Key assumptions when extrapolating data to humans Levels of evidence (in decreasing order of translatability to humans) Human studies—direct clinical and translational data Non-human primates—models with high biological proximity for a number of immune mechanisms Murine models—useful for testing mechanisms but require cautious extrapolation in vitro/ex vivo—allow a mechanism to be isolated but do not reproduce the entire tissue context Principal limitations of interpretation Vγ9Vδ2/BTN3A1–BTN2A1—the pAg-reactive system is characteristic of humans and primates; standard laboratory mice have no direct functional counterpart. Granulysin—participates in the direct anti-mycobacterial activity of cytotoxic cells in humans and primates; absent in ordinary mice. Intravenous BCG—pronounced protection has been shown in macaques, but this route of administration cannot be transferred directly to clinical practice. COPD → γδ T cells → Mtb—direct data are insufficient; the emphysema model with the IL-17A → G-CSF → neutrophil axis is not equivalent to clinical COPD in TB. M1/M2 polarization—a convenient but simplified scheme; single-cell sequencing data reveal a spectrum of coexisting macrophage states.