21produced the components necessary for the mPing burst, not the loss of epigenetic 36 regulation. 37 38 39 majority of TE bursts have been inferred after the fact -via computational analysis of 45 whole genome sequence -the stealth features they require for success have remained 46 largely undiscovered. 47Revealing these stealth features requires the identification of a TE in the midst of a 48 burst. This was accomplished for the miniature inverted-repeat transposable element 49 (MITE) mPing from rice 1,2 . MITEs are non-autonomous Class II (DNA) elements that are 50 the most common TE associated with the non-coding regions of plant genes 3 . To 51 understand how MITEs attain high copy numbers despite a preference for insertion into 52 genic regions, a computational approach was used to identify mPing, and its source of 53 transposase, encoded by the related autonomous Ping element ( Fig. 1a) 1 . 54Ongoing bursts of mPing were discovered in four temperate japonica strains: EG4, 55 HEG4, A119, and A123, whose genomes were sequenced and insertion sites and 56 epigenetic landscape determined 2,4,5 . These analyses uncovered two features of 57 successful bursts. First, mPing targets genic regions but avoids exon sequences, thus 58 minimizing harm to the host 2,5 . Second, because mPing does not share coding 59 sequences with Ping (Fig. 1a), increases in its copy number and host recognition of its 60 sequences does not silence Ping genes, thus allowing the continuous production of the 61 proteins necessary to sustain the burst for decades 4 . 62The contributions of two other features to the success of the bursts could not be 63 assessed previously and are the focus of this study. These features are a single SNP at 64 position 16 (+16G/A) that distinguishes mPing and Ping sequences (Fig. 1a), and a 65 single Ping locus (called Ping16A_Stow) that is the only Ping locus shared by all 66 bursting strains 4 . To understand the origin of these features and their possible role in 67 109
Copy number variation of mPing and Ping elements in domesticated and wild rice 110None of the 3,000 rice strains analyzed in this study have more mPing elements than 111 the 231-503 copies found in the four temperate japonica strains (HEG4, EG4, A119, 112 A123) in the midst of mPing bursts 4 . Of the 3,000 rice strains, 2,780 (92.7%) contain 113 (Table 1 and Supplementary Fig. 2). These data suggest that it is likely that Ping was 129 selected against or lost from most strains during the hypothesized two or more 130 domestication events from O. rufipogon populations 6,16 . 131 132
Origin of a Ping variant and its possible significance 133Analysis of the extensive collection of rice genomes revealed that a SNP distinguishing 134Ping and mPing (+16G/A) located adjacent to the 15-bp terminal inverted repeat (TIR) 135 ( Fig. 3a) and may be implicated in mPing bursts. Pings having these SNPs are referred 136