Neuroimmune Communication in Health and Disease

Abstract

The immune and nervous systems are tightly integrated, with each system capable of influencing the other to respond to infectious or inflammatory perturbations of homeostasis. Recent studies demonstrating the ability of neural stimulation to significantly reduce the severity of immunopathology and consequently reduce mortality have led to a resurgence in the field of neuroimmunology. Highlighting the tight integration of the nervous and immune systems, afferent neurons can be activated by a diverse range of substances from bacterial-derived products to cytokines released by host cells. While activation of vagal afferents by these substances dominates the literature, additional sensory neurons are responsive as well. It is becoming increasingly clear that although the cholinergic anti-inflammatory pathway has become the predominant model, a multitude of functional circuits exist through which neuronal messengers can influence immunological outcomes. These include pathways whereby efferent signaling occurs independent of the vagus nerve through sympathetic neurons. To receive input from the nervous system, immune cells including B and T cells, macrophages, and professional antigen presenting cells express specific neurotransmitter receptors that affect immune cell function. Specialized immune cell populations not only express neurotransmitter receptors, but express the enzymatic machinery required to produce neurotransmitters, such as acetylcholine, allowing them to act as signaling intermediaries. Although elegant experiments have begun to decipher some of these interactions, integration of these molecules, cells, and anatomy into defined neuroimmune circuits in health and disease is in its infancy. This review describes these circuits and highlights continued challenges and opportunities for the field.

I. INTRODUCTION

The nervous and immune systems act together as an integrated physiological system to monitor and respond to infection and inflammation. The concept of neuroimmune communication is not new, with many of the symptoms of inflammation arising from the effects of inflammatory mediators on the nervous system (196), and the detection of acetylcholine released from the spleen 90 yr ago (61). Several prominent studies have resulted in a new appreciation for the innervation of lymphoid organs and the functional consequences of neuronal activation for the immune system (2, 89, 224, 259). Perhaps more intriguingly, immune cells can produce neurotransmitters, functioning as a nonneuronal source of these molecules, with release dependent on signals from the innervation or the local tissue milieu (214, 224).

Communication between the immune and nervous systems is bidirectional, with neuronal signaling activated by exposure to pathogens or inflammation and immune cell function effected by neurotransmitters (2, 31, 214, 224, 259, 265). While there are a number of adaptive and maladaptive physiological responses to inflammation, ranging from a classical sickness response to altered satiety (62), many different stimuli that activate afferent pathways can lead to immunomodulation by autonomic neurons.

Following elegant studies documenting the existence of an anti-inflammatory reflex, there has been a resurgence in interest in the ability of the nervous system to regulate immune function. Demonstrating the power of this pathway, electrical stimulation to the efferent arm of this reflex can significantly reduce morbidity and mortality in a mouse model of septic shock (31, 265). These protective effects appear to be conserved in diverse immunopathologies with recent studies documenting therapeutic efficacy in preclinical models of septic shock, postoperative ileitis, rheumatoid arthritis (RA), inflammatory bowel disease (IBD), and renal ischemia reperfusion injury (31, 93, 119, 122, 126, 138). Such success has also led to the development of electrical nerve stimulators for the treatment of chronic inflammatory conditions. The rapid advances in this field and development of neurostimulators have resulted in numerous clinical trials for diseases ranging from IBD to RA (28, 138). Despite promising early preclinical and open label clinical trial results, new preclinical discoveries continue to highlight that there are a considerable number of unknowns. Factors that could conceivably impact the efficacy of electroceuticals include interspecies and individual variation in neural circuits and the effect of chronic inflammation on peripheral neurons and glia. While the vast majority of studies on the neuroimmune reflex arc have been conducted in mice and rats, it is unknown how applicable these circuits will be to humans.

The field of neuroimmune communication has grown at an exponential rate in recent years. With this rapid growth, there have been a tremendous number of advances and several new controversies that have developed. This review provides a contextual background, highlights these recent advances, and discusses some of the current challenges and controversies in the field.

II. DETECTION OF PATHOGENS, IMMUNE ACTIVATION, AND INFLAMMATION BY THE NERVOUS SYSTEM

How the nervous system becomes activated by bacteria and the immune system remains hotly debated. The source of this controversy likely stems from the use of different animal models influencing experimental outcomes that are then generalized to multiple pathways. In general, the proposed mechanisms of communication from the periphery to the central nervous system (CNS) include a humoral route or direct detection by peripheral afferent neurons. Each of these possibilities will be described in detail.

A. Circumventricular Organs

The humoral route of CNS activation by peripheral inflammation or infection involves the passage of bacterial-derived substances or host-derived cytokines to the brain (11, 12). This transmission of signals from the blood to the brain predominantly occurs in parts of the brain that lack a blood-brain barrier (BBB): the circumventricular organs (CVO). In most parts of the brain, exposure of the CNS to these substances is highly regulated due to the lack of fenestrated capillaries in the majority of CNS vascular beds (279). This BBB is further enhanced by the formation of tight junction proteins between adjacent endothelial cells (11). The consequence of this is reduced paracellular permeability, with blood to cerebrospinal fluid (CSF) transport occurring primarily by transcytosis (11, 12, 279). It is important to note that while transcytosis is the predominant mechanism, endothelial cells in the brain have reduced uptake compared with the periphery. While dissolved gasses are freely exchanged at this surface, this structural arrangement in conjunction with specific receptors and transporter proteins in the luminal and abluminal endothelial membranes permits controlled uptake of small hydrophobic molecules, nutrients, and specific peptides or proteins (11). Complementing the actions of the endothelial barrier, it is well established that astrocytes project feet toward the endothelium, adding another layer of resistance to small molecules transiting from the blood to the brain (12). While this highly selective permeability of this barrier ensures that neurons are protected from potentially neurotoxic substances, these structural features would prevent the central nuclei of the autonomic nervous system from monitoring the periphery. The autonomic nervous system can monitor hormones and metabolites in the periphery that are carried in the blood due to a structural adaptation in endothelium of CVO. In contrast to the highly selective nature of the endothelial cells that provide the BBB, the endothelium of CVO in the CNS allows for detection of substances circulating in the blood. These endothelial cells are highly fenestrated and lack the tight intercellular junctions associated with the endothelium that comprise the BBB (100). The eight CVO organs are divided into two discrete groups by functionality: sensory and secretory. The sensory organs include the subfornical organ (SFO), organum vasculosum, lamina terminalis (OVLT), and the area postrema (AP). The AP is of particular interest in the context of neuroimmune regulation due to the nature of its axonal projections. Neurons originating in the AP innervate the nucleus tractus solitarius (NTS) and the dorsal motor nucleus (DMN) of the vagus nerve, among other targets (206). This anatomy would allow for activation of neurons in the AP and regulation of vagal efferent (parasympathetic) signaling to alter physiology potentially including immune function.

B. Afferent Neural Inputs to Immunomodulatory Circuits

Peripheral primary afferent neurons express receptors for microbial constituents and inflammatory products. These neurons therefore have been proposed as a hardwired neural connection that underpins signaling of inflammation or infection from the periphery to the CNS. Two main populations of neurons comprise this pathway: vagal and spinal afferent neurons. Activation of vagal afferent pathways has been most thoroughly studied, so the mechanisms of vagal afferent neuronal activation will be described in most detail.

As the longest cranial nerve in the body, the vagus nerve innervates a diverse range of tissues including the heart, lung, and gastrointestinal tract. The vagus nerve comprises 70–80% afferent axons, with the cell bodies residing in the nodose and jugular ganglia. It is important to note that unique stimuli can trigger vagal afferent neurons in different tissues, ranging from nutrients in the intestinal tract (212) to mechanosensitive receptors in the lung (234). Alterations of these signals during inflammation at any of these sites could conceivably have profound implications in the activation of the anti-inflammatory reflex arcs. Activation of these afferent fibers results in synaptic transmission in the NTS, where information is integrated and sent to the DMN of the vagus, where the cell bodies of cholinergic parasympathetic preganglionic neurons reside. The axons of efferent neurons in the vagus nerve belong to these parasympathetic preganglionic neurons and will henceforth be referred to as vagal efferent neurons. These vagal efferent axons are thought to be an important target of electroceutical treatments that stimulate vagal axons to suppress inflammation (78).

C. Direct Activation of Vagal Afferent Neurons

1. Pathogen-associated molecular patterns

Direct activation of afferent fibers by pathogens has long been suggested to be an important mechanism of relaying peripheral information to the CNS. In brief, detection of pathogens in the body can occur through host receptors selective for highly conserved repeating motifs common to pathogens. These molecules are termed pathogen-associated molecular patterns (PAMPs) and range from lipopolysaccharide (LPS) in the cell wall of gram-negative bacteria, to unmethylated cytosine and guanine (CpG) in DNA. These motifs serve as ligands for specific Toll-like receptors (TLR), with binding inducing intracellular signaling and gene transcription events. Although TLR expression is considered a feature of innate immune cells, vagal afferent neurons also express TLR that facilitate the detection of PAMPs. While TLR3, 4, 7, and 9 are expressed on spinal afferent neurons and in the dorsal root ganglia (159, 208, 255), only TLR4, a selective receptor for LPS, has been detected in vagal afferent neurons from the nodose ganglia (NG) (65, 113). This section describes the evidence for expression of functional TLR on vagal afferents and evidence of functionality.

Vagal afferents are critical to the peripheral detection of LPS and the communication of this message to the CNS. Prior to the discovery of TLR4 and the development of tools to localize cells expressing this receptor (112, 204), afferent neurons were reported to be activated by LPS (92, 146). Peripheral administration of LPS induced sickness behaviors and neural activation in the NTS that was abrogated in mice subjected to subdiaphragmatic vagotomy. In support of these data that suggested LPS-induced activation of vagal afferents, freshly isolated NG and cultured NG neurons from rats express TLR4 mRNA and protein (65, 113). LPS stimulation can exert a variety of effects on vagal afferents including reduced sensitivity to other physiological stimuli. For example, LPS stimulation increased SOCS3 expression in the NG, preventing leptin-mediated STAT3 activation (65). Despite this ability of LPS to activate and modulate vagal afferent signaling, the mechanisms of LPS-induced neural activation have been elusive. Cultured neurons from the NG stimulated with LPS appear to be polymodal chemosensitive afferents that respond to a variety of nociceptive stimuli. Neural activation determined by Ca²⁺ imaging and patch-clamp experiments further revealed TLR4 independent activation by LPS using TLR4-deficient cells. This TLR4 independent response to LPS was found to be dependent on TRPV1 channel expression and was significantly reduced in TRPV1 knockout (KO) neurons, or wild-type (WT) cells pretreated with a selective TRPV1 blocker. It is however unclear how LPS activates TRPV1 or the ramifications of this in the afferent pathway. Together, these data illustrate the ability of afferent neurons to detect and respond to infection, a key requirement for the nervous system to modulate immune function.

2. Danger-associated molecular patterns

Inflammation and infection can also be detected after the release host-derived intracellular substances. As these substances are typically released as a consequence of inflammation or necrotic cell death, they have been termed danger-associated molecular patterns (DAMPs), or “alarmins” (154, 233). There is a diverse array of substances that serve as DAMPs, including ATP, DNA binding proteins such as high mobility group box 1 (HMGB1), and the alarmin cytokine interleukin (IL)-33 (47, 154, 154). In this section, we describe some of these molecules and the effect they exert on primary afferent neurons (summarized in TABLE 1).

Table 1.

Expression of cytokine, DAMP, and PAMP receptors by neurons and associated cells

Ligand	Receptors	Cell Type	Tissues
ATP	P2X1	Neurons	NG (134, 266)
	P2X2	Neurons	NG (142, 266)
		SGC	NG (79)
	P2X3	Neurons	DRG (152)
			NG (142, 266)
	P2X4	Neurons	NG (142, 266)
	P2X7	SGC	NG (79)
CpG	TLR9	Neurons	DRG (150, 208)
dsRNA	TLR3	Neurons	DRG (208)
HMGB1	RAGE	Neurons	DRG (229, 263)
IL-33	ST2/IL-1RAcP	Neurons	DRG (157)
			NG (246)
LPS	TLR4	Neurons	NG (65, 113)
	TRPV1		DRG (208, 255)
ssRNA	TLR7	Neurons	DRG (159, 208)
TNF-α	TNF-α receptor 1/2	Neurons	DRG (115)
			VAN (107)
			NTS (107)
			DMN (107)

CpG, unmethylated cytosine-phosphate-guanine dinucleotides; HMGB1, high mobility group box 1; IL, interleukin; LPS, lipopolysaccharide; TNF-α, tumor necrosis factor-α; SGC, satellite glial cells; NG, nodose ganglia; DRG, dorsal root ganglia; VAN, vagal afferent neurons; NTS, nucleus tractus solitarius; DMN, dorsal motor nucleus of the vagus.

3. ATP

Like all DAMPs, ATP released from host cells can occur as a consequence of inflammation or necrosis (83). It is also critical to note that ATP is also released as a neurotransmitter within the autonomic nervous system (40) and can be released from epithelial and endothelial cells (134). In the context of inflammation, ATP can be released through Pannexin-1 (Panx1) channels (50, 230), from a variety of cell types including intestinal epithelial cells, keratinocytes, and neutrophils (73). Release of ATP by Panx1 can occur following activation mechanical stress during inflammation (14, 73), or through activation by caspase 3/7-dependent cleavage (50, 230). It is uncertain if distension and stretch during edema that can occur during inflammation is sufficient to induce Panx1-mediated ATP release. Panx1 activation can also occur by caspase 3/7-mediated cleavage during pyroptosis. This nonapoptotic form of programmed cell death occurring during inflammation results in activation of caspase 1 and consequently caspase 7 (174). As such, multiple inflammatory stimuli appear capable of inducing Panx1 activation and extracellular release of ATP that could be detected by afferent neurons.

Extracellular ATP is capable of binding to purinergic (P2) receptors of two discrete types: P2X and P2Y receptors. Depending on the cell type they are expressed on, ATP initiates signaling that results in a range of responses including chemotaxis of neutrophils (51) to activation of neuronal signaling (40). Vagal afferent neurons have been characterized to be responsive to ATP and express the purinergic receptor subunits P2X2 and P2X3. These early experiments demonstrated afferent neuronal activation in response to ATP, or the P2X agonist α,β-methylene-ATP (276), that was blocked with the P2X2/3 receptor-selective antagonist A-317491 (142). Single-cell RT-PCR conducted on neurons from the NG indicated P2X2 and P2X3 expression compared with the predominant expression of P2X2 in jugular neurons of guinea pig (142). Indicative of species-specific differences, although adult rat NG express P2X1, P2X2, P2X3, and P2X4, P2X3 was the predominant receptor subunit (266). These receptors were functional, with application of extracellular ATP activating currents in whole cell patch-clamp experiments (266). As these different receptor subunits engender unique current responses, these data would suggest that ATP could stimulate different patterns of neural activation in unique vagal afferent fibers. In mice, as in other animals, these receptor subunits form heteromeric receptors as evidenced by genetic P2X3 deficiency only reduced ATP evoked responses with slower desensitization compared with WT neurons (241). As a heteromeric receptor with functional changes depending on subunit expression, it would be interesting to consider if peripheral inflammation alters P2X subunit expression. In support of this possibility, P2X2 and P2X3 expression increased in rat NG after myocardial ischemia (267), a trigger of inflammation and necrosis (86). These heteromeric receptors are not restricted to vagal afferents and have also been found on sensory afferents such as the DRG (152). P2X receptors can also be found on neurons that may participate in immunomodulatory reflexes including spinal afferent and enteric afferent neurons (220). These findings suggest the peripheral nervous system possesses the capability to detect DAMP that are released during inflammation.

D. Cytokines

Afferent innervation has long been proposed to be involved in the monitoring of peripheral immune status. Activation of the immune system following infection, or exposure to PAMPs such as LPS, typically leads to the production of proinflammatory cytokines. These cytokines, the expression of their receptors, and the consequence on afferent signaling will be described.

1. Tumor necrosis factor-α

Activation of the immune system results in the production and release of proinflammatory cytokines, including tumor necrosis factor (TNF)-α. This cytokine is produced by a plethora of immune cells including macrophages, T cells, and mast cells following exposure to stimuli such as LPS, or other PAMPs. Detection of TNF-α occurs by binding to TNF-α receptor 1 (TNFR1) or TNF-α receptor 2 (TNFR2). While TNFR1 is expressed constitutively by a wide variety of cells throughout the body, TNFR2 is expressed on selected cell types including macrophages/monocytes, microglia (33), and neurons in the central (274) and peripheral nervous system (115). In addition to differences in receptor expression profiles, the downstream signaling elicited by ligand binding is unique to each receptor. Following binding to TNFR1, a complex series of intracellular signaling events occurs resulting in activation of IKK and phosphorylation of IκBα. This phosphorylation event signals for ubiquitination and proteasome-mediated destruction of this inhibitory protein, allowing for NF-κB activation (35).

Vagal afferent neurons that project into the NTS of the brain stem constitutively express TNFR1, and TNF-α can activate these neurons or enhance the response to otherwise innocuous stimuli (108), stimulation with ATP (222) or capsaicin (115). Expression of TNFR1 was identified in the NTS by immunohistochemistry, with vagotomy rostral to the NG ablating TNFR1 immunoreactivity, suggesting that the immunoreactivity was derived from the central terminals of vagal afferent neurons. This TNFR1 immunoreactivity was also found in the cell bodies in the celiac ganglia but not the distal vagal trunk (107). It is important to note that the identity of TNFR1⁺ cells was not confirmed to be neurons, and as such, expression of TNFR1 by microglia (141), astrocytes, or oligodendrocytes (72) in the NTS cannot be excluded. RNA-Seq data from isolated cells of the cerebral cortex indicates higher basal expression of TNFR1 in microglia and astrocytes compared with neurons (277). These data demonstrate that caution should be taken in interpreting that TNFR1⁺ cells in the NTS are neurons unless other cell types are clearly excluded. Functionally, TNF-α has been shown to activate vagal afferent neurons in the gastrointestinal tract, increasing sensitivity to innocuous stimuli resulting in a vagal-gastroinhibitory reflex and gastric stasis (75). Vagal afferents are therefore capable of detecting and responding to TNF-α, resulting in alterations to host physiology.

2. IL-1β

Similar to TNF-α, IL-1β is produced by a multitude of cells during inflammation or cellular stress. Neurons in the NG that project into the NTS express IL-1β receptor and are activated following intravenous injection of recombinant IL-1β (74, 99). Although these studies suggest neuronal activation of neurons by nuclear c-Fos immunoreactivity, it is conceivable that given the pleiotropic nature of IL-1β, and the multitude of cells that express IL-1βR, activation could be through a paracrine fashion. In support of this contention, activation of vagal afferents was shown to be dependent on IL-1β-induced prostaglandin synthesis (99) from adjacent vascular endothelium (171). Endothelium-derived PGE₂ appears to act presynaptically to augment spontaneous excitatory synaptic input to the NTS. These data suggest that IL-1β stimulation of vagal afferents can act through an indirect mechanism. There is evidence of direct neural activation and sensitization of nociceptive terminal of spinal afferents to other stimuli within the skin by IL-1β (23). Together these findings indicated IL-1β can cause neuronal activation or increase the sensitivity of primary afferents through direct and indirect mechanisms serving to detect inflammation in peripheral tissues.

3. IL-33

Identified as a mediator in T helper 2 (T_h2) responses including allergy and immunity to extracellular parasites, IL-33 has emerged as another cytokine that can fulfill a role as an alarmin molecule (154). There is constitutive IL-33 expression in a wide variety of cells including endothelial, epithelial, and lymph node stromal cells in humans and mice (182). Under basal conditions, IL-33 is localized to the nucleus and is not detected in the cytoplasm or extracellular space. This basal expression can be increased during inflammation, with increased nuclear IL-33 in alveolar epithelial cells following allergic lung inflammation (104), intestinal nematode infection (275), mouse models of colonic inflammation, and patients with IBD (198, 235).

This cytokine functions as an alarmin, with the release of full-length IL-33 from the nucleus during necrosis, tissue damage, or cellular stress. Similar to IL-1β, IL-33 can be cleaved by activated caspase 3 or caspase 9, typically activated during immunologically quiescent apoptosis. This proteolytic cleavage produces noninflammatory fragments of this cytokine (154). The full-length bioactive IL-33 elicits intracellular signaling through the ST2 receptor, that is expressed by a variety of cell types including vagal afferent (246) and DRG neurons (157). Although functional studies were not performed on vagal afferents, it is intriguing to consider that ST2 on vagal afferents could be another mechanism to detect IL-33 released during necrotic cell death. In primary afferents innervating the skin, IL-33 release triggered by the poison ivy antigen, urushiol, resulted in itch through ST2-induced activation, and IL-33 induced Ca²⁺ signaling in DRG neurons. At an organism level, urushiol challenge elicited skin inflammation and pruritic behaviors that were exacerbated with coinjection of IL-33, mimicked by injection of IL-33 alone, and blocked with IL-33 neutralizing antibodies (157). Together these data indicate that afferent neurons express this functional receptor, with activation inducing neuronal activation and behavior changes.

E. Indirect Activation of Vagal Afferents in the Neuroimmune Reflex

A multitude of noninflammatory stimuli can activate vagal afferent neurons, resulting in vagal efferent outflow and consequently inhibition of immune cell activation. In the gastrointestinal tract, ingestion of nutrients activates a specialized subset of the epithelial cells lining the intestinal tract. These enteroendocrine cells (EEC) are activated by sugars, dipeptides, and long-chain fatty acids (144, 211). In response to luminal nutrients, these EEC release hormones, including cholecystokinin (CCK), that activate vagal afferent nerves (211). Elegant studies have demonstrated synapses between EEC and the underlying afferent innervation and that these interactions occur throughout the gastrointestinal tract including in the colon (26, 27). The contribution of EEC to neuroimmune regulation has been suggested with eternal high-fat diet to induce CCK release from EEC, leading to activation of CCK receptors on vagal afferents (164, 165). While CCK actions on vagal afferent neurons is usually associated with satiation, it has also been demonstrated that vagal afferent detection of CCK increased immunomodulatory efferent signaling to the spleen, reducing production of TNF-α and IL-6 production during a model of septic shock (165). Although there are undoubtedly several other mechanisms at work in this model, including the sequestering of endotoxin by increased triacylglycerol-rich lipoproteins in serum (166), these studies highlight that activation of vagal afferents by a diverse array of noninflammatory stimuli can be sufficient to activate an anti-inflammatory reflex.

1. Bitter taste receptors

The family of bitter taste receptors comprises closely related family members encoded by ~28 human and 36 mouse T2R genes with unique sensitivities to select ligands (7, 8). The expression of bitter taste receptors in cells and tissues not involved in taste sensation has been revolutionized by the demonstration that bacterial quorum molecules can serve as ligands for these receptors, aiding host detection of bacteria (251). In the lung, the bacterial pathogen Pseudomonas aeruginosa produces acyl-homoserine lactone as a quorum sensing molecule to establish bacterial population dynamics (197, 240). These molecules are sensed by the host through activation of host T2R, increasing the local immune response against this pathogen (148). Intriguingly, bitter taste receptors have been found to be expressed on EEC and other specialized epithelial cells (205). In the gastrointestinal tract, a wide variety of T2R genes are expressed in a location-specific manner in the stomach antrum and fundus, duodenum (270), ileum, and colon (19, 205). It remains to be determined if these receptors can detect bacterial quorum-sensing molecules at the intestinal mucosal surface and participate in host defense, or if these compounds can directly or indirectly initiate neuroimmune signaling.

F. Inflammation-Induced Changes to Afferent Signaling: Satellite Glial Cells

An additional confounding variable that should be considered in neuroimmune communication is the role of satellite glial cells (SGC). Similar to glia in the CNS, these supporting cells encase neurons within sensory ganglia including the DRG and the vagal afferents of the NG. The SGC from the NG respond to proinflammatory cytokines induced by systemic inflammation by becoming activated as evidenced by increased GFAP expression and Ca²⁺ signaling in response to ATP (79). This increased responsiveness to ATP was due to inflammation-induced expression of the P2X2 and P2X7 receptors, pannexin-1 activation, and increased glia-neuron dye coupling (79). Colonic inflammation also increases glia-neuron dye coupling, spontaneous activity, and hyperexcitability of DRG neurons. This increased neuronal activity in isolated DRG was prevented by the addition of gap junction blockers, suggesting that inflammation-driven glial-neuronal coupling was responsible for aberrant neuronal function (116). This activation of DRG satellite glial cells has been established as a mechanism of inflammation-induced visceral hyperalgesia, that can be reduced with pharmacological blockade of gap junctions (116, 268). These data suggest that SGC in the DRG and NG can increase neuronal signaling, although it is unknown how the activation of SGC during chronic inflammation effects afferent signaling in the anti-inflammatory reflex arc.

G. Other Afferent Neurons: Spinal Afferent Neurons

Spinal afferent neurons, with cell bodies in DRG and trigeminal ganglia, also respond to bacterial products, inflammatory mediators, DAMPs, and PAMPs and transmit information to the CNS. While the primary roles of these neurons are the transfer of afferent and nociceptive information to the brain, spinal afferent neurons may also be important mediators of neural immunoregulation. Activation of the peripheral terminals of spinal afferent neurons leads to the peripheral release of neuropeptides such as substance P and calcitonin gene-related peptide (CGRP) via axon reflexes. These neuropeptides cause vasodilation and lymphocyte extravasation, a process known as neurogenic inflammation. There is also evidence that CGRP can directly suppress macrophage activation (55). Consistent with this, ablation of Nav1.8-expressing afferent neurons increased inflammation in response to Staphylococcus aureus infection (272). The majority of the central terminals of spinal afferent neurons synapse on populations of interneurons in the dorsal horn, which project to higher brain centers and can result in the perception of pain. However, there are well-established local interneuronal projections from the dorsal horn to sympathetic preganglionic neurons within the intermediolateral cell column of the spinal cord (68). An interneuron pathway between spinal afferent and preganglionic splenic sympathetic neurons is particularly intriguing (44). Such a connection could serve as a neuroanatomical substrate of a possible spinal afferent-sympathetic immunomodulatory reflex pathway. This pathway appears to become particularly significant following thoracic spinal cord injury, partly due to the loss of descending inhibitory pathways (257, 278). Therefore, spinal afferent pathways may participate in immunomodulation by release of peripheral immunomodulatory peptides and by activation of autonomic anti-inflammatory pathways.

III. CIRCUITRY OF THE CHOLINERGIC REFLEX ARC

In the cholinergic anti-inflammatory pathway (CAIP), following detection of inflammation by afferent neurons, this information is coordinated in the CNS. In the brain stem, the central terminals of vagal afferent neurons, and those of the neurons in sensory CVOs innervate the NTS. Release of glutamate from these afferent fibers activates neurons within the NTS (5, 145, 248) that project into the DMN, initiating the efferent arm of the anti-inflammatory pathway. This basic functional circuit was established by observing neuronal activation in the NTS, coupled with surgical ablation and neurostimulation techniques. Activation of NTS neurons indicated by increased c-fos expression has been observed during infection with bacteria, nematodes (46), or immune activation elicited by injection of PAMPs such as LPS (156).

The CAIP can also be activated by intracerebroventricular (icv) injection of muscarinic receptor agonists. Injection of muscarine or CNI-1493, an M1 agonist, activates the efferent arm of the anti-inflammatory reflex to reduce carrageen-induced paw inflammation and edema (30). Although it is uncertain how central administration of these muscarinic agonists elicits vagal efferent activity, it has been proposed that neurons within the NTS or DMN express muscarinic receptors, or that these compounds act indirectly by stimulating neurons at other sites that project to the DMN (200). While choline acetyltransferase (ChAT) activity (105) and functional muscarinic receptors are expressed in the NTS (60), vagal preganglionic motor neurons in the DMN do not express muscarinic receptors (111). This would suggest that muscarinic agonists such as CNI1493 could activate interneurons or glia and consequently induce vagus efferent signaling, or alter the response to glutamate. Indirect modulation of sensory inputs has been suggested as there appears to be a lack of tonic ACh signaling. As evidence of this, while injection of ACh into the NTS induces hypotension and bradycardia that can be prevented by muscarinic receptor antagonists, injection of muscarinic receptor antagonists alone fail to induce hypertension (256). It is worth noting that while the anti-inflammatory effects of CNI1493 were abrogated by atropine pretreatment, it is uncertain if central muscarinic signaling occurs during a reflex initiated by peripheral LPS or inflammation.

The role of the vagus nerve in providing protection during systemic inflammation was suggested by experiments using surgical vagotomy. Cutting of the vagus before induction of endotoxemia or polymicrobial sepsis significantly increased mobility and mortality compared with sham-operated animals (31). As the vagus nerve comprises 70–80% afferent neurons, it was not clear if vagal afferents, efferents, or both were part of the functional circuitry. Elegant experiments using electrical vagal neural stimulation (VNS) coupled with vagotomy were central to establishing the importance of the vagal efferent neurons. Application of VNS to the peripheral end of the cut nerve following vagotomy provided protection from endotoxemia, indicating that the efferent innervation of target tissues was responsible for protection (31). These vagal efferents indirectly innervate tissues including the spleen, a site identified as the predominant source of proinflammatory cytokines including TNF-α in mice during endotoxemia (120). These data were interpreted as VNS activation of efferent vagal nerve activity reduces macrophage activation and TNF-α production in the spleen, consequently reducing mortality. Despite this inhibition of aberrant immune responses by the efferent pathway, VNS can also elicit afferent vagal nerve activation, which functions as a potent inhibitor of LPS-induced TNF-α production through an unknown circuit (193a). This study highlights the importance of carefully defining the circuitry being stimulated and could serve as an underlying basis of some of the controversies in the CAIP.

The anatomical connection between the vagus nerve and the sympathetic innervation of the spleen is a highly contentious issue. On the basis of prior neuroanatomical tracing studies, a disynaptic connection was proposed such that the preganglionic parasympathetic axons synapsed onto postganglionic sympathetic neurons in the celiac ganglia (18, 223). These tyrosine hydroxylase (TH)-expressing neurons that produce norepinephrine (NE) are proposed to innervate the spleen. At present, there are several lines of conflicting neuroanatomical evidence that refute or support this pathway. In mice, the injection of cholera toxin B (CTB) into the spleen resulted in labeling of a discrete population of neurons within the DMN that were not found in animals subject to vagotomy (42). In contrast, studies in rats with injection of DiI into the DMN, which contains the cell bodies of parasympathetic preganglionic neurons, and fast blue into the spleen, which contains sympathetic nerve terminals, did not reveal overlap of tracers in neurons in the CG (34). However, analysis of the injected rats revealed abundant fast blue labeling of postganglionic neurons in the suprarenal ganglion without labeling within the CG. Vagal terminals were rare in the suprarenal ganglion, and synaptic contacts between the vagal terminals and the postganglionic sympathetic neurons were not observed (34). These data suggest a complete lack of synaptic connections between parasympathetic preganglionic neurons and sympathetic neurons that innervate the spleen. Electrical stimulation of the efferent vagus in the rat further support this lack of a neural connection from the vagus to the spleen, with electrical VNS unable to elicit electrical activity in sympathetic neurons that projected into the spleen (169). These studies raise the possibility that there are significant neuroanatomical differences in mice compared with rats, or more likely that there are other neural circuits and sympathetic ganglia activated during VNS. As evidence for this multiple circuit hypothesis, injection of pseudorabies virus (PRV) into the spleen results in labeling of the preganglionic sympathetic motor neurons in the spinal cord concurrent with labeling of the DMN. Highlighting that two pathways could innervate the spleen, vagotomy was shown to prevent viral labeling in the DMN, but not the spinal cord (39). In direct opposition to these data, injection of the same PRV strain into the spleen labeled neurons in the intermediolateral nucleus of the spinal cord, but not the DMN (43, 44). Species variation should also not be discounted as functional evidence of vagal-sympathetic anti-inflammatory connections was derived from studies of mice, whereas the neuroanatomical studies were predominantly performed in rats. Discrepancies such as these highlight the continued need to perform comprehensive neuroanatomical and circuitry analysis in a variety of model organisms. The ability to translate these findings to the treatment of inflammation clinically will depend on deciphering not only the neural circuitry but also accounting for the potential existence of multiple immunomodulatory pathways and interspecies variations.

A. Innervation of the Spleen

The innervation of the spleen is critical to the anti-inflammatory reflex in models of sepsis and provides further information about the circuitry involved. The presence of axons in the spleen has long been described in mice (217), rats, cats (25), and humans (109), although the neurotransmitters released and the consequences of the splenic innervation have been contentious.

In mice and rats, splenic innervation is provided by sympathetic fibers that express TH and produce NE. With few exceptions in the literature suggesting sparse innervation by cholinergic fibers, TH⁺ fibers are readily detected in spleen of these animals (15, 186, 223). Lack of ChAT⁺ innervation has been established using both immunohistochemistry and lack of green fluorescence protein (GFP)⁺ axons in ChAT-GFP reporter mice (185). While the cholinergic innervation of the spleen has been reported in humans, it is important to note that this innervation was identified on the basis of acetylcholinesterase (AChE), which is often expressed postsynaptically, and not ChAT expression (140). In support of this lack of cholinergic innervation, vesicular ACh transporter expression was not detected in nerve fibers or bundles by immunohistochemistry on tissue sections of human spleen (110). With the limited number of neuroanatomical studies conducted, it is still not clear if the splenic innervation in humans and non-human primates is similar to other animals. In contrast, the sympathetic innervation of the spleen in mice and rats has been extensively described and characterized. The majority of splenic TH⁺ axons are associated with the vasculature, with isolated fibers branching off and projecting into the white pulp (80, 81). These fibers in the white pulp extend towards macrophages and the T- and B-cell zones, but are absent from germinal centers. Development of this innervation and the local guidance cues received by neurons innervation appears to be dynamic, with reinnervation of the spleen following chemical sympathectomy (161). While it is unknown what factors control these processes, cytokines such as IL-17A could have a significant role. IL-17A induces the outgrowth of sympathetic neurons in the colon (48) and cultures from the SMG (54), although this has not been shown in secondary lymphoid organs, including the spleen. This role is also complicated by the ability of neurons to direct formation of lymphoid tissues during development and infection. Neuronal release of retinoic acid activates lymphoid stromal cells to induce the recruitment of lymphoid tissue inducer cells during development (258). This interplay between neurons and stromal cells is not restricted to development and occurs during the formation of lymphoid aggregates (tertiary lymphoid tissues) during inflammation (192). These studies suggest that the integration of nervous and immune system begins early during development and continues to shape lymphoid organ function throughout life. It is tempting to further speculate that disruption of these early processes could induce alterations in lymphoid structure and functional changes.

Sympathetic innervation of the spleen in humans and non-human primates is considerably less well characterized. For example, while electron microscopy analysis of human spleen identified neuronal fibers associated with the vasculature and trabeculae, no innervation of the white pulp was observed (109). Other studies have confirmed that as in the mouse, sympathetic neurons innervate of lymph nodes along the vasculature and the parenchyma, terminating near lymphocytes (110, 239). Performing these types of comparative anatomical studies would be significant opportunities for discovery and would likely aid the development of new therapeutics. Given the difficulties in obtaining a healthy human spleen, non-human primate models could conceivably provide significant insight.

IV. NONNEURONAL SOURCES OF NEUROTRANSMITTERS

A. T Cells as Non-neuronal Sources of ACh

Despite the controversies over the precise pathways involved, it is generally accepted that activation of vagal efferent axons in mice and rats induces release of NE, resulting in ACh release in the spleen from non-neuronal sources. Electrical stimulation of the spleen has long been known to induce the release of ACh even in mice and rats despite the lack of cholinergic innervation. The source of this ACh is from non-neuronal sources, primarily specialized subsets of B and T lymphocytes (69, 214, 224). Production of ACh had been described in transformed human and mouse lymphocyte cell lines, and in primary rat B and T cells (218). These findings suggested that given the appropriate stimuli, lymphocytes could produce and release ACh to regulate physiological processes.

In the CAIP, T cells in the spleen are critical, serving as the source of ACh released in response to neuronal NE. As evidence of this, mice lacking T cells, or recipients of transferred T cells with short hairpin RNA mediated knockdown of ChAT, were not protected from endotoxemia by VNS. T-cell expression of ChAT and production of ACh were validated using a ChAT-GFP reporter mouse where the ChAT promoter drives GFP expression (247). Analysis of the spleens and lymph nodes of these mice demonstrated a small subpopulation of CD4⁺ T cells were GFP⁺ and made ACh as determined by mass spectrometry. These CD4⁺ ChAT-GFP were found to express the β2-adrenergic receptor (β2AR) and respond to NE by releasing ACh, suggestive of the ability of ChAT⁺ T cells to respond to activation of the sympathetic innervation (224).

Subsequent studies determined that despite this novel ability, CD4⁺ ChAT-GFP⁺ T cells were not a novel T helper subset, or restricted to a specific lineage. Although flow cytometric immunophenotyping indicated ChAT-GFP⁺ T cells were enriched for Foxp3⁺ regulatory T cells, ChAT expression is not restricted to this or other Th lineages (214). Differentiation of naive CD4⁺ ChAT-GFP^- T cells under conditions to induce T_h1, T_h2, T_h17, or T_reg T cells in vitro failed to significantly induce ChAT-GFP expression (214). More recent studies have indicated ChAT is also increased in CD4⁺ T_h17 T cells in the small intestine (69). Curiously, this endogenous population of ChAT⁺ T cells in the lamina propria was not observed by confocal imaging of the intestine of a reporter mouse line where Cre recombinase was expressed under the control of the ChAT promoter allowing for tdTomato expression (91). Although it is tempting to consider that different signals and stimuli experienced by T cells in the intestine versus ex vivo could explain these differences, it is also possible that ChAT⁺ T cells are not a unique subset and do not fit into pre-established Th lineages.

Supporting this contention, a genetic fate map approach revealed that ChAT expression is transient, with CD4⁺ T cells capable of losing and re-expressing ChAT if restimulated through the T-cell receptor (214). The commensal bacteria also have a unique role in inducing expression of ChAT in T cells. Antibiotic-mediated depletion of the microbiota by treatment of mice from weaning to adulthood with a cocktail of broad-spectrum antibiotics significantly reduced the number of ChAT⁺ T cells (214). These cells not only respond to the microbiota, but appear to regulate it as well. Recent studies further suggest that CD4⁺ ChAT⁺ T cells function to alter diversity of the microbiota through altering antimicrobial peptide production (69). Together, these data suggest that ChAT⁺ T cells could be part of a dynamic host-commensal interaction.

These ChAT⁺ T cells have also been proposed to be part of an alternative pathway for a non-neural mechanism of vagal communication. In this proposed model, stimulation of vagal efferents elicits migration of ChAT⁺ T cells to the spleen from the intestinal tract. Once stimulated, the ChAT⁺ T cells are proposed to migrate to the spleen following an as-yet-undefined signal for homing, while continually releasing ACh. In the spleen, ACh derived from these cells then acts on α7R on splenic sympathetic terminals to induce the release of NE, inhibiting macrophage activation (170). In this model, it is uncertain how a neuronal signal could stimulate a T cell in a secondary lymphoid organ to home to the spleen. The local environment of the PP typically programs expression of adhesion molecules and chemokine receptors that will induce homing to the intestinal mucosa (180). Finally, it would seem unlikely that α7R expressed on neurons is required in the transmission of signaling for immune regulation. As evidence of this, the CAIP was abrogated in septic shock (193) and acute kidney injury (97), only in bone marrow chimeric mice lacking the α7R on cells of hematopoietic origin, but not on neurons or other radio-resistant cells. These data suggest that within the spleen, the model proposed by Tracey and colleagues (FIGURE 1), whereby T cells releasing ACh act on target immune cells through activation of α7R, remains the most realistic interpretation of the current experimental data. Critically, this does not preclude the existence of other neural immune circuits within the spleen.

FIGURE 1.

The proposed circuitry of the cholinergic anti-inflammatory pathway. The cholinergic anti-inflammatory pathway is proposed to begin with the detection of inflammation in the periphery. This information is relayed by vagal afferent neurons to the nucleus tractus solitarius (NTS) resulting in vagal efferent (parasympathetic) outflow to sympathetic ganglia that innervate the spleen and release norepinephrine (NE). This NE activates β2-adrenergic receptor (β2AR) on choline acetyltransferase (ChAT⁺) T cells that function as signaling intermediaries causing ACh to be synthesized and released. Adjacent macrophages express the nicotinic α7-receptor (α7R), which when activated by T cell-derived ACh reduces NF-κB signaling and prevents tumor necrosis factor (TNF)-α production. The alternative sympathetic efferent arm is also depicted whereby the efferent arm comprises pre- and postganglionic sympathetic neurons independent of the involvement of parasympathetic neurons. DAMPs, danger-associated molecular patterns; PAMPs, pathogen-associated molecular patterns; DMN, dorsal motor nucleus of the vagus.

B. Other ACh-Producing Immune Cells

Surprisingly, ChAT-GFP reporter mice revealed that other immune cells can produce ACh. B cells in the spleen and lymph nodes are by far a larger population of ChAT-GFP⁺ cells compared with CD4⁺ T cells. In the spleen, we identified that marginal zone and follicular B cells express ChAT at a higher frequency and total number compared with T-cell population (214). The largest proportion of ChAT⁺ B cells was found within the B1 B cells population in the peritoneal cavity. The function of these various ChAT⁺ B cell types is currently unknown, although the signals the B cells respond to are vastly different from the ChAT⁺ T cells. Analysis of the neurotransmitter receptor expression identified that ChAT-GFP⁺ sorted B cells were significantly enriched in peptidergic neurotransmitter receptors and lacked adrenergic receptors (214). These data suggest that ACh from B and T cells fulfill unique physiological roles in response to specific stimuli. Similar to T cells, ChAT expression in B cells can be induced by the commensal microbiota and could suggest a unique role for these cells in host commensal interactions. As with T cells, ChAT expression in B cells was significantly reduced in mice treated with an antibiotic cocktail, and could be rapidly induced by TLR agonists. These data suggest that TLR agonists derived from the commensal microbiota regulate ChAT expression in lymphocytes (214). Further supporting the role for ChAT induction by the microbiota, B cells isolated from the spleen of ChAT-GFP mice at E18.5 were GFP^- unless stimulated with TLR agonists. These data suggest that colonization of animals with the commensal microbiota at birth would appear to exert a unique influence over the expression of ChAT in immune cells. Although a small number of splenic macrophages (214) and dendritic cells (DC) express ChAT (214, 227), a recent publication did not confirm ChAT expression in DC within the Peyer’s patch and mesenteric lymph nodes (69). It is an intriguing and unexplored possibility that these differences could reflect unique populations of DC in these tissues, or that differences in the microbiota in the mouse colonies of the two laboratories. These many questions represent a significant need to better understand not only the regulation of ChAT in non-neuronal cells, but also the stimuli required for ACh release.

C. Immune Cell Production of Noncholinergic Neurotransmitters

Immune cells have been reported to synthesize and release neurotransmitters other than ACh. These include catecholamines such as NE and dopamine, to molecules such as GABA described below.

D. Norepinephrine

As opposed to simply serving as a target for NE, immune cells have been suggested to produce NE as well. Stimulation of rat neutrophils and alveolar macrophages with LPS not only elicited release of NE and epinephrine, but increased transcription of tyrosine hydroxylase and dopamine β-hydroxylase (84). As these genes encode enzymes required for the synthesis of catecholamines, these data suggested catecholamine biosynthesis and release is triggered by PAMPs. More importantly, Flierl et al. (84) demonstrated that macrophage-derived catecholamines exacerbated tissue damage in a model of lung injury in rats via activation of α2 adrenoceptors on alveolar macrophages. Together, these data suggest that phagocytes can synthesize and release catecholamines in response to PAMPs, and that autocrine signaling via α2 adrenoceptors may have proinflammatory effects. Although intriguing, TH staining in immune cells within the spleen or other secondary lymphoid organs has not been reported (185). It is important to note that there are four splice variants of TH in mice, raising the possibility that either immune cells express a unique isoform of TH not detectable by some antibodies or that macrophages cannot produce this neurotransmitter. An inability of macrophages to produce NE is supported by studies using a fate-map reporter system. Mice expressing Cre recombinase under the control of the TH promoter, and possessing a lox-stop-lox tdTomato allele were found to have tdTomato⁺ neurons, but not macrophages (89). These data would suggest that macrophages are not a large source of catecholamines, or that expression of the biosynthetic machinery occurs under specific circumstances.

E. Dopamine

Non-neuronal cells including B and T cells produce, take up, and respond to dopamine (151). T cells isolated from CSF and single-cell propagation of B and T cell clones contained and released dopamine upon T-cell stimulation (16). These results from long-term propagation cultures were early evidence that lymphocytes could produce dopamine, as opposed to simply taking up and storing this molecule. Subsequent studies have indicated that the enzymatic machinery required for synthesis, TH, and dopamine β-hydroxylase are expressed by lymphocytes and phagocytes (84). Again, given the requirement for TH as the rate-limiting enzyme for catecholamine synthesis, lack of TH immunoreactivity and reporter gene expression in mice would seem to contradict these findings. Uptake from the extracellular environment has also been observed with increased intracellular dopamine content after culture in media supplemented with this neurotransmitter (16). In support of this ability, human peripheral blood lymphocytes express the plasma membrane dopamine transporter DAT, and uptake extracellular ³H-labeled dopamine in a DAT-dependent manner (168). These results suggest that if the genes encoding the biosynthetic pathway are not expressed by lymphocytes, these cells could potentially function as a delivery vehicle for dopamine.

F. GABA

Although considered the main inhibitory neurotransmitter in the CNS, GABA is also present in the periphery in a variety of tissues (98) and immune cells (20, 71). Human lymphocytes express glutamate decarboxylase isoforms that synthesize GABA, the proteins involved in GABA vesicular loading, transport, and GABA receptors (71). Lymphocytes not only take up extracellular GABA; these cells also express GABA_A receptors, with GABA inducing currents in patch-clamped lymphocytes and inhibition of phytohaemagglutinin-induced proliferation (71). Prior studies in mice suggested that GABA inhibition of proliferation, that could block immune responses in vivo, is predominantly through GABA_A receptors (250). Although it was presumed that the GABA responsive population was predominantly T cells, enrichment of peripheral blood lymphocytes for specific populations was not performed (71).

G. Vasoactive Intestinal Peptide

In addition to neurons and endocrine cells, immune cells produce and respond to vasoactive intestinal peptide (VIP). Both CD4⁺ and CD8⁺ T lymphocytes express mRNA encoding VIP and secrete VIP in response to antigenic stimulation (66). Receptors for VIP are expressed on almost all immune cell types (67). VIP binds three metabotropic receptors: VPAC1, VPAC2, and PAC1 (90), which are coupled to Gα_s, leading to the generation of cAMP and activation of protein kinase A. In addition, in monocytes and macrophages, PAC1 receptors also couple to phospholipase C and protein kinase C activation. VIP has anti-inflammatory effects on innate immune cells, enhances generation of T_reg cells, and skews CD4⁺ T-cell differentiation towards a T_h2 phenotype (90). These attributes engender a restitutive effect in vivo, with VIP significantly attenuating production of T_h1-driven induced colitis (1). This ability of immune cells to produce and respond to VIP demonstrates the broad array immune functions of classical neurotransmitters and peptide hormones.

H. The Interface Between Neurons and Immune Cells

The physical interface and mechanism(s) of communication between the nervous and immune system are a topic of active investigation. Synapse-like interactions between axons and lymphocytes have been observed in the spleen and are characterized by lymphocytes within 5–15 nm of varicosities containing NE (82). Despite this seminal work, the frequency of interactions between ChAT⁺ B and T cells has only recently been assessed. Confocal microscopy, and CLARITY imaging of spleens from ChAT-GFP reporter mice, has shown very few ChAT-GFP⁺ cells are found in close proximity to TH⁺ axons (185). These data suggest that interactions between sympathetic axons and ChAT⁺ lymphocytes in the spleen are rare, or transient in nature. It is unknown if adhesion molecules maintain this interaction between lymphocytes and axons, or how long these cells interact. These questions and the functional significance of this interaction remain unexplored and are exciting areas of further study.

V. MODULATION OF IMMUNE FUNCTION BY ACETYLCHOLINE

Inhibition of immune responses through an anti-inflammatory reflex arc requires that the neuronal signaling is detected by the target immune cells and impinges on immunological processes. For the anti-inflammatory reflex, identification of the mechanism of silencing macrophage activation has proved challenging. The following sections serve to review the current state of these efforts and identify remaining opportunities and challenges to further advance the field.

A. Target Cells and Mechanisms of ACh-Induced Regulation

Reduced morbidity and mortality afforded by VNS during endotoxemia occurs by ACh-mediated inhibition of signaling in immune cells. In mouse models of endotoxemia, proinflammatory cytokine production is the basis for multiorgan system dysfunction. While injection of LPS or polymicrobial sepsis can activate a number of immune cells, splenic macrophages are implicated as the predominant source of TNF-α (119). Although macrophages express both muscarinic and nicotinic acetylcholine receptors (summarized in TABLES 2 and 3), the nicotinic alpha 7 receptor (α7R) is required for ACh-induced inhibition of TNF-α production. Experiments using the nicotinic receptor antagonist α-bungarotoxin in vitro reduced the ability of ACh to reduce LPS-induced TNF-α production. Further confirming the requirement of α7 in VNS-mediated protection during septic shock, mice lacking the α7R were not protected from LPS-induced TNF-α production with VNS (265). This mechanism of α7R-dependent inhibition has been expanded to numerous types of immune cells including neutrophils and dendritic cells. At this point, it is uncertain how many other immune cell populations express α7R, as there are few antibodies validated in KO mice. Use of fluorescent reporter protein expression to mark cells expressing α7R appears to be restricted to a fate map approach. Such an approach that uses mice with Cre recombinase expression controlled by the α7R promoter and a LoxP-STOP-LoxP-YFP allele has identified 20–25% of myeloid and lymphoid cells in the adult bone marrow, spleen and Peyer’s patches expressed α7R. As Cre-mediated excision of the STOP is irreversible, it is uncertain if these cells expressed α7R at some point or continue to express α7R. It would be interesting to use this approach together with a mouse designed to express a fluorescent protein under the direct control of the α7R promoter in a fate map approach. These types of studies could aid our understanding of α7R expression in these populations of cells and help to determine if expression occurs during immune cell development or can be triggered in the periphery.

Table 2.

Expression and function of muscarinic receptors on immune cells

Primary Cell Type	Receptors	Species	Tissue/Source	Function
B cell	M1	Human	PBMC	ND (249)
	M2	Human	PBMC	ND (249)
	M3	Human	PBMC	ND (249)
	M4	Human	PBMC	ND (249)
	M5	Human	PBMC	ND (249)
T cell	M1	Human	Spleen	↑TNF-α, IFN-γ, IL-6 (130)
		Mouse	Spleen	↑IL-10, ↑IL-17, ↓IFN-γ (209)
	M2	Human	PBMC	ND (53)
	M3	Mouse	Spleen	T cell activation-host defense (63)
			MLN	CD4+ T cell activation, cytokine production, Ca2+ signaling (63)
		Human	PBMC	ND (57, 131)
	M4	Human	PBMC	ND (53, 57)
		Mouse	Spleen	↑IL-10, ↑IL-17, ↓IFN-γ (209)
	M5	Human	Spleen	↑IL-6, ↑IFN-γ, TNF-α (130)
		Human	PBMC	ND (57)
		Mouse	Spleen	↑IL-10, ↑IL-17, ↓IFN-γ (209)
Neutrophils	M1	Human	Peripheral blood	ND (13, 207)
	M2	Human	Peripheral blood	ND (207)
	M3	Human	Peripheral blood	↑NETosis (45), ND (13, 207)
	M4	Human	Peripheral blood	ND (13)
	M5	Human	Peripheral blood	ND (13)
Macrophages	M1	Human	Alveolar	ND (207)
		Mouse	Peritoneal	ND (132)
	M2	Human	Alveolar	ND (207)
		Mouse	Peritoneal	ND (132)
	M3	Guinea pig	Alveolar	↑Inflammation, host defense, ↑IL-8 (203)
		Human	Alveolar	ND (207)
		Mouse	Peritoneal	ND (132)
	M4	Mouse	Peritoneal	ND (132)
	M5	Mouse	Peritoneal	ND (132)
Dendritic cells	M1	Mouse	Bone marrow derived	ND (132)
	M2	Mouse	Bone marrow derived	ND (132)
	M3	Human	Nasal mucosa	↑OX40L (158)
		Human	PBMC derived	Differentiation: ↑TNF-α, ↑HLA-DR; LPS stimulation: ↓TNF-α, ↓HLA-DR (227)
		Mouse	Bone marrow derived	ND (132)
	M4	Human	PBMC derived	Differentiation: ↑TNF-α, ↑HLA-DR; LPS stimulation: ↓TNF-α, ↓HLA-DR (227)
	M5	Human	PBMC derived	Differentiation: ↑TNF-α, ↑HLA-DR; LPS stimulation: ↓TNF-α, ↓HLA-DR (227)

IL, interleukin; LPS, lipopolysaccharide; TNF-α, tumor necrosis factor-α; IFN-γ, interferon-γ; ND, no difference; HLA-DR, human leukocyte antigen-antigen D related; PBMC, peripheral blood mononuclear cells; MLN, mesenteric lymph nodes.

Table 3.

Expression and function of nicotinic receptors on immune cells

Primary Cell Type	Receptors	Species	Source	Function
B cell	α2	Mouse	Spleen	ND (132)
	α3	Human	PBMC	ND (231)*
		Mouse	Spleen	ND (133)
	α4	Mouse	Spleen	↓Antigen specific Ab (238)
	α5	Human	PBMC	ND (231)
		Mouse	Spleen	ND (132)
	α6	Human	PBMC	ND (231)*
		Mouse	Spleen	ND (132)
	α7	Human	PBMC	ND (231)*
		Mouse	Spleen	ND (132, 238), ↓Antigen specific Ab (88)
	α9	Human	PBMC	ND (231)
		Mouse	Spleen	ND (132)
	α10	Human	PBMC	ND (231)
		Mouse	Spleen	ND (132)
	β2	Mouse	Spleen	↓Antigen specific Ab (238)
T cells	α2	Mouse	Spleen	ND (132)
	α3	Human	PBMC	ND (231)*
		Mouse	Spleen	ND (132)
	α5	Human	PBMC	ND
		Mouse	Spleen	ND (132)
	α6	Human	PBMC	ND (231)*
	α7	Human	PBMC	ND (231)*
		Mouse	Spleen	ND (132), ↓Proinflammatory cytokines (88)
	α9	Human	PBMC	ND
		Mouse	Spleen	ND (132)
	α10	Human	PBMC	ND (231)
		Mouse	Spleen	ND (132)
Neutrophils	α2	Mouse	Peritoneal	↑Ca2+ signaling (226)†
	α3β2	Human	PBMC	↑Reactive oxygen intermediates, ↑IL-8 production (121)
		Mouse	Peritoneal	↑Ca2+ signaling (226)†
	α4	Human	PBMC	↑Reactive oxygen intermediates, ↑IL-8 production (121)
		Mouse	Peritoneal	↑Ca2+ signaling (226)†
	α5	Mouse	Peritoneal	↑Ca2+ signaling (226)†
	α6	Mouse	Peritoneal	↑Ca2+ signaling (226)†
	α7	Human	PBMC	↑NETosis (147)
		Mouse	Peritoneal	Regulation of inflammatory responses, ↑Ca2+ signaling (226)†
	α9	Mouse	Peritoneal	↑Ca2+ signaling (226)†
Macrophages	α2	Mouse	Peritoneal	ND (132)
	α4β2	Human	Alveoli	↓IL-6, IL-12, and TNF-α
			Peritoneal and mucosa	↑Phagocytosis, ↓NF-κB signaling, ↓ TNF-α, ↓CCL2, ↑IL-10 (259)
		Mouse	Peritoneal	ND
	α5	Mouse	Peritoneal	ND (132)
	α6	Mouse	Peritoneal	ND (132)
	α7	Human	PBMC	↓NF-κB activation, ↓TNF-α (265)
		Mouse	Peritoneal/spleen	ND (132), ↓NF-κB activation, ↓TNF-α (265); ↓ATP-induced IL-1β cleavage (163)
			Muscularis externa	↓IL-1α, IL-1β, IL-6, and CCL2, ↑ATP-induced Ca2+ transients (173)
	α10	Mouse	Bone marrow derived	ND (132)
Dendritic cells	α2	Mouse	Bone marrow derived	ND
	α5	Mouse	Bone marrow derived	ND
	α6	Mouse	Bone marrow derived	ND (132)
	α7	Mouse	Spleen	↓IL-12, ↓activation of T cells, ↓proinflammatory cytokine release (184)
			Bone marrow derived	ND
	α10	Mouse	Bone marrow derived	ND (132)

^*Mixed population of cells used, impossible to determine precise immune cell phenotype. †Specific receptor not determined. Ab, antibody; IL, interleukin; TNF-α, tumor necrosis factor-α; ND, no difference; PBMC, peripheral blood mononuclear cells.

B. Intracellular Mechanisms of ACh-Mediated Control

Inhibition of TNF-α production by ACh-elicited α7R-dependent signaling has been suggested to involve a wide range of mechanisms from sequestering of signaling proteins, to induction of microRNAs. These multiple proposed mechanisms are likely a consequence of the α7R being a homopentamer of α7 subunits appearing to be an ion channel, with ligand binding also activating intracellular signaling cascades.

As TNF-α production is dependent on NF-κB-mediated signaling and gene transcription, the focus has been on α7R-induced disruptions to this signaling cascade. A key role for the signal transducer of activated T cells 3 (STAT3) has been implicated, in studies with conflicting proposed mechanisms. Nicotine, a nonselective agonist of nicotinic receptors, significantly increased STAT3 phosphorylation while reducing LPS-induced TNF-α production in macrophages (64). Inhibition of TNF-α production by nicotine was dependent on α7R, as pretreatment with selective antagonists restored TNF-α production. These data suggested that α7R activation increases STAT3 phosphorylation, consequently blocking NF-κB-dependent gene expression. Increased STAT3 phosphorylation was also observed in vivo in intestinal muscularis macrophages following VNS (64). VNS-induced activation of STAT3 was crucial to limiting postoperative ileitis, as conditional ablation of STAT3 in macrophages abrogated the protective effect of this treatment. Stimulation of α7R was suggested to increase STAT3 phosphorylation by recruitment and activation of Janus associated kinase-2 (JAK2). Currently, the mechanism responsible for this α7R-induced aggregation of JAK2 is unknown.

In contrast to these findings, α7R agonists prevented LPS-induced STAT3 phosphorylation by JAK2 in the THP-1 macrophage cell line. Inhibition of STAT3 phosphorylation by treatment with α7R agonists or JAK2 inhibitors, siRNA targeting of STAT3, or expression of a STAT3 phosphorylation mutant prevented TNF-α production. These studies suggest that LPS-induced STAT3 phosphorylation is required for optimal NF-κB signaling (202). It is important to note in these studies that the α7R agonist choline reduced TNF-α production in splenocytes and the RAW264.7 cell line at concentrations that did not reduce STAT3 phosphorylation. In support of this proposed mechanism, the nicotinic receptor agonist GTS-21 reduced IL-6-induced JAK2 activation and STAT3 phosphorylation (49). Although induction of STAT3 signaling is not the canonical response to LPS, there is a precedence of LPS-induced JAK2 activation in the RAW264.7 cell line (136). At present, it is not possible to reconcile these proposed roles of STAT3 in α7R-depenant signal transduction, although it is possible that there are multiple downstream signaling pathways induced depending on the strength of the stimulation.

As another alternative, α7R agonists have also been implicated in altering posttranslational processes that would reduce TNF-α release. TNF-α requires cleavage and activation by the enzyme TNF-α converting enzyme (TACE) to become biologically active. Stimuli that reduce TACE expression or activity are therefore able to indirectly block the release of biologically active TNF-α (35). Stimulation through the α7R increased expression of the microRNA mir124, which targeted TACE mRNA, reduced TACE expression, and consequently decreased TNF-α processing (245). Although this study would not account for prior observations of reduced NF-κB DNA binding upon α7R stimulation, it nonetheless highlights that alternative pathways should be explored in immune regulation following α7R stimulation.

Other mechanisms could include a range of cellular responses ranging from activation of kinases to secondary messengers. Activation of T cells through nicotinic receptors can result in intracellular signaling events characterized by kinase activation. T cells treated with nicotine activated T-cell phosphorylation cascade proteins, and this increase was dependent on the SRC family kinases Lck and Fyn leading to Ca²⁺ mobilization (213). While proposed that the α7R and TCR were physically associated, it is unknown if this is a direct or indirect interaction, or how nicotine signaling through α7R induces activation of the T-cell signaling kinases. As T-cell signaling is a highly coordinated process that dictates T-cell function, alterations can have significant ramifications on T-cell-dependent immune function (37).

With the (conflicting) results in the literature, and lack of studies on the structure and signaling elicited by α7R in immune cells, perhaps it would be prudent to examine signaling in neuronal cells. Activation of the α7R following ligand binding allows for opening of the ion channel and influx of sodium or calcium ions, yet it is uncertain if this occurs in non-neuronal cells. Although whole cell patch-clamping experiments have failed to detect α7-induced currents in lymphocytes (262), single-channel openings have been observed in monocyte-derived macrophages following stimulation with nicotine, ACh, and ACh with or without α7R-positive allosteric modulators (9). These studies illustrated that while inward currents can be observed, this occurred with agonist concentrations that are insufficient to prevent TNF-α production. These data indicate that stimulation of immune cells with α7R agonists elicits multiple cell signaling events and that mechanisms other than ion channel opening are critical for regulating the response to DAMPs.

Other mechanisms of α7R elicited signal transduction that have also been described in neurons. Stimulation of hippocampal neurons with α7R agonists increases adenylate cyclase isoform 1, increasing intracellular cAMP (52). Activation of the membrane-associated adenylate cyclase isoform 6 has also been reported following stimulation of PC12 cell line with α7R agonists. While it is by no means certain that a signaling axis in one cell type will be conserved in immune cells, it is worth noting that increased cAMP can inhibit the activation of a variety of immune cells. In T cells, increased cAMP production prevents signaling from the T-cell receptor, followed by increased PKA-induced phosphorylation of Lck at tyrosine 505 (225, 260, 261). This phosphorylation blocks T-cell activation, preventing differentiation, cytokine production, and proliferation. Inhibition of macrophage activation by cAMP-elicited signaling can also occur through induction of c-Fos binding to NF-κB p50. This sequestration prevents NF-κB-mediated gene transcription reducing production of proinflammatory cytokines (137). It is also important to note that increased cAMP in immune cells does not always lead to immune suppression. In T cells, increased cAMP and downstream signaling cascades with prolonged exposure to IL-1β enhances differentiation into IL-17A producing T_h17 T cells (153). Stimulation of α7R should not always be equated with immune inhibition, and the signaling pathways and mechanisms of action require substantial investigation in well-defined primary immune cells.

C. Activation of Other ACh Receptors on Immune Cells

Immune cells have been reported to express a wide array of muscarinic and nicotinic acetylcholine receptors that enhance or decrease the activation. Despite this reported expression of nicotinic and muscarinic receptors, and an ability to respond to these agonists, this expression is not reflected in databases such as the immunological genome project (106). At this point it is unknown if this lack of expression is due to purely technical or biological reasons. For example, is lack of receptor mRNA due to processing induced transcriptional changes that can accompany isolation, or are receptor positive cells a small fraction of the overall population. Despite these concerns, we have included literature that identifies expression by a combination of techniques including functional assays and summarized these findings for muscarinic receptors in TABLE 2 and nicotinic receptors in TABLE 3.

1. Muscarinic receptors

Various immune cells have been found to express muscarinic receptors, with selective agonists impacting cellular functions (TABLE 2). As professional antigen presenting cells that have a unique ability to stimulate naive T cells bearing an appropriate TCR for an antigen, DC are fundamental to immune responses. Human DC derived from blood CD14⁺ monocytes express muscarinic receptor types M3, M4, and M5. Stimulation during differentiation into DC, or stimulation of mature DC cultures with the muscarinic receptor agonist carbachol increased HLA-DR and TNF-α expression. While suggestive of a proinflammatory state, expression of costimulatory proteins such as CD80 expression were unaltered (227). Muscarinic receptor stimulation may therefore regulate specific aspects of DC function and depend on the local environmental stimuli. In support of this, carbachol decreased LPS-induced HLA-DR, IL-12, and TNF-α expression (227), suggesting a highly complex signaling network, and an ability to either enhance or reduce T cell responses depending on coexposure to DAMPs and muscarinic agonists. Complicating our understanding of the effects of muscarinic receptors on adaptive immune function in vivo, T cells have also been identified to express muscarinic receptors that modulate cell function. The M3 receptor is expressed on CD4⁺ T cells and is required to mount an immune response directed against helminths and enteric bacterial pathogens. Mice deficient in M3 receptor (M3R) are more susceptible to infection with Nippostrongylus brasiliensis and Salmonella enterica serovar Typhimurium (63). Deficiency in M3R reduces T-cell Ca²⁺ signaling and activation, culminating in decreased IL-13 and interferon (IFN)-γ production during N. brasiliensis and S. typhimurium infection, respectively. T-cell intrinsic M3R signaling is crucial to this response, as adoptive transfer of WT T cells but not M3R^−/− T cells from previously infected donor mice into M3R^−/− hosts significantly reduced parasite burden after challenge with N. brasiliensis. These data would suggest that ACh signaling through the T cell M3R, and not host cell M3R is responsible for this phenotype (63).

In addition to T cells, muscarinic receptors are also expressed on B cells. Although radiolabeled 3-quinuclidinyl benzilate bound mouse splenocytes and enriched B cells (6), binding was not observed with N-methylscopolamine to human B cells (77). As both agents bind to muscarinic receptors, these data have been interpreted as mouse B cells, but not human B cells are sensitive to muscarinic receptor agonists. More recent immunocytochemistry studies have suggested surface expression of M1-M5 receptor subtypes on human B cells (249). The immunological ramifications of stimulation through these receptors are not well characterized in the steady state or during infection.

Similar to these cells of the adaptive immune system, innate immune cells express muscarinic receptors and have been implicated in host defense and immune function. Deficiency in M3R significantly increases susceptibility to the mouse enteric pathogen Citrobacter rodentium, increasing bacterial burden and reducing clearance. Increased susceptibility in this model of enteric bacterial infection is due to reduced macrophage function and mucus production, and not a T-cell intrinsic defect (176). Expression of M1–3 receptors is not limited to macrophages in the intestine, with expression also found in the lung macrophages of mice and humans (102, 207). Stimulation of alveolar macrophages with ACh induced in vitro migration, that was blocked by cotreatment with the M3 receptor antagonist tiotropium (38). This role for ACh activation of M3R to increase inflammation or host defense has also been suggested in studies where tiotropium administration reduced lung inflammation induced by repeated instillation of LPS as a model of chronic obstructive pulmonary disorder (203). It is important to note, however, that the cell population targeted by ACh in this study was not elucidated. Functionally, stimulation through these receptors have been suggested to aid host defense by increasing macrophage production of leukotriene B₄, a chemotactic signal for neutrophils (207). It is however important to note that increased leukotriene B₄ production was only observed in sputum cells from patients with chronic lung pathology, and not samples from control subjects. Despite clear functional evidence of the role of muscarinic receptors in macrophage function, it is interesting to note that expression may be linked to differentiation within tissues. As evidence of this possibility, circulating monocytes that differentiated into macrophages after recruitment do not express any of the five muscarinic receptor subtypes (102). These findings could suggest that muscarinic receptor expression by macrophages could be tissue specific, occurring in response to local signals from tissue.

Other innate immune cells including neutrophils in peripheral blood express M1R and M2R, with varying reports of M3R (207) as well as M4 and M5 (13) expression. Neutrophils protect against pathogens through a variety of mechanisms including phagocytosis and production of free radicals such as superoxide and peroxynitrate. It is unclear if these fundamental immune functions are enhanced or reduced following stimulation by muscarinic receptor-selective agonists. In addition to phagocytosis and free radical production, neutrophils can also produce neutrophil extracellular traps (NETs) to capture and to kill extracellular pathogens (36). NETs are globular structures comprised of neutrophil DNA, histones, and other host cell proteins (36). Release of NETs has been observed following stimulation with ACh through the M3 receptor (45). While tempting to conclude that this response could aid host defense during infection, NETs release can also result in immunopathology (45, 59, 187). These studies should illustrate that despite the tremendous advances in the role of muscarinic receptors in immune modulation, there are significant gaps in our knowledge.

2. Nicotinic receptors

In addition to muscarinic receptors, numerous studies have identified nicotinic ACh receptor expression on immune cells. These data are summarized in TABLE 3, although it is important to note that for some of these receptors, the functional implications and the precise cell types remain unknown.

Nicotinic receptors containing α4/β2 subunits have been implicated in immunomodulation during intestinal inflammation. Activation of α4/β2 receptors by VNS initiated an anti-inflammatory effect, increasing phagocytic activity and IL-10 production while reducing NF-κB signaling in lamina propria macrophages (259). Thus VNS-induced immunosuppression in the intestinal tract can occur through receptors other than α7R. Activation through nicotinic receptors on macrophages however should not be equated to general host protection, as stimulation of nicotinic receptors in lung alveolar macrophages with nicotine or the nicotinic agonist dimethylphenylpiperazinium reduces bactericidal activity and proinflammatory cytokine production (172).

Similar reductions in macrophages phagocytosis and host defense against bacterial pathogens have been described with cigarette smoke. While nicotine has been implied as the agent of action, cigarette smoke comprises an estimated 7,357 chemical compounds, and as such, it is difficult to determine if the observed effects are due to the presence of multiple agonists of different receptors (221). It is also possible that while alveolar macrophage function is significantly reduced, other populations of macrophages and inflammatory monocytes may not express the same neurotransmitter receptors and may respond in a different manner. In support of this supposition, expression of surface proteins and receptors such as mouse mannose receptor are differentially expressed by alveolar macrophages compared with other macrophage populations (117). Macrophage function and responses to agonists may therefore depend on local tissue environment.

The composition of the nicotinic receptor also has drastic implications for ligand binding and functionality. Nicotinic receptors are comprised of five receptor protein subunits that assemble to form a pentameric structure with a ligand-binding site and channel (4, 181). Unlike the homeric α7R, heteromeric receptors, including α4β2, are comprised of different subunits and can adopt a variety of stoichiometries that impact upon function. The α4β2 receptor comprises three α4 and two β2 [(α4)₂(β2)₃] with an alternative (α4)₂(β2)₃ configuration that has high agonist sensitivity (181). Receptor configuration is not static, with prolonged nicotinic agonist exposure increasing the proportion of (α4)₂(β2)₃ receptors (188). It is uncertain if different immune cell populations or macrophages in lung compared with other organs have unique stoichiometries that could impact signaling characteristics.

Indirect effects of acetylcholine on immune cell populations should also be considered in vivo, and significantly complicate interpretation at the tissue or organism level. Despite reduced activity of alveolar macrophages, nicotinic receptor stimulation significantly increased expression of proteins chemotactic for neutrophils. This suggests that while alveolar macrophage function could be reduced, overall host defense as provided by neutrophils might remain intact. Nicotine as a nonselective agonist can also significantly alter the functionality of neutrophils. While leukocytosis has been well documented in tobacco smokers, some studies have identified increased responsiveness to chemokines (254) as well as production of chemokines and reactive oxygen species (ROS) by neutrophils (121, 125). Stimulation of neutrophils from healthy nonsmoker volunteers with nicotine significantly increased IL-8 and superoxide production in vitro (121). It is worth noting that while the dihydrorhodamine 123 (DHR) reagent has become a gold standard in the evaluation of ROS there are several concerns that complicate the use of DHR to accurately determine ROS generation (127). In neutrophils, nicotine has also been reported to enhance release of lysozyme, neutrophil elastase, β-glucuronidase, prostaglandin E₂, and the leukotrienes. While increased release of these compounds could be viewed as beneficial for host protection, nicotine treatment of macrophages decreased phagocytosis while having no effect on chemotaxis to chemokines, or cell viability (236). At this time, it is unknown what receptors are required for these increased host-defense responses to occur. Curiously, nicotine has also been reported to reduce the bactericidal activity of human peripheral blood neutrophils. Exposure of cells to high concentrations of nicotine that would be found in chewing tobacco or smokeless tobacco products reduced bacterial killing by directly absorbing superoxide (194). Together, these data indicate the continued need to carefully define the receptors required for these effects of nicotine on the vital functions of innate immune cells.

VI. OTHER NEUROIMMUNE CIRCUITS

Although the vagally activated CAIP has been the focus of intense interest and study with application to regulation of immune activity in a variety of conditions, it is becoming increasingly clear that other reflexes occur.

A. Neural Modulation of Intestinal Immune Function

The interactions between the nervous and immune system are particularly interesting in the gastrointestinal tract. The gastrointestinal tract contains a large component of the immune system in close proximity to the nerve terminals of enteric neurons and adrenergic postganglionic sympathetic neurons with cell bodies located in prevertebral ganglia. As an interface between the body and the environment, the immune system in the gastrointestinal mucosal must balance the requirement to protect against pathogenic microbes, while being tolerant to components of the diet and commensal microbiota. Inappropriate immune reactions to the commensal microbiota, or components originating from the diet have the potential to have significant detrimental effects on the health of an individual. While the precise etiology of IBD has remained elusive, it is generally accepted that aberrant immune system activation is an underlying cause (3, 271). There are an estimated 4–4.5 million people in the United States and Europe with IBD (128). This chronic inflammation presents clinically with abdominal pain, fever, diarrhea, fatigue, and rectal bleeding (189). Treatments typically are designed to reduced inflammation in general and include nonsteroidal anti-inflammatory drugs, biologicals targeted to specific cytokines, or surgical intervention (3, 21). Although monoclonal antibodies to TNF-α have been revolutionary in the treatment of IBD, these therapeutics can lose efficacy over time and increase the risk of infection (3, 22, 189). Such limitations demonstrate the continued need for novel therapeutics that target the underlying mechanisms of inflammation (85).

Several well-characterized spontaneous genetic and induced models of IBD have been developed that serve to interrogate unique aspects of the disease process. While spontaneous disease models typically have increased variability and incomplete penetrance, induced colonic inflammation models have been developed that mimic specific aspects of pathology. Here we briefly describe these models; however, it is well beyond the scope of this review to detail the nuances of each system, and interested readers are referred to one of several excellent reviews specifically on this topic (135, 162, 244). Acute or chronic inflammation can be induced by exposure to chemical irritants such as dextran sodium sulfate (DSS), or through the adoptive transfer of colitogenic T cells (215, 216, 244). DSS is toxic to intestinal epithelial cells (IEC) resulting in a mid-distal colonic ulceration of the mucosa, with inflammation due to monocytes/macrophages and neutrophil recruitment and activation (17, 191, 252), without requiring T cells (129). While this colitis model has provided insights into the mechanisms of innate immune cell activation and healing of the colonic mucosa, the lack of T-cell involvement provides an incomplete picture. Consequently, an induced T-cell-dependent colitis model was developed using adoptive transfer of FACS sorted effector T cells (T_E) into immunodeficient (RAG1^−/−) hosts (244). It is important to consider that therapeutics that reduce disease severity or do not appear efficacious in one model may only reflect an inability to modulate a specific aspect of disease.

With intestinal inflammation in preclinical models and patients with IBD occurring due to overexuberant immune activation, immune modulation with VNS has emerged as a potential therapeutic. Endogenous vagal nerve signaling has been implicated in neuroimmune regulation of colonic inflammation; with vagotomy significantly increasing the severity of DSS-induced disease (70, 93, 95). Similarly, activation of the efferent arm through central administration of the cholinesterase inhibitor galantamine significantly reduced the severity of DSS-induced colitis in an α7R-dependent manner (126). This role of α7 nicotinic receptors in mediating the vagal anti-inflammatory effects on colitis remains controversial, with evidence in support of (93, 96) and refuting the importance of α7 receptors (70). It is difficult to reconcile these differences in the requirement of α7R in inhibition of DSS-induced colitis. These conflicting results could be due to a range of factors including the differences in commensal microbes or enteric viruses (41) at different institutions. These studies also suggest clear differences in the ability of the CAIP to modulate innate versus adaptive immune cells during colitis. While vagotomy increased DSS-induced colitis severity, no effect of vagotomy was seen in weight loss, pro-inflammatory cytokine production in a T-cell transfer-induced colitis (70). Since the DSS and T-cell transfer models have distinct immunological mechanisms, these results suggest that separate neuroimmune reflexes could inhibit the function of specific types of immune cells in the gut.

Neuronal inhibition of macrophage function that requires the vagus nerve has also been demonstrated in inflammation-induced inhibition of small intestinal motility. Muscularis macrophages, which have been implicated in the pathogenesis of postoperative ileus, are associated with the enteric nervous system and interstitial cells of Cajal (89, 173, 177, 178, 183). These muscularis macrophages are closely juxtaposed to cholinergic axons, evidenced by ChAT staining, and receive inputs from parasympathetic neurons with cell bodies in the DMN (173). Electrical VNS reduced postoperative ileus independently of the spleen, and required expression of α7R on cells of hematopoietic origin that are presumably macrophages (64, 173). These data were interpreted as vagal efferent axons activate cholinergic enteric neurons, which in turn release ACh to inhibit muscularis macrophage activation (FIGURE 2). The circuitry of this neural anti-inflammatory pathway thus appears to differ from the vagal anti-inflammatory pathways that operate during colitis and sepsis. Further suggestive of multiple neuroimmune circuits that could regulate colonic inflammation, a sympathetic neuronal pathway has been described. Elegant experiments using selective sympathectomy in mice before induction of experimental colitis by DSS administration enhanced disease activity index scores, without a concomitant increase in weight loss, or production of proinflammatory cytokines (269). In a similar manner, implantation of a cuff electrode to facilitate stimulation of the superior mesenteric nerve reduced disease activity scores, without reducing the histopathology, endoscopic damage, or cytokine production (269). Although neural stimulation was sufficient to evoke physiological changes, it is uncertain if the selected stimulation parameters were appropriate for regulation of overt inflammation. It is also conceivable that the DSS model of inflammation was simply too severe for this treatment approach. As this exciting technology matures, it may be possible to tailor nerve stimulation parameters to ameliorate specific types of intestinal inflammation. These data highlight the complexity of neural regulation of immune responses in the intestinal tract, and the value of in-depth characterization of the immune cell populations that interact with neurons and are affected by neuronal signaling in situ.

FIGURE 2.

Vagus nerve mediation regulation of intestinal inflammation. Neural inhibition of intestinal inflammation can also occur by a vagal efferent to enteric nervous system (ENS) circuit. Activation of efferent vagus nerve activity induces activation of cholinergic neurons in the ENS and release of ACh in the intestine, blocking macrophage activation in an α7-receptor (α7R)-dependent manner. TNFα, tumor necrosis factor-α.

The influence of vagal immunomodulatory pathways is also evident in the maintenance of immune homeostasis outside the context of inflammatory diseases. Oral tolerance, which prevents local and systemic immune responses to innocuous antigens that have been administered orally, is a critical determinant of health (195). As further evidence of the role of the vagus in intestinal immune function, vagotomy, but not deficiency of α7 nicotinic receptors, prevented the development of oral tolerance to ovalbumin due to a reduction in antigen specific regulatory T cells (70). These findings would suggest that vagal signaling induces a tolerogenic state to antigens derived from diet and microbiota in the intestinal tract.

Communication between the nervous and immune systems may also be involved in host responses to infectious agents in the intestinal tract. The activity of sympathetic neurons innervating the small intestine were found to increase significantly 2 h post-infection with a strain of Salmonella typhimurium unable to invade host cells. These results would suggest that the presence of a bacterial pathogen in the intestinal lumen can be detected, and integrated by the nervous system. As muscularis macrophages in the duodenum, jejunum, and ileum are closely associated with the sympathetic axons of the myenteric ganglia (89, 183), it was proposed that enteric bacterial pathogens in the intestinal lumen could result in neuronal signaling to muscularis macrophages. The functional consequences of sympathetic neural activation were decreased colonic motility, and induction of an alternatively activated (M2) macrophage differentiation program (89). Sympathetic induction of an M2 profile in muscularis macrophages was dependent on expression of β2AR but not α7R (FIGURE 3). Treatment of WT but not β2AR KO cells with NE or salbutamol increased this gene profile. Supporting in vivo data indicated increased expression of genes common to M2 macrophages including arginase and the scavenger receptor Chi3l3 2 h post-infection with S. typhimurium. These exciting data raise the possibility that pathogen detection can occur rapidly and begin to induce a M2 macrophage transcriptional program at sites distant from the area of infection. Although the afferent mechanism was not elucidated, the commensal microbiota was previously shown to induce muscularis macrophages to activate enteric neurons and induce intestinal contractility (183). It is uncertain if S. typhimurium infection can result in macrophage-induced activation of neurons to effect macrophages distant to the site of infection, or if this neuronal induced M2 profile is truly host protective during enteric bacterial challenge.

FIGURE 3.

Sympathetic innervation of the intestine promotes anti-inflammatory macrophage differentiation. Neuroimmune communication in the intestine requires close approximation of sympathetic axons in and macrophages in the muscularis. Activation of sympathetic innervation following infection induced release of norepinephrine (NE) from nerve terminals and induction of an anti-inflammatory macrophage gene profile. β2AR, β2-adrenergic receptor.

It is unclear if only the sympathetic neurons in the myenteric plexus and muscularis macrophages are activated during infection with S. typhimurium or other pathogens. Prior three-dimensional imaging studies of mouse ileum revealed TH⁺ axons in the submucosa, with some, albeit sparse innervation in the macrophage rich mucosa (87). This sparse innervation in the naive mice does not seem to be static, with DSS-induced inflammation increasing the density of sympathetic axons in the colonic lamina propria (48). As an additional complication, tissue resident macrophages are phenotypically distinct from inflammatory monocyte-derived macrophages that have been recruited to a site of inflammation. It is well appreciated that while tissue resident macrophages in the intestine initially are derived from yolk-sac progenitors, these cells are eventually replaced after weaning by monocytes from the blood throughout life (10) and after inflammation (219). It is unknown how inflammation and changes in macrophage phenotypes would impact the bidirectional macrophage-neuronal communication. As evidence for how different populations of macrophages can uniquely respond to neuronal signals, VNS increases phagocytosis by lamina propria macrophages in a α7R-independent manner (259). It is also uncertain if the enteric glia participate in these reflexes in their capacity as neuronal support cells. Enteric glial cells are tightly integrated with the enteric nervous system, exerting regulatory control over colonic physiology, and are activated in tissues with overt inflammation (232, 237). Activation of these cells has also been suggested to contribute to inflammation during DSS-induced colitis and in patients with IBD (56, 76, 237). There is a significant gap in our knowledge of the influence of inflammation on the neuroimmune circuitry in the intestine, and this represents tremendous opportunity for future studies.

B. Activation of Sciatic-Vagus Adrenal Pathway

The requirement for surgical implantation of electrodes and vagal nerve stimulators has been suggested to limit clinical applications during acute inflammation (253). In response, the ability to activate neuroimmune reflex arcs without device implantation have been championed. Electrical stimulation of the sciatic nerve using electroacupuncture was found to reduce LPS-induced macrophage activation, as indicated by reduced proinflammatory cytokine production in mice (253). Protection afforded by electroacupuncture was voltage dependent and required local nociceptive sensory afferents and an intact sciatic nerve. Functional mapping of this circuitry further identified a requirement for the vagus, as cervical or subdiaphragmatic vagotomy abolished the protective effects. Unlike the vagovagal-splenic CAIP, electroacupuncture-induced protection required the adrenal medullae but not the spleen (253). Functional parasympathetic innervation of the adrenal gland was suggested by prior neuroanatomical tracing in rats, with labeling in the DMN after fast blue injection into the adrenal medulla (58, 190). Although dopamine, norepinephrine, and epinephrine were released with sciatic electrical stimulation, only activation of dopamine D1 receptors were required to inhibit TNF-α production (253). Subsequent studies have demonstrated that D1 receptor activation increases intracellular cAMP, attenuating systemic immune responses (273). Highlighting that this pathway is discrete from the other reflexes, electroacupuncture-induced inhibition of TNF-α production was still observed in α7R knockout mice. From these studies there appear to be other immune modulatory neural circuits, beyond the classically described CAIP that involve the vagus nerve.

C. Neuroimmunomodulation of Renal Injury

Evidence of additional neuronal circuits that can prevent immunopathology have been obtained in models of renal ischemia and reperfusion injury (IRI). Experimentally induced ischemia followed by reperfusion of the kidney results in activation and infiltration of immune cells that contribute to organ damage and physiological dysfunction (29). While activation of the CAIP can prevent renal injury, kidney IRI has provided evidence of alternative functional circuitry regulating immune function (122). In addition to the protective CAIP activated by VNS, activation of afferent vagal and the sympathetic renal nerves was observed. Selective afferent stimulation was equally efficacious in protecting against IRI compared with efferent or intact VNS (122). These studies highlight that while efferent VNS is sufficient to induce protection, it is not necessary.

Prior neuroanatomical studies indicated a group of neurons that reside in the ventrolateral medulla, named C1 neurons, can be activated by inflammation and stress and innervate the DMN of the vagus (2, 101). With this in mind, the contribution of C1 neurons to anti-inflammatory responses in kidney IRI was evaluated using an optogenetic approach. This technique utilizes expression of the light-gated cation channel, Channelrhodopsin-2 (ChR2). Pulsing of neurons expressing this protein with specific wavelengths of light can therefore be used to drive activation (32). With the use of this approach, a protective renal sympathetic circuit was activated in mice by optogenetic activation of adrenergic C1 neurons (2). Similar to the cholinergic anti-inflammatory pathway during sepsis, protective effects required α7 nicotinic, β2 adrenergic receptors, and the spleen. While optogenetic simulation of these catecholaminergic C1 neurons increased action potential discharge in parasympathetic and sympathetic pathways, subdiaphragmatic vagotomy did not abrogate protection from IRI. This suggested that protection was mediated predominantly through activation of sympathetic neurons.

As C1 neurons were known to be involved in the response to stressors, protection against renal IRI could also be activated by brief restraint stress. T cells were implicated as being critical to this pathway, with reduced renal damage occurring following transfer of splenic T cells. This protection afforded by T-cell transfer was further enhanced if donor mice had been subjected to restraint stress, or if CD4⁺ T cells were exposed to NE in vitro before transfer (2). While this has been interpreted as protection is due to activation of these CD4⁺ T cells and may even hint at a “memory” effect, it is uncertain if protection is due to a neuronal anti-inflammatory reflex. These data would suggest that once triggered by NE, ChAT⁺ T cells release ACh and continue to do so following adoptive transfer. Moreover, it is uncertain how many ChAT⁺ T cells would be transferred in this paradigm, given that these cells comprise 1–4% of the CD4⁺ T cells in the spleen (214, 224). These data suggest that sympathetic neurons are sufficient to activate the anti-inflammatory pathway via activation of β2 receptors on splenocytes, which in turn leads to release of ACh from CD4⁺ T cells and the activation of α7 nicotinic receptors (FIGURE 4). This motor pathway is very similar to that described for sympathetic immunomodulation during sepsis. However, although the findings of Abe et al. (2) suggest the improvement of renal ischemia reperfusion injury is due to an anti-inflammatory effect, direct measures of immune cell activation were not reported. Prior studies have noted that VNS affords protection during IRI of the kidney by reducing neutrophil infiltration and increasing M2 macrophage differentiation (122), although the role of these immune cells in this model does not appear to be certain at this point (124). Resolving the mechanism of protection in renal IRI will be crucial in our understanding of a new neuroimmune circuit, or to refine previously proposed models. It will also be important to determine whether stimulation of C1 neurons can suppress inflammation in models of RA and IBD.

FIGURE 4.

A sympathetic pathway in the inhibition of systemic inflammation. C1 neurons residing in the medulla oblongata have been implicated as another coordination center that initiates descending signals to reduce immune activation. It is important to note that the immunomodulatory circuitry of this motor pathway does not require the vagus nerve. NE, norepinephrine; TNFα, tumor necrosis factor-α; β2AR, β2-adrenergic receptor; α7R, α7-receptor.

VII. CLINICAL APPLICATIONS OF VNS

Therapeutic devices that stimulate nerves to alleviate disease have become an intense focus in recent years, captivating the imagination and hopes of the lay public and general scientific community. These hopes have been buoyed by several companies seeking to leverage preclinical data to bring clinical neural stimulators for inflammatory disease to market (78). The abundance of positive results in preclinical models of immunopathologies using electrical VNS have prompted considerable effort in bringing these devices to clinic. Clinically, VNS commonly refers to the surgical implantation of helical or cuff electrodes around the left cervical vagus, with the electrodes connected to a stimulus unit. To date, these trials have yielded promising albeit mixed results. As an example of this promise, pilot studies using VNS for the treatment of IBD in patients reported reduced disease severity. In therapy, naive patients using VNS as an alternative to anti-TNF-α, five of seven patients showed significant improvements in the Crohn’s disease activity score, endoscopic evaluation, reduced serum C-reactive protein, and fecal calprotectin (28). At cessation of the study, four of these five responding patients had entered into remission of disease. While the two patients that were refractive to VNS treatment were removed from the study after 3 mo, this outcome may also provide unique insight into VNS as a therapeutic. It is noted that these patients had the most severe disease of this cohort, suggesting that VNS is potentially better suited for maintenance of remission. It is also worth considering if enough preclinical information is available to understand the functional circuitry, and optimal electroceutical stimulation parameters for immunosuppression.

VNS has also proven efficacious in preventing immune cell activation in response to PAMPs or in patients with RA. In patients with epilepsy that were implanted with a VNS device, active stimulation reduced TNF-α production from isolated leukocytes in response to LPS up to 4 h after VNS. Reduced responsiveness to LPS following VNS was also observed in leukocytes isolated from patients with RA, in addition to reduced disease severity (138). These exciting studies are among the first to demonstrate significant reductions in the clinical severity of RA.

Despite these positive outcomes, the invasive nature of electrode implantation has also led to study of noninvasive techniques for VNS. Preclinical studies transcutaneous vagal nerve stimulation (tcVNS), where the electrode is placed externally on the neck during stimulation, indicated efficacy in the control of inflammation (118). In the clinic, peripheral blood mononuclear cells from healthy volunteers receiving tcVNS produced significantly less proinflammatory cytokines compared with cells from the nonstimulated control group (149). Not all of these “less invasive” VNS approaches are efficacious. In a randomized double-blinded study, transvenous vagal nerve stimulation was safe and well tolerated in healthy human volunteers. Activation of the neural stimulation catheter inserted into the jugular vein failed to protect subjects from increased cytokine production, body temperature, and heart rate elicited by intravenous injection of LPS (139). Moreover, VNS did not reduce cytokine production, or alter neutrophil phagocytosis in leukocytes challenged with LPS ex vivo. This randomized clinical trial used several methods to ensure correct electrode placement, and maximal stimulation without inducing pain or discomfort in subjects, resulting in a wide range of stimulation amplitudes. While the authors note that these are all within the range of voltages used for other preclinical and clinical studies, one cannot help but wonder if the variability of stimulation parameters contributed to the lack of efficacy. It is also conceivable that more than one neural pathway is being stimulated, depending on the electrode placement and stimulation parameters. In support of this contention, vagal nerve stimulation without prior surgical vagotomy or blocking of afferent signaling increases neural activity in the greater splanchnic nerve (199). This neural circuit has previously been shown to augment proinflammatory cytokine production (114). This randomized clinical trial is, however, one of the few to evaluate VNS in healthy human subjects.

VIII. COMPREHENSIVE MAPPING OF FUNCTIONAL CIRCUITS

Application of devices that stimulate neuroimmune reflex arcs to alleviate disease in patients are still in their infancy. At this point, there are few studies that have attempted to document the intrinsic variation in innervation of peripheral organs in outbred animals, let alone in humans afflicted with chronic inflammatory diseases. Of the few studies comparing the innervation between animal species, none has meticulously documented the extent of variation between individual subjects. In target organs such as lymph nodes, axons have been reported in close proximity to lymphocytes in mice, rats, cat (25), and rabbit (179). Unfortunately, these studies did not assess the intersubject variability of innervation or the origin of these neurons. These are important variables to account for in the successful design and implementation of therapies that seek to deliver signals to defined cell populations. Another critical variable that may impact clinical significance is the effect of inflammation on the innervation of the affected organ. Chronic inflammation can significantly alter innervation of effected organs, with immunopathologies such as RA and IBD associated with increased production of cytokines, including IL-17A (3, 175). Sympathetic neurons express IL-17A receptors and IL-17A can promote neurite outgrowth (54, 103). The consequences of this neurite outgrowth would be predicted to include increased sympathetic neural density. In support of this contention that inflammation may alter the innervation, increased sympathetic innervation has been reported in patients with chronic IBD (24, 143), although other studies have reported a decrease in sympathetic innervation (243). Studies from several preclinical colitis models report reduced sympathetic and cholinergic fibers in submucosal (160) and myenteric ganglia (210). Although increased TH immunoreactive fibers are visible in the mucosa, the sparse nature precluded quantitative analysis (210). Similar depletion of nerves, including sympathetic innervation, occurs in the synovium of joints in preclinical models (167) and patients with RA (264). Depletion or a complete absence of sympathetic axons in the spleen has also been observed in spleens of patients with septic shock (110). Together these findings support the need for studies that directly address how inflammation or other preexisting conditions influence the efficacy of neurostimulation devices. Another important consideration is whether inflammation affects the release of neurotransmitters without affecting neuroanatomy (123). These knowledge gaps highlight the continued need for pre-clinical and clinical studies to assess the effect of inflammation on the innervation of various organs. These changes in innervation could be critical factors in the success of a targeted electrical neural stimulation. Consider for instance that increased density could result in increased and unintended effects, while decreased innervation density could conceivably reduce the efficacy. Although the benefit of electroceutical devices is the ability to finely taper the therapy to the individual, increased or decreased neural signaling following inflammation would seem a basic question that should be addressed. Additionally, while stimulation of nerves comprised of multiple axons types, such as the vagus, are likely to yield the same result in different subjects, it is uncertain if next generation devices targeting specific fibers will encounter significant intersubject variability.

Better understanding of the functional and anatomical circuitry in humans will be challenging. Functional circuit tracing and an understanding of variance between subjects in and patients could be benefited with analysis of non-human primates. Technologies including optogenetics have been successfully used in the CNS of Rhesus macaques (242). These techniques could be readily adapted for the peripheral nervous system in an effort to map the neuroimmune connection.

IX. CONCLUSION

The immune and nervous systems are tightly integrated physiological systems that are finally being appreciated as such. While immunology has been regarded as separate or discrete from physiology, progress in this field has been transformed by transdisciplinary studies and research teams. This approach has already led to a resurgence in the field and the dawn of therapeutic approaches that may prove effective in the treatment of chronic inflammatory conditions. Development of technologies in neuroscience and seemingly unrelated fields will continue to drive exciting new opportunities for discovery in this field. The recent surge of findings in neuroimmune communication should not be equated to there being nothing left to discover. From the controversies over established reflex arcs, limited neuroanatomical data, continued discoveries of novel circuits, and the challenges in translating this research to humans, this is an exciting time to engage in studies of the communication between nervous and immune system.

GRANTS

This work was supported by National Institutes of Health Office of the Director SPARC Grant 1OT2OD023871–01 and Crohn's and Colitis Canada.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.