These are divided into three categories: mirtrons that are conserved among diverse mammals and cloned from two or more species (three genes), mirtrons that are conserved among diverse primates and cloned from two or more species (two genes), and mirtrons that were cloned from independent libraries from a single species or three or more times from any single library (14 genes). The mirtron hairpin structures are designated with bracket notation, and exon-intron structure with “ >” and “+” notation. The cloned species are capitalized and highlighted green. Supplementary figures provide more detailed information on the cloned species (Figure S3), their tissue subtype of origin (Figure S4), and possible orthologs in other primates (Figure S5). In addition, information on 46 additional mirtron candidates is found in Figure S6.

(A) Sequence logos of 5′ and 3′ mirtron products. Data represent 19 primate/mammalian mirtrons (this study) and 18 invertebrate (14 fly and 4 worm) mirtrons (Ruby et al., 2007a). (Top row) Mammalian mirtrons generate G-rich 5′ mirtron products and C-rich 3′ mirtron products. Alignment of the 3′ mirtron products by their first nucleotides shows an equal frequency of U and C residues. (Bottom row) Invertebrate mirtrons do not show such overall G:C bias, and their 3′ products are strongly biased toward 5′ U residues.
(B) Typical hairpin-end structures of mammalian and invertebrate mirtrons. These preferred end structures are also evident from the sequence logos presented in (A).
(C) Comparison of the nucleotide composition of mammalian and invertebrate mirtrons with bulk short introns in humans and flies. We analyzed the GC content of 13,453 human introns and 29,120 D. melanogaster introns, each 50–120 nt in length. We also analyzed their 5′-most and 3′-most 24 nt (intron “ends”) as a proxy for miRNA/miRNA* regions. GC content and minimum free energy (mfe, kcal/mol) of straight hairpin structures for the cloned mammalian and fly mirtrons were also assessed; where only one mirtron product was obtained, the miRNA* region was inferred by assuming a 2 nt 3′ overhang. Values are shown ±SD. For comparison, we show the GC content of all human and worm/fly (invertebrate) canonical miRNAs listed in miR-base Release 10.

(A) A well-conserved mammalian mirtron. (Top) The 13th intron of the ATP-binding cassette F-1 gene harbors the mirtron mir-877. This intron is bounded by consensus splice donor and acceptor sequences, and efficient processing of this intron was evidenced by the existence of over 50 spliced cDNA clones in EST databases. The hairpin structure of this mirtron is indicated with bracket notation. Human small RNAs corresponding precisely to the 5′ and 3′ ends of the intron were identified, as were 5′ small RNAs from macaque, mouse, and rat. Cloning frequencies define the left arm product of mir-877 as its “miRNA” and the right arm product as the “miRNA*.” (Bottom) Evolutionary characteristics of this mirtron. Sequence alignment and conservation track were obtained from http://genome.ucsc.edu. mir-877 is highly conserved among diverse eutherian species but exhibits accelerated divergence within the loop region.
(B) A primate-specific mirtron. (Top) The 21st intron of the putative helicase DHX30 gene harbors the mirtron mir-1226. Notation and layout are as described in (A). In this case, cloning frequencies define its right arm product as the miRNA and its left arm product as the miRNA*. (Bottom) This mirtron is identifiable only in primates; the conservation of its 3′-most terminal sequence in other mammals likely reflects the pressure to maintain splice recognition determinants.

Human and macaque introns were binned into 100 nt intervals. We then binned all small RNA reads derived from intron termini by intron length, excluding introns that also generated nonboundary reads (thus excluding cloned small RNAs arising from unannotated intronic noncoding RNA genes such as tRNAs or snoRNAs). It is evident that a majority of intron-terminal small RNAs in human, macaque, chimp, and mouse derive from 1–200 nt introns, and that most of these derive in turn from hairpin introns that we annotated as mirtrons or mirtron candidates.