An Epigenetic Mechanism Underlying Chromosome 17p Deletion-Driven Tumorigenesis

Abstract

Chromosome copy-number variations are a hallmark of cancer. Among them, the prevalent chromosome 17p deletions are associated with poor prognosis and can promote tumorigenesis more than TP53 loss. Here, we use multiple functional genetic strategies and identify a new 17p tumor suppressor gene (TSG), plant homeodomain finger protein 23 (PHF23). Its deficiency impairs B-cell differentiation and promotes immature B-lymphoblastic malignancy. Mechanistically, we demonstrate that PHF23, an H3K4me3 reader, directly binds the SIN3-HDAC complex through its N-terminus and represses its deacetylation activity on H3K27ac. Thus, the PHF23-SIN3-HDAC (PSH) complex coordinates these two major active histone markers for the activation of downstream TSGs and differentiation-related genes. Furthermore, dysregulation of the PSH complex is essential for the development and maintenance of PHF23-deficient and 17p-deleted tumors. Hence, our study reveals a novel epigenetic regulatory mechanism that contributes to the pathology of 17p-deleted cancers and suggests a susceptibility in this disease. SIGNIFICANCE: We identify PHF23, encoding an H3K4me3 reader, as a new TSG on chromosome 17p, which is frequently deleted in human cancers. Mechanistically, PHF23 forms a previously unreported histone-modifying complex, the PSH complex, which regulates gene activation through a synergistic link between H3K4me3 and H3K27ac.This article is highlighted in the In This Issue feature, p. 1.

Figure 1 |. Phf23 deficiencies promoted lymphomagenesis.

A, Top, a schematic diagram of human chromosome 17, showing the relative positions of TP53 and PHF23 on 17p. Bottom, the enrichment folds of Phf23 shRNAs in the shRNA library screening of mouse chromosome 11B3 genes in Myc-driven lymphoma/leukemia in mice. The enrichment folds were calculated by dividing the percentages of each shRNA reads in tumors by those in the library of plasmids. B, The expression levels of PHF23 in DLBCL patients with (Deletion) or without (Diploid) chromosome 17p deletions. RNA-seq data with RSEM were analyzed from TCGA. *** p_adj < 0.001, Wilcoxon signed-rank test. C, Kaplan-Meier tumor-free survival curves of recipient mice transplanted with pre-B cells infected with Myc cDNA and indicated shRNAs. n=5 for each group. *p<0.05, **p<0.01 (log-rank test). D, Representative picture showing enlarged lymph nodes in recipient mice with shPhf23. Scale bar, 5mm. E, Representative flow plots showing the expressions of GFP, B220 and IgM of shPhf23 lymphoma/leukemia cells. F, Relative expression levels of Phf23 in shPhf23 lymphoma/leukemia cells, measured by qPCR. Control, shTrp53;Myc lymphoma/leukemia cells. G, Representative Western blotting picture showing the protein levels of PHF23 in shPhf23 lymphoma/leukemia cells. Control, shTrp53;Myc lymphoma/leukemia cells. H, Kaplan-Meier tumor-free survival curves of recipient mice transplanted with Eμ-Myc fetal liver cells (FLCs) with CRISPR/Cas9 targeting Phf23 or scramble sequence. n=5, sgScr; n=8, sgPhf23. *p<0.05 (log-rank test). I, Gene Set Enrichment Analysis (GSEA) showing the positive enrichment of the shPhf23 lymphoma/leukemia upregulated gene set in 17p deleted TCGA DLBCL, comparing to 17p intact ones (NES=1.75; FDR q=0.00). J, GSEA showing the negative enrichment of the shPhf23 lymphoma/leukemia downregulated gene set in 17p deleted TCGA DLBCL, comparing to 17p intact ones (NES=−1.39; FDR q=0.02).

Figure 2 |. PHF23 was a haploinsufficient TSG and cooperated with p53.

A, Diagram showing Phf23 mutation with a 14bp insertion on one allele introduced by CRISPR/Cas9 in exon 5 in mice. B, Representative picture showing genotyping of Phf23^+/+, Phf23^+/− and Phf23^−/− mice. C, Representative Western blotting picture showing the protein levels of PHF23 in Phf23^+/+, Phf23^+/− and Phf23^−/− E14.5 FLCs. D, Kaplan-Meier tumor-free survival curves of recipient mice transplanted with Myc transduced E14.5 FLCs derived from Phf23^+/+, Phf23^+/− and Phf23^−/− embryos. n=5 for each group. **p<0.01 (log-rank test). E, Representative pictures showing enlarged lymph nodes in recipient mice with Myc transduced Phf23^+/− or Phf23^−/− FLCs, but not Phf23^+/+ FLCs in Fig. 2D. Scale bar, 5mm. F, Representative flow plots showing the expressions of B220 and CD11b/Gr-1 in Phf23^+/+, Phf23^+/− and Phf23^−/− tumor cells from recipient mice in Fig. 2D. G, Western blotting of PHF23 in Phf23^+/+;Myc, Phf23^+/−;Myc and Phf23^−/−;Myc lymphoma/leukemia cells from recipient mice in Fig. 2D. (Note: Phf23^+/+;Myc lymphoma/leukemia cells had Trp53 mutations by CRISPR/Cas9). H, Kaplan-Meier tumor-free survival curves of recipient mice transplanted with pre-B cells infected with Myc-linked tandem shRNAs. n=7, ***p<0.001 (log-rank test). I, Kaplan-Meier tumor-free survival curves of recipient mice transplanted with p53^+/− pre-B cells infected with Myc-linked shRen or shPhf23. *p<0.05 (log-rank test). J, GSEA showing the negative enrichment of the HALLMARK_P53_PATHWAY gene set in shPhf23 lymphoma/leukemia cells, comparing to shRen cells (NES=−1.39; FDR q=0.03).

Figure 3 |. PHF23 co-located with H3K4me3 in pre-B cells and was required for B cell differentiation.

A, Top, profiles showing average H3K4me3 (left) and PHF23 (right) bound peak densities in TSS regions by ChIP-seq analyses. Bottom, heatmaps showing H3K4me3 and PHF23 bound peaks in TSS regions. B, Venn diagram showing overlapping of PHF23 and H3K4me3 bound genes; p, Fisher’s exact test. C, Mean log₂ TPM (Transcripts Per Kilobase Million) of genes with no binding (without H3K4me3 and PHF23 binding), H3K4me3 only, PHF23 only or co-binding (both have H3K4me3 and PHF23 binding) in Ba/F3 cells. p_adj, Wilcoxon signed-rank test. D, Heatmap showing the differentially expressed genes (DEGs) (p_adj<0.05, log₂ Fold change>1 or <−1) in shPhf23 pre-B cells compared to those with shRen. E, GSEA showing the negative enrichments of the HEMATOPOIETIC_STEM_CELL_DOWN (NES=−1.99; FDR q=0.00) and B_CELL_VS_MYELOID_DOWN gene sets (NES=−2.34; FDR q=0.00) in shPhf23 pre-B cells, comparing to shRen cells. F, Integrative Genomics Viewer (IGV) showing binding peaks of H3K4me3 and PHF23 on Cebpb (top) and Hk3 (bottom). G, Schematic diagram of the hematopoiesis analyses of Phf23^+/+ and Phf23^−/− FLCs. H, Percentages of donor-type LSK (Lin⁻ScaI⁺cKit⁺) and CLP (Lin⁻Sca1^lowcKit^low) populations in the bone marrow of recipient mice 4 months after BMT. I, Percentages of donor-type B220⁺CD43⁺ populations in the bone marrow of recipient mice. J, Percentages of donor-type B220⁺CD43⁺CD24⁻ (pre-pro B) populations in the bone marrow of recipient mice. K, Percentages of donor-type B220⁺CD43⁺ CD24⁺ (pro-B) populations in the bone marrow of recipient mice. H-K, *p<0.05, **p<0.01, ns, not significant (t-test).

Figure 4 |. PHF23 directly bound and negatively regulated the SIN3-HDAC complex.

A, Schematic diagram of the PHF23-SIN3-HDAC (PSH) complex. Colors indicated the Mascot scores of each component measured by Co-IP of FLAG-PHF23 and LC/MS. B, Representative Western blotting pictures showing FLAG-PHF23, SIN3A and HDAC1 of input and FLAG-PHF23 Co-IP. C, Representative Western blotting pictures of endogenous PHF23, SIN3A and HDAC1 of input and Co-IP with IgG or PHF23 antibodies. D, Representative Western blotting pictures of FLAG-PHF23, SIN3A and HDAC1 of input and FLAG tagged full length or truncated PHF23 Co-IP. E, Representative Western blotting pictures of PHF23 and H3K27ac in Phf23^+/+ and Phf23^+/− FLCs. F, Venn diagrams showing overlapping of PHF23 and HDAC1 bound genes, analyzed by ChIP-seq. p, Fisher’s exact test. Left, HDAC1 bound at TSS; Right, HDAC1 bound at enhancer regions. G, Mean log₂ H3K27ac levels on genes with no binding (without HDAC1 TSS and PHF23 binding), HDAC1_T (only HDAC1 at TSS), PHF23 only or co-binding (both HDAC1 and PHF23 binding at TSS) in Ba/F3 cells. p_adj, Wilcoxon signed-rank test. H, Mean log₂ H3K27ac levels on genes with no binding (without HDAC1 enhancer and PHF23 binding), HDAC1_E (only HDAC1 at enhancer regions), PHF23 only or co-binding (both HDAC1_E and PHF23 binding). p_adj, Wilcoxon signed-rank test. I, Mean of log₂ TPM of genes with no binding (without HDAC1 TSS and PHF23 binding), HDAC1_T (only HDAC1 at TSS), PHF23 only or co-binding (both HDAC1 and PHF23 binding at TSS) in immature B lymphoblastic malignancy cells. p_adj, Wilcoxon signed-rank test. J, Mean of log₂ TPM of genes with no binding (without HDAC1 enhancer and PHF23 binding), HDAC1_E (only HDAC1 at enhancer regions), PHF23 only or co-binding (both HDAC1_E and PHF23 binding). K, GSEA showing the negative enrichment of the SENESE_HDAC1_TARGETS_UP gene set in shPhf23 pre-B cells, comparing to shRen cells (NES=−1.33; FDR q=0.00). L, Top 5 KEGG pathways enriched in genes co-bound by both PHF23 and HDAC1. M, Venn diagram showing overlapping of the PHF23 bound genes and the p53 pathway genes. p, Fisher’s exact test. N, Percentages of genes with H3K4me3 binding in p53 pathway genes with PHF23 binding or p53 pathway genes without PHF23 binding. O, Percentages of genes with H3K27ac binding in p53 pathway genes with PHF23 binding or p53 pathway genes without PHF23 binding. P, IGV plots showing PHF23, HDAC1, H3K27ac, and H3K4me3 binding density on Cdkn1a.

Figure 5 |. The PSH complex dysregulation was critical for tumorigenesis and maintenance.

A, Knockdown efficiencies of shSin3a, measured by qPCR. B, Kaplan-Meier tumor-free survival curves of recipient mice transplanted with Phf23^+/− FLCs with Myc-linked shSin3a or shRen. n=4. *p<0.05, **p<0.01 (log-rank test). C, Knockdown efficiencies of shSin3b measured by qPCR. D, Kaplan-Meier tumor-free survival curves of recipient mice transplanted with Phf23^+/− FLCs infected with Myc-linked shSin3b or shRen. n=8. ***p<0.001, ****p< 0.0001 (log-rank test). E, Relative cell viabilities of shPhf23 lymphoma/leukemia cells with shSin3a. n=3. F, Relative cell viabilities of shPhf23 lymphoma/leukemia cells with shSin3b. n=3. G, Relative cell viabilities of shPhf23 lymphoma/leukemia cells with shHdac1. n=3. H, Relative cell viabilities of Ba/F3 cells and shPhf23 lymphoma/leukemia cells treated with HDAC inhibitors (entinostat, 0.5μM; mocetinostat, 0.1 μM; chidamide, 0.5 μM). n=3. I, GSEA showing the negative enrichment of the shPhf23 upregulated gene set in chidamide treated shPhf23 lymphoma/leukemia cells, comparing to DMSO treated cells (NES=−1.49; FDR q=0.00). J, GSEA showing the positive enrichment of the shPhf23 downregulated gene set in chidamide treated shPhf23 lymphoma/leukemia cells, comparing to DMSO treated cells (NES=2.05; FDR q=0.00). K, Relative cell viabilities of Phf23^+/−;Myc lymphoma/leukemia cells transduced with full-length or truncated Phf23. n=3. L, Relative cell viabilities of chromosome 11B3 deleted lymphoma/leukemia cells transduced with full-length or truncated Phf23. n=3. M, Relative cell viabilities of 17p intact Ri-1 human lymphoma cells transduced with full-length or truncated PHF23. n=3. N, Relative cell viabilities of 17p intact Namalwa human lymphoma cells transduced with full-length or truncated PHF23. n=3. O, Relative cell viabilities of 17p deleted Granta-519 human lymphoma cells transduced with full-length or truncated PHF23. n=3. P, Relative cell viabilities of 17p deleted JeKo-1 human lymphoma cells transduced with full-length or truncated PHF23. n=3. A, C, E-H and K-P, *p<0.05, **p<0.01, ***p<0.001, ns, not significant (unpaired two-tailed t-test).

Figure 6 |. Schematic diagram showing the working model for the PSH complex in normal and tumor cells.

A, Schematic diagram showing the working models of the PSH complex in B cell differentiation and TSG regulation. B, Schematic diagram showing the working models of the PSH complex in tumor cells with PHF23 deficiency.