A STAT5B-CD9 axis determines self-renewal in hematopoietic and leukemic stem cells

Abstract

The transcription factors signal transducer and activator of transcription 5A (STAT5A) and STAT5B are critical in hematopoiesis and leukemia. They are widely believed to have redundant functions, but we describe a unique role for STAT5B in driving the self-renewal of hematopoietic and leukemic stem cells (HSCs/LSCs). We find STAT5B to be specifically activated in HSCs and LSCs, where it induces many genes associated with quiescence and self-renewal, including the surface marker CD9. Levels of CD9 represent a prognostic marker for patients with STAT5-driven leukemia, and our findings suggest that anti-CD9 antibodies may be useful in their treatment to target and eliminate LSCs. We show that it is vital to consider STAT5A and STAT5B as distinct entities in normal and malignant hematopoiesis.

Graphical abstract

Figure 1.

STAT5B, but not STAT5A, has an important role in HSC dormancy. (A-B) HSC flow cytometry analyses of wt, Stat5a^−/−, or Stat5b^−/− BM. (A) Representative FACS plots showing HSC/multipotent progenitor (MPP)1 (CD150⁺CD48⁻), MPP2 (CD150⁺CD48⁺), MPP3/4 (CD150⁻CD48⁺), and MPP5 (CD150⁻CD48⁻) cell populations gated on LSK cells. (B) Relative quantification of HSC, MPP1-5, and LSK cells (n ≥ 6; mean ± standard error of the mean [SEM]). HSC subpopulations: HSC (LSK, CD34⁻CD48⁻CD150⁺CD135⁻), MPP1 (LSK, CD34⁺CD48⁻CD150⁺CD135⁻), MPP2 (LSK, CD34⁺CD48⁺CD150⁺CD135⁻), MPP3 (LSK, CD34⁺CD48⁺CD150⁻CD135⁻), and MPP4 (LSK, CD34⁺CD48⁺CD150⁻CD135⁺). (C-F) Single cell RNA-Seq of FACS-sorted LSK cells of WT, Stat5a^−/−, or Stat5b^−/− BM (n = 3 pooled/genotype). (C) Experimental workflow. (D) Force-directed graphs of WT, Stat5a^−/−, or Stat5b^−/− LSKs, color code as shown in supplemental Figure 1C. (E) Changes in HSC_1 and HSC_2 cluster sizes of Stat5a^−/− or Stat5b^−/− relative to WT. (F) Absolute differences in the percentage of cells expressing known genes associated with dormant HSCs in pooled clusters HSC_1 and HSC_2 of Stat5a^−/− or Stat5b^−/− compared with WT. (G-H) HSC cell cycle analyses of WT, Stat5a^−/−, or Stat5b^−/− BM. (G) Representative FACS plots (KI67/DAPI) of HSC/MPP1 with gating strategy and indicated percentages, and (H) quantification of cell cycle phase distributions of HSC/MPP1 cells (n = 5; mean ± SEM). Levels of significance were calculated using 1-way analysis of variance (ANOVA) (B,H). *P < .05; **P < .01; ***P < .001.

Figure 2.

STAT5B drives HSC self-renewal. (A-B) Single HSC assay using cell surface markers (lineage⁻ [CD3, CD19, CD11b, Gr-1, Ter-119], c-kit⁺, Sca-1⁺, CD150⁺, CD48⁻). (A) Schematic of single cell in vitro cultures. Single HSC/MPP1 cells of WT, Stat5a^−/−, or Stat5b^−/− BM were FACS-sorted into individual wells and assessed for proliferation and surface marker expression after 10 days of culture. (Bi) Quantification of clonal outgrowth (n = 74/genotype). (Bii) Representative FACS plots of LSK gatings and quantification of total LSK cell numbers (n = 74/genotype, whiskers: Tukey). (C) STAT5A (STAT5A-green fluorescent protein [GFP]) or STAT5B (STAT5B-GFP) overexpression in LSK cells, EV (empty vector-GFP) was used as control. Competitive growth analyses over 28 days (n = 5/genotype, mean ± SEM). (D-F) 5-FU recovery assays of WT, Stat5a^−/−, or Stat5b^−/− mice. (D) Experimental workflow. (EF) Representative FACS plots of LSK gatings of lineage negative BM cells (E) and quantification of LSK cells and HSC/MPP1 cells 8 days after 5-FU injection (n ≥ 5/genotype, mean ± SEM). (G-I) Serial transplantation assays of WT, Stat5a^−/−, or Stat5b^−/− BM cells (F). (G) Experimental workflow: 5 × 10⁶ BM cells (Ly5.2⁺) were injected into lethally irradiated recipient mice (Ly5.1⁺) and reinjected 8 weeks after. TP (transplantation). (H) Representative LSK gatings of lineage negative BM cells 8 weeks after first and fourth transplantation. (I) Quantification of WT, Stat5a^−/−, and Stat5b^−/− LSK and HSC/MPP1 cells 8 weeks after fourth transplantation (n ≥ 6/genotype, mean ± SEM). Levels of significance were calculated using Fisher’s exact test (Bi) or 1-way ANOVA (Bii,F,I) or 2-way ANOVA (C). *P < .05; **P < .01; ****P < .0001.

Figure 3.

Selective STAT5B activation in HSCs drives self-renewal. (A) Flow cytometric analysis of pYSTAT5 levels in HSC subpopulations. Representative histograms of pYSTAT5 signal in WT HSC/MPP1, MPP2, and MPP3/4 20 minutes after TPO stimulation or unstimulated (us) (i) and the quantification of the MFI pYSTAT5 signal (n ≥ 5, mean ± SEM) (ii). (B) Quantification of cells expressing Socs2 or Cish in pooled HSC_1 and HSC_2 clusters (single cell RNA-Seq) of Stat5a^−/− or Stat5b^−/− relative to WT. (C) Flow cytometric analysis of pYSTAT5 levels in different cell types of WT, Stat5a^−/−, and Stat5b^−/− BM after respective stimulations for 20 minutes. (i) Representative histograms of pYSTAT5 signal of HSC/MPP1 cells after TPO stimulation; antibody isotype staining (Ctrl) and unstimulated (us) are shown as control. (ii) B220⁺ B and CD3⁺ T cells were stimulated with IL-7 and IL-2; Gr1⁺CD11b⁺ granulocytes were stimulated with GM-CSF; CD41⁺ megakaryocytes were stimulated with TPO; CD71⁺ erythroid cells were stimulated with EPO; and LSK cells and HSC/MPP1 cells were stimulated with TPO. Log₂ pYSTAT5 levels of Stat5a^−/− or Stat5b^−/− relative to WT (n ≥ 4/genotype, mean ± SEM). (D) pYSTAT5 levels of HSCs derived from hyperactive STAT5A (cS5A^hi) or STAT5B (STAT5B^N642H) transgenic mice (n = 3/genotype, mean ± SEM) determined by intracellular flow cytometry staining. (E) pSTAT5 (Y694/699), STAT5A, and STAT5B immunoblot analysis of cytoplasmic and nuclear fraction of unstimulated (us) or TPO-stimulated (20 minutes) HPC^LSK WT cells (n = 3). α-TUBULIN served as loading control for the cytoplasmic fraction and RCC1 for the nuclear fraction. Representative blot of 3 independent experiments. (F) STAT5A or STAT5B IHC staining of HPC^LSK WT cytospins with or without (us) 20-minute TPO stimulation. Quantification of 3 individual images per condition (1 representative cell line of 4) showing the fold change (fc) of TPO/us of percent nuclear STAT5A or STAT5B. Levels of significance were calculated using 1-way ANOVA (A,D), and using an unpaired Student t test (C). **P < .01; ***P < .001; ****P < .0001.

Figure 4.

Selective STAT5B activation drives the self-renewal of LSCs. (A) Colony numbers in serial plating assay of HPC^LSK cells expressing empty vector (eV), STAT5B, or STAT5B^N642H (n = 6/genotype, mean ± SEM). (B) BCR/ABL^p210+ LSK retransplantation assay: 2 × 10⁴ BCR/ABL^p210-GFP+ FACS-sorted LSK cells were injected IV in NSG mice. Two recipients per donor were injected to use 1 for retransplantation after 13 weeks and 1 for terminal analysis. Experimental workflow (i) and survival analyses of recipients of the primary and secondary transplantation (TP, n = 4/genotype) (ii). (C) pSTAT5 (Y694/699), STAT5A, and STAT5B immunoblot analysis of cytoplasmic (cyt) and nuclear (nuc) fraction of HPC-7 cells expressing JAK2^V617F and HPC^LSK cells expressing FLT3-ITD or BCR/ABL^p210. α-TUBULIN served as loading control for the cytoplasmic fraction and RCC1 for the nuclear fraction. A representative blot of 3 independent experiments is shown. Levels of significance were calculated using 1-way ANOVA (A). **P < .01; ***P < .001.

Figure 5.

CD9 as marker and direct target of pYSTAT5B signaling. (A) 35 STAT5B-specific HSC genes. Color scale indicates % ln of fold-change difference between Stat5a^−/− or Stat5b^−/− and WT. (B) Schematic workflow: genes from panel A were analyzed for their expression levels-based influence on patient survival in STAT5-driven patient cohorts GSE6891 (FLT3-ITD), GSE37642 (Intermediate I), OHSU (FLT3-mut., NPM1-mut.), and the non–STAT5-driven cohort GSE37642 (Intermediate II). (C) Survival analyses comparing CD9-low vs CD9-high expressing patients in the STAT5-driven cohorts GSE6891 (FLT3-ITD), GSE37642 (Intermediate I), and OHSU (FLT3-mut., NPM1-mut.). (D) STAT5B chromatin immunoprecipitation (ChIP)–quantitative polymerase chain reaction: Quantification of STAT5B binding to the Cd9 promoter (prom) or negative regions (neg1 and neg2) in HPC^LSK WT cells, unstimulated (us), or after 30 minutes of TPO stimulation (n = 3, mean ± SEM). (E) CD9 expression changes after culture with TPO. (i) Representative CD9 FACS plots of in vitro cultured WT LSK cells with or without (us) TPO after 96 hours. (ii) Quantification of CD9 expression of WT LSK cells after TPO treatment for 24, 48, 72, and 96 hours (n = 2, mean ± SEM). (F) Analysis of CD9 expression in HSC lines expressing STAT5B-driving oncoproteins. Representative histograms of CD9 expression (i) and quantification of CD9 levels of HPC^LSK WT (ii) (n = 3, mean ± SEM), +FLT3-ITD, +BCR/ABL^p210 (n = 3, ±SEM), and HPC-7 WT, FLT3-ITD, or JAK2^V617F. Levels of significance were calculated using a log-rank test (C), and an unpaired Student t test in (D). Levels of significance were calculated using the Friedman test (E). *P < .05.

Figure 6.

CD9 blocking affects STAT5 activation and impedes self-renewal. (A) pYSTAT5 levels determined by intracellular flow cytometry staining of CD9-low (CD9 MFI < median) and CD9-high (CD9 MFI > median) expressing FLT3-ITD⁺ NPM1⁺ patient BM cells (n = 5). (B) Flow cytometric analyses of pYSTAT5 levels in IgG- or aCD9-treated JAK2^V617F BM cells after respective stimulations for 20 minutes. (i) Differences of pYSTAT5 levels (aCD9 minus IgG) (n ≥ 3/genotype, mean ± SEM). (ii) Representative histograms of pYSTAT5 signal of HSC/MPP1 cells unstimulated (us) or after TPO stimulation; antibody isotype-stained (Ctrl) cells are shown as control. (C-E) Colony formation assays of wt and JAK2^V617F mouse BM cells either treated with IgG or aCD9 (5 µg/mL). (C) (i) Representative pictures of colony formation and (ii) fold change in colony numbers (n ≥ 5/genotype and condition, mean ± SEM). (D-E) Quantification of LSK cells (n ≥ 5/genotype and condition, mean ± SEM) after aCD9 or IgG treatment (D) and representative FACS plots of c-kit/Sca-1 (i) and CD11b expression in JAK2^V617F cells (ii) after aCD9 or IgG treatment (E). Levels of significance were calculated using a paired Student t test (A-B), and an unpaired Student t test (C-D). *P < .05; **P < .01; ***P < .001.

Figure 7.

CD9 as a therapeutic target for pYSTAT5-driven leukemia. (A-B) JAK2^V617F transplantation with aCD9 in vivo treatment. (A) Experimental workflow. (B) Log₂ fold changes (aCD9 vs IgG) of HSC subpopulations and total BM cell numbers of (i) Ly5.2⁺ JAK2^V617F donors or (ii) Ly5.1⁺ WT NSG recipients (n = 4 per treatment, mean ± SEM, 4× 1.25 mg/kg IV IgG or aCD9 were applied). (C) Analysis of CD9 expression in CD34⁺CD38⁻ cells of (i) MPN^JAK2V617F+ (n = 4) and (ii) CML^BCR/ABL1+ (n = 6) patient and control (n = 4) BM. (D-E) In vitro treatment of control and patient BM cells either treated with IgG or aCD9 (2 µg/mL). XY plots showing CD9 levels and (D) CD34⁺CD38⁻ cell numbers or (E) Annexin-V levels of control (n = 4) and (i) MPN^JAK2V617F+ (n = 3) or (ii) CML^BCR/ABL1+ patient BM mononuclear cells (n = 6). (F) Serial plating assays of CML^BCR/ABL1+ patient (n = 6) BM either treated with IgG or aCD9. Fold changes of aCD9/IgG colony numbers in the first and second plating. (G) Serial plating assays of MPN^JAK2V617F+ patient (n = 5) BM either treated with IgG or aCD9. Fold changes of aCD9/IgG colony numbers in the first and second plating. Levels of significance were calculated using an unpaired Student t test (C), and a paired Student t test (F-G). Levels of significance and correlation were calculated using Pearson in panels D-E. *P < .05; ***P < .001.