Fig 2: Identification of LIN28A mRNA targets in mouse ESCs at single-cell resolution.A, schematic depicting the identification of LIN28A targets with scTRIBE in different pluripotency transition processes. NR, non-reprogramming. B, box plot showing the coefficient of variation (CV) percentages of read/UMI counts for edited transcripts between cells. Transcripts were categorized into three levels based on their average edited UMI counts: low (below the first quartile), median (between the first and third quartiles), and high (above the third quartile). p values were generated using a two-sided Wilcoxon test. C, dot plot showing the editing frequency in LIN28A-ADARcd and ADARcd control from scTRIBE results. Editing frequency was obtained by dividing the edited UMIs by the total UMIs per edited site. Only significantly edited sites in LIN28A-ADARcd are plotted. p values were generated using a two-sided Wilcoxon test. D and E, stacked bar plot showing the proportion of transcript types (D) and genomic region types (E) of the edited sites/CLIP peaks as identified by bulk TRIBE, scTRIBE, and CLIP-seq (11). snoRNA, small nucleolar RNA; rRNA, ribosomal RNA; lncRNA, long non-coding RNA; UTR, untranslated region; CDS, coding sequence. F, histogram and fitted curve plot showing the cumulative editing frequency distribution of LIN28A-ADARcd and ADARcd within a ± 500 bp window flanking the CLIP peaks (11). G, integrative genome viewer browser tracks showing the edited sites at the Rps15 loci in bulk, pseudobulk, and single-cell level, all located within the LIN28A CLIP peaks (11). H, illustration of the top 1 enriched motif searched from the ± 50 bp region surrounding significantly edited sites of bulk and scTRIBE, or CLIP peaks (11). I, histogram showing the enrichment of the GGAGA-like motif using the ± 50 bp region surrounding significantly edited sites from randomly sampled cells. Data are the mean ± standard deviation (s.d.) of n = 10 computational trials. Adjusted p values (adj p) were generated using a binomial test and Benjamini-Hochberg correction. J, Venn diagram showing the overlap between targets identified by CLIP-seq (11) and scTRIBE (top left). Pie charts showing the proportion of transcripts that do not overlap (top right) or overlap (bottom) with CLIP-seq data and contain a GGAGA-like motifs within a ± 50 bp region surrounding the edited sites.

Fig 3: Validation of IGF2BP3 as an indirect m6A reader and LIN28A as an anti-reader.a-b, RNA pull-down assays and western blots for (a) IGF2BP3 and (b) LIN28A, using RNA probes that contain unmodified A, m6A, and U, respectively, derived from the indicated positions in the transcripts. m6A sites are marked with a red “m”. Histograms show mean of RNA pull-down from three independent replicates. The error bars represent standard error of mean (s.e.m.).Uncropped blots are shown in Supplementary Data Set 1. c. Density plot of LIN28A binding strength (log ratio) at m6A sites in Mettl3 knockout (KO) versus wild-type mES cells. P-value is calculated by two-sided t-test. The number of transcripts is 145. d-e, Signal tracks of Nanog and Sox2 showing LIN28A binding at specific loci in Mettl3 KO and wildtype mES cells.

Fig 4: Alignment of the amino acid sequences of the predicted Duolang sheep lin-28 homolog B protein with those of human (AAZ38897.1), Japanese quail (XP_015714194), house mouse (AAZ38894.1), and chicken (AAZ38896.1).

Fig 5: SUMOylation of LIN28A augments its affinity with pre‐let‐7 in vitro. (A) Top: Schematic maps of r.LIN28A‐∆14 and r.SUMO1‐LIN28A‐∆14 protein domains. Bottom: Coomassie Blue staining of purified r.LIN28A‐∆14 and r.SUMO1‐LIN28A‐∆14 proteins. (B, C) Binding of r.LIN28A‐∆14 and r.SUMO1‐LIN28A‐∆14 to synthetic preE‐let‐7a‐1 was assessed by EMSA with 5 nm of 5′‐end biotin‐labeled preE‐let‐7a‐1 (B) or preE‐let‐7g (C) and the indicated concentration of recombinant proteins. The band intensities were quantified by imagej software and presented as the fraction of bound preE‐let‐7g RNA in the plots. (D) SPR analysis of the direct binding of r.LIN28A‐∆14 and r.SUMO1‐LIN28A‐∆14 to synthetic preE‐let‐7g using an Biacore T200 instrument. The binding affinity was determined by global fitting to a Langmuir 1 : 1 binding model within the biacore evaluation software. (E) The crystal structures of SUMO1 (PDB: 4WJQ) and LIN28A‐preE‐Let‐7g (PDB: 3TS2) were blindly docked at ClusPro 2.0 docking server with a distance restrain of 20 Angstroms between Ca atoms of Gly96 of Sumo1 and Gln36 of LIN28A. The top solution from the server was selected for presentations here (left). The N‐terminal segment of LIN28A (residues 12–35), which was missing in the crystal structure, is modeled here as a helical structure with the sidechain of K15 (shown in sticks) covalently linked to the carboxyl group of Gly96 of SUMO1. The covalently linked SUMO1 (purple) on LIN28A (cyan) could readily interact with the backbone of preE‐let‐7g (brown) in the complex where the positively charged surface of SUMO1 (right) can form strong electrostatic interactions with the negatively charged phosphate groups of preE‐let‐7g. Therefore, the SUMOylated LIN28A would have higher binding affinity toward preE‐let‐7g. The electrostatic surface (right) of Sumo1 (with positively charged area shown in blue and negatively charged area shown in red) was calculated with software APBS, and the cartoon of the structures was generated using software pymol.