The ubiquitin ligase RNF5 determines acute myeloid leukemia growth and susceptibility to histone deacetylase inhibitors

Abstract

Acute myeloid leukemia (AML) remains incurable, largely due to its resistance to conventional treatments. Here, we find that increased abundance of the ubiquitin ligase RNF5 contributes to AML development and survival. High RNF5 expression in AML patient specimens correlates with poor prognosis. RNF5 inhibition decreases AML cell growth in culture, in patient-derived xenograft (PDX) samples and in vivo, and delays development of MLL-AF9-driven leukemogenesis in mice, prolonging their survival. RNF5 inhibition causes transcriptional changes that overlap with those seen upon histone deacetylase (HDAC)1 inhibition. RNF5 induces the formation of K29 ubiquitin chains on the histone-binding protein RBBP4, promoting its recruitment to and subsequent epigenetic regulation of genes involved in AML maintenance. Correspondingly, RNF5 or RBBP4 knockdown enhances AML cell sensitivity to HDAC inhibitors. Notably, low expression of both RNF5 and HDAC coincides with a favorable prognosis. Our studies identify an ERAD-independent role for RNF5, demonstrating that its control of RBBP4 constitutes an epigenetic pathway that drives AML, and highlight RNF5/RBBP4 as markers useful to stratify patients for treatment with HDAC inhibitors.

Fig. 1. RNF5 expression in AML cell lines and patient samples.

aRNF5 expression data obtained from the CCLE RNA-seq datasets. Transcripts per million (TPM) for protein-coding genes were calculated using RSEM software. Data is log2-transformed, using a pseudo-count of 1w. Box plot is sorted and colored by the average distribution of RNF5 expression in a lineage. Lineages are composed of a number of cell lines. The highest average distributions are shown at left in red. The line within a box represents the mean. b Representative western blot analysis of RNF5 in peripheral blood mononuclear cells (PBMCs) from healthy donors and AML patients (Scripps Health). c Relative abundance of RNF5 protein in PBMCs from AML patients (n = 50) and healthy control subjects (n = 6) from the Scripps Health Center. RNF5 abundance in U937 cells served as a reference for quantification (presented as the mean ± SEM). P = 0.008 by unpaired two-tailed t-test. d Kaplan–Meier survival curve analysis of AML patients stratified by top high (n = 8) versus low (n = 42) RNF5 protein (Scripps Health). P = 0.05 by two-sided Mantel–Cox log-rank test. e Kaplan–Meier survival curve of AML patients stratified by top high (n = 14) versus low (n = 150) RNF5 transcript levels (TCGA dataset). P = 0.009 by tow-sided Mantel–Cox log-rank test. f, h Levels of RNF5 protein and actin in PBMCs from healthy controls and AML patients (Rambam Health Campus Center). D diagnosis, PI post induction, RLP relapse, RMS remission. Quantified data are presented as the mean ± SEM. P = 0.006 by unpaired two-tailed t-test. g Relative abundance of RNF5 protein in PBMCs from AML patients (n = 18) and healthy controls (n = 5) from the Rambam Center. Quantified data are presented as the mean ± SEM. P = 0.006 by unpaired two-tailed t-test. i, j Relative abundance of RNF5 protein in AML patient samples (PBMCs) collected before and after induction treatment (i, n = 8) or before and after relapse (j, n = 5). Lines connect values for the same patient. P = 0.055, P = 0.0074 by paired two-tailed t-test. Source data are provided as a Source Data file.

Fig. 2. RNF5 is required for AML cell proliferation and survival.

a Growth assay of MOLM-13 and U937 cells transduced with empty vector (pLKO) or two different shRNF5 constructs. Cell growth was analyzed using CellTiter-Glo assay. b Cell cycle analysis of MOLM-13 and U937 cell lines 5 days after transduction. c Western blot analysis of cell cycle regulatory proteins in MOLM-13 and U937 AML cells 5 days after transduction. d Representative images (left) and quantification (right) of colonies in soft agar from MOLM-13 and U937 were assessed after 15 days in culture. e Western blot analysis of apoptosis-related proteins in indicated lines 5 days after transduction (C.casp.3 = cleaved caspase-3). f Growth assay of U937 cells expressing doxycycline-inducible Flag-tagged RNF5 WT, RM or empty vector (EV). Cells were induced 48 h with doxycycline (1 µg/ml) and then transduced with either empty vector (pLKO) or shRNF5 for 5 days. g Western blot analysis of indicated proteins in U937 transduced as described in (f). h Growth assay of MOLM-13 cells stably expressing Cas9 and transduced with control Renilla-targeting sgRNA or two RNF5-targeting sgRNAs. CRISPR was performed based on CRISPR knockout cell pool. i Western blot analysis of indicated proteins in MOLM-13 cells described in (h). j Growth assay of PDX AML-669 cells after transduction with empty vector (pLKO) or two independent shRNF5 constructs. Quantified data are presented as the mean ± SD of two independent experiments. (k, l) RT-qPCR (k) and western blot (l) analyses confirming RNF5 KD in PDX AML-669 transduced as described in (j). Quantified data are presented as the mean ± SD (a, f, and h) or SEM (b and d) of n = 5 (a (left)), n = 6 (a (right), b (left)), or n = 4 (b (right), d, f, and h) independent experiments. Western blot data are representative of three experiments. P values were determined using two-way ANOVA followed by Tukey’s multiple comparison test (a, f, and h) or paired two-tailed t-test (b, and d). Source data are provided as a Source Data file.

Fig. 3. RNF5 inhibition sensitizes AML cells to ER stress-induced apoptosis.

a, b Western blot analysis of indicated proteins in MOLM-13 cells expressing empty vector (pLKO) or shRNF5 #1 and treated with thapsigargin (TG, 1 µM) (a) or tunicamycin (TM, 2 µg/ml) (b) for indicated times. c Luminescence-based viability assay of MOLM-13 cells expressing inducible shRNF5 and treated with or without doxycycline (DOX, 1 µg/ml) for 3 days before treatment with TG (100 nM) for indicated times. d RT-qPCR analysis of CHOP and ATF3 mRNA in MOLM-13 cells expressing pLKO or shRNF5 #1 and treated with TG (100 nM) for indicated times. e Western blot analysis of cleaved caspase-3 (C.casp.3) and PARP in MOLM-13 cells expressing pLKO or shRNF5 #1 and treated with bortezomib (BTZ, 5 nM) for indicated times. f Fluorescence-based viability assay of HL-60 cells expressing pLKO or shRNF5 and treated with BTZ (5 nM) for indicated times. Cell viability was determined by flow cytometry of cells stained with annexin V conjugated to fluorescein isothiocyanate and propidium iodide. g Luminescence viability assay of HL-60 cells expressing pLKO or shRNF5 and treated 48 h with indicated concentrations of BTZ. Data are presented as the mean ± SD (c, f, and g) or SEM (d) of n = 3 (c and g), n = 5 (d (left)), or n = 4 (d (right) and f) independent experiments. P values were determined using paired two-tailed t-test (c, d, and f) or two-way ANOVA (g). ns: not significant. Source data are provided as a Source Data file.

Fig. 4. RNF5 suppression impairs leukemia establishment and progression in vivo.

a Graph depicting growth in mice of luciferase-expressing U937-pGFL cells transduced with empty vector (pLKO) or inducible shRNF5. Bioluminescence was quantified to monitor disease burden. Data are presented as the mean ± SD of 7 mice per group. P values were determined using unpaired two-tailed t-test. b Kaplan–Meier survival curves of mice injected with U937-pGFL cells expressing pLKO (n = 7 mice/group) or inducible shRNF5 (n = 7 mice/group). P < 0.001 by two-sided Mantel–Cox log-rank test. c Schematic representation of the experiment. Lin^–Sca1⁺c-Kit⁺ (LSK) cells were purified from the bone marrow of WT or Rnf5^−/− mice, transduced in vitro with a GFP-tagged MLL-AF9 fusion gene, and then either analyzed by colony-forming assays in vitro or intravenously injected into sub-lethally irradiated WT C57BL/6 mice. d Quantification of total colonies (left) or blast-like and differentiated colonies (right) of GFP-MLL-AF9–transformed WT or Rnf5^−/− cells after 7, 14, or 21 days in culture. Data are presented as the mean ± SD of three independent experiments. P values were determined using a paired two-tailed t-test. e Representative pictures of colonies of GFP-MLL-AF9–transformed WT or Rnf5^−/− cells after 7 days in culture. Scale bar 200 µm. f Wright–Giemsa-staining of GFP-MLL-AF9–transformed WT or Rnf5^−/− cells after 7 days in culture. Scale bar 25 µm. g Flow cytometric quantification of GFP+ cells in peripheral blood of mice intravenously injected with GFP-MLL-AF9–transformed WT (n = 4) or Rnf5^−/− (n = 4) cells at days 15 and 28 post-injection. Data are presented as the mean ± SD. P = 0.002 by unpaired two-tailed t-test. The gating strategy is provided in Supplementary Fig. 8b. h Kaplan–Meier survival curves of mice injected with GFP-MLL-AF9–transformed WT or Rnf5^−/− cells. Data are from two independent experiments (n = 4 mice/group per experiment). P < 0.001 by two-sided Mantel–Cox log-rank test. Source data are provided as a Source Data file.

Fig. 5. Transcriptional and survival analysis of RNF5- or RBBP4-deficient AML cells.

a Venn diagram analysis of RNA-seq results showing upregulated (red) and downregulated (green) genes in AML lines following RNF5 KD. Overlapping areas indicate commonly modulated genes. b Heatmap of RNA-seq data performed on pLKO or shRNF5 AML lines, as indicated beneath maps (see Methods). c RT-qPCR validation of genes deregulated by RNF5-KD. Data are presented as the mean ± SD of n = 5 or 6 (ANXA1) or n = 3 (NCF1 and CDKN1A) independent experiments. d Top ten drug screening results from the LINCS matched with transcriptomic data from the shRNF5 MOLM-13 line. Values are overall z-scores from the IPA Analysis Match database. HDAC1 inhibitor results are shown in red. e shRNF5 transduction of MOLM-13 or HL-60 promotes similar changes to those seen in NPC, HEPG2, A549, and PC3 cancer cell lines treated with the HDAC1 inhibitor mocetinostat. Z-score was calculated by IPA: A positive z-score predicts pathway activation; a negative z-score predicts inhibition. f Results of LC-MS/MS analysis present log₂-transformed ratio of proteins in anti-Flag immunoprecipitates of RNF5-overexpressing versus control cells. Green, proteins significantly enriched in RNF5-overexpressing cells; blue, enriched proteins in the ERAD pathway. RNF5 and RBBP4 are indicated in red. g Co-expression of RBBP4 and indicated RNF5 target genes in AML samples analyzed in cBioPortal using the TCGA database (Pearson correlation, P < 0.0001, n = 165). h WB of RBBP4 in PBMCs from healthy subjects and AML patients (Scripps Health). i Overall survival rate (performed using GEPIA and TCGA) of AML patients expressing high (30%) or low (70%) levels of RBBP4 transcripts. TPM: Transcripts Per Million. HR: hazard ratio. j Growth assay of MOLM-13 cells after transduction with indicated constructs. Data are presented as the mean ± SD of n = 3 independent experiments. k WB analysis of indicated proteins in MOLM-13 cells expressing indicated constructs. l RT-qPCR analysis of genes deregulated by RNF5-KD in MOLM-13 cells expressing the indicated constructs. Data are presented as the mean ± SD of n = 5 (RBBP4), n = 4 (NCF1 and CDKN1A), or n = 3 (ANXA1) independent experiments. m Bioluminescent images of representative mice 4 weeks following transplantation of U937-pGFL expressing the indicated constructs. n Kaplan–Meier survival curves of mice injected with U937-pGFL cells expressing indicated vectors. P = 0.02 and P = 0.001 by two-sided Mantel–Cox log-rank test. P values were determined using paired two-tailed t-test (c and l) or two-way ANOVA followed by Tukey’s multiple comparison test (j). Source data are provided as a Source Data file.

Fig. 6. RBBP4 is ubiquitinated by RNF5 and regulates RNF5 target genes.

a Schematic showing full-length and mutants forms of RNF5. b Immunoprecipitation (IP) and Western blot (WB) analysis of HEK293T cells transfected with Flag-tagged forms of full-length RNF5 (WT), the catalytically inactive RING domain mutant (RM), or the C-terminal transmembrane domain deletion mutant (∆CT). Cells were treated with MG132 (10 µm, 4 h) before lysis. c IP and WB of HEK293T cells co-expressing Myc-RBBP4 and Flag-tagged RNF5-WT treated with MG132 (10 µm, 4 h) before lysis. d IP and WB of ectopically expressed doxycycline-inducible Flag-tagged RNF5 and endogenous RBBP4 in MOLM-13 cells. Cells were incubated for 2 days with or without doxycycline (1 µg/ml) and with MG132 (10 µm, 4 h) before lysis. e WB of anti-Myc IP from lysates of HEK293T cells co-expressing Myc-RBBP4, hemagglutinin-tagged ubiquitin (HA-Ub), and indicated Flag-tagged RNF5 constructs. Cells were treated with MG132 (10 µm, 4 h) before lysis. f WB of indicated proteins in MOLM-13 cells expressing empty vector (pLKO) or indicated shRNF5 construct. g WB analysis of indicated proteins in MOLM-13 cells expressing empty vector or doxycycline-inducible Flag-tagged RNF5. h WB of anti-Myc IP and lysates of HEK293T cells co-expressing Myc-RBBP4, Flag-tagged RNF5, and different HA-tagged ubiquitin mutants (K29, K11, K6, K27, and K33). MG132 (10 µm, 4 h) was added before lysis. i IP and WB for the interaction of RBBP4 with HDAC1, HDAC2, or EZH2 in MOLM-13 cells expressing indicated constructs. j IP and WB for RBBP4 interaction with HDAC1, HDAC2, or EZH2 in MOLM-13 cells expressing indicated constructs. MG132 (10 µm, 4 h) was added before lysis. k ChIP and qPCR reveal the enrichment of RBBP4 (normalized to input) at indicated gene promoters in MOLM-13 cells expressing indicated constructs. l–n ChIP and qPCR reveal the enrichment of H3K9ac, H3K27ac, or H3K27me3 (normalized to input) at indicated gene promoters in MOLM-13 cells expressing indicated constructs. Data in k and l are presented as mean ± SEM of n = 4 (ANXA1) or n = 3 (NCF1 and CDKN1A) independent experiment. Data in m and n are the mean of n = 2 independent experiments. The P values were determined using paired two-tailed t-test (k and l). Source data are provided as a Source Data file.

Fig. 7. RNF5 inhibition sensitizes AML cells to HDAC inhibitors.

a Schematic showing the experimental design of the epigenetic screen. b Log₂-transformed ratios of the relative viability of doxycycline-induced (+Dox) versus uninduced (-Dox) U937 cells treated with compounds for 6 days. Red dots represent compounds that altered viability of RNF5-KD more than of uninduced cells, blue dots represent candidate HDAC inhibitors, and grey dots represent the remaining compounds tested. c Viability of U937 cells expressing indicated constructs after treatment for 24 h with CI-994. d U937 cell viability after treatment with 3.5 nM FK228. e Viability of U937 cells or MOLM-13 cells expressing indicated constructs 24 h following FK228 treatment. f WB of apoptotic markers in MOLM-13 and U937 cells expressing indicated constructs and incubated with or without FK228 (4 nM for 24 h). g Viability of U937 cells expressing indicated constructs and treated for 24 h with FK228. EV-pLKO, control cells; EV-shRNF5, cells expressing empty vector and shRNF5; RNF5-pLKO, cells overexpressing RNF5 and pLKO vector; RNF5-shRNF5, cells overexpressing RNF5 and shRNF5. h Viability of MOLM-13 cells expressing indicated constructs 24 h following FK228 treatment. i ChIP and qPCR indicating H3K9ac enrichment (normalized to input) at indicated gene promoters in MOLM-13 cells expressing indicated constructs. Data are presented as the mean ± SD of two independent experiments. j RT-qPCR analysis of indicated genes in MOLM-13 cells expressing indicated constructs following FK228 treatment (4 nM, 15 h). k Viability assay of 4 primary AML blasts (Scripps Health) 48 h following FK228 treatment. l WB analysis of RNF5 and RBBP4 in AML patient samples used for the ex-vivo drug analysis in (l). m WB quantification of RNF5 protein levels in (l) normalized to actin. n Kaplan–Meier plot showing survival analysis of AML patients segregated based on a median synthetic lethality (SL) score. Co-occurrence of low HDAC and RNF5 transcript levels in a patient’s tumor (high SL score; blue line), compared with the rest of the patients (low score, yellow line). Data are presented as the mean ± SD of n = 4 (c and d), n = 3 (e, g, and j), or n = 5 (h) experiments. P values were determined using paired two-tailed t-test (d and j) or two-way ANOVA followed by Tukey’s multiple comparison test (c, e, g, and h). Source data are provided as a Source Data file.