Environmental and Genetic Activation of Hypothalamic BDNF Modulates T-cell Immunity to Exert an Anticancer Phenotype

Environmental factors profoundly influence carcinogenesis, and epidemiologic reports indicate a connection between psychological stress and cancer progression (1–3). Previous research linking stress to cancer growth implicates activation of the sympathetic nervous system (SNS) and the hypothalamic–pituitary–adrenal axis (HPA; refs. 4–6); however, the precise molecular mechanisms by which stress regulates cancer progression are not fully understood. Most research has focused on distress, so the impact of eustress, which is associated with health benefits, has not as yet been well investigated. Both the SNS and the HPA axes regulate lymphocyte development and differentiation, and neurologic control of immune homeostasis is becoming more widely recognized as a major regulator (7–9). Therefore, identification of the mechanisms by which different stresses influence T-cell homeostasis in secondary lymphoid tissue (SLT), and the T-cell's ability to recognize tumors and halt cancer progression, are of significant interest to investigators and clinicians in cancer immunotherapy (10).

To study environmental and psychosocial effects on cancer progression, our laboratory uses environmental enrichment (EE) as a eustress model. EE is a housing environment for laboratory animals, which, in contrast with the standard-environment (SE) cages used in most biomedical research, is socially, physically, and cognitively stimulating. EE has profound impacts on brain structure, function, and progression of neurologic diseases (11, 12). We have reported that EE leads to a distinct metabolic phenotype in mice characterized as decreased adiposity, elevated energy expenditure, induction of beige cells in white fat, enhanced insulin sensitivity, improved adipokine profile, and resistance to diet-induced obesity (13). In addition, an EE inhibits tumor progression in an implantation model of melanoma as well as a genetic model of colon cancer (14). Recent reports from others have shown the anticancer effects of an EE in orthotopic and spontaneous tumor models of breast, colon, pancreatic, and melanoma cancers (15, 16). We have identified one mechanism underlying the effects of an EE on metabolism and cancer, the activation of the hypothalamic–sympathoneural–adipocyte (HSA) axis (14). The environmental stimuli provided by an EE upregulate BDNF expression in the hypothalamus and subsequently elevate sympathetic tone preferentially to the white adipose tissue, which has at least two consequences: induction of beige cells and suppression of leptin production and release. The robust drop of leptin in the circulation plays an important role in EE-induced cancer inhibition. However, an EE is associated with other changes, such as a modest but significant increase of corticosterone, and enhanced CD8 T-cell cytotoxicity (14). The contribution of an adaptive T-cell immune response, however, is unknown in the context of an EE. Here, we investigated how EE modulated T-cell homeostasis in SLT, the role of CD8 cytotoxic T lymphocytes (CTL) in EE-induced cancer inhibition, and the neuronal substrates of the brain–immune interactions induced by EE.

Male 3-week-old C57/BL6 mice were purchased from Charles River and randomized to live in enrichment or standard laboratory conditions (5 mice per cage). For enrichment housing, mice were housed in groups in cages of 1.5 m × 1.5 m × 1.0 m (10–20 mice per cage) or 73 cm × 41 cm × 46 cm (5 mice per cage) supplemented with running wheels, tunnels, igloos, huts, retreats, wood toys, a maze, and nesting material in addition to standard lab chow and water (17). All mice were housed in temperature- (22–23°C) and humidity-controlled rooms with food and water ad libitum. We carried out all mouse experiments in compliance with the regulations of the Ohio State University Institutional Animal Ethics Committees.

The mouse B16-F10 melanoma cell line was purchased from ATCC (ATCC CRL-6475) in 2014 and the culture conditions followed ATCC's instruction. The B16-F10 cell line was expanded three passages, and the frozen stock was thawed and used for implantation. We housed mice in their respective environments for 5 weeks and then intraperitoneally injected 0.2 mg of anti-CD8 (BioXcell, clone YTS 169.4.2) or IgG (BioXcell) in 200 μL PBS. The day after antibody injection, B16-F10 melanoma cells were subcutaneously implanted on the flank (1×10⁵ cells/mouse, n = 10 per group). The antibodies were injected once per week until sacrifice 18 days after melanoma implantation. The tumors were dissected away from neighboring tissues, measured, and weighed.

Spleens and lymph nodes were collected in Eppendorf tubes containing RPMI media. We mechanically dissociated spleens or lymph nodes through a 70-μm strainer to obtain single-cell suspensions. Red blood cell (RBC) lysing buffer were used to lyse RBCs. After washing with PBS, cells were counted using hemocytometer. Cells (2×10⁶) were pipetted for a flow panel. Conjugated antibodies for CD3-PercpCy5 (Cat. #55116, 145-2C11), CD4-APCH7 (Cat. #560181, GK1.5), CD8-V450 (Cat. #560469, 53-6.7), CD44-FITC (Cat. #553133, IM7), and CD122-PE (Cat. #553362, TM-Beta1) were purchased from BD Biosciences. TCRβ-PECy (Cat. #25-5961, H56-597) was purchased from eBioscience. For regulatory T cell (Treg) flow cytometry, we used the Mouse Regulatory T-cell Staining Kit from eBioscience (Cat. #88-8111-40) containing CD4-FITC (RM4-5), CD25-APC (PC61.5), and Foxp3-PE (FJK-16s). All flow cytometry was performed using a BD LSRII. The results were analyzed with FlowJo.

The recombinant adeno-associated virus (rAAV) plasmid contains a vector expression cassette consisting of the CMV enhancer and chicken β-actin (CBA) promoter, WPRE (posttranscriptional regulatory element of woodchuck hepatitis virus) sequence, and bovine growth hormone poly-A region flanked by AAV inverted terminal repeats. Human BDNF cDNA was inserted into the multiple cloning sites between the CBA promoter and WPRE sequence. Enhanced green fluorescent protein (EGFP) was cloned into the rAAV plasmid as a control. rAAV serotype 1 vectors were packaged, purified, and the vectors were adjusted to 2×10¹³ vg/mL in PBS for injection.

Five-week-old male C57BL/6 mice were randomly assigned to receive AAV-BDNF (n = 5–7) or AAV-GFP (n = 5). Mice were anesthetized with ketamine/xylazine and secured via ear bars and incisor bar on a Kopf stereotaxic frame. A midline incision was made through the scalp to reveal the skull, and two small holes were drilled into the skull with a dental drill above the injection sites (−1.2 AP, ± 0.5 ML, −6.2 DV, mm from bregma). rAAV vectors were injected bilaterally into the hypothalamus (0.5 μL per site) at a rate of 0.1 μL per minute using a 10-μL Hamilton syringe attached to Micro4 Micro Syringe Pump Controller (World Precision Instruments). At the end of infusion, the syringe was slowly raised from the brain, and the scalp was sutured. Animals were placed back into a clean cage and carefully monitored after surgery until recovery from anesthesia. After 3 or 5 weeks, those mice were sacrificed.

AAV vectors containing microRNA targeting BDNF (miR-Bdnf) and the scramble control (miR-scr) were reported previously (14). We randomly assigned 6-week-old male C57BL/6 mice to receive AAV-miR-Bdnf (n = 15) or AAV-miR-scr (n = 15). We injected 0.7 μL of AAV vectors (1.4×10¹⁰ per site) bilaterally into the hypothalamus at the stereotaxic coordinates described above. Seven days after surgery, each vector-injected group was split to live in enriched (n = 8 miR-Bdnf, n = 8 miR-scr) or control housing (n = 7 miR-Bdnf, n = 7 miR-scr). Mice were sacrificed after 5 weeks in the EE.

We randomly assigned 40 male C57BL/6 mice, 3 weeks of age, to live in an EE or control cages supplied with propranolol (Roxane Laboratories, Inc. Cat# 0054-3727-63) in drinking water (0.5 g/L).

Mifepristone (Abcam Inc., Cat #ab120356) was dissolved in 0.9% NaCl, 0.25% carboxymethylcellulose, 0.2% Tween 20. Male C57BL/6 mice were housed in an EE or control housing and received daily oral gavage of mifepristone (200 mg/kg body weight) or vehicle control for 1 week.

All data were compared using the Student t test and plotted as mean ± SEM in column charts.

We have previously reported that CTLs from spleens of EE mice display increased cytotoxicity against B16 melanoma cells in vitro (14). We therefore hypothesized that increased EE-CTL cytotoxicity and tumor infiltration contributed to the slower tumor growth observed in EE mice. To test this hypothesis, we eliminated CTLs prior to melanoma implantation to SE and EE mice using CD8-depleting antibody, with IgG as a control. Briefly, 3-week-old C57/B6 (B6) mice were randomly assigned to SE or EE cages for 5 weeks (Fig. 1A), which is the approximate time when EE-associated metabolic changes have been observed. EE mice weighed less than SE mice (Supplementary Fig. S1A), and serum leptin concentration was reduced by ∼70% in EE mice (Supplementary Fig. S1B). Next, EE and SE mice were randomly injected intraperitoneally with either 200 μg of CD8-depleting antibody or IgG control. Twenty-four hours after the first dose of antibody, all mice were subcutaneously injected in the hip flank with 1 × 10⁵ B16 melanoma cells. We confirmed that CTLs were depleted in the blood (Supplementary Fig. S1C). Every 7 days, mice were reinjected with their respective antibody. Eighteen days after melanoma implantation, all mice were sacrificed.

Consistent with our previous findings, EE mice receiving IgG (EE-IgG) showed reduced tumor mass by ∼80% compared with that of SE mice receiving IgG (SE-IgG; Fig. 1B). Although no significant difference was observed between SE mice receiving CD8 (SE-CD8) and IgG (SE-IgG), the CD8 T-cell depletion significantly attenuated EE-induced tumor inhibition (EE-CD8 compared with EE-IgG; Fig. 1B), suggesting an essential role of CD8 T cell in EE-induced tumor inhibition. EE-CD8 tumors were approximately 150% larger than EE-IgG tumors (433 ± 91 mg vs. 174 ± 67 mg).

Within the tumor microenvironment, the mononuclear cells per milligram did not significantly differ between SE-IgG and EE-IgG tumors (Supplementary Fig. S1D), nor did the frequency of tumor-infiltrating CD8 CTLs (Supplementary Fig. S1E). We next measured the surface expression of CD27 and PD-1, both of which are targets for clinical immunotherapy (18, 19). CD27 is a member of the tumor necrosis factor receptor superfamily and is expressed on both naïve and long-lived CD8 T cells, but is lost on terminally differentiated effector CTLs (20–22). In the IgG control groups, EE tumors contained a lower frequency of CD27⁺ effector CTLs when compared with CTLs within SE tumors (Fig. 1C). However, the frequency of CTLs expressing PD-1 or the mean fluorescence intensity of PD-1 expression on CTLs within EE tumors did not change compared with CTLs within SE tumors (Supplementary Fig. S1F). The frequency of Tregs, a helper T cell that inhibits CTLs, was reduced in EE tumors (Fig. 1D).

Total splenocyte (Supplementary Fig. S1G) and total splenic T-cell numbers were reduced in EE mice (Fig. 1E), but the SLT contained T-cell features consistent with greater tumor reactivity. Polarization of CD4 cells toward the helper T cell type 1 (Th1) subset can be defined by surface expression of chemokine receptors CXCR3 and CCR4 (23, 24), and these cells promote protective immunity and induce CTL-mediated tumor eradication (25). CD4 T cells from EE-IgG spleens were polarized toward a Th1 phenotype (Fig. 1F), and this Th1 polarization persisted after CD8 T-cell depletion. The frequency of CTLs was increased in the lymph nodes of EE mice (Fig. 1G), and a greater proportion expressed a memory phenotype CD44^hiCD122⁺ (Supplementary Fig. S1H).

We next sought to determine the early effects of EE exposure on T-cell homeostasis that may contribute to the anticancer phenotype. We randomly assigned 3-week-old B6 mice to either SE or EE for 1 or 4 weeks. After 4 weeks of exposure to an EE, some characteristics of experiencing the EE were confirmed, including the upregulation of hypothalamic BDNF, the key mediator of EE-induced anticancer and anti-obesity effects (refs. 13, 14, 26; Supplementary Fig. S2A), and reduced body weight (Supplementary Fig. S2B). EE mice showed fewer total splenocytes at 4 weeks (Supplementary Fig. S2C).

Exposure to an EE for merely 1 week was sufficient to induce a distinct SLT phenotype in T-cell populations. Consistent with observations from our tumor experiment, total T-cell numbers showed a trend of reduction in spleens of EE mice at 1 week (P = 0.08), and the reduction reached statistical significance by 4 weeks (Fig. 2A). However, lymph nodes in EE mice were measurably larger after 2 weeks (Fig. 2B) and contained substantially more lymphocytes (Fig. 2C). We used flow cytometry to characterize changes in T-cell populations (Fig. 2D). In EE SLT, the frequency of single positive (SP) CD8 CTLs increased, whereas the frequency of SP CD4 T cells declined, resulting in a depressed CD4:CD8 ratio, and the changes of these populations persisted at 4 weeks (Fig. 2E). This change of SP CD4:CD8 ratio was also found in lymph nodes at both time points (Fig. 2F). In contrast to the tumor experiment, we did not observe the change in the frequency of memory CTLs in SLT (Supplementary Fig. S2D) suggesting that the presence of the tumor might be responsible for increased memory CTLs in EE mice.

Hypothalamic BDNF is a key mediator linking EE to the leanness and altered adipokine profile (13, 14). We next investigated whether hypothalamic BDNF also mediated the EE-induced T-cell population changes. Mice were randomized to receive stereotaxic injection of recombinant adeno-associated virus (rAAV) vector containing BDNF or a green fluorescent protein (GFP) vector as a control to the hypothalamus and maintained in SE housing. GFP fluorescence confirmed that the transgene expression was largely confined to the arcuate nucleus, the ventromedial, and the dorsomedial nuclei of hypothalamus (Supplementary Fig. S3A). The overexpression of BDNF was confirmed by qRT-PCR 5 weeks after rAAV injection (Supplementary Fig. S3B). Three weeks after rAAV injection, rAAV-BDNF-mice weighed the same as rAAV-GFP-mice (Supplementary Fig. S3C), but had fewer splenocytes (Supplementary Fig. S3D), similar to the changes observed in EE. A feature of long-term EE exposure is the reduced body weight due to the decrease of adiposity (13). In order to investigate whether the immune modulation was dependent on metabolic regulation, we repeated the hypothalamic BDNF overexpression experiment and examined the immune effects 5 weeks after rAAV injection when significant changes in body weight developed between the two groups (Supplementary Fig. S3C).

The distribution of T cells within the SLT of BDNF mice showed a similar pattern to EE mice, despite the mice being housed in SE. Total splenic T-cell numbers were reduced by 5 weeks (Fig. 3A). Within the SLT, there was a significant increase in CTL frequency within the T-cell population both at 3 and 5 weeks (Fig. 3B and C; Supplementary Fig. S3E, S3F, and S3G). Similar to EE, there were no changes in chemokine receptors CXCR3, CCR6, and CCR4 (Supplementary Fig. S3H), nor differences in IL4, IFNγ, or IL17 by T-helper cells following CD3/CD28 stimulation (Supplementary Fig. S3I). Overall, the results suggested that hypothalamic overexpression of BDNF could mimic the effects of an EE on T-cell populations, and the modulation was independent of changes in body weight.

To confirm whether hypothalamic BDNF expression is necessary for the EE-induced increase in CTL frequency in SLT, we injected mice with either a rAAV vector expressing a microRNA targeting mouse BDNF (miR-Bdnf) or a microRNA targeting a scrambled sequence (miR-scr) that does not interfere with any known genes as a control. The miR-Bdnf vector was fully characterized previously (14). Mice receiving each rAAV vector were then divided, to live in either SE or EE housing for 5 weeks. We confirmed the knockdown of hypothalamic BDNF by quantitative RT-PCR (Supplementary Fig. S4A). miR-Bdnf blocked the EE-induced weight loss (Supplementary Fig. S4B).

Absolute splenocyte (Supplementary Fig. S4C) and T-cell numbers (Fig. 4A) were increased in miR-Bdnf-SE compared with miR-scr-SE housing. Compared with miR-scr-SE mice, the CD4:CD8 ratio in SLT was reduced in miR-scr-EE mice, whereas miR-Bdnf inhibited this change in T-cell frequency that is closely associated with EE (Fig. 4B and C). Collectively, these results showed that knockdown of hypothalamic BDNF prevented the EE-induced changes in CTL frequency within SLT.

We have previously demonstrated that intact SNS signaling is required for EE's metabolic and anticancer phenotype (13, 14). We used the similar approach using the nonspecific β-blocker (βB) propranolol to investigate the role of SNS in the EE-induced changes in T-cell homeostasis in SLT. We repeated the experimental design by assigning 3-week-old mice to live in SE or EE cages for 1 week, and half of the SE or EE mice were supplied with propranolol (0.5 g/L) in the drinking water, and the other half was supplied with regular water. This dose of propranolol is sufficient to block EE-induced anticancer effects and metabolic regulation (13, 14).

Both propranolol-treated SE and EE mice had similar numbers of T cells (Fig. 5A). However, propranolol completely blocked the EE-induced change in CD4 and CD8 populations in the SLT (Fig. 5B and C).

The HPA axis is mildly activated in EE (14). Thus, we investigated the role of the HPA axis using mifepristone, an antagonist of cortisol receptors (27). We randomly assigned 3-week-old mice to SE or EE housing and administered either mifepristone or solvent as vehicle control by daily oral gavage for 1 week until sacrifice. Both SE and EE mice receiving mifepristone had similar numbers of T cells (Fig. 6A), but mifepristone completely blocked the EE-induced reduction of the CD4:CD8 T-cell ratio in the SLT (Fig. 6B and C).

In this study, we investigated the contribution of T-cell immunity to the antitumor effect observed in EE mice and defined changes in SLT T-cell homeostasis induced by EE. Our data showed that CTLs were necessary for the anticancer phenotype of an EE in an orthotropic melanoma model and that the tumor microenvironment contained more tumor-responsive T cells in mice living in EE. The EE induced early changes in T-cell homeostasis, including a reduction of the total number of splenic T cells and a shift to CTLs (decreased CD4:CD8 ratio) in SLT. Mechanistically, overexpressing BDNF or knockdown of BDNF in the hypothalamus reproduced or blocked the key features associated with an EE, respectively, including reduced numbers of T cells in SLT and a depressed CD4:CD8 ratio. All primary and secondary immune organs receive substantial sympathetic innervation from sympathetic postganglionic neurons (28, 29). Norepinephrine (NE) binds to either α- or β-adrenergic receptors expressed on immune or lymphoid stromal cells, which results in changes in gene expression of various immune cell–derived factors (30, 31). Moreover, the HPA axis is also involved in neuro-immune cross-talk (32). Blockade of the SNS and HPA axes eliminated the two key immunomodulatory features of an EE in SLT: reduced numbers of T cells and a depression of the CD4:CD8 ratio, suggesting that both axes contribute to the EE-induced immune phenotypes.

Our previous studies show that metabolic changes are associated with the tumor resistance found after living in an EE (16). In particular, tumor growth is linked with higher β-AR activity in white adipose tissue, leading to reduction of circulating leptin concentrations (14). Blockade of the leptin drop significantly attenuates the EE-induced tumor inhibition, although without full prevention. Our results here demonstrated that T-cell immunity played a critical role in the antitumor effects of an EE. Depletion of CTLs also partially but significantly blocked the reduction of tumor mass in EE, suggesting that the antitumor effects at least partially depend on CTLs. Although CTL expression of PD-1, an immune inhibitory molecule, was unchanged, CTL expression of CD27 was significantly reduced, indicating that the tumor microenvironment in EE mice contains more activated CTLs. In addition, fewer Tregs were found in the tumors of EE mice despite more Tregs in SLT (not shown). No changes of Treg percentage or memory CTL percentage observed in the tumor-bearing mice were found in EE animals in the absence of tumor, suggesting that these differences might be due to T cells responding differently in the context of health and disease. The EE decreased significantly enlarged lymph node mass and led to a higher proportion of CD8 CTLs and lower proportion of CD4 T cells, consistent with our previous finding of enhanced CD8 cytotoxicity in vitro (14).

Overall, the new findings suggest that an EE induces an improved immune microenvironment of tumor, consistent with reports that the increased proportions of CD8⁺ T cells and decreased Tregs correlate with a better prognosis of cancer (33, 34). Taken together, both metabolic regulation and immune modulation contribute to the robust inhibition of melanoma progression induced by an EE (Fig. 7), supporting the notion that combination therapies targeting multiple cancer-promoting capabilities are likely more effective and durable (35). We have previously reported that EE also enhances natural kill (NK) cells activity (14). A recent report from other authors has shown that NK cells play a critical role in the anticancer effect of an EE in a mouse model of glioma (36). Investigations on how an EE regulates additional immune cell populations and their roles in cancer inhibition are under way.

Hypothalamic BDNF expression is necessary for mediating the cancer and metabolic phenotypes seen after living an EE (13, 14). Therefore, we explored whether hypothalamic BDNF could also mediate the EE-induced peripheral CD4:CD8 ratio change. We observed that rAAV-mediated overexpression of BDNF in the hypothalamus leads to a decreased CD4:CD8 ratio in SLT and that depletion of hypothalamic BDNF could eliminate the CD4:CD8 ratio change. These observations indicate that manipulation of a single molecule in the brain could regulate T-cell homeostasis. And this mechanism could potentially be harnessed to strengthen the efficiency of immunotherapies.

Cross-talk between nervous system and immune system has been intensively investigated. However, most research has used models causing distress, such as social isolation, repeated social defeat, and chronic stress models that often are associated with impairment of immune functions (37–40). In contrast, scarce attention has been paid to eustress that is associated with enhanced adaption to environmental demands, improved general health, and resistance to multiple diseases. An EE is thought to be a good model to study eustress response (41). For example, in contrast to repeated social defeat (42), which stimulates cancer growth through stress-induced dysregulation of inflammatory mediators and the tumor microenvironment (43), EE suppresses cancer progression through enhanced immunity as well as regulation of whole body metabolism and cellular metabolism. Our data showed that some immune phenotypes occurred prior to changes of adiposity or weight in response to EE, suggesting that the immune response and the metabolic response might be parallel. However, the precise relationship between the two phenotypes and whether EE could influence immune cell metabolism require further investigations.

Environmental stress is an important confounding factor of preclinical studies. A recent report demonstrates that thermoneutral ambient temperature (30°C) increases sensitivity to chemotherapy in a pancreatic cancer mouse model and the underlying mechanism is the activation of β2-adrenergic receptor in tumors (44). In our EE experiments, all mice were housed at 22°C. It is possible that the extra bedding in EE might provide slightly more additional warmth. However, the EE significantly elevated NE in adipose tissue and showed a trend of increased NE in serum (13, 14). Therefore, EE would not antagonize β-adrenergic receptors in tumor cells; on the contrary, there might be a risk of activation. However, the effects of an EE on metabolism (drop of leptin, increase of adiponectin, and enhanced insulin sensitivity) and immune functions (T cells and NK cells) overweigh greatly the potential negative effect of SNS activation directly on tumor cells. Our preliminary data show that an EE under thermoneutral conditions led to the same metabolic effects (leanness, drop of leptin, and induction of beige cells), even to a greater degree, than that at 22°C (data not shown). This finding suggests that removing cold stress and the associated SNS activation could amplify the HSA axis that elevates the SNS tone preferentially to white fat contributing to the EE-induced tumor resistance. We are investigating other effects associated with an EE, including immune modulations at thermoneutrality.

In summary, an EE beneficially modulated metabolism and immunity, leading to inhibition of tumor progression. Previous research identified hypothalamic BDNF as a key mediator linking social and physical environment to metabolism and cancer progression (13, 14). Our findings showed that hypothalamic BDNF also played an important role in EE-induced immune regulation, suggesting the possibility of manipulating a single molecule in the brain to modulate multiple peripheral systems for therapeutic applications.

No potential conflicts of interest were disclosed.

Conception and design: R. Xiao, S.M. Bergin, S.D. Scoville, J. Yu, M.A. Caligiuri, L. Cao

Development of methodology: R. Xiao, S.M. Bergin, S.D. Scoville

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Xiao, S.M. Bergin, W. Huang, E.-J.D. Lin, K.J. Widstrom

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Xiao, S.M. Bergin, R.T. Judd, M.A. Caligiuri, L. Cao

Writing, review, and/or revision of the manuscript: R. Xiao, S.M. Bergin, R.T. Judd, S.D. Scoville, J. Yu, M.A. Caligiuri, L. Cao

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Huang, X. Liu, R.T. Judd, E.-J.D. Lin, K.J. Widstrom

Study supervision: L. Cao

The study was supported in part by NIH grants CA166590, CA178227, and CA163640 to L. Cao and NIH grants CA163205 and CA068458 to M.A. Caligiuri.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.