JMJD3-mediated senescence is required to overcome stress-induced hematopoietic defects

Abstract

Cellular senescence in stem cells compromises regenerative capacity, promotes chronic inflammation, and is implicated in aging. Hematopoietic stem and progenitor cells (HSPCs) are responsible for producing mature blood cells, however, how cellular senescence influences their function is largely unknown. Here, we show that JMJD3, a histone demethylase, activates cellular senescence by upregulating p16^Ink4a in competition with Polycomb group proteins, and reprograms HSPC integrity to overcome hematopoietic defects induced by replicative and oncogenic stresses. Jmjd3 deficiency does not alter global H3K27me3 levels, indicating that JMJD3 epigenetically regulates specific and limited JMJD3 targets under stress. JMJD3 deficiency also impairs stem cell potential, proper cell cycle regulation, and WNT pathway activation in HSPCs under stress. These impaired phenotypes are rescued through exogenous and retroviral introduction of p16^Ink4a. This JMJD3-p16^INK4a axis in hematopoiesis is age-dependent and is distinct from cellular senescence. Treatment with a selective JMJD3 inhibitor attenuates leukemic potential during cellular senescence. Taken together, these results demonstrate that JMJD3-p16^INK4a mediates cellular senescence and plays critical roles in the functional integrity of HSPCs under stress.

Keywords: Cellular senescence; Hematopoietic stem cell; Histone demethylase; Stress hematopoiesis.

Figure 1. Analysis of Jmjd3^Δ/Δ HSPCs under replicative stress induced by BMT.

(A) Schematic diagram of the serial competitive bone marrow (BM) reconstitution experiment. 2.0 × 10³ LSK cells from Jmjd3^+/+ and Jmjd3^Δ/Δ mice were transplanted into lethally irradiated primary recipients with 2.5 × 10⁵ wild-type competitor mononuclear BM (MNBM) cells. 3.0 × 10⁶ MNBM cells from the first BMT recipients were transplanted in lethally irradiated secondary recipients, and 3.0 × 10⁶ MNBM cells from second recipients were subjected to transplantation into tertiary recipients and colony formation assays. (B) Chimerism and lineage differentiation of donor-derived cells in the peripheral blood (PB) of the first recipients (left and middle left panels, n = 10 each) and chimerism of donor-derived cells in BM subfractions including LSK, LS⁻K (Lin⁻, Sca-1⁻, c-kit⁺), Lin⁻, and Lin⁺ of first recipients at 4 months after BMT (middle right panel) (mean ± SD, n = 6). Flow cytometric profiles of donor-derived Lin⁻ cells in the BM of the first BMT recipients 4 months after BMT (right panel) (mean + SD, n = 5). Student’s t test was used to calculate p value. (C) The same analyses in the second recipients (mean ± SD, n = 10). Student’s t test was used to calculate p value. (D) Chimerism of donor-derived cells in the PB of third recipients (mean ± SD, n = 10) and colony formation assays of LSK cells from the third recipients at 4 months (mean + SD, n = 3). Student’s t test was used to calculate p values. Source data are available online for this figure.

Figure 2. Analysis of Jmjd3^Δ/Δ HSPCs under oncogenic stress.

(A) Schematic diagram of MLL-AF9 oncogene transduction. LK cells from Jmjd3^+/+ and Jmjd3^Δ/Δ mice were transfected with MLL-AF9-IRES-EGFP retrovirus, and EGFP⁺ cells were subjected to the following assays. (B) Colony replating assays. Bars indicate colony number (CFU; colony-forming unit) in each round of plating (mean + SD, n = 3). Student’s t test was used to calculate p value. (C) Flow cytometric profiles of colony-forming cells. Cells were stained with c-kit and lineage markers, and the percentages of c-kit⁺, Lin⁻ cells in each round are shown. (D) Kaplan–Meier survival curves of mice transplanted with MA9 expressing (MA9) Jmjd3^+/+ and Jmjd3^Δ/Δ cells. 4.0 × 10³ (4 K) or 5.0 × 10² (0.5 K) cells were transplanted into lethally irradiated recipients with 2.5 × 10⁵ wild-type competitor MNBM cells (n = 6). A log-rank test was used to calculate p value. (E) BM cells from mice that developed MLL-AF9 induced leukemia were stained with EGFP and lineage markers. The percentage of EGFP⁺, Lin⁻ cells is shown (mean + SD, n = 5). Student’s t test was used to calculate p value. (F) Colony replating assay of Jmjd3^+/+ and Jmjd3^Δ/Δ L-GMPs at the third round of replating (mean + SD, n = 3). Student’s t test was used to calculate p value. (G) Kaplan–Meier survival curves of mice transplanted with Jmjd3^+/+ and Jmjd3^Δ/Δ L-GMPs. 2.0 × 10² L-GMPs were transplanted into lethally irradiated recipients with 2.5 × 10⁵ wild-type competitor MNBM cells for radioprotection (n = 5). A log-rank test was used to calculate p values. Source data are available online for this figure.

Figure 3. Demethylase-dependent regulation at the Cdkn2a locus in Jmjd3^+/+ and Jmjd3^Δ/Δ HPSCs.

(A) qPCR of CDK inhibitor (CDKI) genes in LSK (Steady), LSK (BMT), and L-GMP. Relative fold-changes to LSK (Steady) are shown on a logarithmic scale (mean + SD, n = 3). (B) qPCR of CDKI genes in LSK (Steady), LSK (BMT), and L-GMP of Jmjd3^+/+ and Jmjd3^Δ/Δ mice (mean + SD, n = 3). (C) Immunofluorescence staining (left panels) and relative fluorescence intensity (right panels) of H3K27me3 in LSK (Steady) (n = 175), LSK (BMT) (n = 117), and L-GMP (n = 144) from Jmjd3^+/+ and Jmjd3^Δ/Δ mice. Mean values are indicated as bars. Student’s t test was used to calculate p value. Scale bar, 10 μm. (D) Schematic diagram of the Cdkn2a locus. Promoter regions of p19^Arf and p16^Ink4a are indicated #1–4. (upper panel). H3K27me3 levels in the promoter regions of p19^Arf and p16^Ink4a genes in LSK (Steady), LSK (BMT), and L-GMP of Jmjd3^+/+ and Jmjd3^Δ/Δ mice. Results are shown as fold changes relative to the negative control (Neg ctrl), (mean + SD, n = 3) (lower panel). (E) Box plot showing the global levels of H3K27me3 in LSK (BMT) and L-GMPs from Jmjd3^+/+ and Jmjd3^Δ/Δ mice. Student’s t test was used to calculate p value. (F) Venn diagram showing the distribution of broad H3K27me3 peaks in LSK (BMT) and L-GMPs from Jmjd3^+/+ and Jmjd3^Δ/Δ mice (upper panel) and enrichment analysis from unique H3K27me3 peaks Jmjd3^Δ/Δ L-GMPs with Enrichr on the ENCODE/ChEA database (lower). (G) Genome browser view of H3K27me3 enrichment at Cdkn2a and an untargeted locus in LSK (BMT) and L-GMPs from Jmjd3^+/+ and Jmjd3^Δ/Δ mice. (H) qPCR analysis of Jmjd3 (left panel), p16^Ink4a (middle panel), and Bmi1 (right panel) in LK cells transduced with MLL-AF9. Expression levels are shown relative to day 1 (mean ± SD, n = 3). SDs were calculated with technical duplicates. (I) Generation of Jmjd3-Flag KI mice. In all, 3× Flag sequences were inserted upstream of the stop codon Jmjd3. (J) Immunoblot showing Jmjd3-Flag3 in BM cells of wild-type (WT) and Jmjd3-Flag KI mice (#1 and #2). (K) ChIP-qPCR analysis for the enrichment of Jmjd3 and Bmi1 at the indicated Cdkn2a promoter regions in LK cells transduced with MLL-AF9 from Jmjd3-Flag cKI mice (mean ± SD, n = 3) (see Fig. 3D). Source data are available online for this figure.

Figure 4. Jmjd3 deficiency attenuates cellular senescence under stress.

(A) β-galactosidase staining (left panel) and percentage of β-galactosidase⁺ cells in Jmjd3^+/+ and Jmjd3^Δ/Δ LSK (Steady) (n = 5), LSK (BMT) (n = 8) and L-GMPs (n = 8) treated with adriamycin (ADR) (right panel). Mean values are indicated as bars. Student’s t test was used to calculate p value. Scale bar, 10 μm. (B) Flow cytometric analysis of BrdU incorporation in LSK (Steady) (n = 4), LSK (BMT) (n = 3), and L-GMP (n = 3) of Jmjd3^+/+ and Jmjd3^Δ/Δ mice (mean + SD). Student’s t test was used to calculate p value. (C) Flow cytometric analysis of Annexin V in LSK (Steady) (n = 3), LSK (BMT) (n = 4), and L-GMP (n = 4) of Jmjd3^+/+ and Jmjd3^Δ/Δ mice (mean + SD). Student’s t test was used to calculate p values. Source data are available online for this figure.

Figure 5. Expression and pathway analysis of senescence-associated genes.

(A) Scatter plots comparing normalized expression of individual genes (RPKM > 1) in LSK (Steady), LSK (BMT), and L-GMP of Jmjd3^+/+ and Jmjd3^Δ/Δ mice. Numbers and percent of genes upregulated or downregulated more than twofold in Jmjd3^Δ/Δ cells compared with Jmjd3^+/+ cells are shown as red and blue dots, respectively. (B) The top five most positively and negatively enriched KEGG pathways in LSK (Steady) (upper), LSK (BMT) (middle), and L-GMP (lower) from Jmjd3^Δ/Δ mice compared with Jmjd3^+/+ mice (FDR < 0.25). Common upregulated and downregulated pathways between LSK (BMT) and L-GMP are shown as red and blue bars, respectively. (C) GSEA plots of LSK (Steady), LSK (BMT), and L-GMP in the indicated gene sets (top, genes commonly upregulated in human HSCs and LSCs; bottom, genes commonly upregulated in quiescent human CD34⁺ hematopoietic cells). Enrichment in Jmjd3^Δ/Δ cells relative to Jmjd3^+/+ are shown with NES and FDR values. (D) GSEA plots of LSK (Steady), LSK (BMT), and L-GMP in genes commonly upregulated through a p16^INK4a/RB1 pathway. Results are shown with NES and FDR values. (E) GSEA plots of WNT related pathways (KEGG_WNT_SIGNALING_PATHWAY (upper) and PID_BETA_CATENIN_NUC_PATHWAY (lower) in LSK (Steady), LSK (BMT), and L-GMP. Jmjd3^Δ/Δ cells compared with Jmjd3^+/+ are shown with NES and FDR values.

Figure 6. Rescue of stress-induced defects in Jmjd3^Δ/Δ HSPCs by exogenous and retroviral expression of p16^Ink4a.

(A) Schematic diagram of the p16^INK4a-TAT fusion protein (p16-TAT, left panel). (B) Immunoblot showing time-dependent stability of p16-TAT (50 nM) in cultured LK cells. (C) Flow cytometric profiles and percentages of BrdU incorporation in cultured LSK cells treated with BSA or p16-TAT (50 nM) (mean + SD, n = 3). Student’s t test was used to calculate p value. (D) Schematic of experimental procedure. Jmjd3^+/+ and Jmjd3^Δ/Δ LK cells treated with BSA or p16-TAT (50 nM) in a cytokine cocktail were subjected to colony forming and BMT assays. (E) Colony formation assay of Jmjd3^+/+ and Jmjd3^Δ/Δ LK cells treated with BSA or p16-TAT for 4 days (mean + SD, n = 3). Dunnett’s test was used to calculate p value. (F) Donor-derived chimerism in the PB of recipient mice. 5.0 × 10⁴ Jmjd3^+/+ or Jmjd3^Δ/Δ LK cells treated with BSA or p16-TAT for 4 days were transplanted into lethally irradiated recipients with 2.5 × 10⁵ wild-type competitor MNBM cells for radioprotection (mean + SD, n = 4). Dunnett’s test was used to calculate p value. (G) Percent of donor-derived, lineage committed (Thy1.2⁺, B220⁺, or Mac-1⁺) cells in the PB of recipient mice 4 months after BMT (mean + SD, n = 4). Dunnett’s test was used to calculate p value. (H) Percent of donor-derived LSK, LS⁻K, Lin⁻, and Lin⁺ cells in the BM of recipient mice at 4 months after BMT (mean + SD, n = 4). (I) Flow cytometric profiles and percentages of Lin⁻ cells in the donor-derived BM of recipient mice 4 months after BMT (mean + SD, n = 4). Dunnett’s test was used to calculate p value. (J) Schematic diagram of MLL-AF9 and p16^Ink4a co-transduction. LK cells from Jmjd3^+/+ and Jmjd3^Δ/Δ mice were transfected with the MLL-AF9-IRES-EGFP retrovirus. EGFP⁺ cells were further transfected with empty- or p16^Ink4a-IRES-KO (Kusabira Orange) retrovirus, and double-positive cells were subjected to the following assays. (K) Flow cytometric profiles of Lin⁻ cells in MLL-AF9 Jmjd3^+/+ and Jmjd3^Δ/Δ leukemic cells transfected with empty or p16^Ink4a (mean + SD, n = 3). Dunnett’s test was used to calculate p value. (L) Colony forming assay of Jmjd3^+/+ and Jmjd3^Δ/Δ L-GMPs transfected with empty or p16^Ink4a. Bars indicate the colony numbers at the third round of replating (mean + SD, n = 3). Dunnett’s test was used to calculate p value. (M) Kaplan–Meier survival curves of mice transplanted with Jmjd3^+/+ and Jmjd3^Δ/Δ L-GMPs transfected with empty or p16^Ink4a. 2.0 × 10² L-GMPs were transplanted into lethally irradiated recipients with 2.5 × 10⁵ wild-type competitor MNBM cells (n = 6–7). A log-rank test was used to calculate p values. Source data are available online for this figure.

Figure 7. JMJD3 contributes to cellular aging by regulating p16^Ink4a.

(A) Schematic diagram of the experiments with young and aged LSK cells and LT-HSCs. (B) qPCR analysis of p16^Ink4a in LSK cells of young (2 M) and aged (26 M) Jmjd3^+/+ and Jmjd3^Δ/Δ mice (mean + SD, n = 3). (C) H3K27me3 enrichment in the promoter region of Cdkn2a (see Fig. 3D) in LSK cells from aged (26 M) Jmjd3^+/+ and Jmjd3^Δ/Δ mice. Results are shown as fold changes relative to a negative control (Neg ctrl) (mean + SD, n = 3). (D) Competitive repopulation assays. 5.0 × 10² LT-HSCs from young (2 M) (n = 6–7), mature (18 M) (n = 6–7), and aged (26 M) (n = 4) Jmjd3^+/+ and Jmjd3^Δ/Δ mice were transplanted into lethally irradiated recipients with 2.5 × 10⁵ competitor MNBM cells (mean ± SD). Student’s t test was used to calculate p values. Source data are available online for this figure.

Figure EV1. Targeting strategy, genotyping of ES clones, and deletion of Jmjd3.

(A) Exons 15–17 of the mouse Jmjd3 gene were encompassed by two loxP sites (black triangles), and a neomycin-resistance gene (Neo) was flanked by two Frt sites (white arrows). After removing Neo by Flpe, floxed exons were deleted by crossing with MxCre⁺ mice and pIpC treatment. The position of the genomic probe for the 5′ Southern blot (5′ probe), primers for the 3′ genomic PCR (P1 and P2), and the positions of restriction enzymes (BamHI and Eco437) are shown. BamHI with an asterisk (*BamHI) is an artificial enzyme site introduced by in vitro mutagenesis. Gray boxes indicate the exons that encode the JmjC domain. (B) Homologously recombined ES clones (#1 and #2) identified by 5′ Southern blot and 3′ genomic PCR. Germline (GL) and targeted bands in 5′ Southern blot are indicated by arrows (left panel) and PCR products for the 3′ genomic PCR are indicated by arrowheads (right panel). (C) Immunoblot showing JMJD3 protein in bone marrow (BM) cells of Jmjd3^flox/flox; MxCre⁺ and Jmjd3^flox/flox; MxCre⁻ mice at 4 weeks (4 wks) and 6 months (6 mos) after pIpC treatment.

Figure EV2. Analysis of Jmjd3^Δ/Δ hematopoietic cells at steady state.

(A) Analysis of peripheral blood (PB) parameters in Jmjd3^+/+ and Jmjd3^Δ/Δ mice at 4 weeks after pIpC treatment. White blood cell (WBC) counts, hemoglobin concentration (Hgb), and platelet (Plt) number in the PB of Jmjd3^+/+ (n = 13) and Jmjd3^Δ/Δ mice (n = 12) are plotted as dots, and the mean values are indicated as bars. Student’s t test was used to calculate p value. (B) Analysis of lineage differentiation (Thy1.2⁺, B220⁺, and Mac-1⁺ cells) in the PB cells of Jmjd3^+/+ (n = 13) and Jmjd3^Δ/Δ mice (n = 12) (mean ± SD). Student’s t test was used to calculate p value. (C) Flow cytometric profiles of lineage^- (Lin⁻) cells in the BM of Jmjd3^+/+ and Jmjd3^Δ/Δ mice (mean + SD, n = 5). Student’s t test was used to calculate p value. (D) Absolute numbers of HSPC subpopulations (LT-HSC, ST-HSC, and MPP) and myeloid progenitors (CMP, GMP, and MEP) in the BM of Jmjd3^+/+ and Jmjd3^Δ/Δ mice (mean + SD, n = 5). Student’s t test was used to calculate p values.

Figure EV3. JMJD3 competitively regulates Polycomb targets under stress.

Comparison of gene set enrichment between Polycomb proteins and JMJD3. Gene sets upregulated in cells deficient in the indicated Polycomb genes were compared with those downregulated in Jmjd3-deficient LSK (Steady), LSK (BMT), or L-GMP. NES and FDR are indicated. Blue boxes show significantly enriched pathways (FDR < 0.25).

Figure EV4. HSPC defects caused by Jmjd3 loss under replicative stress are rescued by retroviral expression of a Ccnd1 mutant.

(A) A schematic diagram of empty, Ccnd1^WT, and Ccnd1^T156A transduction. LK cells from Jmjd3^+/+ and Jmjd3^Δ/Δ mice were transfected with empty-, Ccnd1^WT-, or Ccnd1^T156A-IRES-EGFP retrovirus, and 5.0 × 10⁴ EGFP⁺ cells were transplanted into lethally irradiated recipients with 2.5 × 10⁵ wild-type competitor MNBM cells for radioprotection (mean + SD, n = 4). (B) Chimerism of donor-derived PB cells in the recipients. Results of mice transplanted with Jmjd3^+/+/empty, Jmjd3^Δ/Δ/empty, Jmjd3^Δ/Δ/Ccnd1^WT, and Jmjd3^Δ/Δ/Ccnd1^T156A cells are shown (mean ± SD, n = 4). Dunnett’s test was used to calculate p value. (C) Lineage differentiation in the donor derived PB cells in recipients 4 months after BMT (mean + SD, n = 4). Dunnett’s test was used to calculate p value. (D) Chimerism of donor-derived BM cells in recipients 4 months after BMT (mean + SD, n = 4). Dunnett’s test was used to calculate p values.

Figure EV5. Inhibition of JMJD3 suppresses LSC potential by regulating p16^Ink4a expression.

(A) Schematic diagram of GSK-J4 treatment after MLL-AF9 (MA9) transduction. LK cells from wild-type mice were transfected with the MLL-AF9-IRES-EGFP retrovirus. EGFP⁺ cells were further exposed to DMSO or GSK-J4 for 24 h and subjected to the following assays. (B) Flow cytometric profiles of c-kit⁺ fractions in MA9 cells treated with DMSO or GSK-J4 (5 or 10 μM) for 24 h. (C) qPCR analysis of Cdkn2a genes in L-GMPs exposed to DMSO or GSK-J4 (10 μM) for 24 h. (mean + SD, n = 3). (D) Immunofluorescence staining (left panel) and relative fluorescence intensity (right panel) of H3K27me3 in L-GMPs exposed to DMSO or GSK-J4 (10 μM) for 24 h. Mean values are indicated as bars (n = 105). Student’s t test was used to calculate p value. Scale bar, 10 μm. (E) H3K27me3 levels in the promoter region of Cdkn2a (see Fig. 3D) in L-GMPs exposed to DMSO or GSK-J4 (10 μM) for 24 h. Results are shown as fold changes relative to a negative control (Neg ctrl) (mean + SD, n = 3). (F) Venn diagrams showing the overlap of negatively enriched KEGG pathways in L-GMPs exposed to GSK-J4 (10 μM) for 24 h and Jmjd3^Δ/Δ L-GMPs. The overlapped pathways are listed in Table EV2. (G) GSEA plots of L-GMPs exposed to DMSO or GSK-J4 (10 μM) for 24 h in the indicated gene sets (left, genes commonly upregulated in human HSC and LSC; middle, genes commonly upregulated in quiescent human CD34⁺ hematopoietic cells; right, genes commonly upregulated through the p16^INK4a/RB1 pathway. Results are shown with NES and FDR values. (H) Flow cytometric analysis of BrdU incorporation in L-GMPs exposed to DMSO or GSK-J4 (5 or 10 μM) for 24 h. (mean + SD, n = 3). Student’s t test was used to calculate p value. (I) Schematic diagram of in vitro GSK-J4 (10 μM) treatment for 24 h after MLL-AF9 transduction (left panel) and Kaplan–Meier survival plots of mice transplanted with these cells (right panel). In all, 1.0 × 10⁵ MA9 cells (Ly5.2⁺) were transplanted into lethally irradiated recipients with 2.5 × 10⁵ wild-type competitor MNBM cells (n = 13). A log-rank test was used to calculate p value. (J) Schematic diagram of in vivo GSK-J4 treatment after MLL-AF9 transduction (left panel) and Kaplan–Meier survival plots of mice transplanted with MA9 cells and treated in vivo (right panel). In all, 1.0 × 10⁵ MA9 cells (Ly5.2⁺) were transplanted into lethally irradiated recipients with 2.5 × 10⁵ wild-type competitor MNBM cells. 10 days after BMT, DMSO or GSK-J4 (50 mg/kg/day) was intraperitoneally injected into the recipients for 5 consecutive days (n = 12). A log-rank test was used to calculate p values.

Figure EV6. Analysis of Jmjd3^Δ/Δ HSPCs in replicative stress caused by 5-FU treatment and Utx^Δ/Δ HSPCs at steady state.

(A) Changes in WBC count in Jmjd3^+/+ and Jmjd3^Δ/Δ mice every 8 days after 5-FU injection (150 mg/kg) (mean ± SD, n = 3). Student’s t test was used to calculate p value. (B) Absolute numbers of BM cells from Jmjd3^+/+ and Jmjd3^Δ/Δ mice 24 days after 5-FU treatment (mean + SD, n = 3). Student’s t test was used to calculate p value. (C) Absolute numbers of HSC subpopulations (LT-HSC, ST-HSC, and MPP) in the BM of Jmjd3^+/+ and Jmjd3^Δ/Δ mice 24 days after 5-FU treatment (mean + SD, n = 3). Student’s t test was used to calculate p value. (D) Scatter plots comparing normalized expression of individual genes (RPKM > 1) in LSK cells of Utx^Δ/Δ mice compared with Utx^+/+ mice at steady state. Genes more than twofold upregulated and downregulated are plotted as red and blue dots, respectively. (E) Venn diagrams showing the overlap of genes more than twofold upregulated or downregulated in Jmjd3^Δ/Δ and Utx^Δ/Δ LSK cells at steady state.