Abstract
Regulating immunity is a leading target for cancer therapy. Here, we show that the anti-tumor immune response can be modulated by the brain’s reward system, a key circuitry in emotional processes. Activation of the reward system in tumor-bearing mice (Lewis lung carcinoma (LLC) and B16 melanoma) using chemogenetics (DREADDs), resulted in reduced tumor weight. This effect was mediated via the sympathetic nervous system (SNS), manifested by an attenuated noradrenergic input to a major immunological site, the bone marrow. Myeloid derived suppressor cells (MDSCs), which develop in the bone marrow, became less immunosuppressive following reward system activation. By depleting or adoptively transferring the MDSCs, we demonstrated that these cells are both necessary and sufficient to mediate reward system effects on tumor growth. Given the central role of the reward system in positive emotions, these findings introduce a physiological mechanism whereby the patient’s psychological state can impact anti-tumor immunity and cancer progression.
Introduction
Epidemiological evidence supports a connection between the patient’s mental state and cancer survival1,2. Nevertheless, many of these studies have yielded inconsistent results3,4, and our understanding of the central neuronal mechanisms underlying the effect of emotional states on cancer is limited. Moreover, most research in this field has been focused on negative emotional states, such as stress and depression5,6, while the impact of positive mental attributes on cancer biology is largely unknown.
The brain’s reward system, specifically the dopaminergic neurons in the ventral tegmental area (VTA), constitutes a key neuronal network whose activation mediates positive emotions, expectations, and motivation7,8,9. The dopaminergic projections from the VTA to components of the limbic system are causally associated with motivated behavior and reward perception10,11. Pharmacological studies indicated a connection between reward system activity and immune modulation12,13,14, and we recently showed that reward system activity can boost antibacterial immunity15.
Immune system activity is crucial for controlling the initiation and progression of tumors. However, the immune system can also act as a double-edged sword. On the one hand, it generates effector cells, such as CD8 T cells and NK cells that can eliminate tumors16,17. On the other hand, some immune cell subsets, such as myeloid derived suppressor cells (MDSCs), act to support tumor growth by suppressing the anti-tumor immune response and by generating a favorable environment for the tumor (e.g., promoting angiogenesis)18.
Thus, given the importance of the immune system in tumor biology, and since reward system activity affects immunity, we test here the hypothesis that reward system activity could affect tumor growth. We used chemogenetics, which enables targeted neuronal manipulation19, to reveal a causal connection between reward system activity and alterations in anti-cancer immunity. We demonstrate, using two murine tumor models (Lewis lung carcinoma (LLC) and B16 melanoma), that chemogenetic activation of the reward system attenuates tumor growth. This manipulation also resulted in reduced sympathetic activity in the bone marrow, evident by attenuated noradrenaline (NA) levels. We further showed that MDSCs that develop in the bone marrow are functionally affected by the noradrenergic input. Thus, following reward system activation, MDSCs exhibited an attenuated immunosuppressive profile, which manifested in vivo by increased expression of Granzyme B by tumor CD8 T cells. By depleting and adoptively transferring MDSCs, we showed that these cells are both necessary and sufficient to mediate the effects of reward system activation on tumor growth.
Results
Specific and functional DREADD expression in the VTA
Given the central role of the immune system in fighting cancer, and given the effects of reward system activity on immunity, we hypothesized that reward system activity could also affect tumor growth. To test this hypothesis, we used Designer Receptor Exclusively Activated by Designer Drugs (DREADDs) to specifically control reward system activity. DREADDs are mutated muscarinic receptors that no longer respond to their endogenous ligand20. Instead, upon exposure to a synthetic ligand, clozapine-N-oxide (CNO), stimulatory DREADDs (hM3D(Gq)) elicit an intracellular cascade that leads to neuronal activation21. DREADDs were expressed in VTA dopaminergic neurons, using an adeno associated virus (AAV)-based vector. The virus carried a gene encoding the DREADD receptor and an mCherry fluorescent reporter. We used stereotactic injections to deliver the virus directly to the VTA, and a Cre-dependent system to ensure virus expression specified by the VTA dopaminergic neurons22 (Fig. 1a). As controls in this study, we injected mice with the same virus encoding the fluorescent reporter, mCherry, but lacking the DREADD gene (control virus). This enabled us to control for any potential local inflammatory response induced by the viral infection, effects of surgery, and CNO administration.
The efficiency of virus-expression among the VTA dopaminergic neurons was validated using immunohistochemistry. We analyzed the co-labeling for tyrosine hydroxylase (TH), a marker for dopaminergic neurons23,24, and mCherry, indicating virus expression (Fig. 1b). We found that 57.3 ± 5% of the TH+ neurons in our control mice (injected with the control virus that lacks the information for DREADDs) and 64.7 ± 3.4% of TH+ neurons in the experimental mice expressed the viral vector (P = 0.29; Supplementary Fig. 1; Student’s t-test), rendering them potentially amenable for manipulation using DREADDs. We also confirmed that DREADDs expression was restricted to the VTA as demonstrated by the lack of mCherry labeling in additional brain regions (nucleus accumbens, lateral hypothalamus and frontal cortex) (Supplementary Fig. 2). To validate neuronal activation following daily CNO injection, we used immunohistochemical analysis of c-Fos, an early activation marker. Indeed, 60.9 ± 4% of the DREADD-expressing VTA neurons were c-Fos+, compared to 34.5 ± 6% in the mice injected with the control virus (P < 0.005; Fig. 1c; Supplementary Fig. 3; Student’s t-test).
VTA activation attenuates tumor growth
To analyze the effect of VTA activation on tumor growth, we used the solid tumor model LLC. We subcutaneously (s.c.) injected the LLC cells to the lower right flank of both DREADD-expressing mice and their controls (expressing the control virus). Following the tumor cell injection, we treated mice daily with CNO to induce VTA activation (controls were also treated with CNO; Fig. 1a). After 14 days of repeated VTA activation, tumor size was reduced by 46.5 ± 17.6% and tumor weight by 52.4 ± 15.1% (P < 0.014, P < 0.003, respectively) in the VTA-activated mice when compared to the controls (Fig. 1d–f). Reward system activation was also effective in reducing tumor size in an additional murine cancer model, the B16 melanoma model. As with the LLC model, we s.c. injected mice with B16 tumor cells and activated their reward system daily. In this cancer model, reward system activation reduced tumor weight by 38 ± 11.99% when compared to controls (P < 0.03; Fig. 1g), indicating that the effects of VTA activation were not limited to a single cancer model. Yet, as the effect was more profound with the LLC, we focused on this model for the remainder of the study. With this model, we tested an additional reward system activation regimen, every other day (instead of every day). To maintain the same number of VTA activation sessions, we continued this experiment for 28 days (instead of 14 days). As shown in Fig. 1h, this less frequent manipulation regimen, also reduced tumor weight by 28 ± 11.7% (P < 0.006). Taken together, the use of DREADDs, which enables causal connections to be established25, revealed that VTA activation reduces tumor growth. However, it is unlikely that this is a direct effect, as dopamine does not cross the blood-brain barrier (BBB), raising the question of how the signal is transmitted from the brain to the tumor.
The necessity of the sympathetic nervous system for VTA’s effect on tumor
Previous findings indicate that reward system activity affects the sympathetic nervous system (SNS)15,26. Other studies demonstrate that sympathetic activity affects cancer progression27,28. Thus, we hypothesized that the SNS is likely to be involved in mediating the signal from the VTA to affect tumor growth. To directly test this hypothesis, we treated DREADD-expressing mice and their controls with 6-hydroxydopamine (6OHDA; intraperitoneal injection). 6OHDA systemically ablates catecholaminergic neurons comprising the SNS29. 6OHDA does not cross the BBB29, and thus, when injected to the periphery, its effect is considered to be limited to the peripheral SNS innervations. Accordingly, 6OHDA had no effect on the number of TH+ cells in the VTA of the tumor-bearing mice (P < 0.69; Supplementary Fig. 4). In these sympathetically ablated mice, VTA activation had no effect on tumor weight (P < 0.8825; Fig. 2a) demonstrating the requirement of the SNS in mediating VTA effects on the tumor. We further validated the SNS involvement in mediating reward system effects by treating the VTA-activated mice and controls with a blocker to one of the main NA receptors, the β-adrenergic receptor. Nadolol is a β-adrenergic blocker that does not cross the BBB30 and thus, inhibits β-adrenergic receptors in the periphery. Similarly to 6OHDA treatment, Nadolol eliminated the effects of VTA-activation on tumor weight (P < 0.4283; Fig. 2b). Thus, we conclude that sympathetic activity is required to mediate the reward system effects on tumor growth.
SNS activity is commonly associated with the stress response. Therefore, we analyzed the levels of the stress hormone, corticosterone, in VTA-activated mice and their controls. We could not detect any difference in plasma corticosterone between the two groups (P < 0.81; Fig. 2c). However, the SNS also directly innervates organs relevant to cancer biology, including some tumors27,31 and all immune organs32,33. These innervations enable local control of sympathetic activity at the target site34,35. Thus, we analyzed whether VTA activation had any effect on these local sympathetic innervations, measuring the levels of the primary SNS neurotransmitter, NA, in various target sites. VTA activation had no significant effect on plasma and spleen NA levels (determined by ELISA; P < 0.89, P < 0.16; Fig. 2d, e; Supplementary Fig. 5), though at the tumor site, we observed some, non-significant (P < 0.107) reductions in NA level (Fig. 2f; Supplementary Fig. 5). However, in the bone marrow of VTA-activated mice, NA levels were significantly reduced by 24.6 ± 0.1% (P < 0.044; Fig. 2g; Supplementary Fig. 5). We verified this finding by staining the bone marrow for TH, an enzyme that participates in NA synthesis and is expressed by sympathetic fibers in the periphery36. We found a 50.5 ± 22.3% reduction in TH expression in the bone marrow of tumor-bearing VTA-activated mice compared to their controls (P < 0.044; Fig. 2h, i).
Functional β2-adrenergic receptor on bone marrow MDSCs
The bone marrow is a particularly important site for the anti-tumor immune response37,38. During cancer progression, the bone marrow is characterized by extensive proliferation of myeloid cells, especially of MDSCs (identified by Gr-1+ CD11b+ expression)39,40. In the LLC model, this heterogeneous population, including neutrophils41, comprises a major fraction of cells in the bone marrow of tumor-bearing mice39,42 (on average 81.6 ± 1.6% of the bone marrow cell population 28 days after s.c. injection of LLC; Fig. 3a). MDSCs support tumor progression by various mechanisms including promoting angiogenesis and suppressing anti-tumor immunity42,43. Given that VTA activation specifically reduced NA levels in the bone marrow and given the high abundance of MDSCs in this site, we decided to focus on these cells. Moreover, previous studies showed that SNS activity affects myeloid population in the bone marrow44,45,46,47.
Bone marrow MDSCs express a functional β2-adrenergic receptor and are affected by VTA activation. a Representative dot plot demonstrating staining for Gr-1 and CD11b in the bone marrow of LLC tumor-bearing mice. b qPCR analysis of β2-adrenergic receptor (ADRB2) mRNA expression levels by bone marrow M-MDSCs and PMN-MDSCs of tumor-bearing mice (P < 0.004; Mann–Whitney test; n = 5, 6 in each group). c Phospho-flow analysis of intracellular CREB phosphorylation levels in bone marrow MDSCs following incubation with isoproterenol (1 µM; 15 min). Values represent median fluorescence intensity (MFI) (M-MDSCs P < 0.023, PMN-MDSCs P < 0.0004; Student’s paired t-test; mean ± s.e.m; n = 8). Data represent two independent repeats. d qPCR analysis of ADRB2 mRNA expression levels by MDSCs sorted from the bone marrow or tumor of LLC tumor-bearing mice (P < 0.015; Mann–Whitney test; n = 5). Data represent two independent repeats. eMDSCs were isolated from the bone marrow of tumor-bearing mice and incubated with isoproterenol (1 µM). Changes in gene expression were analyzed by qPCR. f qPCR analysis of TNFα mRNA expression levels by bone marrow MDSCs of tumor-bearing mice, and incubated with isoproterenol (1 µM) (P < 0.016; Mann–Whitney test; n = 5). Data represent one experiment out of two independent repeats; *P < 0.1. g Intracellular TNFα levels in bone marrow MDSCs from VTA-activated mice and controls (injected with virus lacking the DREADD gene). Values represent MFI (P < 0.027; Student’s t-test; mean ± s.e.m; n = 4). h Intracellular TNFα levels in spleen MDSCs from VTA-activated mice and controls (injected with virus lacking the DREADD gene). Values represent MFI (P < 0.039; Student’s t-test; mean ± s.e.m; n = 5, 4). I Analysis of suppression assay (as described in the methods) using bone marrow MDSCs from tumor-bearing mice following treatment with isoproterenol or vehicle (P < 0.007; Student’s t-test; mean ± s.e.m; n = 7, 6). Data represent two independent repeats. j Schematic representation of the experiment measuring the effect of VTA activation on MDSCs immunosuppressive capacity. k Analysis of suppression assay using tumor MDSCs from VTA-activated mice and controls (injected with virus lacking the DREADD gene; P < 0.004; Student’s t-test; mean ± s.e.m; n = 5, 4). Data represent two independent repeats. l CD69 expression levels on tumor CD4+ T cells from VTA-activated mice and controls, indicated by MFI (P < 0.001; Student’s t-test; mean ± s.e.m; n = 7). Data represent two independent repeats
First, we evaluated whether MDSCs could be affected by the change in NA levels, and analyzed their expression of the β2 adrenergic receptor. We focused on this receptor because it was already shown that it is commonly expressed on immune cells48, and it was consistent with our finding that the β-blocker, Nadolol, abrogated VTA-activation effects on tumor growth (Fig. 2b). We compared β2 adrenergic receptor mRNA levels expressed by two main subsets of bone marrow MDSCs: monocytic MDSCs (M-MDSCs; CD11b+ Ly6C+) and polymorphonuclear MDSCs (PMN-MDSCs; CD11b+ Ly6G+). While both MDSCs subsets expressed the β2 adrenergic receptor, receptor levels on PMN-MDSCs were 3.2-fold higher compared to M-MDSCs (P < 0.004; Fig. 3b). Next, we tested whether the β2 adrenergic receptor was not only expressed by these cells but also functionally active. NA signaling is known to activate an intracellular cascade leading to CREB phosphorylation49, thus to evaluate the activity of the receptor, we assessed pCREB levels. We exposed PMN-MDSCs and M-MDSCs to a β-adrenergic agonist (isoproterenol) in vitro. PMN-MDSCs manifested a significantly greater increase in pCREB levels compared to M-MDSCs following exposure to the β-agonist (P < 0.0004, P < 0.023, respectively; Fig. 3c; Supplementary Fig. 6), in agreement with their elevated expression of the β2 adrenergic receptor (Fig. 3b). Although PMN-MDSCs were more reactive than M-MDSCs to NA signaling, both populations responded to this signal. Moreover, VTA activation had no effect on the relative abundance of the two subpopulations (P < 0.22 for PMN-MDSCs and P < 0.64 for M-MDSCs; Supplementary Fig. 7); therefore, we continued our analysis on the total MDSC population. We found that the overall expression levels of the β2 adrenergic receptor on MDSCs isolated from the bone marrow were 2.49-fold higher compared to MDSCs isolated from the tumor (P < 0.015; Fig. 3d), suggesting that bone marrow MDSCs are more susceptible to changes in NA levels. Interestingly, VTA activation reduced NA levels mainly in the bone marrow (Fig. 2g–i).
To determine the direct effects of NA on MDSCs, we isolated these cells from the bone marrow of tumor-bearing mice and incubated them in vitro with the β-adrenergic agonist (Fig. 3e). We screened for changes in the expression of known mediators that can affect tumor growth. Specifically, we analyzed mRNA levels of the angiogenic factor VEGF42, the pro-tumorigenic cytokine TGFβ50, the immunosuppressive cytokine IL-10 as well as the regulatory factors iNOS and TNF-α50,51,52. This analysis revealed that following β-adrenergic agonist treatment, the main significant effect was a reduction in TNF-α expression levels by MDSCs (P < 0.016; Fig. 3f; Supplementary Fig. 8). TNF-α is a potent anti-tumor cytokine that has a disputed, yet central role in cancer biology53. Given that the β-adrenergic agonist decreased TNF-α expression in the bone marrow MDSCs, we expected that in the VTA-activated mice, which had reduced NA levels, MDSCs would manifest an increase in TNF-α. Accordingly, flow cytometry analysis revealed a significant increase in TNF-α levels in MDSCs from the bone marrow and spleen (P < 0.027, P < 0.039, respectively; Fig. 3g, h). This increase was specific to TNF-α, as other markers such as IFN-γ, iNOS, Arginase, IDO, PDL-1, and VEGF were not affected (P < 0.98, P < 0.44, P < 0.34, P < 0.35, P < 0.36, and P < 0.95, respectively; Supplementary Fig. 9). Moreover, CD31 mRNA levels in the tumor site were unchanged following VTA activation (P < 0.41; Supplementary Fig. 10). Taken together, following VTA activation, we observed a decrease in NA levels in the bone marrow but not in the spleen or tumor site. Thus, it is possible that this change in the bone marrow, the milieu, where the MDSCs develop, affected their subsequent functional profile at other sites.
One of the pro-tumorigenic functions of MDSCs is their ability to suppress T-cell activation and proliferation. Therefore, we treated MDSCs with the β-adrenergic agonist and incubated them with activated T cells in vitro. We found that the β-adrenergic agonist increased the immunosuppressive effect of MDSCs, manifested by a reduction in CD4 T-cells proliferation (P < 0.007; Fig. 3i). In agreement with this in vitro finding, MDSCs isolated from tumors of VTA-activated mice were less effective in suppressing CD4 T cell proliferation compared to MDSCs derived from controls (P < 0.004; Fig. 3j, k; Supplementary Fig. 11). Moreover, CD4 T cells derived from the tumor site of VTA-activated mice manifested elevated levels of the activation marker, CD69 (P < 0.001; Fig. 3l).
Necessity and sufficiency of MDSCs on VTA’s effect
Although CD69 is indicative of cell activation, it is not directly associated with effector and cytotoxic functions. Therefore, we characterized CD4 and CD8 populations in the spleen and tumor site. We evaluated their abundance and expression of the cytokines IFN-γ, TNF-α, and Granzyme B (on CD8 cells) (Supplementary Fig. 12). We found a significant increase in Granzyme B expression on CD8 T cells in the VTA-activated mice (P < 0.0159; Fig. 4a; Supplementary Fig. 13). Granzyme B is known to have a cytotoxic effect on target cells54 and thus, the increase in its levels further supports the induction of anti-tumor immune response following VTA activation. To determine whether this effect on Granzyme B levels was dependent on MDSCs activity, we depleted the MDSC population. We injected an anti-Gr-1 antibody to deplete the Gr-1-positive populations in tumor-bearing VTA-activated mice and controls. As shown in Fig. 4a, depletion of MDSCs eliminated the effect of VTA activation on Granzyme B levels on CD8 T cells (P < 0.1188). Moreover, we found that the tumor weight in these anti-Gr-1-treated mice did not differ between the two experimental groups, indicating the necessity of MDSCs in mediating these effects (P < 0.3272; Fig. 4b).
MDSCs are necessary and sufficient to mediate the effects of VTA activation on tumor weight. For all experiments in vivo, the control mice were injected with the control virus carrying the gene encoding mCherry but lacking the information for DREADD. These controls were treated, as the experimental group, with daily CNO injections. a, b Mice were daily injected (starting the day of tumor cell injection) with an anti-Gr-1 depletion antibody or an isotype control antibody, and with CNO to daily activate their VTA for 14 days. a Analysis of intracellular Granzyme-B levels in tumor CD8 T cells from VTA-activated mice and their controls. Data are shown as fold change in Granzyme B antibody staining (Gr-1 depletion (P < 0.1188; n = 8, 7), isotype control (P < 0.0159; n = 10, 9); Student’s t-test; mean ± s.e.m; NS not significant). Data represent two independent repeats. b Weights of tumors isolated from DREADDs-expressing mice and their controls. Data are shown as fold change in tumor weight (Gr-1 depletion (P < 0.3272; n = 11, 9), isotype control (P < 0.0202; n = 6, 9); Student’s t-test; mean ± s.e.m; NS not significant). Data represent two independent repeats. c Schematic representation of the experimental design used to determine whether the difference in MDSCs function induced by VTA activation is sufficient to recapitulate the effect of the VTA on tumor weight. d Representative tumors extracted from mice inoculated with naive LLC cells together with MDSCs isolated from mice that expressed the control virus (treated daily with CNO as the experimental group; upper row), versus LLC tumors inoculated with MDSCs isolated from VTA-activated animals (lower row). e Weights and f size of tumors extracted from mice inoculated with naive LLC cells together with MDSCs isolated from control animals (injected with virus lacking the DREADD gene), or isolated from VTA-activated mice (tumor weights P < 0.019, tumor size P < 0.008; Student’s t-test; mean ± s.e.m; n = 6, 9). Data represent two independent repeats
Although we do not assume that MDSCs are the only cell population affected by VTA activation, a change in these cells might be sufficient to induce the reduction in tumor weight. To directly address this question, we performed an adoptive transfer experiment; MDSCs were isolated (based on their expression of Gr-1 and CD11b) from tumors of VTA-activated mice and their controls (carrying the control virus and treated with CNO). The cells were then co-injected along with new tumor cells into naïve recipient mice. As the recipient mice were injected with an equal number of LLC cells and MDSCs (isolated from control or VTA-activated mice), any difference in the evolving tumors could be directly attributed to functional changes in the transferred cells (Fig. 4c). After validation that the transferred MDSCs could survive in the recipient mice (Supplementary Fig. 14) and that the tumor cells were not immediately rejected in our experimental paradigm (P < 0.24; Supplementary Fig. 15), we compared the tumors in both groups. We found that when MDSCs were derived from VTA-activated mice, the tumor size was reduced by 42.9 ± 13.4% and tumor weight was reduced by 43.6 ± 16.3% compared to the controls (P < 0.008, P < 0.019, respectively; Fig. 4d–f), indicating that the effect on MDSCs was sufficient to mediate the VTA effects on tumor growth. Taken together, the transfer and depletion experiments demonstrated that MDSCs are both sufficient and necessary to mediate VTA-induced effects on tumor growth.
Leave a Reply