Lipid Nanoparticles Elicit Reactogenicity and Sickness Behavior in Mice via Toll-Like Receptor 4 and Myeloid Differentiation Protein 88 Axis

Abstract

mRNA therapeutics encapsulated in lipid nanoparticles (LNPs) offer promising avenues for treating various diseases. While mRNA vaccines anticipate immunogenicity, the associated reactogenicity of mRNA-loaded LNPs poses significant challenges, especially in protein replacement therapies requiring multiple administrations, leading to adverse effects and suboptimal therapeutic outcomes. Historically, research has primarily focused on the reactogenicity of mRNA cargo, leaving the role of LNPs understudied in this context. Adjuvanticity and pro-inflammatory characteristics of LNPs, originating at least in part from ionizable lipids, may induce inflammation, activate toll-like receptors (TLRs), and impact mRNA translation. Knowledge gaps remain in understanding LNP-induced TLR activation and its impact on induction of animal sickness behavior. We hypothesized that ionizable lipids in LNPs, structurally resembling lipid A from lipopolysaccharide, could activate TLR4 signaling via MyD88 and TRIF adaptors, thereby propagating LNP-associated reactogenicity. Our comprehensive investigation utilizing gene ablation studies and pharmacological receptor manipulation proves that TLR4 activation by LNPs triggers distinct physiologically meaningful responses in mice. We show that TLR4 and MyD88 are essential for reactogenic signal initiation, pro-inflammatory gene expression, and physiological outcomes like food intake and body weight—robust metrics of sickness behavior in mice. The application of the TLR4 inhibitor TAK-242 effectively reduces the reactogenicity associated with LNPs by mitigating TLR4-driven inflammatory responses. Our findings elucidate the critical role of the TLR4-MyD88 axis in LNP-induced reactogenicity, providing a mechanistic framework for developing safer mRNA therapeutics and offering a strategy to mitigate adverse effects through targeted inhibition of this pathway.

INTRODUCTION

The development of messenger RNA (mRNA) vaccines demonstrates the power of mRNA therapeutics in preventing infectious diseases. ^{1, 2} However, the potential therapeutic applications of mRNA-based treatments go beyond vaccine development. mRNA therapeutics encompass transient protein replacement and supplementation achieved through the delivery and expression of the in vitro transcribed (IVT) mRNA. ³ Diseases that require personalized and precise interventions, such as various cancers, neurodegenerative disorders, and certain autoimmune conditions, stand to benefit from the innovative capabilities of mRNA therapeutics. ^4–6 The therapeutic action of IVT mRNA is predominately accomplished through lipid nanoparticle (LNP)-mediated delivery and intracellular translation of mRNA molecules. ⁷ mRNA LNP vaccines promote immunity via the immunogenic properties of antigen proteins synthesized from IVT mRNA. In IVT mRNA-based protein replacement therapies, conversely, the expressed proteins serve to replace aberrant or deficient endogenous proteins, thereby providing therapeutic benefits. While immunogenicity is an anticipated characteristic of mRNA LNP vaccines, mRNA LNPs’ reactogenicity results in suboptimal therapeutic outcomes, decreased vaccine potency, and adverse effects, including pain, fever, chills, and fatigue. ^8–12 Therefore, reactogenicity is undesirable, especially in IVT mRNA-based protein replacement therapies that necessitate repeated administrations.

Real-world data on mRNA vaccines reveal that the reactogenicity profiles of mRNA LNP formulations, including adverse effects, significantly influence patients’ decisions about vaccination and adherence to immunization schedules. ¹³ Vaccine hesitance and partial immunization were linked to COVID-19 vaccine side effects and response severity after the initial dose. ¹⁴ The inflammatory nature of LNPs used in mRNA vaccines is well-documented. For instance, LNPs can cause severe injection site inflammation, broad biodistribution, and are found in multiple tissues including the brain in both humans and laboratory animals. ^{15, 16} Additionally, LNPs can induce rapid and robust inflammatory responses characterized by neutrophil infiltration and activation of inflammatory pathways, which contribute to the side effects observed in humans. ¹¹ Other studies have linked mRNA LNP vaccines to various adverse events including cardiac tissues inflammation, anaphylaxis, arthralgias and myalgias, due to the pro-inflammatory nature of ionizable cationic lipids used in these formulations. ^17–19 Furthermore, systemic off-target gene expression has been identified as a primary cause of acute adverse reactions and side effects associated with nucleoside-modified mRNA vaccines. ²⁰ Co-administration of cyclooxygenase (COX) inhibitors or dexamethasone can alleviate specific reactogenicity-related symptoms following mRNA LNP administration. However, this transient immunosuppression does not address the concerns related to decreased vaccine potency caused by mRNA degradation or reduced mRNA translation due to translational stalling, which are also consequences of the reactogenicity of the administered formulations. ²¹ Even in the context of a single mRNA vaccine administration, where an adaptive immune response is the foremost goal, innate immune responses initiate mRNA degradation and diminish mRNA translation to immunizing protein. ^{22, 23} Notably, these manifestations are not dependent on IVT mRNA as a sole reactogenic species.

Empty lipid nanoparticles (eLNPs) possess intrinsic adjuvant activity, as evidenced by their ability to induce immune responses in the absence of mRNA cargo. ²⁴ Reactogenic and immunologic responses to LNP components, including polyethylene glycol (PEG) and ionizable lipids, reduce the expression of therapeutic mRNA. ^{25, 26} Additionally, the research found that eLNPs containing ionizable lipids, used in the Moderna COVID-19 vaccine, led to an increase in levels of pro-inflammatory cytokines in purified monocytes, including interleukin 1B (IL-1β). ²⁴ Moreover, the LNP vehicle’s adjuvanticity and pro-inflammatory characteristics are hypothesized as probable causes of COVID-19 mRNA vaccines’ side effects, as opposed to the immunological responses induced by the translated immunogen. ¹¹ Importantly, ionizable lipids cause inflammation characterized by the expression of signature pro-inflammatory cytokines, including interleukin 1B and 6 (IL-1β, IL-6), and chemokines, such as chemokines CC motif ligand 2 and 4 (CCL2, CCL4), and chemokines (C-X-C motif) ligand 2 and 10 (CXCL2, CXCL10).¹¹ Although the precise mechanism by which LNPs initiate the chemokine-cytokine cascade has yet to be determined, some cationic lipids activate the immune system via binding to toll-like receptors (TLR). ^{22, 27, 28} Additionally, ionizable lipids in several Food and Drug Administration (FDA)-approved LNP formulations share structural similarities with lipid A, a key component of lipopolysaccharide (LPS) and a known TLR4 agonist. Moreover, saturated lipids, including those used as helper lipids in LNP formulations, including distearoylphosphatidylcholine (DSPC), have been shown to activate TLR4. Previous studies demonstrated that saturated fatty acids, such as palmitate and lauric acid, trigger TLR4-mediated proinflammatory signaling cascades, contributing to inflammation and metabolic disorders like insulin resistance and metabolic syndrome. ^29–31 Given these structural and functional parallels between LNP components and known TLR4 activators, we hypothesized that TLR4 serves as a critical pattern recognition receptor for LNPs, mediating their immunostimulatory effects. This hypothesis provides a mechanistic link between the lipid composition of LNPs and their potential to elicit innate immune responses, which may have significant implications for the safety and efficacy of LNP-based therapeutics.

The activation of TLR4 is implicated in the production of cytokines associated with the inflammation-driven adverse effects of LNP formulations and in the diminishing expression of mRNA delivered by LNP carriers. ²⁵ Furthermore, TLR4 over-activation has been implicated in various inflammatory and autoimmune diseases, including sepsis, atherosclerosis, liver and renal fibrosis, and rheumatoid arthritis. ^{30, 32–38} Additionally, excessive TLR4 signaling can exacerbate existing conditions involving inflammation, including cancers, acute kidney injury, obesity-associated insulin resistance, colitis, liver ischemia/reperfusion injury, and lipid-induced insulin resistance. ^{29, 38–43} Therefore, the administration of LNPs to individuals with pre-existing inflammatory conditions could potentially lead to disease exacerbation through TLR4 activation. While it is established that concurrent TLR4 activation by LPS decreases mRNA LNP efficacy by reducing mRNA expression, it remains unknown whether LNPs themselves activate TLR4 signaling.

Since TLR4 adaptor proteins are gatekeepers to the expression of multiple cytokines, including IL-1β, IL-6, tumor necrosis factor-alpha (TNF-α), and lipocalin 2 (LCN2), as well as secretion of type I interferons, deconvolution of the complex pathways downstream of TLR4 is important for improving mRNA LNP reactogenicity profiles. It is well established that distinct adaptors promote TLR4 dimerization in different cell compartments. The adaptor protein myeloid differentiation primary response 88 (MyD88) is required for TLR4 dimerization on the plasma membrane, whereas the adaptor protein TIR-domain-containing adapter-inducing interferon-β (TRIF) is involved in TLR4 dimerization in the endosomal compartment, as reviewed in ^{44, 45}. Differential TLR4 dimerization and association with distinct adaptor proteins could provide insights into whether LNPs activate TLR4 on the plasma membrane or inside the endosome, revealing the preferred pathway of reactogenic signal propagation. Furthermore, MyD88 adaptor is crucial for the adjuvant properties of eLNPs, as MyD88 deficient mice showed no response in terms of follicular T-helper and germinal center B-cell upregulation. ²⁴ Given the importance of MyD88 in mediating the adjuvant effects, further investigation is needed to elucidate the full pathway of innate immune signal initiation and propagation through both MyD88 and TRIF-dependent pathways, and how these ultimately lead to the activation of adaptive immunity. Understanding these signaling cascades could provide valuable insights into the mechanisms by which eLNPs enhance vaccine efficacy and potentially guide the development of more potent adjuvant formulations.

In addition to investigating the role of TLR4 and its adaptors in LNP-induced reactogenicity, a gap exists in understanding the physiologically relevant innate immune responses triggered by empty LNP (eLNP) vehicles activating TLR4. Our study focuses on animal sickness behavior as a measurable physiological response to acute or chronic inflammation, elucidating TLR4 downstream signaling leading to sickness outcomes induced by LNP administration. We establish that the activation of TLR4 by an eLNP formulation induces distinct physiological responses in mice. Moreover, our findings highlight the indispensable role of the MyD88 adaptor protein in the development of sickness behavior triggered by LNP administration in mice. Altogether, our work fills a knowledge gap and emphasizes the crucial need to understand these details for enhancing the effectiveness and safety of mRNA LNP therapeutics, especially for treatments that require repeated administrations.

RESULTS AND DISCUSSION

Dosing strategies for induction of murine sickness behavior

To enhance LNP formulation safety, we aimed to distinguish immune responses triggered by eLNPs from those induced by mRNA cargo. Although therapeutic considerations often treat mRNA cargo and LNP vehicles as a combined entity, it is essential to recognize the unique contribution of eLNPs in initiating and amplifying reactogenic manifestations (Figure 1 A). This understanding is crucial to supplement existing safety measures applied for mRNA, such as the incorporation of pseudouridine substitutions. Therefore, we carefully selected the eLNP formulation, administration route, and dosage regimens to systematically investigate eLNP-dependent reactogenicity in mice.

Figure 1. MC3-containing eLNPs exhibit the most robust reactogenic profile, as evidenced by comprehensive molecular analysis and physiological responses.

(A) Schematic representation of mRNA LNPs (left) and eLNPs (right) (B) Representative bioluminescence images of organs from the wild-type mice injected with Luc mRNA LNP formulation following 6 hours after IP administration of three LNP formulations, containing ALC-0315, SM-102, and MC3 ionizable lipids. (C) Bioluminescence signal from peritoneal lavage samples following administration of MC3-, SM-102, and ALC-0315-containing Luc mRNA LNPs at 2-, 4-, and 6-hours post-injection. (D) Hepatic gene expression following IP administrations of Luc mRNA LNP formulations, normalized to PBS-treated mice whose expression is equal to one. (E) Body weight change following IP administration of eLNP formulations. Statistics: (D) Expressed as mean ± SEM, n=5, *p<0.05, analyzed by one-way ANOVA followed by Bonferroni’s post hoc test. (E) Expressed as mean ± SEM normalized to baseline, n=5, *p<0.05 for MC3 vs. PBS, analyzed by two-way repeated measures ANOVA.

In investigating LNP formulation reactogenicity, we focused on changing a single variable, the ionizable lipid, aiming to understand its influence on the overall LNP’s reactogenic characteristics. We investigated DLin-MC3-DMA (MC3), ALC-0315, and SM-102, all of which are integral to FDA-approved LNP formulations. Specifically, MC3 is associated with Patisiran, siRNA therapy for hereditary transthyretin amyloidosis, ALC-0315 is a constituent of the Pfizer COVID-19 vaccine, and SM-102 is utilized in Moderna’s COVID-19 vaccine. ^{46, 47}

We first used Luciferase (Luc) mRNA LNPs to monitor the biodistribution of each formulation, distinguished by the ionizable lipid they incorporated and the administration route (SI Figure 1 A–D). Both ALC-0315- and MC3-containing formulations produced a high bioluminescence signal in the liver following 6-hours-post intraperitoneal (IP) administration (Figure 1B). Notably, through IP administration, we observed that the Luc mRNA LNP formulation containing SM-102 resulted in the highest influx of innate immune cells that were extravasating to the peritoneum, validated by the bioluminescence in peritoneal lavage (Figure 1 C, SI Figure 1 E). Additionally, SM-102 LNP administration prompted a pronounced immune response, recruiting considerably higher numbers of innate cells with greater proportions of macrophages and neutrophils, assessed with fluorescence-assisted cell sorting (FACS), and induced local overexpression of pro-inflammatory cytokines by infiltrating cells (SI Figure 1 F, G). Following IP injection, the MC3 formulation resulted in a moderate influx of innate immune cells to the peritoneum while concurrently exhibiting high mRNA transfection in hepatic tissues, as evidenced by bioluminescence data. (Figure 1 B, C; SI Figure 1 D–F).

Further testing using eLNPs containing MC3 ionizable lipid demonstrated the most pronounced hepatic reactogenic outcome following IP injection, as evidenced by elevated levels of Il1b, Tnf, Il6, and Orm1 after three consecutive administrations (Figure 1 D). The activation of hepatic pro-inflammatory genes was mirrored by the onset and progression of murine sickness behaviors, exemplified by a decrease in body weight observed across all three studied eLNPs (Figure 1 E). Notably, the most pronounced reduction in body weight was observed in mice administered MC3 eLNPs (Figure 1 E). To strike a balance between adequately stimulating the immune system to induce behavioral changes and avoiding an excessive molecular response, we opted to evaluate the reactogenicity of MC3-containing eLNPs by administering doses equivalent to 5 μg of Luc mRNA LNPs per mouse through IP injection. While acknowledging the potential limitations of generalizing findings from a single eLNP formulation to the overall reactogenicity of LNPs, it is important to note that not only are the ionizable lipids in FDA-approved LNP formulations similar featuring tertiary amines with ester tails, but other LNP components also share structural similarities. Additionally, the methods used for LNP preparation are consistently applied across different formulations. ²¹

MyD88 is an indispensable adaptor protein propagating reactogenic signals following eLNP injections

We adopted a methodological approach to investigate TLR4 involvement in eLNP-induced reactogenicity. Utilizing genetic ablation studies targeting the toll-like receptor adaptors MyD88 and TRIF (Figure 2 A) allowed us to address fundamental questions regarding their roles in reactogenic induction. Additionally, these studies provided a platform to explore further questions, such as whether LNPs interact with TLR4 on the plasma membrane or within endosomes. This distinction is significant because MyD88 serves as an adaptor protein for TLR4 when it is located on the plasma membrane. However, when TLR4 undergoes conformational rearrangements due to acidification of the maturing endosome, MyD88 can be replaced by TRIF. ⁴⁸ Regardless of the location of the adaptor association, the absence of MyD88 or TRIF prevents the propagation of signals downstream of TLR4 in the membrane or in the endosome, respectively. Therefore, our genetic ablation studies, incorporating mice lacking these adaptor proteins, serve to elucidate the precise contributions of MyD88 and TRIF, providing insights into the dynamics of TLR4 interaction with eLNPs across various cellular compartments.

Figure 2. Adaptor protein MyD88 is necessary for reactogenic signal transduction following eLNP administration.

(A) Schematic representation of TLR4 interactions with MyD88 and TRIF adaptors, emphasizing proteins of interest for the current section. (B) A methodologic approach using genetic ablation to study acute and chronic reactogenic manifestations in WT, MyD88 KO, and TRIF KO mice. Percent change in (C) Daily food intake and body weight following MC3 eLNP chronic administration in WT, MyD88 KO, and TRIF KO mice. (D) Relative gene expression in MyD88 KO, TRIF KO, and WT murine livers at the baseline and after single and chronic MC3 eLNP injections. To simplify the graphics, genotype-matched controls treated with PBS were excluded, as their average consistently fluctuated around the 100% baseline. (E) Serum IL-1β levels in WT, TRIF KO, and MyD88 KO mice at the baseline (PBS) and following chronic MC3 eLNP administration (eLNP). (F) Serum LCN2 levels in WT, TRIF KO, and MyD88 KO mice at the baseline (PBS) and following single (1X) and chronic (3X) MC3 eLNP administrations. (G) Serum IL-6 levels in WT, TRIF KO, and MyD88 KO mice at the baseline (PBS) and following chronic MC3 eLNP administration. (H) Innate immune cell populations in peritoneal lavage samples at baseline (PBS) and following IP administrations of Luc mRNA LNP formulations (1X and 3X) for WT and MyD88 KO mice. (C) Expressed as mean ± SEM normalized to baseline, n=5, ****p<0.0001 analyzed by two-way repeated measures ANOVA. (D) Expressed as mean ± SEM, n=5, *p<0.05, **p<0.01, ***p<0.001, ****p<0.001 analyzed by two-way ANOVA followed by Bonferroni’s post hoc test. Intragroup significance compared to PBS normalized to a relative expression set to 1 within each genotype. Additional information on inter-group comparison is provided in SI Table 1. (E-G) Expressed as mean ± SEM, n=5, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 analyzed by one-way ANOVA followed by Bonferroni’s post hoc test. (H) Expressed as a proportion of total cells.

In our experimental setup, we included MyD88 knock-out (MyD88 KO), TRIF knock-out (TRIF KO), and wild-type (WT) mice, all of which shared a C57Bl/6 background (Figure 2 B). We subjected each genotype group to either a single injection or three consecutive daily IP injections, allowing us to assess both acute and chronic reactogenic manifestations of eLNPs. Liver, peritoneal lavage, and serum samples were collected 6 hours after the last injection for all groups.

Given that food intake and body weight are robust indicators of an animal’s sickness response, we utilized these metrics to observe how mice responded to three consecutive daily injections of eLNPs (Figure 2 C). Notably, the deletion of MyD88 completely normalized food intake and body weight to the pre-injection baseline. However, both WT mice and TRIF KO mice exhibited a decrease in both parameters. Using food intake reduction as a sensitive method to detect ongoing sickness in mice, we observed no significant difference between WT and TRIF KO mice. However, there was a statistically significant difference between these two groups and MyD88 KO mice. These findings underscore the specific influence of MyD88, but not TRIF, on the sickness response in mice following eLNP injections.

Furthermore, chronic eLNP administration triggered a pronounced neuroinflammatory response, which varied significantly across genotypes. WT mice exhibited the most robust response, followed by TRIF KO mice, while the response was markedly attenuated in MyD88 KO mice (SI Figure 2). This differential pattern was evident in the expression of key inflammatory mediators in the hypothalamus. Following a single eLNP treatment, both WT and TRIF KO mice showed significantly elevated levels of Il1β and Lcn2 compared to MyD88 KO mice. Lcn2 was upregulated by 58-fold and 14-fold, while Il1β was upregulated by 14-fold and 3 fold in WT and TRIF KO mice, respectively. The inflammatory response in WT mice persisted with chronic eLNP injections, as evidenced by sustained elevation of Il1β (8-fold) and increased Il6 expression. Notably, Il6 levels were elevated in WT mice after both single (624-fold) and chronic (27-fold) eLNP administrations. These pro-inflammatory cytokines, well-established markers of neuroinflammation, play crucial roles in modulating neural function and behavior. ⁴⁹ Additionally, the immediate-early gene Fos, encoding the c-FOS protein, serves as a robust marker of neuronal activation. Our analysis revealed a distinct pattern of Fos expression across genotypes in response to eLNP administration. WT and TRIF KO mice exhibited a significant 2-fold decrease in Fos expression (SI Figure 2), indicative of suppressed neuronal activity associated with prolonged chronic stress. ^{50, 51} In stark contrast, MyD88 KO mice maintained stable Fos expression levels, suggesting that the MyD88-dependent pathway is crucial for eLNP-induced neuronal suppression. This differential Fos response further underscores the pivotal role of MyD88 in mediating the neuroinflammatory effects of eLNPs and their subsequent impact on neuronal activity. The observed differential responses across genotypes underscore the importance of MyD88-dependent signaling in mediating the neuroinflammatory response to eLNPs.

Given the critical role of MyD88 adaptor protein in LPS-induced hypothalamic inflammation and associated anorexia, ⁵² we investigated the expression of appetite-regulating neuropeptides in MyD88 KO and WT mice following single and chronic eLNP injections. The observed neuroinflammatory state in WT mice was associated with significant alterations in the expression of key appetite-inducing (orexigenic) hypothalamic neuropeptides, neuropeptide Y (NPY) and agouty-related peptide (AgRP). Release of cytokines, such as IL-1β, IL-6, and TNF-α decreases NPY and AgRP expression to decrease food intake. ⁴⁹ In addition to the observed upregulation of Il1β and Il6 in the hypothalamus of WT mice, we detected a significant 10-fold and 4-fold increase in Tnf expression following single and chronic eLNP injections, respectively (SI Figure 3). Tnf, originally termed “cachectin” due to its critical role in cachexia, is associated with the wasting syndrome frequently observed in chronic inflammatory conditions. ⁵³ Cachexia is characterized by progressive loss of body mass, diminished appetite, and reduced physical function. ⁵⁴ The concurrent elevation of Tnf alongside Il1β and Il6 not only corroborates the robust inflammatory response elicited by eLNPs but also suggests a potential mechanism underlying the observed alterations in feeding behavior and body composition (Figure 2 C). Importantly, following chronic eLNP injections, WT mice exhibited a marked reduction in Npy (5-fold) and Agrp (10-fold) expression, while MyD88 KO mice maintained stable expression of Npy with 4-fold increase in Agrp expression (SI Figure 3). NPY and AgRP, produced in the arcuate nucleus of the hypothalamus, are orexigenic neuropeptides crucial for stimulating appetite. ⁴⁹ The reduction in Npy and Agrp gene expression observed in WT mice suppresses the drive to feed, leading to reduced food intake and weight loss, that we observe in our food intake and body weight measurements (Figure 2 C). These results highlight the complex interplay between eLNP-induced inflammation, hypothalamic gene expression and feeding behavior in mice. The chronic inflammatory state induced by repeated eLNP administration likely disrupts the normal functioning of hypothalamic neurons, interfering with signaling pathways that regulate appetite and energy homeostasis.

Given the pronounced tropism of eLNPs for the liver, we focused on assessing the expression of hepatic inflammatory genes to investigate the primary site of immune response induction (Figure 2 D, SI Table 1). Notably, Il1B exhibited a more significant upregulation in WT and TRIF KO mice than in MyD88 KO mice, particularly under acute administration conditions with single eLNP administration. Il6 showed a slight upregulation in WT and TRIF KO mice, but no apparent increase was observed in MyD88 KO mice (SI Figure 4). Lcn2, a marker of propagating inflammation expressed by neutrophils and endothelial cells, ⁵⁵ demonstrated robust and sustained upregulation in WT and TRIF KO mice after single and chronic eLNP injections. Conversely, in MyD88 KO mice, Lcn2 expression remained at the same level as in control MyD88 KO mice treated with PBS after a single injection. An increase was observed only after chronic injections, yet it remained approximately seven times lower than WT and TRIF KO animals.

The expression of Tnf demonstrated a pronounced upregulation in WT and TRIF KO mice compared to MyD88 KO mice following single eLNP administration. Our analysis revealed a potential delayed upregulation of Tnf expression in MyD88 KO mice after chronic eLNP administration, although this trend did not reach statistical significance (p=0.086). While MyD88-dependent signaling appears to be the primary driver of acute inflammatory responses, alternative or compensatory mechanisms may also emerge during chronic eLNP exposure. Notably, although Tnf is well-recognized as an anorexigenic factor, ⁵³ we observed decreased food intake and body weight after the first eLNP injection in WT and TRIF KO mice, highlighting the importance of the initial pro-inflammatory gene induction. This effect was not observed in MyD88 KO mice despite some Tnf upregulation following chronic dosing. Further investigation into these potential parallel signaling pathways could provide valuable insights into the long-term effects of eLNP administration and inform strategies to mitigate chronic inflammation in LNP therapeutic applications.

When administering LNP formulation that included mRNA cargo, we noted an increase in Il1b, Tnf, Il6, and Lcn2 levels in MyD88 KO mice after a single administration of Luc mRNA LNPs, mirroring the levels observed in TRIF KO mice (SI Figure 5). In contrast, WT mice demonstrated relatively lower levels of these markers, likely due to either a rapid, substantial overexpression of pro-inflammatory genes, which was followed by a quicker dampening of the signal, or a potential compensatory mechanism in MyD88 KO and TRIF KO mice, leading to substantial overexpression of these cytokines. Additionally, this observation underscores the importance of alternative innate defense mechanisms triggered by mRNA cargo, activating a distinct subset of innate immunity receptors beyond those requiring MyD88 and TRIF adaptors, such as RIG (Retinoic Acid-Inducible Gene) and MDA5 (Melanoma Differentiation-Associated protein 5). ⁵⁶ Despite increased levels of pro-inflammatory cytokines in MyD88 and TRIF KO mice, these animals exhibited increased quantities of Luc mRNA delivered by LNPs relative to WT control mice, implying that the activation of MyD88- and TRIF-dependent reactogenic pathways by both mRNA cargo and LNP vehicles diminishes the efficacy of mRNA LNP therapeutics. The observed increase in Luc mRNA levels in MyD88 KO mice may be attributed to the influence of eLNP-induced reactogenicity via MyD88-dependent pathways. This underscores the essential task of disentangling reactogenic responses induced by eLNPs from mRNA-dependent manifestations, particularly those induced by activation of TLR3 and either TLR7 (in mice) or TLR8 (in humans) that require MyD88 and TRIF adaptors. ⁵⁷

Returning to the discussion of eLNP-induced reactogenicity, we evaluated serum cytokine levels in mice subjected to eLNP injections. Following a single eLNP administration, serum IL-1β levels were undetectable; however, they increased in WT and TRIF KO mice after three chronic injections. Serum IL-1β was elevated in WT mice and increased to an even greater extent in TRIF KO mice (Figure 2 E). IL-1β was undetectable after single and chronic injections in MyD88 KO mice. TNF-α levels were below the limit of quantification in the serum of all three groups for both administration approaches (data not shown). Serum LCN2 levels in WT animals increased from 100 to almost 1000 ng/mL after a single injection (Figure 2 F). Following chronic injection, LCN2 levels continued to rise in WT mice, reaching up to 2500 ng/mL, and increased to 1500 ng/mL in TRIF KO mice. It is worth noting that we observed a positive trend in serum LCN2 increase in MyD88 KO mice, but the difference was not statistically significant. Additionally, we evaluated LCN2 levels in peritoneal lavage after single and chronic eLNP injections in WT, MyD88 KO, and TRIF KO mice (SI Figure 6). We noted increased LCN2 levels in peritoneal lavage samples from WT animals, indicating potential variations in the dynamics or magnitude of the response, a trend in line with LCN2 levels in the serum. The difference between MyD88 KO and TRIF KO became apparent after a single eLNP administration, resulting in lower LCN2 levels observed in the peritoneal lavage of MyD88 KO mice, which then plateaued after chronic administrations. While IL-6 increased in MyD88 KO mice, the elevation was notably lower compared to the other groups (Figure 2 G). These findings demonstrated the differential responses of inflammatory markers in serum after eLNP injections, with notable distinctions among MyD88 KO in comparison to TRIF KO and WT mice, and underscore the pivotal role of the MyD88 adaptor in propagating reactogenic signals, as evidenced by the observed differences between MyD88 KO mice and TRIF KO and WT mice.

To further explore the differences between MyD88 KO and WT animals, we examined the dynamics of innate immune cell recruitment to the peritoneum, comparing WT and MyD88 KO mice with FACS following eLNP injections. Due to the limited number of innate immune cells in the peritoneal cavity of healthy, unstimulated mice (SI Figure 7), insufficient for robust biochemical studies, and its conventional consideration as a sterile environment, we used PBS injections as our control. ⁵⁸ In WT mice, a single eLNP injection resulted in a significant increase in immune cell infiltration in the peritoneum, with counts increasing by 81 times compared to those following PBS injection. Similarly, chronic eLNP injections led to an increase in immune cell counts by 86 times in WT mice. In MyD88 KO mice, immune cell counts increased by 21 times following a single injection and by 27 times after chronic injections. Importantly, in WT mice, a single eLNP injection resulted in a marked increase in neutrophils (20.0% and 71.7% following PBS and single eLNP injection, respectively), making them the predominant population of innate cells in the peritoneum (Figure 2 H). In contrast, in MyD88 KO mice, the primary cell type remained macrophages/monocytes, maintaining a proportion similar to that observed following PBS injection (63.8% and 72.2% following PBS and single eLNP injections, respectively). Chronic injections in WT mice led to the recruitment of more macrophages/monocytes and dendritic cells, whereas in MyD88 KO mice, there was only a slight increase in monocytes/macrophages. Of note, single and chronic eLNP injections led to a rapid increase in blood leukocyte count in WT mice in comparison to MyD88 KO mice with a higher proportion of neutrophils in WT animals (SI Figure 8, SI Figure 9). Furthermore, we observed increased expression of Cxcl10 (4.5-fold), Tnf (12-fold), and Lcn2 (2.8-fold) in sorted neutrophils from WT murine peritonea after single eLNP injections with subsequent normalization following chronic eLNP injections (SI Figure 10). Strikingly, Il1b exhibited substantial upregulation (570-fold) in peritoneal neutrophils from WT mice after single injections and remained elevated (130-fold) after chronic injections. In contrast, peritoneal neutrophils from MyD88 KO mice displayed upregulated Tnf (7.5-fold) and Il1b (4-fold) only after chronic eLNP administration. Therefore, our observations suggest that WT mice exhibit a faster and more pronounced response of innate immune cells after single and repeated exposure to eLNPs, indicating a heightened reactivity and sensitization of the immune system to the formulation that requires an intact MyD88 adaptor.

In summary, we show that MyD88 is crucial in governing pro-inflammatory responses at the molecular level and propagating behavioral responses caused by eLNP reactogenicity. Our findings suggest that the propagation of reactogenic signals from eLNPs takes place earlier in the endosomal maturation process rather than later when MyD88 is replaced by TRIF; however, the addition of mRNA cargo changes this dynamic.

TLR4 ablation modifies immune response and gene expression patterns in mice subjected to eLNP administration

After establishing the crucial role of MyD88 in propagating reactogenic signals, we turned our attention to the TLR4 receptor itself (Figure 3 A). We started with genetic ablation studies using TLR4 knock-out (TLR4 KO) mice. Employing a similar dosage approach, both TLR4 KO and WT mice received either single or chronic injection regimens (Figure 3 B). As expected, food intake exhibited a marked decrease in WT mice following eLNP administration. This was attenuated, but not completely reversed, in TLR4 KO mice (Figure 3 C). Furthermore, TLR4 KO mice displayed reduced sensitivity to eLNP-induced changes in body weight, with no statistically significant difference observed after the third injection compared to WT mice subjected to PBS injections (Figure 3 C).

Figure 3. Genetic ablation studies with TLR4 KO mice demonstrated TLR4’s involvement in eLNP-induced reactogenicity in mice.

(A) Schematic representation of the TLR4 receptor’s role in reactogenic signal transduction, as a central focus for the current section. (B) A methodologic approach using genetic ablation to study acute and chronic reactogenic manifestations in WT and TLR4 KO mice. (C) Food intake and body weight following MC3 eLNP chronic administration in WT and TLR4 KO mice. WT mice subjected to PBS act as a negative control, and TLR4 KO mice receiving PBS were omitted from the graph for simplicity as they follow the same trend as the WT-PBS group. (D) Differentially expressed genes (DEGs) in WT and TLR4 KO murine livers normalized to their corresponding genetic control groups following a single MC3 eLNP injection. (E) Enriched terms based on DEGs for WT and TLR4 KO mice receiving single MC3 eLNP injection. (F) The heatmap depicts the hierarchical clustering of DEGs in WT and TLR4 KO murine livers following chronic MC3 eLNP injections, with the color scale from lowest to highest log2 Fold Change. DEG phylogenetic relationship is shown on the left of the heatmap. The top clustering indicates the relationship of the samples. Each heatmap is accompanied by the name list for DEGs. Accompanying each heatmap is a comprehensive list of the DEG names. On the right - volcano plots of significantly changed DEGs (p<0.05). (C) Expressed as mean ± SEM normalized to baseline, n=5, ****p<0.0001 analyzed by two-way repeated measures ANOVA. (D) and (F) Log2 Fold Change in the list is an average for n=3.

Given the LNPs’ high tropism for the liver, we investigated hepatic differentially expressed genes (DEGs) using a curated gene list associated with inflammation assessed using NanoString nCounter technology. We compared the hepatic gene expression profiles of eLNP-treated mice and PBS-treated controls of identical genotypes. We first looked at the DEGs in the liver in the setting of acute inflammation following a single eLNP injection and identified 50 and 34 DEGs in WT and TLR4 KO mice, respectively (Figure 3 D). Notably, Ifna1, Maff, and Mx1 displayed consistent downregulation in WT and TLR4 KO mice after a single injection. These genes are associated with the antiviral response, and their downregulation suggests a suppression of antiviral defenses, possibly influenced by the fact that LNPs mimic the milieu of viruses. Additionally, both WT and TLR4 KO mice exhibited upregulation in Ifl44 and Irf7 in the acute inflammation setting. In WT mice, there was a pronounced upregulation of complement cascade molecules (C6, C8a, C8b, and C9). The “C” molecules denote components of the complement system, integral to immune responses. Additionally, we observed a substantial increase in Cxcl1 in the WT group, a chemokine that plays a crucial role in inflammation, particularly the recruitment of neutrophils to sites of injury or infection. ⁵⁹ In contrast, TLR4 KO mice did not show upregulation in Cxcl1 and did not upregulate complement effector molecules, except for C3. TLR4 KO mice also exhibited an upregulation of the chemokine transcript Cxcl10, which subsequently returned to normal levels after chronic injections. Furthermore, TLR4 KO mice uniquely upregulated Ifit1, Ifit2, Ifit3, Oasl1, and Oas1a, responses not observed in WT mice.

NCATS BioPlanet ⁶⁰ enrichment analysis of differentially expressed genes (DEGs) revealed distinct pathway activation profiles in WT and TLR4 KO mice. In WT mice, we observed significant enrichment in pathways associated with the activation of the classical, lectin, and alternative complement cascades (Figure 3E, SI Table 2). Conversely, TLR4 KO mice exhibited a pronounced enrichment in interferon signaling pathways, encompassing both type I and type II interferons. Notably, following chronic eLNP injection, both WT and TLR4 KO mice showed an upregulation of DEGs associated with complement-related pathways, as evidenced by BioPlanet enrichment analysis (SI Table 2). These findings were further corroborated by additional pathway analyses, including Reactome ⁶¹ and Wiki Pathways ⁶², which consistently highlighted the differential activation of immune-related pathways between genotypes and in response to eLNP treatment (SI Table 2).

The MSigDB ⁶³ immune signature enrichment analysis revealed significant differences in immune cell type representation between TLR4 KO and WT mice (p < 0.05; SI Table 3). In acute settings following single eLNP injection, TLR4 KO mice exhibited enrichment of signatures associated with T cells (50 terms), dendritic cells (25 terms), B cells (12 terms), and regulatory T cells (10 terms). In contrast, WT mice demonstrated a more diverse immune profile, encompassing T cells (20 terms), dendritic cells (12 terms), macrophages (10 terms), neutrophils (5 terms), monocytes (5 terms), B cells (5 terms), NK cells (3 terms), and mast cells (3 terms). This broader representation suggests a more generalized acute immune response in WT mice. In chronic injection settings, the immune cell signature profiles became more comparable between the two groups. TLR4 KO mice showed enrichment in signatures related to T cells (87 terms), dendritic cells (22 terms), B cells (13 terms), peripheral blood mononuclear cells (PBMCs; 12 terms), macrophages (11 terms), thymocytes (11 terms), neutrophils (5 terms), monocytes (5 terms), mast cells (3 terms), and NK cells (1 term). Similarly, WT mice exhibited enrichment in signatures associated with T cells (41 terms), dendritic cells (9 terms), macrophages (8 terms), neutrophils (6 terms), B cells (5 terms), thymocytes (4 terms), monocytes (3 terms), microglia (3 terms), and NK cells (2 terms). These findings suggest that TLR4 deficiency alters the landscape of immune cell activation in response to eLNP administration, with notable differences in acute versus chronic exposure scenarios.

Notably, in WT mice, most DEGs were associated with the lectin-induced complement pathway, while TLR4 KO mice displayed genes linked to interferon signaling. Following chronic eLNP administration, the persistent sickness behavior in mice prompted us to identify hepatic DEGs after chronic eLNP injections (Figure 3 F). Chronic eLNP injections induced even more pronounced upregulation of complement pathway genes in WT mice, specifically C8a, C6, C8b, and C4a. These specific genes were absent in the battery of upregulated DEGs in TLR4 KO mice, with only C3 showing a 1.6-fold upregulation. Furthermore, Cxcl1 remained elevated three-fold in WT mice after chronic injection as compared to five-fold after a single administration. Mirroring the results of DEGs in acute inflammation, both WT and TLR4 KO mice exhibited a consistent downregulation in Ifna1, Mx1, and Maff following chronic injections. Additionally, the continued upregulation of Irf7 in TLR4 KO mice after chronic injections may indicate a compensatory mechanism in response to the absence of TLR4 signaling.

In summary, the comparison between WT and TLR4 KO responses reveals distinct patterns in gene expression induced by eLNP administration. The aforementioned genes belong to various functional categories, including interferon-stimulated genes (ISGs: Ifna1, Mx1, Ifl44, Irf7, Ifit44, Ifit1, Ifit2, Ifit3, Oasl1, Oas1a), complement cascade molecules (C6, C8a, C8b, C9, C3), and chemokines (Cxcl1 and Cxcl10). While both groups exhibit sustained downregulation of Ifna1, Maff, and Mx1, TLR4 KO mice show a unique upregulation in Cxcl10, Ifit1, Ifit2, Ifit3, Oasl1, and Oas1a, suggesting a specific response in the absence of TLR4 signaling. Upregulation of Irf7 and ISGs in TLR4 KO mice prompted us to use a different strategy of pharmacological receptor manipulation due to possible compensatory mechanisms that often accompany germline receptor deletions.

Pharmacological receptor manipulation selectively inhibiting TLR4 improves the reactogenicity of eLNPs

To validate TLR4’s role in eLNP reactogenicity and account for potential compensatory mechanisms in TLR4-deficient mice (Figure 4 A), we employed the TLR4 inhibitor TAK-242 for focused pharmacological receptor manipulation. TAK-242 is a small molecule that acts as a potent inhibitor of TLR4 signaling. This compound is designed to selectively inhibit TLR4 activation and act as a non-competitive antagonist toward LPS. ⁶⁴ TAK-242 exerts its inhibitory effects by disrupting the interaction between TLR4 and its adaptor proteins, thereby suppressing downstream signaling cascades. The level of specificity inherent in TAK-242 ensures a targeted impact on TLR4, distinguishing it from other receptors and signaling pathways. ⁶⁵

Figure 4. TLR4 antagonist rescues food intake and body weight of eLNP-treated mice.

(A) Schematic representation of the TLR4 receptor’s role in reactogenic signal transduction as a central focus for the current section. (B) A methodologic approach using pharmacologic receptor manipulation to study acute and chronic reactogenic manifestations in WT and WT mice subjected to TLR4 inhibitor, TAK-242. (C) Food intake and body weight following MC3 eLNP chronic administration in WT mice pre-treated with TAK-242 inhibitor. WT mice subjected to PBS with and without TAK-242 pre-treatment act as negative controls, and WT mice receiving MC3 eLNP act as positive control. In daily food intake, ^{a, b, c} highlights statistically significant difference of TLR4_INH + eLNP compared to PBS, ^{d, e} – statistically significant difference of eLNP compared to PBS, ^f – statistically significant difference of eLNP compared to PBS, TLR4_INH, TLR4_INH + eLNP. In body weight, ^{a, b} – highlights statistically significant difference of TLR4_INH + eLNP compared to PBS, ^{c, d} – statistically significant difference of eLNP compared to PBS, ^e – statistically significant difference of eLNP compared to PBS, TLR4_INH, TLR4_INH + eLNP. (D) Hepatic DEGs in WT mice receiving TAK-242 pre-treatment (left) and DEGs in WT mice receiving TAK-242 pre-treatment and MC3 eLNP treatment (right) all normalized to WT mice receiving PBS. (E) Serum immune markers following chronic eLNP injection to WT, TLR4 KO, and TLR4 antagonized mice (TLR4_INH). (C) Expressed as mean ± SEM normalized to baseline, n=5, ****p<0.0001 analyzed by two-way repeated measures ANOVA. (D) Fold change is an average for n=3. (E) Expressed as mean ± SEM, n=5, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 analyzed by one-way ANOVA followed by Bonferroni’s post hoc test.

In the experimental setup (Figure 4 B), WT mice were either pre-treated with TAK-242 two hours before eLNP injections or received eLNP injections without pre-treatment. We only tested chronic responses in the setting of three daily eLNP injections. The comparison group did not undergo pre-treatment and received three eLNP injections, while a negative control group received PBS. Additionally, we assessed whether TAK-242 itself influenced murine sickness behavior. Notably, there was no significant change in food intake and body weight in the PBS and TAK-242-injected mice (Figure 4 C). In contrast, eLNP-treated animals experienced a substantial decrease in food intake by almost 75% and a nearly 15% drop in body weight. Although mice pre-treated with TAK-242 exhibited a statistically significant decrease in food intake (−23%) and body weight (−7%) compared to the negative control, these reductions were considerably less severe than those observed in the eLNP-treated group. This indicates that specific TLR4 inhibition in WT mice rescues these crucial physiological parameters, underscoring the pivotal role of the TLR4 receptor in inducing reactogenic responses to eLNPs.

Pre-treatment with TAK-242 slightly perturbed the equilibrium of complement effector molecules, evident in the upregulation of C6, C8a, C8, C9, and Cxcl1 cytokine following TAK-242 administration (Figure 4 D). Strikingly, the analysis revealed only six DEGs in TAK-242-pretreated mice after eLNP administration compared to the mice receiving PBS. These genes included C3, C4a, and Hc, which encodes complement factor C5, all associated with the activation of the complement system. Additionally, Itgb2, Kng1, and Tyrobp showed upregulation that did not achieve statistical significance. The limited number and magnitude of DEGs, especially the singular upregulation of C3 with a fold change higher than 1.5, stand in stark contrast to the diverse and pronounced gene expression changes observed in WT mice receiving eLNPs without pre-treatment, as depicted in Figure 3.

We conducted a comprehensive analysis of thirty-sixty immune markers in murine serum (Figure 4 E, SI Table 5). Notably, the most consistently altered markers were pro-inflammatory cytokines (IL-1β, TNF-α, IL-6, IL-17A), anti-inflammatory cytokines (IL-4, IL-5, IL-22), chemokines (CXCL-10, CCL5, CCL7, CCL2), and the hematopoietic growth factor GCS-F, following chronic eLNP injections. In WT mice, there was a consistent upregulation of these markers, indicating an intensified immune response. However, TLR4 KO and TLR4-inhibited mice did not exhibit significant changes in these markers, except for an increase in IL-22, CCL7, and TNF-α in TLR4 KO mice. This again suggests potential compensatory mechanisms triggered by TLR4 KO mice, evident in the specific markers that increased despite the absence of TLR4 signaling.

Our study reveals critical insights into TLR4 role in LNP-induced reactogenicity and sickness behavior, highlighting significant differences between WT mice and those with inhibited TLR4 activity, either through TLR4 genetic ablation or pharmacological inhibition with TAK-242. Our findings suggest that compensatory mechanisms in TLR4 KO animals may affect LNP responses, with pharmacological inhibition providing better outcomes. In TLR4 KO mice, genetic ablation leads to an absence of TLR4-mediated signaling. However, immune system plasticity often results in compensatory pathway activation to counterbalance TLR4 function loss. This could involve upregulation of other TLRs or alternative pattern recognition receptors (PRRs) that partially mimic TLR4’s role in sensing and responding to eLNPs. Conversely, TAK-242 pharmacological inhibition specifically targets TLR4 signaling without inducing compensatory receptor upregulation. TAK-242 effectively blocks TLR4-mediated pathways, leading to more pronounced reductions in inflammatory responses and sickness behaviors, such as decreased food intake and body weight loss. This specific inhibition likely avoids alternative PRR activation, providing a clearer effect compared to genetic ablation. Genetic ablation of MyD88 adaptor leads to similar outcomes. Our observations suggest that MyD88 KO mice exhibit a marked reduction in reactogenicity to eLNPs, highlighting the critical role of MyD88-dependent signaling in LNP-induced inflammatory responses. As MyD88 serves as a convergence point for multiple PRR signaling pathways, its genetic ablation may circumvent the compensatory upregulation of alternative receptors often observed in TLR4 KO models, potentially offering a more comprehensive inhibition of LNP-induced inflammatory responses. However, the persistence of some inflammatory markers, such as Tnf, underscores the involvement of additional signaling pathways in the complex immune response to LNPs. These insights into the mechanisms of LNP-induced inflammation not only advance our understanding of innate immune responses but also have profound clinical implications. The involvement of TLR4 and MyD88 in LNP-induced reactogenicity raises important considerations for the application of LNP-based therapeutics in various patient populations, particularly those with underlying immune dysregulation or chronic diseases associated with heightened inflammatory states.

Our findings have significant implications for the application of LNP-based formulations in patients with pre-existing inflammatory conditions and diseases characterized by aberrant TLR4 signaling. Diseases such as autoimmune disorders, including rheumatoid arthritis and systemic lupus erythematosus, inflammatory bowel disease like ulcerative colitis and Crohn’s disease, certain cancers such as glioblastoma, cardiovascular diseases including atherosclerosis, neurodegenerative disorders, and metabolic disorders like obesity-induced insulin resistance, exhibit aberrant TLR4 signaling pathways. ^{29, 30, 32–43} Our research direction is crucial for developing safe nanomedicine therapeutic strategies for vulnerable populations and may inform the need for personalized approaches in LNP-based therapeutic applications. Furthermore, TLR4 activation in the liver has been extensively documented to induce liver failure and fibrosis. ⁶⁶ Given our observation of TLR4 activation by eLNPs and eLNPs tendency to have high tropism to liver, there is a pressing need to investigate the potential hazardous effects of LNP-based therapies administration in patients with liver diseases such as hepatitis, fatty liver disease, and non-alcoholic steatohepatitis (NASH). Future studies should focus on elucidating the specific impacts of LNP-induced TLR4 activation on compromised liver function and the potential exacerbation of existing pathologies.

CONCLUSION

Our investigation provides a comprehensive analysis of the reactogenicity associated with the administration of LNP formulations, highlighting the complex interplay between ionizable lipid components and immune system responses. We reveal that although specific ionizable lipids can modulate the extent of reactogenic effects, none of the tested LNP formulations are entirely free from such responses. This insight is pivotal for refining LNP carriers, underscoring the importance of adjusting lipid components to reduce adverse immune reactions, all while preserving the delivery effectiveness of mRNA LNP therapies. Utilizing a detailed experimental framework that includes genetically modified mice and pharmacological interventions, our research investigates the immune mechanisms triggered by eLNPs. A key discovery is the role of TLR4 and its adaptor protein, MyD88, in mediating immune responses to eLNPs lacking mRNA cargo. Our investigations with MyD88 and TRIF KO mice have delineated their respective roles in the propagation of reactogenic signals, identifying MyD88 as a crucial factor in managing inflammatory responses to eLNPs. Activation of TLR4 by eLNPs induces a surge in cytokine and chemokine production, triggers the complement system, and drives the recruitment of innate immune cells. This cascade of events illuminates the molecular and cellular processes underlying LNP-induced reactogenicity via the TLR4 receptor. Significantly, our findings suggest that selectively inhibiting TLR4 may be an effective strategy to reduce reactogenicity, offering a path to improve the safety and therapeutic efficacy of mRNA-based treatments.

A primary limitation of our study is the lack of assessment of LNP reactogenicity via TLR4 activation in the context of mRNA-loaded LNPs. This omission was due to the inherent challenges in distinguishing between the inflammatory responses induced by the empty LNP carrier and those triggered by the LNP carriers with mRNA payload. Both eLNPs and mRNA payload can activate overlapping innate immune pathways, leading to the expression of similar pro-inflammatory cytokines and chemokines. This similarity in effector molecules makes it challenging to differentiate between the sources of inflammation. The situation is further complicated by the existence of multiple positive feedback loops and feeder pathways that can amplify the expression of reactogenicity-related effectors, potentially masking the specific contributions of individual pathways. While eLNPs primarily engage TLR4, mRNA-loaded LNPs may additionally activate nucleic acid-sensing TLRs such as TLR3, TLR7, and TLR9, all ultimately leading to NF-κB activation and cytokine expression. Furthermore, the dynamic nature of mRNA-LNP interactions within cellular compartments, particularly the potential separation of lipids and mRNA in endosomes, further complicates the isolation of LNP-specific effects. The complexity is further amplified by the involvement of adaptor proteins TRIF and MyD88 in multiple signaling pathways. These proteins not only function in TLR-4 signaling but also serve as adaptors for nucleic acid-sensing TLRs. TRIF is involved in TLR3 signaling, which recognizes double-stranded RNA (dsRNA), while MyD88 is crucial for TLR7 activation by single-stranded RNA (ssRNA). Given these intricate interactions and overlapping pathways, we acknowledge the need for continued investigation to unravel the specific contributions of LNPs and mRNA to the overall inflammatory response, which will be crucial for optimizing the safety and efficacy of mRNA-LNP based therapeutics.

In conclusion, our study underscores the essential requirement for continued research into the reactogenic properties of LNP-based therapies as a critical pathway to enhance both their safety and efficacy profiles.

METHODS

Materials

Lipid Nanoparticle Components for eLNP and mRNA LNP synthesis, including ALC-0315, SM-102, DLin-MC3-DMA, cholesterol, DSPC, and DMG-PEG2k, were sourced from BroadPharm Inc. (San Diego, CA), BioFine International (Vancouver, BC, Canada) or Avanti Polar Lipids (Alabaster, AL, USA). For mRNA LNP synthesis, Luc mRNA was obtained from TriLink Biotechnologies Inc. (San Diego, CA, USA). For LNP concentration, Amicon ultra centrifugal filters were obtained from MilliporeSigma (MilliporeSigma, Burlington, MA, USA). For experimental consumables, Gibco (Gaithersburg, MD, USA) supplied RPMI, DMEM media, FBS, PBS, and penicillin-streptomycin. Nuclease-free water was obtained from Cytiva (Hyclone Laboratories, South Logan, UT, USA). For chemical reagents, Thermo Fisher Scientific (Waltham, MA, USA) provided a d-luciferin reagent, Halt Protease Inhibitor Cocktail, Pierce BCA protein assay, Quant-iT RiboGreen RNA kit, and RNA standards. Ambion (Carlsbad, CA, USA) supplied a DNA-free Kit and TRIzol reagent. Proteinase K reagent and Qiagen RNeasy Mini kit were purchased from Qiagen Corporation (Hilden, Germany). Applied Biosystems (Foster City, CA, USA) supplied molecular biology reagents, including TaqMan Reverse Transcription Reagents Kit, TaqMan Gene Expression Master Mix, Power SYBR Green Master Mix, and other TaqMan and SYBR assay reagents. Nanostring nCounter Mouse Inflammation Assay and consumables were obtained from NanoString Technologies (Seattle, WA, USA). Primers for Luc mRNA detection were ordered from Integrated DNA Technologies (IDT, Coralville, IA, USA). R&D Systems (Minneapolis, MN, USA) provided IL-6, IL-1β, TNF, and LCN2 ELISA kits. Assay for polyplex cytokine and chemokine assessment in mouse sera, ProcartaPlex^™ Mouse Cytokine & Chemokine Panel 1A, was obtained from Thermo Fisher Scientific (Waltham, MA, USA). TAK-242 was obtained from MilliporeSigma (MilliporeSigma, Burlington, MA, USA).

LNP preparation

The eLNP and luciferase (Luc) mRNA LNP formulations were prepared using a microfluidic method previously established in the literature. ⁶⁷ The synthesis involved combining ionizable lipids, cholesterol, DSPC, and DMG-PEG-2k in proportions of 50:38.5:10:1.5, using NanoAssemblr Spark or Benchtop (Precision Nanosystems Inc., Vancouver, Canada). In our examination of ionizable lipids as the only variable that was changed for the initial study of eLNP reactogenicity, we investigated DLin-MC3-DMA (MC3), ALC-0315, and SM-102, all integral to FDA-approved LNP formulations. ^{46, 47} For Luc mRNA LNPs, we maintained an N:P ratio of 5.67 between ionizable lipids and nucleic acids.

After the formulations’ preparation, LNPs were dialyzed using a buffer exchange approach with PBS (pH 7.4). eLNP and Luc mRNA LNP characterization, including hydrodynamic size, PDI, and ζ potential, was evaluated using Zetasizer Nano ZS (Malvern Analytical, Malvern, UK) and nanoparticle-tracking analysis using NanoSight LM-20 (Malvern Instruments, Worcestershire, UK). All mRNA LNPs employed in the study exhibited mRNA encapsulation efficiency exceeding 90%, as determined by a modified Quant-iT RiboGreen assay.

Animal Studies

All mouse studies were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The Institutional Animal Care approved all animal experiments at Oregon Health and Science University and Oregon State University (Approval Number: IP00000690).

1. Animal model

The animal studies were performed using wild-type mice (C57BL/6J, IMSR_JAX:000664), MyD88 KO mice (B6.129P2(SJL)-Myd88tm1.1Defr/J, IMSR_JAX:009088), TRIF KO mice (C57BL/6J-Ticam1Lps2/J, IMSR_JAX:005037), and TLR4 KO mice (B6(Cg)-Tlr4tm1.2Karp/J, IMSR_JAX:029015), all on a C57BL/6 background, obtained from The Jackson Laboratory. The experiments included both male and female mice aged between 8 to 10 weeks. All mice were housed in a dedicated mouse room with an euthermic environment set to murine body temperature (26°C) to avoid any cold-related stress and a 12-hour light/dark cycle. Animals had ad libitum access to water and standard food, Purina rodent diet 5001 (Purina Mills, St. Louis, MO, USA).

2. Administration routes, doses and dosages

Luc mRNA LNPs were administered at doses of 5 μg per mouse, while eLNPs were administered in volumetric equivalents to 5 μg of mRNA LNPs after confirming particle concentration through nanoparticle tracking analysis (NTA). LNPs were delivered in 100 μL PBS. Negative control groups administered with PBS were injected with equivalent volumes as the treatment groups. Biodistribution studies involved administration through intravenous (IV), subcutaneous (SQ), and intraperitoneal (IP) routes, as further discussed in Dosing strategies for induction of murine sickness behavior.

For eLNP reactogenicity studies, a single IP eLNP injection was administered for the single dosage group, with mice euthanized 6 hours post-injection. In the chronic dosage group, mice received three IP injections 24 hours apart, with euthanasia performed 6 hours post the last injection.

The TAK-242 inhibitor (3 mg/kg) was administered two hours before eLNP treatment, with both injections given IP three times, 24 hours apart. We used a dosage of 3 mg/kg of TAK-242 in mice to decrease inflammation through TLR4 inhibition based on its documented efficacy in various studies. ^68–74 Mice were euthanized 6 hours post-last injection. TAK-242 was initially dissolved in DMSO, followed by dilution in PBS, and subjected to 1 hour of sonication. A 1% DMSO solution in PBS was administered to mice as a negative pre-treatment control.

3. Biodistribution analysis

Conducting biodistribution studies, we administered 5 μg doses of Luc mRNA LNPs to mice through intraperitoneal (IP), intravenous (IV), and subcutaneous (SQ) routes, each in 100 μL PBS volumes. At 2, 4, and 6 hours post-injection, mice were exposed to d-luciferin (150 mg/kg in 100 μL PBS). Following a 15-minute interval, bioluminescent signals (BLS) were captured using the IVIS Lumina XRMS imaging system (Perkin Elmer, Hopkinton, MA, USA). Analysis of BLS data was performed with Living Image software (Perkin Elmer, Hopkinton, MA, USA). Organs and ascitic fluid were promptly collected and subjected to assessment for BLS distribution.

4. Feeding behavior assessment

In preparation for behavioral studies, animals underwent a 7-day acclimation period, during which they were individually housed. Baseline measurements for food intake and body weight were then established following this acclimation period. Murine weight and food consumption were assessed each day at a consistent time to ensure accuracy and reliability. Additionally, food was pre-dried to ensure consistency in measurements. The baseline data was recorded before the administration of eLNP formulations and then continued throughout the duration of LNP administration. Food intake and body weight data post injections were normalized to mean baseline measurements.

Fluorescence-assisted cell sorting (FACS)

Six hours after the last Luc mRNA LNP or eLNP injection in the chronically injected mice, the mice were euthanized, and the peritoneal cavity was washed to collect immune cells that were then centrifuged (1800 rpm, 2 minutes) and resuspended in equal volumes of PBS for initial counting. For FACS, pooled samples (n=5) were prepared to investigate the composition of immune cell populations. The cell preparation involved washing to achieve a single-cell suspension, adjusting the cell number to a concentration of 1.0 × 10⁷ cells/ml in ice-cold FACS buffer (PBS, 1% FBS). Staining with ice-cold reagents and conducting the procedure at 4°C prevented modulation and internalization of surface antigens. Staining was conducted in polystyrene round-bottom 12 × 75 mm BD Falcon tubes. Cell viability was confirmed to be around 95%. Each tube received 100 μl of cell suspension. Blocking antibody step using Fc block for 30 minutes was included to mitigate non-specific binding. After washing with ice-cold FACS buffer (2X 2mL) and resuspension in 100 μl PBS/FBS, primary labeled antibodies (1 μg per test) were added and incubated for 30 minutes at room temperature. Subsequently, after ice-cold FACS buffer (3X2mL) washes, cells were resuspended in ice-cold FACS buffer with the addition of DNase (50 μg/mL) and analyzed on FACSymphony^™ S6 Cell Sorter (BD Bioscience, Franklin Lakes, NJ, USA).

The gaiting included a live/dead discrimination step, utilizing an Aqua dye to distinguish viable and non-viable cells. Subsequently, the CD45 marker was used to encompass all CD45-positive cells, providing a broad group of various immune cell types. Myeloid cell identification was achieved through the use of CD11B staining. Further classification of myeloid cells involved specific markers, including Ly6G, for the segregation of neutrophils. Monocytes, macrophages, and dendritic cells within the myeloid population were identified using F4/80. Additionally, the characterization of monocytes and macrophages specifically was accomplished by staining the Ly6C marker. The staining of CD11C allowed for the identification of dendritic cells within the myeloid subset. All FACS reagents are listed in SI Table 6.

Quantitative rtPCR

For total RNA extraction, we employed the TRIzol (Ambion Co., Carlsbad, USA) and Chloroform (Sigma-Aldrich, Merck KGaA, Darmstadt, Germany) method, followed by Qiagen RNeasy Mini kit (Qiagen Corporation, Hilden, Germany) extraction with modified manufacturer’s instructions. The removal of genomic DNA was accomplished using the DNA-free Kit (Ambion, Invitrogen, Carlsbad, CA, USA). For cDNA synthesis, the TaqMan Reverse Transcription Reagents Kit (Applied Biosystems, Foster City, CA, USA) was employed. Relative quantification for all targets, excluding Luc mRNA, utilized the TaqMan Gene Expression Master Mix (Applied Biosystems, Foster City, CA, USA) and TaqMan assay reagents. The rtPCR on Luc mRNA was conducted using the SYBR Green PCR Master Mix. Data analysis was performed using a QuantStudio 3 real-time PCR system (Applied Biosystems, Foster City, CA, USA). Expression levels were normalized to 18S. The primer sequences for Luc mRNA were as follows: forward: ACTTCGAGATGAGCGTTCGG, reverse: CCAACACGGGCATGAAGAAC. All TaqMan reagents are listed in the SI Table 7.

NanoString experiments

Liver samples were homogenized in Trizol Reagent, and RNA extraction was carried out following the previously outlined procedure. RNA integrity and concentration were assessed using conventional agarose gel electrophoresis with ethidium bromide staining and the NanoDrop^™ 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Subsequently, 100 ng of RNA per sample was employed for hybridization, following the published NanoString Technologies nCounter platform protocol. ^{75, 76} The analysis was conducted with the XT_PGX_MmV2_Inflammation_CSO panel (NanoString Item No. 115000082, NanoString Technologies, Seattle, WA), enabling the assessment of multiple hepatic RNA markers associated with murine inflammation. The assay was performed on the NanoString nCounter SPRINT Profiler system (NanoString Technologies, Seattle, WA). NanoString nCounter data was analyzed using ROSALIND software (Rosalind, Sand Diego, CA, USA). Before proceeding with gene expression profiling, quality control measures, and data normalization were undertaken following the manufacturer’s guidelines, using NanoString nSolver (NanoString Technologies, Seattle, WA). Additionally, the data was further checked for quality and processed using ROSALIND software, with no samples flagged as poor quality (SI Figure 11).

Normalization of raw data in the ROSALIND software was performed using the geNorm algorithm. ROSALIND selected the optimal subset of housekeeping probes for normalization. For each sample, the geometric mean of the counts for the selected housekeeping genes was calculated. The overall geometric mean across all samples, referred to as the global geometric mean, was then computed. A normalization factor for each sample was derived by dividing the global geometric mean by the sample-specific geometric mean. The raw counts for each gene in every sample were normalized by multiplying them by the sample-specific normalization factor, and the normalized data were subsequently log2 transformed. Differential expression between groups of samples was analyzed using the Generalized Linear Model (GLM) for count data as part of the ROSALIND software pipeline. This model assumes a negative binomial distribution and utilizes estimates of noise and dispersion across all samples to calculate fold-change and p-value for each gene. The limma R library, ⁷⁷ also integrated into the ROSALIND pipeline, was used to calculate fold changes and p-values. Upregulated and downregulated genes were selected according to fold-change ≥1.5 and adjusted p-value ≤0.05. P-value adjustment is performed using the Benjamini-Hochberg approach for estimating false discovery rates (FDR).

Protein quantification in serum

Blood was collected by cardiac puncture in BD Microtainer serum separator tubes (BD Biosciences, Franklin Lakes, NJ) and was let to sit at room temperature for a minimum of 30 min. Blood samples were centrifuged for 90 sec at 15000 × g within 1 h of collection. Following serum separation, a protease inhibitor (Halt Protease Inhibitor Cocktail; Thermo Scientific, Waltham, MA, USA) was added to the serum sample. Mouse IL-6, IL-1β, LCN2, and TNF-α ELISAs were performed according to the manufacturer’s protocols (R&D Systems, Minneapolis, MN, USA). To evaluate LCN2 levels in peritoneal lavage, mice were euthanized, and their peritoneal cavities were washed with 2 mL of PBS. The collected peritoneal lavage fluid was then utilized in ELISA analyses.

Mouse inflammatory factors in serum were quantified using the ProcartaPlex assay and measured with a Luminex 200 Instrument System (MAGPIX; Thermo Fisher Scientific, Waltham, MA, USA). To investigate the mouse reactogenic response, a ProcartaPlex Mouse Cytokine and Chemokine Panel 1A 36-plex (Thermo Fisher Scientific, Waltham, MA, USA) was employed. The ProcartaPlex assay was carried out following the manufacturer’s instructions to ensure consistency and accuracy in the measurements.

Data analysis

All experimental protocols employed a sample size (n) of 5, unless otherwise specified. The determination of statistical significance utilized various analyses, including paired t-tests, multiple comparison t-tests, two-way ANOVA, and repeated-measures one-way ANOVA, conducted with GraphPad Prism software. Significance levels were indicated as follows: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

ABBREVIATIONS

CCL2

Chemokine (C-C motif) ligand 2

CCL4

Chemokine (C-C motif) ligand 4

CCL5

Chemokine (C-C motif) ligand 5

CCL7

Chemokine (C-C motif) ligand 7

COX

Cyclooxygenase; Cxcl1 - Chemokine (C-X-C motif) ligand 1

CXCL10

Chemokine (C-X-C motif) ligand 10

CXCL2

Chemokine (C-X-C motif) ligand 2

GCS-F

Granulocyte Colony-Stimulating Factor

Hc

Complement factor H (gene symbol)

Ifit1

Interferon-induced protein with tetratricopeptide repeats 1

Ifit2

Interferon-induced protein with tetratricopeptide repeats 2

Ifit3

Interferon-induced protein with tetratricopeptide repeats 3

Ifl44

Interferon-induced protein 44

Ifna1

Interferon alpha 1

IL-1β

Interleukin 1 Beta

IL-17A

Interleukin 17A

IL-22

Interleukin 22

IL-4

Interleukin 4

IL-5

Interleukin 5

IL-6

Interleukin 6

IP

Intraperitoneal

Irf7

Interferon regulatory factor 7

ISGs

Interferon-Stimulated Genes

Itgb2

Integrin beta 2

IVT

In Vitro Transcribed

Kng1

Kininogen 1

LCN2

Lipocalin 2

LNP

Lipid Nanoparticle

Luc

Luciferase

Maff

Musculoaponeurotic fibrosarcoma oncogene family protein F

MC3

MC3 (DLin-MC3-DMA), an ionizable lipid component used in lipid nanoparticles

mRNA

Messenger RNA

Mx1

Myxovirus resistance protein 1

MyD88

Myeloid differentiation primary response 88

Oas1a

2’-5’-Oligoadenylate synthetase 1A

Oasl1

2’-5’-Oligoadenylate synthetase-like 1

Orm1

Orosomucoid 1

PEG

Polyethylene Glycol

TAK-242

A small molecule inhibitor of TLR4 signaling

TLR

Toll-Like Receptor

TLR4

Toll-Like Receptor 4

TNF

Tumor Necrosis Factor

TNF- α

Tumor Necrosis Factor Alpha

TRIF

TIR-domain-containing adapter-inducing interferon-β

Tyrobp

TYRO protein tyrosine kinase-binding protein

WT

Wild Type