Clustered regularly interspaced short palindromic repeat (CRISPR)-Cas systems are bacterial defences that target bacteriophages and mobile genetic elements. How these defences evolve in novel host environments remains largely unknown. We studied the evolution of the CRISPR-Cas system in Mycoplasma gallisepticum (also named Mycoplasmoides gallisepticum ), a bacterial pathogen of poultry that jumped into a passerine host ~30 years ago. Over the decade following the host shift, all isolates displaying a functional CRISPR-Cas system were found not only to harbour completely new sets of spacers, but the DNA protospacer adjacent motif recognized by the main effector M. gallisepticum Cas9 (MgCas9) was also different. These changes in CRISPR-Cas diversity and specificity are consistent with a change in the community of phages and mobile elements infecting M. gallisepticum as it colonized the novel host. In the years following the host shift, we also detected a gradual rise in isolates displaying non-functional MgCas9. After 12 years, all circulating isolates harboured inactive forms only. This loss of CRISPR-Cas function comes at a time when the passerine host is known to have evolved widespread resistance, which in turn drove the evolution of increasing M. gallisepticum virulence through antagonistic coevolution. Such striking concordance in the rise of inactivated forms of CRISPR-Cas and the evolution of host resistance suggests that the inactivation of the CRISPR-Cas system was necessary for enabling adaptive bacterial responses to host-driven selection. We highlight the need to consider both host and pathogen selection pressures on bacteria for understanding the evolution of CRISPR-Cas systems and the key factors driving the emergence of a pathogenic bacterium in a novel host.< Réduire