How early-life colonization and subsequent exposure to the microbiota affect long-term tissue immunity remains poorly understood. Here, we show that the development of mucosal-associated invariant T (MAIT) cells relies on a specific temporal window, after which MAIT cell development is permanently impaired. This imprinting depends on early-life exposure to defined microbes that synthesize riboflavin-derived antigens. In adults, cutaneous MAIT cells are a dominant population of interleukin-17A (IL-17A)-producing lymphocytes, which display a distinct transcriptional signature and can subsequently respond to skin commensals in an IL-1-, IL-18-, and antigen-dependent manner. Consequently, local activation of cutaneous MAIT cells promotes wound healing. Together, our work uncovers a privileged interaction between defined members of the microbiota and MAIT cells, which sequentially controls both tissue-imprinting and subsequent responses to injury.

Cas1 integrase is the key enzyme of the clustered regularly interspaced short palindromic repeat (CRISPR)-Cas adaptation module that mediates acquisition of spacers derived from foreign DNA by CRISPR arrays. In diverse bacteria, the cas1 gene is fused (or adjacent) to a gene encoding a reverse transcriptase (RT) related to group II intron RTs. An RT-Cas1 fusion protein has been recently shown to enable acquisition of CRISPR spacers from RNA. Phylogenetic analysis of the CRISPR-associated RTs demonstrates monophyly of the RT-Cas1 fusion, and coevolution of the RT and Cas1 domains. Nearly all such RTs are present within type III CRISPR-Cas loci, but their phylogeny does not parallel the CRISPR-Cas type classification, indicating that RT-Cas1 is an autonomous functional module that is disseminated by horizontal gene transfer and can function with diverse type III systems. To compare the sequence pools sampled by RT-Cas1-associated and RT-lacking CRISPR-Cas systems, we obtained samples of a commercially grown cyanobacterium-Arthrospira platensis Sequencing of the CRISPR arrays uncovered a highly diverse population of spacers. Spacer diversity was particularly striking for the RT-Cas1-containing type III-B system, where no saturation was evident even with millions of sequences analyzed. In contrast, analysis of the RT-lacking type III-D system yielded a highly diverse pool but reached a point where fewer novel spacers were recovered as sequencing depth was increased. Matches could be identified for a small fraction of the non-RT-Cas1-associated spacers, and for only a single RT-Cas1-associated spacer. Thus, the principal source(s) of the spacers, particularly the hypervariable spacer repertoire of the RT-associated arrays, remains unknown.IMPORTANCE While the majority of CRISPR-Cas immune systems adapt to foreign genetic elements by capturing segments of invasive DNA, some systems carry reverse transcriptases (RTs) that enable adaptation to RNA molecules. From analysis of available bacterial sequence data, we find evidence that RT-based RNA adaptation machinery has been able to join with CRISPR-Cas immune systems in many, diverse bacterial species. To investigate whether the abilities to adapt to DNA and RNA molecules are utilized for defense against distinct classes of invaders in nature, we sequenced CRISPR arrays from samples of commercial-scale open-air cultures of Arthrospira platensis, a cyanobacterium that contains both RT-lacking and RT-containing CRISPR-Cas systems. We uncovered a diverse pool of naturally occurring immune memories, with the RT-lacking locus acquiring a number of segments matching known viral or bacterial genes, while the RT-containing locus has acquired spacers from a distinct sequence pool for which the source remains enigmatic.

Although millions of distinct virus species likely exist, only approximately 9000 are catalogued in GenBank's RefSeq database. We selectively enriched for the genomes of circular DNA viruses in over 70 animal samples, ranging from nematodes to human tissue specimens. A bioinformatics pipeline, Cenote-Taker, was developed to automatically annotate over 2500 complete genomes in a GenBank-compliant format. The new genomes belong to dozens of established and emerging viral families. Some appear to be the result of previously undescribed recombination events between ssDNA and ssRNA viruses. In addition, hundreds of circular DNA elements that do not encode any discernable similarities to previously characterized sequences were identified. To characterize these 'dark matter' sequences, we used an artificial neural network to identify candidate viral capsid proteins, several of which formed virus-like particles when expressed in culture. These data further the understanding of viral sequence diversity and allow for high throughput documentation of the virosphere.