Rb family proteins enforce the homeostasis of quiescent hematopoietic stem cells by repressing Socs3 expression

Hematopoietic stem cells (HSCs) are highly quiescent and can occasionally enter the cell cycle to either self-renew or differentiate into intermediary progenitor cells (Morrison and Scadden, 2014). In vivo studies have shown that progression through a unique round of cell cycle is not detrimental for HSC homeostasis (i.e., the size of the HSC pool). In contrast, prolonged exit from quiescence progressively impairs the self-renewal and engraftment capacity of proliferative HSCs, which ultimately disrupts their homeostasis in the BM. These observations have fueled the hypothesis that a common yet undefined mechanism enforces both the cellular quiescence and homeostasis of HSCs in the BM (Orford and Scadden, 2008).

The HSC niche provides growth factors and cytokines to modulate HSC fate. In particular, HSCs rely on the binding of the thrombopoietin (Tpo; Tong et al., 2007) cytokine to its receptor Mpl to promote their self-renewal and homeostasis in the BM. Mpl is devoid of any kinase activity and thus recruits the Jak2 kinase to activate several intracellular cascades (Mapk, Akt, and Stat pathways) upon Tpo binding (Vila-Coro et al., 1999; Bersenev et al., 2008). Accordingly, genetic inactivation of Mpl (Kimura et al., 1998), Jak2 (Akada et al., 2014), and Stat5 (Wang et al., 2009) leads to impaired HSC homeostasis and progressive BM failure. In addition to these positive cues, Jak2 is also negatively regulated by suppressor of cytokine signaling (Socs) proteins (Kershaw et al., 2013) and Lnk. Inactivation of Lnk increases Jak2 activity and the size of the HSC pool in the BM (Bersenev et al., 2008). Therefore, it appears that Jak2 plays a central role in the regulation of HSC pool size and that a balance of positive and negative regulators of Jak2 activity controls HSC homeostasis in the BM.

Rb proteins (Rb, p107, and p130) enforce the cellular quiescence of HSCs by repressing the activity of E2f transcription factors through physical interaction (Burkhart and Sage, 2008; Chen et al., 2009). Mitogen stimulation of quiescent HSCs leads to dissociation of the Rb/E2f complex, followed by E2f-mediated activation of a transcriptional program that drives the progression of HSCs through the G₁/S restriction point, by which the fate (self-renewal vs. differentiation) of the daughter cells is thought to be determined (Pietras et al., 2011). However, whether and how E2f factors also govern cell fate determination during progression through the cell cycle is unknown (Chen et al., 2009). In addition, proliferative HSCs are mobilized into the peripheral circulation, suggesting that their retention in the niche may be altered upon entry into the cell cycle (Passegué et al., 2005).

Taking advantage of conditional Rb family–deficient mice (triple KO [TKO]), we previously demonstrated that Rb protein inactivation in adult HSCs leads to their robust proliferation and impaired engraftment (Viatour et al., 2008). Using these TKO mice, we now show that Rb proteins collectively maintain HSC homeostasis by promoting the activity of Jak2 downstream of Tpo signaling through repression of E2f-mediated activation of Socs3 expression. Accordingly, inactivation of the Rb family in HSCs progressively impairs their homeostasis, which is rescued upon repression of Socs3 expression in TKO HSCs. Collectively, our results elucidate a long-awaited mechanism by showing that Rb proteins enforce the homeostasis of quiescent HSCs in the BM by repressing distinct transcriptional programs regulated by E2f factors.

We inactivated the entire Rb family of genes in all hematopoietic cells by deleting Rb and p130 alleles in Rb^Lox/Lox p130^Lox/Lox p107^−/− mice using a tamoxifen-regulated Cre recombinase expressed from the Rosa26 locus (Rosa26-CreER^T2). Here, we refer to hematopoietic cells with Rb family deletion as TKO cells. We observed unaltered frequency of phenotypic TKO progenitors (lineage⁻ Kit⁺ Sca1⁺ Cd48⁺ Cd150⁻) and HSCs (lineage⁻ Kit⁺ Sca1⁺ Cd48⁻ Cd150⁺) relative to control (CT; provided by tamoxifen-treated Rb^Lox/Lox p130^Lox/Lox p107^−/− mice, which are phenotypically and functionally indistinguishable from WT mice; Fig. S1) populations 2 wk after deletion (Fig. 1 A) despite their increased proliferative activity (Fig. 1, B and C). To assess the growth potential of TKO HSCs in vitro, we plated unfractioned BM and purified HSCs isolated from CT and TKO mice 2 wk after tamoxifen treatment in semisolid culture for colony assay. TKO cells initially exhibited increased colony-forming activity, but serial replating of a fixed number of cells resulted in faster exhaustion of their colony-forming activity compared with CT (Fig. 1 D and Fig. S2 A). To rule out nonhematopoietic priming of TKO HSCs before collection from BM, we isolated unfractioned BM cells from untreated CT and TKO mice and determined their colony-forming activity in the presence of 4-OH-tamoxifen. Upon successful recombination of floxed alleles (not depicted), TKO cells displayed impaired colony-forming activity compared with CT (Fig. S2 B). Altogether, these data suggest that, after an initial phase of expansion, TKO HSCs and their downstream progenitors exhaust prematurely in vitro.

Expression of the Rosa26 locus is ubiquitous (Casola, 2010), leading to a widespread Rb family inactivation and the subsequent early mortality of Rosa26Cre-ER^T2 TKO mice. To circumvent this early mortality (which prevents long-term in vivo study) and avoid potential non–cell-autonomous phenotypes, we transplanted unrecombined BM cells from CT and TKO mice into lethally irradiated immunocompromised hosts. (Rag1^−/− mice were used as recipients to overcome the transplantation issues associated with the mixed 129/BL6 background of TKO mice; these mice are defined as CT and TKO chimeras based on the origin of the transplanted BM cells.) Upon establishment of hematopoiesis (5 wk), recipient mice were treated with tamoxifen and killed at predetermined time points to analyze hematopoiesis and the HSC compartment (Fig. 1 E). Whereas the BM pool of progenitors in both groups was stable during the course of the experiment, TKO HSC frequency initially remained stable during the first 2 wk, then progressively decreased to approximately a tenth of the CT HSC frequency by the 8-wk time point (Fig. 1, F and G; and Fig. S2 C). HSCs isolated from TKO chimeras at 8 wk displayed decreased colony-forming activity compared with CT, whereas the colony-forming activity of progenitors was unaffected (Fig. 1 H and Fig. S2 D). However, TKO cells did not display increased apoptosis compared with CT (Fig. S2 E). These data indicate that, in addition to decreased frequency, TKO HSC are also functionally impaired in their capacity to form colonies.

Recent data have highlighted the predominant role of HSCs for hematopoiesis upon transplantation (Sun et al., 2014; Busch et al., 2015). However, histological analysis reveals a modest decrease in the BM cellularity of TKO chimeras compared with CT (Fig. S2, F and G), with a reduced percentage of lymphoid cells in TKO chimeras compared with CT (Fig. 1 I). In the spleen, expansion of the red pulp, which is suggestive of extramedullary hematopoiesis (O’Malley et al., 2005), was observed in TKO chimeras at 4 wk but not 18 wk (Fig. 1 J). At the 18-wk time point, TKO chimera spleens were enlarged and contained a higher number of HSCs and progenitor compared with CT (Fig. 1, K and L), which is associated with higher colony-forming activity of TKO splenocytes compared with CT (Fig. 1 M). Yet, the colony-forming activity from TKO chimera spleen was marginal compared with the activity observed from CT and TKO chimera BM (Fig. 1 N). When cellularity from both BM and spleen is combined, 10–20% of TKO HSCs were still present in the BM, and another 20–25% were present in the spleen, compared with CT (Fig. 1 O). Collectively, these data indicate that upon Rb family inactivation, a significant fraction of HSCs exit the BM to occupy the spleen. However, after a brief period of hematopoiesis, their contribution to blood cell formation appears to be limited.

Finally, we generated additional chimeras and, 5 wk after tamoxifen treatment, performed weekly injection with the myeloablative agent 5-fluorouracil (5-FU; Fig. 1 P). In these conditions, TKO chimeras displayed early mortality compared with CT (Fig. 1 Q), indicating that inactivation of Rb family proteins impairs the resistance of HSC to stress-induced hematopoiesis.

To elucidate the molecular mechanism by which Rb proteins maintain the homeostasis of quiescent HSCs, we performed array analyses on flow-purified HSCs from CT and TKO mice 2 wk after tamoxifen treatment (Table S1). GSEA analysis of the genome-wide array data (see Tables S2 and S3 for full results) identified cell cycle and E2f target genes as the most enriched gene sets in TKO HSCs compared with CT. The third most highly enriched gene set corresponded to regulators of cytokine signaling (Fig. 2 A) and, from this gene set, Socs3 was the most highly up-regulated mRNA in TKO HSCs compared with CT (Fig. 2 B; and Table S1). RT-qPCR analysis confirmed increased expression of Socs3 in TKO HSCs compared with CT and showed that Socs3 transactivation is predominantly restricted to TKO HSC, as Socs3 expression barely increases in TKO progenitors (Fig. 2 C). Because Socs3 overexpression enforces the differentiation of embryonic stem cells by inhibiting Jak2 activity downstream of Lif signaling (Li et al., 2005), we reasoned that Socs3 may represent a valid candidate for the regulation of proliferative HSC homeostasis. To understand the molecular mechanism of Socs3 up-regulation in TKO HSCs, we screened the Socs3 promoter region and identified an E2f binding site in the proximity of the transcription start site that is conserved among several Socs gene promoters (Fig. 2 D). Analysis of previously published chromatin immunoprecipitation (ChIP)–Seq data identified the binding of E2f1 to the Socs3 promoter in breast cancer (Cao et al., 2011) and liver cancer (Tarangelo et al., 2015) cell lines, but analysis of expression data showed a lack of Socs3 transactivation in Rb family–deficient hepatocellular carcinoma (TKO HCC; Tarangelo et al., 2015), suggesting a cell type–specific transactivation of Socs3 by E2f. In addition, ChIP experiments in unfractioned primary BM cells and myeloid progenitor–derived 32D cells, as well as luciferase assays in 293 cells, confirmed that E2f factors bind to and transactivate the Socs3 promoter (Fig. 2, E–G), thereby establishing Socs3 as a novel E2f target gene.

We recently identified two classes of E2f target genes defined as early and late targets (Tarangelo et al., 2015). Although late E2f target genes (primarily activated upon prolonged E2f activity and enriched for non–cell cycle functions) harbor only a minimal E2f core binding site (-CGCGC-), early E2f target genes (promptly activated upon transient E2f activity and predominantly enriched for cell cycle functions) harbor a thymidine (T)-stretch at the 5′ extremity of a minimal E2f core binding site that confers high E2f affinity binding (Tarangelo et al., 2015). The E2f binding site located in the Socs3 promoter includes a degenerated T-stretch, suggesting that Socs3 is a late E2f target gene; accordingly, inserting a T-stretch in the E2f binding site improved the binding affinity of recombinant E2f1 (Fig. 2 H). In addition, minimum competition (1:2 ratio) by a cold E2f consensus probe was sufficient to disrupt the E2f–Socs3 probe complex (Fig. 2 I). In agreement with the delayed activation kinetics of late E2f target genes, Socs3 expression was not activated in proliferative TKO HSCs 4 d after Rb family inactivation (Fig. 2 J). Accordingly, expression analysis of HSCs isolated at different time points after their stimulation with 5-FU showed the rapid induction of a cell cycle signature peaking at day 6 (in agreement with the findings of the authors) in the absence of any significant Socs3 transactivation (Fig. 2 K; Venezia et al., 2004). Collectively, these data identify Socs3 as a novel E2f target gene in HSCs that is activated only in the context of sustained E2f activity.

To test the consequences of increased Socs3 expression on TKO HSC homeostasis, we determined the activity of cytokines Tpo and Sdf1 (which is critical for HSC engraftment and retention in the BM; Sugiyama et al., 2006) that recruit Jak2 in HSCs and other cell types (Fig. 3 A; Vila-Coro et al., 1999; Pello et al., 2006; Bersenev et al., 2008), in CT and TKO cells isolated 2 wk after tamoxifen treatment. Stimulation with Tpo led to reduced Jak2 phosphorylation in TKO KLS (lineage⁻ Kit⁺ Sca1⁺) cells, compared with CT (Fig. 3 B). Using p-Stat5 and p-S6 as readouts for Jak2 activity in progenitors and HSCs, we found that stimulation with Tpo led to decreased phosphorylation of Stat5 and S6 in TKO cells compared with CT (Fig. 3, C and D). These results therefore show that Tpo/Mpl/Jak2 signaling is impaired in TKO HSCs and progenitors. In addition, TKO KLS cells exhibited impaired migration capacity when placed in a Sdf1 gradient (Fig. 3 E), indicating defective Sdf1 signaling in these cells. However, intracellular signaling analysis revealed a lack of Jak2 and Fak (Glodek et al., 2007; another Socs3 target [Le et al., 2007]) activation upon Sdf1 stimulation in KLS cells, suggesting that Sdf1 may activate an alternative signaling pathway in hematopoietic progenitors (Fig. 3 F).

In contrast to TKO HSCs, TKO progenitors only exhibited reduced Jak2 activity upon exposure to a high Tpo dose (20 ng/ml; Fig. 3, C and D), which correlates with the limited transactivation of Socs3 in TKO progenitors (see Fig. 2 C) and KLS cells (which contain a majority of progenitors; Fig. 3 G). Remarkably, global comparison of gene signatures in TKO HSCs and KLS cells versus their respective CTs indicated that the transcriptional output is extremely different between TKO HSCs and progenitor cells, with only 21 genes up-regulated in both HSC and KLS cells upon Rb family inactivation. In contrast, 336 and 364 genes are uniquely up-regulated in TKO HSC and TKO KLS cells, respectively (Fig. 3 H; see Tables S4, S5, and S6 for full gene lists). These differentiation stage–specific gene signatures each combine distinct cell cycle genes (Table S7) as well as genes associated with non–cell cycle functions, such as Socs3. These data suggest that the transcriptional consequences of Rb family inactivation, predominantly driven by E2f, include both the activation of a common cell cycle program (although driven by different cell cycle genes) and the cell type–specific activation of non–cell cycle features.

Finally, fetal liver HSCs retain self-renewal potential despite their sustained proliferation, suggesting that the mechanisms linking quiescence and homeostasis may be different in fetal versus adult HSCs. Accordingly, whereas fetal liver HSCs activated a robust cell cycle signature compared with their adult counterparts, they failed to transactivate Socs3 expression (Fig. 3 I; Venezia et al., 2004). Collectively, these results suggest that Socs3 transactivation by E2f factors is restricted to adult HSCs.

To functionally test the consequences of increased Socs3 expression on the regulation of TKO HSC homeostasis, we genetically manipulated Socs3 expression in CT and TKO KLS cells in vitro and in vivo. We found that Socs3 overexpression decreased the number and size of colonies formed by transduced CT and TKO cells in methylcellulose culture (Fig. 4, A and B) and increased the frequency of lineage-specific colonies (granulocyte, macrophage, granulocyte-macrophage, colony-forming unit erythrocyte, burst-forming unit, and megakaryocyte) at the expense of mixed-lineage colonies (granulocyte-erythrocyte-monocyte-megakaryocyte; Fig. 4 B). Similarly, overexpression of Socs3 in CT lineage⁻ Kit⁺ (KL) cells followed by their transplantation into recipient mice decreased the frequency of HSCs but not their downstream progenitors, as shown by the unaltered frequency of KLS cells (Fig. 4 C). Together, these data indicate that overexpression of Socs3 impairs HSC homeostasis in the BM.

We then performed the opposite experiment via shRNA-mediated silencing of Socs3 expression (using two different hairpins to silence Socs3 in parallel with a Scramble hairpin; Cui et al., 2011) in CT and TKO KLS cells. Partial repression (∼65% knockdown) of Socs3 expression had no significant effect on the colony-forming capacity of transduced CT and TKO cells in vitro, except for a mild twofold reduction in the number of colonies generated by transduced TKO cells upon expression of the Socs3 hairpin #2 (Fig. 4 D). However, Socs3 repression increased the frequency of mixed-lineage colonies from both transduced CT and TKO cells at the expense of lineage-specific colonies (Fig. 4 E). Silencing of Socs3 in TKO KL cells followed by their transplantation into recipient mice led to an approximately 10-fold expansion of both TKO progenitors and TKO HSCs compared with the corresponding TKO populations infected with the Scramble vector 9 wk after tamoxifen injection of recipient mice (Fig. 4 F; compare with ∼10-fold reduction of HSC frequency at 8 wk in Fig. 1 G). However, long-term (30 wk) follow-up did not show any aberrant accumulation of Socs3 shRNA-transduced TKO cells in the BM of recipient mice compared with Scramble shRNA-transduced TKO cells (Fig. 4 G). This latest result suggests that the silencing of Socs3 in proliferative TKO HSCs is sufficient to restore their homeostasis in the BM without inducing cellular transformation.

Through its regulation of Tpo signaling, Jak2 plays a critical role in preserving HSC homeostasis in the BM (Pello et al., 2006; de Graaf and Metcalf, 2011; Anthony and Link, 2014). The expression of Socs3, a potent inhibitor of Jak2, is low in HSCs, which correlates with the absence of HSC phenotype in Socs3-deficient mice (Marine et al., 1999). Proliferative HSCs display impaired homeostasis in the BM, and our data identify a molecular connection establishing Rb proteins as a central hub for the maintenance of quiescence and homeostasis in adult HSCs through the combined repression of cell cycle genes and Socs3 (Fig. 4 H). In particular, the identification of Socs3 as a late E2f target gene that requires prolonged E2f activity for its transactivation correlates with the progressive disruption of homeostasis observed by us (Fig. 1 G) and others (Orford and Scadden, 2008) in proliferative HSCs.

Proliferative HSCs are mobilized in the peripheral circulation, and their transplantation potential is compromised (Passegué et al., 2005). Interestingly, our data also show that Sdf1 signaling, which is critical for HSC engraftment in their niche, is impaired in TKO HSCs, although the molecular mechanism remains unclear at this point. Future work will include tailored functional and molecular assays to determine the mechanism and consequence of impaired Sdf1 signaling in TKO HSCs.

Jak2 regulates the activity of multiple cytokines in different blood populations, which suggests that its regulation must be cell type specific to avoid multipopulation consequences (Meyer and Levine, 2014). Our transcriptional data indicate that E2f transcriptional output is highly cell type specific: although activation of a cell cycle program is a common consequence of E2f activity, the variability of E2f transcriptional output through hematopoietic differentiation leads to the HSC-specific activation of Socs3 and repression of Jak2 activity. Although the molecular basis for this transcriptional variability is not currently known, we hypothesize that it is influenced by the local chromatin conformation. However, more experimental evidence is needed to support this model.

Finally, myeloproliferative neoplasms are predominantly driven by constitutively active Jak2 mutants (Staerk and Constantinescu, 2012; Meyer and Levine, 2014), and the Socs3 promoter is often methylated during the course of the disease (Capello et al., 2008). Similarly, we hypothesize that activation of Socs3 expression in proliferative HSCs acts as a tumor suppressor mechanism by enforcing the differentiation of proliferative HSCs to prevent their unnecessary accumulation and potential transformation.

Rosa26-CreER^T2 Rb^Lox/Lox p130^Lox/Lox p107^−/− mice were previously described (Viatour et al., 2008), and Cre expression was induced by intraperitoneal injection of 1 mg tamoxifen (Sigma-Aldrich) in corn oil for five consecutive days. Mice were housed in the Children’s Hospital of Philadelphia (CHOP) barrier facility. All experiments were approved by CHOP Institutional Animal Care and Use Committee (protocol 969).

Cells from the BM (femur, tibia, and hips) and spleen were filtered, and red blood cells were lysed with ACK buffer (NH₄Cl/KHCO₃). The remaining white blood cells were stained with a cocktail of lineage antibodies (CD3, CD4, CD8, B220, Ter119, Mac1, and Gr1) and antibodies against Kit, Sca1, Cd48, and Cd150 (eBioscience). Flow analysis and sorting were performed on an LSR Fortessa and Mowflo, respectively. Cell cycle analysis was performed by injecting mice with BrdU 4 h before death. Cells were processed as previously described (Viatour et al., 2008) and analyzed on an analogue analyzer. Apoptosis assay was performed with the Annexin V Apoptosis Detection kit (eBioscience), according to the manufacturer protocol.

Two million unfractioned BM cells were collected from CT and TKO mice and retro-orbitally transplanted into lethally irradiated (8G) Rag1^−/− mice. 5 wk after transplantation, mice were injected with 1 mg tamoxifen for five consecutive days. 5-FU treatment (150 mg/kg) was performed weekly, starting 8 wk after tamoxifen treatment.

KL cells were infected with supernatant from PlatE cells previously transfected with retroviral vectors modulating Socs3 expression (MSCV-Socs3-Thy1; LMP-shRNA^Scramble-GFP, LMP-ShRNA^Socs3#1-GFP, and LMP-ShRNA^Socs3#2-GFP; Cui et al., 2011). 2 d after the final round of infection (two to three rounds were performed, 1 round/d), infected cells (as identified by the expression of Thy1.1 and GFP) were isolated by flow cytometry and transplanted into recipient mice together with 500,000 helper BM cells.

100 CT/TKO HSCs and progenitors and 10,000 unfractioned CT/TKO BM cells were plated in duplicate in 1 ml methylcellulose (STEMCELL Technologies) and 1 µM 4-hydroxy tamoxifen (Fig. S2 B). Colonies were counted after 9–12 d in culture. For serial replating, 10,000 cells were isolated from every condition and replated in duplicates every 9 d. Three independent experiments were performed for each assay and the result from one representative experiment is displayed. For colony assay displayed in Fig. 4 (B and E), flow-isolated KLS cells from untreated CT/TKO mice were infected with retroviruses modulating Socs3 expression (as described in the previous paragraph), and 250 Thy1⁺ (Fig. 4 B) and GFP⁺ (Fig. 4 E) cells were plated in triplicate in 1 ml methylcellulose and 1 µM 4-hydroxy tamoxifen. Colonies were counted after 9 d in culture. Three independent experiments were performed for each assay, and the mean result from the three experiments is displayed.

Spleen and bone sections were fixed in formalin and mounted, and slides were stained with hematoxylin/eosin. Cytospins were obtained by spinning 20,000 cells at 500 rpm (∼30 g) against glass slides. Cells were stained with Giemsa, as described previously (Viatour et al., 2008). All histological and morphological analyses were performed by G. Wertheim.

KLS cells, progenitors, and HSCs were flow-isolated from CT and TKO mice 2 wk after tamoxifen treatment and serum-starved for 1 h before stimulation with various amounts of Tpo for 10 min and Sdf1 for 2 min (both PeproTech). Cells were fixed, permeabilized, and stained for either p-Jak2 (Tyr1007-8)/total Jak2 and p-Fak (Tyr397)/total Fak for immunoblotting or anti–p-Stat5 (Tyr694) and p-S6 (Ser235/236; all Cell Signaling Technology) for phospho-flow. KLS cells were flow-isolated and placed in the upper compartment of a Boyden chamber. Cells were allowed to migrate in an Sdf1 gradient for 4 h. After incubation, the cells present in the lower compartment were counted by flow analysis. The data displayed in Fig. 3 (C–E) represent the mean of three independent experiments.

Molecular biology assays were performed as recently described (Tarangelo et al., 2015). Purified hematopoietic populations were sorted in TRIzol LS, and mRNA was extracted with the direct/zol RNA miniprep kit (Zymo Research), followed by DNaseI digestion to minimize the risk of genomic contamination. Reverse transcription was performed with the ProtoScript First Strand cDNA Synthesis kit (New England Biolabs, Inc.), and qPCR was performed in duplicate with SYBR Green PCR Master Mix (Thermo Fisher Scientific) on the Viia7 Real-Time qPCR system (Thermo Fisher Scientific). Data were normalized using gapdh as a reference gene. Primer sequences are available upon request. Each HSC/progenitor sample is a pool of cells collected from three mice to increase cell number.

Electrophoretic mobility shift assay was performed with increasing amount of recombinant GST-E2f1 produced in bacteria and isolated with glutathione beads. Probes were labeled with biotin and detected with streptavidin-conjugated antibody. Competition with a cold E2f consensus probe was performed by incubating increasing amounts of the cold probe before adding the biotin-labeled probe. Annealed probes were run on an agarose gel to confirm similar annealing efficiency.

The vector containing a 1.6-kb fragment of the Socs3 promoter upstream of the luciferase cDNA was described previously. Four independent luciferase assays were performed, each in duplicate (Jiang et al., 2013). For immunoblotting, unconjugated lysis samples were run on a 4–12% polyacrylamide gel for 1.5 hours at 150 V. Proteins were transferred to a nitrocellulose membrane, blocked in 5% milk for 1 h, and incubated at 4°C with primary antibody overnight. Membrane was washed in TBST three times and probed with an HRP-conjugated secondary antibody (anti–mouse, anti–rabbit, anti–goat) for 1 h. Membrane was washed again three times in TBST. Amplified signal was detected via enhanced chemiluminescence autoradiography. Antibody against Socs3 was purchased from Cell Signaling Technology. For ChIP, ACK-treated primary BM cells and 32D cells were collected into 0.5% SDS lysis buffer at a density of 50 × 10⁶ cells/ml and cross-linked in 2% formaldehyde for 10 min. Lysates were sonicated on a BioLogics Model 3000 Ultrasonic Homogenizer for 30 s at 40% power on ice to produce chromatin fragments of ∼400–700 bp. Antibodies against IgG (negative control), E2f1, and E2f3 were purchased from Santa Cruz Biotechnology, Inc.

To perform array analysis, cells were isolated in TRIzol and extracted using the QIAGEN RNA miniprep kit. cDNA was amplified before array analysis (Stanford Functional Genomics Facility, Stanford, CA). Array data generated in this publication are publicly available (GEO accession no. GSE98311). In short, GMAP/GSNAP was used to align the fragments to mouse assembly GRCm38. Normalization and differential expression analysis was performed with the Bioconductor package “DESeq2.” Microarray data were obtained from GEO with accession nos. GSE1559 (Figs. 2 K and 3 I) and GSE11253 (Fig. 3, G and H). All microarray data were processed and normalized with the RMA algorithm implemented in the “affy” R package. Heat maps for both microarrays and RNA-Seq were plotted using R packages “pheatmap” and “ggplot2.”

Fig. S1 shows CT HSCs and progenitor cells are immunophenotypically and functionally undistinguishable from WT counterparts. Fig. S2 shows impaired homeostasis of TKO HSCs. Tables S1–S7 are included in a separate Excel file. Table S1 shows gene profiling in TKO versus CT HSCs. Table S2 shows GSEA for genes up-regulated in TKO versus CT HSCs. Table S3 shows GSEA for genes down-regulated in TKO versus CT HSCs. Table S4 shows genes up-regulated only in TKO versus CT HSCs. Table S5 shows genes up-regulated only in TKO versus CT KLS cells. Table S6 shows genes up-regulated in both TKO HSCs and KLS cells versus their respective controls. Table S7 shows the cell cycle signature in TKO HSCs and KLS cells versus their respective controls.

We thank the members of the Cancer PathoBiology Division of the Department of Pathology at CHOP, as well as Peter Klein, Warren Pear, and Ken Zaret for their suggestions and their critical reading of the manuscript. We thank Nathanael Lo and Rebecca Teng for excellent technical work. We thank Susan Kaech and Frank Gonzalez for the kind gift of vectors to modulate Socs3 expression and the luciferase vector containing Socs3 promoter, respectively. We also thank the members of the Flow Cytometry and Animal Facility at CHOP.

P. Viatour is supported by the W.W. Smith Charitable Trust, a Foerderer Award, the Alex's Lemonade Stand Foundation, the Canuso Foundation, the American Cancer Society (grant RSG-16-233-01-TBE), and start-up funds from the Center for Childhood Cancer Research at the Children’s Hospital of Philadelphia.

The authors declare no competing financial interests.