Improving antibody language models with native pairing.

Existing antibody language models are limited by their use of unpaired antibody sequence data. A recently published dataset of ∼1.6 × 106 natively paired human antibody sequences offers a unique opportunity to evaluate how antibody language models are improved by training with native pairs. We trained three baseline antibody language models (BALM), using natively paired (BALM-paired), randomly-paired (BALM-shuffled), or unpaired (BALM-unpaired) sequences from this dataset. To address the paucity of paired sequences, we additionally fine-tuned ESM (evolutionary scale modeling)-2 with natively paired antibody sequences (ft-ESM). We provide evidence that training with native pairs allows the model to learn immunologically relevant features that span the light and heavy chains, which cannot be simulated by training with random pairs. We additionally show that training with native pairs improves model performance on a variety of metrics, including the ability of the model to classify antibodies by pathogen specificity.

Deep repertoire mining uncovers ultra-broad coronavirus neutralizing antibodies targeting multiple spike epitopes.

The development of vaccines and therapeutics that are broadly effective against known and emergent coronaviruses is an urgent priority. We screened the circulating B cell repertoires of COVID-19 survivors and vaccinees to isolate over 9,000 severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-specific monoclonal antibodies (mAbs), providing an expansive view of the SARS-CoV-2-specific Ab repertoire. Among the recovered antibodies was TXG-0078, an N-terminal domain (NTD)-specific neutralizing mAb that recognizes diverse alpha- and beta-coronaviruses. TXG-0078 achieves its exceptional binding breadth while utilizing the same VH1-24 variable gene signature and heavy-chain-dominant binding pattern seen in other NTD-supersite-specific neutralizing Abs with much narrower specificity. We also report CC24.2, a pan-sarbecovirus neutralizing antibody that targets a unique receptor-binding domain (RBD) epitope and shows similar neutralization potency against all tested SARS-CoV-2 variants, including BQ.1.1 and XBB.1.5. A cocktail of TXG-0078 and CC24.2 shows protection in vivo, suggesting their potential use in variant-resistant therapeutic Ab cocktails and as templates for pan-coronavirus vaccine design.

Deciphering functional roles and interplay between Beclin1 and Beclin2 in autophagosome formation and mitophagy.

The degradation of macromolecules and organelles by the process of autophagy is critical for cellular homeostasis and is often compromised during aging and disease. Beclin1 and Beclin2 are implicated in autophagy induction, and these homologs share a high degree of amino acid sequence similarity but have divergent N-terminal regions. Here, we investigated the functions of the Beclin homologs in regulating autophagy and mitophagy, a specialized form of autophagy that targets mitochondria. Both Beclin homologs contributed to autophagosome formation, but a mechanism of autophagosome formation independent of either Beclin homolog occurred in response to starvation or mitochondrial damage. Mitophagy was compromised only in Beclin1-deficient HeLa cells and mouse embryonic fibroblasts because of defective autophagosomal engulfment of mitochondria, and the function of Beclin1 in mitophagy required the phosphorylation of the conserved Ser15 residue by the kinase Ulk1. Mitochondria-ER-associated membranes (MAMs) are important sites of autophagosome formation during mitophagy, and Beclin1, but not Beclin2 or a Beclin1 mutant that could not be phosphorylated at Ser15, localized to MAMs during mitophagy. Our findings establish a regulatory role for Beclin1 in selective mitophagy by initiating autophagosome formation adjacent to mitochondria, a function facilitated by Ulk1-mediated phosphorylation of Ser15 in its distinct N-terminal region.