Heteromultimeric sarbecovirus receptor binding domain immunogens primarily generate variant-specific neutralizing antibodies

Significance Multimeric vaccines containing variant proteins from related viruses are expected to generate broader neutralizing antibody responses. Combining variants in a single heteromultimeric immunogen particle or polypeptide may preferentially stimulate B cells that recognize multiple viral variants and subsequently produce cross-reactive antibodies. We show that cross-reactive B cells and antibodies can indeed be elicited using immunogens built from two distinct SARS-like coronaviruses receptor binding domains. However, most antibodies with virus neutralizing activity generated by such immunogens recognize viruses corresponding to only one of the two immunogen components. This is likely because sequence variation is concentrated in areas recognized by neutralizing antibodies; thus, the cross-reactive B cells mostly recognize sequences conserved in multiple viral variants and primarily generate antibodies lacking neutralizing activity.

Vaccination will likely be a key component of strategies to curtail or prevent future sarbecovirus pandemics and to reduce the prevalence of infection and disease by future SARS-CoV-2 variants.A "pan-sarbecovirus" vaccine, that provides maximum possible mitigation of human disease, should elicit neutralizing antibodies with maximum possible breadth.By positioning multiple different receptor binding domain (RBD) antigens in close proximity on a single immunogen, it is postulated that cross-reactive B cell receptors might be selectively engaged.Heteromultimeric vaccines could therefore elicit individual antibodies that neutralize a broad range of viral species.Here, we use model systems to investigate the ability of multimeric sarbecovirus RBD immunogens to expand cross-reactive B cells and elicit broadly reactive antibodies.Homomultimeric RBD immunogens generated higher serum neutralizing antibody titers than the equivalent monomeric immunogens, while heteromultimeric RBD immunogens generated neutralizing antibodies recognizing each RBD component.Moreover, RBD heterodimers elicited a greater fraction of cross-reactive germinal center B cells and cross-reactive RBD binding antibodies than did homodimers.However, when serum antibodies from RBD heterodimer-immunized mice were depleted using one RBD component, neutralization activity against the homologous viral pseudotype was removed, but neutralization activity against pseudotypes corresponding to the other RBD component was unaffected.Overall, simply combining divergent RBDs in a single immunogen generates largely separate sets of individual RBD-specific neutralizing serum antibodies that are mostly incapable of neutralizing viruses that diverge from the immunogen components.
Vaccination will likely be a key component of strategies to curtail or prevent future sarbecovirus pandemics and to reduce the prevalence of infection and disease by future SARS-CoV-2 variants.A "pan-sarbecovirus" vaccine, that provides maximum possible mitigation of human disease, should elicit antibodies with maximum possible breadth, ideally with pan-sarbecovirus neutralizing activity (11).Such a vaccine might prevent transmission of new SARS-CoV-2 variants and curtail nascent epidemics of sarbecoviruses that may emerge from animal reservoirs.However, antigenic variation among animal sarbecoviruses is considerable, and antigenic drift in SARS-CoV-2 is ongoing (12).
Neutralizing antibody-based protection against diverse sarbecoviruses could, in prin ciple, be achieved by vaccination in distinct ways.In one scenario, many different, narrowly specific antibodies would provide protection against many well-matched extant viruses (11).This scenario would necessitate the administration of many, or highly multivalent, vaccines.While potentially effective against known virus threats, such a strategy would be unlikely to provide protection against any divergent virus or variant whose emergence was not anticipated and thus was not included in the vaccine cocktail.In another scenario, protection could be provided by a smaller number of broadly neutralizing antibody lineages that are effective against many different viruses with diverse sequences (13)(14)(15)(16).Such antibodies would need to target invariant neutralizing epitopes or tolerate sequence variation in the targeted epitopes.Broadly cross-reactive antibodies would be more likely to be capable of neutralizing viruses whose spillover was unantic ipated.However, such a scenario would require the development of vaccines that are capable of efficiently generating such broadly cross-reactive neutralizing antibodies.
In the case of sarbecoviruses, the spike protein receptor binding domain (RBD) constitutes the major target of potent neutralizing antibodies and includes epitopes that vary to differing degrees among sarbecoviruses (5,17).For example, the majority of anti bodies elicited by SARS-CoV-2 do not neutralize SARS-CoV, but a subset that neutralizes both SARS-CoV and SARS-CoV-2 target less variable portions of the RBD (15,18,19).Notably, SARS-CoV-2 RBD-specific neutralizing antibodies in the B cell memory com partment can acquire breadth during months of affinity maturation following infection and vaccination (20)(21)(22)(23) and, thus, the ability to neutralize SARS-CoV-2 variants or other sarbecoviruses (such as SARS-CoV) whose epitope target sequences differ from the original antigen (24).Similarly, after multiple antigen exposures, polyclonal sera from SARS-CoV or SARS-CoV-2 infected and later vaccinated individuals can neutralize SARS-CoV, other sarbecoviruses and some SARS-CoV-2 variants (17,25,26).
Heteromultimeric vaccines such as mosaic nanoparticles could potentially selectively elicit individual antibodies with the ability to neutralize a broad range of viral species by positioning multiple dif ferent antigens in close proximity on a single immunogen (27)(28)(29)(30)(31)(32).In principle, selective stimulation and expansion of B cells whose antigen receptors target conserved epitopes or, following diversifica tion, are preferentially cross-linked by heteromultimeric antigens might be achieved.Here, we use model systems to investigate the ability of multimeric sarbecovirus RBD immunogens to elicit cross-reactive B cells and generate broadly cross-reactive serum antibodies.

Results
Heteromultimeric Sarbecovirus RBD Immunogens.We first com pared the ability of monomeric and multimeric homotypic SARS-CoV-2 RBD-based immunogens to elicit neutralizing antibodies.A SARS-CoV-2 RBD protein from the Wuhan-hu-1 strain (RBD Wu ) was expressed as a monomeric protein, as a genetically fused single polypeptide dimer (RBD Wu -RBD Wu ), or a dimer by fusion to an IgG1 Fc domain, or as a high order 60-mer nanoparticle by fusion to a modified Lumazine Synthase (LzS) (33) (Fig. 1A and SI Appendix, Fig. S1A).Mice were inoculated with each protein immunogen at week 0 and week 3. Neutralizing titers against SARS-CoV-2 Wu pseudotypes plateaued at ~week 5 and remained constant until at least week 12 (Fig. 1B).Even though the same total mass of protein was given in all cases, each of multimeric proteins generated mean SARS-CoV-2 pseudotype neutralization titers that were sevenfold to 14fold greater than those generated by the monomeric RBD Wu protein (P = 0.0079, Fig. 1C).The homomultimeric RBD Wu immunogens generated similar neutralizing antibody titers, irrespective of how multimerization was mediated.
We next generated a fusion dimer, Fc and LzS based homomul timeric immunogens containing the RBD of SARS-CoV (RBD 1 ) as well as heteromultimeric immunogens containing both RBD 1 and RBD Wu (SI Appendix, Fig. S1 A and B) and compared the ability of homomultimers and heteromultimers to generate anti bodies that neutralized SARS-CoV and SARS-CoV-2 Wu .The mean NT 50 values elicited by RBD 1 homomultimers were 8346 to 49464 against SARS-CoV but only <10 to 215 against SARS-CoV-2 Wu (Fig. 1D).Conversely, NT 50 values generated by RBD Wu homomultimers were 23,870 to 73,325 against SARS-CoV-2 Wu but only 79 to 1,027 against SARS-CoV (Fig. 1D).The heteromultimers combining both RBD 1 and RBD Wu , elicited high titer neutralizing antibodies against both viruses (mean NT 50 = 24,840 to 55,079 against SARS-CoV and 4,257 to 21,675 against SARS-CoV-2 Wu ) (Fig. 1D).Overall, the homomultimeric immunogens elicited neutralizing antibodies that were largely specific for the autologous virus, while the heteromultimers elic ited antibodies that neutralized both viruses.

Expansion of Cross-Reactive B Cells by RBD Heterodimers.
Although each of the various heteromultimeric immunogens generated similar neutralization titers against both SARS-CoV and SARS-CoV-2 Wu (Fig. 1D), the Fc fused and LzS nanoparticle fused RBD proteins differ from the fusion dimer in that they generate a mixture of heteromultimeric immunogens of undefined composition (Fig. 1A).Specifically, the coexpressed Fc fused RBD 1 and RBD Wu are expected to generate mixtures of homoand heterodimers, while the juxtaposition of RBD 1 and RBD Wu on a mixed LzS nanoparticle will inevitably be heterogeneous.These properties might reduce the ability of the heteromultimeric immunogen to selectively cross-link B cell receptors that crossreact with both of the RBD components and thus reduce the selective expansion of cross-reactive B cells.Therefore, we focused our attention on the RBD 1 -RBD Wu fusion dimers, where the composition and juxtaposition of RBD 1 and RBD Wu are defined (Fig. 1A).

SARS-CoV and SARS-CoV-2 RBD Cross-Reactive Antibody Binding.
To assess the degree to which the RBD-binding serum antibodies elicited by homodimers and heterodimers were cross-reactive with both RBD 1 and RBD Wu , we established a competition binding assay (SI Appendix, Fig. S3 A-C).RBD proteins were expressed as a fusion with NanoLuc luciferase, and each of the RBD-NanoLuc fusion proteins (SI Appendix, Fig. S3A) was incubated with serum from mice immunized with RBD 1 -RBD 1 , RBD Wu -RBD Wu , and RBD 1 -RBD Wu beginning 11 to 12 wk previously.Then, the antibody:RBD-NanoLuc complexes were magnetically separated with protein G Dynabeads (SI Appendix, Fig. S3B), and the RBD 1 -NanoLuc and RBD Wu -NanoLuc binding capacity of each mouse serum was determined in the absence of competitor (SI Appendix, Fig. S3C).Thereafter, serum volumes containing equivalent amounts of RBD-NanoLuc binding antibody were simultaneously incubated with a fixed, saturating concentration of RBD-NanoLuc (10 ng) along with a variable (0.1× to 50×) concentration of an unlabelled RBD 1 or RBD Wu competitor (SI Appendix, Figs.S3B and S4 and Fig. 3 A-C).As expected, homologous competition binding assays (RBD 1 competition with RBD 1 -NanoLuc and RBD Wu competition with RBD Wu -NanoLuc) yielded mean IC 50 values of 4.2 ± 1.1 to 19.9 ± 2.7 ng/mL, regardless of whether animals were immunized with RBD and RBD 1 -RBD Wu (Fig. 3 A-C and SI Appendix, Fig. S4).Conversely, heterologous binding competition assays (RBD Wu competition with RBD 1 -NanoLuc and RBD 1 competition with RBD Wu -NanoLuc) yielded results that were highly dependent on the immunogen.In RBD 1 -RBD 1 immunized mice, RBD 1 competed with RBD Wu -NanoLuc (mean IC 50 = 3.8 ± 1.6 ng/mL) but RBD Wu did not compete with RBD 1 -NanoLuc (mean IC 50 > 1,000 ng/mL) (Fig. 3A and SI Appendix, Fig. S4).In RBD Wu -RBD Wu immunized mice, RBD 1 did not compete with RBD Wu -NanoLuc (mean IC 50 = >1000 ng/mL) but RBD Wu competed with RBD 1 -NanoLuc (mean IC 50 = 6.4 ± 2.7 ng/mL) (Fig. 3B and SI Appendix, Fig. S4).These results suggest that even though RBD 1 and RBD Wu share 78% sequence identity (SI Appendix, Fig. S1C, the majority of RBD binding serum antibodies elicited by RBD 1 -RBD 1 or RBD Wu -RBD Wu homodimers were not crossreactive with the heterologous RBD. In mice immunized with RBD 1 -RBD Wu heterodimers, there was competition between unfused RBD and heterologous RBD-NanoLuc serum antibodies that was variable among individual mice.Specifically, RBD 1 competed for RBD Wu -NanoLuc binding (mean IC 50 = 54 ± 79 ng/mL) and in one of five mice, and RBD Wu competed for RBD 1 -NanoLuc binding (mean IC 50 = 784 ± 429 ng/mL) (Fig. 3C and SI Appendix, Fig. S4).Thus, the heterodi meric RBD 1 -RBD Wu immunogen exhibited a greater propensity to elicit cross-reactive antibodies that bound to both RBD 1 and RBD Wu than did either homodimer.

Antibodies Elicited by SARS-CoV-2 Variant Homodimers and
Heterodimers.We next constructed RBD homodimers and heterodimers based on SARS-CoV-2 variants that are more closely related to each other than SARS-CoV and SARS-CoV-2 Wu are (SI Appendix, Fig. S1).Specifically, we chose SARS-CoV-2 Wu and SARS-CoV-2 BA.5 , whose RBDs share 93% amino acid identity, but differ in susceptibility to many (but not all) SARS-CoV-2 neutralizing monoclonal antibodies (7).We immunized groups of mice with RBD Wu -RBD Wu or RBD BA.5 -RBD BA.5 homodimers or RBD Wu -RBD BA.5 heterodimers.Then, using sera collected 11 to 12 wk after the first immunogen injection, we performed the RBD:RBD-NanoLuc fusion protein competition binding and RBD-6xHis-Dynabead neutralization-depletion assays to determine the extent to which homodimer and heterodimer immunized mice generated RBD Wu and RBD BA.5 cross-reactive binding and neutralizing antibodies.
In sera from mice immunized with RBD Wu -RBD BA.5 heterod imers, competition for antibody binding between heterologous RBDs was evident in sera from some but not all mice.Overall, RBD BA.5 competed for RBD Wu -NanoLuc binding (mean IC 50 = 429 ± 496 ng/mL) and RBD Wu competed for RBD BA.5 -NanoLuc binding (mean IC 50 = 576 ± 388 ng/mL) (Fig. 4C and SI Appendix, Fig. S5).We conclude that the RBD Wu -RBD BA.5 heterodimer exhibited greater propensity to elicit cross-reactive RBD binding antibodies than did either of the two homodimers.

Long-Term Heteromultimeric RBD Exposure Does Not Broaden
Serum Neutralizing Antibody Specificity.The neutralization breadth of individual SARS-CoV-2 antibodies in the memory B cell compartment following infection or vaccination increases with time, as a consequence of ongoing somatic mutation, presumably driven by residual persisting antigen (20)(21)(22)(23)(24).In principle, therefore, restimulation of B cells by persistent, heterogenous, antigen exposure might broaden the specificity of serum antibodies through the selective expansion and affinity maturation of B cells that acquire the ability to react with multiple components of a heteromultimeric antigen.
To expose B cells to with heterogeneously juxtaposed, heteromul timeric RBD antigens for prolonged periods, we constructed recom binant adeno-associated virus vectors (rAAVs) expressing RBD-LzS nanoparticles.Immunization of mice with a single intramuscular dose of an rAAV expressing RBD Wu -LzS, but not control vector expressing mNeonGreen, generated similar titers of SARS-CoV-2 Wu neutralizing antibodies (mean ± SD NT 50 = 33,040 ± 22,254) as did two doses (at week 0 and week 3) of the corresponding recom binant RBD Wu -LzS protein with adjuvant (mean NT 50 = 48,497 ± 21,360) (SI Appendix, Fig. S7A).These neutralizing titers were meas ured at 7 to 8 wk after initial immunization, at approximately the time they reached a plateau for both immunization approaches.The rAAV immunized mice were maintained for a total of ~6 mo and neutralizing titers remained approximately constant at this level (SI Appendix, Fig. S7A).
As was the case with recombinant RBD proteins (Fig. 3), neu tralizing antibodies elicited by rAAV vectors expressing RBD 1 -LzS or RBD Wu -LzS at 8 wk after administration, were largely specific for the corresponding virus pseudotype (Fig. 5 A and B).Control rAAV immunizations elicited no pseudotype neutralizing activity (SI Appendix, Fig. S7B).Challenge of human ACE2 expressing mice (K18-hACE2) with SARS-CoV-2 USA-WA/2020, which shares an identical RBD sequence with RBD Wu , 8 wk after immu nization with rAAV RBD Wu -LzS resulted in strong protection, with lung viral RNA levels reduced by >5 orders of magnitude compared to controls (P = 0.0009, Welch's t test) (Fig. 5C).Conversely, immunization with rAAV RBD 1 -LzS gave a nonsig nificant (P = 0.09) reduction in lung SARS-CoV-2 USA-WA/2020 load.Infection of similarly rAAV RBD 1 -LzS or rAAV RBD Wu -LzS immunized mice with SARS-CoV-2 BA.5 gave no discernable reduc tion in viral load (Fig. 5D).Thus, a single injection of a single rAAV RBD-LzS immunogen generated strongly protective neu tralizing antibodies against the homologous virus, but these anti bodies lacked sufficient breadth to confer protection against a heterologous virus at 8 wk after immunization.
To attempt to broaden the neutralizing antibody response elicited by rAAV-RBD 1 -LzS and rAAV-RBD Wu -LzS, we generated a sce nario in which mouse B cells would be exposed, over an extended period, to heteromultimeric antigen in the form of LzS particles bearing a mixture of RBD 1 and RBD Wu .To generate mixed RBD particles, individual cells would need to be transduced with both rAAV RBD 1 -LzS and rAAV RBD Wu -LzS.To test whether this would be the case, we administered the same dose of a single rAAV encod ing either mNeonGreen or mScarlet-I fluorescent proteins, or a mixture of the two to a single hind limb muscle, and then used microscopy to determine whether myotubes expressed either or both fluorescent reporters.In coinjected mice, all of the imaged myotubes expressed both mNeonGreen and mScarlet-I, indicating that they received both constituents of a mixed, 2-component rAAV-injection (SI Appendix, Fig. S8A).Moreover, rAAV-driven expression of both mNeonGreen and mScarlet-I was maintained in myotubes for at least 6 mo (SI Appendix, Fig. S8B).
Some neutralizing activity against SARS-CoV2 BA.5 was acquired in mice immunized with rAAV RBD Wu -LzS alone and in mice immunized with both rAAV RBD 1 -LzS and rAAV RBD Wu -LzS (Fig. 5A and SI Appendix, Fig. S10 B and C).This neutralizing activity was acquired at later time points, after titers against SARS-CoV-2 Wu and SARS-CoV had reached a plateau, suggesting a role for extended B cell evolution in its generation.However, coadministration of rAAV RBD 1 -LzS with rAAV RBD Wu -LzS to generate the mixed nanoparticles did not result in increased titers against SARS-CoV-2 BA.5 (mean NT 50 = 196 ± 256) compared to immunization with rAAV RBD Wu -LzS alone (mean NT 50 = 149 ± 183, P = 0.856) (Fig. 6A and SI Appendix, Fig. S10 B and C).Moreover, challenge of rAAV RBD 1 -LzS + rAAV RBD Wu -LzS immunized K18-hACE2 mice with SARS-CoV-2 BA.5 did not result in reduced viral load compared to mice immunized with rAAV RBD Wu -LzS alone (P = 0.27) or nonimmunized controls (P = 0.88) (Fig. 6B).
Overall, these data suggest that coadministration of rAAV RBD 1 -LzS with rAAV RBD Wu -LzS generated largely separate sets of SARS-CoV and SARS-CoV-2 Wu neutralizing antibodies.A period of 12 to 24 wk of antigen exposure was accompanied with the gen eration of some level of antibodies capable of neutralizing a divergent virus (SARS-CoV-2 BA.5 ) that was not included in the immunogen.However, these antibodies were insufficient for protection against infection and were generated at equivalent levels whether mice were subjected to extended exposure to nanoparticles bearing single het erologous RBD antigen or two diverse heterologous RBD antigens.

Discussion
It is thought that vaccines which display multiple, physically linked but diverse sarbecovirus RBDs would selectively cross-link B cell receptors (BCRs) that bind conserved epitopes or BCRs that tolerate epitope variation.Selective BCR cross-linking in this way might selectively stimulate B cells that can go on to secrete broadly reactive antibodies (28)(29)(30)(31).Indeed, in animals that have been immunized with mosaic RBD nanoparticles, antibodies were directed to more conserved epitopes (30) and individual broadly neutralizing antibodies can be generated (34).However, in a pol yclonal response to physically linked heterologous RBDs, it is unclear what fraction of antibodies cross-react with multiple RBDs and what fraction are specific to a single RBD component.Similarly, multivalent mRNA spike vaccines such as the Wuhanhu-1/BA.5 mixture deployed in 2022, in which mRNAs were mixed prior to administration, would generate a mixture of homo multimeric and heteromultimeric trimers (35,36).In a naive recipient, these trimers could, in principle, selectively stimulate cross-reactive B cells or stimulate separate Wuhan-hu-1-specific and BA.5-specific neutralizing antibody generating cells.We found that multimerized RBD immunogens containing a single ancestral SARS-CoV-2 RBD species generated higher serum neutralizing antibody titers than the equivalent mono meric RBD.The superiority of the multimerized RBD in the generation of high serum neutralizing activity was evident irre spective of whether multimerization was mediated by genetic fusion to a second RBD in a single polypeptide, fusion to IgG1 Fc that spontaneously dimerizes, or fusion to LzS that sponta neously assembles into nanoparticles.This finding is consistent with the idea that artificial multimerization augments stimulation of naive B cells through BCR cross-linking (37).Analysis of the RBD binding specificity of germinal center B cells and selective depletion of RBD binding antibodies in serum generated using genetically fused SARS-CoV-2 Wu RBD and either SARS-CoV or SARS-CoV-2 BA.5 RBD suggests that heterodimeric immuno gens can bias the response, to some degree, toward B cells and serum antibodies that bind both components of a heterodimer and are therefore cross-reactive.Again, this is consistent with the notion that a multimeric immunogen is better able to cross-link cross-reactive BCRs the depletion with one antigen, and heterod imeric antigens can elicit antibodies that cross-react with both components.
However, when antibodies from RBD heterodimer-immunized mice that bound one of the two RBD components were depleted from serum, neutralizing activity against the homologous viral pseu dotype was removed, but neutralization activity against pseudotypes corresponding the other dimer component was unaffected.Thus, the polyclonal antibody pool generated by RBD heterodimers appears to contain two sets of neutralizing antibodies that are almost entirely separate and does not consist of cross-reactive antibodies.Notably, the SARS-CoV and SARS-CoV-2 BA.5 RBD amino acid sequences are 78% and 93% identical to the SARS-CoV-2 Wu RBD sequence respectively, but variation is con centrated in the neutralizing epitopes (12,17).Greater divergence in the neutralizing epitopes of an RBD heteromultimer would reduce the probability that the subset of naive BCRs that can be cross-linked by heteromultimers represent those that recognize neutralizing epitopes and therefore can go on to generate neu tralizing antibodies.the initial B cell response to a heteromul timeric RBD immunogen may be biased toward conserved, but mostly non-neutralizing, epitopes.
Naive BCRs that are capable of reacting with two different neu tralizing epitopes on two components of an RBD heteromultimer may be extremely rare.However, individual SARS-CoV-2 RBDbinding neutralizing antibodies that appear in the memory B cell compartment acquire diversity and greater affinity in the months following infection or vaccination (20)(21)(22)(23)(24).In some cases, this mat uration enables recognition and neutralization of sarbecoviruses that were not neutralized by the earlier antibodies and are divergent from any antigen to which the subject has been exposed (20-22, 24, 25).This property presumably reflects the evolution of BCRs in long-lived germinal centers (38,39) and is evident to a greater degree following SARS-CoV-2 infection, where antigen may be more persistent (20,40) than following vaccination (23).We used rAAVs to prolong exposure of the murine immune system to het eromultimeric RBD antigens in an attempt to drive affinity matu ration that might broaden the reactivity of BCRs and serum antibodies.While there was some evidence that some neutralization breadth was acquired over months in mice with exposure to a per sistent antigen, this effect was not greater in mice immunized with heterodimers compared to homodimers.In alternative "prime boost" strategy, animals might be first immunized with one antigen component to generated diversified BCRs that bind one antigen and then boosted with a second variant to selectively expand cross-reactive antibodies.Whether such an approach is better able to generate cross-neutralizing antibodies than prolonged exposure to heterodimers will require further experimentation.Overall, while heteromultimeric RBD immunogens may bias a murine B cell response toward cross-reactive antibodies to some extent, the results herein suggest that simply combining divergent RBDs results in a B cell response consisting of largely separate sets of single RBD-specific neutralizing serum antibodies.These anti bodies are mostly incapable of neutralizing viruses that diverge from the immunogen components.Careful consideration of the genetic and antigenic distance between the components of mul timeric immunogen-based vaccines may be required to generate antibodies with maximum possible neutralizing breadth, and par ticularly with activity against viruses whose sequences differ sub stantially from the immunogen components.Such considerations are of key importance when designing vaccines in the context of an evolving virus such as SARS-CoV-2 and in preparation for potential future viral pandemics.

RBD Expression
Plasmids.Synthetic DNA sequences encoding the RBD of SARS-CoV-2 (Wuhan-hu-1 variant), termed RBD Wu , human IgG1Fc, and Lumazine Synthase were purchased from GeneArt (ThermoFisher).Sequences encoding the RBD of SARS-CoV (RBD 1 ) and other SARS-CoV-2 variants (RBD BA.1 , RBD BA.5 ) were amplified from previously described full-length spike expression plasmids (6,7,41).The RBD 1 , RBD Wu , RBD BA.1 , RBD BA.5 sequences were PCR amplified to introduce a secretion signal pSecTag2 (METDTLLLWVLLLWVPGSTGD) and a 6xHisTag at the N and C termini of the RBD encoding sequences.Overlap extension PCR was used to fuse RBD Wu to IgG1Fc with an N-terminal pSecTag2 sequence and a C-terminal 6xHisTag or to Lumazine Synthase with an N-terminal pSecTag2-6xHisTag.The RBD, RBD-Fc, and RBD-LzS constructs were inserted into the NcoI and XhoI sites of pCAGGS to generate the mammalian protein expression plasmids.Plasmids expressing RBD-Avi-tag fusions for generating biotinylated RBD proteins were generated in the same way except that sequences encoding an Avi-tag GLNDIFEAQKIEWHE followed by a 6xHisTag were included at the C terminus.Plasmids expressing RBD-RBD fusion protein dimers (RBD Wu -RBD Wu , RBD 1 -RBD 1 , RBD 1 -RBD Wu , RBD BA.5 -RBD BA.5 , RBD Wu -RBD BA.5 , and RBD Wu -RBD BA.1 ) were constructing by amplifying the N-terminal RBD sequence with oligonucleotides encoding an N-terminal pSecTag2 sequence and a C-terminal glycineserine-linker (GGSGG) incorporating a NotI site, while the and the C-terminal RBD sequence was amplified to introduce a 5′ NotI site and a C-terminal 6xHisTag.Plasmids expressing the RBD-RBD fusion dimers were built using three fragment ligation, whereby the two RBD encoding sequences(NcoI-RBD-NotI-RBD-XhoI) were inserted into the NcoI and XhoI sites of pCAGGS.To express RBD-NanoLuc fusion proteins constructs, NanoLuc encoding DNA sequences were amplified from pHIV-1 NL4-3 DEnv-NanoLuc (41) introducing 5′ NotI and 3′ 6xHisTag-XhoI sequences and inserted into pCAGGS, in the same manner as for the RBD-RBD fusion dimers but with NanoLuc in place of the C-terminal RBD.
AAV Constructs.Synthetic DNA sequences encoding pSecTag2, fused to RBD 1 , or RBD Wu and Lumazine synthase-6xHis were amplified to introduce EcoRI and HindIII restriction sites at 5′ and 3′ ends of the amplicon.These DNA fragments were inserted into adeno-associated virus (AAV) expression vector (pAAV-MCS, VPK-410, Cell Biolabs, Inc.) at the multiple cloning site (MCS) flanked by the two inverted terminal repeats (ITRs), generating pAAV-RBD 1 -LzS and pAAV-RBD Wu -LzS expressing plasmids respectively.mNeonGreen and mScarlet-I fluorescent protein coding sequences were PCR amplified to introduce EcoRI and HindIII restriction sites and inserted into the AAV vector in the same way generating pAAV-mNeonGreen and pAAV-mScarlet-I.The AAV packaging plasmid, pAAV-DJ Rep-Cap expressing the recombinant capsid and AAV2 Rep proteins, and helper plasmid (pAAV-Helper) expressing the E2A, 4B, and VA RNA from adenovirus were obtained from Cell Biolabs (Cat# VPK-400-DJ).
Recombinant Protein Expression and Purification.All 6xHis tagged RBD proteins, RBD-RBD fusion dimers, and other RBD-fusion proteins were produced using the Expi293 Expression System (ThermoFisher A14635) according to the manufacturer's protocol.Briefly, 75 to 600 million Expi293 were transfected with pCAGGS-based plasmids expressing RBD, RBD-Fc, RBD-LzS, or RBD dimers, using ExpiFectamine, with the addition of Enhancer 1 and 2 at 16 to 18 h posttransfection (ThermoFisher A14524).Culture supernatants were harvested at 4 to 6 d posttransfection depending on cell viability.The supernatants were passed through 0.22 μm filters (Millipore S2GPU05RE), and 1/10 volume of 500 mM NaH 2 PO 4 , 1.5 M NaCl, and 100 mM imidazole pH 8.0 was added.The supernatants were batch bound to 1 to 2 mL Ni-NTA Agarose beads (Qiagen 30210) for 2 to 16 h at 4 °C, and the beads were collected by gravity flow into a column.Beads were washed with 10 column volumes of 50 mM NaH 2 PO 4 , 300 mM NaCl, and 20 mM imidazole pH 8.0.The proteins were then eluted in 3 column volumes of 50 mM NaH 2 PO 4 , 300 mM NaCl, and 250 mM imidazole pH 8.0.Purified fractions were dialyzed into PBS using Slide-A-Lyzer Cassettes (ThermoFisher 66382), quantified by measuring A 280 with a Nanodrop instrument (ThermoFisher), and analyzed for purity by separation on 4 to 12% Bis-Tris NuPage Gels (ThermoFisher NP0323BOX) followed by visualization with Gel Code Blue Stain (ThermoFisher 24590).Avi-Tagged RBD proteins were biotinylated using the BirA500 biotin-protein ligase standard reaction kit (Avidity EC 6.3.4.15) according to the manufacturer's instructions, as described previously (44).
Mouse Experimentation.Ten-week-old C57BL/6 J male and female mice or K18-hACE2 transgenic mice expressing the human ACE [B6.Cg-Tg (K18-ACE2)2Prlmn/J,034860] were obtained from the Jackson Laboratory (JAX) and housed in a standard BSL1 facility at the Rockefeller Comparative Biosciences Center and regular chow diet, water ad libitum under a 12-h light-dark cycle.The mouse studies were performed in compliance with the animal protocol 21042-H and the Rockefeller University Institutional Animal Care and Use Committee (IACUC).The mice were prebled by the submandibular route and arranged into groups of mixed sex.
Recombinant Protein Immunization.Ten-week-old C57BL/6 (The Jackson Laboratory Strain #000664) mice of both sexes, cared for as described above, were immunized using RBD, RBD-RBD fusion dimers, RBD-Fc, or RBD-LzS fusion proteins mixed with Enhanced Magic Mouse Adjuvant using the manufacturer's protocol (Creative Diagnostics Cat#CDN-A001E).Mice were randomly grouped and antigens were administered by intramuscular injection (a total of 10 µg RBD in 50 µL PBS mixed with 50 µL Enhanced Magic Mouse Adjuvant) with 5 µg injected per hind limb.The mice were injected again with the same antigen/ adjuvant combinations 3 wk later.Blood samples were taken at 0, 3, 5, 7, 11, and 12 wk after the first antigen injection.
To determine the persistence and coexpression of antigens in individual cells in mixed AAV-based immunogen experiments, 10-wk-old C57BL/6 J or K18-hACE2 were injected intramuscularly with a single dose of a single rAAV encoding mNeonGreen or mScarlet-I at 50 μL 2.5 × 10 10 vg/mouse.Alternatively, 1.25 × 10 10 vg of the AAVs encoding mNeonGreen and mScarlet-I were combined in a single 50 μL injection (2.5 × 10 10 vg/mouse total) and injected in the left hind limb of each mouse.These mice were euthanized at each time point and muscle tissue dissected for embedding, thin-sectioning and microscopy to detect mNeonGreen and mScarlet-I expression.
Sectioning and Microscopy.Mice were euthanized with a CO 2 overdose, and the musculature of the left and right hind limbs was dissected.Muscle tissues were placed in dry ice mixed with 2-methylbutane (isopentane, Fisher Scientific, Cat# 60-048-072), transferred into plastic molds (Fisher Scientific, Tissue-Tek Cryomold, Cat# NC9511236), embedded in optimal cutting temperature (OCT) compound (Fisher Scientific, Tissue-Tek OCT, Cat# NC1029572), cryosectioned at 5 μm, and stained with DAPI (Fisher Scientific, Cat# 62247, Lot SF3605141) nuclear stain at 1:1,000 dilution in PBS for 5 min.Images were acquired at a fixed exposure on a BX60 fluorescence microscope (Olympus America Inc.) equipped with a motorized stage (Prior Scientific Instruments Ltd.) and a CC12 camera (Olympus America Inc.).
Analysis of Cross-Reactive B Cell Elicitation.Nine-week-old female C57BL/6 J mice (The Jackson Laboratory, Strain #000664) cared for as described previously were randomly sorted into three groups of 5. Mouse groups were immunized with 3 μg of RBD 1 -RBD 1 , RBD Wu -RBD Wu , or RBD 1 -RBD Wu fusion dimers.Three control mice were immunized with 12.5 μg of an irrelevant NP-OVAL antigen (Biosearch Technologies N-5051-10).Immunizations were carried out in both footpads, using in a 25-μL volume containing 8.3 μL of 2% Alhydrogel (Invivogen VAC-ALU-250) and 1× PBS.On day 14 postimmunization, the draining popliteal lymph nodes from each mouse were pooled in 200 μL of 1× PBS and mechanically disrupted to create a single-cell suspension and cells recovered by centrifugation at 350 g for 5 min.
Neutralization Assays.Mouse serum was tested for neutralizing potency against SARS-CoV, SARS-CoV-2, and its variants using the spike-pseudotyped HIV-1-based assay described (41) previously.Briefly, following heat inactivation at 55 °C for 30 min, mouse serum samples were fourfold or fivefold serial diluted using a Pipetmax (Gilson GFAM0072) and incubated with spike-pseudotyped HIV-1 reporter virus for 1 h at 37 °C.The serum-pseudotype virus mixture was then added to 1 × 10 4 HT1080/ACE2.cl14cells seeded in 96-well black plates (Costar 3916).At 48 h after infection, the cells were washed with PBS and lysed in Luciferase Cell Culture Lysis reagent (Promega E1531).Luciferase activity was measured using the Nano-Glo Luciferase Assay System (Promega N1150) and a Glomax Navigator luminometer (Promega) or the ClariostarPlus Microplate Reader (BMG LabTech).The relative luminescence units (RLUs) were plotted as a decimal fraction of those obtained in wells on the same plate that were infected in the absence of mouse serum.The normalized data were plotted in GraphPad Prism, and the NT 50 was calculated using four-parameter nonlinear regression.
Neutralization Depletion Assays.Aliquots (4 µL) of mouse serum were incubated with 10 µg RBD-6xHis in 30 µL PBS at 4 °C with shaking.After 30 min, 10 µL His-Tag Dynabeads (ThermoFisher 10103D) were added, and the mixture was incubated for an additional 30 min at 4 °C with shaking.After magnetic separation of the beads, the supernatant was transferred to a fresh 96-well plate and serially diluted for assessment of neutralization potency against pseudotyped virus using the assay described above.
Antibody: RBD-NanoLuc Binding and Competition Assays.Fourfold serial dilutions of mouse serum were incubated with 10 ng of an RBD-NanoLuc fusion protein at 4 °C with shaking.After 30 min, 10 µL Protein G Dynabeads (ThermoFisher 10004D) were added, and the samples were incubated for 15 min at 4 °C with shaking.The Dynabeads were magnetically separated, washed three times with 100 µL PBS, and then incubated in 50 µL Luciferase Cell Culture Lysis reagent (Promega E1531).Bead-bound NanoLuc luciferase activity was measured using the Nano-Glo Luciferase Assay System (Promega N1150) and a Glomax Navigator luminometer (Promega).
Serum concentrations yielding RBD-NanoLuc binding capacities of 1 × 10 5 to 1 × 10 6 RLU were used for competition assays.For the competition assays, mouse serum was incubated with a fourfold serially diluted unlabelled RBD (0 to 500 ng) and 10 ng of an RBD-NanoLuc fusion protein for 30 min at 4 °C with shaking; then, 10 µL Protein G Dynabeads were added, and the samples were incubated for an additional 15 min at 4 °C with shaking.The Dynabeads were separated using a magnet, washed and then bead-bound NanoLuc luciferase activity measured as described above.Bead-bound NanoLuc luciferase activities were plotted as decimal fraction of those obtained in the absence of unlabelled RBD competitor.The normalized data were plotted in GraphPad Prism, and the IC 50 was calculated using four-parameter nonlinear regression.
Statistical Analysis.Statistical analysis was performed using Prism 9.0 (GraphPad Software).Mean values were compared between groups by Student's t test and two-tailed Mann-Whitney U tests.
Data, Materials, and Software Availability.Numerical data have been deposited on Figshare (DOI: 10.6084/m9.figshare.24585138)(45).All other study data are included in the article and/or SI Appendix.

Fig. 1 .
Fig. 1.Multimeric sarbecovirus RBD immunogens.(A) Schematic representation of genetically fused RBD dimers, RBD-IgG Fc fusions, and LzS nanoparticles.For the LzS nanoparticles, the N terminus of each LzS monomer RBD fusion site is indicated in red.(B) SARS-CoV-2 Wu pseudotype neutralizing titers in mouse sera following a 2-dose immunization with the indicated proteins (week zero = first dose, week 3 = second dose), n = 5 individual mice for each immunogen.(C) SARS-CoV-2 Wu pseudotype neutralizing titers in mouse sera (pooled week 7/8 bleeds) following a 2-dose immunization with the indicated proteins.(D) SARS-CoV and SARS-CoV-2 Wu pseudotype neutralizing titers in mouse sera at week 11/12 following initiation of a 2-dose immunization with the indicated homomultimeric or heteromultimeric RBD dimers (Left), RBD-IgG fusions (Center) and LzS nanoparticles (Right).In (C and D), horizontal lines indicate group means, and symbols indicate individual mice n = 5 per group.The dotted line indicates the lowest sera dilution tested (1:50).

Fig. 6 .
Fig. 6.Neutralizing antibodies and heterologous challenge 24 wk after rAAV RBD-LzS immunization.(A) Neutralizing titers (NT 50 ) against SARS-CoV, SARS-CoV-2 Wu , and SARS-CoV-2 BA.5 pseudotypes at 24 wk after immunization with a single dose of rAAV RBD 1 -LzS, rAAV RBD Wu -LzS or a mixture of both in a single injection.The dotted line indicates the lowest sera dilution tested (1:50).(B) Lung viral load (SARS-CoV-2 RNA copies per μg total RNA) at day 3 after challenge in mice following immunization with rAAV RBD 1 -LzS, rAAV RBD Wu -LzS or a mixture of both in a single injection, as indicated, and challenge at 24 wk after immunization with SARS-CoV-2 BA.5 Each symbol = 1 mouse, lines = group mean, n = 8 mice.
1 -RBD 1 , RBD Wu -RBD Wu , (F) Percentage of GC B cells that bound one RBD 1 , or RBD Wu that also bound the other RBD following immunization with the homodimers or heterodimer (F).For (E) and (F) each symbol = 1 mouse, lines group mean, n = 5 mice.