A decades-old TB vaccine read-out doesn't match human lungs, study finds

For decades, tuberculosis vaccine research has leaned on a single chemical read-out to decide whether a candidate is working. A 2026 study says the read-out still fires in mice. In human lungs, it barely flickers, and it shows up in the wrong place.

The signal in question is inducible nitric oxide synthase, usually shortened to iNOS. When the immune messenger interferon-gamma (IFNγ) arrives, scavenger-type white blood cells called macrophages are supposed to switch on iNOS and flood the area with nitric oxide, a small, reactive molecule that restricts bacterial growth. That step has anchored the field's idea of what a "protective" TB response looks like. The trouble, according to a new paper in the American Journal of Respiratory Cell and Molecular Biology, is that the step largely does not happen in people.

The researchers ran a multi-pronged comparison across seven mammalian species, including humans, non-human primates, and cattle. In mouse macrophages, IFNγ pretreatment still drove iNOS expression and restricted the growth of Mycobacterium tuberculosis (Mtb) in culture. In human and non-human-primate macrophages, IFNγ did not turn on the gene (called NOS2), did not raise iNOS protein, and did not slow the bacteria.

That gap has been reported before in narrow settings. The new work widens it in two ways. First, the team reanalyzed published single-cell RNA-sequencing data from human and NHP tuberculous granulomas, the organized clusters of immune cells that form around the bacteria in infected lungs. NOS2 transcripts were nearly absent in the myeloid cells at the granuloma core. Second, the researchers used multiplexed ion beam imaging (MIBI), a tissue-staining technique that maps dozens of proteins at single-cell resolution, on human lung biopsies. iNOS protein showed up in the respiratory epithelial cells, the lining cells of the airways, sitting adjacent to the lesions, rather than at the core where the macrophages work.

Monocytes from cattle also failed to mount a robust iNOS response to IFNγ, leaving mice as the outlier. The authors are careful with the language. They do not write off mouse models, and they do not say mouse TB research is wrong. They argue instead that one specific correlate of protection, the IFNγ-to-iNOS readout that vaccine developers have used to rank candidates for decades, is built on a step that does not reliably transfer to humans.

That matters because TB still kills about a million people a year, and the vaccine pipeline has been thin since BCG, the century-old childhood vaccine that does little for adults. IFNγ responses are a standard checkpoint in TB vaccine development, and IFNγ-treated macrophages are a common lab proxy for whether a candidate can restrict Mtb growth. If that lab proxy is a poor stand-in for what human lungs actually do, the field has been over-weighting one signal and under-weighting others.

The paper's constructive move is to point at where the better-aimed next step sits: human lung tissue, primate models, and non-IFNγ readouts. The authors note the same uncoupling may apply to other pneumonias, though they do not test that and the implication is speculative. The question is whether vaccine developers can fold lung-tissue endpoints into clinical readouts fast enough to matter for the next candidates in the pipeline.