PU.1 eviction at lymphocyte-specific chromatin domains mediates glucocorticoid response in acute lymphoblastic leukemia

Abstract

The epigenetic landscape plays a critical role in cancer progression, yet its therapeutic potential remains underexplored. Glucocorticoids are essential components of treatments for lymphoid cancers, but resistance, driven in part by epigenetic changes at glucocorticoid-response elements, poses a major challenge to effective therapies. Here we show that glucocorticoid treatment induces distinct patterns of chromosomal organization in glucocorticoid-sensitive and resistant acute lymphoblastic leukemia xenograft models. These glucocorticoid-response elements are primed by the pioneer transcription factor PU.1, which interacts with the glucocorticoid receptor. Eviction of PU.1 promotes receptor binding, increasing the expression of genes involved in apoptosis and facilitating a stronger therapeutic response. Treatment with a PU.1 inhibitor enhances glucocorticoid sensitivity, demonstrating the clinical potential of targeting this pathway. This study uncovers a mechanism involving PU.1 and the glucocorticoid receptor, linking transcription factor activity with drug response, and suggesting potential therapeutic strategies for overcoming resistance.

Fig. 1. HiC multi-omics annotation and changes in ALL-54S and ALL-50R pre- and post-treatment with DEX.

Violin plots of HiC contact intensities quantified using principal component analysis (PCA) in genomic regions for A (i) 0–9 DHS domains (DHS-Low), (ii) 10-19 DHS domains (intermediate, DHS-Mid), (iii) ≧20 DHS domains (DHS-High), (iv) lymphocyte-specific open regions (LSOs), (v) lymphocyte-specific closed regions (LSCs), (vi) LSOs with GR binding (GR-LSOs), (vii) LSOs without GR binding (Non-GR-LSOs), (viii) 42 activated GR-LSOs and (ix) 61 repressed GR-LSOs; for B genomic regions with different abundance of LSOs. Statistics by Wilcoxon matched-pairs signed rank test, ***p < 0.001; ****p < 0.0001; NS, not significant. See Supplementary Data 1 for exact p values. C Gene expression levels measured by RNA-seq at the regions defined in A. Data are presented as normalized transcript counts across transcription start sites (TSS) to transcription end sites (TES) of all genes located in each bin category of genomic regions. D DEX-induced HiC contact intensity changes in ALL-50R and ALL-54S in vivo. Dynamic bins are genome regions that changed in HiC intensity. IN, increase from negative to positive PCA values or >1.5-fold increase; DE, decrease from positive to negative PCA values or >1.5-fold decrease). Stable bins (ST) are genome regions with no change of HiC contact intensities, including ST-Low Contacts with negative PCA values and ST-High-Contacts with positive PCA values. E Volcano plots depicting differential expressed bins of HiC contact intensities between control and dex-treated groups in ALLs. A generalized linear model “estimateGLMCommonDisp” in R package EdgeR (version 4.2.1) was applied to calculate the statistical significance for the exponential transformed PCA values of each HiC bin (n = 56380). F Heatmap of HiC contact intensities in ALLs. The top row indicates the 103 (i.e., 42 GR-activated or 61 GR-repressed) LSOs as well as their adjacent 50 kb regions (42 GR-LSOs +/− nearby bin) and the bottom row highlights genes identified by a ΔΔHiC analysis (ΔΔHiC = ΔHiC of ALL-54S (Dex – Ctrl) - ΔHiC of ALL-50R (Dex – Ctrl)). See also Supplementary Data 1–3 and Supplementary Fig. 1. Source data are provided as a Source Data file.

Fig. 2. Identification of PU.1 as a key regulator for genome regions associated with HiC changes in ALL-54S and ALL-50R pre- and post-treatment with DEX.

A Motif enrichment analysis using i-cisTarget at genomic regions associated with HiC contact intensity changes in ALL-54S and ALL-50R pre- and post-treatment with DEX. The normalized enrichment scores (NES) for all TF motifs are shown in blue and the top 10 TF motifs are highlighted in red and annotated. B Violin plots of the NES for the top 10 ranked TF motifs, two lineage-specific TFs EBF1 and TCF3, and structural protein CTCF. C GR-interacting proteins identified using Rapid Immunoprecipitation Mass Spectrometry (RIME). Interacting affinities were calculated as a differential abundance of the GR-bound proteins in DEX-treated vs. control samples in ALL-50R and ALL-54S. GR co-regulators, previously reported GR-interacting proteins^– to mediate GR actions. TF, transcription factor; FC, fold change. Statistics were performed using limma through the FragPipe-Analyst (http://fragpipe-analyst.nesvilab.org/). N = 3, the data was generated from PDX cells in 3 separate mice of each group. D Relative changes of TF-GR interacting affinities in ALL-54S compared with ALL-50R upon DEX treatment. Δlog2FC = log2FC in ALL-54S – log2FC in ALL-50R based on the data from C. N = 3. E Violin plots of PU.1 and CTCF ChIP-seq enrichment in ALL-50R and ALL-54S pre- and post-treatment with DEX at genomic regions described in Fig. 1D. ST = genomic bins with stable HiC-contact intensities after DEX treatment, IN = bins with increased HiC-contact intensities after DEX treatment; DE = bins with decreased HiC-contact intensities after DEX treatment. ALL-50R: IN, n = 3956; DE, n = 3299; ST, n = 49125. ALL-54S: IN, n = 7036; DE, n = 6578; ST, n = 42766. See also Supplementary Data 4-5. Source data are provided as a Source Data file.

Fig. 3. Combinatorial binding of PU.1 and other TFs in ALL-50R and ALL-54S pre- and post-treatment with DEX.

A GR, PU.1, and CTCF ChIP-seq binding profile plots at GR-LSOs, non-GR-LSOs, LSOs, and LSCs in DEX-treated ALL-54S. BS, binding sites. B PU.1 and CTCF ChIP-seq binding profile plots at GR-LSOs and non-GR-LSOs in ALL-54S and ALL-50R pre- and post-DEX treatment. C Pie chart showing the binding and co-occurrence of PU.1, CTCF, and EBF1 in ALL-54S pre- and post-treatment with DEX. D Summary of binding dynamics of PU.1, CTCF, and EBF1 at GR-LSO in ALL-54S pre- and post-treatment with DEX. See also Supplementary Data 6 and Supplementary Fig. 2. Source data are provided as a Source Data file.

Fig. 4. PU.1 binding at gene regulatory elements and associated network modules in ALL-50R and ALL-54S pre- and post-treatment with DEX.

A Epigenome categorization in ALL-54S and ALL-50R pre- and post-DEX treatment. Top panel: Combinatorial code of histone modifications for epigenome categorization. Pro, promoter with H3K4Me3; Enh, enhancer with H3K4Me1; four sub-categories under Pro or Enh: S, silent; A, active; R, repressed; Bi, bivalent. RO, repressed mark H3K27Me3 only; AO, active mark H3K27Ac only; NS, no signal (regions lacking ChIP enrichment). Lower panel: GR, PU.1, and CTCF binding at defined epigenome categories in ALL-54S and ALL-50R pre- and post-treatment with DEX. B RNA-seq signal of GR-LSO-associated genes and H3K27Ac/H3K27Me3 ChIP-seq signal at GR-LSOs in ALL-54S and ALL-50R pre- and post-DEX treatment. HiC data was used to associate LSOs with genes and the five genes with the highest scores in each LSO were used for the analysis. H3K27Ac and H3K27Me3 ChIP-seq profiles are +/− 5 kb surrounding GR-LSOs. Red box, clusters 2 showing relatively higher H3K27Ac enrichment after DEX treatment in ALL-54S; yellow box, genomic regions showing lower H3K27Me3 enrichment in ALL-54S compared to ALL-50R. C Bar-graph of -log (p-values) calculated using the Fisher exact test in Ingenuity Pathway Analysis to determine the molecular pathways enriched between ALL-54S and ALL-50R upon DEX treatment in vivo based on genes expression levels (RNA-Seq; N = 87), protein interactions (RIME; N = 125), and gene regulatory elements (ChIP-seq; N = 247). D Cystoscope model integrating i-cisTarget motifs, TF ChIP-seq, and histone marks ChIP-seq at LSOs. See also Supplementary Data 6-7 and Supplementary Fig. 3. Source data are provided as a Source Data file.

Fig. 5. Multi-omics annotation of two de novo TADs at the BIM and ZBTB16 loci in ALL-50R and ALL-54S pre- and post-treatment with DEX.

HiC interaction map and ΔHiC analysis at BIM TAD (A) and ZBTB16 TAD (C) of ALL-50R and ALL-54S pre- and post-treatment with DEX (top) and directionality index (DI; bottom). Triangles indicate TADs. DI shows a preference to interact with either an up or downstream region. (B, D) UCSC Genome Browser tracks at BIM locus (B) and ZBTB16 locus (D) showing LSO-interacting genomic regions by HiC and CTCF, PU.1, GR, H3K27Ac, H3K4Me1, and H3K27Me3 ChIP-seq. Blue and green highlights are the up and downstream TAD borders, yellow highlight is LSO, pink highlight is super-enhancer (SE). E UCSC Genome Browser tracks of HiC interaction at ZBTB16 SE in ALL-50R and ALL-54S pre- and post-DEX treatment. F Heatmap of RNA-seq data showing gene expression within the BCL2L11 (BIM) TAD and ZBTB16 TAD in ALL-50R and ALL-54S pre- and post-DEX administration in vivo. G UCSC Genome Browser tracks of ATAC-seq data showing accessible genomic regions at the ZBTB16 locus in a panel of DEX-sensitive and resistant ALL PDXs pre- and post-treatment with DEX. T-ALL 8I has intermediate resistance to glucocorticoids. See also Supplementary Fig. 4.

Fig. 6. Combination treatment with PU.1 inhibitor DB2313 in ALL.

A Left: Schematic of PU.1 motif that is most responsive to DB2313 treatment (DB-sensitive) and showed PU.1 binding loss after DB2313 treatment. Right: log-odds ratio for enrichments of DB-sensitive PU.1 motif in GR-LSOs (n = 1510) and non-GR-LSOs (n = 10707). Statistics were performed using receiver operating characteristics (ROC) analysis. B PU.1 ChIP-seq profiles and heatmap plots (DeepTools) at LSOs in a B-ALL cell line (kopn8) exposed to DEX (Dex, 1 μM) alone or a combination treatment of 1 μM Dex and 1 μM of a PU.1 inhibitor DB2313 (DB) for 16 h in vitro. C ATAC-seq analysis of Nalm6 and ALL-54S cells treated with DEX ± DB2313 for 48 h in vitro. Upper left panel normalized ATAC-seq intensity at LSOs. Gene-set enrichment analysis was performed using the “RHEIN ALL GLUCOCORTICOID THERAPY UP” gene set (RAGTU, lower left panel) and the top activated signaling pathways (right panel) comparing Dex + DB with Dex-treated Nalm6 cells. See the Methods section for statistics. D RNA transcriptional activities in Nalm6 treated with DEX ± DB2313 for 48 h in vitro. The data are presented as normalized transcript counts across TSS to TES of genes associated with LSOs (LSO-related Genes), GR-LSOs (GR-LSO Genes), GR-LSOs with PU.1 Loss (PU.1 Loss at GR-LSOs), RAGTU gene set, and differentially expressed genes comparing Dex treatment and control Nalm6 cells (DEG). E Time course study of DEX ± DB2313 induced gene expression in Nalm6 cells. Cells were exposed to DEX (1 μM) ± DB2313 (1 μM). GILZ is a direct target of GR and was used as a positive control. F Cytotoxicity assays of DEX (1 μM) ± DB2313 (1 μM) in two ALL cell lines (Nalm6 and 697) and two ALL PDXs (RJ-1 and RJ-10) treated for 72 h in vitro. G Cytotoxicity assay to determine synergistic effects of DEX and DB in inducing cell death of Nalm6 Cells. Interactive analysis is performed using SynergyFinder. A synergy score >10 is considered significant. Data in E, F, and G are presented as the mean ± SEM of three biological replicates. P value was calculated using two-sided student t-test. See also Supplementary Fig. 5–7. Source data are provided as a Source Data file.

Fig. 7. PU.1 Knockdown in ALL-54S in vivo.

Engraftment (A) and Event-free Survival (EFS) curves (B) of ALL-54S cells with BIM KO, PU.1 KO or Non-Targeting Control (NTC) guide RNA in mouse peripheral blood (PB). Gene KO was induced by doxycycline (Dox) feeding. Leukemia event is defined as 25% human cells in mouse PB. Gene KO was induced by doxycycline (Dox) feeding. Leukemia event is defined as 25% human cells in mouse PB. Statistics were performed using log-rank (Mantel-Cox) test, n = 6/group except for the BIM KO DEX-treated group has n = 5 due to mouse loss from non-leukemia related disease. Insertion and deletion profiles of relapsed BIM KO (C) and PU.1 KO (D) ALL-54S clones following saline (Ctrl)/DEX treatment in vivo. Genomic DNA of PDX cells from two independent mice were sequenced and representative profiles from each group were shown. The TIDE webtool (http://tide.nki.nl) was used for sequencing data analysis. See also Supplementary Fig. 7. Source data are provided as a Source Data file.