A census of nearly 800 million CRISPR spacers finds rapid turnover and warns that a viral match does not always identify a true host.
Microbes store short snippets of invading DNA in CRISPR arrays, molecular records that can guide defense against later viral attacks. Simon Roux and colleagues developed a method for extracting these 'spacers' from mixed-community DNA sequencing reads. They screened 467,018 publicly available metagenomes and recovered about 794 million spacers from 367,609 samples. The dataset contained roughly 410 million distinct sequences. Most were rare, and, among well-sampled repeat groups, most spacer sequences appeared in only one sample. In contrast, spacers that were common within a microbial population were more likely to persist across samples collected from the same location or host. The patterns suggest frequent acquisition, turnover, and selective retention.
Comparing spacers with viruses and plasmids yielded more than a billion sequence matches. These matches can reveal encounters with foreign DNA, but they are not straightforward host assignments. Some viruses matched immune records from distantly related microbes, especially viruses carrying diversity-generating retroelements that can modify proteins used to contact cells. The authors propose that viral DNA sometimes enters a non-host cell without completing infection, yet still contributes a spacer. That explanation remains inferential rather than experimentally established. Most well-supported virus matches implicated one microbial class. The dataset is also unevenly sampled, with many human gut metagenomes, limiting global generalizations.