in eliciting neutralizing antibodies at the equivalent dosages, Lipid 5 LNP was less reactogenic than SM-102 LNP at the dosages inducing comparable levels of neutralizing antibody titers (Fig. 2h). Together, we identified Lipid 5 LNP as an mRNA platform that offers an optimal balance between immunogenicity and reactogenicity compared to conventional SM-102 LNP in the ability to induce multiple humoral immune parameters.

Lower inflammatory profiles and distinct cell recruitment at the injection site of Lipid 5 LNP

Lipid 5 was previously identified as highly tolerable when administered intravenously²⁴. However, the previous toxicity assessments have been restricted to liver injury indicators, providing an incomplete assessment of potential inflammatory responses²⁴. To investigate the differences in immune responses induced by SM-102 and Lipid 5 LNP formulations, we first compared the transcriptomic profiles at the injection site 6 h after booster immunization (Fig. 3a). Bulk RNA-seq analysis revealed broad transcriptional activation in the SM-102 group, with 562 genes significantly upregulated compared with 167 genes in the Lipid 5 group (Fig. 3b). In-depth analysis of selected immune-related transcripts demonstrated higher expression of pro-inflammatory cytokines, including Il1a, Il1b, Il6, Ifng, and Tnf, in the SM-102 group relative to the Lipid 5 group (Fig. 3c).

Fig. 3. Lower inflammatory profiles and distinct immune cell recruitment at the injection site of Lipid 5 LNP.

a Experimental design for assessing immune responses induced by SM-102 or Lipid 5 LNPs in C57BL/6 mice. b Volcano plot showing differential gene expression in gastrocnemius muscle 6 h after vaccination with SM-102 LNP versus Lipid 5 LNP (n = 3 mice per group). Blue, genes upregulated in Lipid 5; red, genes upregulated in SM-102. c Normalized expression levels of selected genes at the injection site 6 h after LNP–mRNA vaccination. d Z-scored serum cytokine and chemokine levels at indicated time points (n = 8 mice per group). e Absolute counts of immune subsets at the injection site (n = 9 mice per group). f Plasma HMGB1 levels (n = 9 mice per group). g Heatmaps showing the frequencies of cytokine-producing cells among indicated immune subsets following 6 h HMGB1 stimulation (n = 10 mice per group). Heatmaps in (d, g) display group means. In (c, e, f), each dot represents an individual mouse; horizontal lines indicate means. Statistical analyses were performed using DESeq2 implemented in iDEP (b, c), unpaired two-tailed Welch’s t-test with Benjamini–Hochberg correction (d–f), and one-way ANOVA with Dunnett’s multiple-comparison test (g). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

We next quantitated cytokine protein expression in the serum and cellular composition at the injection sites longitudinally following the booster dose (Fig. 3a). Consistent with transcriptome analysis at the injection sites, we detected a broad range of pro-inflammatory cytokines and chemokines in the serum from SM-102 group (Fig. 3d). However, the pro-inflammatory responses were highly attenuated in the Lipid 5 group (Fig. 3d). This difference was particularly pronounced 6 h post-vaccination, with substantial reductions observed in IL-6 (3.2×), IL-1RA (4.5×), IFN-γ (1.5×), and monocyte chemoattractant protein-1 (MCP-1) (4.8×) (Supplementary Fig. 3).

About 6 h post-injection, SM-102 LNP led to a marked fivefold increase in neutrophil infiltration compared with Lipid 5 LNP over 24 h (Fig. 3e and Supplementary Fig. 4a). In contrast, Lipid 5 LNP predominantly recruited monocyte subsets, including macrophages and Ly6C^hi monocytes (Fig. 3e). Thus, vaccination with a poorly reactogenic LNP influences not only the pro-inflammatory cytokine/chemokine production but also the early cellular dynamics at the injection site.

Innate immune cells recognize damage-associated molecular patterns (DAMPs) released by host cells and initiate early immune responses, including inflammation²⁵. We hypothesized that the differences in DAMP release following booster vaccination might account for the observed changes in the early cell recruitment. High mobility group box 1 (HMGB1), a well-known DAMP²⁶, rapidly increased > 7-fold by 2 h of SM-102 LNP vaccination, with the peak response at 6 h, whereas HMGB1 induction was less than 2-fold in the Lipid 5 LNP (Fig. 3f). To assess the pro-inflammatory potential of HMGB1, we stimulated mouse splenocytes with recombinant HMGB1 and analyzed intracellular cytokine production (Fig. 3g and Supplementary Fig. 4b). HMGB1 induced tumor necrosis factor (TNF)-α secretion from macrophages and Ly6^hi monocytes in a dose-dependent manner, but it failed to induce other cytokines above the detectable levels (Fig. 3g and Supplementary Fig. 4c–e). Thus, a reactogenic LNP-mRNA formulation rapidly triggers HMGB1 release, with the ability to promote TNF-α secretion from monocytes.

Although neutrophil infiltration was observed at the injection sites of reactogenic LNPs (Fig. 3e), their contribution to systemic ARs remains to be elucidated. To this end, we depleted the neutrophils by anti-Ly6G/Ly6C antibody treatment; however, no improvements in fever or locomotor activity were observed in the experimental conditions tested (Supplementary Fig. 5), showing the dispensability of neutrophil recruitment on systemic ARs.

IL-1RI deficiency reduces reactogenicity without impairing humoral immunity

To explore the role of inflammatory cytokines in mRNA vaccine-related ARs, receptor-blocking or neutralizing antibodies targeting IL-1, IL-6, IFN-γ, and TNF-α were administered to mice prior to LNP-mRNA vaccination (Fig. 4a). We observed the most prominent inhibitory effects on fever after the treatment with IL-1 receptor (IL-1R) antibodies (Fig. 4b). The treatment with IL-6R antibody and TNF-α antibody also reduced the levels of fever reactions (Fig. 4b). The IL-1R blockade tended to restore the locomotor activity inhibited by reactogenic LNP, but the effects were quite modest (Fig. 4c). The role of IL-1 pathway was further investigated using IL-1 receptor type I knockout (IL-1RI KO) mice. In line with antibody blockade studies, IL-1RI KO mice did not elevate body temperature above the detectable levels (Fig. 4d), and the inhibitory effect on locomotor activity was significantly attenuated relative to the control mice (Fig. 4e). However, IL-1R deficiency did not fully restore locomotor activity even in the IL-1R-deficient mice, suggesting that additional factors contribute to these ARs (Fig. 4e). Despite these strong impacts on the systemic ARs, the loss of IL-1R did not alter the production of cytokines and chemokines following SM-102 LNP vaccination (Fig. 4f and Supplementary Fig. 6). HMGB1 release also remained fairly intact (Fig. 4g). These findings suggest that the early DAMP and cytokine/chemokine production proceeds IL-1 pathway independently of IL-1-regulated network.

Fig. 4. IL-1RI deficiency reduces reactogenicity without impairing humoral immunity.

a Experimental design assessing the effects of monoclonal antibodies against cytokines or cytokine receptors on vaccine-induced adverse reactions. b Mean body temperature during the light period after secondary immunization (n = 5 mice per group). c Relative locomotor activity during the dark period after secondary immunization (n = 5 mice per group). d Mean body temperature during the light period in wild-type and Il1r1^–/– mice after secondary immunization (n = 4–5 mice per group). e Relative locomotor activity during the dark period in wild-type and Il1r1^–/– mice after secondary immunization (n = 4–5 mice per group). f Z-scored serum cytokine and chemokine levels at indicated time points (n = 4–6 mice per group). g Plasma HMGB1 levels at indicated time points (n = 4–6 mice per group). h Frequencies of ancestral RBD-specific germinal center (GC) B cells (left) and memory B cells (right) within B220⁺CD138^– B cells on day 28 (n = 8 mice per group). i Plasma IgG titers against Wuhan spike (left), ancestral RBD (center), and pseudovirus neutralization titers (PVNT) (IC₅₀, right) (n = 8 mice per group). In (b–e, g–i), each dot represents an individual mouse; horizontal lines indicate means. Heatmaps in (f) display group means. Statistical significance was determined by one-way ANOVA with Dunnett’s multiple-comparison test (b, c), one-way ANOVA with Tukey’s test (d, e), unpaired two-tailed Welch’s t-test with Benjamini–Hochberg correction (f, g), and unpaired two-tailed Welch’s t-test (h, i). ns not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

We also examined the role of IL-1 on the immunogenicity in eliciting humoral immunity. In contrast to the profound impact on the systemic ARs on IL-1R deficiency, the levels of RBD-reactive GC and B_mem cells as well as spike/RBD-reactive IgG and neutralizing antibodies were comparable to those observed in wild-type mice (Fig. 4h, i). Together, these results indicate that IL-1 pathway plays an essential role on the systemic ARs following mRNA vaccination but it is not required for vaccine-induced humoral immunity.

Potential role of IL-1 on systemic ARs in vaccinees

Clinical samples and reactogenicity-related data from a well-controlled vaccination cohort were previously collected²⁷. In the present study, factors associated with ARs following mRNA vaccination were investigated by reanalyzing 90 individuals who received three doses of BNT162b2 (Pfizer-BioNTech). The analysis focused on data collected 1 day after the third dose, along with questionnaire-based adverse event scores²⁷. Correlation analysis revealed a significant association between systemic AR scores and plasma levels of MIG (CXCL9) and IP-10 (CXCL10), which are downstream of IFN-γ signaling, and IL-1R antagonist (IL-1RA), a competitive inhibitor of IL-1 (r = 0.42–0.47, p < 0.0001; Fig. 5a). These results suggest that multiple soluble factors are involved in LNP-mRNA-induced ARs. Of note, the correlation of IL-1RA was relatively specific to the systemic ARs, but not applicable to the antibody titers at day 28 after vaccination, such as RBD-reactive IgG titers, RBD–ACE2 inhibitory titers, and pseudovirus neutralizing titers (Fig. 5b). These findings in clinical setting further support the notion that IL-1 pathway regulates ARs specifically without affecting the elicitation of humoral immunity.

Fig. 5. Potential role of IL-1 in systemic adverse reactions in vaccinees.

a Spearman’s correlation coefficients between plasma cytokine and chemokine levels measured 1 day after the third BNT162b2 vaccination and adverse event (AE) scores recorded over 7 days (n = 90 individuals). Color scale represents correlation coefficients (r). Asterisks denote significance based on p-values adjusted by Benjamini–Hochberg correction (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). b Correlations between plasma IL-1RA concentrations and peak anti- ancestral RBD IgG titers, relative RBD–ACE2 inhibition, or pseudovirus neutralization titers (PVNT; IC₅₀) on day 68 (n = 90 individuals). Each dot represents one participant. Statistical significance was determined using Spearman’s rank correlation test with Benjamini–Hochberg correction.

Discussion

The relationship between the inflammatory responses triggered by LNP-mRNA vaccines, protective immunity, and adverse reactions (ARs) remains poorly understood. Using a murine model that replicates vaccine-induced ARs in humans, we identified an ionizable lipid, Lipid 5, as a poorly reactogenic component. Indeed, LNP-mRNA formulations containing Lipid 5 elicited multiple components of humoral immunity, such as antigen-reactive antibodies, GC/B_mem cells, Tfh cells, and spike-specific CD4⁺ effector T cells at comparable levels to those induced by reactogenic LNPs. Whereas we did observe lesser activity in eliciting neutralizing antibodies from poorly reactogenic LNP under the same vaccine dosage, reactogenicity was consistently lower in Lipid 5 LNP at the dosages eliciting the equivalent neutralizing antibodies to reactogenic LNP. Thus, we conclude that Lipid 5 LNP indeed has more preferable balance in the immunogenicity/reactogenicity under the evaluation of mouse model. LNP formulations with reduced reactogenicity may be particularly advantageous for populations at higher risk of ARs or for applications in which improved tolerability is prioritized. At the same time, the potential requirement for higher mRNA doses with Lipid 5 underscores the need for careful dose-escalation and long-term safety evaluation before clinical application.

As previously described in ref. ¹⁴, we confirmed that the IL-1/IL-1RA axis is involved in lipid-induced inflammatory responses. Further extending the findings, we demonstrated the pivotal and selective role of IL-1 pathway in vaccine-induced reactogenicity, represented by elevated body temperature and reduced locomotor activity, under our mouse model. While isotype-matched control antibodies were not employed for each blocking antibody in adherence to the 3Rs principle, the use of IL-1RI-deficient mice provided stringent genetic validation, confirming that the observed effects on reactogenicity were specifically IL-1R-dependent. Furthermore, IL-1 pathway serves as a key innate event that distinguishes the reactogenic activity from immunogenicity observed between Lipid 5 LNP and reactogenic LNP-mRNA formulations, a notion further supported by clinical data. These findings in both mice and humans indicate that reactogenicity and immunogenicity can be at least partially uncoupled, with IL-1 pathway playing a central role in this process. This insight offers promising avenues for the development of LNP-mRNA vaccines that maintain robust immunogenicity while minimizing ARs.

One possible mechanism for IL-1-mediated fever reaction is that IL-1RI signaling in vascular endothelial cells within muscle promotes local prostaglandin E2 production²⁸, which subsequently crosses the blood–brain barrier and acts on the central nervous system to thermoregulatory set point. Together with other possible pathways, we consider that IL-1 is the principal driver mediating vaccine-induced fever. However, the mechanisms behind IL-1-mediated reduction of locomotor activity are more complex. Because IL-1R deficiency failed to fully restore the activity, it is conceivable that multiple pathways and factors other than IL-1 pathway are involved in the process. Indeed, in addition to IL-1²⁹, other cytokines including IL-6, IFN-γ, and TNF-α have been implicated in sickness behaviors^30–33. These reports suggest that the decline in locomotor activity may result from a combination of systemic and local ARs, including fatigue, pain, and headache, that are mediated by multiple pathways. Accordingly, while locomotor activity provides a useful quantitative parameter of systemic reactogenicity in our murine model, it does not fully capture the multifaced nature of clinical ARs observed in humans. Further investigation is required to elucidate the mechanisms by which IL-1 and other pathways contribute to reduced locomotor activity.

Our study also identified a complementary HMGB1–TNF-α pathway associated with early inflammatory responses. SM-102 LNP induced rapid release of HMGB1 at the injection site, which correlated with neutrophil recruitment. In vitro, HMGB1 directly and prominently induced TNF-α production in macrophages and monocytes, whereas other cytokines were not detected. Given that TNF-α is known to drive IL-1 production in multiple contexts^34,35, HMGB1/TNF-α axis may contribute to IL-1–mediated febrile responses induced by highly reactogenic LNPs. Thus, the HMGB1–TNF-α–IL-1 axis may represent a key amplification loop in LNP-induced ARs. Bulk RNA-seq analysis at the injection site also revealed broader transcriptional activation including pro-inflammatory cytokines such as IL-1, IL-6, IFN-γ, and TNF-α, in reactogenic LNP group compared with Lipid 5 group. These local transcriptomic data support the broad activation of multiple pro-inflammatory pathways, and indicate that HMGB1 may act either as an upstream inflammatory mediator or as a downstream marker of tissue damage rather than as a sole driver of reactogenicity. Although neutrophil depletion did not significantly affect systemic reactogenicity parameters in our model, our current readouts primarily reflect systemic and behavior-based responses and do not capture local reactogenicity at the injection site. Therefore, we cannot exclude the possibility that neutrophils contribute to local inflammatory symptoms, tissue damage, or tissue repair processes following LNP-mRNA vaccination. Future studies focusing on local reactogenicity will be required to clarify the role of neutrophils and other innate immune cells in shaping the local inflammatory environment.

Despite the subtle structural difference between Lipid 5 and SM-102, namely the positions of the two ester bonds (Supplementary Fig. 1a), Lipid 5 LNPs consistently formed smaller particles than SM-102 LNPs (Supplementary Fig. 1c). These altered ester bond positions are expected to yield distinct metabolites upon degradation, although their physiological impact was not addressed in this study. A recent report suggested that LNP-induced endosomal pores can be sensed by galectins to trigger inflammation³⁶, and rapidly biodegradable lipids such as Lipid 5 may generate smaller, more readily repairable pores, thereby attenuating such responses. Thus, while the precise mechanism remains unclear, the chemical structure of ionizable lipids may account for differential reactogenicity. Importantly, defining the mechanistic basis of Lipid 5’s reduced reactogenicity-including potential differences in lipid metabolism, endosomal escape, and innate immune sensing-represents a critical limitation of the present study and will be essential for maximizing its translational impact and informing rational LNP design. The low-reactogenic profile of Lipid 5 has not yet been directly evaluated in humans. To date, the safety of Lipid 5 has been assessed in rodents and non-human primates²⁴, but clinical validation will be required to establish its translatability.

Differences in antigen expression efficiency could influence reactogenicity. However, we observed no correlation between antibody responses and systemic ARs across multiple LNP formulations (Supplementary Fig. 1f, g), suggesting that differences in reactogenicity are not solely driven by antigen expression levels. It should be noted, however, that our analyses in this study focused primarily on the contribution of ionizable lipids, whereas other vaccine components, such as mRNA modifications and encoded antigens, were not systemically examined. Moreover, direct measurements of antigen expression at the injection site will be required to fully address the relative contribution of antigen expression efficiency to reactogenicity.

Beyond humoral immunity, our analysis of spike-specific effector T cell responses revealed that Lipid 5 LNP induces CD4⁺ effector T cells at levels comparable to SM-102, suggesting that attenuated reactogenicity does not necessarily compromise CD4⁺ effector priming. In contrast, CD8⁺ effector T cell responses were slightly reduced with Lipid 5 LNP relative to SM-102, indicating that cellular immunity may be sensitive to the magnitude of inflammation. Together, these findings raise the possibility that inflammatory pathways such as IL-1 shape the quality of vaccine-induced T cell responses. Further studies will be required to determine how such inflammatory signals influence T cell differentiation, polarization, and durability, which were beyond the scope of the present study.

Notably, clinical analyses revealed that IL-1RA correlated with systemic ARs but not with antibody responses, consistent with the mouse findings. In our model, SM-102 LNP triggered early release of HMGB1, which activated macrophages and monocytes to produce TNF-α. TNF-α, in turn, is known to promote IL-1 production, suggesting that HMGB1 functions as an upstream amplifier of IL-1-driven febrile responses. Taken together, these results support a model in which IL-1 signaling acts as the central driver of fever, whereas HMGB1-TNF-α-IL-1 interactions further contribute to amplification of inflammatory pathways underlying reactogenicity. These pathways collectively shape reactogenicity without impairing adaptive immunity (Fig. 6).

Fig. 6. Proposed model of inflammatory and immune pathways triggered by LNP–mRNA vaccination.

LNP–mRNA vaccination activates innate pathways that contribute to both systemic adverse reactions and antibody responses. LNPs induce IL-1 production, which drives systemic adverse reactions. In parallel, LNPs promote HMGB1 release from the injection site, stimulating macrophages and monocytes to produce TNF-α. TNF-α can further enhance IL-1 production, leading to febrile reactions. LNPs also induce chemokines that recruit neutrophils and monocytes to the injection site. Recruited monocytes likely migrate to draining lymphoid tissues, where they support antibody responses independently of IL-1 signaling.

This study has several limitations. First, the human analyses were restricted to a single vaccine platform and correlative in nature. Because adverse event scores were derived from self-reported questionnaires, they are subject to recall bias and inter-individual variability in symptom perception; this limits direct mechanistic conclusions regarding the causal role of IL-1 signaling in humans. Second, the generalizability of our findings to other populations and contexts remains uncertain, as experiments were limited to a single antigen and mouse strain using intramuscular administration in young adult mice. Immune responses to LNP–mRNA vaccines may vary based on experimental parameters (e.g., antigen properties and administration routes) and host-intrinsic factors (e.g., genetic background, age, species-specific differences in innate immune sensing, and prior immune history). These factors may influence both immunogenicity and reactogenicity of vaccines. Consequently, future studies across diverse vaccine platforms, antigens, and host contexts are essential to clarify how IL-1-mediated reactogenicity shapes vaccine-induced ARs.

Methods

Human subjects

This human observational study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Boards of the National Institute of Infectious Diseases and Tokyo Shinagawa Hospital (permit numbers: 1322 and 21-A-11). The cohort consisted of 184 Japanese participants enrolled in a phase 2/3 randomized, observer-blinded, noninferiority trial of the S-268019-b recombinant SARS-CoV-2 spike protein vaccine (jRCT2031210470). For this study, 90 participants who received BNT162b2 as their third vaccination were analyzed. Informed consent was obtained as previously reported.

Animal models

All animal experiments were approved by the Institutional Review Board of the National Institute of Infectious Diseases (permit numbers: 122080-V, 122114-II, 122159-II, 123105-II, and 123127). Seven-week-old female C57BL/6 mice (Japan SLC) and Il1r1^−/− mice were used. Animals were randomly allocated to each group, and no blinding was applied. All surgical procedures were performed under isoflurane anesthesia. At the end of the study, mice were humanely euthanized by cervical dislocation under deep isoflurane anesthesia, in accordance with the guidelines of the Animal Care and Use Committee of the National Institute of Infectious Diseases.

mRNA synthesis and LNP formulation

The complete sequence of the SARS-CoV-2 ancestral spike protein, incorporating two key mutations (K986P and V987P) to enhance structural stability, was codon optimized and cloned into a plasmid containing a T7 promoter, a 5′ UTR derived from human α-globin, a 3′ UTR from amino-terminal enhancer of split (AES) mRNA and mitochondrially encoded 12S rRNA (mtRNR1)³⁷, and a poly-A tail designed for mRNA synthesis. mRNA was synthesized using the IVTpro T7 mRNA Synthesis Kit (Takara Bio, #6144) with a linearized plasmid digested by Esp3I (New England Biolabs, #R0734L) or BsmBI-v2 (New England Biolabs, #R0739L). N1-methylpseudouridine-5′-triphosphate (TriLink, #N-1081) was used in place of the uridine-5′-triphosphate supplied with the IVTpro T7 kit. The mRNA was capped with CleanCap Reagent AG (TriLink, #N-7113) to obtain cap1 mRNA. The integrity and length of synthesized mRNAs were confirmed by agarose gel electrophoresis using the DynaMarker® RNA High for Easy Electrophoresis (BioDynamics Laboratory, #DM170), and aliquots were subsequently stored at −80 °C.

Tetrakis(8-methylnonyl) 3,3′,3′′,3′′′-(((methylazanediyl)bis(propane-3,1-diyl))bis(azanetriyl))tetrapropionate (306Oi10, #36698); 2-(((4-(dimethylamino)butanoyl)oxy)methyl)-2-((((Z)-tetradec-9-enoyl)oxy)methyl)propane-1,3-diyl (9Z,9′Z)-bis(tetradec-9-enoate) (TCL053, #37045); 2-hexyl-decanoic acid, 1,1′-[[(4-hydroxybutyl)imino]di-6,1-hexanediyl] ester (ALC-0315, #34337); 8-[[8-[(1-ethylnonyl)oxy]-8-oxooctyl][3-[[2-(methylamino)-3,4-dioxo-1-cyclobuten-1-yl]amino]propyl]amino]-octanoic acid, 1-octylnonyl ester (Lipid 29, #35337); 8-[(2-hydroxyethyl)[8-(nonyloxy)-8-oxooctyl]amino]-octanoic acid, 1-octylnonyl ester (Lipid 5, #34372); 8-[(2-hydroxyethyl)[6-oxo-6-(undecyloxy)hexyl]amino]-octanoic acid, 1-octylnonyl ester (SM-102, #33474); 1,1′-[[2-[4-[2-[[2-[bis(2-hydroxydodecyl)amino]ethyl](2-hydroxydodecyl)amino]ethyl]-1-piperazinyl]ethyl]imino]bis-2-dodecanol (C12-200, #36699); N,N,N-trimethyl-2,3-bis[[(9Z)-1-oxo-9-octadecenyl]oxy]-1-propanaminium monochloride (DOTAP, #15110); 4-(dimethylamino)-butanoic acid, (10Z,13Z)-1-(9Z,12Z)-9,12-octadecadien-1-yl-10,13-nonadecadien-1-yl ester (DLin-MC3-DMA, #34364); 9Z-octadecenoic acid, 1,1′-[1-[(dimethylamino)methyl]-1,2-ethanediyl] ester (DODAP, #25726); 9-[4-(dimethylamino)-1-oxobutoxy]-heptadecanedioic acid, 1,17-di-(2Z)-2-nonen-1-yl ester (L-319, #35051); N,N,N-trimethyl-2,3-bis[(9Z)-9-octadecen-1-yloxy]-1-propanaminium monochloride (DOTMA, #25926); N,N-dimethyl-2,3-bis[(9Z,12Z)-9,12-octadecadien-1-yloxy]-1-propanamine (DLin-DMA, #36701); N,N-dimethyl-2,2-di-(9Z,12Z)-9,12-octadecadien-1-yl-1,3-dioxolane-4-ethanamine (DLin-KC2-DMA, #34363); N,N-dimethyl-2,3-bis[(9Z)-9-octadecen-1-yloxy]-1-propanamine (DODMA, #15109); 1-methyl-4,4-bis[(9Z,12Z)-9,12-octadecadien-1-yloxy]-piperidine (YSK05, #35786); (3β)-cholest-5-en-3-ol 3-[N-[2-(dimethylamino)ethyl]carbamate], monohydrochloride (DC-Chol, #16943); N-[2-[(2-hydroxyethyl)amino]ethyl]-(3β)-cholest-5-ene-3-carboxamide (OH-Chol, #26586); and α-[(2 R)-2,3-bis[(1-oxotetradecyl)oxy]propyl]-ω-methoxy-poly(oxy-1,2-ethanediyl)(DMG-PEG2000, #33945) were purchased from Cayman Chemical Company. Meanwhile,(9Z,12Z)-3-((4,4-bis(octyloxy)butanoyl)oxy)-2-((((3-(diethylamino)propoxy)carbonyl)oxy)methyl)propyl octadeca-9,12-dienoate (LP-01, #DC31000) and 1,19-Bis(2-butyloctyl) 10-[[3-(dimethylamino) propyl](1-oxononyl)amino]nonadecanedioate (Lipid A9, #DC57100) were purchased from DC Chemicals. 3,6-bis[4-[Bis (2-hydroxydodecyl)amino]butyl]-2,5-piperazinedione (cKK-E12, #N-1012) was purchased from Echelon Biosciences. COATSOME SS-E (#SS-E) and COATSOME SS-OP (#SS-OP) were purchased from Yuka Sangyo Co., Ltd.

Cholesterol (#C8667-5G) and 1,2-dioctadecanoyl-sn-glycero-3-phosphocholine [DSPC (18:0 PC), #850365P-500MG] were purchased from Sigma-Aldrich.

For the LNP formulation, lipids were dissolved in ethanol at molar ratios of 50:38.5:10:1.5 (ionizable lipid: cholesterol: DSPC: DMG-PEG2000). This lipid mixture was combined with 50 mM sodium citrate buffer (pH 4.0) containing mRNA at a 3:1 aqueous-to-ethanol ratio using the NanoAssemblr Ignite nanoparticle formulation system (Precision NanoSystems).

The resulting LNP formulations were dialyzed against 0.25 × Tris buffered saline (Nacalai Tesque, #12748-31) supplemented with 254 mM sucrose (Nacalai Tesque, #30403-55) for 18 h using Slide-A-Lyzer G3 Dialysis Cassettes (Thermo Fisher Scientific, #A52971). Following dialysis, the LNP formulations were stored at −80 °C. The size and polydispersity of the mRNA-loaded particles were characterized via dynamic light scattering using a Zetasizer Ultra (Malvern Panalytical). The encapsulation efficiency was assessed using the Quant-iT RiboGreen RNA Assay Kit (Thermo Fisher Scientific, #R11490).

In vivo evaluation of LNP-mRNA

Seven-week-old female C57BL/6 mice (Japan SLC) and Il1r1^−/− mice were used. To monitor reactogenicity, mice were surgically implanted intraperitoneally with three-axis accelerometers (nano tag; KISSEI COMTEC CO., Ltd.), followed by a 5-day recovery period. Young adult female mice were used to provide a controlled and homogeneous model for assessing reactogenicity, as group-housed male mice frequently display agonistic behaviors that can confound locomotor activity-based measurements. Mice were immunized intramuscularly in the left gastrocnemius muscle with either saline (mock control) or LNP-mRNA (the indicated amount of mRNA in 50 μL) on days 0 (prime) and 21 (boost). Vaccinations were performed between 10:00 and 11:00 AM.

For reactogenicity assessment, body temperature and locomotor activity were monitored from Day 20 to Day 22. In cytokine blockade/neutralization and neutrophil depletion experiments, monitoring was extended to Day 19–22 to account for possible effects of antibody treatment. Body temperature was calculated as the mean from the time of booster vaccination until the end of the light period (8:00 PM). Locomotor activity was quantified as the ratio of vibration frequency during the dark period on the day of boost relative to the preceding day.

Blood samples for serological assays were collected on Day 35. To assess local immune responses, blood and injection site tissues were collected immediately before (0 h) and at 2, 6, and 24 h after booster vaccination on Day 21.

Cytokine blockade/neutralization and neutrophil depletion

To assess the contribution of inflammatory cytokines to vaccine-induced adverse reactions, mice received monoclonal antibodies against IL-1R (Bio X Cell, #BE0256), IL-6R (Bio X Cell, #BE0047), IFNγR (Bio X Cell, #BE0029), or TNF-α (Bio X Cell, #BE0058). Antibodies [200 µg in 200 µL phosphate buffered saline (PBS; Fujifilm Wako Pure Chemicals, #166-23555)] were administered intravenously on Day 20 (one day before the boost) and Day 21 (immediately before vaccination). Body temperature and locomotor activity were monitored from Day 19 to Day 22 to account for potential effects of antibody treatment prior to vaccination. For the cytokine receptor-blocking experiments, isotype-matched control antibodies were omitted in adherence to the 3Rs principle (Replacement, Reduction, and Refinement). Given that the blocking antibodies targeted multiple distinct receptors and originated from various host species and isotypes, including individual control groups would have required a substantial increase in animal numbers. To prioritize animal welfare and minimize usage, we instead relied on independent validation, such as the use of IL-1RI-deficient mice, to confirm experimental specificity.

Neutrophils were depleted by intravenous injection of anti-Ly6G/Ly6C monoclonal antibody (clone RB6-8C5, Bio X Cell, #BE0075; 200 µg in 200 µL PBS) on Day 20 and Day 21. To confirm depletion efficiency, peripheral blood was collected on Day 21 and analyzed by flow cytometry as described below. Body temperature and locomotor activity were monitored from Day 19 to Day 22 to capture possible effects of antibody-mediated depletion prior to boosting.

Sample processing

Serum and plasma samples were prepared as follows. After collection, whole blood was centrifuged at 15,000 rpm for 10 min at 4 °C. The supernatant was collected as plasma or serum and stored at −20 °C until further analysis. Plasma samples were used for antibody response and HMGB1 measurements, whereas serum samples were used for cytokine and chemokine analyses, including IL-1RA. Plasma samples for antibody response assays were heat-inactivated by incubation at 56 °C for 30 min before measurement. After thawing, serum or plasma samples were immediately processed for the corresponding assays.

To analyze B cells in the draining lymph nodes, a digestion buffer was prepared by adding 1 mg/mL collagenase D (Roche, Cat#11088858001) and DNase I (Roche, Cat#10104159001) to RPMI 1640 medium (Fujifilm Wako Pure Chemical Corporation, #189-02025) supplemented with 10% heat-inactivated FBS (Nichirei Biosciences, #171012), 100 U/mL penicillin and 100 μg/mL streptomycin (Thermo Fisher Scientific, #15140-122). The inguinal lymph nodes were incubated in 2 mL of digestion buffer at 37 °C for 1 h with stirring; then, the mixture was passed through a 100 µm cell strainer to obtain a single-cell suspension.

For transcriptome analysis at the injection site, the gastrocnemius muscle was immediately immersed in RNAlater stabilization solution (Thermo Fisher Scientific, #AM7021) and incubated overnight at 4 °C. Tissues were subsequently stored at −80 °C until shipment for RNA extraction and sequencing.

For immune cell analysis at the injection site, the gastrocnemius muscle was cut into 1-mm sections, and single-cell suspensions were obtained using the gentleMACS Octo Dissociator with Heaters (Miltenyi Biotec, #130-096-427) and the Skeletal Muscle Dissociation Kit for mice and rats (Miltenyi Biotec, #130-098-305), following the manufacturer’s instructions.

For splenocyte preparation, spleens were aseptically removed and minced with scissors. Tissues were digested in 5 mL of digestion buffer for 30 min at 37 °C with stirring. The resulting cell suspension was passed through a 70 µm cell strainer and pelleted by centrifugation at 1,500 rpm for 5 min. Red blood cells were lysed with BD Pharm Lyse™ Lysing Buffer (BD Biosciences, Cat#555899), washed with PBS, and pelleted again. Cells were finally resuspended in RPMI supplemented with 10% heat-inactivated FBS and used immediately for in vitro assays.

For neutrophil depletion experiments, peripheral blood (100 µL) was collected into 1 mL PBS containing 2.5 mM EDTA (Thermo Fisher Scientific, #15575), followed by centrifugation at 1500 rpm for 5 min. Red blood cells were lysed using 1 mL BD Pharm Lyse™ Lysing Buffer. After washing with PBS, leukocytes were processed for flow cytometry.

In vitro HMGB1 stimulation

Splenocytes were plated at 1 × 10⁶ cells per well in 96-well U-bottom plates (TPP, #92097) and stimulated with recombinant HMGB1 (Sino Biological, #50913-M01H) at 1 or 5 µg/mL, or left unstimulated. Cells were incubated for 2 h at 37 °C in 5% CO₂, after which brefeldin A (BioLegend, #420601) was added at a final 1× concentration. Cultures were continued for an additional 4 h, followed by harvest and analysis by intracellular cytokine staining as described below.

Electrochemiluminescence immunoassay

The inhibition of the RBD- ACE2 interaction was quantified using a V-PLEX SARS-CoV-2 Panel 22 IgG kit (Meso Scale Discovery, #K15559U). For the RBD-ACE2 inhibition assay, Panel 22 plates were blocked with MSD Blocker A reagent (Meso Scale Discovery, #R93AA) for 1 h, washed with washing buffer, and incubated for 1 h with plasma diluted at a 1:10 ratio in Diluent 100 (Meso Scale Discovery, #R50AA). Following the incubation, SULFO-TAG Human ACE2 Protein (Meso Scale Discovery, #D21ADG) was added without washing, followed by a 1 h incubation at room temperature (15–25 °C). The plates were then washed with washing buffer before adding MSD Gold read buffer B (Meso Scale Discovery, #R60AM). Electrochemiluminescence was immediately measured using the MESO QuickPlex SQ120 (Meso Scale Discovery, #AI0AA). ACE2-binding signals were normalized to those obtained in the blank wells (where only the diluent was added instead of the diluted plasma).

Plasma IgG titers

Plasma IgG titers were measured using a Luminex-based bead assay. For bead preparation, biotinylated spike protein or RBDs (produced in-house) were incubated with MagPlex-Avidin microsphere beads (Luminex, #MA-A012-01, #MA-A014-01, #MA-A-019-01, and #MA-A025-01) at 4 °C overnight. Following incubation, the beads were washed three times with Luminex wash buffer [PBS containing 1% bovine serum albumin (BSA; Sigma-Aldrich, #A2153), 0.05% Tween (Fujifilm Wako Pure Chemicals, #167-11515), and 10 μM biotin (Tokyo Chemical Industry)]. The prepared beads were then mixed with plasma samples diluted in milk buffer [PBS containing 1% skim milk (Fujifilm Wako Pure Chemicals, #190-12865), 1% BSA, 0.05% Tween, and 10 μM biotin] in a MultiScreenHTS BV Filter Plate (Millipore, #MSBVN1250) and incubated overnight at 4 °C. After incubation, the plates were washed three times with Luminex wash buffer using a MultiScreenHTS Vacuum Manifold (Millipore). The beads were subsequently incubated with phycoerythrin (PE)-conjugated goat anti-mouse IgG (Southern Biotech, #1030-09) at room temperature for 1 h. After another three washes with Luminex wash buffer, the beads were suspended in the same buffer. The median fluorescence intensity of each bead was measured using the Bio-Plex 3D System (Bio-Rad) or xMAP INTELLIFLEX DR-SE (Thermo Fisher Scientific). Plasma IgG levels were quantified using the monoclonal anti-RBD antibody SA55³⁸ as a reference. Concentrations were calculated by nonlinear regression (four-parameter logistic curve fitting) using GraphPad Prism software (GraphPad).

Pseudovirus neutralization assay

SARS-CoV-2 pseudotyped viruses bearing the ancestral strain spike protein (WK-521), including a 19-amino-acid truncation at the C terminus, were produced as previously described³⁹. The spike sequence was cloned into pCAGGS, transfected into 293 T cells (ATCC, #CRL-3216), and after 24 h, infected with G-complemented VSVΔG/Luc (MOI = 0.5). Virus supernatants were collected 24 h later, clarified, and stored at −80 °C.

VeroE6/TMPRSS2 cells (JCRB Cell Bank, #JCRB1819) were maintained in D-MEM (low glucose) (Fujifilm Wako, #041-29775) with 10% heat-inactivated FBS (Nichirei Biosciences), 1 mg/mL geneticin (Thermo Fisher Scientific, #10131-027), and 100 U/mL penicillin-streptomycin (Thermo Fisher Scientific). For neutralization assays, cells were switched to D-MEM (high glucose) (Fujifilm Wako Pure Chemicals, #044-29765) during the infection period. Heat-inactivated plasma was serially diluted (five 6-fold dilutions starting at 1:10) and incubated with pseudotyped virus for 1 h at 37 °C. Mixtures were then added to VeroE6/TMPRSS2 cells plated in 96-well white flat-bottom plates (Corning, #3917) and incubated for 24 h at 37 °C with 5% CO₂.

Luminescence was measured using the Bright-Glo Luciferase Assay System (Promega, #E2610) on a GloMax Discover microplate reader (Promega, #GM3000). IC₅₀ values were calculated using Prism 10 (GraphPad). Titers below the detection limit (IC₅₀ = 20) were set to 20, with the detection threshold indicated in the figures.

Cytokine/chemokine and HMGB1 measurements

Serum cytokine and chemokine concentrations were measured using Luminex-based multiplex assays. These included the Bio-Plex Pro Mouse Chemokine Panel 31-Plex (Bio-Rad, #12009159), the Mouse IL-1α ProcartaPlex™ Simplex Kit (Invitrogen, #EPX01A-20611-901), and the Mouse IFN-α/IFN-β ProcartaPlex™ Panel 2-plex (Invitrogen, #EPX02A-22187-901). Assays were performed according to the manufacturers’ instructions. Serum samples were diluted as specified by each kit and incubated with antibody-conjugated beads. Following incubation and washing steps, biotinylated detection antibodies and streptavidin-PE were sequentially added. Beads were resuspended in assay buffer, and fluorescence intensities were acquired on a Bio-Plex 3D System (Bio-Rad) or xMAP INTELLIFLEX DR-SE (Thermo Fisher Scientific). Serum IL-1RA concentrations were measured using the Mouse IL-1RA Sandwich ELISA Kit (Proteintech, #KE10052) according to the manufacturer’s instructions. Concentrations were calculated by nonlinear regression (four-parameter logistic curve fitting) using GraphPad Prism software.

Plasma HMGB1 levels were measured using a mouse HMGB1 ELISA kit (Wuhan Fine Biotech, #EM0382) according to the manufacturer’s instructions. Briefly, standards and plasma samples (diluted 1:2) were added to the plate and incubated at 37 °C for 90 min. After washing with the provided wash buffer, anti-mouse HMGB1 was added and incubated at 37 °C for 1 h. Plates were washed and incubated with HRP-conjugated antibody for 30 min at 37 °C. Following additional washes, TMB substrate solution was added and incubated in the dark for 15 min. The reaction was stopped, and absorbance was measured at 450 and 570 nm using a microplate reader (BioTek Epoch 2, Agilent Technologies). Concentrations were calculated by nonlinear regression (four-parameter logistic curve fitting) using GraphPad Prism software.

Flow cytometry

For the analysis of immune cells at the injection site, single-cell suspensions were washed with PBS (Fujifilm Wako, #166-23555) containing 0.5% BSA (Sigma-Aldrich,) and 5 mM EDTA (Thermo Fisher Scientific) (FACS buffer). After Fc block with mouse TruStain FcX (anti-mouse CD16/32, BioLegend, #101320) for 5 min at room temperature, cells were stained in FACA buffer and Brilliant Stain Buffer Plus (BD Biosciences, #566385) with the following antibodies for 2 h at 4 °C: Zombie Yellow™ Fixable Viability Kit (BioLegend, #423104; 1:300), BUV395-conjugated anti-CD45 (30-F11, BD Biosciences, #564279; 1:150), AF700-conjugated anti-TER-119 (TER-119, BioLegend, #116220; 1:300), PE/Cy5-conjugated anti-CD3ε (145-2C11, BioLegend, #100310; 1:300), APC/Cy7-conjugated anti-CD19 (6D5, BioLegend, #115530; 1:300), BUV805-conjugated anti-NK1.1 (PK136, BD Biosciences, #741926; 1:300), BUV496-conjugated anti-CD4 (GK1.5, BD Biosciences, #612952; 1:300), BUV615-conjugated anti-CD8α (53–6.7, BD Biosciences, #613004; 1:300), BV750-conjugated anti-Ly6G (1A8, BD Biosciences, #747072; 1:300), PE-conjugated anti-Siglec-F (S17007L, BioLegend, #155506; 1:300), BUV737-conjugated anti-CD317 (927, BD Biosciences, #749272; 1:300), APC-conjugated anti-Siglec-H (551, BioLegend, #129612; 1:300), BV480-conjugated anti-CD11b (M1/70, BD Biosciences, #566117; 1:300), PE/Cy7-conjugated anti-F4/80 (BM8, BioLegend, #123114; 1:300), BB700-conjugated anti-CD11c (HL3, BD Biosciences, #566504; 1:300), and BV605-conjugated anti-Ly6C (AL-21, BD Biosciences, #563011; 1:300). Cells were washed, passed through a 100-µm mesh (CS CRIE, #R-CSC019-01), and analyzed on a FACSymphony A3 (BD Biosciences) using FACSDiva v9.1 (BD Biosciences).

For the analysis of antigen-specific GC/memory B cells, antigen probes were prepared as previously described in ref. ⁴⁰. Briefly, biotinylated spike (Acrobiosystems, #SPN-C82E9-200ug) and RBD proteins (prepared in-house) were conjugated with APC-streptavidin (Invitrogen, #S32362), R-PE streptavidin (Invitrogen, #S21388) in a 4:1.5 ratio, and incubated overnight at 4 °C. After Fc block, lymph node cells were stained overnight at 4 °C with spike/RBD probes together with the following antibodies in RPMI 1640 supplemented with 10 µM biotin: Zombie Yellow Fixable Viability Kit (BioLegend; 1:300), BV510-conjugated anti-IgM (II/41, BD Biosciences, #743324; 1:500), BV510-conjugated anti-IgD (11–26 c.2a, BioLegend, #405723; 1:500), BUV737-conjugated anti-CD11b (M1/70, BD Biosciences, #612800; 1:500), BUV737-conjugated anti-TER-119 (TER-119, BD Biosciences, #741736; 1:500), BUV737-conjugated anti-CD5 (53-7.3, BD Biosciences, #612809; 1:500), BUV737-conjugated anti-CD43 (S7, BD Biosciences, #612840; 1:500), BUV615-conjugated anti-B220 (RA3-6B2, BD Biosciences, #569724; 1:500), BV750-conjugated anti-CD138 (281-2, BD Biosciences, #747070; 1:500), PE/Dazzle594-conjugated anti-CD38 (90, BioLegend, #102730; 1:500), and Pacific Blue-conjugated anti-GL7 (GL7, BioLegend, #144614; 1:500), along with decoy streptavidin-BV786 (BioLegend, #405249; 1:500). Cells were analyzed on a FACSymphony A3 with FACSDiva v9.1 (BD Biosciences).

For intracellular cytokine staining (ICS), splenocytes stimulated in vitro with recombinant HMGB1 were processed as described above, washed with FACS buffer, and stained with the following antibodies: LIVE/DEAD Fixable Yellow Viability Dye (Thermo Fisher Scientific, #L34968; 1:300), APC/Cy7-conjugated anti-CD19 (6D5, BioLegend; 1:150), BUV395-conjugated anti-CD45 (30-F11, BD Biosciences; 1:300), BUV805-conjugated anti-NK1.1 (PK136, BD Biosciences; 1:300), BB700-conjugated anti-CD11c (HL3, BD Biosciences; 1:300), PE/Cy5-conjugated anti-CD3ε (145-2C11, BioLegend; 1:300), BV480-conjugated anti-CD11b (M1/70, BD Biosciences; 1:300), BV605-conjugated anti-Ly6C (AL-21, BD Biosciences; 1:300), BV750-conjugated anti-Ly6G (1A8, BD Biosciences; 1:300), and RB744-conjugated anti-F4/80 (T45-2342, BD Biosciences, #570508; 1:300). Cells were then fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific, #00-5523) and stained intracellularly with PE-conjugated anti-IL-1α (ALF-161, Invitrogen/eBioscience, #12-7011-82; 1:300), PE/Cy7-conjugated anti-IL-1β (NJTEN3, Invitrogen/eBioscience, #25-7114-82; 1:300), Alexa Fluor 488-conjugated anti-IL-6 (MP5-20F3, BD Biosciences, #561363; 1:300), BV421-conjugated anti-TNF-α (MP6-XT22, BD Biosciences, #563387; 1:300), and Alexa Fluor 647-conjugated anti-IFN-γ (XMG1.2, BD Biosciences, #557735; 1:300).

For neutrophil depletion assays, leukocytes from peripheral blood were stained with Zombie Yellow Fixable Viability Kit (BioLegend; 1:300), BV510-conjugated anti-TER-119 (TER-119, BioLegend, #116237; 1:300), BUV395-conjugated anti-CD45 (30-F11, BD Biosciences; 1:150), PE/Cy5-conjugated anti-CD3ε (145-2C11, BioLegend; 1:300), APC/Cy7-conjugated anti-CD19 (6D5, BioLegend; 1:300), PE-conjugated anti-NKp46 (29A1.4, BioLegend, #137604; 1:300), and BV750-conjugated anti-Ly6G (1A8, BD Biosciences; 1:300). Neutrophils were identified as CD45⁺TER-119^-CD3⁻CD19⁻NKp46⁻Ly6G⁺ cells.

For T cell analysis, stimulated cells were washed with PBS containing 2% heat-inactivated FBS and 0.01% NaN3, and incubated on ice for 60 min with Fc block (anti-CD16/32, in-house clone 2.4G2, 5 µg/mL). Cells were then stained with the following antibodies: LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (Thermo Fisher Scientific, #L10119; 1:1000), BUV805-conjugated anti-CD3 (17A2, BD Biosciences, #569192; 1:100), BV510-conjugated anti-CD4 (RM4-4, BioLegend, #116025; 1:100), BUV496-conjugated anti-CD8a (53–6.7, BD Biosciences, #569181; 1:100), BV711-conjugated anti-CD25 (PC61, BioLegend, #102049; 1:50), PE/Cy7-conjugated anti-CXCR5 (L138D7, BioLegend, #145516; 1:100), BV785-conjugated anti-PD-1 (29 F.1A12, BioLegend, #135225; 1:100), FITC-conjugated anti-CD90.2 (30–H12, BioLegend, #105306; 1:100), and BV605-conjugated anti-OX40 (OX-86, BioLegend, #119419; 1:100). Intracellular staining for Foxp3 was performed after fixation and permeabilization using the Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific), followed by BV421-conjugated anti-Foxp3 (FJK-16s, Thermo Fisher Scientific, #404-5773-82; 1:250). Data were acquired on a FACSymphony S6 (BD Biosciences) and analyzed with FlowJo software (BD Biosciences). Spike-specific Tfh, effector CD4⁺, and effector CD8⁺ cells were defined as CD3⁺CD4⁺CD8^-CXCR5⁺Foxp3^-PD-1⁺CD90⁺OX40⁺CD25⁺, CD3⁺CD4⁺CD8^-CXCR5^-Foxp3^-OX40⁺CD25⁺, and CD3⁺CD4^-CD8⁺IFN-γ⁺TNF-α⁺ cells, respectively.

Activation-induced marker (AIM) and intracellular cytokine staining assays

Single-cell suspensions obtained from lymph nodes were plated at 1.5 × 10⁶ cells per well in 96-well U-bottom plates in RPMI 1640 medium containing 10% heat-inactivated FBS (HyClone, #SH30396.03), 55 µM 2-mercaptoethanol (Thermo Fisher Scientific, #21985023), 1% minimum non-essential amino acid (Thermo Fisher Scientific, #11140050), 1% GlutaMAX Supplement (Thermo Fisher Scientific, #35050061), 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (Nacalai Tesque, #17557-94), 1 mM sodium pyruvate (Nacalai Tesque, #06977-34), and 100 U/mL penicillin-streptomycin (Thermo Fisher Scientific). The cells were stimulated with overlapping peptide pools (15 oligomers with an 11-amino acid overlap) spanning the spike glycoprotein of the wild-type strain (JPT Peptide Technologies, #PM-WCPV-S-2) at a final concentration of 1 µg/mL for 20 h at 37 °C. An equivalent volume of dimethyl sulfoxide solvent alone was used as a mock negative control. Following stimulation, the cells were harvested, stained as described above, and used for flow cytometry analysis. OX40 and CD25 were used as activation-induced markers to identify antigen-specific Tfh and effector CD4⁺ T cells, whereas antigen-specific effector CD8 + T-cell responses were assessed by intracellular cytokine staining (ICS) for IFN-γ and TNF-α. Antigen-specific T cell responses were quantified by subtracting the frequency of marker-positive cells in the mock negative control condition from that in the peptide-stimulated condition.

Transcriptome analysis

Total RNA was extracted from the mouse gastrocnemius muscle. RNA quality assessment, library preparation using the TruSeq Stranded mRNA kit (Illumina), and paired-end sequencing (2 × 100 bp, ~4 Gb per sample) on the NovaSeq 6000 platform were performed by Macrogen Japan according to the company’s standard protocols. Raw sequencing reads (FASTQ files) were quality-checked using FastQC v0.11.9, and adapter trimming and quality filtering were performed with Trimmomatic v0.39. The filtered reads were aligned to the mouse reference genome (GRCm38) using STAR v2.7.3a. Gene expression was quantified using featureCounts v2.0.1 to generate a count matrix. Differential expression analysis was conducted using DESeq2 implemented in iDEP v1.13⁴¹.

Quantification and statistical analysis

All statistical analyses were performed in GraphPad Prism 10. Statistical methods included Spearman’s rank correlation, two-way ANOVA with Greenhouse–Geisser correction, one-way ANOVA with Tukey’s or Dunnett’s test, and Welch’s t-test as appropriate. Multiple testing corrections were performed using the Benjamini–Hochberg method. RNA-seq analysis employed DESeq2, with FDR < 0.1 and fold change ≥ 2 as thresholds. Significance was defined as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Details of replicates and statistical tests are provided in figure legends.

Author contributions

T.T., T. Matsumura, and Y.T. conceptualized the study and drafted the original manuscript. T.T., K. Kumagai, H.I., K.T., A.M., E.S., Y.A., R.K., S.M., K. Saito, M.I., K. Soga, H.N.A., and S.H. developed the methodology. T.T., K. Kumagai, H.I., K.T., E.S., Y.A., R.K., M. Fujisawa., T. Mizukami, and T. Matsumura performed the experiments. A.M., K. Kobiyama and K.J.I. provided the experimental materials. S.O. and M.S. provided the clinical data and samples. T.T., K. Kumagai, H.I. and K.T. analyzed the data. T.T. visualized the data. T.T., M.H., T. Matumura, and Y.T. acquired funding. M.H., M. Fukasawa, M.S., T. Matsumura, and Y.T. supervised this study. T.T., K. Kumagai, H.I., K.T., T. Matsumura, and Y.T. revised the original draft and contributed to manuscript finalization. All authors critically reviewed the manuscript and approved the final version.

Data availability

This study did not generate new unique reagents. All mRNAs and LNPs were synthesized as described, and plasmid constructs are available upon request. RNA-seq data have been deposited in the DNA Data Bank of Japan (DDBJ) under BioProject accession number PRJDB19162, with individual run accession numbers DRR614477–DRR614512. Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Yoshimasa Takahashi (takahashi.yos@jihs.go.jp).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41541-026-01441-9.