Neural regulation of hematopoiesis, inflammation and cancer

Abstract

Although the function of the autonomic nervous system (ANS) in mediating the “flight-or-fight” response was recognized decades ago, the crucial role of peripheral innervation in regulating cell behavior and response to the microenvironment has only recently emerged. In the hematopoietic system, the ANS regulates stem cell niche homeostasis, regeneration and fine-tunes the inflammatory response. Additionally, emerging data suggest that cancer cells take advantage of innervating neural circuitry to promote their progression. These new discoveries outline the need to redesign therapeutic strategies to target this underappreciated stromal constituent. Here, we review the importance of neural signaling in hematopoietic homeostasis, inflammation and cancer.

Introduction

The crosstalk between central nervous system (CNS) and periphery to maintain basal and stress-related body homeostasis is largely exerted by the hypothalamic-pituitary-adrenal (HPA) axis and the autonomic nervous system (ANS). The ANS is comprised of two branches, the sympathetic division emerging from the thoraco-lumbal spinal cord, and the parasympathetic division derived from cranial nerves and the sacral spinal cord (Figure 1). Functionally, the sympathetic nervous system (SNS) prepares the body for alert situations by activating cardio-vascular organs increasing blood flow to skeletal muscles and lungs, and simultaneously suppressing gastro-intestinal activity. By contrast, the parasympathetic nervous system (PNS) is more active in periods of rest and decreases muscle contractility and heart rate, while increasing gastrointestinal function.

Figure 1. Autonomic neural circuits.

The autonomous nervous system (ANS) is comprised of a sympathetic division (right), emerging from the thoraco-lumbal spinal cord, and a parasympathetic division (left) derived from cranial nerves and the sacral spinal cord. The ANS requires preganglionic and postganglionic neurons, forming synapses in the autonomic ganglion to connect the central nervous system to its targets. (I) Sympathetic nerve fibers in the bone marrow locally release neurotransmitters that activate adrenergic receptors expressed on niche cells and thus regulate the maintenance and retention of hematopoietic stem and progenitor cells in the bone marrow. (II) Neural circuits comprised of cholinergic and adrenergic neurons, splenic T cells and splenic macrophages regulate the innate immune response. Activation of the HPA axis may also counter-balance inflammation through glucocorticoids released from the adrenal cortex.

In recent years, the ANS emerged as an essential microenvironmental component of a variety of vertebrate tissues, regulating homeostasis and influencing various pathologies. In the liver, sympathetic and parasympathetic innervations regulate both apoptosis and regeneration following partial hepatectomy (Kiba, 2002). Autonomic innervation of the endocrine pancreas also controls proliferation of pancreatic β cells in response to insulin resistance, modulating pancreatic islet mass under pathological conditions (Kiba, 2004, Imai et al., 2008). Brown adipose tissue (BAT) regulates thermogenesis and energy expenditure via CNS-regulated sympathetic nerve activity (Morrison et al., 2014). Additionally, peripheral nerves regulate the function of multiple secretory organs. For example, innervation of the prostate by both sympathetic and parasympathetic nerves regulates epithelial cell growth and secretion by exerting contraction on smooth muscle cells (Lujan et al., 1998, Ventura et al., 2002).

Like most, if not all vascularized tissues, the sympathetic branch of the ANS highly innervates the bone marrow, affecting both hematopoiesis and bone formation. Initial studies revealed that bone formation via osteoblast function was regulated by central leptin-mediated signals that activated the SNS and acted on β2 adrenergic receptors on osteoblasts (Takeda et al., 2002). Further studies showed that the mobilization of hematopoietic stem cells (HSCs) into peripheral blood depended on adrenergic modulation of the HSC microenvironment (Katayama et al., 2006). Here, sympathetic signals traveling through adrenergic nerve fibers of the bone marrow, locally release neurotransmitters that activate adrenergic receptors expressed on niche cells. Moreover, further studies have revealed that leukocyte trafficking is influenced by a central molecular clock that coordinates circadian secretion of noradrenaline from nerve terminals. Conveyed through the hypothalamic-pituitary adrenal axis or the vagus and adrenergic nerves, these signals induce release of neurotransmitters in the adrenal medulla or the spleen (Mendez-Ferrer et al., 2008, Scheiermann et al., 2012). The ANS thus serves as a major modulator of immune responses by exerting both pro and anti-inflammatory functions through influence on immune effector cells. In this perspective, we review the neural regulation of hematopoietic stem and immune cell behavior and discuss its emerging role and implications in malignant pathology.

Neural regulation of hematopoiesis

Hematopoietic homeostasis is maintained through coordinated proliferation, self-renewal and differentiation of HSCs in the bone marrow leading to the development of the entire blood system. HSCs are the only cell type capable of producing all blood cell lineages throughout life, and their behavior is tightly regulated by the local microenvironment, also termed the niche (Morrison and Scadden, 2014). Regulatory elements within this specialized niche are derived from surrounding stromal cells and innervating sympathetic nerve fibers that coordinate HSC quiescence, a prerequisite for stem cell maintenance (Frenette et al., 2013). To maintain homeostasis, HSCs are kept in a dormant/quiescent state within the niche, however, in response to stress, such as infection or radiation, they become activated in order to replenish damaged blood cells and restore homeostasis (Wilson et al., 2009). As HSCs are highly dependent on their microenvironment for normal function, damaging effects to the bone marrow, such as radiation injury or high-dose chemotherapy, contribute to diminish their function and impair their regenerative capacity (Lucas et al., 2013, Cao et al., 2011). The SNS represents a critical regulatory component of the bone marrow niche, in which sympathetic nerve fibers and neural crest-derived cells serve as a major niche constituents, essential for maintaining HSC behavior in both homeostasis and stress (Katayama et al., 2006, Mendez-Ferrer et al., 2008, Yamazaki et al., 2011).

Neural activity in the bone marrow niche

The bone marrow microenvironment is formed by a complex network of blood vessels, sympathetic nerve fibers and stromal cellular components integrating membrane-bound and locally secreted signals that coordinate quiescence, proliferation and retention of HSCs and progenitor cells. The HSC niche was initially identified as an osteoblastic unit (Zhang et al., 2003, Calvi et al., 2003), where bone lining osteoblasts and osteolineage cells were suggested to maintain HSC activity trough secretion of angiopoietin-1 (Arai et al., 2004) and osteopontin (Opn) (Nilsson et al., 2005, Stier et al., 2005). However, recent development in imaging technology has revealed that HSCs are primarily located in close proximity to perivascular stromal cells in the bone marrow (Entschladen et al., 2004, Mendez-Ferrer et al., 2010b, Kunisaki et al., 2013). In addition, conditional deletion of key HSC maintenance factors, the CXCL12 chemokine and stem cell factor (SCF), in osteolineage cells indicated that these cells had a negligible effect on HSC maintenance (Ding and Morrison, 2013, Ding et al., 2012). The bone marrow vasculature is comprised of two types of blood vessels; sinusoids that are evenly distributed throughout the marrow cavity and arterioles which are primarily located near the endosteal bone region and highly innervated by sympathetic nerves (Kunisaki et al., 2013). Several candidate perivascular stromal cells have been suggested to regulate healthy HSCs, including CXCL12-abundant reticular (CAR) cells (Sugiyama et al., 2006), Nestin expressing cells (Mendez-Ferrer et al., 2010b) and Leptin receptor (LepR) expressing cells (Ding et al., 2012), that exhibit significant overlap among each other (Pinho et al., 2013). During development, neural crest-derived cells migrate to the bone marrow and reside along the blood vessels of the inner bone surface, where until adulthood, they retain neural crest-associated gene expression including nestin (Nagoshi et al., 2008, Takashima et al., 2007). Nestin-expressing perivascular cells contain all bone marrow mesenchymal stem and progenitor cell (MSPC) activity and are highly enriched for HSC maintenance and retention factors (Vcam1, Cxcl12, Angpt1, Scf and Opn) (Mendez-Ferrer et al., 2010b, Pinho et al., 2013). In a transgenic reporter mouse, differential nestin expression identifies distinct MSPC subset populations in the perivascular niche. MSPC-expressing high nestin levels associate with HSCs within arterioles and are characterized by the expression of the NG2 pericyte marker. In contrast, lower nestin expression identifies a more abundant reticular MSPC population located around bone marrow sinusoids and enriched for LepR expression (Kunisaki et al., 2013). These stromal LepR⁺ cells exhibit an MSPC phenotype and contribute to adipocyte and bone regeneration (Mizoguchi et al., 2014, Zhou et al., 2014). Interestingly, the rare arteriolar NG2⁺ pericytes are found to be intimately associated with sympathetic nerve fibers and GFAP⁺ Schwann cells and may harbor significantly higher MSPC and HSC maintaining activity (Kunisaki et al., 2013).

Sympathetic nerve fibers in the bone marrow are spatially closely associated with perivascular MSPCs around arteriolar blood vessels, forming a structural network termed the neuro-reticular complex (Kunisaki et al., 2013, Yamazaki and Allen, 1990, Mendez-Ferrer et al., 2010b) (Figure 2). Peripheral adrenergic signals released to the bone marrow negatively regulate the mesenchymal fate and the proliferative state of nestin⁺ MSPCs (Mendez-Ferrer et al., 2010b). Sympathetic denervation leads to proliferation and expansion of normally quiescent MSPCs while adrenergic stimulation, mediated through the adrenergic β3-receptor, leads to decreased osteoblastic differentiation and expression of HSC-regulating genes (Lucas et al., 2013, Mendez-Ferrer et al., 2010b). Bone forming osteolineage cells are also highly regulated by sympathetic nerve signals (Katayama et al., 2006). The first indication that the SNS might regulate bone turnover came from studies of leptin, a small polypeptide hormone, secreted by adipocytes. Leptin was identified to control body weight and gonadal function by binding to specific receptors located within the hypothalamus (Friedman and Halaas, 1998). Surprisingly, leptin was also demonstrated to act as a potent anti-osteogenic factor, indicated by increased bone mass in leptin-deficient (ob/ob) and LepR-deficient (db/db) mice (Ducy et al., 2000). Further studies found that sympathetic signaling via β2-adrenergic receptors present on osteoblasts controls bone formation downstream of leptin signaling (Takeda et al., 2002). Thus, similar to MSPCs, sympathetic stimulation negatively regulates the function of osteoblasts and presumably also the more abundant osteocyte population in the compact bone (Asada et al., 2013, Elefteriou et al., 2005). However, in contrast to perivascular MSPCs that regulate the maintenance and retention of HSCs, osteolineage cells appear to create a niche for early lymphoid progenitors (Ding and Morrison, 2013, Zhu et al., 2007). Still, the influence of osteoblast-mediated sympathetic signaling on lymphoid progenitor development remains to be fully characterized. Blood vessels lining endothelial cells play an important role in promoting HSCs maintenance by secreting SCF (Ding et al., 2012). In addition, stress-induced hematopoietic recovery following myeloablation seems to require endothelial cells for proper regeneration and replenishment of the HSPC population (Kobayashi et al., 2010). To date, the neural regulation of endothelial cells in the bone marrow niche has not been systematically addressed. However, endothelial cell and MSPC numbers appear to recover in parallel during bone marrow regeneration or after sympathetic denervation (Lucas et al., 2013), suggesting a similar neural regulation for both of these niche constituents.

Figure 2. Autonomic signals modulate steady-state hematopoiesis.

Different stromal cell types, including nestin-expressing perivascular cells, endothelial cells and CAR cells, regulate HSC maintenance. Although osteoblasts are dispensable for HSC maintenance, osteolineage cells may contribute to HSC regulation and for lymphoid progenitor cells maintenance. The neuronal components of the HSC niche comprise peripheral sympathetic neurons and non-myelinating Schwann cells that maintain HSC dormancy through activation of the TGF-β/SMAD signaling. Circadian noradrenaline secretion from sympathetic nerves leads to circadian expression of CXCL12 by nestin⁺ MSPCs, resulting in rhythmic release of HSCs to the periphery. The adrenergic signals in this case are mediated through the β3-adrenergic receptors (Adrβ3). Quiescent HSCs are located in close proximity to arteriolar blood vessels, ensheathed with sympathetic nerve fibers and Nestin^high NG2⁺ pericytes, however, upon activation, relocate near the Nestin^low LepR-expressing perisinusoidal area. Similar to MSPCs, sympathetic signals also regulate bone formation, via β2-adrenergic receptor (Adrβ2) signaling in osteoblasts.

Apart from sympathetic nerve fibers, other neural crest derivatives have been shown to regulate HSC homeostasis. Nonmyelinating Schwann cells that ensheathe nerve fibers of the bone marrow were suggested to preserve HSC quiescence through activation of TGF-β and SMAD signaling (Yamazaki et al., 2011). Autonomic nerve denervation resulted in significant decrease in bone marrow Schwann cells which was accompanied by drastic increase in HSC proliferation. However, it remains unclear how sympathetic nerves can signal to Schwann cells and to what extent bone marrow denervation, independent from Schwann cells and TGF-β/SMAD signaling, contributed to the effects observed on HSCs.

In addition to neural crest derivatives, neurotrophic factors and neuropeptides, released by innervating nerve fibers and surrounding cells, have also been suggested to participate in formation of hematopoietic environment in the bone marrow (Liu et al., 2007). For instance, substance P and neurokinin-A, a tachykinin family neuropeptides, have been suggested to stimulate production of hematopoietic cytokines by BM stromal cells as well as serving as essential modulators of both normal and malignant hematopoiesis (Nowicki et al., 2007).

Autonomic regulation of hematopoietic homeostasis

Preliminary evidence that sympathetic signals might regulate hematopoietic cells emerged several years ago when circadian oscillations of noradrenaline content in murine bone marrow was suggested to positively correlate with proliferation of bone marrow hematopoietic cells (Maestroni et al., 1998). Initial implications of the SNS in maintaining HSPC homeostasis were triggered by the discovery that a selectin glycomimetic inhibitor, fucoidan, significantly mobilized HSPCs independent of selectin itself (Frenette, 2000, Sweeney et al., 2000). This raised the possibility, that sulfated glycans in the bone marrow microenvironment modulate HSPC motility. Galactocerebroside (GalC) and its sulfated derivative, sulfatide, are the major component of myelin sheaths responsible for adequate nerve conduction (Norton and Cammer, 1984). Mice deficient for the UDP-galactose ceramide galactosyltransferase (Cgt^−/− mice) displayed no egress of HSPC from the bone marrow to the periphery upon enforced mobilization with granulocyte colony-stimulating factor (G-CSF) (Katayama et al., 2006). Interestingly, Cgt^−/− mice also exhibited defects in lymphopoiesis associated with deficit in stromal fractions in long-term bone marrow cultures (Katayama and Frenette, 2003). The defect in HSPC mobilization in Cgt^−/− mice was not due to the lack of GalC production but rather impaired adrenergic regulation of niche cells involved in the mobilization of HSPCs (Katayama et al., 2006). Mobilization of HSPCs is regulated by a gradient of the chemokine CXCL12, whose levels in the bone marrow are downregulated following G-CSF administration, leading to HSPC migration out to the periphery (Hattori et al., 2001, Petit et al., 2002). Sympathetic denervation in genetic and pharmacological models prevented the CXCL12 downregulation after G-CSF treatment resulting in impaired HSPC mobilization (Katayama et al., 2006).

The aforementioned circadian oscillations of adrenergic activity in murine bone marrow are associated with circadian oscillation of CXCL12 expression and consequently rhythmic release of HSPCs (Mendez-Ferrer et al., 2008). This postulates a model where core genes of the molecular clock in the CNS act through circadian noradrenaline secretion delivered by local sympathetic nerve fibers. Sympathetic signals are then mediated through adrenergic receptors expressed on bone marrow stromal cells and lead to downregulation of CXCL12 expression. CXCL12 is mostly secreted by perivascular MSPCs and to less extent by osteolineage cells (Ding and Morrison, 2013, Greenbaum et al., 2013). The adrenergic-stimulated release of HSPCs is mediated by β3-adrenergic receptors that are almost exclusively expressed within the nestin⁺ perivascular MSPC population in the bone marrow stroma (Mendez-Ferrer et al., 2010b). However, the enforced mobilization of HSPCs depends on adrenergic signals mediated through both β2 and β3-adrenergic receptors (Mendez-Ferrer et al., 2010a). Moreover, it is not only the downregulation of CXCL12 in nestin⁺ MSPCs but also the suppression of macrophage function that contribute to G-CSF-mediated HSC mobilization (Semerad et al., 2005, Chow et al., 2011). This is also associated with suppression of osteolineage cell function (Asada et al., 2013, Katayama et al., 2006, Semerad et al., 2005, Christopher and Link, 2008). In addition, G-CSF-induced mobilization leads to formation of larger gap spaces between bone marrow endothelial cells (Szumilas et al., 2005), accompanied by downregulation of the connexin-43 and 45 gap-junction proteins and lower CXCL12 expression in the niche (Dar et al., 2006, Gonzalez-Nieto et al., 2012). At steady state, connexin-43 and 45 were shown to influence the secretion of functional CXCL12, whereas their inhibition resulted in impaired homing of leukocytes to the bone marrow (Schajnovitz et al., 2011). Efferent sympathetic nerve terminals and stromal cells are inter-connected by gap junctions adjacent sinusoids in the bone marrow (Yamazaki and Allen, 1990). Therefore it cannot be excluded that sympathetic signals are propagated through intercellular communication between niche cells thus regulating HSPC homeostasis.

Similar to rodents, mobilized human HSPC counts fluctuate in a circadian manner, although the circadian rhythm in humans is inverted with significant higher yields of HSPCs toward the evening (Lucas et al., 2008). This finding may have clinical relevance for the transplantation of HSPCs in benign and malignant disease (Gratwohl et al., 2013). Interestingly, human hematopoietic progenitors can also be directly regulated by the sympathetic signals, as immature human CD34⁺ progenitors express dopamine receptors, which can be induced following G-CSF mobilization (Spiegel et al., 2007). Catecholamine neurotransmitters supported both motility and proliferation of human CD34⁺ progenitors suggesting that catecholamines can exert both HSC-autonomous and non-autonomous (niche-mediated) effects.

Autonomic regulation of hematopoietic regeneration

The success of any hematopoietic stem cell transplantation depends on harvesting a sufficient number of mobilized HSPCs from the peripheral blood. In a subset of donors, the yield of mobilized HSPCs is insufficient. In many cases, it appears to be due to prior myelotoxic anti-cancer therapies that correlate with poor mobilization (Perseghin et al., 2009, Robinson et al., 2000). In murine type 1 and 2 diabetes models, enforced mobilization of HSPCs to the blood is impaired, and this observation was confirmed in a small cohort of diabetic patients (Ferraro et al., 2011). These changes were mainly referred to alterations within the bone marrow niche with a reduction of osteoblasts and a surprising increase in nerve fibers in calvarial bone marrow. Moreover, sensitization of β-adrenergic receptors was altered by the presence of diabetes. In line with these findings, an impaired circadian release of endothelial progenitor cells was observed in a rat type 2 diabetes model (Busik et al., 2009), but by contrast to the above study, the authors reported sympathetic denervation of the bone marrow. More recently, impaired mobilization of HSPCs and hematopoietic regeneration was linked to sympathetic neuropathy of the bone marrow (Lucas et al., 2013). Sympathetic denervation -- either pharmacologically induced or through specific genetic deletion of sympathetic neurons -- led to impaired hematopoietic regeneration after challenging denervated mice with irradiation or myeloablation. The proposed model suggests that sympathetic neuropathy damages the bone marrow niche by driving nestin⁺ MSPCs and endothelial cells into the cell cycle, making them more vulnerable to genotoxic insults. After exposure to cytotoxic agents, the niche size decreases and fails to support hematopoietic regeneration (Figure 3). Administration of the neuroprotective agents 4-methylcatechol (4-MC), or the glial-derived neurotrophic factor (GDNF) concomitantly with the neurotoxic drugs rescued sympathetic nerve fibers (Lucas et al., 2013). This combined treatment saved bone marrow niche cells from genotoxic insults and allowed normal hematopoietic regeneration and efficient enforced mobilization of HSPCs despite prior neurotoxic treatment. These data clearly indicate that diseases or therapies that are accompanied with peripheral neuropathy might have a so far unappreciated long-term damaging effect on hematopoiesis that may be preventable.

Figure 3. Sympathetic denervation compromises hematopoietic regeneration.

Sympathetic neuropathy of the bone marrow damages the HSCs niche by driving proliferation of nestin⁺ MSPCs and endothelial cells. SNS injury to the bone marrow disrupts HSC and progenitor mobilization, and upon damaging evens such as chemotherapy or radiation, the niche size decreases and fails to support hematopoiesis.

Neural regulation of inflammation

During an innate immune response, sensory nerves are activated by inflammatory signals, leading to the release of neurotransmitters and neuropeptides (Deutschman and Tracey, 2014). Neuropeptides, including corticotropin-releasing hormone, substance P and calcitonin gene-related peptide, trigger the release of pro-inflammatory mediators (Pavlov and Tracey, 2012, Chiu et al., 2013, von Hehn et al., 2012) that promote the inflammatory response through increased vasodilatation, blood flow vascular leakiness and pain. The resulting activation of the HPA axis via afferent nerves or indirectly by inflammatory mediators, may counter-balance inflammation through glucocorticoids released from the adrenal cortex (Sternberg, 2006).

The SNS provides signals in the initial phase of inflammation. Some studies suggest that noradrenaline mediates stimulation of immune response by influencing immune cell migration (Sternberg, 2006) indicating that regional proinflammatory signals may serve to sustain local innate immune response. At the same time, in vitro experiments have shown an immunosuppressive effect of noradrenaline on monocytes and dendritic cells by suppression of pro-inflammatory cytokines and enhanced production of anti-inflammatory cytokines (Maestroni and Mazzola, 2003, Van der Poll et al., 1994). This complexity becomes obvious in denervation experiments in murine models of arthritis where in the early phase of disease the SNS shows proinflammatory effects, but at later stages, adrenergic signals have an anti-inflammatory impact (Harle et al., 2005). Noradrenaline has proinflammatory effects at low concentrations, mediated through α-adrenergic receptors, and predominantly anti-inflammatory effects at high concentrations, mediated through β-adrenergic receptors depending on the time point of activation relative to inflammation (Pongratz and Straub, 2013). The expression pattern of both α- and β-adrenergic receptors on innate immune cells depends on the cell type and its activation state. For example, T and B lymphocytes exclusively express the β2-adrenergic receptors, while murine TH2 cells lack expression of adrenergic receptors (Padro and Sanders, 2014). Furthermore, the expression levels of adrenergic receptors as well as the receptor binding capacity on immune cells are adaptive to environmental requirements. For example, peripheral blood mononuclear cells have been suggested to express lower levels of β2-adrenergic receptors in rheumatoid arthritis patients compared with healthy subjects (Baerwald et al., 1992). In active juvenile rheumatoid arthritis, leukocytes have a lower cAMP response to β2-adrenergic agonist indicating defective neuroimmune response (Kuis et al., 1996), whereas functional α1-adrenergic receptors are upregulated (Heijnen et al., 1996). Together with altered adrenergic receptor expression patterns during chronic inflammation, the local innervation pattern of sensory and sympathetic nerve fibers is altered with a shift towards proinflammatory sensory nerves (Miller et al., 2000, Reynolds and Fitzgerald, 1995).

A localized immune response depends on a regulated recruitment of leukocytes to the inflammatory site. Adrenergic signals have been suggested to regulate leukocyte trafficking to sites of inflammation (Viswanathan and Dhabhar, 2005). This action is mediated at least in part by an influence of the SNS on adhesion molecule expression on venular endothelial cells (Scheiermann et al., 2012). Adrenergic activity is delivered locally by nerve terminals and is controlled by the molecular clock resulting in a circadian oscillation in adhesion molecule expression and leukocyte recruitment with the peak recruitment occurring at night, the period of activity in rodents. At times of higher SNS tone, mice exhibited an increased sensitivity to inflammatory stimuli. In contrast, cholinergic stimulation has been reported to suppress the expression of endothelial cell adhesion molecules via the nicotinic acetylcholine receptor alpha7 subunit (α7nAChR), (Saeed et al., 2005), suggesting the potential for opposing action of the SNS and the parasympathetic nervous system (PNS) in the inflammatory response.

The PNS senses the inflammatory reaction through IL-1β receptors expressed on vagus nerve afferents and chemosensory cells in paraganglia surrounding afferent vagus nerve endings (Goehler et al., 2000). This appears to represent an initial step in the inflammatory reflex of early inflammation where the PNS counteracts the systemic inflammatory response through the release of acetylcholine. The increased release of acetylcholine inhibits endotoxin-induced release of pro-inflammatory but not anti-inflammatory cytokines from macrophages through α7nAChR (Borovikova et al., 2000, Wang et al., 2003). Anti-inflammatory activity is supported by acetylcholine-mediated inhibition of the protein high mobility group box 1 (HMGB1) (Wang et al., 2004). HMBG1 is secreted by activated immune cells and activates prototypical inflammatory responses in immune cells and endothelial cells (Lotze and Tracey, 2005). Electrical stimulation of the peripheral vagus nerve attenuates serum TNFα levels and prevented development of shock during lethal endotoxemia (Borovikova et al., 2000). Interestingly the anti-inflammatory effect of vagus nerve stimulation also requires a functional β2-adrenergic receptor on CD4⁺CD25⁻ lymphocytes (Vida et al., 2011). In contrast to SNS that innervates primary and secondary lymphoid organs (Bellinger et al., 2008), there is no definite evidence for efferent vagal or parasympathetic innervation of the immune system (Figure 1) (Nance and Sanders, 2007, Schafer et al., 1998). Indeed, vagus nerve stimulation appears to use sympathetic nerve relay to induce acetylcholine production by memory CD4⁺ T cells in the spleen (Rosas-Ballina et al., 2011). Thus, the neural circuits regulating immune responses require both parasympathetic and sympathetic branches of the ANS. Interestingly, another recent study has demonstrated that activation of the sensory sciatic nerve by electro-acupuncture attenuated inflammation in an experimental sepsis model via dopamine production in the adrenal glands (Torres-Rosas et al., 2014), suggesting a new anti-inflammatory mechanism mediated by the sciatic and vagus nerves modulating the production of catecholamines.

The increasing understanding of neural circuit modulation of immune responses and SNS-regulated mechanisms in leukocyte trafficking open new therapeutic options to manage inflammation-mediated diseases. Electrical stimulation and pharmacological approaches have already shown encouraging results. Considering circadian fluctuations, time-dependent application of therapies might finally boost anti-inflammatory effects.

Neural regulation of Cancer

Consistent with complex neural influence, behavioral factors including stress, depression and social isolation have been suggested to contribute to pathogenesis (Chrousos and Gold, 1992). Epidemiological studies have attempted to identify correlations between chronic stress and clinical outcome in cancer patients. For example, women who suffered traumatic life events had a two-fold higher risk of developing breast cancer (Lillberg et al., 2003). Moreover, chronic depression has been linked to poor survival and higher mortality in a large variety of tumor types (Chida et al., 2008). More recently, the follow up of a large cohort of patients with localized prostate cancer showed reduced survival in patients who were also diagnosed with depression (Prasad et al., 2014). Similarly, chronic stress induced by obesity and diabetes has been shown to correlate with increased risk of cancer and poor survival among cancer patients with pre-existing diabetes (Calle and Kaaks, 2004). Taking into account the role of autonomic nerves in hematopoiesis and tissue morphogenesis, it is reasonable to think that neural-mediated stress signals may affect tumor cell growth and vice versa. In this section, we will focus on recent developments with regards to hematologic and solid malignancies.

Leukemia

The deterioration of bone marrow microenvironment is suggested to be an important step in the development of myeloid leukemia (Walkley et al., 2007a, Walkley et al., 2007b, Raaijmakers et al., 2010, Zhang et al., 2012). Both chronic myeloproliferative neoplasms (MPNs) and acute myeloid leukemia (AML), have been demonstrated to remodel their microenvironment to form a self-reinforcing oncogenic unit at the expense of normal hematopoiesis. MPN and AML are stem cell-driven diseases where leukemic stem cells (LSCs) arising from mutated HSCs or committed progenitors preserve their capability to self-renew and can contribute to relapse of the disease after therapy (Ishikawa et al., 2007, Krause and Van Etten, 2007). Bone marrow MSPCs and osteolineage cells play a central role in remodeling the bone marrow niche to accommodate LSC expansion. Leukemic myeloid cells target MSPCs by secreting specific chemokines and growth factors that differentiate them into functionally altered osteolineage precursors (Schepers et al., 2013). Once expanded, they accumulate within the endosteal niche as inflammatory myelofibrotic cells, severely compromised in their ability to maintain normal HSC activity. However, a BCR-ABL blast-crisis model was reported to lead to severe reduction of both mature osteoblasts and osteoprogenitors (Frisch et al., 2011). Activation of the osteoblast function in the BCR-ABL model led to a significantly attenuated course of disease, and by contrast, in MLL-AF9 AML, osteoblast specific activation of parathyroid hormone (PTH) signaling led to an acceleration of disease (Krause et al., 2013). Further, recent studies with the MLL-AF9 model revealed that AML development expanded osteolineage-primed MSPCs at the expense of HSC-maintaining NG2⁺ periarteriolar cells. The documented loss of MSPC quiescence was accompanied by bone marrow sympathetic neuropathy (Hanoun et al., 2014). Notably, the expanded osteolineage-committed cells were blocked in their ability to differentiate into mature osteoblasts, contributing to significant bone loss. Adrenergic signals reinforcing AML were mediated through the β2-adrenergic receptors expressed on stromal leukemic niche cells. In line with a more differentiated niche with reduced β3-adrenergic signaling, maintenance and retention of healthy HSCs was significantly impaired. Sympathetic neuropathy of the bone marrow was also found to be essential for JAK2(V617F)-driven MPN (Arranz et al., 2014). Development of chronic MPN in this model was accompanied with drastic reduction in bone marrow sympathetic nerve fibers, and ensheathing Schwann cells. However, in contrast sympathetic neuropathy was accompanied by reduced nestin⁺ MSPCs numbers through increased apoptosis, possibly triggered by IL-1β production by LSCs. This appears to contrast with the described MSPC expansion and enforced differentiation contributing to the common myelofibrotic tissue formation in BCR-ABL-driven MPN (Schepers et al., 2013). The adrenergic signal in this case was propagated through the β3-adrenergic receptor, as the administration of β3-adrenergic agonist rescued nestin⁺ MSPC numbers, blocked myelofibrosis and MPN development. Thus, the blockade of adrenergic signaling has overall disease-promoting effects in both MPN and AML, but it appears to differentially alter the healthy HSC niche to the benefit of a putative LSC (Figure 4). Clarification is required about the distinct niche requirements for different subsets of leukemia. However, protection of the sympathetic innervation of the bone marrow may represent a common niche-targeted approach to preserve healthy HSCs while contributing to the emerging anti-leukemic arsenal.

Figure 4. Sympathetic neuropathy promotes hematopoietic malignancy.

(A) AML induces sympathetic neuropathy of the leukemic niche. Development of MLL-AF9-driven AML disrupts SNS nerves within the leukemic bone marrow niche. The SNS neuropathy is accompanied with expansion of nestin⁺ MSPCs primed for osteoblastic differentiation at the expense of HSC-maintaining Nestin^high NG2⁺ pericytes. In addition, AML is also associated with increased vascular density. The adrenergic signals regulating LSCs are transduced by the β2-adrenergic receptor (Adrβ2) expressed on stromal cells in leukemic bone marrow. (B) Bone marrow neuropathy is essential for the development of MPN. Sympathetic nerve fibers, non-myelinating Schwann cells and Nestin⁺ MSPCs are consistently reduced in MPN bone marrow driven by JAK2(V617F) mutations in HSCs. In contrast to AML, the reduction in Nestin⁺ MSPCs is not due to differentiation, but neural damage coinciding with Schwann cell and Nestin⁺ MSPCs apoptosis triggered by IL-1β production by mutant HSCs. Treatment with β3-adrenergic agonist blocked MPN progression, suggesting that the adrenergic signals were transduced via the β3-adrenergic receptor (Adrβ3) present on MPN bone marrow stromal cells.

Solid tumors

In his anatomy book, the French anatomist Cruveilhier describes a process by which cancer cells invade the space surrounding the nerves, called perineural invasion (Cruveilheir, 1835), now known to be common in several cancer types (Liebig et al., 2009b). Perineural space is thought to provide cancer cells a pro-survival and anti-apoptotic environment that allows extracapsular spread of the growing tumor (Ayala et al., 2006, Bockman et al., 1994). Furthermore, certain cancer types take advantage of neurotransmitters released by nerve fibers to boost their growth and metastatic capacity (Entschladen et al., 2004). Cancer cells can also attract nerves surrounding the tumor by expression and secretion of axon guidance molecules and nerve attractants (Ayala et al., 2008). This process implies an increase in the number and extensions of axons, in contrast to neurogenesis, which is associated with an increase in neuronal cell bodies. Such recruitment of nerves into the tumor has been described in prostate, colon, bladder, and esophageal cancers (Liebig et al., 2009a, Ayala et al., 2008, Palm and Entschladen, 2007, Magnon et al., 2013, D and F., 2007). It is intriguing that the development of new nerves resembles the development of new vascularization (angiogenesis) (Magnon et al., 2013), raising important implications for future combination therapies.

The mechanisms of axonogenesis and the identity of neurotrophic factors secreted by the tumors to initiate their own innervation are not well understood. Some tumor types and cancer cell lines are more neurotrophic than others by secreting neural growth factor (NGF), brain-derived neurotrophic factor (BDNF) and glial-derived neurotrophic factor (GDNF) to induce the regeneration of neo-fibers in the tumors (Geldof et al., 1997, Kowalski and Paulino, 2002, Okada et al., 1999). In prostate cancer, patients exhibit increased prostate nerve density (Ayala et al., 2008, Magnon et al., 2013), and in vitro studies have revealed that the attraction of PC-3 prostate cancer cells to neural cells from dorsal root ganglion (DRG) was driven by semaphorin 4F, a well characterized axon guidance molecule (de Wit and Verhaagen, 2003). More recently, the recruitment of both sympathetic and parasympathetic fibers within the tumor has been shown in vivo in human prostate cancer orthotopic xenografts (Magnon et al., 2013). Data from a cohort of prostate cancer patients showed a significant correlation between patients with high risk of prostate cancer and increased sympathetic fiber (tyrosine hydroxylase-positive) densities in healthy tissue and parasympathetic fiber (vesicular acetylcholine transporter-positive) densities in malignant areas. In addition, analyses of primary pancreatic tumors have revealed an enhanced neural density of the pancreas (Ceyhan et al., 2009). Similar to prostate and pancreatic cancers, increased sprouting of both sensory and sympathetic nerve fibers was observed in the tumor-bearing regions and the periosteum of a bone lesion model, in which breast, prostate or sarcoma cancer cells were intrafemorally injected (Bloom et al., 2011, Halvorson et al., 2005, Jimenez-Andrade et al., 2011, Mantyh et al., 2010).

In several tumor types, autonomic nerves promote tumor growth and metastasis by acting on both cancer cells and the tumor microenvironment (Chida et al., 2008). Adrenergic signals can directly influence cancer development via stimulation of β-adrenergic receptors expressed by tumor cells. For instance, in breast and ovarian cancers, catecholaminergic signals are predominantly mediated via the β2-adrenergic receptor, expressed on the cancer cells, and can directly modulate their invasive potential (Penninx et al., 1998, Sood et al., 2006). Furthermore, norepinephrine has a stimulatory effect on the migratory capacity of breast cancer cells (Lillberg et al., 2003) and colon carcinoma (Prasad et al., 2014), which was inhibited by β-blockers. In vivo studies of chronic behavioral stress show that tumor growth was promoted in a xenograft model of peritoneal ovarian cancer via direct signaling of β2-adrenergic receptor expressed by the tumor cells (Thaker et al., 2006). In this case, the activation of adrenergic receptors led to an increased secretion of vascular endothelial growth factor (VEGF) that facilitated angiogenesis and enhanced tumor progression. At the same time, adrenergic signals protect tumor cells from programmed cell death in both prostate and ovarian cancers (Hassan et al., 2013, Sood et al., 2010). In support of a cell-autonomous adrenergic regulation, catecholamine signaling mediated by the β2-adrenergic receptor triggers DNA damage and suppression of p53 levels in a variety of human and murine cells (Hara et al., 2011). Recent studies also show that cholinergic signals from the vagus nerve regulate gastric tumorigenesis by stimulating Wnt signaling into a subset of tumor stem cells (Zhao et al., 2014). This effect is mediated by the muscarinic receptor 3 (Chrm3) expressed in Lgr5⁺ stem cells, and it is abrogated by unilateral vagotomy or local botox injection.

As described for hematopoietic malignancies, solid tumor progression and metastasis do not solely depend on cell-autonomous signals, but rely on cues from a remodeled microenvironment. Adrenergic signaling to stromal populations was recently shown to be essential for tumor initiation in transgenic and orthotopic xenograft models of prostate cancer, where cancer cell growth was inhibited in recipient mice deficient for β2- and β3-adrenergic receptors (Magnon et al., 2013). However, the underlying mechanisms as to how stromal adrenergic signaling supports tumor growth have not yet been resolved. Potential candidates in the tumor stroma are macrophages, immune cells, smooth muscle cells or endothelial cells. All express adrenergic receptors and are functionally regulated by the ANS (Galasso et al., 2013, Lujan et al., 1998, Sanders et al., 1997, Sloan et al., 2010) (Figure 5). In vitro studies suggest that macrophages may contribute to adrenergic signaling in the tumor microenvironment. Human and murine-derived macrophages express VEGF and matrix metallopeptidase 9 (Mmp-9) upon norepinephrine stimulation (Lutgendorf et al., 2008, Sloan et al., 2010). Upregulation of these molecules induces the formation of de novo blood vessels and promotes tumor cell invasive phenotypes by remodeling of the extracellular matrix (Sood et al., 2006). Other subsets of immune cells are known to express adrenergic receptors and respond to stress-mediated signals. B-lymphocytes, CD4⁺ naïve cells and Th1 cells preferentially express the β2-adrenergic receptor (Calle and Kaaks, 2004). However, the effect of catecholamines on these cell populations remains controversial due to result discrepancies reported for different experimental models (Elenkov et al., 2000). In the context of cancer, only two studies have addressed chronic stress-mediated signals on immune cells. In rats harboring breast cancer xenografts, chronic stress promoted tumor proliferation in parallel to the decrease of leukocytes and T-suppressor cells (Renehan et al., 2008). Moreover, in a syngeneic model of lymphoma, chronic restraint stress reduced the survival of mice and was accompanied with a decrease in CD4⁺ T-lymphocyte numbers and a reduced production of TNFα and interferon gamma (IFNγ) (Maestroni, 2000). Further work identifying the stromal SNS target cell is required in order to fully understand the contribution of the microenvironment in stress-promoted cancer progression. Similar to macrophages and immune cells, the tumor vasculature can also be affected by sympathetic signals. Dopamine has been shown to participate in tumor progression by affecting the tumor vasculature (Basu et al., 2001, Elenkov et al., 2000). Dopamine levels are reduced in ovarian carcinoma from stressed mice, and the exogenous administration of dopamine counteracts the tumor promoting effects of sympathetic catecholamines by inhibiting angiogenesis and apoptosis of tumor cells (Moreno-Smith et al., 2011). Moreover, in a syngeneic melanoma model, ablation of dopaminergic nerve fibers or deficiency of the D2 dopamine receptor was associated with increased angiogenesis and tumor growth (Basu et al., 2004). These results have important implications in therapy, primarily because the administration of dopamine both stabilized and normalized tumor vasculature increasing the delivery and efficiency of chemotherapy (Chakroborty et al., 2011).

Figure 5. Autonomic nerves are an active component of the tumor microenvironment.

Tumor microenvironment is heterogeneous and composed of fibroblasts, blood vessels, nerves fibers, macrophages and other immune cells that actively interact with tumor cells to regulate cancer progression. Under stress situations, nerve fibers recruited within and around the tumor release catecholamines in the tumor stroma. Both tumor cells and stromal cells express β2-adrenergic receptors (Adrβ2), and respond to norepinephrine which promotes tumor growth. Pericytes, within the tumor microenvironment, express dopamine receptor type 2 (DR2) and respond to dopamine which regulates angiogenesis. Moreover, tumor cells invade autonomic nerves (perineural invasion) in several cancer types causing pain and may facilitate systemic spreading.

Another element of communication between neurons and the surrounding tumor environment is the presence of neuropeptides, a very heterogeneous group of molecules released by nerve terminals in both the CNS and the periphery. Different neuropeptides have been shown to regulate several tumorigenic functions. For example, neuropeptide Y increases angiogenesis in xenograft models of neural-crest derived tumors (Kitlinska et al., 2005) and bombesin, neurotensin and endothelin-1 are regulating survival and migratory functions of androgen-independent prostate tumors (Lee et al., 2001, Sumitomo et al., 2000, Zheng et al., 2006).

The action of the ANS is not restricted to the primary tumor and may also affect the ability of tumor cells to metastasize (Figure 6). Different stress regimes including swim-stress, surgical stress, social confrontation and hypothermia promote lung metastasis in breast cancer rat models (Vida et al., 2011, Borovikova et al., 2000, Goldstein et al., 2007). These effects were mediated by adrenergic signals because β-adrenergic agonist injections promoted mammary tumor metastasis. Importantly, the stress effects on metastasis were prevented by combination pre-treatment with the β-adrenergic antagonist nadolol and indomethacin, a nonsteroidal anti-inflammatory drug (Saeed et al., 2005). In an orthotopic breast cancer model, adrenergic stimulation induced the infiltration of CD11b⁺ F4/80⁺ macrophages, which in turn promoted metastatic seeding without differences in primary tumor growth (Sloan et al., 2010). Moreover, inhibition of the macrophage colony-stimulating factor receptor reverted the stress-induced metastasis to distant organs.

Figure 6. Metastases at different sites are regulated by autonomic nerve signals.

Both branches of the autonomic nervous system can promote metastatic spread of solid tumors. (Left) Lung metastases are increased in mice harboring breast tumor xenografts when subjected to chronic stress. Stress-mediated signals stimulate the recruitment of CD11b⁺ F4/80⁺ macrophages in breast tumor orthotopic xenografts. These macrophages express β2-adrenergic receptors and enhance the lung metastatic capacity of tumor cells without affecting primary tumor growth. (Right) Muscarinic receptor 1 (Chrm1) expressed in the tumor stroma promotes prostate tumor cell invasion and metastasis to lymph nodes and distant organs. However, it is not yet known which cell of the stroma is targeted by parasympathetic signals.

Parasympathetic signals also mediate specific metastatic activity. In prostate cancer, carbachol stimulation mediated by muscarinic receptor type-1 (Chrm1), promoted visceral lymph node metastasis in orthotopic xenografts models. Moreover, in transgenic mice overexpressing the oncogene c-Myc in the prostate, Chrm1 expression in the tumor microenvironment was dispensable for primary tumor growth but critical for distant metastases (Magnon et al., 2013). Thus, accumulating evidence suggests an active role of the ANS in progression and dissemination of solid tumors, both by acting directly on tumor cells and also by signaling through niche components.

Outlook

From the description by Walter Cannon of the fight-or-flight response (Cannon, 1929) to the well described functions in maintaining steady-state hemodynamics or metabolism, the role of the ANS in regulating stress response and basal functions has been appreciated for many years. However, only recently has the influence of neuronal signals on stem cell behavior, inflammation, or cell migration emerged as an important physiological regulating pathway that may have clinical implications. For example, knowledge of the circadian release of stem cells may have consequences for the optimal time to harvest HSCs for transplantation. That tumors induce the growth of new axons contributing to the cancer opens a new field to study in the cooperation of nerves and tumor cells and can potentially serve as novel targets for therapeutic intervention. In this regard, nerve density assessment merits exploration as a possible predictive marker of cancer aggressiveness. Much of research in cancer biology has focused on the genetic basis of cancer which is extremely complex and heterogeneous even within the same individuals. Therapies directed at common alterations of the niche may thus provide a novel approach to harness cancer progression. Recent clinical studies have already demonstrated that the intake of β-blockers benefited prostate and breast cancer patients (Barron et al., 2011, Grytli et al., 2013, Melhem-Bertrandt et al., 2011, Perron et al., 2004, Powe et al., 2010), but not colorectal cancer and melanoma patients (Hicks et al., 2013, McCourt et al., 2014), indicating that therapies may need to be suited for the different cancer types. Treatments with adrenergic β-receptor inhibitors could also be complemented with agents blocking parasympathetic receptors to prevent metastasis development or as a preventive strategy in patients with cancer predisposition. As sympathetic nerves promote hematopoietic recovery (Lucas et al., 2013), niche-targeted therapies may also be beneficial in promoting hematopoietic recovery after neurotoxic chemotherapeutic treatments. Devices and new methods to stimulate or inhibit nerve function may one day join the therapeutic options available to treat chronic intractable diseases (Famm et al., 2013). Beneficial effects have already been observed in pilot clinical trials testing vagus nerve stimulation for pharmacoresistant epilepsy or rheumatoid arthritis. With improved delivery of opsins to selected nerve fibers, one could envision further progress using optogenetics or other emerging technologies to manipulate selectively local neural input for therapeutic purposes.