Antibody landscapes reveal tuberculosis beyond single antigens

Detecting tuberculosis (TB) in the blood usually depends on finding antibodies that recognize pieces of the TB bacterium. But researchers asked a different question: does chronic TB change the overall shape of a person’s antibody repertoire, not just the presence of antibodies to specific antigens? Led by corresponding author Alexander Nesterov‐Mueller, the team used high-density peptide arrays to probe this idea directly. They measured serum antibody binding to 6,936 peptides in 105 individuals from three countries, using two complementary peptide libraries: Mycobacterium tuberculosis peptides (TBC) and a resemblance-ranking library representing the human self-proteome (RRL). Rather than looking only at which specific peptides drew the strongest responses, the group summarized the full distributional patterns of binding into a five-dimensional immune state vector and then placed each person into a mapped immune phase space. This approach let them treat each serum sample as a point in a multi-dimensional landscape, opening the door to detecting changes in the global antibody topology that might accompany chronic TB infection.

To translate these patterns into a diagnostic signal the researchers built what they call a remodeling classifier and tested how well it could distinguish TB from non-TB samples. Strikingly, the classifier performed similarly whether it used pathogen-derived peptides (TBC) or host-derived peptides (RRL), with AUC values in the range 0.63–0.73. That parity indicates the signal comes from broad repertoire restructuring rather than from antigen-specific recognition alone. HIV co-infection partially masks this remodeling signal; when the analysis was restricted to HIV-negative individuals the AUC rose to 0.73 (permutation p = 0.005). With that restriction the method could even detect smear-negative TB cases with AUC = 0.83 and a reported specificity of 0.95 using just three peptides. Phase-space projections based on the five-dimensional immune state vector showed TB severity lying along a continuous remodeling gradient, with smear-negative patients falling between healthy controls and smear-positive cases.

These findings shift how we might think about serological diagnostics for chronic infections like TB. Instead of relying only on the presence or absence of antibody responses to a few pathogen antigens, high-density peptide arrays can act as sensors of the overall antibody repertoire topology, revealing a remodeled immune state associated with disease. That remodeling signal appears in both pathogen-derived and host-derived peptide libraries, which suggests diagnostic approaches could exploit global antibody patterns rather than searching for single diagnostic antigens. The observed impact of HIV co-infection highlights that other chronic conditions may obscure or alter this repertoire remodeling, so clinical use would need to account for comorbidities. Finally, the ability to detect smear-negative TB — a clinically challenging group — suggests this approach could fill gaps where traditional antigen-specific serology or smear tests fall short, although further work will be needed to translate these array-based signatures into routine diagnostics.