The redox-sensitive transcription factor Nrf2 regulates murine hematopoietic stem cell survival independently of ROS levels

Reactive oxygen species (ROS) are required for normal cellular homeostasis because they serve as critical mediators of cytokine signaling and antimicrobial host defenses. However, excessive ROS levels can induce cellular damage that compromises cellular survival and generates genetic mutations that eventually lead to cancer.¹  ROS levels influence normal hematopoiesis; therefore, identifying the cellular processes that regulate responses to oxidative stress may improve our understanding of BM failure states and hematopoietic malignancies. In mature blood cells, ROS regulates the production and life span of erythrocytes,²  the proliferation of leukocytes in response to cytokines,^3-6  and the maturation of megakaryocytes.⁶  Hematopoietic stem progenitor cells (HSPCs) have been found to have relatively low ROS levels compared with their mature progeny.⁷  Furthermore, HSPCs may be particularly sensitive to oxidative stress, because transgenic mouse models have demonstrated that the functional loss of various genes (ATM, FOXO3a, p38 MAPK, Prdm16, and Mdm2) results in defective HSPC self-renewal, apoptosis, and differentiation associated with increased ROS.^8-13  However, the precise mechanisms by which ROS coordinate signals critical to HSPC survival and function are not clear.

Nuclear factor erythroid-2–related factor 2 (Nrf2), a bZIP transcription factor, acts as a master regulator of the cellular response to increased oxidative states.¹⁴  Nrf2 establishes a reduced intracellular redox state by inducing the expression of a myriad of cytoprotective genes, including those involved in the glutathione and thioredoxin systems, xenobiotic metabolism, and the members of the glutathione-S-transferase family.^14-21  Nrf2-deficient (Nrf2^−/−) mice are born in expected Mendelian ratios and develop normally, but succumb prematurely to multiorgan autoimmune inflammation. Moreover, responses to chemical and biologically induced oxidative stress are impaired, as evidenced by increased disease severity in models of emphysema,^17,22  asthma,¹⁶  sepsis,²³  rheumatoid arthritis,^24,25  gastritis,²⁶  and colitis.²⁷  Nrf2 has been recently implicated as key regulator of Drosophila intestinal stem cells, because its loss is associated with increased intracellular ROS and stem cell dysfunction.²⁸  Therefore, Nrf2 signaling plays an essential role in attenuating oxidative damage and may play such a role in stem cells as well. Given the critical requirement for appropriate levels of ROS during hematopoiesis and the essential role of Nrf2 in regulating ROS, we hypothesized that Nrf2 would be required for normal hematopoiesis and HSPC function.

In the present study, we characterized the HSPC compartment in Nrf2^−/− mice and demonstrated that the loss Nrf2 results in defective differentiation, decreased survival, and impaired engraftment of HSPCs after BM transplantation. Unexpectedly, Nrf2^−/− HSPCs did not display elevated basal levels of ROS, although they were more sensitive to apoptosis by induced oxidative stress. Gene-expression profiling suggested the loss of Nrf2 leads to global defects in cytokine signaling, and the administration of the exogenous G-CSF improved HSPC survival despite increasing intracellular ROS. Although increased ROS levels have been generally associated with impaired HSPC function, our findings indicate that increased levels of ROS are not always deleterious. Furthermore, our data suggest an essential role for Nrf2 in regulating HSPC survival independently of its role in regulating ROS.

Nrf2-deficient C57BL6 mice (SLC Japan) were generated as described previously.²⁹  Mice were fed an AIN-76A diet and water ad libitum and housed under controlled conditions (23 ± 2°C; 12-hour light/dark periods). All experimental protocols conducted on the mice were performed in accordance with the standards established by the US Animal Welfare Acts as set forth in National Institutes of Health (NIH) guidelines and the Policy and Procedures Manual of the Johns Hopkins University Animal Care and Use Committee.

Mice used for analysis were between 8 and 12 weeks old. Peripheral blood was obtained by retroorbital puncture in heparinized capillaries, and complete blood counts were measured on a Hemavet hematology analyzer (Drew Scientific).

Mice used for analysis were between 8 and 12 weeks old. BM was flushed from femurs and tibia with staining medium (RPMI with 2% FBS), filtered, and resuspended at a concentration of 10⁸ cells/mL in staining medium. The following Abs were used for staining the BM cells: CD34 (clone: RAM34)–allophycocyanin (APC) or FITC, FcRγ (clone: 93)–PE or Flt3 (clone: A2F10)-PE, c-Kit (clone: 2B8)–APC-Alexa Fluor 750, Sca1 (clone: D7)–PE-cyanin 7, CD150 (clone: TC15-12F12.2)–APC (BioLegend), CD48 (clone: HM48-1)–PE (all BD Biosciences) and biotin streptavidin-peridinin-chlorophyll-protein complex-cyanin 5.5 (PerCP-Cy5.5) or eFlour450-labeled lineage cocktail (CD3e, Gr1, B220, and Ter119, all eBioscience). Apoptotic cells were identified using the FITC Annexin V Apoptosis Kit (BD Biosciences). Fractions were analyzed on a FACSCalibur, 4 laser LSRII and/or sorted on a 2-laser FACSAria flow cytometer (BD Biosciences).

Cell populations were isolated by FACS as described in “Analysis of BM fractions and cell sorting.” RNA was isolated using the RNeasy Microkit (QIAGEN), and cDNA was made using the Sensiscript cDNA synthesis kit (QIAGEN) primed with oligo dT primers as per the manufacturer's protocol. Quantitative RT-PCR for selected genes was carried out using predesigned Taqman real-time primer/probes from Applied Biosystems. Expression levels were normalized to β-actin and GAPDH RNA and compared using the threshold cycle method.

CD45.2 whole BM or sorted c-Kit⁺Sca1⁺Lin⁻ (KSL) cells were obtained and transplanted into lethally irradiated (1100 rads in 2 divided doses at least 4 hours apart) CD45.1 hosts. Peripheral blood was obtained at the time points indicated and analyzed for WBC chimerism (CD45.2) and multilineage engraftment. Myeloid engraftment assessed by expression of Mac-1 and Gr1, B-cell engraftment by B220, and T-cell engraftment by CD3 on a FACSCalibur flow cytometer (BD Biosciences).

Mice were treated with single dose (100 mg/kg) of bromodeoxyuridine (5-bromo-2-deoxyuridine [BrdU]; BD Biosciences) by IP injection.¹¹  Three to 14 hours after injection, BM was isolated, filtered, labeled with Abs, fixed, and labeled with an FITC-conjugated anti–BrdU Ab using the BrdU Flow Kit (BD Biosciences) according to the manufacturer's protocol.

Five- to 6-week-old donor Nrf2^+/+ and Nrf2^−/− mice were treated with N-acetyl-L-cysteine (NAC; 100 mg/kg/d; Sigma-Aldrich) or saline solution by subcutaneous administration for 3 weeks. At the end of 3 weeks, mice were killed and the BM was isolated, analyzed, and transplanted in the recipient mice. The recipient mice were treated with NAC for 1 week after BM transplantation.⁹ 

Mice were treated with 100 μg/kg of recombinant G-CSF (Filgrastim; Amgen) by daily IP injection.

Whole BM or KSL cells were plated in methylcellulose medium supplemented with cytokines (Methocult M3434; StemCell Technologies) according to the manufacturer's directions, and colonies were scored at 12-14 days.

Statistical significance was calculated using a Student t test, and P < .05 was considered significant.

We initially analyzed the peripheral blood counts of Nrf2^+/+ and Nrf2^−/− mice and observed no differences in the total leukocyte count (7.1 vs 8.3 × 10⁴/μL, P = .41), absolute neutrophil count (1.6 vs 2.0 × 10⁴/μL, P = .30), or absolute lymphocyte count (5.3 vs 6.0 × 10⁴/μL, P = .44) between Nrf2^+/+ and Nrf2^−/− mice. However, consistent with previous reports,^6,30 Nrf2^−/− mice had reductions in RBC counts (9.40 vs 7.67 × 10⁶/μL, P = .005), hemoglobin concentration (14.2 vs 11.9 g/dL, P = .002) and platelet count (845 vs 550 × 10⁴/μL, P = .018). We studied the expression of Nrf2 and select target genes in HSPCs isolated from Nrf2^+/+ mice using quantitative RT-PCR. The KSL compartment is highly enriched for HSPCs and demonstrated the highest expression of Nrf2 and target genes, followed by committed myeloid (c-Kit⁺Sca1⁻Lin⁻) and lymphoid (c-Kit^lowSca1⁺Lin⁻) progenitors (Figure 1A). To determine whether Nrf2 is required for HSPC function, we performed competitive transplantation of BM from CD45.2 Nrf2^+/+ and Nrf2^−/− mice into CD45.1 congenic wild-type recipients. Mice receiving Nrf2^−/− BM displayed significantly lower levels of long-term (> 12 weeks) CD45.2 peripheral blood chimerism compared with mice receiving Nrf2^+/+ BM (Figure 1B). To confirm that Nrf2^−/− cells are defective during transplantation, we repeated these studies using an HSPC-enriched population and transplanted equal numbers of Nrf2^+/+ or Nrf2^−/− CD45.2 KSL cells into wild-type CD45.1 recipients. Mice transplanted with Nrf2^−/− KSL cells showed a 5-fold decrease in long-term peripheral blood chimerism, confirming that the loss of Nrf2 leads to HSPC dysfunction (Figure 1C).

To better characterize the hematopoietic defect in Nrf2^−/− mice, we examined the BM cellularity and distribution of cells in the stem and progenitor compartments. Total cellularity was increased by approximately 15% in Nrf2^−/− BM (Figure 1D), as was the frequencies of hematopoietic stem cells (HSCs)/progenitors (KSL; Figure 1E). Identification of HSCs using SLAM markers (CD150⁺CD48⁻), however, revealed a decrease in the numbers of HSCs (supplemental Figure 1A, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). Analysis of the most primitive long-term HSC (CD150⁺CD34⁻KSL) compartment showed an approximately 20% reduction in Nrf2^−/− BM (supplemental Figure 1B).³¹  This modest reduction was not statistically significant and is unlikely to account for the 5-fold difference in engraftment seen during KSL transplantation (Figure 1C). Therefore, the loss of Nrf2 leads to both qualitative and quantitative defects in HSPCs. Interestingly, there appeared to be a bias toward CD150^low lymphoid primed HSCs and away from CD150^high myeloid primed HSCs, which is consistent with the defect in myeloid development we observed (supplemental Figure 1B).^32,33 

To determine whether the changes in the BM compartment were due to increased proliferation, we compared in vivo BrdU incorporation by Nrf2^+/+ and Nrf2^−/− KSL cells (Figure 1F). There was a trend toward increased KSL proliferation in the Nrf2^−/− mice (P = .06). To determine whether the increased BM cellularity and proliferation were associated with increased cell turnover, we measured basal rates of apoptosis in the KSL compartment of freshly isolated BM from Nrf2^+/+ and Nrf2^−/− mice. Although absolute levels of apoptosis were low in both groups, there was a 65% relative increase in apoptosis in Nrf2^−/− KSL cells (Figure 1G). Therefore, the combination of normal to reduced peripheral blood counts, increased BM cellularity, and increased BM proliferation all suggest BM compensation in response to decreased HSPC survival in Nrf2^−/− mice.

Nrf2 and target gene expression has been shown to increase with myeloid maturation.³⁴  Therefore, we examined the effects of Nrf2 loss on the frequency of myeloid progenitors. In Nrf2^−/− mice, we observed a trend (P < .1) toward increased common myeloid progenitors (FcRγ^lowCD34⁺c-Kit⁺Sca1⁻Lin⁻) and decreased granulocyte-monocyte progenitors (FcRγ^highCD34⁺c-Kit⁺Sca1⁻Lin⁻) (Figure 1H), suggesting impaired differentiation. This was correlated with a significant decrease in early myeloid engraftment (at 4 weeks) in mice transplanted with Nrf2^−/− BM (Figure 1I). Furthermore, total in vitro colony formation was also impaired, with decreased macrophage, granulocyte-macrophage, and granulocyte-erythrocyte-macrophage-megakaryocyte colonies (Figure 1J).

Nrf2 loss results in increased ROS levels in most tissues. Therefore, we hypothesized that HSPC dysfunction in Nrf2^−/− cells was because of increased oxidative stress. We quantified ROS by flow cytometry using H₂-CM-DCFDA, and found that ROS levels were lowest in the long-term HSC compartment of Nrf2^+/+ mice and increased with differentiation, as described previously (Figure 2A).⁷  Unexpectedly, Nrf2^−/− BM cells displayed ROS levels similar to those of Nrf2^+/+ cells, suggesting that the functional defect in Nrf2^−/− HSPCs is not because of increased oxidative stress (Figure 2A). When we examined the expression of redox regulatory genes in KSL cells from Nrf2^+/+ and Nrf2^−/− BM, we observed that the levels of the Nrf2 targets GCLM and GCLC were, as expected, decreased and the HO-1 levels were unchanged, whereas Trx1 levels were increased (supplemental Figure 2). We suspect that the increased Trx1 expression may represent a compensatory mechanism in HSCs and may explain why ROS levels are not increased at baseline in Nrf2^−/− HSCs. Although baseline levels of ROS did not change with the loss of Nrf2 expression, we speculated that Nrf2 could be important for protecting HSCs from induced oxidative stress. To determine whether Nrf2^−/− cells handle ROS differently, we treated Nrf2^+/+ and Nrf2^−/− cells with increasing concentrations of H₂O₂. At low doses of H₂O₂, Nrf2^−/− KSL cells displayed higher levels of ROS, which is consistent with the role of Nrf2 in attenuating oxidative stress (Figure 2B). However, at the highest doses of H₂O₂, higher levels of ROS were induced in the Nrf2^+/+ KSL compartment, suggesting that the ability of Nrf2^−/− cells to tolerate high levels of induced ROS was impaired. We plated Nrf2^+/+ and Nrf2^−/− BM cells in methylcellulose following treatment with H₂O₂, and found that colony formation was significantly reduced in Nrf2^−/− cells (70% vs 40%, Figure 2C). These results demonstrated that the loss of Nrf2 increases the sensitivity of myeloid progenitors to induced oxidative stress in vitro. To examine whether Nrf2 similarly protects HSPCs from induced ROS in vivo, we exposed wild-type CD45.1 mice stably engrafted (> 20 weeks after transplantation) with CD45.2 BM from Nrf2^+/+ or Nrf2^−/− mice to sublethal radiation and quantified peripheral blood chimerism (Figure 2D). Radiation exposure is known to generate oxidative stress, and treatment with antioxidants can prevent radiation-induced injury.³⁵  Four months after irradiation, the percentage of CD45.2 peripheral blood cells was significantly decreased in mice transplanted with Nrf2^−/− BM compared with Nrf2^+/+ BM (Figure 2E). Therefore, the survival of Nrf2^−/− HSPCs is reduced after the induction of oxidative stress both in vitro and in vivo.

Because Nrf2 appears to protect HSPCs from induced oxidative stress, increased ROS during the peritransplantation period could potentially explain the decreased engraftment of Nrf2^−/− cells. Therefore, we treated donor mice with the free radical scavenger NAC for 3 weeks before BM harvesting and recipient mice for 1 week after transplantation (Figure 3A). NAC, a thiol antioxidant and GSH precursor, has been shown to rescue the loss of HSPC function caused by increased ROS,⁸  and NAC treatment can partially suppress oxidative stress in Nrf2-deficient mice.^16,23,36  NAC treatment for 3 weeks did not change basal ROS levels in HSPCs from Nrf2^+/+ or Nrf2^−/− mice, but limited the induction of ROS after ex vivo treatment with H₂O₂ (Figure 3B). We transplanted 500 KSL cells isolated from NAC- or vehicle-treated donors and subsequently treated recipient mice with NAC for 1 week after transplantation. NAC treatment failed to rescue the impaired long-term chimerism of Nrf2^−/− cells, demonstrating that their transplantation defect is not because of ROS induced during the peritransplantation period (Figure 3C).

To better understand the differentiation and survival defect of Nrf2^−/− HSPCs, we performed gene-expression studies on KSL cells isolated by FACS from Nrf2^+/+ and Nrf2^−/− mice. BM cells from 5 mice per genotype were pooled and FACS sorted in 3 replicate experiments (total of 15 mice). RNA isolated from the first sort was profiled using Affymetrix GeneChip Mouse Gene ST 1.0 chips, and changes in gene expression were verified by quantitative RT-PCR on RNA from the second and third sorts. The microarray data were analyzed using the Partek Genomics suite, and differentially expressed genes were subjected to pathway analysis using Ingenuity pathway analysis. All microarray data are available in the Gene Expression Omnibus (GEO) under accession number GSE33139. The differentially expressed genes were primarily involved in the development of the hematologic system and immune cell trafficking (Table 1). The expression of genes mediating the inflammatory response (CCL2, CCL5, CXCL10, IL1B, IL1R, IL10, and Cox-2) and myeloid development (CSF1, CSF1R, and TREM1) were reduced in Nrf2^−/− HSPCs (Figure 4A). These findings were unexpected, because previous studies showed that Nrf2 suppresses the inflammatory response and that Nrf2 deficiency results in amplified inflammatory signaling and myeloid cell activation.^16,17,23  However, these differences may be explained by our examination of HSPCs rather than mature myeloid cells.

Cytokines are known to have an important prosurvival effect on BM cells; therefore, we hypothesized that the loss of Nrf2 leads to decreased cytokine activity and cell survival. To test this hypothesis, we treated Nrf2^+/+ and Nrf2^−/− mice with G-CSF for 1 week before harvesting BM for study. G-CSF treatment resulted in a marked increase in the numbers of phenotypic HSPCs in both Nrf2^+/+ and Nrf2^−/− mice (Figure 4B). BrdU uptake by CD34⁻Flt3⁻KSL LT-HSCs was not significantly affected by G-CSF in Nrf2^+/+ and Nrf2^−/− mice (supplemental Figure 3). In contrast, the proliferation of CD34⁺Flt3⁻KSL shortening-HSCs and CD34⁺Flt3⁺KSL multipotent progenitors was significantly increased by G-CSF in Nrf2^+/+ mice, but not in Nrf2^−/− mice. These findings suggest that the increase in HSPC frequency after G-CSF is because of enhanced survival rather than changes in proliferation. To test this hypothesis, we cultured G-CSF–treated and control BM in cytokine-free medium and measured cell viability (annexin V⁻/propidium iodide⁻) in control and H₂O₂-treated cells (Figure 4C). The numbers of viable cells were equal in untreated Nrf2^+/+ and Nrf2^−/− cells; however, after exposure to H₂O₂, Nrf2^−/− HSPCs showed reduced survival that could be rescued with G-CSF. In accordance with these findings, we also found that G-CSF restored the expression of multiple genes involved in cytokine signaling and cell survival (including the antiapoptotic gene BCL2A1) to wild-type levels (Figure 4D and supplemental Figure 4).

G-CSF enhances cell survival, but also generates ROS,³  therefore, we examined Nrf2^−/−HSPC function after cytokine treatment. As expected, ROS levels increased 3-fold in the KSL compartment after G-CSF treatment, but this increase was similar to that of Nrf2^+/+ cells (Figure 5A). This degree of increased ROS was similar to the levels reported in the HSPCs of FOXO-knockout mice, which are thought to be the cause of the marked HSPC dysfunction observed in that model.⁹  However, we found that G-CSF–treated cells from Nrf2^−/− mice competed equally well during transplantation with G-CSF–treated Nrf2^+/+ HSPCs despite the dramatic increase in intracellular ROS (Figure 5B). In contrast, G-CSF treatment failed to rescue defective in vitro colony formation (Figure 5C), suggesting that Nrf2 function may be different between HSPCs and more differentiated myeloid cells.

Our findings are the first to implicate a role for Nrf2 in HSPC function and myeloid development. Surprisingly, Nrf2 does not control basal levels of ROS in BM cells; however, Nrf2^−/− BM cells are more sensitive to oxidative stress. To delineate the antioxidative function of Nrf2 in ROS-mediated HSPC dysfunction, we used an in vitro model of ROS induction with H₂O₂ and an in vivo model of ROS injury using ionizing radiation treatment (Figure 6). H₂O₂, a potent oxidizing agent, is naturally produced as a by-product of oxidative metabolism and generates ROS that play an important role in cellular signaling. Exposure of primary cells to sublethal doses of H₂O₂ has been commonly used as a model of oxidative injury. Conversely, exposure to ionizing radiation induces hydrolysis of water and thus triggers the formation of free radicals that can further damage cellular DNA and oxidize protein thiols and lipids. A study by Wang et al demonstrated that total body irradiation induces a persistent and prolonged increase in ROS in the HSC compartment.³⁷  Using both of these models, we show herein that Nrf2^−/− HSPCs are more susceptible to oxidative stress. However, the failure of NAC to rescue the engraftment of Nrf2^−/− HSPCs suggests that induced ROS is not the cause the transplantation defect we observed. It is possible that ROS detection performed using H₂-CM-DCFDA is relatively insensitive to superoxide anion- and nitrogen-containing ROS, but similar methods have been used previously to study HSPCs.⁷ 

Our findings that G-CSF treatment could enhance HSPC survival after H₂O₂ treatment and during BM transplantation despite causing a 3-fold increase in ROS suggest that the role of Nrf2 in promoting survival is more important in HSPCs than its role in reducing levels of intracellular ROS. The linkage of ROS generation (via cytokines), the detoxification of electrophiles, and the expression of antiapoptotic genes (eg, IL-10 and BCL2A1)—all via Nrf2 function—suggests an interesting fail-safe mechanism whereby HSPCs would be well equipped to survive large increases in ROS associated with sudden bursts of cytokine-driven hematopoiesis. Our finding that H₂O₂ induced the highest levels of intracellular ROS in Nrf2^+/+ KSL cells supports this idea. Therefore, optimal survival during oxidative stress requires both detoxification of ROS and activation of prosurvival pathways, and Nrf2 participates in both functions to a greater or lesser degree in a tissue-specific manner.

In the myeloid compartment, loss of Nrf2 impairs the terminal differentiation and growth of mature cells. This was demonstrated in both colony-forming assays and in vivo during early myeloid engraftment after BM transplantation. Whereas the administration of G-CSF rescued HSPC function in Nrf2-deficient mice, it failed to restore myeloid progenitor function. Interestingly, G-CSF failed to restore levels of CSF1 and TREM1, both critical factors for myeloid differentiation, to wild-type levels in Nrf2^−/− mice.^34,38,39  Because the differentiation of BM cells is associated with increases in intracellular ROS, and becaise mature myeloid cells generate ROS as part of their host defense function, we speculate that coupling of Nrf2 signaling with myeloid differentiation programs protect developing myeloid cells from ROS-induced toxicity. Further evidence for a link between myeloid differentiation in Nrf2 was recently reported in a myeloid leukemia model.⁴⁰  A precise analysis of the transcriptional targets of Nrf2 at various stages of hematopoietic development is needed to understand how Nrf2 regulates myeloid differentiation.

Nrf2 has been shown to confer chemoresistance in several malignancies,^41-44  and its role in regulating HSPC survival and myeloid differentiation suggest that it may also play an important role in myeloid neoplasms. In particular, Nrf2 may contribute to the chemoresistance of myeloid leukemia stem cells, making it an attractive therapeutic target. Conversely, Nrf2 agonists may have utility in cases of ineffective hematopoiesis and impaired differentiation, such as myelodysplastic syndrome and refractory anemia.

Finally, our findings demonstrate that increased ROS levels are not necessarily correlated with HSPC dysfunction, because we observed decreased HSPC function in Nrf2^−/− HSPCs with normal ROS levels and improved HSPC function in Nrf2^−/− BM after treatment with G-CSF despite increased ROS levels. Similarly, a recent study demonstrated that HSPC dysfunction resulting from the loss of AKT1/AKT2 function is associated with reduced ROS levels, and in that study, HSPC function could be restored by the pharmacologic induction of ROS.⁴⁵  Undoubtedly, ROS can be deleterious to cells, but given the diversity of cellular functions that generate—and in some cases require—ROS, simply detecting elevated levels of ROS in HSPCs can no longer be accepted as an adequate explanation for HSPC dysfunction. Future work is needed to correlate these findings with measures of oxidative stress and, ideally, markers of cellular damage.

The authors thank Dr Thomas Kensler (Johns Hopkins University, Baltimore, MD) and Dr M. Yamamoto (Tohoku University Graduate School of Medicine, Sendai, Japan) for providing the Nrf2^−/− mice used in this study.

This work was supported by the National Institutes of Health (grants RO1 CA140492 to S.B., P50 CA058184 to S.B., R33 AI080541 to S.B., R01CA127574 to W.M., and P01CA015396 to W.M.), the Leukemia & Lymphoma Society (to W.M.), the Gabrielle's Angel Foundation for Cancer Research (to W.M.), the Sidney Kimmel Foundation for Cancer Research (to W.M.), an American Society of Hematology Scholar Award (to A.A.M.), an American Society for Clinical Oncology Young Investigator Award (to A.A.M.), an American Association for Cancer Research-Astellas USA Fellowship (to A.A.M.), the Maryland Stem Cell Research Fund (to S.B. and W.M., and the Flight Attendant Medical Research Institute (to S.B. and A.S.).

National Institutes of Health

Contribution: A.M. and A.S. performed the experiments and analyzed the data, and all authors designed the experiments and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Shyam Biswal, PhD, Department of Environmental Health Sciences, Johns Hopkins School of Public Health, 615 N Wolfe St, Baltimore, MD 21205; e-mail: sbiswal@jhsph.edu; or William Matsui, MD, The Sidney Kimmel Comprehensive Cancer Center, Department of Oncology, Johns Hopkins University School of Medicine 1650 Orleans St, Baltimore, MD 21287; e-mail: matsuwi@jhmi.edu.