Interstitial cells of Cajal in diabetic gastroenteropathy

Abstract  Gastroenteropathy causes considerable morbidity in patients with diabetes mellitus and represents a major healthcare burden. Current treatments are largely symptomatic and frequently ineffective. Development of new therapeutic options is hampered by poor understanding of the underlying pathomechanisms. Experimental studies and sparse human data indicate that diabetic gastroenteropathy is multifactorial and involves not only parasympathetic and sympathetic autonomic nerves, but also enteric neurons, smooth muscle cells and interstitial cells of Cajal (ICC). ICC are mesenchymal cells that occur throughout the muscular coat of the gastrointestinal tract and provide functions critical for normal gastrointestinal motility including generation and propagation of electrical slow waves and mediation of bidirectional communication between the autonomic nervous system and smooth muscle cells. Through these functions, and in concert with other cell types of the gastrointestinal muscles, ICC support basic gastrointestinal functions such as digestion, absorption and waste removal. Loss or dysfunction of ICC in various dysmotilities and their animal models has been shown to lead to gastric dysrhythmias, gastroparesis, slow intestinal transit, impaired neuroeffector mechanisms and altered visceral afferent signalling that are considered hallmarks of diabetic gastroenteropathy. These findings and an increasing body of evidence indicating disruptions of ICC networks in diabetes suggest that the loss of ICC in this disorder is probably of functional significance and could even be a major pathogenetic factor. Future research should focus on the identification of the molecular and cellular mechanisms underlying ICC loss in diabetes and the translation of the experimental findings into treatments.

and may involve all aspect of gastrointestinal motility. 1,2,[4][5][6] Many investigators view these changes as consequences of irreversible autonomic neuropathy affecting primarily the vagus and sympathetic nerves. 2,3 However, they are likely multifactorial and also involve enteric neurons, mucosal endocrine cells, smooth muscle cells and interstitial cells of Cajal (ICC). The pathogenesis of gastroparesis and diabetic gastrointestinal autonomic neuropathy has been in the focus of recent reviews. 5,6 This study addresses the role of ICC loss and dysfunction in the pathophysiology of diabetic gastroenteropathy.

THE ICC PHENOTYPE
ICC are mesenchymal cells that have been described throughout the muscular coat of the gastrointestinal tract of all vertebrates studied to date. Although they only represent 5% of cells that make up the tunica muscularis, the functions they provide are critical for normal gastrointestinal motility. ICC develop independent of neural crest-derived enteric neurons or glia and mainly originate from Kit + mesenchymal mesodermal precursors. [7][8][9] In the chick, some ICC in the stomach and duodenum may also develop from multipotent, ventrally emigrating neural tube cells. 10 Whether the same occurs in mammals remains to be established. ICC have been classified according to their localization within the muscle layers (submuscular, intramuscular, myenteric and subserosal), their basic morphology (stellate vs bipolar) and primary function (pacemakers vs cells that mediate neuromuscular neurotransmission and mechanoreception). 8,11 There is some correspondence between morphology, localization and function 12 but the division of labour between the various morphological classes is not as clear-cut as previously believed. 13 ICC classes share several ultrastructural features such as abundant mitochondria, endoplasmic reticulum and intermediate filaments; close contacts with other ICC, smooth muscle cells and nerves; and lack of contractile proteins. 11 They also express the type III receptor tyrosine kinase Kit, the receptor for stem cell factor (SCF or Kit ligand; Kitl). The membrane-associated form of this cytokine is the most important developmental, growth and survival factor for ICC. 8,14 These characteristics identify various ICC classes as manifestations of the same, distinct cell type.

LOSS OF ICC IN DIABETES MELLITUS
Over the last couple of years several publications have demonstrated loss of ICC in diabetic gastroenteropa-thy. However, human studies using biopsy specimens may suffer from potential sampling errors due to inhomogeneities in ICC network densities 15 and difficulties in obtaining control tissues from anatomically corresponding regions. These potential problems can be avoided by using tissues from smaller animals, which can be studied in their entirety and compared to appropriate controls. Fortunately, well-characterized rodent models of human diabetes mellitus are readily available and have made it possible to take a more analytical approach to characterizing the involvement of ICC than would have been possible solely from human tissues.
In non-obese diabetic (NOD) mice, a well-established model of human type 1 diabetes mellitus, gastric emptying of solids is delayed after 6-8 weeks of untreated diabetes. 16 Gastroparesis has also been described in NOD mice kept on limited insulin treatment that was only sufficient to prevent severe hyperglycaemia (>500 mg dL )1 ) and ketosis but did not normalize blood glucose levels. 17 In two studies, 20 of 21 long-term (60 days) diabetic NOD mice had clearly reduced ICC networks by Kit immunofluorescence in their gastric corpus and antrum relative to the corresponding regions in their age-matched, non-diabetic littermates. 16,18 Both myenteric and intramuscular ICC were affected. The lesions were variable in size and locations but were more frequent distal to the midcorpus. ICC numbers were also decreased by electron microscopy. The remaining ICC only had minor abnormalities and, similarly to ICC after Kit blockade, 8 showed no signs of apoptosis or necrosis and there was no evidence of mononuclear infiltration. 16 Depletion was not detected in the fundus although the ICC lacked the typical close associations with enteric nerve terminals due to the accumulation of extracellular matrix. Fluorescent microscopy also revealed an invasion of myenteric ICC deep into the fundus, which normally lacks this class in mice. 16 ICC loss was accompanied by a significant decrease in Kit expression detected by oligonucleotide microarrays and quantitative RT-PCR. Significant decline in Kit mRNA only occurred after 60 days of untreated diabetes. 18 Others reported a decrease in Kit protein by Western blotting. 17 In the type 2 model db/db (leptin receptor mutant) mice, there was a modest decline in both myenteric and intramuscular ICC throughout the stomach, small intestine and colon. The authors found no evidence of apoptosis. 19 ICC loss was also detected by electron microscopy in the gastric antrum of 12week streptozotocin (STZ)-diabetic rats with significant gastric dysrhythmias. The remaining ICC were separated from other cells by wide extracellular spaces and had reduced gap junctions and organelles, swollen mitochondria, dilated endoplasmic reticula, vacuoles, myelin figures and broad perinuclear spaces. 20 Similar degenerative changes including swollen mitochondria, lamina bodies, partial depletion of cell bodies and processes, and loss of synapse-like connections with enteric nerves have also been described in intramuscular ICC of the gastric fundus and corpus of STZ rats along with a depletion of this ICC class in both the circular and longitudinal muscle layers. 21 Myenteric ICC were preserved in this model. ICC loss has also been documented in the stomachs of diabetic patients with gastroparesis. Forster et al. 22 examined gastric wall biopsies taken from the greater curvature of the gastric antrum. Profound loss of ICC was found in four of nine patients and modest depletion was noted in one. The damage was most evident in the circular muscle layer but the immunostaining technique used preferentially labelled intramuscular ICC in the controls so comparisons of myenteric ICC populations may not have been reliable. In a recent update, the authors reported that nine of 23 diabetic patients with refractory gastroparesis had no ICC in their antrum biopsies. 23 ICC loss also occurred in the corpus but was less frequent. Iwasaki et al. 24 examined ICC in antral tissues from 42 diabetic patients who had had gastrectomy for gastric cancer. Intramuscular ICC in the circular muscle layer, but not myenteric ICC, were significantly reduced in eight patients with severe diabetes but not in patients with milder forms of the disease. The authors also noted that close associations between ICC and nerves positive for neuronal nitric oxide (NO) synthase were reduced in patients with poorly controlled diabetes. Unfortunately, the reported changes in ICC density were not verified by electron microscopy in the aforementioned human studies, thus the possibility that some ICC with reduced or undetectable Kit immunoreactivity might have survived cannot be excluded.
There have been fewer reports of ICC loss in the small intestines and the colon. In a jejunal biopsy taken from a patient with 15 years, poorly controlled type 1 diabetes complicated by nephropathy, neuropathy, severe gastroparesis and abnormal small intestinal motility, He et al. 25 detected a marked reduction of Kit immunofluorescence associated with ICC in the myenteric region and the outer third of the circular muscle. ICC were completely missing from the inner two-thirds of the circular layer (Fig. 1). Nakahara et al. 26 studied colon tissues from seven colon cancer patients with type 2 diabetes. Four patients had constipation. ICC were significantly depleted in the diabetic group (to 40% of non-diabetic controls with or without cancer). However, the ICC loss did not correlate with constipation. Electron microscopy was not performed in either study. Intramuscular and myenteric ICC were also reduced in the small and large intestines of db/db mice. 19 In long-term diabetic NOD mice, constipation and accumulation of fecal pellets in the distal colon were noted while the proximal colon was filled with liquid feces. ICC were reduced in large foci across the entire thickness of both parts of the colon and these findings were also verified by electron microscopy. 27

ROLE OF ICC LOSS AND DYSFUNCTION IN THE PATHOGENESIS OF DIABETIC GASTROENTEROPATHY
A role for ICC in diabetic gastroenteropathy cannot be established solely on the basis of detecting their loss or dysfunction in the affected tissues. Similarly to several other gastrointestinal neuromuscular disorders with significant ICC involvement, 8,28,29 ICC lesions in diabetes tend to occur long after the onset of the problems and without further evidence it would be difficult to conclude that these defects were the cause of the dysmotilities. Moreover, gastrointestinal sensory and motor functions depend on cooperation between smooth muscle cells, ICC, the enteric and systemic autonomic nerves and even the mucosa 6,12,30 and it is difficult to relate complex pathophysiological changes at the organ level to specific alterations in various cell types. Therefore, establishing a role for ICC loss or dysfunction in any particular aspect of diabetic dysmotilities also requires the demonstration that ICC normally contribute to the organ-level function that is impaired in these tissues; and that loss of these cells occurring independently of diabetes results in dysfunctions also found in diabetes. However, it is important to note that even a demonstrated role for ICC does not exclude the potential involvement of other cell types in the observed pathophysiological changes. After a brief overview of basic functions of ICC, this section will discuss the impact of ICC depletion or dysfunction at the organ level and in the context of pathophysiological changes underlying manifestations of diabetic gastroenteropathy.

Elementary functions of ICC
ICC are an integral part of the gastrointestinal neuromuscular apparatus and support digestion, absorption and waste removal in close coordination with smooth muscle cells, which perform mechanical work, and the systemic autonomic and enteric nervous systems, which convey sensory information and generate motor patterns. 30 In phasic muscles of the gastrointestinal tract, ICC produce slow electrical oscillations termed slow waves. They conduct to smooth muscle cells and cause corresponding depolarizations, which trigger voltage-dependent Ca 2+ influx and resultant contractions. Slow waves provide a relatively monotonous ÔbackgroundÕ activity. However, the degree of their coupling to mechanically productive contractions varies greatly depending on the fasting or fed state of the individual under the influence of neural, hormonal and paracrine inputs. In vivo blockade of ICC and slow waves in experimental animals leads to paralytic ileus. 31 In mutant animals lacking pacemaker ICC and slow waves, motility and the survival of the animals depend on spontaneous action potentials generated by the smooth muscles, 30,32 which may emerge as a result of developmental compensation. This activity is relatively disorganized but can also drive spontaneous and distention-induced peristaltic waves, albeit less effectively than normal slow waves. 30 Subcellular mechanisms of electrical pacemaking may vary between species, organ or tissue regions and ICC classes. Details of these mechanisms have been reviewed elsewhere. 13,33 It is important to emphasize that the generation of unitary potentials, the elementary spontaneous depolarizations underlying slow waves, depends on mitochondrial oxidative metabolism and glucose utilization. 33 These unitary potentials are synchronized by a voltage-sensitive, dihydropyridine-insensitive Ca 2+ current to produce the plateau of the slow wave. 33 Activation of these channels, e.g. by depolarization of electrically coupled neighbour cells can also phase-advance slow waves and thus lead to active propagation of activity. Propagation involves tuning the intrinsic frequencies of the individual pacemaker cells to the frequency of the dominant (fastest) pacemakers. For directional propagation to occur, ICC in the propagation pathway must have lower intrinsic frequencies. Indeed, orad-to-aborad frequency gradients are demonstrable both in dissected muscles and isolated ICC and reestablishing cell-tocell connections in cell cultures leads to slow wave entrainment. 15 In most species, multipolar ICC forming networks in the myenteric region of phasic muscles represent the dominant pacemakers, although other ICC classes also contribute. 13,32,33 Unitary potentials occur even in elongated, intramuscular ICC populating non-phasic mucles (e.g. the murine fundus), although they do not synchronize to elicit slow waves. 33 In contrast, in the phasic stomach, morphologically similar intramuscular (and, possibly, septal) ICC do contribute to the generation of slow waves and provide a pathway for their rapid, circumferential propagation. 13 Intramuscular/septal ICC temporarily become dominant pacemakers under excitatory, cholinergic stimulation (mediated by postjunctional M3 receptors) and during passive stretch (an effect mediated by prostanoids acting via EP 3 receptors), when they can entrain slow waves generated by the primary, myenteric pacemaker ICC. 12 In the guinea pig, where myenteric ICC are missing from the most orad part of the phasic stomach, intramuscular ICC may normally be the dominant pacemakers. 13 Mediation of neuromuscular neurotransmission and mechanosensitive responses are functions of intramuscular ICC and related ICC in the deep muscular plexus region of the small intestines and these functions are not shared by other ICC classes. Intramuscular ICC mediate both cholinergic excitation and nitrergic inhibitory neuromuscular neurotransmission. These functions are supported by anatomical/morphological features and specialized gene expression. 12,34,35 ICC may transmit the effects of neural inputs to the smooth muscle cells by electrical coupling or paracrine mediators. 12 Cholinergic and nitrergic responses to electrical stimulation are reduced or missing in the stomachs of Kit and Kitl mutant mice (W/W v and Sl/Sl d respectively), which lack intramuscular ICC. 12,34 However, there have been reports of residual nitrergic neuromuscular neurotransmission in both W/W v mice and Ws/Ws rats (which also have reduced ICC populations) and both purinergic inhibition and non-cholinergic (peptidergic) excitation are rather preserved. 12,32,36 These observations may be explained by parallel innervation of the smooth muscle and/or by compensatory gene expression in the mutant animals. 12 Mediation of mechanosensitive responses are the least well understood functions of ICC. These functions include the aforementioned transduction of passive stretch into excitatory input to pacemaker ICC in the stomach possibly via prostaglandin release, 12 and mediation of effects of distention to local enteric 36 and vagal neuronal circuits. 37 The latter depends on a trophic effect of intramuscular ICC on the development and, perhaps, maintenance of vagal intramuscular arrays. 37 The mechanisms underlying mechanosensory functions of ICC are not clear but likely involve expression of mechanosensitive ion channels. 38

Oesophageal dysfunction
Dysphagia and heartburn are usually mentioned as symptoms of diabetic gastroenteropathy. 2,6 Oesophageal transit may be delayed due to reduced peristalsis and basal lower oesophageal sphincter tone is diminished. In W/W v mice, which lack intramuscular ICC in the stomach, the tone of the lower oesophageal sphincter and the entire proximal stomach is reduced 34,39,40 and there is evidence of impaired nitrergic inhibitory neuromuscular neurotransmission. 34 However, relaxation in response to swallow has been shown to occur. 39 In Ws/Ws rats, which also lack intramuscular ICC in the proximal stomach, the lower oesophageal sphincter was not hypotensive as in W/W v mice but, rather, exhibited higher basal tone and spontaneous contractile activity. 41 Ultrastructural damage and loss of intramuscular ICC in the oesophagus have been reported in achalasia 29 but not in diabetes. Therefore, a role for ICC in the latter remains uncertain.

Proximal gastric dysfunction
Dyspepsia and symptoms arising from rapid gastric emptying (particularly of liquids, e.g. in early stages of type 2 diabetes), such as abdominal pain, diarrhoea, dizziness and sweating after a meal, are typically attributed to impaired receptive relaxation and accom-modation of the fundus. 5 The reduced volume change occurs in the presence of a lower fasting tone. 42 These changes are usually interpreted as manifestations of impaired vagal input to the smooth muscle. In W/W v mice lacking intramuscular ICC, myogenic tone is similarly reduced in the proximal stomach and the gastric muscles are more compliant, 34,40 possibly as a result of reduced cholinergic neuromuscular neurotransmission. 34 Although these cells have also been found to mediate nitrergic inhibitory neuromuscular neurotransmission, 34 the proximal stomach of W/W v mice can still relax in response to swallow or liquid infusion, which prompted some investigators to conclude that intramuscular ICC are not required for these responses. 39,40 However, as the relaxation occurred against a backdrop of a reduced tone, both the maximum pressure changes and their rate of development were lower, 39,40 and abolishing nitrergic or purinegic neurotransmission had larger effects in the wild type than in the mutant mice. 40 Thus, a role of intramuscular ICC in these complex responses cannot be excluded. Depletion of intramuscular ICC and degenerative changes in the residual cells have been described in the fundus and corpus of STZ rats. 21 Although no ICC loss was found in the fundus of longterm diabetic NOD mice, the typical close associations between ICC and enteric nerve terminals were missing due to the accumulation of extracellular matrix and both excitatory and inhibitory postjunctional electrical responses were significantly reduced. 16 These findings support the idea that reduced tone and impaired nitrergic relaxation in diabetes may in part be due to ICC involvement. However, reduction of intramuscular ICC in diabetic patients has only been reported in the antrum 22,24 and it remains to be investigated if similar changes also occur in the fundus.
In some diabetic NOD mice, there was also an invasion of myenteric ICC deep into the fundus and rhythmic slow waves appeared. 16 Although acquisition of phasic contractile activity could potentially contribute to impaired reservoir function in this normally electrically quiescent and tonic organ, further studies are needed to establish the significance of this observation.
In patients without demonstrable changes in gastric motility, dyspeptic symptoms may arise from increased perception of gastric distention. 5 Changes in ICC may also contribute to visceral hypersensitivity. For example, intramuscular ICC and vagal intramuscular mechanoreceptors are missing from the fundus of W/W v and Sl/Sl d mice and these mice maintain normal food intake by consuming smaller meals more frequently. 37 These eating patterns resemble those that diabetic patients with gastroenteropathy can tolerate. 3 It is unclear what exactly underlies this peculiar eating behaviour in mice but altered visceral afferent signalling likely play a role. However, eating patterns of diabetic mice has not been studied.
On the whole, the experimental data tend to support a role for impaired proximal gastric functions and dyspepsia-like symptoms, but more studies are needed to identify the specific mechanisms involved.

Gastroparesis
Delayed emptying of both digestible and indigestible solids and nutrient liquids occurs commonly in diabetes. 4 Delayed emptying of solids is more common than slow liquid emptying. 3 The latter depends mainly on the fundic Ôpressure pumpÕ mechanism controlled by pyloric opening. Impaired pyloric function (pylorospasm) arising primarily from reduced nitrergic relaxation has been described both in patients and mouse models of diabetes. 5 ICC are greatly reduced in patients with infantile hypertrophic pyloric stenosis 28 and in the absence of intramuscular ICC, nitrergic relaxation of the pyloric sphincter is diminished. 34 However, no ICC loss has so far been reported in the pylorus in diabetes and the exact role of these cells in diabetesassociated pylorospasm remains unclear.
Gastric stasis of solids is caused primarily by antral hypomotility. 43 Reduced smooth muscle contractions may arise from myopathy, 5,18 loss or reduction of pacing by electrical slow waves, 4,44 or impaired electromechanical coupling, which may occur when slow wave frequency is abnormally high and the slow wave plateaus are too short to allow sufficient Ca 2+ entry into the smooth muscle cells. 45 Gastric dysrhythmias including bradygastrias (slower-than-normal rhythm), tachygastrias (faster-than-normal rhythm) and mixed dysrhythmias are commonly found in diabetic patients with meal-related symptoms. 4 Both postprandially and during phase III of the migrating interdigestive motor complex, the orderly orad-to-aborad propagation of peristaltic ÔringsÕ of contraction depends on a similar, organized spread of electrical slow waves. 13 Thus, if dysrhythmias develop at any point along the propagation pathway, the normal pattern of slow wave propagation and peristalsis will be adversely affected. Indeed, gastric myoelectric dysrhythmias have been found to predict delayed gastric emptying with an accuracy of up to 78%. 46 Abnormal slow wave activity has also been reported in STZ-diabetic rats and in Otsuka Long-Evans Tokushima Fatty (OLETF) rats with type 2-like diabetes. 47 In diabetic NOD mice focal losses of electrical slow waves have been shown, primarily in the antrum. 16 The role of ICC (both myenteric, intramuscular and septal ICC) in the generation and propagation of gastric electrical slow waves is well-established. 13,15,33,44 Focal or diffuse reduction of ICC has been reported in the distal stomach of mice and rats with type 1-or type 2-like diabetes [16][17][18][19][20][21] as well as in patients. 22,24 ICC loss in diabetic rodents was associated with electrical dysrhythmias 16,20 and gastroparesis. 16,17 Patients with diabetic gastroparesis and severe ICC depletion had more tachygastria, higher symptom scores and showed less improvement to electrical stimulation than patients with normal ICC. 22 ICC depletion that occurs independent of diabetes, e.g. in idiopathic gastroparesis or after in vitro blockade of Kit signalling or in W/W v mice, can also lead to a variety of gastric dysrhythmias including loss of slow waves, bradygastrias, tachybradyarrhythmias and antral tachygastrias, e.g. by interrupting normal propagation pathways and allowing the emergence of dysrhythmic, ectopic pacemakers. 15,22,44,45 These electrical abnormalities closely resemble those typically found in diabetes. 4 Thus, reductions in ICC most likely play a significant role in the pathogenesis of diabetic gastroparesis.
In diabetes, ICC dysfunction without ICC loss may also lead to abnormal electrical slow wave activity and its consequences. For example, acute hyperglycaemia has been shown to cause various dysrhythmias (particularly, tachygastrias and irregular rhythms) in normal volunteers, type I diabetic patients and rats with type 2-like diabetes. 47,48 These dysrhythmias may also lead to hypomotility, reduced phase III activity and delayed gastric emptying. 6 The effects of high glucose on pacemaking may be mediated by prostaglandins 48 via EP 3 prostanoid receptors 33 but a role for altered mitochondrial Ca 2+ uptake also cannot be excluded. However, gastric dysfunction in diabetes cannot be explained solely on the basis of acute elevations in blood glucose. 2

Small intestinal dysmotilities
The small intestine is frequently affected in diabetes and both delayed and rapid transit may occur. 6 In db/ db mice, irregular small intestinal contractile activity has been reported. 19 In a patient with severe gastroparesis and abnormal small intestinal motility, He et al. 25 detected a marked reduction in ICC throughout the entire thickness of the jejunum. ICC loss has also been described in patients with chronic idiopathic pseudo-obstruction 29 and paralytic ileus can be induced by depleting small intestinal ICC by blocking Kit signalling with neutralizing antibodies. 31 ICC loss induced by Kit blockade also inhibits intestinal electrical slow waves. 49 Small intestinal transit is delayed in W/W v mice lacking myenteric ICC and slow waves in the small intestines. 30 Thus, ICC loss probably contributes to the loss-of-function type changes observed in a subset of patients.

Colorectal dysmotilities
Colorectal dysfunction is common in diabetes. Of patients attending specialized diabetes clinics, up to 60% reported constipation, 22% had diarrhoea and 20% had fecal incontinence. 1 The pathophysiology underlying these problems is not well understood but impaired gastrocolic reflex, rapid or delayed transit, abnormal internal anal sphincter tone, as well as impaired rectal compliance and sensation may contribute. 6 In diabetic rodents, constipation was accompanied by reduced neuromuscular neurotransmission in the distal colon, 27,50 whereas a paradoxical increase in contractile and underlying spike complex activity was noted in the proximal colon. 27,51 The latter occurred in the absence of reduced inhibitory control and may have reflected functional compensation or a response to small intestinal bacterial overgrowth. 6 ICC were reduced in the colon of mice with both type 1-and type 2-like diabetes 19,27 as well as in type 2 diabetic patients. 26 However, the depletion of ICC did not correlate with constipation in the latter. Nevertheless, ICC loss occurring independent of diabetes has also been demonstrated in idiopathic slow transit constipation 29 and in Kit mutant Ws/Ws rats, the profound reduction in colonic ICC was associated with loss of slow waves and reduced nitrergic neuromuscular neurotransmission. 32 Thus, ICC loss may in part be responsible for loss-of-function type changes in colorectal motility although its role in the observed dysfunctions is less clear than in the stomach and the small intestines.
There has been no report of neuromuscular dysfunction or ICC loss in the anorectum of diabetic rodents. In W/W v mice, depletion of ICC from the internal anal sphincter region caused a reduced distention-induced relaxation and this abnormality arose from impaired afferent, rather than efferent, mechanisms. 36 It remains to be established whether ICC loss also occurs in the internal anal sphincter in diabetes or its models.

MECHANISMS OF ICC DEPLETION
In both patients and animal models, cell loss appears to be the main cause of ICC-related pathologies. ICC depletion could potentially arise from autoimmune attack, hyperglycaemia and associated oxidative damage, dystrophic changes consequent to reduced insulin and growth factor signalling, or combinations thereof. Lymphocytic infiltration of the myenteric region has been described in the oesophagus of both type 1 and type 2 diabetic patients 52 but inflammation does not appear to be common in diabetic gastroenteropathy. Circulating autoantibodies have been detected in both type 1 and type 2 diabetes even though the latter is regarded as a metabolic and not autoimmune disease. Loss of ICC caused by anti-Kit autoantibodies has been reported in a patient with paraneoplastic gastroparesis and intestinal pseudo-obstruction. 53 However, an autoimmune attack on ICC in diabetes has not been demonstrated.
Chronic complications of diabetes are generally attributed to recurring episodes of hyperglycaemia and resultant oxidative damage, increased formation of advanced glycation end-products, glucose-induced activation of protein kinase C and NFjB, and increased glucose flux through the aldose reductase pathway. All these changes may be the consequence of increased production of superoxide by the mitochondrial electron transfer chain when glucose metabolism is accelerated. 54 Increased oxidative stress has recently been reported in diabetic NOD mice with gastroparesis along with reduced expression of neuronal NO synthase and the carbon monoxide (CO)-producing enzyme heme oxygenase-1, 17 which are sources of factors potentially cytoprotective for ICC. 55 Consistent with its proposed role, heme oxygenase-1 was indeed upregulated in diabetic NOD mice that did not develop gastroparesis but the significance of reduced NO synthase expression is less clear. 17 Because independent control of insulin and glucose concentrations is difficult in chronic in vivo studies, we developed long-term organotypic cultures of murine gastric tunica muscularis tissues to study the relative significance of hyperglycaemia and reduced insulin and growth factor signalling in diabetic ICC loss. ICC and electrical slow waves in these cultured tissues can be maintained with normoglycaemic, unsupplemented, serum-free basal media for up to several weeks, indicating that the tunica muscularis has intrinsic reserves (e.g. SCF production) ICC can draw on for normal function. However, ICC networks eventually undergo significant depletion and slow waves disappear following a time course similar to the ICC loss in untreated diabetic NOD mice. Blockade of SCF/Kit signalling accelerates these changes indicating that the cultured tissuesÕ capacity to produce SCF determines and limits the survival of ICC and that maintenance of endogenous production of SCF beyond a certain time probably requires some serum-born factors. 44 Using such cultures we found that insulin and insulin-like growth factor-I (IGF-I) could be such factors because they completely prevented the loss of ICC and Kit expression (Fig. 2). 56 Slow wave activity was also rescued by these treatments. The effects of insulin and IGF-I were not enhanced by serum administration. Hyperglycaemia failed to accelerate the demise of ICC or negatively affect the rescue of ICC by insulin. Thus, diabetesassociated depletion of ICC seems unlikely to be caused by chronic hyperglycaemia per se. Rather, maintenance of ICC may require insulin, which is reduced or ineffective depending on the type of diabetes; and IGF-I, which is reduced in both type 1 and type 2 diabetes. 56 It is unclear why ICC in these cultures were not damaged by hyperglycaemia but it is remarkable that these cells contain an abundance of mitochondria and rely on oxidative metabolism for electrical pacemaking. 33 ICC also appear to possess efficient mechanisms for elimination of superoxide and other reactive oxygen species or counter their apoptotic effects. 35 However, protective mechanisms that are either intrinsic or extrinsic to ICC (e.g. CO produced by heme oxygenase-1) may themselves be negatively affected by diabetes. Loss of protection, in turn, may permit hyperglycaemia to exert effects on ICC not seen in normal tissues. It remains to be studied whether the decline in heme oxygenase-1 expression in diabetic mice 17 could have such a potentiating effect.
The pro-survival effects of insulin and IGF-I were probably indirect because we failed to detect expression of receptors for these hormones in mature ICC. The membrane-bound isoform of SCF, the most important developmental, growth and survival factor for ICC, appeared to mediate the effects of insulin and IGF-I on these cells as evidenced by the decline of its expression in diabetic stomachs and its rescue in vitro by insulin and IGF-I treatments that also prevented ICC loss. 18 Even though both neurons and smooth muscle cells produce SCF, we only found the latter to express this protein on their membrane, which explains why in the presence of smooth muscle cells, neurons are not required for the development or maintenance of ICC. 34 Indeed, the reduction in SCF mRNA in diabetic mouse stomachs mirrored the development of smooth muscle dystrophy and occurred in spite of the maintenance of enteric neurons. 18 The proposed interplay between insulin, IGF-I, SCF, hyperglycaemia and CO in diabetes-associated ICC loss is illustrated in Fig. 3. Figure 2 Long-term effects of hyperglycaemia, insulin, and insulin-like growth factor-I (IGF-I) on interstitial cells of Cajal (ICC) networks and electrical slow waves in organotypic cultures of murine gastric corpus and antrum tunica muscularis. (A) Confocal images of Kit immunofluorescence in freshly dissected controls (Ctrl) and in tissues cultured for 10 weeks with one of the following media: unsupplemented normoglycaemic (NG; 100 mg dL )1 ) or hyperglycaemic (HG; 1000 mg dL )1 ) media, normoglycaemic media containing insulin (Ins; 5 lg mL )1 ), hyperglycaemic media supplemented with insulin (InsHG), or normoglycaemic media containing IGF-I (IGF-I; 100 ng mL )1 ). Note the depletion of ICC in normoglycaemic and hyperglycaemic cultures and the complete prevention of ICC loss by insulin or IGF-I. High glucose did not attenuate the effect of insulin. (B) Electrical slow waves in organotypic cultures maintained as above for 12 weeks. Note the lack of slow waves and depolarization of resting membrane potentials in the unsupplemented normoglycaemic or hyperglycaemic cultures. Both changes were prevented by insulin or IGF-I. Copyright Ó 2005 American Diabetes Association from Diabetes Ò , Vol. 54, The finding that SCF may serve as a paracrine mediator of the actions of other growth factors raises the possibility that SCF/Kit signalling may represent the final common pathway for several pathological processes causing ICC loss or dysfunction. 18 SCF is ideally suited for this role because ICC mass in the tissues is regulated, to a great extent, by this pathway. Firstly, the development of ICC depends on SCF signalling via Kit. 49 This role is further evidenced by the significant quantitative reduction of ICC in mutant rodents, which have reduced (but not lost) Kit activity (e.g. W/W v mice and Ws/Ws rats) or can only produce soluble, secreted SCF but not its membrane-anchored isoforms, which are much more effective in stimulating Kit signalling in ICC (Sl/Sl d mice). 8,12 Secondly, SCF is also required for the survival of established ICC. 8,18,44,49 This role may not be exclusive as mouse strains differ in their sensitivity to in vivo Kit blockade. 31 These differences may reflect varying capacities of compensating mechanisms including signalling via other receptor tyrosine kinases 35 and factors such as NO and CO, which may be cytoprotective for ICC. 17,55 Different levels of resistance to in vivo Kit blockade may also reflect different rates at which ICC could regenerate. SCFÕs third function is that it also is a key proliferation factor for ICC. 28 It is not known whether ICC regeneration could be regulated by other factors as well. Recent in vitro results suggest that serotonin 57 and interleukin-9 58 can stimulate the proliferation of murine small intestinal ICC but it is uncertain whether they have the same effect in vivo. It also remains unclear whether ICC regeneration is primarily from mature cells or from progenitors and whether the size of the progenitor pool or the capacity of the precursors to differentiate into ICC is affected in various pathological conditions including diabetes. These are some of the least explored areas of ICC biology and should be in the focus of future research.

CONCLUSIONS
An increasing body of evidence indicates that ICC networks undergo significant depletion in diabetic gastroenteropathy. From basic physiological studies, it is now well-established that ICC perform key functions required for normal gastrointestinal motility including electrical pacemaking and mediation of neuroeffector and afferent signalling. Furthermore, loss or dysfunction of ICC in various dysmotilities and animal models has been shown to lead to organlevel dysfunctions that are considered hallmarks of diabetic gastroenteropathy. Thus, disruption of ICC networks may play a key role in gastrointestinal complications of diabetes, particularly in gastroparesis and slow intestinal transit. Future research should focus on identifying the molecular and cellular mechanisms involved and translate the findings into treatments. Figure 3 Hypothetical model for the pathogenesis of interstitial cells of Cajal (ICC) loss in diabetes. Maintenance of ICC requires circulating insulin and insulin-like growth factor-I (IGF-I), which is in part secreted by smooth muscle cells (SMC). The effects of these hormones are mediated by membrane-bound SCF (mb SCF) presented to ICC by SMC. Hyperglycaemia in normal tissues does not harm ICC due to the protective effects of carbon monoxide (CO) produced by the stress-induced enzyme heme oxygenase-1 (HO-1) and of some unidentified factors and mechanisms produced or possessed by ICC (???). Diabetes independent of hyperglycaemia reduces insulin and IGF-I signalling and SCF expression and may inhibit HO-1 and other protective mechanisms allowing the expression of the cytotoxic actions of hyperglycaemia. ICC loss is the combined consequence of a reduction of survival factors and increase of pro-death factors. Arrows indicate stimulation, blocked lines, inhibition. Dotted blocked line indicates that a deleterious effect of hyperglycaemia on ICC is not demonstrable in normal tissues.