Background
There is a need to design novel HIV-1 vaccine candidates that could elicit broadly neutralizing antibodies (bnAbs) and that reflect currently circulating viruses. Here we tested a subtype B envelope (Env) consensus designed in silico with shortened hypervariable loops and a native B Env to compare their Ab binding and neutralizing responses in rabbits.
Methods
A subtype B Env consensus sequence was derived from circulating viruses sampled since 2011. Env hypervariable loops were modeled using AlphaFold2. HIV-1 env sequences were designed as membrane-bound stabilized trimers and expressed as mRNA encapsulated in lipid nanoparticles (mRNA/LNP). Rabbits were immunized at weeks 0, 4 and 8 with 20ug of mRNA/LNP encoding either a consensus or a native env (n=8/arm) and blood was collected every other week until week 12. Binding Abs were measured with a multiplex Luminex assay (using fold-change over baseline) and neutralization with a TZM-bl assay using tier 1 (n=5) and 2 (n=12) pseudoviruses.
Results
AI-models were used to compare different designs for the hypervariable loops V1, V2, V4 and V5 of the subtype B consensus Env. Modified loops were created to promote Ab accessibility, including by shortening V1, V2 and V5. A native subtype B sequence was selected because it was the Env most sensitive to bnAb neutralization among 32 persons living with HIV-1 in early infection. mRNA transfection showed that both proteins were well-expressed as transmembrane proteins and recognized by bnAbs targeting sites of vulnerability. The contemporary consensus B Env induced higher Ab binding responses than the native Env vaccine (average fold change against 10 heterologous Env-gp120s=1579.2 vs 917.2, p<0.001). Rabbits immunized with the consensus elicited broader and more potent neutralizing responses than those who received the native Env after the second or third immunization. At week 10, 3-4 rabbits in the consensus B vaccine group compared to 1-2 rabbits in the native Env group neutralized the 4 Tier 1 strains. Neutralization titers tended to be higher in the group that received the consensus B Env (p=0.081 against MN3; p=0.041 against SF162). Similar results were obtained when comparing consensus and native subtype C sequences.
Conclusions
Contemporary HIV-1 Env that are redesigned and optimized with AI tools are well expressed and immunogenic. They elicited superior antibody responses than native Env, warranting further testing as vaccine candidates.
Table or Figure
