Accelerated Blood Clearance of Lipid Nanoparticles Entails a Biphasic Humoral Response of B-1 Followed by B-2 Lymphocytes to Distinct Antigenic Moieties

mRNA therapeutics were first described in 1992 (1), but initial clinical progress was limited because of challenges delivering mRNA to target tissues. Naked mRNA molecules require complete encapsulation to avoid being degraded by ribonucleases and many putative mRNA therapeutics require chronic dosing. Lipid nanoparticles (LNP) were first described as vectors for nucleic acid delivery 20 y ago and have made progress with small interfering RNA payloads (reviewed in Ref. 2) and, more recently, with mRNA vaccines (3). Most LNP delivery systems combine different amounts of an ionizable cationic lipid, cholesterol, phospholipids (often distearoyl phosphorylcholine [DSPC]), and polyethylene glycol (PEG), lipids that are necessary to stabilize the surface of the nanoparticles during manufacture and storage (4).

A limitation of polyethylene glycosylated (PEGylated) lipid-delivery systems is the induction of accelerated blood clearance (ABC), a phenomenon of immune activation and rapid clearance of drugs on two or more doses that may require comedication with immunosuppressants (5–9). Previous studies have shown that a common mechanism of ABC of LNP is an increase in serum IgM against the PEGylated surface lipids, leading to clearance by the mononuclear phagocyte system (MPS) (10). However, a recent study showed that phosphorylcholine (PC)–bearing liposomes stimulated B-1 lymphocytes, leading to production of anti-PC IgM (11). Based on these observations, there has been debate over what epitopes initiate and maintain ABC for PEGylated liposomes or PEGylated LNP.

In this study, we report a rapid onset of ABC for mRNA-formulated LNP (mRNA-LNP) and investigate its mechanism in vivo and in vitro. We present evidence for a previously unrecognized role of B-1 lymphocytes in ABC induction of LNPs, in which rapid production of anti-PC IgM by B-1 lymphocytes initiates ABC and increases immune sampling via redistribution of LNP to the MPS. Upon repeated injections, B-2 lymphocytes also become activated to drive an adaptive anti-PEG IgM immune response against PEGylated surface lipids. We propose a model whereby sequential innate and adaptive responses against distinct surface Ags drive ABC of LNP. These results have significant potential implications for nanoparticle delivery systems beyond mRNA therapeutics.

Specific pathogen-free, 6–8-wk-old C57BL6/J, IFNAR knockout (KO) (no. 32045), secretory IgM mutant mice (sIgM) KO (no. 003751), J_HT (no. 002438), C57BL6/J nude (000819), and BALB/c (no. 000651) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Mice were housed in microisolator cages in a BSL-2 facility. All mice were provided sterile water and food ad libitum, and all research involving animals was conducted in accordance with Moderna’s Animal Care and Use guidelines. For injection, 1.0 or 5.0 μg of mRNA-LNP in Dulbecco’s PBS were injected into animals i.v. (100 μl) with 3/10-cc insulin syringes (BD Biosciences) using standard techniques (12).

mRNA was synthesized in vitro by T7 RNA polymerase–mediated transcription from a linearized DNA template and complete uridine replacement with N1-methylpseudouridine, as described previously (13). The final mRNA uses Cap1 to increase mRNA translation efficiency, and after purification, the mRNA was diluted in citrate buffer and frozen until use.

LNP formulations were prepared using a modified procedure of a method previously described (14). Briefly, lipids were dissolved in ethanol at molar ratios of 50:10:38.5:1.5 (ionizable/helper/structural/PEG), 50:11.5:38.5 (ionizable/helper/structural) for PEG-less (PEGLess) LNP, or 50:48.5:1.5 (ioniable/structural/PEG) for PC-less (PCLess) LNP and were nanoprecipitated by mixing with mRNA in acetate buffer (pH 5.0) in a ratio of 3:1 (aqueous/ethanol). For labeled particles, 0.1 mol percentage of rhodamine/1,2-dioleoyl-sn-3-phosphatidylehanolamine was incorporated into the composition in place of the helper lipid. Formulations were dialyzed against PBS (pH 7.4) in dialysis cassettes for at least 18 h. Formulations were concentrated using Amicon Ultra Centrifugal filters (Millipore Sigma), passed through a 0.22-μm filter, and stored at 4°C until use. All formulations were tested for particle size, RNA encapsulation, and endotoxin and were found to be 80–120 nm in size, with >80% encapsulation and <10 endotoxin units/ml.

Anti-mouse Abs were obtained from eBiosciences: CD19 (clone ebio103), CD5 (clone 57-7.3), CD3 (clone 145-2C11), CD86 (clone GL10), and CD69 (clone H1.2F3). Anti-IgM (clone II/41) was obtained from BD Biosciences. Anti-V_Η11 and anti-V_κ9 were a generous gift from R. R. Hardy (Fox Chase Cancer Center, Philadelphia, PA).

Streptavidin microspheres (catalog no. 24158; Polysciences) were coupled to Biotin PC (catalog no. 860563P; Avanti) in PBS (pH 7.4) plus 2% MesoScale Diagnostics (MSD) Blocker A (catalog no. R93AA) for 30 min. After coupling beads were suspended in 800 μl PBS plus 2% MSD Blocker A for direct usage or storage at 4°C.

Carboxy-modified latex beads (catalog no. C37259; Molecular Probes, Life Technologies) were aliquoted at 100-μl/tube and washed three times with 50 mM MES 1 mM EDTA (pH 6.0). Beads were coupled with 5 mg 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide HCl (catalog no. 22980; Thermo Fisher Scientific) for 15 min at room temperature (RT) on a plate shaker at maximum rpm. 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride–coupled beads were incubated with PEG2K-DMG in PBS for 2 h. Beads were washed two times with PBS and resuspended in 2% MSD Blocker A/PBS 0.05% Thermisol for direct usage or storage at 4°C.

Beads were incubated with serum samples diluted 1/100 at RT for 1 h. After washing, beads were incubated with 1:1000 anti–anti-mouse IgM IgG APC (clone Il-41; BD Pharmingen) for 30 min in the dark. After washing, beads were analyzed by flow cytometry. Ab levels were quantified using a standard curve obtained with a monoclonal mouse anti-PEG IgM (AG no. AGP4-PABM-A; Academia Sinica, Taipei, Taiwan).

Splenocyte populations were assessed by flow cytometry. Briefly, staining was performed in FACS buffer (PBS plus 2% FBS) at RT. Following staining, cells were washed, resuspended in FACS buffer, and stored at 4°C until analyzed. Samples were collected using a BD LSRFortessa (BD Biosciences). Analysis gates were set on viable unstained cells and were designed to include all viable cell populations. Approximately 100,000 gated events were analyzed for each sample. After gating on single cells (doublets removal) and viable cells, B cells were gated on expression of CD19 and lack of expression of CD3 (CD19⁺CD3⁻), and the opposite for T cells (CD3⁺CD19⁻). Macrophages were gated on the expression of CD11b and the lack of expression of CD19 and CD3 (CD11b⁺CD3⁻CD19⁻). Isotype control Abs were included when analyses and panels were first being performed to assure specificity of staining but were not routinely included with each experiment. Data were analyzed using FlowJo software (Tree Star, Ashland, OR).

Cells were stained for CD5 and CD19 as described above. B-1a lymphocytes (CD5⁺CD19⁺) were collected by FACS using an BD FACSAria II and cultured immediately after collection. Sorted B-1a cells were >98% pure.

Human erythropoietin (Epo) levels were measured using Epo Human ELISA Kit (no. BMS2035; Thermo Fisher Scientific) following manufacturer’s recommendations. IL-6 and TNF-α were measured using V-PLEX Proinflammatory Kit (no. K152A0H-4; Meso Scale Discovery) following manufacturer’s recommendations.

ELISPOT was performed with Mouse IgM Single-Color ELISPOT (ImmunoSpot) following manufacturer’s protocol. In brief, sort-purified CD5⁺CD19⁺ B cells were distributed onto plates and incubated with PC-containing LNP, PC-containing liposomes, or LPS for 24 h. Plates were treated with anti–anti-murine IgM and developed. IgM-secreting B cells were enumerated using ImmunoSpot Analyzer (ImmunoSpot).

Sorted B-1a lymphocytes were loaded with Calcium Sensor Dye eFluor 514 (no. 65-0859-39; Thermo Fisher Scientific). After acquisition of the baseline (30 s) reading, cells were incubated with PC-containing LNP or liposomes, or anti–anti-IgM and increases in free intracellular calcium in gated B cell populations were measured in real time on a BD LSRFortessa (BD Biosciences) for 120 s. Total intracellular calcium was evaluated by incubation with ionomycin, and changes in free calcium were measured for an additional 30 s. Raw data files were analyzed with FlowJo software.

Splenocytes were incubated with LNP for 24 h. The cells were stained with Abs against surface markers and fixed in 4% paraformaldehyde for 10 min. To facilitate cell adherence to the glass surface, glass coverslip-bottom dishes (MatTek, Ashland, MA) were coated with poly-l-lysine by incubating the dishes with 20 μg/ml poly-l-lysine (Sigma-Aldrich) for 1 h at 37°C; the solution was then removed, and the glass allowed to dry overnight. The stained cell suspension was loaded onto the poly-l-lysine–coated dish and incubated for 30 min at RT to allow cell adherence before proceeding with imaging.

Structured Illumination Imaging was performed on a Zeiss Elyra S1 microscope using a 63× magnification (numerical aperature = 1.46) oil-immersion objective and 405, 488, 561, and 647 nm laser excitation. Emission bandpass filters were 420–480, 495–550, and 570–620 nm, with a 655-nm longpass filter used for the 647 laser. The excitation/emission channels were recorded consecutively; within each channel, the raw data contained three rotations, five phases, and 0.1-μm-spaced z-stack images. The super-resolution images were then reconstructed from raw images using ZEISS Structured Illumination Imaging processing software and a z-section from the middle of the stack was used for the images shown in this study.

Liver tissues from mice were collected 24 h postdose and from cynomolgus monkeys at 9–10 h postdose. Lobes were fixed in 10% neutral-buffered formalin for 24 h before being processed and embedded into paraffin blocks. Tissue blocks were cut into 5-μM sections and mounted onto glass slides.

In situ hybridization was performed using the RNAscope 2.5 LS Reagent Kit–BROWN (catalog no. 322100; Advanced Cell Diagnostics [ACD], Hayward, CA) for use with Leica Biosystems’ BOND RX System per the manufacturer’s instructions. An exclusive target probe with proprietary sequences was designed by ACD to target human Epo. Control probes to the housekeeping gene peptidylprolyl isomerase B (Ppib) mRNA (catalog no. 313918; ACD) were used as positive control and the bacterial gene dihydrodipicolinate reductase (DapB) (catalog no. 312038; ACD) as negative control and were also used as a quality control check for tissues. Briefly, slides were baked for 30 min at 60°C prior to use, then placed on Leica Bond RX autostainer (Leica Microsystems, Buffalo Grove, IL) and baked and dewaxed. Next, slides were processed using a Leica staining protocol per the ACD user manual (document number 322100-USM). Images were captured at 20× magnification with the Panoramic 250 Flash II digital slide scanner (3DHISTECH, Budapest, Hungary).

Stefin A quadruple mutant—Tracy (15) protein expression was detected by Ab to FLAG-tag. Immunohistochemistry was performed on formalin-fixed, paraffin-embedded sections using the Leica Bond RX autostainer (Leica Microsystems). FLAG Ab (catalog no. F7425; Sigma-Aldrich, St. Louis, MO) was used at 0.5 μg/ml and was detected with the Bond Polymer Refine Detection (Leica Microsystems), followed by hematoxylin and bluing reagent (Leica Microsystems) counterstain. All images were captured at 20× magnification with the Pannoramic 250 Flash II digital slide scanner (3DHISTECH).

The statistical significance of differences between groups was determined by Student t tests. Data were analyzed with Prism (version 5.0; GraphPad Software) and shown as mean ± SD. A p value <0.05 was considered statistically significant.

We sought to determine if the ABC phenomenon described in the literature (16, 17) occurs with mRNA-LNP. Weekly injections of mRNA formulated in PEGylated LNP into wild-type (WT) mice resulted in a dramatic decrease in protein expression (Fig. 1A), a redistribution of mRNA away from hepatocytes by day 22 (Fig. 1B) and a significant increase in anti-PEG IgM Abs (Fig. 1C). To further elucidate the mechanism, we evaluated the pharmacology of mRNA-LNP in T cell–deficient (nude) mice (18) and mature B cell–deficient (J_HT mice) (19). Similar to WT mice, T cell–deficient mice showed a significant reduction in protein expression (Fig. 1A), a reduction of mRNA distribution to hepatocytes by day 22 (Fig. 1B), and a significant induction of anti-PEG IgM Abs (Fig. 1C). In contrast, mature B cell–deficient animals showed no evidence of ABC: no significant decrease of protein expression (Fig. 1A), normal distribution of mRNA in hepatocytes by day 22 (Fig. 1B), and no significant anti-PEG IgM Ab response (Fig. 1C). These observations show that mRNA-LNP are subject to ABC in WT mice via a T cell–independent mechanism. In addition, dosing had a significant effect on the kinetics of ABC, as intervals of 2 wk and/or wash out periods of similar duration resulted in mRNA expression that was indistinguishable from mice injected only once (Fig. 1D, 1E). Thus, although ABC was acquired, the duration appeared to be short-lived and dependent upon repeated booster injections for maintenance.

Having demonstrated the central role of B cells in the ABC mechanism of mRNA-LNP, we next sought to identify if LNP interact directly with B cells. Freshly harvested splenocytes were isolated from naive WT mice and incubated ex vivo for 24 h with PE-rhodamine–labeled LNP for evaluation by flow cytometry. Flow cytometry data show that LNP only associated with CD19⁺ B cells and not with CD3⁺ T cells (Fig. 2A, Supplemental Fig. 1A, 1B), which was independent of mRNA payload, as empty LNP behaved similarly to mRNA loaded LNP (Supplemental Fig. 1D). Independent evaluation by fluorescence microscopy demonstrated B cell–LNP association was restricted to the plasma membrane with no evidence of endosomal uptake (Fig. 2B). The lack of internalization for CD19⁺ cells contrasted with CD11b⁺ cells where internalization through mRNA-mediated enhanced GFP protein expression were observed upon flow analysis after 24 h incubation (Supplemental Fig. 1C). Given the direct interactions of LNP with B cells, we investigated whether LNP might directly activate B cells by measuring induction of CD86 (also known as B7-2) (20–24). CD86⁻-sorted B cells incubated ex vivo with LNP showed a significant upregulation of CD86 and a strong positive correlation with LNP association and activation at the cellular level (r² = 0.8882) (Fig. 2C), suggesting a direct mechanism of action on a subset of CD19⁺ splenocytes. IFN-α transcripts were modestly and transiently upregulated in total purified B cells (Supplemental Fig. 2).

To identify which LNP surface epitopes were driving association and activation of CD19⁺ cells, we formulated mRNA in PEGLess LNP. PEGLess LNP showed a dramatic reduction in B cell association following ex vivo incubation (Fig. 2D, 2E) by flow cytometry and fluorescent microscopy and in vivo 4 h after i.v. injection (Fig. 2F). These results demonstrate that LNP can directly activate B cell subsets following surface association via distinct PEG-dependent and PEG-independent mechanisms.

Both ex vivo and in vivo experiments suggested a small subpopulation of CD19⁺ cells that continued to associate with PEGLess LNP. These B cells were determined to be CD5⁺ by cell phenotyping analysis (Fig. 3A). CD5 expression is limited largely to the B-1 cell subset of B cells ((25, 26)), a distinct lineage with unique characteristics including constitutive production of natural IgM Ab against autoantigens and pathogens (reviewed in Ref. 27). Several natural IgM specificities have been described for B1a cells, including autoreactive anti-PC IgMs encoded preferentially by V_Η11/V_κ9 or V_Η12/V_κ4/5 subsets in the mouse (25, 28–32). Given our earlier findings, we hypothesized that LNP might directly activate B1a cells via the PC epitopes on phospholipids such as DSPC. First, we confirmed the prevalence of these cell types in our models by incubating naive mouse B cells with PE-rhodamine–labeled, PC-containing LNP and measuring LNP binding by flow cytometry using specific Abs for V_Η11 and V_κ9 (33). CD19⁺CD5⁺ cells accounted for >80% of cells binding dioleylphosphatidylcholine liposomes, and >20% of those were specific for the V_κ9 subset (Fig. 3A). We then confirmed the activation of CD5⁺ B-1 cells by measuring calcium flux during incubation with PC-containing LNP (control), PCLess LNP, and PC-containing liposomes and anti-IgM control (Fig. 3B). PC-containing LNP and liposomes triggered calcium flux equivalent to that of direct BCR stimulation with anti-IgM, whereas PCLess LNP leads to a moderate calcium flux. Additional experiments confirmed the induction of IgM by PC-containing LNP by ELISPOT (Supplemental Fig. 3A) and intracellular colocalization of CD5 with LNP by fluorescent microscopy (Supplemental Fig. 3B).

To confirm these findings in vivo, mice were injected weekly with PC-containing LNP and PCLess LNP. After the fourth injection, B-1 cell population was increased in mice injected with PC LNP relative to PCLess LNP or negative controls (Fig. 3C). This corresponded with a significant increase in anti-PC IgM levels (Fig. 3D). B-1 cell responses have been reported to be type I IFN independent, in contrast to B-2 cells (34). IFNAR KO mice were therefore used to isolate B-1 contributions to ABC. After five weekly injections of human Epo mRNA formulated in LNP, we observed a 66% drop in human Epo expression in the serum of IFNAR KO mice compared with WT mice in which we observed complete (100%) loss of expression as measured by human Epo in serum (Fig. 3E). Intriguingly, WT and IFNAR KO animals showed similar expression and anti-PC IgM levels through day 15, when anti-PEG IgM responses became detectable in the WT animals and drove complete ABC (Fig. 3F, 3G). Collectively, these results demonstrate that LNP can directly activate B-1 cells to produce anti-LNP (PC) IgMs that drive the early phase of ABC prior to induction of an early adaptive response through anti-PEG IgM.

To definitively show the role of anti-PC IgM and anti-PEG IgM cooperation in ABC, we used sIgM KO, in which B cells cannot secrete IgM, but still express surface IgM and IgD and undergo class switching to express other Ig isotypes (35), including IgG. After repeated weekly injections of LNP, sIgM KO animals showed sustained Epo protein expression, a complete absence of anti–PC IgM and anti–PEG IgM but significant levels of anti–PEG IgG3, and sustained biodistribution of human Epo mRNA to hepatocytes in contrast to WT controls (Fig. 4). Based on our findings in this study, we propose a model for ABC for PEGylated lipid-delivery systems, including mRNA formulated in LNP (Fig. 5).

The use of mRNA molecules to express proteins arose 25 y ago, with the injection of vasopressin-encoding mRNA into rats with diabetes insipidus (1). Despite progress in increasing expression and limiting the immunogenicity of the mRNA (reviewed in Ref. 36), efficacy in preclinical and clinical applications has been limited because of challenges with in vivo delivery (reviewed in Ref. 37). One of the major limitations is the ABC of LNP and liposomal delivery vehicles (16, 17).

As described elsewhere, ABC is an apparent tachyphylaxis of response to PEGylated LNPs following repeated injections into the same animal, where generation of IgM Abs in response to the first injection of PEGylated LNPs accelerates clearance of subsequent doses of PEGylated LNPs (38). In this study, we demonstrate the generation of an ABC response to PEGylated LNP containing mRNA in WT mice (Fig. 1A).We confirmed that this response is dependent on IgM, as sIgM KO mice, which do not secrete IgM, but express surface IgM and IgD and undergo class switching to express other isotypes (35), did not show an ABC response to such LNP (Fig. 4).

Recent studies have shown that PEGylated liposomes are recognized after the first injection by marginal zone B cells. Following opsonization by anti–PEG IgM and subsequent complement activation, the liposomes are transported to the follicular zone, where Ag-presenting dendritic cells are located (39). In this report, we showed that LNP can interact directly with a large amount of B cells (Fig. 2A), specifically the B-1 cell subset, and induce their activation (Fig. 3A) and apparent expansion in vivo (Fig. 3C).

It is important to note that LNP association with B cells has been measured in mice that were never exposed to PEG molecules. In humans, a growing body of evidence suggests that the induction of anti-PEG Abs is possible (40–42). In contrast to most animal studies, the anti-PEG Ab response in humans is more skewed toward the development of IgG isotype Abs (40). Interestingly, Armstrong et al. (43) have found that a significant fraction of the normal population actually possesses pre-existing anti-PEG Abs (i.e., the presence of PEG-specific Abs in the absence of treatment with PEGylated therapeutics), which may become even more prevalent in subsequent years. Both pre-existing and induced anti-PEG Abs should present significant challenges to the clinical efficacy of PEGylated therapeutics and will require strategies to overcome the effects of anti-PEG Abs.

The B-1 cell population in mice contains a specific subset displaying an anti-PC specificity through expression of the V_Η11V_κ9 or V_H12V_κ4/5 BCR (25, 28, 32). Using an Ab against the PC-specific BCR, we were able to demonstrate the association with and internalization of LNP by this subset of B-1 cells (Fig. 4A, Supplemental Fig. 3B). The LNP/B-1 association resulted in activation of the B-1 cells, an increase in anti-PC IgM secretion, and an increase in the fraction of B-1 cells in the spleen following injection of the LNP in WT mice (Figs. 3C, 4B). Although the induction of anti-PC IgM was relatively modest (∼2-fold above baseline), the absolute quantity of Ab in the serum 7 d postdose was >100-fold higher than the anti-PEG IgMs response (Fig. 4C versus 4B). This correlated with a 50% drop in the pharmacology as measured by mRNA-encoded human Epo expression upon second dose (Fig. 4A). These results are consistent with previously published findings regarding PC-containing liposomes (44) but, to our knowledge, are the first example of this pathway impacting LNPs. Upon third and subsequent doses, we demonstrated the consolidation of ABC resulting in complete loss of pharmacology. Although anti-PEG IgM responses continued to rise modestly, this final phase of ABC to the LNPs appear to correlated with the rise of anti-PEG IgM responses. Notably, in the sIgM KO mice, the lack of IgM responses abrogated the ABC response despite the apparent induction of anti-PEG IgG by week 5 (Fig. 4).

Based on these and other findings, we propose a model for ABC for PEGylated lipid-delivery systems, including mRNA formulated in LNP (Fig. 5). On first dose in naive animals, PC lipids directly activate B-1 cells to induce natural IgM. Upon repeated dosing, anti-PC IgMs binding to LNP drive delivery to APC (e.g., follicular dendritic cells and marginal zone B cells) that initiate an adaptive anti-PEG response resulting in first anti-PEG IgM and anti-PEG IgG responses. Thus, similarly to Cruz-Leal et al.’s data (40), we highlighted a coupled interaction between B-1 cells and LNPs leading to anti-PEG IgM response, both humoral responses being critical in driving the clearance of the LNP from the blood and into the MPS of the lymphoid (and other) tissues.

It is tempting to extend these findings to other delivery technologies that might similarly be subject to natural IgM opsonization. For instance, gene therapies often rely on viral vectors that include repeating PC-like epitopes, and other oligonucleotide therapeutics that use LNP similar to those described in this study (17, 45). By defining the molecular and immunological basis of ABC in this study, we highlight the need for future delivery systems, including for mRNA, which minimize surface immune epitopes that drive innate, as well as adaptive, humoral responses.

We thank R. Levy and L. Marini for technical assistance, B. Sorensen for manuscript editing, and the entire in vivo pharmacology team in Moderna, Inc. The authors thank R.R. Hardy for PC-specific V_k9 and V_H11 Abs. Finally, the authors thank Prof. Ulrich Von Andrian and Dr. Michel S. Diamond for critical reading of the manuscript and helpful discussions.

All authors are or have been employees of Moderna, Inc. and receive salary and stock options from Moderna, Inc.