ESCRT recruitment to SARS-CoV-2 spike induces virus-like particles that improve mRNA vaccines

Summary Prime-boost regimens for COVID-19 vaccines elicit poor antibody responses against Omicron-based variants and employ frequent boosters to maintain antibody levels. We present a natural infection-mimicking technology that combines features of mRNA- and protein nanoparticle-based vaccines through encoding self-assembling enveloped virus-like particles (eVLPs). eVLP assembly is achieved by inserting an ESCRT- and ALIX-binding region (EABR) into the SARS-CoV-2 spike cytoplasmic tail, which recruits ESCRT proteins to induce eVLP budding from cells. Purified spike-EABR eVLPs presented densely arrayed spikes and elicited potent antibody responses in mice. Two immunizations with mRNA-LNP encoding spike-EABR elicited potent CD8+ T cell responses and superior neutralizing antibody responses against original and variant SARS-CoV-2 compared with conventional spike-encoding mRNA-LNP and purified spike-EABR eVLPs, improving neutralizing titers >10-fold against Omicron-based variants for 3 months post-boost. Thus, EABR technology enhances potency and breadth of vaccine-induced responses through antigen presentation on cell surfaces and eVLPs, enabling longer-lasting protection against SARS-CoV-2 and other viruses.

INTRODUCTION mRNA vaccines emerged during the COVID-19 pandemic as an ideal platform for the rapid development of effective vaccines. 1 Currently approved SARS-CoV-2 mRNA vaccines encode the viral spike (S) trimer, 2 the primary target of neutralizing antibodies during natural infections. 3 Clinical studies have demonstrated that mRNA vaccines are highly effective, preventing >90% of symptomatic and severe SARS-CoV-2 infections 4,5 through both B and T cell responses. 6 mRNA vaccines in part mimic an infected cell since expression of S within cells that take up S-encoding mRNAs formulated in lipid nanoparticles (LNP) 7 results in cell surface expression of S protein to stimulate B cell activation. Translation of S protein inside the cell also provides viral peptides for presentation on MHC class I molecules to cytotoxic T cells, which does not commonly occur in protein nanoparticle-based vaccines 8 that resemble the virus by presenting dense arrays of S protein; e.g., the Novavax NVX-CoV2373 vaccine. 9,10 However, comparisons to COVID-19 mRNA vaccines showed that NVX-CoV2373 elicits comparable neutralizing antibody titers, 11,12 a correlate of vaccine-induced protection, 13 suggesting that potent B cell activation can be achieved through presentation of viral surface antigens on cell surfaces or virus-resembling nanoparticles. Achieving higher antibody neutralization titers is desirable as antibody levels contract substantially over a period of several months, 11 and SARS-CoV-2 variants of concern (VOCs) that are less sensitive to antibodies elicited by vaccines or natural infection have been emerging. [14][15][16] An optimal vaccine might therefore combine attributes of both mRNA-and protein nanoparticlebased vaccines by delivering a genetically encoded S protein that gets presented on cell surfaces and induces self-assembly and release of S-presenting nanoparticles.
Here, we describe a new technology that engineers membrane proteins to induce self-assembly of enveloped virus-like particles (eVLPs) that bud from the cell surface. This is accomplished for the SARS-CoV-2 S protein by inserting a short amino acid sequence (termed an endosomal sorting complex required for transport [ESCRT]-and ALG-2-interacting protein X [ALIX]binding region or EABR) 17 at the C terminus of its cytoplasmic tail to recruit host proteins from the ESCRT pathway. Many enveloped viruses recruit ESCRT-associated proteins such as TSG101 and/or ALIX through capsid or other interior viral structural proteins during the budding process. 18,19 Thus, fusing the EABR to the cytoplasmic tail of a viral glycoprotein or other membrane protein directly recruits TSG101 and ALIX, bypassing the need for co-expression of other viral proteins for eVLP self-assembly. Cryoelectron tomography (cryo-ET) showed dense coating of spikes on purified S-EABR eVLPs, and direct injections of the eVLPs elicited potent neutralizing antibody responses in mice. Finally, we demonstrate that an mRNA vaccine encoding the S-EABR construct elicited at least 5-fold higher neutralizing antibody responses against SARS-CoV-2 and VOCs in mice than a conventional S-encoding mRNA vaccine or purified S-EABR eVLPs. These results demonstrate that mRNA-mediated delivery of S-EABR eVLPs elicits superior antibody responses, suggesting that dual presentation of viral surface antigens on cell surfaces and on extracellular eVLPs has the potential to enhance the effectiveness of COVID-19 mRNA vaccines.

RESULTS
ESCRT recruitment to the spike cytoplasmic tail induces eVLP assembly To evaluate the hypothesis that direct recruitment of ESCRT proteins to the cytoplasmic tail of a SARS-CoV-2 S protein could result in self-assembly and budding of eVLPs, we fused EABRs derived from different sources to the truncated cytoplasmic tail of the S protein, separated from its C terminus by a short Gly-Ser linker ( Figures 1A and 1B). The S protein contained the D614G substitution, 20 a furin cleavage site, two proline substitutions (2P) in the S2 subunit to stabilize the prefusion conformation, 21 and the C-terminal 21 residues were truncated to optimize cell surface expression by removing an endoplasmic reticulum (ER)-retention signal (DCT) 22 ( Figure 1B). We evaluated the EABR fragment from the human CEP55 protein that binds TSG101 and ALIX during cytokinesis 17 ( Figure 1B). For comparisons, viral late domains that recruit early ESCRT proteins during the viral budding process were obtained from the Equine infectious anemia virus (EIAV) p9 protein, 23 residues 1-44 of the Ebola virus (EBOV) VP40 protein, 24 and the HIV-1 p6 protein 25 (Figure S1A). We hypothesized that eVLP production could be enhanced by preventing endocytosis of EABR-fusion proteins to extend the duration that proteins remain at the plasma membrane to interact with ESCRT proteins. We therefore added an endocytosis prevention motif (EPM), a 47-residue insertion derived from the murine Fc gamma receptor FcgRII-B1 cytoplasmic tail ( Figures 1A and 1B) that tethers FcgRII-B1 to the cytoskeleton to prevent coated pit localization and endocytosis. 26 The abilities of the S-EABR, S-p9, S-VP40  , and S-p6 constructs to generate eVLPs were evaluated by transfecting Ex-pi293F cells and measuring eVLP production in supernatants from which eVLPs were purified by ultracentrifugation on a 20% sucrose cushion. Western blot analysis showed that the highest S protein levels were detected for the S-EABR construct, suggesting that the CEP55 EABR induced efficient self-assembly of S-containing eVLPs ( Figures 1C and S1B). At a sample dilution of 1:400, the S-EABR construct produced a similarly intense band compared with the S-p9 construct at a 1:40 dilution, suggesting that S protein levels were 10-fold higher. The CEP55 EABR binds both ALIX and TSG101, 17 whereas EIAV p9 only binds ALIX, 23 suggesting that optimal recruitment of both ESCRT proteins is required for efficient eVLP assembly. The S-p6 and S-VP40 1-44 samples contained little or no S protein, suggesting that eVLP assembly was inefficient, possibly resulting from lower affinities for ESCRT proteins ( Figures 1C and S1B).
We further characterized the S-EABR construct by experimenting with different EABR sequences ( Figure S1A), finding that addition of a second EABR domain (S-2xEABR) reduced eVLP production ( Figure 1D). To investigate whether S-EABR eVLP assembly is dependent on ESCRT recruitment, we generated S-EABR mut by substituting an EABR residue (Tyr187 in CEP55) that is essential for interacting with ALIX 17 ( Figure S1A). While the purified S-EABR eVLP sample produced an intense band at a 1:200 dilution, no band was detected for S-EABR mut at a 1:20 dilution, suggesting that eVLP production was abrogated for S-EABR mut and highlighting the importance of ALIX recruitment for eVLP assembly ( Figure 1D). To identify the minimal EABR sequence required for eVLP assembly, we designed S constructs fused to the complete EABR domain (CEP55 170-213 ), EABR min1 (CEP55 180-213 ), and EABR min2 (CEP55 180-204 ) (Figure S1A). While S-EABR eVLP yields were diminished for EABR min2 , production efficiency was retained for EABR min1 (Figure 1E). To assess the effects of the EPM within the cytoplasmic tail of the S-EABR construct, we evaluated eVLP production for an S-EABR construct that did not include the EPM. Western blot analysis demonstrated that increased amounts of S protein were detected after eVLP purification from cells transfected with S-EABR compared with S-EABR/no EPM, suggesting that the EPM enhances eVLP production ( Figure 1F).
We also compared the S-EABR construct with other eVLP approaches 28 that require co-expression of S protein with structural viral proteins, such as HIV-1 Gag 29 or the SARS-CoV-2 M, N, and E proteins. 30 Western blot analysis showed that purified S-EABR eVLP fractions contained at least 10-fold more S protein than eVLPs produced by co-expression of S and Gag or S, M, N, and E ( Figure 1G), suggesting that S-EABR eVLPs assemble and/or incorporate S proteins more efficiently than the other eVLP approaches. Purified S-EABR eVLPs also contained higher levels of S protein compared with S-ferritin nanoparticles purified from transfected cell supernatants, which have been shown to elicit potent immune responses in animal models 31,32 ( Figure 1G).
3D reconstructions derived from cryo-ET showed purified S-EABR eVLPs with diameters ranging from 40 to 60 nm that are surrounded by a lipid bilayer and the majority of which were densely coated with spikes ( Figures 1H and 1I; Video S1). To estimate the number of S trimers, we counted trimer densities in 4 nm computational tomographic slices of individual eVLPs, finding 10-40 spikes per particle that were heterogeneously distributed on the surface of eVLPs. The upper limit of the number of spikes on eVLPs roughly corresponds to spike numbers on larger SARS-CoV-2 virions (>100 nm in diameter) 33 ; thus, the spike densities on the majority of eVLPs exceed those on authentic viruses. Spikes on eVLPs were separated by distances of 20-26 nm (measured between the centers of trimer apexes) for densely coated particles ( Figures 1H and 1I). To assess the general applicability of the EABR approach, we also generated EABR eVLPs for HIV-1 Env, which produced eVLPs with higher Env content than co-expression of Env and HIV-1 Gag (Figure S1C), and for the multi-pass transmembrane protein CCR5 ( Figure S1D). Taken together, these results are consistent with efficient incorporation of S proteins into S-EABR eVLPs that are released from transfected cells and suggest that the EABR technology can be applied to a wide range of membrane proteins.

S-EABR eVLPs induce potent antibody responses in immunized mice
The potential of purified S-EABR eVLPs as a vaccine candidate against SARS-CoV-2 was evaluated in C57BL/6 mice (Figure 2A).
S-EABR eVLPs were purified from transfected cell supernatants by ultracentrifugation on a 20% sucrose cushion followed by size exclusion chromatography (SEC), and S protein concentrations were determined by quantitative western blot analysis ( Figures  S2A and S2B). Immunizations with S-EABR eVLPs were compared with purified soluble S and to soluble S covalently attached to SpyCatcher-mi3 protein nanoparticles (S-mi3). 34 0.1 mg doses (calculated based on S protein content) were (C-G) Western blot analysis detecting SARS-CoV-2 S1 protein on eVLPs purified by ultracentrifugation on a 20% sucrose cushion from transfected Expi293F cell culture supernatants.
(C) Cells were transfected with S-EABR, S-p9, S-VP40 1-44 , or S-p6 constructs. The purified S-EABR eVLP sample was diluted 1:400 (left), while S-p9, S-VP40 1-44 , and S-p6 samples were diluted 1:40 (right). Comparison of band intensities between lanes suggests that the S-EABR eVLP sample contained 10-fold higher levels of S1 protein than the S-p9 sample and >10-fold higher levels than the S-VP40 1-44 and S-p6 samples. (I) Model of a representative S-EABR eVLP derived from a cryo-ET reconstruction (Video S1). Coordinates of an S trimer (PDB: 6VXX) 27 were fit into protruding density on the best resolved half of an eVLP and the remainder of the eVLP was modeled assuming a similar distribution of trimers. The position of the lipid bilayer is shown as a 55-nm gray sphere. See also Figure S1 and Video S1.  Figures 2B and 2C). In contrast, no neutralizing antibody responses were detected for soluble S protein immunization after the prime. Neutralizing antibody titers elicited by S-EABR eVLPs and S-mi3 increased by >10-fold after boosting and were >20fold higher than titers measured for soluble S ( Figure 2C). S-EABR eVLPs elicited potent antibody responses targeting the receptor-binding domain (RBD) of the S protein ( Figure S2C), a primary target of anti-SARS-CoV-2 neutralizing antibodies. 35 Serum responses were also evaluated against authentic SARS-CoV-2 by plaque reduction neutralization tests (PRNTs), showing robust neutralizing activity against SARS-CoV-2 WA1 ( Figure S2D). Neutralization titers dropped 4and 2-fold against the SARS-CoV-2 Beta and Delta variants, respectively, consistent with studies of licensed vaccines that encode the SARS-CoV-2 WA1 S protein. 36 These results demonstrate that purified S-EABR eVLPs elicit potent immune responses in vivo and represent an alternative technology for producing nanoparticle-based vaccines that does not involve detergent-mediated cell lysis and separation of membrane protein antigens from cell lysates, as required for protein nanoparticle vaccines such as NVX-CoV2373, a COVID-19 vaccine, 9,10 or FluBlok, an influenza vaccine. 37

mRNA-encoded S-EABR construct induces cell surface expression and eVLP budding
A key advantage of the EABR eVLP technology over existing nanoparticle-based vaccine approaches is that S-EABR con- Dashed horizontal lines correspond to the background values representing the limit of detection for neutralization assays. Significant differences between cohorts linked by horizontal lines are indicated by asterisks: *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S2.
structs can be easily delivered as mRNA vaccines since both eVLP assembly and cell surface expression only require expression of a single genetically encoded component. While conventional COVID-19 mRNA vaccines induce antibody responses through cell surface expression of S protein ( Figure 3A, top), mRNA-mediated delivery of an S-EABR construct could enhance B cell activation because S-EABR proteins will not only be expressed at the cell surface-they will also induce assembly of eVLPs that bud from the cell and distribute inside the body to activate immune cells ( Figure 3A, bottom).
To investigate whether genetic encoding of S-EABR eVLPs enhances the potency of a SARS-CoV-2 S-based mRNA vaccine, we started by synthesizing nucleoside-modified mRNAs encoding S, S-EABR, S-EPM, or S-EABR/no EPM. Cell surface expression and eVLP assembly were evaluated by flow cytometry and western blot analysis 48 h after in vitro transfection of mRNAs in HEK293T cells, demonstrating higher surface expression for S compared with the S-EABR fusion protein ( Figure 3B). While addition of the EPM had little effect on S surface expression, removal of the EPM lowered surface levels for the S-EABR construct. Western blot analysis of supernatants confirmed that the S and S-EPM transfections did not generate detectable eVLPs in supernatants, whereas eVLPs were strongly detected in supernatants from S-EABR transfected cells (Figure 3C). eVLP production was decreased for S-EABR/no EPM, which, together with the flow cytometry results ( Figure 3B), suggests that EPM addition enhances both S-EABR cell surface expression and eVLP assembly.
The observed reduction in S cell surface expression in the S-EABR versus S mRNA transfections could be caused by lower overall cell surface expression of the S-EABR fusion protein, incorporation of S-EABR proteins into eVLPs that bud from the cell surface, or both. To evaluate these possibilities, we calculated approximate numbers of S trimers expressed from the S-EABR construct. Assuming that 3 3 10 6 cells were transfected (6-well plate) and up to 1 3 10 5 S trimers were expressed on the surface of each cell (based on the approximate number of B cell receptors on a B cell 38 ), transfected cell surfaces would contain 0.5 pmol or 70 ng of total S protein. Supernatant samples for western blots were concentrated to a final volume of 200 mL of which 1.2 mL was loaded onto a gel. As the detection limit for S1 is 20 ng, the western blot analysis suggested that purified S-EABR eVLPs from transfected cell supernatants contained at least 17 ng/mL S protein, corresponding to >3 mg S protein in the purified transfected cell supernatant. These calculations suggested that the observed reduction in cell surface expression for the S-EABR construct was at least partially caused by incorporation of S-EABR proteins into budding eVLPs that were released into the supernatant. Given that the estimated S protein content on released eVLPs exceeded the approximate amount of S protein presented on cell surfaces, it is possible that the S-EABR construct induces higher overall expression of S antigens compared with S for which expression is restricted to cell surfaces. Taken together, the mRNA transfection results demonstrate that the mRNA-encoded S-EABR construct enables dual presentation of S antigens on cell surfaces and released eVLPs.

S-EABR mRNA-LNP elicit superior antibody titers compared with conventional vaccines
The effect of eVLP production on mRNA vaccine potency was evaluated in BALB/c mice by comparing mRNAs encoding S or S-EABR constructs that were encapsulated in LNP ( Figure 4A).
As described for preclinical studies of a COVID-19 mRNA vaccine in mice, 1 mRNA-LNP were administered intramuscularly (IM) at a dose of 2 mg mRNA on days 0 and 28. mRNA-LNP immunizations were also compared with purified S-EABR eVLPs that were injected IM in the presence of Addavax adjuvant. After the prime, S and S-EABR mRNA-LNP elicited significantly higher antibody binding responses against the SARS-CoV-2 S protein than purified S-EABR eVLPs ( Figure 4B). However, the highest neutralizing antibody titers were elicited by purified S-EABR eVLPs, which were significantly higher than titers elicited by the S mRNA-LNP ( Figure 4C).
We also evaluated serum neutralizing activity against SARS-CoV-2 VOCs. S-EABR mRNA-LNP elicited 4.9-and 6.5-fold higher mean neutralizing responses against the Delta variant compared with S mRNA-LNP, as well as 4.6-and 9.4-fold higher  Article titers compared with purified S-EABR eVLPs on days 56 and 112, respectively ( Figures 4D and 4E). Against Omicron BA.1, neutralizing antibody responses dropped markedly for all groups, except for mice that received S-EABR mRNA-LNP, which elicited 15.1-and 9.5-fold higher neutralizing titers than S mRNA-LNP and 20.7-and 15.4-fold higher titers than purified S-EABR eVLPs on days 56 and 112, respectively ( Figures 4D and  4F). Against Omicron BA.2, mean neutralization titers measured for mice that received S-EABR mRNA-LNP were also 10.9-and 8.2-fold higher compared with S mRNA-LNP and 7-and 12.2fold higher compared with purified S-EABR eVLPs on days 56 and 112, respectively, but these differences narrowly failed to reach statistical significance ( Figures 4D and 4G). Together, these results demonstrate that mRNA-mediated delivery of S-EABR eVLPs enhances the potency and breadth of humoral immune responses in mice compared with conventional mRNA and protein nanoparticle-based vaccine approaches. The observed improvements in neutralizing activity against Omicron-based VOCs were substantially larger than the 1.5-fold increases reported for recently approved bivalent mRNA booster shots, 39 suggesting that S-EABR mRNA-LNP-based booster immunizations could induce more effective and lasting immunity against Omicron-based and emerging VOCs than current COVID-19 vaccines.

S-EABR mRNA-LNP induce potent T cell responses
On day 112 (3 months post-boost), splenocytes were isolated from immunized mice to analyze T cell responses by enzymelinked immunosorbent spot (ELISpot) assays. 40 Splenocytes were stimulated with a pool of SARS-CoV-2 S-specific peptides, and INF-g and IL-4 secretion were measured to evaluate T cell activation. mRNA-LNP encoding S and S-EABR constructs induced potent INF-g responses, consistent with the presence of S-specific cytotoxic CD8 + T cells and T helper 1 (T H 1) cellular immune responses ( Figure 5A). In contrast, INF-g responses were almost undetectable for mice immunized with purified S-EABR eVLPs ( Figure 5A). These results were expected as mRNA-LNP immunizations result in intracellular expression of S or S-EABR immunogens and MHC class I presentation of antigenic peptides that activate CD8 + T cells, which does not commonly occur for protein nanoparticle-based vaccines. 8 S-EABR mRNA-LNP induced significantly stronger IL-4 responses compared with S mRNA-LNP and purified S-EABR eVLPs ( Figure 5B), consistent with potent T H 2 cellular immune responses. While T H 1-and T H 2-biased responses were observed for S mRNA-LNP and purified S-EABR eVLPs, respectively, S-EABR mRNA-LNP induced a balanced T H 1/T H 2 response, thereby potently stimulating cellular and humoral immune responses. Thus, S-EABR mRNA-LNP retain the ability of conventional S mRNA-LNP to activate potent cytotoxic CD8 + T cell responses, while also potently activating T H 2 CD4 + T cell responses to enhance humoral immune responses leading to increased antibody potency and breadth.

DISCUSSION
Here, we present a new technology to generate eVLPs for vaccine and other applications. The approach harnesses the ESCRT pathway that is involved in cell division and viral budding 18,19 to drive assembly and release of eVLPs that present membrane proteins containing a cytoplasmic ESCRT-recruiting motif, the EABR sequence from the human centrosomal protein CEP55. 41 Our results demonstrate that the EABR-based platform produces eVLPs that incorporate higher levels of membrane antigens compared with approaches that require co-expression of the antigen with viral capsid proteins such as Gag or with the SARS-CoV-2 M, N, and E proteins. Purified S-EABR eVLPs elicited potent antibody responses against SARS-CoV-2 in mice that were similar in magnitude to those elicited by a 60-mer protein nanoparticle displaying S trimers. Compared with existing protein nanoparticle-based vaccine approaches, the EABR technology exhibits attractive manufacturing properties as (1) eVLP production requires expression of only a single component, (2) transmembrane proteins are retained in their native membrane-associated conformation to ensure optimal protein expression and stability, and (3) fully assembled eVLPs can be purified directly from culture supernatants without requiring detergent-mediated cell lysis and separation of membrane protein antigens from cell lysates. The lipid bilayer surrounding eVLPs also prevents offtarget antibody responses against a nanoparticle scaffold that have been reported for protein nanoparticle-based immunogens. 42 Due to its modularity, flexibility, and versatility, the EABR technology could potentially be used to generate eVLPs presenting a wide range of surface proteins for vaccine and therapeutic applications.
To optimize the EABR technology, we evaluated several ESCRT-recruiting motifs for their ability to drive eVLP assembly, including viral late domains from EIAV, HIV-1, and EBOV. The EABR from CEP55 generated eVLPs 10-fold more efficiently than the EIAV late domain p9. The EABR binds to ESCRT proteins ALIX and TSG101, 17 while p9 binds only to ALIX, 23 suggesting that efficient eVLP assembly requires recruitment of both proteins. HIV-1 p6 contains motifs that interact with both TSG101 and ALIX, 23,25 but S-p6 constructs did not induce detectable eVLP budding in our experiments, perhaps because reported affinities are relatively low 23,43 compared with TSG101 and ALIX affinities reported for the EABR. 17 eVLP production might be optimized by designing ESCRT-binding motifs with increased affinities for ESCRT proteins. We were able to enhance eVLP production by including an EPM derived from the FcgRII-B1 cytoplasmic tail 44 to reduce endocytosis of EABR-fusion proteins, which increased S-EABR cell surface expression and eVLP production.
An advantage of the EABR technology is that constructs can be easily delivered as mRNA vaccines since eVLP assembly requires expression of only a single component. This strategy results in presentation of viral surface antigens on the cell surface (E-G) Neutralization data from the indicated time points for antisera from individual mice (colored circles) presented as the geometric mean (bars) and standard deviation (horizontal lines). Neutralization results against SARS-CoV-2 Delta (E), Omicron BA.1 (F), and Omicron BA.2 (G) pseudoviruses are shown as ID 50 values. Dashed horizontal lines correspond to the background values representing the limit of detection for neutralization assays. Significant differences between cohorts linked by horizontal lines are indicated by asterisks: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

OPEN ACCESS
and on released eVLPs that could distribute throughout the body, thereby combining immune responses elicited by both conventional mRNA and protein nanoparticle-based vaccines. S-EABR mRNA-LNP elicited significantly higher binding and neutralizing antibody responses compared with conventional S-based mRNA-LNP analogous to current COVID-19 mRNA vaccines and to purified S-EABR eVLPs, suggesting that dual presentation of viral surface antigens on cell surfaces and eVLPs potentiates B cell activation. Presentation of viral surface antigens on cell surfaces alone potentially restricts expression for conventional mRNA vaccines due to a finite, and presumably limited, environment for insertion of both delivered and endogenous membrane proteins. Thus, combining cell surface expression and eVLP release for the S-EABR mRNA-LNP may increase overall presentation of viral surface antigens to the immune system. It is also possible that mRNA-mediated S-EABR eVLP production expands the biodistribution of viral surface antigens to more effectively engage B cells in lymph nodes distant from the injection site. The enhanced humoral immune responses elicited by S-EABR mRNA-LNP were consistent with potent T H 2 cellular responses observed in S-EABR mRNA-LNP-immunized mice, which were more pronounced than in mice immunized with S mRNA-LNP or purified S-EABR eVLPs. Importantly, cytotoxic CD8 + T cell responses were maintained in S-EABR mRNA-LNP compared with S mRNA-LNP-immunized animals. Thus, S-EABR mRNA-LNP potently stimulate both cellular and humoral immune responses.
The higher peak antibody levels elicited by the S-EABR mRNA-LNP would likely impact the durability of protective antibody responses. Notably, differences in serum antibody titers across different immunizations were maintained until 3 months post-boost, suggesting that antibody levels might contract at similar rates for the tested vaccine types. Hence, the elevated peak antibody titers elicited by the S-EABR mRNA-LNP could result in markedly prolonged periods of immune protection compared with conventional vaccine approaches, which could minimize the need for frequent booster immunizations. Longterm studies that monitor antibody levels for several months are needed to elucidate the relationship between peak antibody titers and durability of responses.
Two immunizations with S-EABR mRNA-LNP also elicited potent neutralizing antibody responses against SARS-CoV-2 Delta and Omicron-based VOCs, suggesting that higher antibody responses could lead to enhanced protection against viral escape variants. The conventional S-based mRNA-LNP immunization only elicited weak responses against Omicron-based VOCs, consistent with outcomes reported in humans in which weak Omicron-specific responses to WA1-based vaccines were enhanced after a 3 rd immunization. 13,45 S-EABR mRNA-LNP elicited >10-fold higher neutralizing antibody titers against Omicron BA.1 and BA.2 VOCs compared with S mRNA-LNP after only two immunizations, suggesting that the simple addition of a short EABR-encoding sequence to the spike gene in current mRNA vaccines could have limited the global spread of Omicron-based VOCs. Our results also suggest that S-EABR mRNA-LNP-based booster immunizations would induce superior immunity against Omicron-based and emerging VOCs compared with current boosting strategies, as bivalent booster shots that contain ancestral and Omicron-based variants improve neutralizing antibody titers by only 1.5-fold compared with conventional booster shots. 39 Future studies need to investigate whether the observed increase in neutralization activity against Omicron-based VOCs results from higher overall antibody levels and/or increased antibody targeting of sub-immunodominant conserved epitopes on S trimer.
Enhanced antibody responses compared with S mRNA-LNP have also been reported for co-delivery of mRNAs encoding SARS-CoV-2 S, M, and E proteins, which should result in dual presentation of S on cell surfaces and released eVLPs. 46 However, higher mRNA doses (10 mg) were needed to deliver all three mRNAs, and only modest improvements (2.5-fold) in neutralizing antibody titers were achieved. Our results showed that S-EABR eVLPs assemble more efficiently in vitro than eVLPs driven by co-expression of S, M, N, and E proteins, potentially explaining why S-EABR mRNA-LNP induced larger increases in antibody titers at lower doses. Co-delivery of multiple mRNAs also poses an obstacle for vaccine manufacturing, whereas COVID-19 and other mRNA vaccines could be easily modified to generate eVLPs by adding a short sequence containing EABR and EPM motifs to the cytoplasmic domains of the Significant differences between cohorts linked by horizontal lines are indicated by asterisks: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. encoded immunogens. mRNA delivery of a trimerized RBDferritin fusion construct, which should result in secretion of non-enveloped ferritin nanoparticles displaying trimeric RBDs without cell surface expression of RBDs, has also been reported. 47 This approach was not compared with a conventional S mRNA-LNP-based immunogen, highlighting the need for comparison studies of different vaccine approaches to elucidate the individual effects of antigen presentation on cell surfaces and virus-like nanoparticles on the magnitude and quality of immune responses.
In summary, we present a new technology to efficiently generate eVLPs for vaccine and other therapeutic applications. We demonstrate that an mRNA vaccine encoding SARS-CoV-2 S-EABR eVLPs elicits antibody responses with enhanced potency and breadth compared with conventional vaccine strategies in mice, which warrants further investigation in other preclinical animal models and humans as a vaccine strategy.
Limitations of the study Since our study involves immunization studies performed in mice, future studies will need to evaluate whether S-EABR mRNA immunizations also elicit more potent and broad antibody responses in non-human primates and humans compared with conventional mRNA vaccine strategies. Because binding and neutralizing antibody responses correlate with protection in humans and animals vaccinated with COVID-19 mRNA vaccines, 48-50 the strong antibody responses elicited by S-EABR mRNA immunizations are predictive of protection. However, viral challenge studies in animals could provide further evidence that S-EABR mRNA immunizations induce more effective protection against Omicron-based variants. In addition, although in vitro experiments showed that S-EABR protein is presented on cell surfaces and on released eVLPs, we have no direct evidence that mRNA-encoded delivery of the S-EABR construct resulted in eVLP production in vivo. Thus, future studies are needed to confirm eVLP production and distribution in vivo and investigate how released eVLPs affect immune cell activation. The effects of S-EABR mRNA immunizations on T cells, Fcmediated effector functions, and other aspects of the immune response are also needed to fully assess the potential of the EABR vaccine approach. Finally, the effectiveness of the EABR vaccine platform against other viral pathogens needs to be evaluated.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We thank J. Vielmetter and the Caltech Protein Expression Center for assistance with protein production; K. Dam for biotinylated proteins for ELISAs; M. Anaya for a BirA expression plasmid; and J. Bloom (Fred Hutchinson) and P. Bieniasz (Rockefeller University) for neutralization assay reagents. We thank J. Keeffe, Y. Tam (Acuitas Therapeutics), C. Barnes

INCLUSION AND DIVERSITY
We support inclusive, diverse, and equitable conduct of research.

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contacts, Magnus A.G. Hoffmann (mhoffman@caltech.edu) and Pamela J. Bjorkman (bjorkman@caltech.edu).

Materials availability
All expression plasmids generated in this study are available upon request through a Materials Transfer Agreement.
Data and code availability All data are available in the main text or the supplemental information. Materials are available upon request to the corresponding authors with a signed material transfer agreement. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. This paper does not report original code. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Bacteria E. coli DH5 Alpha cells (Zymo Research) used for expression plasmid productions were cultured in LB broth (Sigma-Aldrich) with shaking at 250 rpm at 37 C. E. coli BL21-CodonPlus (DE3)-RIPL cells (Agilent Technology) used for producing SpyCatcher003-mi3 were cultured in 2xYT media with shaking at 220 rpm at 37 C, IPTG was added at OD of 0.5 and induction lasted for 5 hours at 30 C.
Cell lines HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Sigma-Aldrich) and 1 U/ml penicillin-streptomycin (Gibco) at 37 C and 5% CO 2 for pseudovirus production. Expi293F cells (Gibco) for protein expression were maintained at 37 C and 8% CO 2 in Expi293 expression medium (Gibco). Transfections were carried out with an Expi293 Expression System Kit (Gibco) and maintained under shaking at 130 rpm. All cell lines were derived from female donors and were not specially authenticated.

Viruses
Pseudovirus stocks were generated by transfecting HEK293T cells with pNL4-3DEnv-nanoluc and SARS-CoV-2 S constructs 52 using FuGENE HD (Promega); co-transfection of pNL4-3DEnv-nanoluc with a SARS-CoV-2 S construct will lead to the production of HIV-1-based pseudovirions carrying the coronavirus S protein at the surface. Eight hours after the transfection, cells were washed twice with phosphate buffered saline (PBS) and fresh media was added. Pseudoviruses in the supernatants were harvested 48 hours post-transfection, filtered, and stored at -80 C until use. Infectivity of pseudoviruses was determined by titration on HEK293T-ACE2 cells.

Supplemental figures
(C) ELISA data from day 42 for antisera from individual mice (colored circles) immunized with soluble S (purified S trimer) (gray), S-mi3 (S trimer ectodomains covalently attached to mi3, a 60-mer protein nanoparticle) (blue), or S-EABR eVLPs (green). Results are shown as area under the curve (AUC) and presented as the geometric mean (bars) and standard deviation (horizontal lines). Significant differences between cohorts linked by horizontal lines are indicated by asterisks: *p < 0.05; **p < 0.01. (D) PRNT assay results from day 56 for antisera from individual mice (colored circles) immunized with S-EABR eVLPs. Results against the SARS-CoV-2 WA1 (green), Beta (orange), and Delta (brown) variants are shown as TCID 50 values 72 and presented as the geometric mean (bars) and standard deviation (horizontal lines).