Transcriptional control of CBX5 by the RNA binding proteins RBMX and RBMXL1 maintains chromatin state in myeloid leukemia

Abstract

RNA binding proteins (RBPs) are key arbiters of post-transcriptional regulation and are found to be found dysregulated in hematological malignancies. Here, we identify the RBP RBMX and its retrogene RBMXL1 to be required for murine and human myeloid leukemogenesis. RBMX/L1 are overexpressed in acute myeloid leukemia (AML) primary patients compared to healthy individuals, and RBMX/L1 loss delayed leukemia development. RBMX/L1 loss lead to significant changes in chromatin accessibility, as well as chromosomal breaks and gaps. We found that RBMX/L1 directly bind to mRNAs, affect transcription of multiple loci, including CBX5 (HP1α), and control the nascent transcription of the CBX5 locus. Forced CBX5 expression rescued the RBMX/L1 depletion effects on cell growth and apoptosis. Overall, we determine that RBMX/L1 control leukemia cell survival by regulating chromatin state through their downstream target CBX5. These findings identify a mechanism for RBPs directly promoting transcription and suggest RBMX/L1, as well as CBX5, as potential therapeutic targets in myeloid malignancies.

Extended Data Fig. 1. RBMX is required for murine leukemia maintenance but is dispensable for leukemia initiation

(a) Log₂ fold depletion of control shRNAs (n = 5 independent shRNAs) and Rbmx-specifc shRNAs (n = 4 independent shRNAs) in MLL-AF9 leukemic BM cells in pooled in vivo shRNA screen. (b) qRT-PCR showing efficient Rbmx deletion in MLL-AF9 leukemia cells 4 days post transduction. n = 3 independent experiments. (c) Representative FACS plots of Fig. 1c. (d) Representative FACS plots of Fig. 1d. (e-f) qRT-PCR of Rbmx in normal c-Kit⁺ BM cells (control and shRNA97: n = 4, shRNA96: n = 3 independent experiments) and in MLL-AF9 leukemia cells (n = 3 independent experiments). (g) qRT-PCR of Rbmx in AML1-ETO leukemia cells. n = 3 independent experiments. (h) Colony formation assay of cells in h. n = 3 independent experiments. (i) Diagram for generating Rbmx conditional knockout mice. (j) Genotyping PCR detecting Rbmx deletion in BM cells isolated from Rbmx^Δ/Δ and Rbmx^Δ mice before leukemic transformation. Cre bands indicate the presence of Mx1-Cre allele. Representative image of 2 independent experiments with similar results. (k) qRT-PCR showing Rbmx deletion of cells in j. n = 3 independent experiments. (l) Experimental scheme for leukemia initiation experiments. (m) Kaplan-Meier survival analysis of mice injected with female MLL-AF9-transformed Rbmx^f/f and Rbmx^Δ/Δ cells. Rbmx^f/f n = 9, Rbmx^Δ/Δ n = 10 mice; two-sided Mantel-Cox test. (n) Spleen and liver weights from mice that succumbed to disease in m. Rbmx^f/f n = 8, Rbmx^Δ/Δ n = 9 mice. (o) Kaplan-Meier survival analysis of mice injected with male MLL-AF9 transformed Rbmx^f and Rbmx^Δ cells. Rbmx^f/f n = 10, Rbmx^Δ/Δ n = 12 mice; two-sided Mantel-Cox test. (p) Spleen and liver weights from mice that succumbed to disease in o. Rbmx^f/f n = 8, Rbmx^Δ/Δ n = 6 mice. (q) Immunoblot analysis of MLL-AF9 Rbmx^f/f Cre-ER cells expressing ectopic human RBMX and knocking down of endogenous RBMX when treated with 400nM 4-OHT. The experiment was performed 4 times with similar results. (r) qRT-PCR of the cells from q. n = 3 independent experiments. Data presented in a-b, e-h, k, n, p, r as mean +/− s.e.m. p value determined by two-tailed Student’s t test, unless stated otherwise.

Extended Data Fig. 2. Retrogene Rbmxl1 is functionally redundant with Rbmx in vitro and in vivo.

(a) Immunoblot analysis of RBMX/L1 protein in the BM cells isolated from Rbmx^f/f and Rbmx^f mice and RBMXL1 from Rbmx^Δ/Δ and Rbmx^Δ before leukemic transformation. The experiment was performed once. (b) qRT-PCR of Rbmx and Rbmxl1 expression by shRNAs specific to Rbmx (but not Rbmxl1) in MLL-AF9 leukemia cells. Control and shRNA96: n = 3, shRNA97: n = 4 independent experiments. (c) qRT-PCR of Rbmx and Rbmxl1 expression in MLL-AF9 Rbmx^f/f Cre-ER cells treated with 4-OHT for 24 hours. n = 3 independent experiments. (d) Quantitative FACS analysis summary of Gr-1⁺Mac-1⁺ and CD115⁺F4/80⁺ cells in control and RBMXL1-knockdown leukemic Rbmx^Δ/Δ cells 4 days post transduction. Control and shRNA587: n = 6, shRNA932: n = 4 independent experiments. (e) Representative FACS plots of Fig. 2j. (f) Representative FACS plots of Fig. 2k. (g) Spleen and liver weights from mice that succumbed to disease in Fig. 2l. Control: n = 9, shRNA587: n = 5, shRNA932: n = 4 mice. (h) Immunoblot analysis of leukemic BM cells isolated from mice that succumbed to disease in Fig. 2l. The experiment was performed once. Data presented in b-d, and g as mean +/− s.e.m. p value determined by two-tailed Student’s t test, unless stated otherwise.

Extended Data Fig. 3. RBMX and RBMXL1 overexpression in myeloid leukemia.

(a)RBMX expression in AML patients with indicated status of NPM1 and FLT3-ITD mutation. Data are presented as mean of normalized RPKM +/− s.e.m, on the basis of data from the BeatAML vizome dataset (Tyner JW et al. Nature. 2018). Double negative n = 286, FLT3-ITD positive n = 54, NPM1 mut positive n = 59, double positive n = 49 patient samples; two tailed Student’s t test with Welch’s correction. (b)RBMX expression in leukemic stem cell (LSC⁺) and non-leukemic stem cell population (LSC⁻) in AML patient samples on the basis of data from Ng SW et al. Nature. 2016 (GSE76009). LSC⁻ n = 89, LSC⁺ n = 138 patient samples; data as mean +/− s.e.m, two-tailed Student’s t test. (c-d) qRT-PCRs showing RBMX and RBMXL1 mRNA levels in multiple myeloid leukemia cell lines. For RBMX CB-CD34⁺, THP-1, HL-60, U-937 and K562: n = 3, MOLM13, Nomo-1 and NB4: n = 5, Kasumi-1, KCL-22, KG-1 and TF-1: n = 4 independent experiments; For RBMXL1 CB-CD34⁺, Nomo-1, THP-1, HL-60, U-937 and TF-1: n = 3; MOLM13, Kasumi-1, K562 and KG-1: n = 4; NB4 and KCL-22: n = 5 independent experiments. Data as mean +/− s.e.m, two-tailed Student’s t test. (e) CRISPR score rank of RBMX, its retrogenes (RBMXL1, RBMXL2, and RBMXL3), and its paralog RBMY1A1. CRISPR score is the average log₂ fold-change in the abundance of all sgRNAs targeting the gene after 14 population doublings. (f) CRISPR score of RBMX, its retrogenes, and its paralog across the 14 tested leukemia cell lines.

Extended Data Fig. 4. RBMX and RBMXL1 are required for human myeloid leukemia cell survival.

(a) Giemsa staining of control and RBMX/L1-knockdown MOLM13 cells from Fig. 4k. Original magnification 400X. Scale bar, 20μm. The experiment was performed once; control and shRNA1: 9 images, shRNA2: 16 images were collected. (b-d) Representative FACS plots of Fig. 4m, o and p, respectively. (e) Immunoblot analysis of RBMX/L1 in KCL-22 transduced with EV or RBMX overexpressing vector (RBMX-R2) and shRNA control or shRNAs against RBMX/L1. The experiment was performed 3 times with similar results. (f) Proliferation assay of cells in e. n = 3 independent experiments. (g) Quantitative FACS analysis of apoptotic cells in control and RBMX/L1-knockdown KCL-22-EV and KCL-22-RBMX-R2 cells 5 days post transduction. n = 3 independent experiments. (h) Immunoblots for RBMX/L1 in MOLM13-Cas9 cells transduced with sgRNAs targeting RBMX/L1 (sg-1 and sg-2). The experiment was performed 4 times with similar results. (i) Proliferation assay of cells in h. n = 4 independent experiments. (j) FACS analysis of myeloid differentiation markers CD11b and CD33 in cells from h. n = 4 independent experiments. (k) Quantitative FACS analysis summary of apoptosis of cells in h, 3 days post transduction. n = 4 independent experiments. (l) Immunoblots analysis of GFP⁺ PDX AML-1 cells. The experiment was performed once. (m) qRT-PCR showing depletion of RBMX/L1 in GFP⁺ PDX AML-13 cells. n = 1 experiment. (n) Immunoblots for RBMX/L1 upon RBMX/L1 depletion in GFP+ PDX AML-11 cells. The experiment was performed once. (o) Representative FACS plots of Fig. 4t. (p) Immunoblot and band densitometry of RBMX/L1 in BM cells from animals transplanted with PDX AML-1 cells that succumbed to leukemia. The experiment was performed once. (q-s) Representative FACS plots of Fig. 5c, g and h, respectively. Data presented in f-g, i-k as mean, error bars, s.e.m. p values were calculated using two-tailed Student’s t test, unless indicated otherwise.

Extended Data Fig. 5. Loss of RBMX and RBMXL1 results in a dysregulated chromatin state in leukemia cells.

(a) Immunoblot analysis of RBMX/L1, MYC, and HOXA9 in RBMX/L1 depleted MOLM13 cells. The experiment was performed 3 times with similar results. (b) Heatmap of the top 1,000 peaks from ATAC-sequencing from control and RBMX/L1-knockdown MOLM13 cells. n = 3 independent experiments. (c) Scatterplot showing accessibility changes at pericentric and telomeric heterochromatin upon RBMX/L1-knockdown. n = 3 independent experiments. (d) Location of increased accessible and decreased accessible ATAC-sequencing peaks in RBMX/L1-knockdown MOLM13 cells. (e) Location of increased accessible and decreased accessible ATAC-sequencing pericentric and telomeric heterochromatin peaks in RBMX/L1-knockdown MOLM13 cells. (f) Gene expression heatmap of the top 99 upregulated and downregulated genes from RNA-sequencing analysis of MOLM13 cells upon RBMX/L1 knockdown. n = 4 independent experiments. (g) Tables showing alternative splicing events and genes MOLM13 cells upon RBMX/L1 depletion. n = 4 independent experiments. (h) mRNA expression of the 11 genes from the overlap in Fig. 7a. n = 4 independent experiments; data as mean +/− s.e.m, two-tailed Student’s t test. (i) qRT-PCR of recovered RNA in RNA-IP at 11 candidate target genes shown in Fig. 7a. n = 4 independent experiments; data as mean +/− s.e.m, two-tailed Student’s t test. (j) Overall survival analysis of AML patients with low versus high expression of RBMX/L1 direct regulated pathway including 8 down-regulated targets validated by PAR-CLIP, RNA-IP and RNA-sequencing (CBX5, CBS, DACH1, SEPT11, UNG, XBP1, PABPC4, and SLC38A1). Data from TCGA database.

Extended Data Fig. 6. Loss RBMX/L1 in MOLM13 cells leads to decreased CBX5 transcript expression.

(a) Immunoblot analysis and band densitometry of RBMX/L1, DACH1, CBX5, CBS, and SEPT11 upon RBMX/L1 depletion in MOLM13 cells. The experiment was performed 3 times with similar results. (b)CBX5 mRNA expression in leukemic RBMXL1 depleted Rbmx^Δ cells. n = 5 independent experiments. (c) Immunoblots and band densitometry of RBMX/L1 and CBX5 in leukemic RBMXL1 depleted Rbmx^Δ cells. The experiment was performed 3 times with similar results. (d)CBX5 mRNA expression upon RBMX/L1 depletion in MOLM13-Cas9 cells. n = 4 independent experiments. (e) Immunoblots and band densitometry of RBMX/L1 and CBX5 upon RBMX/L1 depletion in MOLM13-Cas9 cells. Same immunoblot as Extended data Fig. 4h with longer exposure for RBMX/L1 band. The experiment was performed 3 times with similar results. (f) qRT-PCR of CBX5 mRNA expression at indicated exon and exon-exon junction. Exon 3 amplicon: n = 3, all other amplicons: n = 6 independent experiments; data as mean +/− s.e.m, two-tailed Student’s t test. (g) mRNA stability of RBMX and CBX5 upon RBMX/L1 depletion in MOLM13 cells. 0 min and 90 min: n = 5, 30 min and 270 min: n = 4, 150 min: n = 3 independent experiments. (h) qRT-PCR of nascent mRNAs of RBMX/L1 candidate targets upon RBMX/L1 depletion. DACH1: n = 4, SEPT11 and PABPC4: n = 5, CBS control and shRNA1 n = 6 and shRNA2 n = 5, XBP1 control n = 6 for, shRNA1 and shRNA2 n = 4, UNG control and shRNA1 n = 6, shRNA2 n = 4 independent experiments. (i) Quantitative summary of smFISH with CBX5-Intron 1 probe (Fig. 7f). Control: n = 229, shRNA1: n = 71, shRNA2: n = 38 foci. (j) qRT-PCR measuring absolute number of CBX5-Intron 1 and Renilla luciferase (Rluc) nascent mRNA transcripts in CBX5-Intron 1 luciferase reporter assay. n = 4 independent experiments. (k) qRT-PCR of CBX5 and RBMX mRNA upon CBX5 depletion in MOLM13 cells. n = 3 independent experiments. (l) qRT-PCR of RBMX and CBX5 mRNA upon RBMX/L1 depletion in MOLM13-EV and MOLM13-CBX5 cells. EV-shRNA2 n = 4, all other groups: n = 3 independent experiments. (m) Cells from Fig. 8e were plated for proliferation assay and counted 4 days post transduction. n = 3 independent experiments. Data presented in b, d, f-m as mean +/− s.e.m. p values were calculated using two-tailed Student’s t test, unless indicated otherwise.

Figure 1. RBMX is required for murine leukemia maintenance in vitro and in vivo.

(a) Immunoblot showing knockdown of RBMX in mouse MLL-AF9 bone marrow (BM) leukemia cells. The experiment was performed 3 times with similar results. (b) Colony formation assay in RBMX-knockdown mouse MLL-AF9 leukemia cells. Number of colonies was normalized to that of control MLL-AF9 leukemia cells. n = 6 independent experiments. (c) Quantitative FACS analysis summary of myeloid differentiation markers in control and RBMX-knockdown MLL-AF9 leukemia cells. Control and shRNA96: n = 5 independent experiments, shRNA97: n = 4 independent experiments. (d) Quantitative summary of FACS analysis of apoptotic (Annexin V⁺) cells in control and RBMX-knockdown MLL-AF9 leukemia cells 5 days post transduction. n = 4 independent experiments. (e) Colony formation assay in normal c-Kit BM cells, n = 5 independent experiments, and MLL-AF9 leukemia cells, n = 6 independent experiments. (f) Experimental scheme for investigating the role of RBMX in leukemia maintenance in vitro and in vivo by generating inducible MLL-AF9 Rbmx^f/f Cre-ER cells. (g) Immunoblot showing depletion of RBMX after 24 hours of 200nM 4-OHT (TAM) treatment. The experiment was performed 3 times with similar results. (h) Colony formation assay in MLL-AF9 Rbmx^f/f Cre-ER cells treated with 4-OHT. n = 3 independent experiments. (i) Kaplan-Meier analysis of leukemia maintenance survival of mice injected with MLL-AF9 Rbmx^f/f Cre-ER cells with or without TAM treatment. NO TAM: n = 14 mice, TAM: n = 13 mice; two-sided Mantel-Cox test. (j) Colony formation assay in control and MLL-AF9 Rbmx^f/f Cre-ER cells depleted of endogenous RBMX and overexpressing ectopic human RBMX-FLAG. EV control: n = 10, RBMX control: n = 9, EV 4-OHT: n = 7, RBMX 4-OHT: n = 8, independent experiments. (k) Quantitative summary of FACS analysis of Gr-1⁺Mac-1⁺ cells in control and MLL-AF9 Rbmx^f/f Cre-ER cells depleted of endogenous RBMX and overexpressing ectopic human RBMX-FLAG. n = 3 independent experiments. Data presented in b, c, d, h, j, and k as mean +/− s.e.m. p value determined by two-tailed Student’s t test for all experiments in this figure, unless stated otherwise.

Figure 2. Retrogene Rbmxl1 is functionally redundant with Rbmx in leukemia maintenance in vitro and in vivo.

(a) Immunoblot analysis with RBMX/L1 protein abundance in leukemic BM cells isolated from Rbmx^f/f mice and RBMXL1 from Rbmx^Δ/Δ mice that succumbed to leukemia in Extended data Fig. 1n and p. The experiment was performed once. (b) Immunoblot analysis with RBMX/L1 in leukemic BM cells isolated from mice untreated with TAM and RBMXL1 from mice treated with TAM that succumbed to disease in Fig. 1h. The experiment was performed once. (c-e) qRT-PCR showing efficient Rbmx deletion and detectable Rbmxl1 expression in leukemic BM cells isolated from (c) female mice (Extended data Fig. 1n), (d) male mice that succumbed to disease in leukemia initiation experiment (Extended data Fig. 1p), and (e) mice that succumbed to disease in leukemia maintenance experiment (Fig. 1h). Rbmx^f/f: n = 8, Rbmx^Δ/Δ: n = 8, Rbmx^f: n = 6, Rbmx^Δ: n = 6, NO TAM: n = 10, TAM: n = 7 mice. (f) Experimental scheme for RBMXL1 depletion using Rbmxl1-specific shRNAs in leukemic Rbmx^Δ/Δ cells. (g) Immunoblot analysis showing knockdown of RBMXL1 in leukemic Rbmx^Δ/Δ cells 4 days post transduction. The experiment was performed 3 times with similar results. (h) qRT-PCR showing depletion of Rbmxl1 in leukemic Rbmx^Δ/Δ cells 4 days post transduction. Control and shRNA932: n = 3, shRNA587: n = 4 independent experiments. (i) Colony formation assay in RBMXL1 depleted and control leukemic Rbmx^Δ/Δ cells. Number of colonies was normalized to that of the control. n = 3 independent experiments. (j) Quantitative summary of FACS analysis showing myeloid differentiation (Gr-1 and Mac-1) in leukemic Rbmx^Δ/Δ colonies upon RBMXL1 depletion. Control: n = 5, shRNA587, shRNA932: n = 4 independent experiments. (k) Quantitative summary of FACS analysis of apoptotic cells in control and RBMX-knockdown leukemic Rbmx^Δ/Δ cells 5 days post transduction. n = 3 independent experiments. (l) Kaplan-Meier survival analysis of mice injected with control or RBMXL1-knockdown leukemic Rbmx^Δ/Δ cells. Control: n = 13, shRNA587: n = 10, shRNA932: n = 10 mice; two-sided Mantel-Cox test. Data presented in c-e and h-k as mean +/− s.e.m. p value determined by two-tailed Student’s t test for all experiments in this figure, unless stated otherwise.

Figure 3. RBMX and RBMXL1 are overexpressed in myeloid leukemia

(a)RBMX mRNA expression in AML compared to other cancers, on the basis of data from The Cancer Genome Atlas (TCGA) database. Data are presented as mean log₂ expression with range. AML, pink dots. Ordinary one-way ANOVA test with multiple comparisons. (b) Graph shows that RBMX is elevated in samples from patients with AML as compared to normal HSPCs. AML: n = 48, AML t(15;17): n = 54, AML inv(16)/t(16;16): n = 47, AML t(8;21): n = 60, AML t(11q23)/MLL: n = 43 patients. HSC: n = 6, GMP: n = 7, MEP: n = 4 healthy individuals (Bloodspot data of RBMX probe 1556336_at from the U133 Plus 2.0 array). Data as mean +/− s.e.m, two-tailed Student’s t test. (c)RBMX mRNA expression in different phases of chronic myeloid leukemia (CML), on the basis of data from ONCOMINE database (Radich dataset). CP: n = 42, AP: n = 15, BC: n = 36 CML patients. Data are presented in arbitrary unit, mean +/− s.e.m, two-tailed Student’s t test. (d) Immunoblot and band densitometry showing RBMX and RBMXL1 protein abundance in multiple human myeloid cells lines compared to normal HSPCs. The experiment was performed twice with similar results. (e) Two immunoblots (from two separate runs) measuring RBMX/L1 protein abundance in healthy CB-CD34⁺ cells, primary AML patient samples, healthy peripheral blood (PB), healthy BM cells, and aged (>60 years old) BM (Aged-BM). CB-CD34⁺ (1), (2), and (3) samples were pooled from 10, 4, and 5 individuals, respectively. PB, BM, and Aged-BM sample were each pooled from 3 individuals. Each AML patient sample represents one individual. Band densitometry presented below each blot, with blast percentage indicated above each bar.

Figure 4. RBMX and RBMXL1 promote human leukemogenesis.

(a-e) Immunoblot showing RBMX and RBMXL1 knockdown in MOLM13, THP-1, Kasumi-1, KCL-22, and K562 cells. The experiment was performed 3 times for each cell line with similar results. (f-j) Proliferation assay upon RBMX/L1 depletion in indicated cell line. MOLM13 Control and shRNA2: n = 5, shRNA1: n = 4; THP-1: n = 4; Kasumi-1: n = 4; KCL-22: n = 7; and K562: n = 3 independent experiments. (k-o) Quantitative FACS summary of differentiation markers upon RBMX/L1 depletion in indicated cell line. MOLM13 Control and shRNA2: n = 6, shRNA1: n = 4; THP-1: n = 6; Kasumi-1: n = 4; KCL-22: n = 5; and K562: n = 4 independent experiments. (p) Quantitative FACS summary of apoptotic cell death in indicated RBMX/L1-depleted cell line. MOLM13 Control: n = 5, shRNA1: n = 3, shRNA2: n = 4; THP-1: n = 6; Kasumi-1: n = 4; KCL-22 Control and shRNA1: n = 6, shRNA2: n = 5; and K562 Control: n = 4; shRNA1 and shRNA2: n = 3 independent experiments. (q) Kaplan-Meier analysis of NSG mice injected with MOLM13 cells transduced with RBMX/L1-specific shRNAs or control. Control: n = 8, shRNA1: n = 9, shRNA2: n = 10 mice; two-sided Mantel-Cox test. (r-s) Kaplan-Meier analysis in NSG mice injected with patient-derived xenograft (PDX) cells (patient AML-1 and patient AML-13 from Fig. 3e) transduced with RBMX/L1-specific shRNAs or control. PDX AML-1 Control: n = 3, shRNA1: n = 3, shRNA2: n = 3 mice. PDX AML-13 Control: n = 5, shRNA1: n = 4, shRNA2: n = 3 mice; two-sided Mantel-Cox test. (t) RBMX/L1 depletion resulted in decreased engraftment potential of PDX AML-11 (Fig. 3e). Quantitative FACS summary of indicated human CD45 marker and GFP (marker of shRNA expression vector). Control: n = 6, shRNA1: n = 4, shRNA2: n = 5 mice. Data presented in f-p and t as mean +/− s.e.m. p value determined by two-tailed Student’s t test for all experiments in this figure, unless stated otherwise.

Figure 5. RBMX and RBMXL1 differentially regulate human HSPCs.

(a) Immunoblot showing RBMX/L1 knockdown in CB-CD34⁺ cells. The experiment was performed 3 times with similar results. (b) Colony formation for cells from a. n = 3 independent experiments. (c) Quantitative FACS summary of CD33 and CD11b for cells from a after 2 days in myeloid differentiating media. n = 4, independent experiments. (d) Quantitative FACS analysis for apoptosis in CB-CD34⁺ cells. n = 6 independent experiments for day 4; n = 3 independent experiments for day 5. (e) Immunoblot showing CB-CD34⁺ cells overexpressing RBMX (RBMX-R2) compared to control (EV). The experiment was performed 3 times with similar results. (f) Colony formation assay for cells in e. n = 3 independent experiments. (g-h) Quantitative FACS analysis summary showing CD33⁺ CD11b⁺ population, n = 6 independent experiments, and CD34⁺ CD38⁻ population, n = 6 independent experiments, for cells from e at day 7 in myeloid differentiation promoting media. (i) Proliferation assay for cells from e. n = 3 independent experiments. Data presented in b-d and f-h as mean +/− s.e.m. p value determined by two-tailed Student’s t test for all experiments in this figure, unless stated otherwise.

Figure 6. RBMX and RBMXL1 regulate the chromatin state in leukemia cells.

(a) Immunoblot analysis showing that RBMX and RBMXL1 are predominantly localized in the chromatin-bound fraction. The experiment was performed once. GAPDH: cytoplasmic fraction loading control. Histone H3: chromatin-bound fraction loading control. (b) Complex karyotyping of cells depleted of RBMX/L1 and control 3 days post transduction. The experiment was performed 5 times with similar results, 450 images collected. (c) Quantitative summary of complex karyotyping of b. Control and shRNA1: n = 5, shRNA2: n = 3 independent experiments. Data as mean +/− s.e.m, two-tailed Student’s t test. (d) ATAC-sequencing identified major changes in chromatin state in RBMX/L1-knockdown cells compared to control 3 days after transduction. Significant changes (peaks) determined by two-tailed Student’s t test, q-value < 0.05. n = 3 independent experiments. (e) Heatmap of all pericentric and telomeric heterochromatin peaks from the ATAC-sequencing in d. n = 3 independent experiments. (f) Representative tracks of the ATAC-sequencing in d with increased and decreased compaction at pericentric heterochromatin.

Figure 7. RBMX/L1 directly bind to CBX5 transcripts.

(a) Venn diagram showing overlap between genes differentially expressed after RBMX/L1 depletion (RNA-seq) and genes bound by RBMX/L1 (PAR-CLIP). (b) RNA-sequencing and PAR-CLIP profiles of CBX5 transcript; arrows mark two significant RBMX enrichment sites at the start of intron 1 and in the 3’UTR. (c) Downregulated steady state CBX5 mRNA expression upon RBMX/L1 depletion in cells 2 days post transduction. n = 3 independent experiments. (d) Experimental scheme depicting nascent mRNA capture protocol. (e) Downregulated nascent CBX5 mRNA expression upon RBMX/L1 depletion in cells 2 days post transduction. Control and shRNA1: n = 6, shRNA2: n = 5 independent experiments. (f) Representative smFISH images of cells 2 days after RBMX/L1 depletion. Pre-mature and mature full-length CBX5 mRNA/cell measured by CBX5–3UTR probes and nascent transcript abundance measured by CBX5-Intron1 probes. The experiment was performed twice with similar results. (g-h) Quantitative summary of smFISH number of CBX5-3UTR foci per cell and number of CBX5-Intron1 foci per cell. Control: n = 433, shRNA1: n = 271, shRNA2: n = 362 cells. (i) Diagram of full-length RBMX and RBMX mutants used in luciferase reporter assay. (j) Diagram of vector used in luciferase reporter assay. (k) Luciferase reporter assay, using the original intron 1 or the mutated CBX5 intron 1, with different RBMX mutants. EV CBX5 and RBMX CBX5: n = 15, ΔSRR EV ΔCBX5: n = 6, RBMX ΔCBX5: n = 5, ΔRRM CBX5: n = 7, ΔRRM ΔCBX5: n = 3, ΔSRR CBX5: n = 7, ΔSRR ΔCBX5: n = 3, ΔTRR CBX5 n = 6, ΔTRR ΔCBX5 n = 7, ΔRBD CBX5 and ΔRBD ΔCBX5: n = 6, RGG mutant CBX5 and RGG mutant ΔCBX5: n = 7 independent experiments. Data presented in c, e, g-h, and k as mean +/− s.e.m. p value determined by wo-tailed Student’s t test, unless stated otherwise.

Figure 8. RBMX/L1 acts through direct regulation of CBX5 transcription.

(a) Immunoblot showing CBX5 depletion in MOLM13 cells. The experiment was performed 3 times with similar results. (b) Proliferation assay of cells from a. n = 4 independent experiments. (c) Quantitative apoptosis analysis summary of cells from a. n = 4 independent experiments. (d) Quantitative FACS summary of CD11b and CD33 of cells from a. n = 4 independent experiments. (e) Immunoblots of RBMX/L1, endogenous CBX5, and CBX5-Flag in RBMX/L1 depleted MOLM13 cells expressing CBX5-Flag. The experiment was performed 5 times with similar results. (f) Proliferation assay of cells from e. n = 5 independent experiments. (g) Quantitative apoptosis analysis summary of cells from e. n = 4 independent experiments. (h-i) Quantitative FACS summary of CD11b (h) and CD33 (i) of cells from e. n = 5 independent experiments. (j)CBX5 mRNA levels in high versus low RBMX expressing AML patients (OHSU dataset, Nature 2018) (high RBMX defined as a value > mean + 1 s.d. and low RBMX as a value < mean – 1 s.d.). High group: n = 82 patients, low group: n = 74 patients. (k) Immunoblot analysis and band densitometry of CBX5 in AML patient samples compared to CB-CD34⁺ cells, normal PB, BM, and aged BM. Same immunoblot as Fig. 3e left panel. The experiment was performed once. (l) Correlation of RBMX/L1 protein and CBX5 protein abundance quantified in Fig. 3e (right blot) and Fig. 8l. Simple linear regression, R square indicated. p value (F test) indicating the slope significantly different than zero. (m-n) Immunoblot analysis and band densitometry of RBMX/L1 and CBX5 upon RBMX/L1 depletion in PDX AML-1 (m) (same immunoblot as Extended data Fig. 4l) and PDX AML-11 (n) (same immunoblot as Extended data Fig. 4n). Each immunoblot was performed once. (o) Model for the RBMX/L1 requirement in AML. RBMX/L1 binds to CBX5 RNA and regulates the transcriptional activity of the CBX5 locus, which then maintains proper chromatin compaction in leukemia cells. Data presented in b-d and f-j as mean +/− s.e.m. p value determined by two-tailed Student’s t test, unless stated otherwise.