The lipid nanoparticle component of the mRNA-LNP platform used in preclinical vaccine research is highly inflammatory

Version 2. bioRxiv.Preprint.2021 Mar 4 [revised 2021 Jul 23]. doi: 10.1101/2021.03.04.430128
PMCID:PMC7941620 Other versions PMID:33688649

Sonia Ndeupen, 1 Zhen Qin, 1 Sonya Jacobsen, 1 Henri Estanbouli, 1 Aurélie Bouteau, 1 Botond Z. Igyártó1,*

Original article here.

Related data
Supplementary material

Abstract

Vaccines based on mRNA-containing lipid nanoparticles (LNPs) are a promising new platform used in two major vaccines for coronavirus disease 2019 (COVID-19).
Clinical trials and ongoing vaccinations have presented very high levels of protection and varying degrees of side effects. However, the nature of the reported side effects has not yet been well defined.
Here, we present evidence that LNPs, which are used in many preclinical studies, are highly inflammatory in mice. Intradermal injection of these LNPs triggered a rapid and robust inflammatory response characterized by massive neutrophil infiltration, activation of diverse inflammatory pathways, and production of various inflammatory cytokines and chemokines.
When the same amount of LNPs was administered intranasally, a similar inflammatory response occurred in the lungs, resulting in high mortality.

In summary, here we show that LNPs used in many preclinical studies are highly inflammatory.
Therefore, the potent adjuvant activity of LNPs and their reported superiority compared to other adjuvants in supporting the induction of adaptive immune responses likely stem from their inflammatory nature.
Furthermore, since preclinical LNPs are similar to those used in human vaccines, this platform may also explain the side effects observed in humans.

Keywords: mRNA-LNP vaccine, SARS-CoV-2 vaccine, LNP, inflammation, side effects

Introduction

The nucleoside-modified mRNA-LNP vaccine platform used by Pfizer/BioNTech and Moderna in SARS-CoV-2 vaccines has been widely tested in preclinical studies, and its effectiveness in supporting Tfh cells and protective humoral immune responses is equal to or better than other vaccines (Alameh et al., 2020). The mRNA component of these vaccines is nucleoside-modified to reduce potential innate immune recognition (Karikó et al., 2005, 2008). LNPs were selected as carrier vehicles to protect mRNA from degradation and to assist in intracellular delivery and endosomal escape. LNPs are composed of a mixture of phospholipids, cholesterol, PEGylated lipids, and cationic or ionizable lipids. Phospholipids and cholesterol have structural and stabilizing roles, and PEGylated lipids support long-term circulation. Cationic/ionizable lipids are included to allow for the complexation of negatively charged mRNA molecules and to enable the exit of mRNA from the endosome to the cytosol for translation (Samaridou et al., 2020). Data support that some LNPs containing ionizable/cationic lipids are highly inflammatory and likely cytotoxic (Samaridou et al., 2020). Preclinical studies have shown that mRNA complexed with Acuitas Therapeutics' proprietary LNPs has adjuvant activity (Pardi et al., 2018a). However, the potential inflammatory nature of these LNPs was not evaluated (Alameh et al., 2020; Pardi et al., 2018b, 2018a).

In human clinical trials of Pfizer/BioNTech and Moderna vaccines, side effects often associated with inflammation, such as pain, swelling, fever, and drowsiness, have been reported (Jackson et al., 2020; Sahin et al., 2020; Walsh et al., 2020). Under the assumption that this vaccine platform is non-inflammatory, the reported side effects were interpreted as the vaccine being potent and generating an immune response. However, no studies have been conducted to identify the potential causes of local and systemic side effects.
In this study, we took a systematic approach to focus on the injection site and analyze the inflammatory properties of LNPs used in preclinical vaccine research (Awasthi et al., 2019; Laczkó et al., 2020; Lederer et al., 2020; Pardi et al., 2017a, 2017b, 2018c, 2018a). Using complementary techniques, we show that intradermal or intranasal delivery of LNPs used in preclinical studies in mice induces inflammation characterized by leukocyte infiltration, activation of different inflammatory pathways, and secretion of a diverse pool of inflammatory cytokines and chemokines.
Therefore, the inflammatory environment induced by LNPs may be part of the side effects of mRNA-LNP-based SARS-CoV-2 vaccines reported in humans, and may contribute to the high potency in inducing protective immunity.

Results

Intradermal inoculation of LNPs induces robust inflammation
mRNA combined with LNPs has been used in many preclinical studies and is a major component of recent Pfizer/BioNTech and Moderna SARS-CoV-2 vaccines (Alameh et al., 2020; Jackson et al., 2020; Sahin et al., 2020; Walsh et al., 2020). The mechanism of action of this mRNA-LNP platform is not well defined. Although the mRNA component has been modified to reduce activation of the interferon pathway (Karikó et al., 2005, 2008), mRNA complexed with LNPs has been shown to have adjuvant activity (Pardi et al., 2018a). The mRNA-LNP platform promotes a robust humoral immune response, and vaccinated humans often presented with typical inflammatory side effects such as pain, swelling, and fever (Jackson et al., 2020). Based on these observations, we hypothesized that mRNA-LNP adjuvant activity and side effects reported in humans could stem from the inflammatory properties of LNPs. mRNA complexed with LNPs was used in preclinical studies at doses ranging from 3 to 30 μg/mouse (Laczkó et al., 2020; Pardi et al., 2018a). Therefore, we intradermally injected 10 μg (4 spots; 2.5 μg/spot) of these empty LNPs formulated in phosphate-buffered saline (PBS) or control PBS into adult wild-type (WT) C57BL/6 (B6) mice. Mice were sacrificed at different time points after injection, and ~1 cm2 skin samples were collected from the injection site. Skin samples injected with LNPs macroscopically showed strong signs of inflammation, such as redness and swelling (Fig. 1A). Single-cell suspensions were prepared from these samples and analyzed by flow cytometry for infiltration (Fig. 1B and Supplementary Fig. 1). Flow cytometry revealed massive and rapid leukocyte infiltration dominated by neutrophils, which slowly disappeared by day 14 (Fig. 1B). When the ionizable lipid component was removed from the LNPs, visible skin inflammation (Fig. 1C) and leukocyte infiltration (Fig. 1D) disappeared. Thus, LNPs used in preclinical studies, promote a rapid inflammatory response at the injection site, which was found to be dependent on the ionizable lipid component.

Figure 1

To better understand the global changes caused by intradermal injection of LNPs, we repeated the above experiment using LNPs complexed with control non-coding polycytosine mRNA. Skin samples collected 1 day after injection were divided into two and analyzed using Luminex® and bulk RNA-seq (Fig. 2A). Luminex® data confirmed the flow cytometry findings, demonstrating the presence of various inflammatory cytokines and chemokines compared to control samples (Fig. 2B, 2C and Suppl. Fig. 2). Chemokines that attract and promote the function of neutrophils and monocytes, such as CCL2, CCL3, CCL4, CCL7, CCL12, CXCL1, and CXCL2, dominated the panel (Fig. 2B). Furthermore, large amounts of IL-1β, GM-CSF, and IL-6, which are signature cytokines of the inflammatory response, were detected (Fig. 2C). RNA-seq analysis revealed that LNP injection upregulated thousands of genes (Fig. 2D). 9,508 genes were differentially expressed at p<0.05 and 8,883 genes at FDR<0.05. More importantly, confirming the flow cytometry and Luminex® data, genes related to monocyte/granulocyte development, recruitment, and function (Cxcl1, Cxcl2, Cxcl5, Cxcl10, Ccl2, Ccl3, Ccl4, Ccl7, Ccl12, Csf2, Csf3) and inflammation (IL1b and IL6) showed the highest fold changes compared to control samples (Fig. 2E). We also observed a significant increase in gene transcripts related to inflammasome activation, such as Il1b and Nlrp3, and a downregulation of Nlrp10, which is known to suppress the inflammasome (Fig. 2E). Gene Set Enrichment Analysis (GSEA) showed activation of many different inflammatory pathways, including but not limited to viral infection, RIG-I, NOD-like, and Toll-like receptor signaling (Fig. 2F). Pro-apoptotic and necroptotic gene sets were also significantly upregulated, as was interferon signaling (Fig. 2F).

Figure 2

In short, we found that LNPs, alone or complexed with control non-coding polycytosine mRNA, cause strong inflammation in mice.

Intranasal inoculation of 10 μg of LNPs into mice causes high mortality.

To determine whether LNP-induced inflammation is independent of the inoculation route, we further verified the potency in the respiratory tract. Since mice are susceptible to intranasal inoculation of inflammatory compounds, it is a preferred route for examining side effects. We intranasally inoculated adult WT B6 mice with LNPs ranging from 2.5 μg to 10 μg/mouse and monitored their health and body weight for up to 8 days. The results showed that ~80% of mice administered 10 μg of LNPs died within 24 hours (Fig. 3A). A 5 μg dose killed ~20% of mice by that time, but all mice treated with 2.5 μg survived and showed no significant clinical signs of weight loss (Fig. 3B) and distress (Fig. 3C). For the 5 and 10 μg doses, surviving mice showed notable clinical scores of distress such as shivering/trembling, and they lost significant weight during the first 2 days of treatment (Fig. 3B and C). After the first ~3 days, these mice no longer continued to show notable clinical scores, and their body weight slowly began to normalize (Fig. 3C).

Figure 3

Next, we verified whether intranasal inoculation leads to inflammation. For this, lung samples from PBS and 10μg LNP-treated mice were prepared for macroscopic analysis at 9 and 24 hours post-inoculation, and for flow cytometry at 9 hours post-inoculation. Within hours, the lungs turned red (Figure 3D). As observed in the skin, flow cytometry analysis showed significant leukocyte infiltration dominated by neutrophils and eosinophils, as well as a decrease in macrophages and specific DC subsets (Figure 3E and Supplementary Figure 3). To determine whether LNPs reach the lungs via intranasal administration, mice were intranasally inoculated with 10μg of DiI-labeled LNPs. Six hours later, histological examination confirmed a homogeneous distribution of LNPs in the lung tissue (Figure 3F).

Thus, similar to skin inoculation, intranasal administration of LNPs causes massive inflammation. Furthermore, the inflammatory properties of LNPs are not site-specific and showed rapid diffusion, dispersion, and distribution rates within the tissue.

Discussion

Here, we reveal that the LNPs used in many preclinical studies (Freyn et al., 2020; Laczkó et al., 2020; Lederer et al., 2020; Pardi et al., 2017b, 2018c, 2018a) are highly inflammatory. This may explain their potent adjuvant activity and their superiority compared to other adjuvants in supporting the induction of adaptive immune responses.

Previous preclinical mouse data suggested that mRNA complexed with LNPs has adjuvant activity (Pardi et al., 2018a). The mRNA is nucleoside-modified and specifically purified to reduce the activation of certain innate inflammatory pathways (Karikó et al., 2005, 2008, 2011). Our analysis focused on the injection site revealed the inflammatory nature of these LNPs, which may be the basis for their adjuvant properties. The cationic/ionizable lipid component of LNPs is often inflammatory and cytotoxic (Samaridou et al., 2020). Indeed, we found that the unique ionizable lipid component of these LNPs is also inflammatory (Figure 1C and D). mRNA-LNP vaccines are administered intramuscularly in humans. We reasoned that with intramuscular administration, the immediate inflammatory response in deep tissues might remain hidden from the observer. This is especially true in fur-covered mice. Therefore, we chose to inject these constructs intradermally into shaved skin or inoculate them intranasally. The reason is that skin inflammation is visible to the naked eye, and airway inflammation causes observable distress in mice. Indeed, in both cases, clear signs of inflammation were seen, confirming that LNPs cause inflammation regardless of the route of inoculation. Therefore, it is highly likely that intramuscular injection of LNPs causes a similar inflammatory response in the muscle.

Humans present with various side effects after intramuscular inoculation with Pfizer/BioNTech or Moderna vaccines, most frequently pain, swelling, fever, and chills (Jackson et al., 2020; Sahin et al., 2020; Walsh et al., 2020). These are typical symptoms associated with inflammation caused by cytokines such as IL-1β and IL-6 (Dinarello, 2018; Tanaka et al., 2014). Along with causing local inflammatory responses, these cytokines act as major endogenous pyrogens (Conti, 2004) and instruct the hypothalamus to raise body temperature (fever) to overcome potential infections. Consistent with this, intradermal inoculation of mice with LNPs resulted in the massive secretion of major and minor pyrogens, IL-1β/IL-6 and macrophage inflammatory protein-α (CCL3) and macrophage inflammatory protein-β (CCL4), respectively (Figure 2B and C). Furthermore, the activation of other inflammatory pathways and cell death were observed, which may further emphasize the side effects experienced. However, further research will be needed to determine the exact nature of the inflammatory response caused by mRNA-LNP vaccines in humans and the extent to which it may overlap with the inflammatory signature recorded here for mice.

How these LNPs and their ionizable lipid components activate different inflammatory pathways remains to be elucidated. Theoretically, LNPs could activate multiple pathways, or they could be involved in only one pathway that initiates the inflammatory cascade. Some cationic/ionizable lipids bind to and activate TLRs (Lonez et al., 2012, 2014; Samaridou et al., 2020; Tanaka et al., 2008; Verbeke et al., 2019). Our GSEA analysis revealed that these unique LNPs are also highly likely to activate TLR pathways in particular (Figure 2F). We also observed increased expression of inflammasome components such as Nlrp3 and enrichment of genes involved in necroptosis. Inflammatory cell death such as necroptosis and pyroptosis is thought to cause the release of DAMPs and further exacerbate inflammation.

Intranasal inoculation of LNPs resulted in significant mortality, which is thought to be due to the massive inflammatory response induced in the lungs (Figure 3D and E). However, some clinical symptoms such as tail stiffness, tremors, and lack of coordination (data not shown) also suggest the involvement of the brain as a confounding factor. As lipid particles, LNPs can diffuse rapidly (Figure 3F) and may access the central nervous system via the olfactory bulb or blood. Whether these LNPs can cross the blood-brain barrier has not yet been determined. However, there is also a possibility of entering the CNS from the hypothalamus, which lacks a blood-brain barrier. Even if peripherally injected LNPs could reach the CNS through the blood, the amount reaching the CNS would be small and thus would not induce major inflammation, but the possibility remains that they could induce hypothalamus-driven side effects such as fever, nausea, and drowsiness.

After booster injections, people often show more severe systemic side effects. This raises the possibility that the adaptive immune response somehow amplifies vaccine-related side effects. One of the causes identified so far is immunogenic PEG. Antibodies formed against PEG have been reported to support so-called anaphylactoid, complement activation-related pseudo-allergy (CARPA) reactions (Kozma et al., 2020; Szebeni, 2005, 2014). Notably, since PEG is a compound frequently used in cosmetics and toothpaste, many people may have anti-PEG antibodies. We discuss other possible mechanisms in a recent opinion article (Igyártó et al., 2021). Briefly, mRNA primarily infects cells near the injection site, but it could potentially reach any cell in the body (Maugeri et al., 2019; Pardi et al., 2015). As a result, the translated protein may be presented on MHC-I in the form of peptides or displayed as a whole protein on the cell membrane. In either case, cells bearing vaccine peptides/proteins on their surface could be targeted and killed by cells of the adaptive and innate immune systems, CD8+ T cells and NK cells, respectively (via ADCC).

From the above, it was found that side effects of the first vaccination, excluding CARPA, are likely associated with the robust inflammation induced by LNPs. On the other hand, after the second vaccination, side effects may be further exacerbated by immune responses targeting cells expressing vaccine proteins or their peptide derivatives. Whether innate memory responses to LNPs (Nea et al., 2011) also contribute to the amplification of side effects has not yet been determined (Figure 4). Overall, the robust inflammatory environment induced by LNPs, coupled with the presentation of vaccine-derived peptides/proteins outside of antigen-presenting cells, may cause tissue damage and exacerbate side effects. Since self-antigen presentation in an inflammatory environment is associated with the development of autoimmune diseases (Charles A. Janeway et al., 2001), although not detected here, further investigation is required.

Figure 4

mRNA-LNPs support very robust adaptive immune responses in animal models and humans (Alameh et al., 2020; Le Bert et al., 2020; Jackson et al.). So far, the induction mechanisms of these immune responses have not been fully elucidated. Our findings reveal that the LNPs used in several preclinical studies are inflammatory, which may explain their superiority over FDA-approved adjuvants in supporting the development of Tfh cells and humoral immune responses. Notably, their high efficacy likely depends on the activation of different inflammatory pathways exacerbated by direct cytotoxicity. The inflammatory properties of these LNPs should certainly be further utilized as an adjuvant platform in combination with proteins, subunit vaccines, or existing attenuated vaccines (Bernasconi et al., 2021; Debin et al., 2002; Martins et al., 2007; Shirai et al., 2020; Swaminathan et al., 2016). Unlike other adjuvants, LNPs can therefore serve a dual purpose as a delivery vehicle for different cargoes and as an adjuvant. However, as LNP-related vaccines advance, it will be necessary to balance positive adjuvant and negative inflammatory properties. Since the doses of vaccines used in rodents are much higher than those used in humans (Nair and Jacob, 2016), detailed dose-response studies are required. Since some DCs can support humoral immune responses in the absence of adjuvants, non-inflammatory LNPs may also be developed (Bouteau et al., 2019; Kato et al., 2020; Li et al., 2015; Yao et al., 2015).

Materials and Methods

Mice
WT C57BL/6J mice of different ages and sexes were purchased from Jax® or bred in-house. All experiments were performed on 6-12 week old mice. Mice were housed in microisolator cages and fed autoclaved sterilized food. All mouse protocols were approved by the Institutional Care and Use Committee.

Reagents
In our study, we used LNP formulations owned by Acuitas Therapeutics, as described in US Patent US10,221,127. These LNPs were previously carefully characterized and widely tested in preclinical vaccine research in combination with nucleoside-modified mRNA (Laczkó et al., 2020; Lederer et al., 2020; Pardi et al., 2017a, 2018c, 2018a). The following LNP formulations were used: empty LNPs with or without proprietary ionizable lipids, DiI-labeled LNPs, and LNPs complexed with non-coding polycytosine mRNA.

Intradermal inoculation and flow cytometry analysis
The day before injection, hair on the dorsal skin of WT B6 adult mice was removed with electric clippers, and the injection site was wet-shaved with a Persona razor. The next day, mice were intradermally injected with 2.5μg/spot of LNPs (4 spots, total 10μg) in PBS or PBS. At different time points after injection, mice were sacrificed, and ~1cm2 of skin around the injection site was harvested. The skin samples were then minced into small pieces using curved scissors and exposed to collagenase/hyaluronidase digestion as previously described (Kashem and Kaplan, 2018). Single-cell suspensions were stained for the following markers: fixable viability dye (Thermo Fisher), MHC-II, CD11b, CD11c, CD45, CD64, F4/80, and Ly6G (all from BioLegend). Stained samples were processed on an LSRFortessa™ (BD Biosciences), and the resulting data were analyzed with FlowJo 10.

Luminex®
The day before injection, hair on the dorsal skin of WT B6 adult mice was removed with electric clippers, and the injection site was wet-shaved with a Persona razor. The next day, mice were intradermally injected with 2.5μg/spot LNPs complexed with non-coding polycytosine mRNA or PBS (Figure 2A). 24 hours later, skin samples were harvested and processed for Luminex® and RNA-seq. Samples for Luminex were weighed and homogenized using a Dounce tissue grinder in 1.5 ml of 10 mM Tris pH 7.4, 150 mM NaCl, 1% Triton-X-100 per gram of tissue in the presence of Roche protease inhibitor cocktail. After incubating on ice for 30 minutes, samples were spun at 10K RPM at 4°C for 10 minutes. The supernatant was collected and filtered through a 0.22μm Eppendorf tube filter (MilliporeSigma). The supernatant was aliquoted and stored at -80°C until further use. Samples were tested using the Bio-Plex Pro™ Mouse Chemokine Panel 33-Plex according to the manufacturer's instructions.

RNA preparation, sequencing, data analysis, and visualization
Total RNA was isolated from tissue lysates using the RNeasy Mini Kit (Qiagen) (including on-column DNase digestion). The quantity and quality of total RNA were analyzed using the RNA 6000 Pico Kit (Agilent). Sequencing, data analysis, and visualization were performed as previously described by us and others (Kanehisa, 2000; Liberzon et al., 2011; Love et al., 2014; Su et al., 2020; Subramanian et al., 2005).

Intranasal inoculation and flow cytometry analysisMice were anesthetized by intraperitoneal injection of a xylazine/ketamine mixture. LNPs were administered intranasally by gently dropping 2.5, 5, and 10 μg doses in 30 μL of sterile PBS into the left nostril, allowing the mice to inhale them. Clinical performance of the mice was scored daily for 8 days as previously described (Shum et al., 2014). In addition, body weight was measured daily. A subset of mice from the 10 μg LNP dose and corresponding PBS controls were sacrificed at the indicated time points post-inoculation, and lung samples were collected for histology and flow cytometry. For histology, samples were fixed overnight in 4% PFA and then embedded in OCT. 8-micrometer thick sections were prepared using a cryostat and counterstained with DAPI. Stitched confocal images were taken using a Nikon A1 microscope. Lung samples for flow cytometry were digested using the collagenase/hyaluronidase technique also used for skin samples (Kashem and Kaplan, 2018). The resulting single-cell suspensions were stained with the following markers: fixable viability dye (Thermo Fisher), MHC-II, CD11b, CD11c, CD24, CD45, CD64, Ly6G, and Ly6C (all from BioLegend) (Yu et al., 2016).

Statistical analysis
All data were analyzed using GraphPad Prism version 9.0.0, and the statistical methods used to determine significance are described below each figure.

Supplementary materials

Supplementary 1
Supplementary Figure 1. Gating strategy for the skin infiltration data shown in Figure 1B.

Click to view (220K, pdf)

Supplementary 2
Supplementary Figure 2. Extended cytokine and chemokine panel.

Click to view (35K, pdf)

Supplementary 3
Supplementary Figure 3. (A) Gating strategy for the lung leukocyte data shown in Figure 3E. (B) H&E staining of lungs collected from PBS and 10 μg LNP-treated mice 9 hours post-inoculation.

Click here (8.0M, pdf)

Acknowledgments

This research was supported by departmental startup funds and R01AI146420 to B.Z.I. We thank the following core facilities for their cooperation and support: Genomics, Proteomics, Flow Cytometry, and Imaging Cores. Figures were created with BioRender. RNA-seq data are available under the following GEO accession number: GSE167521.

Footnotes

Conflict of interest

The authors declare no conflict of interest of any kind.

References 1-51
Original reference