IntroductionEliciting high-titer antibody responses to conserved epitopes is critical for vaccines against highly diverse viruses1,2,3,4,5,6,7,8,9. However, these epitopes are often sub-immunodominant10,11,12,13,14. Although many advanced vaccine technologies can enhance overall immune responses15,16,17,18,19,20,21, it remains essential to maximize the on-target, subdominant response. Sequential heterologous immunization is a widely used strategy to enhance cross-reactive antibody responses against HIV-1, influenza, and SARS-Cov-22,6,22,23,24,25,26,27,28,29,30,31,32,33,34. While most off-target epitopes present on the priming immunogen can be avoided by using boost immunogens from heterologous strains, some off-target epitopes persist and may compete with on-target epitopes for B cell engagement. For example, soluble HIV-1 Env trimers expose non-neutralizing epitopes at the trimer base and at sites where N-linked glycans are incompletely occupied10,11,12,13,14,35,36. This raises a practical question: to what extent should recurrent off-target epitopes be eliminated across sequential immunizations?The HIV-1 fusion peptide (FP) serves as an ideal model epitope for investigating this question37,38, as it is a short linear sequence that can be readily distinguished from carrier-derived off-target epitopes when evaluating antibody responses. Importantly, FP is a promising target for HIV-1 vaccine development39,40. We previously demonstrated that immunizing non-human primates (NHPs) with an eight-amino acid FP conjugated to keyhole limpet hemocyanin (KLH), followed by booster immunizations with the HIV-1 Env trimer, elicited FP-directed antibodies capable of neutralizing up to 59% of HIV-1 strains39. This prime-boost strategy has since been reproducibly validated in multiple follow-up studies using varied immunization regimens and animal models41,42,43,44,45,46. Despite some success with conjugated vaccine approaches, it is still important to explore alternative strategies and broader rationales to enhance both the magnitude and quality of the FP-directed response prior to Env trimer boosting.To address both the vaccinology question, to what extent recurrent off-target epitopes should be eliminated to enhance on-target antibody responses, and the specific need for high-titer FP-directed antibody responses in HIV-1 vaccine development, we developed a new set of FP immunogens using three alphavirus-like particle (VLP) carriers, including Chikungunya (CHIKV), eastern equine encephalitis (EEEV), and venezuelan equine encephalitis (VEEV) VLPs. These VLPs have been shown to be safe, well-tolerated, and highly immunogenic in clinical trials47,48. We also developed glycan-engineered VLPs in which off-target epitopes shared across carriers were masked. Across two independent immunization studies, FP-directed antibody responses were enhanced by using heterologous carriers in sequential immunizations and further improved when shared off-target epitopes were masked. These findings demonstrate that maximizing on-target antibody titers requires minimizing recurrent off-target epitopes across sequential immunizations. In parallel, we selected three FP variants and evaluated them in sequential immunizations, which improved the quality of the FP-directed response after three doses. When the two strategies, off-target epitope masking and heterologous FP variants, were combined, we observed a further increase in the magnitude of FP-binding and neutralizing antibodies, along with the induction of FP-directed serum antibodies capable of neutralizing multi-clade wild-type HIV-1 strains in most animals.ResultsHIV-1 fusion peptide immunogen design using alphavirus-like particlesCHIKV, EEEV, and VEEV VLPs are approximately 60 nm in diameter and contain 240 envelope protein E1-E2 heterodimers, which form 80 spikes49,50,51,52,53. The N-terminus of the E2 subunit is solvent-accessible, as indicated by existing VLP structures (CHIKV: PDB 6nk554, EEEV: PDB 6xo455, and VEEV: PDB 3j0c53). We designed new HIV-1 fusion peptide immunogens (CHIKV-FP8.1, EEEV-FP8.1, and VEEV-FP8.1) by genetically inserting an eight-residue FP between the E3 and E2 subunits, just C-terminal to the E3/E2 cleavage site (Fig. 1A). Each VLP is thus expected to present 240 FP epitopes at the N-terminus of E2 (Fig. 1B). Compared to unmodified VLPs, these VLP-FPs exhibited similar elution volumes in size exclusion chromatography (SEC) (Fig. 1C) and retained comparable size and shape in negative-stain electron microscopy (EM) (Fig. 1D). ELISA assays confirmed that VLP-FPs bind to both the VLP carrier-directed mAb SKT0556 and FP-directed mAb DFPH-a.0139, indicating FP epitope exposure (Fig. 1E). Collectively, all three VLP-FPs maintained the expected size, morphology, and antigenicity.Fig. 1: HIV-1 fusion peptide immunogen design using alphavirus-like particles.A Graphical schematic of the polyprotein derived from alphavirus sub-genomic RNA (termed 26S RNA) showing the HIV-1 FP insertion site between E3 and E2. B Ribbon diagram of the mature CHIKV T = 4 icosahedral surface glycoprotein shell (left, PDB: 2XFB) and E1/E2 heterodimer (right, PDB: 3N42) with FP (from PDB: 6NC3) modeled at the N-terminus of E2. C VLP purification by size exclusion column. D VLP analysis by negative-stain electron microscopy. E VLP binding with anti-FP mAb DFPH-a.01, anti-alphavirus mAb SKT05, and anti-HIV CD4bs mAb VRC01 in ELISA.Full size imageGlycan engineering to mask off-target epitopes shared across VLPsTo enhance the overall magnitude of FP-directed antibody response, we planned sequential FP immunizations using heterologous VLP carriers to minimize competing off-target epitopes. However, the E1 and E2 subunits of CHIKV, EEEV, and VEEV share 43-54% amino acid sequence identity (Supplementary Fig. 1). These conserved residues (Fig. 2A, Supplementary Fig. 2A, B) may form off-target epitopes that compete for B cell responses during boost immunizations. To mask these shared off-target epitopes, we performed multiple iterations of glycan engineering (Fig. 2B, Supplementary Fig. 2C–E), adding one glycan at a time. Using structure-based analysis57, we designed 50-70 potential glycan addition sites per VLP as described in the methods. Each glycan design was individually evaluated for VLP yield and VLP carrier-specific polyclonal serum binding. The top candidate was advanced to the next iteration for further glycan addition. After 3–4 iterations, we generated three new VLP-FPs: CHIKV-3g-FP8.1 (3g: three added glycans E1-69, E1-99, and E2-158), EEEV-3g-FP8.1 (3g: three added glycans E1-63, E2-28, and E2-181), and VEEV-4g-FP8.1 (4g: four added glycans E2-117, E2-150, E2-185, and E2-201) (Fig. 2A, Supplementary Figs. 3 and 4). Glycan occupancy at each of the newly introduced glycosylation sites was assessed by liquid chromatography-mass spectrometry (LC-MS) and confirmed to be high (77–99%) (Fig. 2C and Supplementary Fig. 5).Fig. 2: Glycan engineering to mask off-target epitopes shared across VLPs.A Structure of CHIKV Capsid-E1-E2 (PDB 6NK5). Conserved residues across CHIKV, EEEV, and VEEV are highlighted. Preexisting and newly added glycan sites are highlighted. B Flow chart illustrating the strategy for adding new glycans to the surfaces of VLPs. C Glycan occupancy analysis for newly added glycans. D mAb and animal serum binding to VLPs in ELISA. A representative animal serum is shown. WEVEEV indicates a mixture of WEEV, EEEV, and VEEV. E Fold change in VLP binding titers after glycan addition. VLP binding titers are defined as the serum dilution (ED1.5) or mAb concentration (EC1.5) that reaches an OD450 of 1.5. Greater than one-fold change indicates reduced binding. Each circle indicates an animal serum or mAb. Red bar and error indicate geometric mean and SD. See also Supplementary Figs. 1–5.Full size imageFollowing glycan addition, CHIKV-3g-FP8.1 VLP exhibited a geometric mean reduction of 8-fold in binding to sera from EEEV/VEEV/WEEV-immunized NHPs56 (Fig. 2D, E), indicating that about 88% of cross-reactive binding activity was blocked. Similarly, EEEV-3g-FP8.1 and VEEV-4g-FP8.1 showed geometric mean reduction of 3.5 and 4.6 fold, respectively, in binding to sera from CHIKV-immunized guinea pigs (Fig. 2D, E), blocking about 71% and 78% of binding activity. Importantly, glycan engineering did not affect FP-directed mAb binding (Fig. 2D), and the modified VLPs maintained expected profiles in SEC (Supplementary Fig. 3) and negative-stain EM (Supplementary Fig. 4). These results demonstrate that glycan engineering effectively masked most off-target epitopes shared across three VLP carriers without compromising particle quality or FP epitope exposure.Improved FP-binding antibody titers in guinea pigs after minimizing recurrent off-target epitopes across sequential immunizationsTo assess the impact of sequential heterologous carrier immunization with or without glycan engineering on FP-directed antibody responses, we conducted a three-group guinea pig study (Fig. 3A). Group 1 received three immunizations with the same immunogen, CHIKV-FP8.1. Group 2 received three immunizations, each with a different VLP carrier. Group 3 received sequential heterologous VLP carriers with glycan engineering. Overall, Group 3 exhibited the highest FP-binding serum antibody titers after two or three immunizations (Fig. 3B). After two immunizations (week 6), Group 3’s geometric mean titer (70,749) was 4.3-fold higher than Group 1’s titer (16,375, P = 0.0087) and 2.2-fold higher than Group 2’s titer (32,559). Although Group 2 titers were higher than Group 1, the difference was not statistically significant. After three immunizations (week 10), Group 3’s titer (59,595) was 2-fold higher than Groups 1’s titer (29,442) and 2.2-fold higher than Group 2’s titer (27,056). These results demonstrate that while sequential heterologous carrier immunization enhances FP-directed antibody responses, minimizing recurrent off-target epitopes across immunizations is critical to maximizing this enhancement.Fig. 3: Improved FP-binding antibody titers in guinea pigs after minimizing off-target epitopes across sequential immunizations.A Immunization scheme for guinea pig groups 1, 2, and 3. (B-G) Serum binding to free FP (B), CHIKV VLP (C), EEEV VLP (D), VEEV VLP (E), VLP carriers that match the most recent immunization (F), and HIV-1 Env trimer (G). FP9.1 (AVGIGAVFL) contains the same N-terminal eight amino acids as FP8.1, with one additional C-terminal residue corresponding to position 520 of HIV-1 Env sequence. ED1.5 indicates the serum dilution that reaches an OD450 of 1.5. Median values are shown as red bar. Kruskal–Wallis test and Dunn’s multiple-comparisons test were performed. P values from significant pairs are shown. H Serum neutralization against HIV-1 BG505.N611Q.Full size imageTo evaluate the impact of glycan engineering on off-target VLP-specific antibody responses, longitudinal serum samples were analyzed by ELISA for binding to unmodified CHIKV, EEEV, and VEEV VLPs (i.e., without FP insertion). Group 3 exhibited the lowest CHIKV binding after the second and third immunizations (Fig. 3C). Additionally, Group 3 showed little to no binding to EEEV and VEEV, whereas Groups 1 and 2 displayed detectable binding after the first immunization. This suggests that glycan engineering on CHIKV-3g-FP8.1 effectively prevented the induction of antibodies against conserved VLP carrier-specific epitopes (Fig. 3D, E). After the second immunization, Group 3’s geometric mean titer for EEEV binding (8,993) was 5.4-fold lower than Group 2’s titer (48,251, P = 0.0132). Similarly, Group 3’s geometric mean titer for VEEV binding (1,073) was 47.1-fold lower than Group 2’s titer (50,487, P = 0.0049). These results indicate that glycan engineering on CHIKV-3g-FP8.1 and EEEV-3g-FP8.1 effectively reduced antibody responses to conserved VLP carrier-specific epitopes (Fig. 3D, E). Notably, no significant difference in EEEV and VEEV binding was observed between Groups 2 and 3 after the third immunization. This is likely because glycan engineering on VEEV-4g-FP8.1 was originally designed to mask off-target epitopes shared with CHIKV (Fig. 2) but not specifically with EEEV. Considering that glycan engineering may introduce new epitopes not detected in ELISA using unmodified VLP carriers, we also compared antibody titers against the VLP carriers that were used in the most recent immunization (Fig. 3F). For example, week 6 sera from Group 3 were tested on EEEV-3g (EEEV VLP with 3 added glycans), and week 10 sera were tested on VEEV-4g (VEEV VLP with 4 added glycans). After the second immunization, Group 3’s geometric mean titer (6,363) was 36.8-fold lower than Group 1’s titer (234,474, P