Fig 2: Models of regulatory element clusteringWe propose that control of cell identity and housekeeping genes employs a common strategy. To regulate cell identity genes, multiple Es loop to a P, facilitated by bundling factors such as YY1, and concentrate the assembled transcription apparatus (left). In the case of housekeeping genes, multiple Ps are clustered, facilitated by Ronin, and concentrate the assembled transcription apparatus at these genes (right).

Fig 3: Ronin regulates transcription of housekeeping genes and evolved from an ancient transposon to loop DNA(A) Expression level of genes (with more than 3 reads) that are not bound by Ronin (NRB; n = 6,149), bound by Ronin but not looped by Ronin (RBNL; 0–2 Ronin PETs, n = 7,089), or Ronin bound and looped (RBL/anchors; = 3 Ronin PETs, n = 3,104). p = 0 (NRB/RBNL), 0 (NRB/RBL), 0.99328 (RBNL/RBL) by t test.(B) Change in gene expression (log2 fold change) in Ronin KO cells after 4 days of tamoxifen treatment for all genes plotted against expression in Ctrl cells. Genes that displayed significant changes in expression (false discovery rate [FDR]-adjusted p < 0.05) are colored, with upregulated genes plotted in red and downregulated genes plotted in blue.(C) Heatmaps displaying the gene expression changes in Ronin KO cells (left) and Ronin occupancy in a ±5-kb region centered on the transcription start site (TSS) of each gene (right). Each row represents a single gene, and genes are ranked by their adjusted p value for change in expression in Ronin KO cells compared with Ctrl cells.(D) Changes in gene expression between Ronin KO and Ctrl cells, reported as absolute shrunken log2 fold change for genes that are NRB, RBNL (0–2 PETs), and RBL (=3 PETs). p = 2.3e–109 (NRB/RBNL), 1.870 (NRB/RBL), and 1.224 (RBNL/RBL), as determined by t test.(E) Fold change of genes within individual clusters that are related to the hub genes (red circles, named on the y axis) that are downregulated the most within each cluster after Ronin KO. Blue circles represent genes that are controlled by the same P (bilateral) as the hub gene.(F) Venn diagram of differentially expressed protein-coding genes that are bound by Ronin, housekeeping genes, and metabolic genes. Only protein-coding genes were considered for the overlap for consistency. Statistical enrichment for housekeeping genes (p = 1.7e–133) and metabolic genes (p = 7.18) was determined by hypergeometric testing.(G) Same as (C) but only considering genes whose Ps are involved in P-P interactions. Statistical enrichment for housekeeping genes (p = 4.037) and metabolic genes (p = 0.01) was determined by hypergeometric testing.(H) AFM images of linear substrate DNA (top right) that contains two Ronin-bound Ps with several RBMs (black bars) that are separated by a short DNA fragment (top left). In the presence of recombinant Ronin protein, we observed circular monomers held together by Ronin binding to the RBMs at the opposite ends of the DNA (bottom left)(white arrows indicate the 328- and 412-bp overhangs) and Ronin protein bound to multiple DNA molecules in trans (bottom right). FntbP, Fntb P.(I) Representative image (top) and quantification (bottom) of gel shift circularization assays using a linearized 5-kbp substrate harboring two Ps with RBMs in the absence or presence of different recombinant proteins. YY1, which is known to loop DNA but cannot bind the DNA template used in this experiment, served as an additional negative Ctrl. Data are represented as mean ± SD; n = 3, p = 0.00028 by t test. CCM, covalently closed circular monomer; L, linear substrate; ThD, Thap domain (amino acids 1–80 of the Ronin protein); Zfp, Zfp143.Related to Figure S7 and Tables S1 and S3.

Fig 4: Prediction of Ronin-mediated loop formation based on RBM distribution(A) Correlation plots showing the relation between RBMs and peak and loop densities. The correlation coefficients between the RBMs and peak density and the RBMs and loop density are 0.801 and 0.353, respectively.(B) Heatmap of Ronin-associated interactions, simulated based on the distribution of RBMs in a representative region on chromosome 13 (“dephasing” distance [LD] = 10 nodes, simulation temperature [T] = 4,000 K, and time step [dt] = 0.15625 s; see STAR Methods for details) (top) compared with interactions experimentally detected by ChIA-PET in the same region (bottom).(C) Comparison of PET pair length distributions between the simulated (top) and experimental results (bottom). MSE, mean-squared error; PET, paired-end tag.

Fig 5: Ronin and the Ronin binding motif are necessary and sufficient for P clustering(A) RNA FISH with probes targeting the Gsk3a and Rabac1 loci (illustrated in Figure 1F) in control and Ronin knockout ESCs after 4 days of tamoxifen treatment. White boxes represent the magnified areas shown at the bottom. Gsk3a, Rabac1, and an overlay of both signals are shown from left to right.(B) Quantification of RNA FISH analyses as shown in (A). Boxplots show the distance between the closest green and red signals per cell measured after detecting indicated loci that are looped by Ronin and orthogonal controls that have been described previously to interact but are not bound by Ronin (looped in a Ronin-independent fashion). n = 208, 208, 78, 44, 61, 76, 97, 101, 74, 108, 110, 108; p = 6.8 × 1015, 0.00329, 0.04587, 2.8 × 107, 0.20614, 0.58125 by t test from left to right). The distance between the TSSs of the genes is reported under the gene names.(C) 3C-PCR analyses of Ronin-looped Ps in comparison with genomic sites that are known to interact (e.g., Myc and Pvt Ps, Phc1 E and its P) but are not bound by Ronin. Representative gel images (bottom) and quantifications as boxplots (top) are shown. n = 3; p = 0.02211, 0.03567, 0.01146, 0.03557, 0.91768, and 0.52944 by t test from left to right. Note that the direct interaction between Gpbp1l1 and Toe1 was predicted to occur when the Mmachc P interacts with both Ps in the same cell. The distance between the TSSs of the genes is reported under the gene names.(D) Illustration of the strategy to test whether the RBM is necessary and sufficient for P looping using CRISPR-Cas9 deletion of one RBM in interactions that are mediated by a single RBM in one of two interacting loci.(E) Illustration of the Snx14/Zfp949 interactions (top) and the Snx14 alleles (bottom) that were targeted by CRISPR-Cas9 in comparison with the wild-type (WT) allele.(F) Ronin chromatin immunoprecipitation results of the Snx14 locus in wildtype cells and cells after CRISPR-Cas9 targeted deletion of the RBM in the Snx14 P (left). Data are represented as mean ± SD; n = 4, p = 0.00828 by t test) and boxplots showing the distance between the closest green and red signals per cell measured after RNA FISH detecting the Snx14 and Zfp949 loci (right) in wildtype control cells (n = 65) or cells with RBM deletion (delRBM). n = 82;p = 1.5×107 by t test. Related to Figures S4-S5.