Scheme of urea ChIP-seq and difference from standard ChIP-seq

A) An illustration for a BUR anchored to the high-salt extraction resistant SATB1 nuclear architecture (39). DNA-FISH signals (red dots) for a cloned SATB1-bound DNA in both alleles are shown in Satb1+/+(WT) and Satb1−/−(KO) thymocytes on slides treated with 2M NaCl solution to produce DNA halos around the nuclei (blue loops). SATB1 and the SATB1-bound DNA remain in the residual nucleus in WT thymocytes after salt treatment. However, in KO thymocytes the DNA regions normally bound by SATB1 are located in the DNA halos after the treatment. B) A comparative overview of the urea ChIP-seq and standard ChIP-seq protocols. In urea ChIP-seq, entire crosslinked chromatin purified by 8M urea ultracentrifugation is solubilized (see Materials and Methods) before chromatin immunoprecipitation. By contrast, in standard ChIP-seq, crosslinked chromatin in whole nuclei is sonicated, centrifuged and the soluble fraction is isolated for chromatin immunoprecipitation. These critical steps that are different between urea and standard ChIP-seq are highlighted in boxed texts. C) Differences in the profiles of the peaks generated from urea ChIP-seq versus standard ChIP-seq. A diagram is shown for ChIP-seq peaks using antibodies (α) against a DNA direct-binding protein (pink), a DNA indirect-binding protein (blue), and DNA direct-binding protein that resides in the inaccessible chromatin region (green) Note: This diagram illustrate a hypothetical scenario where some of the proteins (green) in the inaccessible chromatin region may also reside in the accessible chromatin binding indirectly to chromatin.

Identification of BURs as direct targets of SATB1 in vivo by urea ChIP-seq.

A) Contrasting distributions of SATB1-binding profiles determined by urea ChIP-seq (Urea) compared to standard ChIP-seq (Std) in mouse thymocytes. B) A zoomed-in view of the gray-highlighted region shown in (panel A). The track 3 (panels A and B) represents all potential BUR sites serving as a reference for BUR sites. For both panels A and B, two independent urea and standard ChIP-seq experiments show high reproducibility (tracks 1 and 2 and tracks 4 and 5). Tracks 6 and 7 [ENCODE, Bring Ren’s Laboratory at the Ludwig Institute for Cancer Research (LICR)] and tracks 8 and 9 (ENCODE, the University of Washington (UW) group]. C) The percentages of urea (Urea) and standard (Std) SATB1 ChIP-seq peaks that intersect with the BUR reference map (see supplementary Table 1). The significance (p values) of the overlap in each of the four cases (urea and standard ChIP-seq samples with two replicates each) with BURs is estimated as the fraction of the overlap counts greater than or equal to the observed count over 1000 random bootstrapped genomic features. p<0.001: none of 1000 random fragment showed greater overlap with BURs than observed in ChIP-seq samples, p=1 means all 1000 random fragments showed greater overlap with BURs than observed in ChIP-seq, indicating avoidance of BURs (see Methods). D) Normalized genome-wide average intensity distribution profiles (normalized relative score) of SATB1 binding sites determined by standard ChIP-seq, urea ChIP-seq as well as BURs, over a +/− 3kb window centered at BUR peaks (top). Relative distance, projection, and Jaccard tests all indicate strong correlation (p< 0.05) between urea ChIP-seq peaks for SATB1 and BURs, but no overlap (p< 0.05) between standard ChIP-seq peaks for SATB1 and BUR peaks.

Identical CTCF-binding profiles are produced by urea and standard ChIP-seq, independent of SATB1.

A) Urea ChIP-seq (Urea) profiles for CTCF in Satb1+/+ (WT) thymocytes and Satb1−/− (KO) thymocytes (track 2 and 3), and for SATB1 in WT thymocytes (track 1). These binding patterns were compared against CTCF-binding profiles generated by standard ChIP-seq (Std) from ENCODE (tracks 4-6) for testis and thymocytes (ENCODE, LICR), along with the DNase1 HS profile (ENCODE, UW) as well as the SATB1-binding profile from standard ChIP-seq (track 7). B) Genome-wide average intensity distribution (relative score) of peaks from CTCF and SATB1 urea ChIP-seq for Satb1+/+ thymocytes (WT), CTCF urea ChIP-seq for Satb1−/− thymocytes (KO), and SATB1 standard ChIP-seq (WT) over a +/− 3kb window centered at CTCF peaks (WT). Relative distance, projection, and Jaccard tests all indicate strong correlation (p< 0.05) between CTCF urea ChIP-seq peaks for WT and KO thymocytes, between these peaks and standard ChIP-seq peaks for SATB1, but no overlap (p< 0.05) between SATB1 urea ChIP-seq peaks. C) Urea-ChIP-Western results for CTCF, RAD21, and SATB1, showing that SATB1 does not associate with CTCF nor RAD21 on chromatin. Also, CTCF and RAD21 co-binding to chromatin appears to be infrequent, suggested by the weak ChIP-Western signals. D) Hi-C interaction heatmaps for Satb1+/+:Cd4-cre (WT) (top right triangle) and Satb1 F/F:Cd4-cre (KO) thymocytes (bottom left triangle) show essentially identical patterns. E) Genome-wide comparison of distributions of cis-Eigenvector 1 values shows no difference between WT and KO thymocytes (high correlation at r=0.998), indicating that genomic compartmentalization is unaffected in Satb1 KO thymocytes. Experiments were done in two replicates each for WT and KO samples. The data shown in D and E were generated by combining data from the replicates for each condition.

SATB1-bound sites identified by standard ChIP-seq at regulatory regions are SATB1 dependent.

SATB1-binding profiles in immature thymocytes determined by urea ChIP-seq (Urea) and standard ChIP-seq (Std) covering three loci at Rag1/Rag2 (top) Foxp3 (middle), and Runx3 (bottom) are shown. For each of the three loci, the standard SATB1 ChIP-seq generates peaks in Satb1+/+ (WT) thymocytes (track 3), but not in Satb1−/− (KO) thymocytes (tracks 9 for Rag 1/2 and Runx3, and tracks 11 for Foxp3). Input chromatin for standard ChIP-seq in these regions did not produce any major peaks (track 10 in Rag1/Rag2 and Runx3, and track 12 in Foxp3). The SATB1 peaks from standard ChIP-seq in Rag1/Rag2 and Runx3 loci largely overlap with enhancers (marked by H3K4me1 and H3K27ac). At the Foxp3 locus in immature thymocytes SATB1 peaks (track 3) appear in a region marked by low levels of H3K27ac and H3Kme1 (tracks 4 and 5) where super-enhancer is established upon differentiation into Treg cells (tracks 6 and 7)(47). SATB1-bound peaks produced by urea ChIP-seq (track 2) are consistently located in distinct positions from those peaks obtained by standard ChIP-seq in these gene loci. Urea ChIP-seq did not generate peaks in KO thymocytes with anti-SATB1 antibody. Input chromatin also did not produce any peaks. H3K27ac and H3Kme1 data (tracks 4 and 5) are from Encode, generated by Bing Ren’s laboratory at LICR, and data for Treg shown in (tracks 6 and 7) are from Kitagawa(47).

SATB1 mediates dense chromatin interactions over Rag1/2-containing gene-rich region.

A) Chromatin interactions detected by urea 4C-seq using a SATB1-bound BUR (BUR-1) and another site (BUR-2) as baits. These baits are located near the border of a gene-poor region and a gene-rich region containing SATB1-regulated Rag1 and Rag2 genes (pink vertical bar with arrowhead). BUR-2 was found to be one of the interacting sites of BUR-1. Therefore, it was used as a reciprocal control for BUR-1 to check the reproducibility of BUR-1 interactions. We used Satb1+/+ (WT) and Satb1−/− (KO) thymocytes from 2 weeks old mice for urea 4C-seq. BUR-1 and BUR-2 interacts with many sites within the ∼5.8Mb gene-rich region, containing Rag1 and Rag2 (tracks WT-BUR1 and WT-BUR2). These interactions were greatly reduced upon SATB1 deletion (track KO-BUR1). B) A zoom-out view, focusing on the Rag1 and Rag2 region, shows that BUR-1 interacts extensively over this region in WT thymocytes. Such interactions were virtually diminished in the absence of SATB1. The pink vertical bars with arrowheads indicate the positions of the BUR-1 and BUR-2 bait primers. Each of these primer sequences is located at the end of the HindIII fragment containing the primer. These HindIII fragments are indicated by pink bars adjacent to pink arrowheads pointing primer positions. The map of HindIII fragments is shown for interacting regions (track HindIII sites). BURs bound by SATB1 (confirmed by SATB1 urea ChIP-seq peaks and BUR reference) found within those HindIII fragments that interact with BUR-1 or BUR-2 are marked by small red stars and these HindIII fragments are light gray shaded. Note: Reads to the HindIII fragment immediately adjacent to each of the bait primers (BUR-1 and BURs) were removed as they comprised more than 2% of the total reads, typically derived from re-ligation of digested chromatin fragments to their original neighboring sequences and/or from a small undigested chromatin fraction.

SATB1 binds to BURs in a cell-type dependent manner.

A) SATB1-binding profiles in a ∼2.4 Mb region in chromosome 6 containing Cd4 in a gene-rich region from standard ChIP-seq (Std) (track 2) and urea ChIP-seq (Urea) (tracks 3-8). Urea ChIP-seq-derived SATB1 profiles in thymocytes (track 3), brain (track 4) and skin (track 5) show difference in peak numbers depending on cell type (see supplementary Table 1 and 2). This contrasts with mostly identical CTCF patterns in these cell types (tracks 11-13). B) SATB1-binding profiles in a 406kb region in chromosome 5 containing Gabra2 and Gabrg1, highly expressed in brain, also shows much increased BUR binding in brain, compared to thymocytes and skin similar to A). In this region, there is no SATB1-binding site detected by standard ChIP-seq (track 2), and CTCF binding sites are mostly absent (track 9). C) Percentages of SATB1 urea-ChIP-seq peaks in skin, thymocytes and brain intersecting with the BUR reference map, indicating that the majority of the peaks are BURs (supplementary Table 1). D) The percentage of BURs among all BURs identified in the BUR reference map that intersected with urea SATB1 ChIP-seq peaks identified in skin, thymocytes and brain. DNase1 HS data are from ENCODE (UW), the CTCF-binding profile and H3K27ac data from standard ChIP-seq are from ENCODE (LICR).

BURs largely co-map with LADs

A) The BUR distribution (BUR reference map, track 2) was compared against Lamin B1-DAM-ID sites(81) (track 1) and SATB1-binding sites derived from urea-ChIP-seq (Urea) (track 3 and 4) over a 8.97 Mb region in chromosome 5 containing neuronal genes, Gabrg1, Gabra2, Gabra4, and Gabrb1. In this region, co-mapping of BURs and Lamin B1-DAM-ID regions is confirmed. Among BURs, SATB1-bound BURs derived from urea ChIP-seq (Urea) show difference in the number of peaks depending on cell type (see supplementary Table 1 and 2). SATB1-binding sites derived from urea ChIP-seq avoid CTCF-binding sites, H3Kme1, H3K27ac sites (from Encode LICR). B) Average Lamin B1 DamID signal (black dotted), standard ChIP signal for SATB1 (black) and urea ChIP for SATB1 (blue) signal over LADs and flanking regions (interLAD sized at10% of adjacent LAD). C) Average Lamin B1 DamID signal (dotted black), BUR signal (red), and urea ChIP for CTCF (orange) signal over LADs (gray shaded) and flanking regions (interLAD sized at 10% of adjacent LAD). D) Average BUR signal (red), urea ChIP for SATB1 (blue) signal, Lamin B1 DamID (black, dotted), H3K4me1 (light green) and H3K27ac (light orange) signals over a +/− 3kb window centered at H3K4me1 peaks within LADs (LADs are gray shaded and DiPC that reside inside large LADs are unshaded, see text). E) Average BUR signal (red), urea ChIP for SATB1 (blue) signal, H3K4me1 (light green) and H3K27ac (light orange) signals over a +/− 3kb window centered at H3K4me1 peaks outside LADs. Relative distance, projection, and Jaccard tests all indicate strong correlation (p< 0.05) between urea ChIP-seq peaks for SATB1 and LADs, between BURs and LADs, but no overlap (p<0.05) between urea ChIP-seq SATB1 and H3K27ac and H3K4me3 peaks. The results in B to E represent normalized average intensities of signals.

A model for SATB1-mediated chromatin organization with direct and indirect binding of SATB1 with chromatin.

The model shown is based on the results from urea ChIP-seq showing direct binding of SATB1 to BURs and standard ChIP-seq showing presumably indirect binding of SATB1 to gene-rich regions. In the inaccessible chromatin regions of nuclei, selected BURs are tightly anchored to the SATB1 nuclear architecture that resists high salt extraction. Therefore, we speculate that SATB1-BUR interactions are presumably stable, forming a chromatin scaffold. Such SATB1 chromatin scaffold further interacts dynamically with active gene-rich regions through indirect binding to form functional chromatin complexes (or hubs) to regulate gene expression. Future studies are required to determine the potential role of soluble SATB1 proteins in these interaction events.

Similar DNA binding profiles for PRC2 core subunits ained by stardard ChIP-seq and Urea ChIP-seq

The DNA binding profiles of PRC2 core subunits, Jarid2, Suz12, and Ezh2 obtained by standard ChIP-seq (64) [Peng et al., Cell, 2009, GEO repository (GSE18776)] were compared against the DNA binding profiles of these proteins by urea ChIP-seq. Their binding profiles in the Hoxd cluster region on chromosome 2 (mm9) are shown. Their DNA binding profiles by urea ChIP-seq and by standard ChIP-seq are very similar.

SATB1 does not directly co-bind chromatin with CTCF and RAD21 or has a role in TAD formation.

(A) Urea ChIP-Western studies for CTCF, RAD21 and SATB1. Original data for Figure 2C is shown. The excised areas are indicated by dotted boxes (first three panels). Western blot with the anti-H3 antibody shows that the original urea-purified chromatin samples before ChIP were similar in amount (right panel). (B) Read alignment statistics for Hi-C datasets, as absolute reads (left) and relative reads (in %, right). “ds.accepted.intra” are all intra-chromosomal reads used for all downstream analyses. (C) Distribution of TAD sizes and number of TAD calls per sample using hic-ratio TAD caller at 10kb bin resolution and the insulation score of 100kb windows. (D) Scatter plot of pair-wise comparisons of TAD activities (the sum of normalized Hi-C chromatin contacts inside the TAD in WT) between WT and KO thymocyte samples, showing well-conserved intra-TAD activity in KO thymocytes. (E) The Pearson correlation estimates computed between the counts for every pair of 10kb bins on chromosome 2 shows that WT (control: upper right triangle) and KO (observed:lower left triangle) thymocytes have no difference in compartmentalization. The compartment A (red) and compartment B (blue). (F) The scaling plots displaying how the Hi-C contacts decrease as a function of genomic distance.

Standard ChIP-seq for SATB1 produces non-specific peaks in Satb1−/− thymocytes.

(A) Urea ChIP-seq (Urea) and std ChIP-seq (Std) profiles for SATB1 and input chromatin in a randomly chosen, 3.12 Mb region in chromosome 17 shows appearance of some peaks in input chromatin from std ChIP-seq that match with peaks for SATB1 in std ChIP-seq (top). For Satb1−/− thymocytes (KO), using either anti-SATB1 antibody 1583 (rabbit polyclonal, SATB1-1) or rabbit monoclonal ab109122 (SATB1-2), urea and std ChIP-seq experiments for SATB1 were performed. With std ChIP-seq, multiple peaks were generated in KO thymocytes in the similar regions observed in WT thymocytes (panel A track 7 versus 9 and 10), while urea ChIP-seq produced only few peaks (panel A top, track 2 versus 4 and 5). A zoom-in view of blue-highlighted SATB1-clustered regions (bottom) shows that for standard ChIP-seq, multiple peaks in KO thymocytes co-mapped with peaks in WT thymocytes, but others did not (panel A bottom, track 7 versus 9 and 10). (B) Another randomly chosen 3.45 Mb region in X chromosome (panel B top) and its zoom-in view (panel B bottom) shows high incidence of std ChIP-seq peaks in KO thymocytes co-mapping with std ChIP-seq peaks in WT thymocytes (tracks 7 versus 9 and 10).

Urea 4C-seq protocol.

Starting with urea-ultracentrifugation purified formaldehyde crosslinked chromatin: (1) A bait sequence is designed adjacent to a HindIII site of a HindIII–HindIII fragment containing SATB1 bound site (e.g. BUR-1). (2) Ligation is performed to ligate the original SATB1-bound fragment (bait) with other fragments whose close proximity was fixed by formaldehyde at the time of crosslinking and presumably retained after urea purification. The fixed chromatin structure with proximity is illustrated by pink shade. (3) Chromatin is decrosslinked and purified. (4) Ligated genomic DNA fragments are digested with HaeIII (secondary restriction enzyme). (5) A Linker fragment containing an Xho1 site and Biotin labeled on one strand is blunt-end ligated to HaeIII sites of the ligated genomic DNA fragments shown in step 4 to generate A, B, C, D double stranded fragments. o 1 sites are indicated by red arrowhead. (6) The ligated products are digested with Xho1. (7)Biotin-labeled double stranded DNA (B, C, and D) are captured by Streptavidin-Dynabeads. (8) The beads are treated with NaOH to convert DNA to single strand DNA. This will result in dissociation and elution of non-biotinylated single strand DNA (derived from B and C) from the Dyanbeads. (9) Using a specific primer including the HindIII of the bait sequence at the end and another primer from the Linker, captured DNA fragments are PCR amplified. (10) Control experiment is performed to confirm that Linker primer alone does not generate any amplified products. Once this is confirmed, the experiment is expected to contain minimum background signals from non-specific sequences.

Long-range chromatin interactions from BUR-1 traverse TADs.

BUR-1 interactions determined by urea4C-seq for wild type (WT) and SATB1-KO thymocytes covers the entire 5.75 Mb gene-rich region and these interactions bypass multiple small TADs revealed by the Hi-C heatmap (mm10) generated in the same region. Dotted line indicates the position of BUR-1

Long-distance interactions covering distal gene-rich regions by BUR-1.

(A) A wide view (44.61 Mb) covering distal gene-rich regions shows BUR-1 (red bracket) interacts with distal gene-rich DNase 1 HS regions marked by clusters of H3Kme1 and H3K27a (indicated by blue horizontal bars). These interactions were greatly reduced in SATB1 deficient thymocytes (KO). A gene-rich region (indicated by a black horizontal bar) that does not correspond to DNase 1 HS is enriched in olfactory receptor genes. DNase 1 hypersensitive (HS) regions, which indicate highly accessive chromatin (or “open” chromatin). H3Kme1 and H3K27ac data are from ENCODE (LICR) and DNase 1 HS data are from ENCODE (University of Washington). (B) SATB1 peaks identified by standard (Std) ChIP-seq in this region are mostly real and only detected in wildtype (WT) and not in SATB1-null (KO) thymocytes.

Genes within the gene-rich region (chr2:98,542,232-108,455,449) that exhibit SATB1-dependent expression.

Genes showing differential expression for SATB1 with p<0.05 are shown in red (SATB1 activated) and in blue (SATB1 repressed). Data are from Zelenka et al., Nat Commun.,13, 6954, 2022 (54).

BURs and BUR-bound SATB1 largely co-map with LaminB1-DamID.

The BUR distribution assessed by the BUR reference map (tracks 2) was compared against Lamin B1-DamID (track 1) over a 47.4Mbp region in chromosome 3. Lamin B1-DamID data are from Hao et al., J. Exp. Med, 212, 809, 2015 (45). The BUR distribution shows very similar overall distribution of LADs mapped by Lamin B1-DamID. BUR-binding frequency of SATB1 in thymocytes (track 3) is lower than that in brain (track 4). Distribution of BURs (track 2) and SATB1-binding regions determined by urea ChIP-seq (Urea) (track 3, 4, and 5) largely overlap with LADs (track 1). These regions mostly avoid DNase hypersensitive regions (HS) (track 10), which overlap with SATB1-binding sites determined by standard ChIP-seq (Std) (track 6), CTCF-binding sites (track 7), enhancer marks (H3Kme1 and H3K27ac (tracks 8 and 9, respectively). Data for CTCF, H3Kme1 and H3K27ac are from ENCODE (LICR) and for DNase 1 hypersensitive sites (HS) is from ENCODE (University of Washington).

Some SATB1-bound BURs located in SATB1-dependent gene loci are outside of LADs.

SATB1-bound BURs shown in Figure 4 (indicated by black box) in the Rag1/Rag2, Foxp3, and Runx3 gene loci are found outside LADs mapped by LaminB1-DamID. These are close to H3Kme1 and H3K27ac sites. H3Kme1 and H3K27ac data are from ENCODE (LICR).

Interaction sites from BUR-1 or BUR-2 are mostly confined to the inter-LAD region.

The region containing BUR-1 and BUR-2 as baits is light gray shaded with a pink arrowhead. This region is near the border of a LAD and an inter-LAD region, shown by LADs identified by Lamin B-DamID. Interaction profiles from either BUR-1 and BUR-2 obtained in wild-type thymocytes (WT) are found in a region surrounded by LADs. These interactions are greatly reduced in Satb1-KO thymocytes (KO). Many interaction sites are close to H3Kme1 and H3K27ac sites. H3Kme1 and H3K27ac data are from ENCODE (LICR).

Intersection analysis of ChIP-seq samples

Intersection of peaks were determined by GeneTool between two samples of choice. Percent of peaks in each sample on the left column intersected with peaks in each of the samples on the top row are indicated.

ChIP-seq sample identification and peak number

Identification of each sample and their total number of peaks shown.

Intersection analysis of WT and KO ChIP-seq samples

Intersection of peaks were determined by GeneTool between two samples of choice. Percent of peaks in each sample on the left column that intersected with peaks in each sample on the top row are indicated. Important sets of data for comparison are highlighted in blue for urea ChIP-seq and pink for standard ChIP-seq.

WT and KO ChIP-seq sample identification and total peak number

Identification of each sample and their total number of peaks are shown.