Self-organized division of labor and selection for long-term antiviral defense in CRISPR arrays.

Abstract

CRISPR-Cas is an adaptive immunity system that protects prokaryotes from viruses and other genetic parasites. CRISPR-Cas immunity is guided by short DNA sequences, called spacers, that are acquired upon encounters with parasites and stored in the CRISPR array. Although the mechanisms driving spacer acquisition are relatively well understood, the timespan of CRISPR immune memory and, consequently, the very nature of CRISPR-Cas as a short-term or long-term defense mechanism remain controversial. Because CRISPR arrays have limited size, CRISPR-Cas systems face a fundamental trade-off between acquiring new spacers against recent infections and retaining old spacers for long-term defense. To investigate how CRISPR-Cas systems resolve this "fast response-versus-memory" trade-off, we developed a stochastic model that captures CRISPR array dynamics under varying abundances of endemic and episodic viruses. Simulations reveal three distinct evolutionary regimes that result in short-term, long-term, and "dual memory" arrays. These regimes are governed by the ratio of two measurable parameters: the length of the CRISPR array and the time required for transient epidemics to decay below basal endemic viral abundance. Spontaneous division of labor emerges in the dual memory regime, with spacers near the leader and distal ends specialized in short-term and long-term memory, respectively. Within-array division of labor results in a U-shaped longitudinal profile for the probability of finding the targets of CRISPR immunity among the local virome. Such profiles, that have been empirically observed in CRISPR arrays from the human gut microbiome, could inform future research on the environmental persistence of poorly characterized viromes.