Enteroendocrine and Neuronal Mechanisms in Pathophysiology of Acute Infectious Diarrhea

Introduction

Acute infectious diarrhea is generally considered to be a manifestation of enterocyte secretion. The best example is diarrhea due to Vibrio cholerae. Thus, cholera toxin (CT) consists of a toxic active A subunit (CT_A), and a B subunit pentamer (CT_B), which binds the toxin to the enterocyte brush border membrane via a ganglioside (GM₁) receptor. The bound toxin is internalized into the intestinal epithelial cell, where the A subunit is cleaved into the A1 peptide which ADP-ribosylates G_s, the stimulatory subunit of heterotrimeric G protein, leading to activation of adenylate cyclase. The resultant increase in intracytoplasmic cAMP concentration leads to increased chloride secretion by crypt cells and reduced absorption of sodium and chloride ions by villous cells; these actions result in diarrhea [1].

However, the “doctrine” of diarrheal disease through enterocyte secretion has been challenged. Thus, Lucas [2] noted that, for almost 40 years, one of the principal causes of diarrheal disease has been considered fluid secretion from the enterocytes and colonocytes. This enterocyte secretion hypothesis rapidly displaced all other alternatives, such as vasodilatation coupled with enhanced paracellular permeability. Indeed, Ussing chamber measurements of short circuit current (Isc), potential difference, flux measurements, or measurements of elevated second messenger concentrations that are linked to chloride ion secretion in vitro were supportive of the enterocyte secretion doctrine. However, there were several observations that were not supportive of enterocyte secretion as the sole mechanism of secretion in acute infectious diarrhea. For example, in contrast to its effects in vitro in the Ussing chamber, the heat-stable enterotoxin of E. coli (STa) was not secretory when fluid flux was measured by volume recovery [3]. In addition, knock-out studies fail to confirm the enterocyte secretion hypothesis, since deletion of the essential Na–K–Cl (NKCC1) co-transporter does not prevent the action of STa [4].

These observations suggest that alternative mechanisms must be considered to explain the mechanism of acute diarrhea induced by infectious organisms.

Intestinal Secretion as Part of an Enteroendocrine-Neural Reflex in Cholera

A hypothetical involvement of enteroendocrine–neural reflexes as the diarrheagenic “messenger” in cholera was proposed by Osaka et al. [5]. In their experiments, CT was introduced into the duodenum of young rabbits; this caused severe degranulation of the enteroendocrine cells as revealed by electron microscopy: basal granules were swollen up and opened to the basal and lateral cell surface. Osaka et al. [5] proposed the hypothesis that CT stimulates the apical receptor of the enteroendocrine cells, and in response to the stimulus, amine (such as serotonin or 5-HT) and peptide (possibly motilin) products of the cells were released and mediated the diarrheagenic action of cholera.

There was no controversy regarding the first step: the localization of CT in the mucosa was shown by immunohistochemistry in guinea pig jejunum [6]. Subsequent studies in animals and humans demonstrated release of 5-HT from the enteroendocrine cells and intestinal secretion in cats and humans.

Nilsson et al. demonstrated a relationship of relative 5-HT fluorescence in enteroendocrine cells and net fluid transport across the intestinal epithelium induced by CT in cats [7]; these effects were not replicated with heat-inactivated CT. Subsequently, the relationship of CT, 5-HT and intestinal secretion in human jejunum was also demonstrated in vivo [8]. The role of enteroendocrine cells could conceivably be paracrine or neurocrine.

Subsequent studies, so simple in design but so convincing and important, explored the role of neural elements through in vivo studies conducted in animal denervated jejunum. Lundgren et al. studied the effects of lidocaine (exploring nerve activity as it blocks sodium currents that are essential for generating action potentials) or the sodium channel blocker tetrodotoxin (TTX) and showed that CT-induced secretion in cat or rat denervated jejunum involved a neural mechanism [9, 10]. In addition to the known release of 5-HT from the cat small intestine exposed to CT, these authors also demonstrated release of vasoactive intestinal peptide (VIP) [11] in the cat small intestine.

The next major step was the proof that the VIP and other transmitters were originating from submucosal secretomotor neurons. Thus, for example, CT was shown to act in the mucosa and to induce specific and sustained hyperexcitability of secretomotor neuron in guinea pig jejunum [6]. In the same species, CT increases the number and the duration of action potentials fired during prolonged depolarizations in VIP and neuropeptide Y (NPY) neurons [6]. Overall, it became clear that effects of CT are neurally-mediated: TTX and hexamethonium (nicotinic acetylcholine receptor antagonist) blocked effects of CT in VIP and NPY neurons [6], and the neurally-mediated actions depend on 5-HT₃ and the neurokinin 1 (NK₁) receptors [6] in addition to the nicotinic ACh receptor.

Hubel [12] summarized in a review the secretory reflex circuit, acknowledging that much of the model was based on the body of work of Cassuto and colleagues, which was predominantly on cat small intestine [13]. In this model (Fig. 1), epithelial sensory cells (the “antecedent” to the tasting role of the enteroendocrine cells) liberate 5-HT or other chemical transmitters in response to intraluminal stimuli and stimulate afferent neurons. Through nicotinically-mediated reflexes, a submucosal neuron secretes a neuro-transmitter (e.g., acetylcholine) that influences ion transport in the enterocyte in the villus or crypt. In this model, extrinsic nerves, such as adrenergic nerves can also affect the enterocyte and enteroendocrine cells though the precise site of action was not clearly understood.

Enteroendocrine-Neural Reflex with Other Bacterial Toxins

E. coli STa activates guanylate cyclase C (GCC) receptors to induce chloride secretion. The receptors are confined to the luminal surface (microvillus border) of the intestinal epithelium throughout the length of both the small intestine and colon of several mammalian species examined [14]. In GCC receptor-deficient mice, baseline net bicarbonate absorption is comparable with that in normal mice. However, whereas in normal mouse proximal colon, mucosal 50 nM E. coli STa increased Isc, in receptor-deficient mice, neither mucosal nor serosal 500 nM STa affected electrolyte transport in proximal or distal colon of mice [15]. There is little secretion of 5-HT from the pig intestinal mucosa in response to STa [16].

C. difficile toxin B activates human VIP submucosal neurons [17], and this increased expression is partly mediated by the cytokine, IL-1β.

Lundgren and colleagues summarized the literature demonstrating the effects of neuropharmacological agents tetrodotoxin, lidocaine and hexamethonium on net fluid secretion induced by various secretagogues including rotavirus in newborn mouse intestine [18], and parasites such as Nippostrongylus brasiliensis in inflamed rat jejunum in vivo [19, 20].

The Role of Goblet Cells

The “rice water” stool of cholera is an expression of the induction of mucin secretion. The release of mucin from goblet cells may be regulated by innate and adaptive immune responses; thus, the most potent effects are induced by T cell cytokines: interleukin-4 (IL-4) and IL-13 produced by T helper 2 (T_H2) cells in response to parasitic infections. Interferon-γ (IFNγ) and IL-17, which are classically produced by T_H1 cells and T_H17 cells in response to intracellular and extracellular pathogens, also induce mucin production by goblet cells [21].

CT induces mucin secretion from goblet cells. As in the enterocyte, where CT induces cAMP production, Epple et al. [22] showed that CT caused a 116-fold increase of intracellular cAMP and stimulated the secretion of both preformed and newly synthesized mucin in a human goblet cell line HT-29/B6. These actions involved de novo synthesis of mucin molecules and microtubule-mediated secretion. This contrasts with carbachol, a cholinergic agonist, which only triggered the release of preformed mucin immediately after addition of this drug.

When CT was added to cloned human goblet cell monolayers (without submucosal neurons), it did not induce mucin secretion in contrast to effects of the calcium ionophore, A23187. This suggests that mucin secretion by CT involves submucosal nerves [23].

Guanylin is produced in the rat small intestinal and colonic goblet cells [24] and is structurally related to the STa. Given the production by goblet cells of the endogenous ligands of GCC receptor (guanylin and uroguanylin) on enterocytes that result in secretion of chloride ions, an intriguing question is: does goblet cell mucin secretion result in secretion of guanylin or uroguanylin? Guanylin-like immunoreactivity is present in the intracellular compartment occupied by the secretory granules in goblet cells. Li et al. [25] suggested that guanylin is packaged into these granules along with mucin in the rat intestine, and is released into the lumen when goblet cells are stimulated to secrete. The guanylin then has access to the GCC receptor on the luminal domain of the enterocyte. This results in activation of intracellular mediators including guanylate cyclase and cGMP, which ultimately influence chloride ion fluxes through the cystic fibrosis transmembrane regulator-chloride channel (CFTR, Fig. 2).

Effects on Bile Acid Absorption During Acute Infectious Diarrhea

Acute infectious diarrhea due to shigellosis in humans is associated with increased fecal bile acid excretion. The increased total bile acid excretion in shigellosis is associated with a greater proportion of cholic acid in stool. The fact that stool content of this primary bile acid is increased relative to the proportion of secondary bile acids suggests that bacterial dehydroxylation in the colon is deficient during the acute diarrheal stage [26].

Fecal bile acid and neutral sterol patterns were also studied by the same group in eight healthy adults challenged with a Vibrio cholerae strain. In contrast to the results with shigellosis, bile acid handling appears not to be affected in cholera: first, bacterial 7 α-dehydroxylation of primary bile acids, cholic and chenodeoxycholic acid, was not altered despite the greatly increased fecal wet weight; and second, total concentrations of both bile acids and cholesterol in mg/g of fecal wet weight were decreased as a result of dilution due to diarrhea. However, total bile acid and neutral steroid excretions in mg/kg/day in subjects do not appear to be different with or without diarrhea [27]. Results obtained with cholera are similar to those observed in experimentally-induced travelers’ diarrhea associated with toxigenic E. coli [27].

When bile acids are malabsorbed or cause intestinal secretion, there is evidence of the involvement of serotonin. Thus, it has been shown that sodium deoxycholate (4 mM) caused a release of 5-HT into the rat intestinal lumen, and this release was inhibited by calcium channel blockade; in addition, granisetron, a 5-HT₃ receptor blocker, partly inhibited the fluid secretion caused by bile salts [28].

The Role of Myenteric Neurons in Acute Infectious Diarrhea

There is evidence that the myenteric plexus is also involved during CT-induced net fluid secretion in rat small intestine. The role of the myenteric plexus was considered critical since treatment of the serosal surface of the intestinal loop with benzalkonium chloride (BAC), which selectively ablated the myenteric plexus, eliminated the ability of CT to induce rat intestinal secretion [29]. While the authors suggested that all afferent fibers in the intramural secretory reflex activated by CT are probably conveyed via the myenteric plexus, which functions as the integrating center in the enteric nervous system, it is conceivable that the topically active BAC may have penetrated beyond the myenteric plexus and also affected submucosal neurons. If the penetration of the BAC reached submucosal neurons, the secretory reflex at the submucosal level could be abolished. In fact, these careful semi-quantitative studies of Jodal et al. [29] showed that the calcitonin gene-related peptide (CGRP) neurons in the submucosa were depleted by BAC treatment relative to controls, even though the VIP, galanin and substance P submucosal neurons were not apparently altered. CGRP is an important transmitter in the control of gut submucosal arterioles [30], and it is conceivable that depletion of CGRP in these preparations may have influenced the vasodilatation that is ultimately involved in the secretion in response to CT.

Mathias et al. [31] used a different approach to assess the role of motility in CT-induced intestinal secretion. They studied repetitive action potentials or migrating action potential complexes (MAPCs) that may constitute a specific motor repertoire in response to these toxins or the non-specific effects of intestinal secretion or diarrhea. Mathias et al. compared the number of MAPCs of the intestinal musculature per hour for 10 loops of rabbit intestine infected with the live Vibrio culture, the whole cell lysate, or the purified enterotoxin. They showed that CT elicits MAPCs recorded by electrodes placed on the serosal surface distal to the segment exposed to CT. In addition, Nocerino et al. [32] showed that, in the rat, MAPCs can migrate past the ileocecal valve and into the colon. The MAPCs reflect contractile waves of the circular layer moving in an anal direction. They are eliminated by the same types of drugs as those that influence the local secretory reflex elicited by CT, i.e., by local anesthetics, nicotinic receptor agonists and atropine.

Stimulation of repetitive bursts of action potentials in the rabbit small intestine was also observed with E. coli STa application [33].

Figure 3 shows a schematic figure of the model proposed for the secretory nervous reflex activated by cholera toxin and incorporating the influence of the myenteric cholinergic neurons [28].

Increased Intestinal Permeability in Diarrhea Due to Invasive Organisms

The enteric nervous system, more specifically the enteric glial cells, has recently been demonstrated to significantly reduce the mucosal lesions and inflammatory response caused by Shigella flexneri in an ex vivo human colonic mucosa model [34]. These protective effects were mediated by S-nitrosogluthatione, the nitrosylated form of reduced glutathione, which is secreted by the glial cells. The involvement of the enteric glia in protection against intestinal secretion could be potentially relevant to other infectious diarrheas.

Infectious diarrhea caused by invasive pathogens results in increased intestinal permeability as well as inflammation. This was originally demonstrated by comparing the permeability to ⁵¹Cr-EDTA in controls and patients with infectious diarrhea [35]. Thus, acute gastroenteritis caused by infection with Yersinia enterocolitica is associated with altered tight junction proteins and induction of cell necrosis, both resulting in a decrease in transepithelial resistance in colonic HT-29/B6 cell monolayers [36]. This fall in trans-epithelial resistance was manifested by an increased permeability to mannitol and fluorescein in a human colonic cancer cell line. In amebic dysentery caused by Entamoeba histolytica, the constitutive production and secretion of prostaglandin E₂ by the parasite alters ion permeability of the paracellular tight junctions of T84 cells grown to 100% confluence on polyethylene membrane inserts, resulting in increased sodium ion permeability towards the lumen in addition to the luminal chloride secretion by activation of CFTR [37]. Thus, the authors were able to demonstrate that E. histolytica is also capable of initiating disease in a contact-independent manner, in contrast to its direct effects on colon epithelium causing hemorrhage and inflammation that also contributes to diarrhea.

Noninvasive infectious diarrhea due to cholera is also associated with increased mucosal permeability in the small intestine. Fasano et al. showed that V. cholerae secretes an additional toxin (different from the CT) that affects the structure of the zonula occludens in the intercellular tight junctions, causing secretion in rabbit intestine in vitro [38].

Novel Treatment Strategies Targeting Enteroendocrine and Neural Mechanisms

Given the involvement of enteroendocrine and at least submucosal neurons in the development of acute infectious diarrhea, a number of experimental approaches suggest that new treatment strategies may be developed to enhance the traditional sugar-electrolyte-water approaches to treat acute dehydration.

Somatostatin exerts antisecretory effects by suppressing the firing of VIP secretomotor neurons rather than via a direct action on mucosal enterocytes in guinea-pig jejunum [39]. Somatostatin also decreases chloride secretion in response to 5-HT and a nicotinic agonist, dimethylphenylpiperazinium, a response that is sensitive to tetrodotoxin and reflects actions on both cholinergic and non-cholinergic secretomotor neurons in guinea-pig jejunum and rat colon [39, 40].

In rat distal colonic mucosa/submucosa preparations, pre-treatment with 5-HT₃ receptor antagonists enhanced 5-HT-induced increases in Isc; in mucosa-only preparations without retained neural elements, pre-treatment with 5-HT₃ receptor antagonist inhibited 5-HT-induced increase in Isc [41]. The overall effect of a 5-HT₃ receptor antagonist in the intact organ would be predicted to be deleterious in acute infectious diarrhea. Activation of 5-HT₃ receptors does not lead to secretion in human intestine, but it does in many animal models. For example, Michel et al. [42] showed that 2-methyl-5-HT (a moderately selective full agonist at the 5-HT₃ receptor) did not evoke chloride secretion in the human intestine, even though specific 5-HT_3A and 5-HT_3B receptor subunit immunoreactivity, as well as 5-HT_3A and 5-HT_3B receptor–specific messenger RNA were detected in the human intestinal tissue samples. On the other hand, 2-methyl-5-HT induced secretion in the guinea-pig intestine [42]. Others showed that mucosal stroking of muscle-stripped human jejunum mounted in modified Ussing chambers resulted in 5-HT release and chloride secretion mediated through a neural reflex that involved 5-HT₄ receptors [43]. In human jejunal perfusion studies, Bearcroft et al. showed a benefit of alosetron (a5-HT₃ receptor antagonist) on basal fluid and sodium secretion, but did not reverse secretion in response to CT in human jejunum [44]. Overall, therefore, 5-HT₃ receptors activate human submucous neurons, but these neurons do not appear to play a significant role in reflexes controlling secretion. The lack of functional evidence for 5-HT₃ receptors in human intestine argues against a potential therapeutic role of 5-HT₃ antagonists in intestinal secretory states.

Acetorphan, an orally administered inhibitor of enkephalinase in the wall of the digestive tract (Fig. 4), prevents inactivation of endogenous opioid peptides released by submucosal and myenteric neurons. Acetorphan completely prevented water and electrolyte secretion induced by CT in human jejunum [45].

Another experimental approach involves the inhibition of release of 5-HT from enteroendocrine cells. Kojima et al. [46] demonstrated that a lipophilic nitric oxide (NO)-releasing compound 5-amino-3-(3, 4-dichlorophenyl) 1,2,3,4-oxatriazolium (GEA3162) can suppress 5-HT release from colonic mucosa through a mechanism that does not involve cyclic GMP system or peroxynitrite generation. The effects of this agent on 5-HT secretion in response to infection have not been reported to date.

Other neurotransmitters that may be the target for future therapy are VIP and NPY. In perfused rat jejunum, PG 97-269 abolished the effects of VIP on fluid and electrolyte transport and attenuated cholera toxin and Escherichia coli heat labile toxin-induced net fluid and electrolyte secretion [47].

Y2 receptors are involved in mediating rat intestinal secretion in response to VIP [48]. Conversely, NPY Y2 receptor activation (e.g., with PYY_3–36) inhibited colonic motor function in mice, thus reducing intestinal fluid secretion and slowing colonic transit [49]; however, the efficacy in intestinal infections is unclear.

A novel approach to inhibition of secretion involves ligands that inhibit binding of the CT to the enterocytes’ GM₁ glycoside receptor [50] (Fig. 4). Such an inhibition of toxin binding is a very attractive approach to the prevention of diarrhea, because it can be applied either prophylactically or after the infection has occurred [51]. However, these agents are at a preliminary stage of development.

In view of the pivotal role of CFTR in chloride secretion from the enterocyte, another approach in the treatment of cholera targets the CFTR chloride channel. The thiazolidinone CFTR inhibitor (CFTR_inh-172) reduced CT-induced fluid secretion in the mouse small intestine when injected intraperitoneally [52]. Despite being homologous to the sulfonylureas, such as glibenclamide that are used in the treatment of type II diabetes mellitus, CFTR_inh-172 did not inhibit other ion channels (including the ATP-sensitive K⁺ channels that are blocked by glibenclamide). Another class of compounds, called glycine hydrazide compounds (GlyH), blocks the CFTR: after intraluminal administration, GlyH-101 reduced intestinal fluid secretion caused by CT by approximately 80% [53]. Moreover, GlyH-101 is highly water-soluble, with rapid onset of action, and it binds close to the external pore entrance of the CFTR ion channel, in contrast to CFTR_inh-172, which is less water-soluble and binds to the nucleotide binding domain at the cytoplasmic surface of the CFTR ion channel.

Studies of pharmacological inhibition of chenodeoxycholate-induced fluid and mucus secretion and mucosal injury in the rabbit colon suggested that atropine markedly reduced and carbachol potentiated the fluid secretion, mucus output, and mucosal damage observed during bile acid perfusion. In contrast, pre-treatment of the colonic mucosa with lignocaine and parenteral administration of methysergide (5-HT antagonist) and somatostatin produced a modest reduction in the fluid secretory response without apparent effects on mucus output or mucosal damage. These data suggest that these effects were independent of the actions on mucosal protective factors, specifically mucus [54].

Binding of bile acids with resins such as cholestyramine has been tested in pediatric infectious diarrhea [55, 56]. Poor patient compliance precludes its widespread use, but tablet alternatives such as colesevelam (which is effective in diarrhea-predominant irritable bowel syndrome [57]) may be tested in the future.

Approaches to reduce intestinal permeability include the experimental drug 2,4,6-triaminopyrimidine [58]. The potential for agents to reduce intestinal permeability in infectious secretory diarrheas requires further study.

Summary and a Look to the Future

Enteroendocrine and neuronal mechanisms participate in the development of acute infectious diarrhea due to bacterial toxins (e.g., cholera, E. coli heat-stable enterotoxin, C. difficile) and rotavirus. These agents are all associated with secretion of transmitters from enteroendocrine cells (e.g., 5-HT) and activation of afferent neurons stimulating submucosal secretomotor neurons that secrete ACh (nicotinic) and VIP. Specific organisms also modify other mechanisms that may contribute to development of acute diarrhea, including mucin secretion, increased permeability and inhibition of bile acid absorption. New therapies targeting neural and transmitter mediation of acute infectious diarrhea could usher in novel approaches in the treatment of acute diarrhea.