INTRODUCTION

TOP ABSTRACT INTRODUCTION ICC ISEMF FUNCTIONS OF ISEMF CONCLUSIONS REFERENCES

FUNDAMENTAL BIOLOGICAL PROCESSES such as cell motility, proliferation, differentiation, apoptosis, morphogenesis, tissue repair, inflammation, and the immune response are initiated, maintained, and terminated by local interactions between cells. These interactions are brought about by contact of cells with each other or with the extracellular matrix (ECM) or through the elaboration of and response to soluble mediators. Although myofibroblasts were identified morphologically a century ago, it is now recognized that they constitute a family of paracrine cells that play an important role in the regulation of these fundamental processes [see part I of this review, which appeared in the July issue (163)] (181, 207). This article focuses on intestinal subepithelial myofibroblasts (ISEMF), the most recently recognized myofibroblasts in the intestine, with a brief review of the interstitial cells of Cajal (ICC), the other intestinal myofibroblasts.

The ICC are located in the submucosa and muscularis propria in association with the smooth muscle layers of the gut. The ISEMF are located in the lamina propria under the epithelial cells. It is not known if these two myofibroblasts are derived from a common precursor (stem) cell. Both exist as a syncytium, but it has not been determined whether the ICC network is physically connected to the ISEMF network. The ICC was discovered over 100 years ago and 50 years ago was shown convincingly not to be a specialized neuron but a fibroblast-like cell (176). Twenty years ago these cells were proposed to be the electrical pacemakers that control motility of the gastrointestinal tract, and recently a large body of evidence affirms this concept (176). Over 30 years ago Pascal, Kaye, and Lane (148) described a sheath of lamina propria fibroblasts in tight apposition to the crypts in the intestine. This structure was confirmed by others and was found also in other gastrointestinal tissue, such as the gallbladder and stomach (70, 95, 100, 101, 110, 111, 127, 133, 178). This fibroblastic sheath or pericryptal cells (fibroblasts), as they were called then (178), are now known to be a syncytium of cells that extend throughout the lamina propria (95).

	ICC

TOP ABSTRACT INTRODUCTION ICC ISEMF FUNCTIONS OF ISEMF CONCLUSIONS REFERENCES

Morphologically, the ICC have the classic appearance of myofibroblasts and exist in several separate locations within the gastrointestinal tract (84, 176). Classically, they are located in the intermuscular space between the circular and longitudinal layers of the muscularis propria in the stomach, small bowel, and colon. However, they also reside, with slight differences in morphology, along the submucosal surface of the circular muscle bundles of the colon, within the deep muscular plexus region of the small intestine, and intramuscularly in the esophagus, stomach, and colon (176, 199). These cells have a stellate appearance and are connected in a syncytium via gap junctions (31, 110). They have all the morphological characteristics of myofibroblasts [see part I of this review (163) as well as Ref. 205]. The biology of ICC has been reviewed excellently and in detail by Sanders (176), a review with many original references not listed here.

Antibodies against the receptor c-kit localize ICC in the mouse embryo (220) and suggest that the ICC originate from the mesenchyme of the gut and not from neural crest cells, the precursors of intestinal neurons (10, 201, 229). However, the true origin of these cells is still uncertain. The embryonal cells responsible for the synthesis and secretion of stem cell factor (SCF), the ligand that promotes the growth and differentiation of c-kit-positive ICC, have also not been clearly identified (209).

Considerable experimental evidence suggests three major functions of the ICC: 1) they are pacemakers for gastrointestinal smooth muscle motility, 2) they facilitate active propagation of electrical events, and 3) they modulate neurotransmission (176). As myofibroblasts, it is likely, however, that they have other functions [see part I of this review (163)], such as immune modulation, growth, repair, and fibrosis. These additional functions should be investigated in ICC.

The mammalian intestine demonstrates a wave of electrical activity with a characteristic frequency somewhere between 6 and 12 cycles per minute, depending on the segment studied (85, 176). These slow waves are accompanied by corresponding peristaltic contractions during the intercibal periods, and this activity is thought to serve a housekeeper function of clearing the intestinal lumen between meals. The pattern is disrupted by eating. Both freshly isolated and cultured ICC demonstrate oscillations of membrane potential due to varying Ca²⁺ conductivity, i.e., activation or closure of voltage-dependent Ca²⁺ channels (176). This property constitutes the primary evidence for assigning the pacemaker role to the interstitial cells (see Ref. 176 for original references). The lack of propagation of the slow wave in the bulk of smooth muscle, plus the fact that removing a thin strip of tissue that contains the syncytium of ICC also destroys the intestinal slow-wave activity, give strong credence to the idea that the ICC syncytium is the origin of the slow wave. Furthermore, chemical lesions of the ICC also inhibit both the pacemaker potentials and slow-wave propagation.

Evidence that ICC are involved in neurotransmission is both anatomic (their intermediate position between neural varicosities and the smooth muscle cells) and physiological. ICC are responsive to a host of enteric neurotransmitters, including acetylcholine, nitric oxide, vasoactive intestinal peptide, ATP, and substance P (176). Nitric oxide and, recently, carbon monoxide have been proposed as inhibitory neurotransmitters generated by the ICC and other myofibroblasts (8, 57, 131, 136).

Additional evidence for a pacemaker role of ICC comes from experiments using antibodies directed against c-kit protein (85, 200). When neutralizing c-kit (ACKII) antibodies are injected in neonatal animals for several days following birth, intestinal c-kit immunoreactivity disappears and intestinal smooth muscle activity becomes abnormal (85, 144, 200). Furthermore, the electrical rhythmicity of the embryo and neonate is absent until c-kit-positive cells appear (201).

Mutants of c-kit expression clarify the function of ICC (85, 90, 176). The white-spotting (W) locus in mice and a hypopigmentation disorder in humans known as piebaldism are allelic with c-kit. A number of spontaneous mutations of the W locus are available, which either completely or partially block c-kit protein expression. Mutations that completely block expression of c-kit are fatal; homozygotes (W/W) die in utero. However, there are point mutations of the locus, such as W^v, that do not completely abolish the receptor expression. W^v/W^v homozygotes or compound W/W^v heterozygotes lose electrical slow-wave rhythmicity and have abnormalities of intestinal smooth muscle function. Humans born with piebaldism often develop congenital megacolon.

Mutants of SCF, so-called Steel mutants, also clarify the role of the ICC in gut motility (176). The classic Steel (Sl) mutants represent a complete deletion of the genomic region that codes for SCF, and homozygotes (Sl/Sl) do not survive because of development of severe anemia. However, nonlethal mutants such as Sl-dickie (Sl^d) and compound heterozygotes such as Sl/Sl^d retain some c-kit signaling activity (16). The ICC of these animals are histologically abnormal, and their electrical and contractile patterns of the intestinal smooth muscle are abnormal as well (217).

The recognized human diseases associated with abnormalities of the ICC are diseases of gastrointestinal motility and inflammation (84): congenital megacolon of piebaldism and Hirschsprung's disease (176, 209), infantile hypertrophic pyloric stenosis (208), intestinal pseudoobstruction (18, 84, 89), possibly achalasia (218), and ulcerative colitis (173). Hirschsprung's is a polygenic disease that results in enteric nervous system aganglionosis and disturbed motility. Mutant genes associated with Hirschsprung's disease include 1) the RET protooncogene, a tyrosine kinase receptor, which prevents ganglion cell migration from the neural crest (151); 2) the endothelin-3 (ET-3) gene (172); and 3) the gene for the ET-3 receptor (172). The ICC in Hirschsprung's disease are present, but morphologically abnormal, indicating that disturbed ICC function accompanies aganglionosis in this polygenic disease (176, 209). Perhaps the presence of ICC is required for the migration and full development of ganglionic neurons (115) (or vice versa) or, alternatively, some defect in the embryological mesenchyme affects both the migration of neural crest cells and the differentiation of the ICC.

Infantile hypertrophic pyloric stenosis is a relatively common disease of newborns, characterized by hypertrophy of the muscular region of the pyloric sphincter (208). Various pathological disturbances have been demonstrated, including abnormalities of the enteric nervous system and, more recently, the reduction or lack of the neuronal isoform of nitric oxide synthase. c-Kit immunoreactive ICC have been shown to be absent in the longitudinal muscle layer and in most of the hypertrophic muscle of all 26 patients investigated (208). The exact role of the ICC in the muscular abnormality of this disease is unclear.

Recently, abnormalities in nitric oxide-dependent inhibitory neurotransmission in the lower esophageal sphincter (LES) of W/W^v mutant mice raise the interesting possibility of a role of the ICC in human achalasia, a disease in which the LES fails to relax (218). This work needs to be confirmed and extended in the human disease.

Abnormalities of the colonic ICC at an ultrastructural level have been demonstrated in patients with severe ulcerative colitis (173). The role these abnormalities may play in the deranged motility of colonic inflammation remains to be defined. It is possible that the histological abnormalities of the ICC in ulcerative colitis might be secondary to high-dose corticosteroid therapy rather than the inflammation. Glucocorticoids decrease the number of tissue mast cells (also c-kit-positive cells) by downregulating the amount of SCF produced by fibroblasts (55). Therefore, reductions in the amount of available SCF could result in histological abnormalities and even apoptosis of the ICC.

	ISEMF

TOP ABSTRACT INTRODUCTION ICC ISEMF FUNCTIONS OF ISEMF CONCLUSIONS REFERENCES

ISEMF exist in a subepithelial location throughout the gastrointestinal tract from esophagus to anus and in the gallbladder and pancreas, but they are best described in the small intestine and colon (70, 95, 100, 101, 111, 148). Although initially thought of as a sheath of fibroblasts, more dense in the region of crypts than at the surface of the colon or in the villi of the small intestine (70, 100, 127, 148), it is now clear that they exist as a syncytium that extends throughout the lamina propria of the gut, merging with the pericytes surrounding the blood vessels that course through the tissue (95). In the region of the crypts, the myofibroblasts are oval and scaphoid in appearance and appear to overlap like shingles on a roof (100, 137). Even here they are attached one to the other with both gap junctions and adherens junctions, as they are throughout the syncytium. In the upper regions of the colonic crypts and in the small intestinal villi, the ISEMF take on a stellate morphology (Fig. 1) (95, 111, 207).

View larger version (99K): [in this window] [in a new window]   Fig. 1.   Scanning electron micrographs of intestinal subepithelial myofibroblasts (ISEMF) in rat intestinal villi. A: lower-power view showing ISEMF syncytium (*) underneath palisading columnar epithelial cells (EP). Arrows indicate apical surface of goblet cells. Bar, 20 µm. [From Desaki et al. (35).] B: higher-power view demonstrating anastomosing syncytium of stellate fibroblast-like cells (FLC), which are the ISEMF. Bar, 5 µm. [From Gannon and Perry (61a).]

Ultrastructural studies demonstrate close contacts between ISEMF and synaptic vesicle-containing nerve terminals (70). We have found that 18Co colonic myofibroblasts contain carbachol receptors coupled to prostaglandin (PG) E₂ synthesis and reversal of stellate transformation (77, 205). These data suggest that ISEMF can be modulated by cholinergic inputs, and structures resembling dendritic spines on stellate 18Co processes may be the location of these acetylcholine receptors (205).

At transmission electron microscopic resolution, ISEMF have the typical appearance of myofibroblasts: a cell membrane with multiple caveolae, a well-developed rough endoplasmic reticulum and Golgi complex, and a cytoplasm filled with microfilament bundles (stress fibers) and associated dense bodies [see part I of this review (163) and Ref. 205]. In the small intestinal villus, the processes of the ISEMF are even more attenuated and encircle the capillaries that extend throughout the lamina propria (111). Here ISEMF are essentially indistinguishable from pericytes (35) (and indeed may be one and the same).

The basal lamina contains numerous fenestrations, particularly on the upper two-thirds of the villus, and cell processes of the myofibroblasts and/or the epithelium can extend through these fenestrae (100, 111, 202). The myofibroblasts are embedded also in a subepithelial sheet of reticular fibers that also contains fenestrae or foramina through which lymphocytes and macrophages traverse (202). Just below the surface epithelium of the colon, the reticular sheet is called the collagen table. Processes of the myofibroblasts extend through this table and abut the basal lamina under the surface epithelial cells with foot processes reminiscent of those in the kidney glomerulus (100, 148). Thus the subepithelial space appears to have two fenestrated barriers: the basal lamina and the subepithelial reticular sheet in the small intestine (or collagen table in the colon), both of which are formed by connective tissue fibrils secreted by the myofibroblasts. This anatomic feature may have functional implications for water transport by the epithelium (see Water and Electrolyte Transport).

Antibodies that react against -smooth muscle (-SM) actin or against glial fibrillary acidic protein will stain ISEMF as well as ICC, hepatic stellate cells, pancreatic stellate cells, and other stellate-type myofibroblasts such as those in Wharton's jelly of the umbilical cord [see part I of this review (163) and Ref. 61]. The identifying cytoskeletal staining characteristic of ISEMF, however, is their reaction to -SM actin antibodies. ISEMF can be distinguished from smooth muscle of the muscularis mucosae, which is also -SM actin positive, by absence of staining for desmin (Fig. 2) (129, 165, 166). Thus both ISEMF and muscularis mucosae stain for -SM actin and variably for myosin, but the ISEMF, in contradistinction to the muscularis mucosae, is negative (normal intestine) or only weakly positive (severely inflamed intestine) for desmin (129, 165, 166, and A. B. West, unpublished observations).

View larger version (144K): [in this window] [in a new window]   Fig. 2.   A: normal human colonic mucosa stained for -smooth muscle (-SM) actin. Note weak staining of ISEMF and strong staining of muscularis mucosae. B: muscularis mucosae stains strongly for desmin but ISEMF do not. C: cross section of human colon of a patient with radiation colitis. Note that ISEMF are "activated" and strongly express -SM actin. D: note lack of ISEMF desmin staining of the same tissue as in C.

Several independent studies in mice reveal the presence of ICC, as determined by c-kit expression, in the embryo from days 12.5-13.5 onward (10, 176, 199, 220). With the use of -SM actin as a marker, ISEMF in the human embryo are detectable in the intestine at 21 wk of gestation (178). The origin of these cells may be from the neural tubes or neural crest stem cells, migrating along the vagus nerve to the gut (14, 92). Alternatively, ISEMF may transdifferentiate from resident fibroblasts or smooth muscle cells [see part I of this review (163)]. The cells are first visible in immediate juxtaposition with the muscularis mucosae at the base of the intestinal crypts (97). Between the 21st and 39th wk, the number of ISEMF increases progressively in the region between the base and the middle of the crypts, in synchrony with the proliferation and differentiation of the epithelial cells. At birth, the ISEMF line the lower two-thirds of the crypts of the colon, and all of the cells avidly express -SM actin. These studies suggest that antibodies against -SM actin are currently the best markers for ISEMF and, furthermore, that ISEMF may originate from the muscularis mucosae or from cells adajcent to them. It is important to note that c-kit antibody staining, which defines the ICC, has not been studied extensively in ISEMF and, conversely, -SM actin staining, which defines ISEMF, has not been explored well in ICC.

[³H]thymidine incorporation into primary cultures of ISEMF indicates proliferative responses to platelet-derived growth factor (PDGF)-BB, epidermal growth factor (EGF), basic fibroblast growth factor (bFGF), insulin-like growth factors I and II (IGF-I and IGF-II), interleukin-1 (IL-1), and tumor necrosis factor- (TNF-) (94). In this study, synergism between PDGF-BB and IL-1 was demonstrated, and the proliferative response was inhibited by elevation of cellular cAMP levels. The effect of SCF was not reported. mRNA for c-kit can be detected in 18Co cells, human myofibroblasts cultured from a mucosal biopsy of human neonatal colon (Mifflin, Saada, and Powell, unpublished observations). The functional significance of this remains to be determined. These cells also express SCF message and protein (109), as well as message for PDGF- receptor (Mifflin, Saada, and Powell, unpublished observations). The growth factor(s) and receptor set(s) crucial for the development or proliferation of ISEMF remain to be more clearly determined.

	FUNCTIONS OF ISEMF

TOP ABSTRACT INTRODUCTION ICC ISEMF FUNCTIONS OF ISEMF CONCLUSIONS REFERENCES

Mucosal Growth and Development

Fibroblasts capable of driving either epithelial proliferation or differentiation have been isolated from the gut lamina propria (58, 72). Over 15 years ago, Haffen et al. (72) showed that isolated intestinal fibroblasts were as potent as dissected intestinal mesenchyme in inducing intestinal endoderm. More recently, they have shown that the intestinal mucosa contains different clones of "fibroblasts" that will drive either epithelial proliferation or differentiation (58). Transforming growth factor-1 (TGF-1) was capable of transforming one set of isolated fibroblasts into -SM-actin-expressing myofibroblasts and inhibiting their proliferation. In contrast, proliferation of the second, typical, fibroblast cell line was inhibited by IL-2 but not by TGF-. When the typical fibrocyte cell line was implanted into fetal endoderm, the endoderm formed deep crypts due to proliferation of the epithelial cells. In contrast, similar coculture containing the -SM-actin-expressing myofibroblasts demonstrated differentiation rather than proliferation of the epithelial cells: the endoderm developed villi with mature differentiated enterocytes, goblet cells, and endocrine cells (58). There were differences in the amount of basement membrane proteins (laminin 1 and type IV collagen) produced by the two (myo)fibroblast cells lines, but it is unclear whether matrix or growth factor secretion caused these different growth patterns.

Coculture of fibroblasts or myofibroblasts with epithelial cells will cause the epithelial cell to proliferate and differentiate (33, 74). Foreskin and lung fibroblasts (74) and primary cultured ISEMF (33) are capable of inducing cellular differentiation of HT-29 and T84 human colon cancer cells. When grown in collagen gels without (myo)fibroblasts, T84 cells formed round, ball-shaped colonies. When these (myo)fibroblasts were included, a basal lamina formed around the T84 colonies, they differentiated into one-cell-layer hollow balls with a lumen, and the epithelial cells changed from thin, cuboidal cells with occasional microvilli to columnar enterocyte-like epithelial cells with well-developed microvilli (74). In separate culture studies, it was shown that hepatocyte growth factor (HGF) caused proliferation but not differentiation of the T84 cells, whereas TGF- had the opposite effect (74). Furthermore, antibodies against TGF- or against the TGF- latency-associated protein abrogated the differentiating effect of fibroblasts cocultured with the T84 cells (74). Thus (myo)fibroblast-derived HGF appears to be a proliferative growth factor for the intestinal epithelial stem cell, whereas (myo)fibroblast-secreted TGF- appears to promote epithelial cell differentiation.

The studies detailed above suggest that phenotypically different lamina propria fibroblasts are capable of causing either proliferation or differentiation of the intestinal epithelium, that basement membrane formation is an important ingredient in differentiation, and, finally, that HGF is a proliferating growth factor, whereas TGF- is a differentiating growth factor for the intestinal epithelium. Studies of keratinocyte growth factor (KGF) and KGF receptor expression during fetal development suggest that this factor, also of fibroblast origin, is equally important in mediating the morphogenic mesenchymal-epithelial interactions (53).

A third axis of function and differentiation is a proximal-distal gradient from duodenum to distal colon. Plateroti et al. (155) have shown that myofibroblasts from proximal jejunum, distal ileum, and proximal colon secrete different and characteristic amounts of HGF, TGF-1, and epimorphin and have differing effects on endodermal growth in coculture. These studies suggest that the ISEMF may be important in proximal-distal gut differentiation as well as crypt-villus morphogenesis (65).

There is evidence that myofibroblasts, like intestinal epithelial cells, proliferate, migrate, and differentiate along the crypt-villus axis. In [³H]thymidine pulse-labeling experiments of rabbit colon and adult mouse jejunum, both pericryptal myofibroblasts and crypt epithelial cells took up thymidine after a single pulse administration, i.e., both were proliferating (127, 148). Both the labeled cell populations appeared to migrate up the basement membrane over the course of 2-4 days, disappearing at the tip of the small intestine villi or the surface of the colon through the processes of exfoliation and/or apoptosis (73). These experiments suggest that myofibroblasts differentiate as they move from a discoid morphology at the base of the crypt to the stellate shape as they move toward the intestinal lumen. Perhaps maturation and differentiation of the myofibroblasts drive the maturation and differentiation of the epithelial cells that are moving pari passu up the basement membrane. It is also conceivable that there are signals from the epithelial cells that drive myofibroblast differentiation. It should be noted, however, that this finding of crypt myofibroblast labeling and migration could not be confirmed in studies with mouse small intestine and colon (142).

Protection and Wound Healing Functions of ISEMF

It can be speculated also that the ISEMF network alone, or through its connections to the smooth muscle cells of the muscularis mucosae that extend up the villus core, is responsible for the rhythmic villus contractions observed in the small intestine (224). This phenomenon was first described over 150 years ago but has received little study in the past two decades. Almost certainly the ISEMF syncytium accounts for the mucosal contractions in the stomach, which lead to expulsion of gastric acid from the gastric pits in a jetlike fashion (79). Although these cells were described as "smooth muscle cells" originating from the muscularis mucosae, their anatomic location in the gastric mucosa suggests that they are the gastric counterpart of the ISEMF (194).

Intestinal myofibroblasts display receptors for contractile and relaxing agonists such as endothelin and atrial natriuretic peptide (59, 60, 205). Of interest, ISEMF fail to respond to other guanylate cyclase stimulants such as guanylin, heat-stable Escherichia coli enterotoxin (Sta), or sodium nitroprusside, suggesting that these receptors and/or the specific guanylate cyclases are absent from the ISEMF cells (205). The role of the ISEMF, which have the anatomic appearance of pericytes, in the autoregulation of lamina propria blood flow remains to be determined.

Restitution. The rapid migration of epithelial cells over a denuded basement membrane is an important intestinal repair response to minor/moderate injury (214). It is likely that ISEMF play an important role in this process, both because their location under the basement membrane is ideal for paracrine action and because myofibroblasts secrete the agents that thus far have been shown to enhance epithelial cell migration in experimental disease states in vivo and in wounding models in vitro. Factors promoting restitution include PGE₂ (in synergism with prostacyclin) (13), IL-1 (37), interferon- (IFN-) (37), TGF- (37), EGF (37, 157, 170), acidic FGF (aFGF) and bFGF (39), TGF- (especially TGF-3) (25, 130), and HGF (143, 195). The most compelling evidence for a role of ISEMF in restitution comes from the laboratories of Mahida and Podolsky (130). They have shown that the active factor in conditioned media from primary cultures of ISEMF that promote IEC-6 or T84 epithelial cell motility is TGF-3. The PGs may have their action by virtue of stimulating HGF secretion, since anti-HGF antibodies abrogate PG acceleration of gastric epithelial cell restitution in a coculture model of gastric epithelial cells and fibroblasts (195). Similarly, anti-TGF- antibodies block the restitution-promoting effects of IL-1, IFN-, TGF-, EGF, aFGF, and bFGF (37, 39). Thus these agents may not act directly on the epithelial cell to promote cell motility; rather, they may augment the secretion of and/or the expression of receptors for TGF- and HGF. Of interest, TGF- may also enhance the barrier integrity of newly formed tight junctions between epithelial cells and protect against the barrier-destroying effects of IFN- (154). Trefoil peptides also stimulate intestinal epithelial restitution (106), but these factors are secreted by epithelial mucus cells (goblet cells), not by ISEMF (24). It is of interest that PDGF, an agent that stimulates myofibroblast and smooth muscle motility, was without effect on the motility (restitution property) of the intestinal epithelial cell (37).

Protection and repair. Tissue repair is a complex, coordinated event [see part I of this review (163)], in which there is release of various lipid mediators such as eicosanoids, gases such as nitric oxide, cytokines such as TNF-, IL-1, IL-6, IL-2, and IL-15, and the various growth factors mentioned in the previous section. Many of these factors activate myofibroblasts, resulting in myofibroblast motility and the release of ECM proteins and other growth factors. Remodeling of intestinal tissue (villous atrophy/crypt hyperplasia) is also an important response to gut injury (121). There is evidence that myofibroblasts take part in this process through the secretion of matrix metalloproteinases and other proteases (32, 121), as well as secretion of TGF- and KGF (5). In a cascade fashion, growth factors induce angiogenesis and epithelial proliferation and differentiation. The process is terminated by myofibroblast apoptosis [see part I of this review (163)].

Myofibroblasts take part in the healing of gastrointestinal ulcerations. Myofibroblasts are identified as -SM-actin-positive cells at the base of gastric and duodenal ulcers (140, 141) and in the colon after experimental lesions (211). In vitro studies of primary ISEMF cultures suggest that the important factors inducing myofibroblast activation and proliferation are TNF-, IL-6, EGF, bFGF, IGF-I and IGF-II, PDGF-BB, and IL-1 (the latter two synergistically) (94).

Many of the factors secreted by the activated myofibroblasts, as well as their respective receptors, are upregulated in the intestine in various disease states. Examples include PGs via cyclooxygenase (COX)-2 activity (104, 132, 196), TGF- (4, 38, 191), EGF (117), TGF- (4, 38), bFGF (86), HGF (107, 180), and KGF (23, 54, 107) in experimental colitis, small bowel injury (38), or gastric ulcer models or disease (86, 107, 180), and also in naturally occurring inflammatory bowel disease (4). Furthermore, animals lacking TGF- are more susceptible to colonic injury, and this susceptibility is ameliorated by exogenous administration of the factor (48). In fact, the exogenous administration of many growth factors such as EGF (156, 170), bFGF (87), IGF-I (152), KGF (230), and IL-II (118, 145) results in amelioration of experimental or natural gastrointestinal injury.

Not only are these growth factors upregulated in intestinal disease or injury, they have proliferative effects on normal or transformed epithelial cell lines in culture and cause epithelial proliferation in vivo. ET-1 (184), IL-2 (184), TGF- (96), IL-15 (167), aFGF (39), bFGF (39), KGF (39, 107), and HGF (107) all have mitogenic activity in epithelial cell culture models. Administration of EGF and TGF- (81, 158), IGF-I and II (158), and KGF (82) to animals causes intestinal epithelial cell and, in some cases, hepatocellular proliferation. Transgenic animals overproducing IGF-I have predominantly muscle layer proliferation (215), suggesting that this factor may be more important in muscle repair.

A protective role for nitric oxide cannot be overlooked. Inducible nitric oxide synthase-deficient mice have poor wound healing that is corrected by exogenous administration of nitric oxide (226). Furthermore, administration of nitric oxide donors prevents gastrointestinal injury from nonsteroidal anti-inflammatory drugs (NSAIDs) (50). While other myofibroblasts (e.g., the renal mesangial cell) are significant producers of nitric oxide (112), it is unclear whether ISEMF or ICC are significant sources in the intestine (131, 225).

Fibrosis

A major question regarding intestinal fibrosis is whether the increase in matrix production emanates from activated resident fibroblasts, smooth muscle cells, or myofibroblasts. Normally, intestinal smooth muscle expresses -SM actin, desmin, and tropomyosin. The "smooth muscle cells" isolated and cultured from Crohn's disease bowel demonstrate increased amounts of -SM actin, -enteric actin, and desmin (66). In situ, normal ISEMF are desmin negative, but staining characteristics can change during culture or disease states, and in situ immunohistochemical studies of Crohn's disease strictures are not available. We have found increased expression of desmin in ISEMF in Crohn's disease intestine (West, unpublished observations). The intestinal "smooth-muscle-like" cells responsible for intestinal fibrosis are more stellate in appearance, have more prominent stress fibers, and contain more extensive, dilated endoplasmic reticulum than conventional intestinal smooth muscle cells (66), all characteristics of myofibroblasts [see part I of this review (163)]. We believe, therefore, that the cells responsible for fibrosis in Crohn's disease are activated myofibroblasts (either ISEMF or ICC) and not conventional intestinal smooth muscle. Furthermore, Western analysis demonstrates that cultured ISEMF express several ECM proteins, including type IV collagen and 1- and 1-laminin and fibronectin (122). Perhaps this is semantics, because it is entirely possible that myofibroblasts transdifferentiate from the smooth muscle cells of the muscularis mucosae [see part I of this review (163)].

There are a number of interesting characteristics of the Crohn's disease, collagen-producing, smooth-muscle-like cells. The mesenchymal cells isolated from the lamina propria of Crohn's strictures produce more total collagen and more type III collagen than mesenchymal cells isolated from nonstrictured or normal intestine (66). TGF- stimulates (67), while IL-1 and PDGF-BB downregulate, procollagen secretion but not its synthesis (69). In contrast to the similar effect of IL-1 and PDGF on collagen formation and secretion, IL-1 caused a marked concentration-dependent increase in collagenase mRNA levels, whereas PDGF had no effect (69). IL-1 and dexamethasone inhibit collagen synthesis in dermal fibroblasts, yet they increase procollagen synthesis by intestinal smooth muscle cells (128). At certain concentrations in vitro, glucocorticoids actually increase procollagen gene expression by human intestinal smooth muscle cells (68), a finding with potentially important therapeutic implications for the treatment of inflammatory bowel disease with corticosteroids.

Collagenous colitis is a disease that appears to be due to abnormalities of or activation of the subepithelial myofibroblasts of the colon (1, 231). This is a form of "inflammatory bowel disease," causing watery diarrhea predominantly in middle-aged or older women, who often have other manifestations of autoimmune diseases. The pathological abnormality in collagenous colitis is defined entirely at the microscopic level, where it is characterized by an expansion of the normal reticular band (collagen table) that exists in the subepithelial region of the surface epithelium of the colon, i.e., in the location of the ISEMF (231). Along with the collagen band, there is an increased number of the mononuclear inflammatory cells in the lamina propria. The possibility of an abnormality in collagen type or secretion in this disease has been previously studied with somewhat controversial findings (1). It has been suggested that in this disease there is a normal distribution of type IV collagen, increased type III collagen, and reduced or abnormal type I collagen production (1). Normal type IV collagen distribution in this disease indicates that the band is not an expanded basement membrane. More recent studies suggest increases in type VI collagen and tenascin in the collagen band under the surface epithelium, with focal increases in types I, III, and VI collagen in the area of the crypts (1). However, no increase in type VI collagen mRNA was found in the myofibroblasts around and entrapped in the collagen table. These studies suggest that the accumulation of type VI collagen is not caused by increased collagen synthesis by the myofibroblasts but rather by abnormalities in the degradation of the intracellular matrix, e.g., decreased secretion of metalloproteinases or increased secretion of tissue inhibitors of metalloproteinases (1).

Granulomas of the intestine are basically of two types: accumulations of only macrophages as seen often in Crohn's disease or -SM-actin-positive myofibroblasts surrounding macrophages and/or giant cells, which are thought to be of either macrophage or myofibroblast origin, as commonly seen in sarcoidosis or foreign body reactions (174). A model for granuloma formulation in the intestine is the injection of sterile bacterial peptidoglycan-polysaccharide polymers into the bowel wall (232). The bacterial cell wall polymers increased both TGF- immunostaining and collagen deposition, as well as collagen -1 mRNA in the granulomas (211). Furthermore, TGF- directly stimulated collagen type I, TGF-1, IL-1, and IL-6 mRNA in myofibroblasts cultured from the granulomas (211). Thus granuloma myofibroblasts represent another potential origin for the intestinal fibrosis of Crohn's disease.

Immunology and Inflammation

At the electron microscopic level, it is possible to find close apposition of ISEMF and lymphocytes, suggesting that the myofibroblasts may take part in the growth and development of the T cell (202). Fiocchi (56), Toyoda et al. (202), and Roberts et al. (171) have shown that experimentally induced proliferative responses in T cells are enhanced by coculture with intestinal fibroblasts or with smooth muscle cells.

PGs derived from either constitutive (COX-1) or inducible (COX-2) forms of cyclooxygenase (PGH₂ synthase) have fundamental regulatory roles in gastrointestinal barrier function, inflammation, the immunophysiology of intestinal electrolyte transport and motility, and gastrointestinal neoplasia (13, 21, 26, 42, 46, 159, 210). The cellular sources of PG in normal and diseased intestine are poorly defined. It does seem likely, however, that the epithelium is the least contributor. When the intestinal mucosa is separated with a combination of sharp dissection and Ca²⁺ chelation into epithelial and subepithelial (lamina propria, submucosa, and muscularis propria) fractions, 95-99% of rat mucosal PG synthesis emanates from the subepithelial fractions (27, 28, 114). Myofibroblasts are but one of several tissue elements in the subepithelium capable of producing PGs (11, 77, 104, 122, 197, 205); white blood cells of various types, smooth muscle cells, endothelial cells, and conventional fibroblasts are among the other PG-producing constituents (159). The relative proportions of PGs secreted in normal or inflamed states by intestinal myofibroblasts compared with these other elements remain to be determined. Studies of nonstimulated intestine show COX-1 to be the predominant isoform producing basal secretion of PGs in the gastrointestinal tract (98). In the experiments noted above, in which intestine was dissected into epithelial and subepithelial fractions, it is likely that COX-2 is activated by the trauma of the experimental procedures.

Immunohistochemical studies show constitutive expression of COX-1 in the crypt epithelium of the small intestine (26) and predominantly in mucous neck cells of the gastric glands (88). However, epithelial COX-2 expression can be induced. Experimental injury (injection of endotoxin) (52) and bacterial (Salmonella) invasion (47), as well as natural disease states such as Crohn's colitis, ulcerative colitis (76, 187), and colonic adenocarcinoma (45), result in expression of COX-2 in intestinal epithelial cells. Furthermore, COX-2 has been cloned from nontransformed rat intestinal epithelial (RIE-1) cells, a cryptlike intestinal epithelial cell line (43).

ISEMF also express COX-1 and COX-2 (Fig. 3) (77, 104, 122, 205). IL-1 increased COX-1 mRNA in 18Co ISEMF after 24 h exposure, but the threefold increase in message was not accompanied by detectable increases in protein, as determined by Western analysis (175). In contrast, IL-1 induces an exponential increase in COX-2 mRNA and protein levels in 18Co cells. Under these conditions, COX-2 accounts for 95% of the total PGE₂ synthesis (77).

View larger version (58K): [in this window] [in a new window]   Fig. 3.   Northern analysis of 18Co total RNA after stimulation with interleukin-1 (IL-1) for 5 and 24 h. A: 2.7-kb cyclooxygenase (COX)-1 hybridization shows constitutive expression of message and a 3-fold increase only after 24 h of incubation with IL-1. B: 4.2-kb COX-2 message is increased from barely detectable levels to >10-fold by within only 5 h of incubation with IL-1. [From Hinterleitner et al. (77).]

Immunocytochemical studies localize COX-2 to the lamina propria of the inflamed stomach and intestine (52, 168, 187). This subepithelial localization of COX-2 is usually ascribed to either macrophages or blood vessels, and expression in ISEMF has not been sought or clearly demonstrated. In contrast, COX-2 expression has been definitively located to lamina propria interstitial mesenchymal cells (probably ISEMF) in the early stages of polyp formation in the adenomatous polyposis coli (APC) knockout mouse (146). Thus the relative roles of epithelial, white blood cell, endothelial, and ISEMF PG production, as well as the relative roles of COX-1 vs. COX-2 enzymatic PG production, in intestinal barrier function, inflammation, immunophysiology, and neoplasia remain to be determined. Most studies suggest COX-1 is the predominant enzyme responsible for barrier function and protection against damage, whereas COX-2 expression is increased in gastrointestinal inflammation (13, 26, 29, 46, 47, 52, 210, 212). Whether COX-2 inhibition is sufficient to alleviate pain and inflammation is still a topic of debate (168, 213). Nevertheless, this concept of differential functions of COX-1 and COX-2 has been the theoretical basis for the development of COX-2-selective NSAIDs (210). This concept has been challenged by experiments that show that COX-1 contributes to the inflammatory responses of the gastrointestinal mucosa and that inhibition of COX-2 can exacerbate inflammation in the colon (168, 213). To date, initial reports suggest that selective COX-2 inhibition does, indeed, cause less gastrointestinal mucosal damage than conventional NSAIDs and is effective in symptom relief (75, 120, 185).

Water and Electrolyte Transport

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Short-circuit current response (Isc) of T84 colon carcinoma cells mounted in an Ussing chamber either alone (A) or cocultured with neonatal piglet jejunal myofibroblasts (P2JF) (B), or of separately cultured T84 and P2JFs that were acutely juxtaposed (C). Note augmentation of Isc in response to H2O2 when myofibroblasts are present. Indomethacin (INDO) inhibits augmented Isc (B and C). [From Powell and Berschneider (162).]

View larger version (23K): [in this window] [in a new window]   Fig. 5.   A: preincubation (24 h) of human colonic myofibroblasts (18Co) with IL-1 augments Isc response of T84 cells acutely juxtaposed with 18Co cells and then stimulated with histamine or carbachol. B: PGE2 production from 18Co cells is correspondingly increased by IL-1 preincubation. Both augmented Isc response and PGE2 production induced by IL-1 are prevented by simultaneous incubation with IL-1 receptor antagonist (IL-1ra). [From Hinterleitner et al. (77).]

The studies above were performed in a reductionist model: cell culture of epithelium and myofibroblasts. There is, however, evidence from studies of native tissues that PG release is part of immune mediator, neurotransmitter, and cholera toxin-induced Cl secretion (7, 21, 93, 103, 123, 153, 159, 162, 189, 219). These studies also raise the question whether some of the abnormal electrolyte secretory responses in the intestine of W/W^v mice are due to loss of ISEMF rather than loss of mast cells (216). Some of these abnormal electrolyte transport responses (e.g., the cholera toxin response) were not reconstituted by bone marrow transplantation that replaced mast cells in these c-kit-negative animals (216).

Absorption. Recently, Naftalin and colleagues (137, 138) have published intriguing experiments that suggest ISEMF also have a critical function in intestinal water and electrolyte absorption. These investigators had previously suggested that colonic crypts act as suction devices to dehydrate feces in the distal segment of the mammalian colon (150, 160). Such a process requires that the colon absorb fluid hypertonically (which it is known to do) (12), but the mechanism of hypertonic absorption by an epithelium has heretofore not been evident (161). Naftalin and colleagues showed that the ISEMF sheath represents a diffusion barrier to Na⁺ transport that creates a significantly hypertonic middle compartment between the epithelial tight junctions and vascular system. It has been shown experimentally that such a three-compartment model will permit hypertonic transport in keeping with the Curran-MacIntosh model of water movement across epithelial tissues (30, 137, 138). Furthermore, they demonstrated proliferation of the ISEMF sheath in response to dietary Na⁺ depletion, which would account for the increase in Na⁺ and water absorption in high ANG II-high aldosterone states. ANG II (and/or aldosterone) has been demonstrated to cause proliferation of cardiac myofibroblasts and to increase their secretion of ECM fibrils (19).

ISEMF may be responsible for differentiation of epithelial cells from a secretory to an absorptive phenotype. When T84 colon cells, a secretory cell line, are cocultured with 18Co human myofibroblasts, fluid-filled domes develop under the T84 cells (205). Dome formation in epithelial cell culture systems is usually thought to result from Na⁺ and water transport from the culture media to the space between the epithelium and the plastic culture dish.

Intestinal Polyps and Neoplasia

View larger version (124K): [in this window] [in a new window]   Fig. 6.   -SM actin (A) and desmin (B) staining of a human colonic tubular adenoma. Note the activated (-SM actin positive, desmin negative) myofibroblasts in the lamina propria of this benign neoplasm.

GISTs are the most common of the mesenchymal tumors of the gastrointestinal tract. Recently, it has been shown that they make up a very high percentage of conventional mesenchymal tumors and can be differentiated from classic smooth muscle tumors (leiomyomas) by the expression of c-kit (78, 105, 139, 188). These c-kit-expression cells are also CD34⁺, suggesting a stem cell origin and, furthermore, that they originate from ICC. Because ISEMF may well be c-kit-positive cells as well, it remains to be determined which myofibroblast is the cell of origin. Alternatively, GISTs may be derived from some progenitor stem cell of both intestinal myofibroblasts. GISTs often contain mutations in the transmembrane or tyrosine kinase domain of c-kit. The stable introduction of a mutant c-kit cDNA into murine lymphoid cells induces malignant transformation. This suggests that these mutations contribute to tumor formation and, as such, are "gain of function" mutations.

Myofibroblasts appear to be the primary mesenchymal element in the lamina propria of sporadic, hyperplastic, hamartomatous, and adenomatous polyps (108, 113, 178, 181). They are also the principal mesenchymal cell of inherited neoplasms such as juvenile polyposis, Peutz-Jeghers syndrome, and familial polyposis (91, 108, 113). This has led Kinzler and Vogelstein (108) to classify juvenile polyps and the hamartomas of ulcerative colitis as members of the "landscaper defect tumors." The hypothesis is that the stromal cells secrete growth factors that make the epithelium more susceptible to neoplastic transformation, much in the same way cultivated soil allows for the better growth of plants (108). Primary defects in the mesenchymal cells of juvenile polyposis tumors are loss of a suppressor locus in chromosome 10q22 (91) or germ line mutations of the gene SMAD4 located on chromosome 18q21 (83). Hyperplastic polyps are characterized by abnormally shaped or branched crypts (3). It is intriguing to speculate that myofibroblasts, also through secretion of growth factors, might induce this abnormal crypt morphology.

Pericryptal fibroblasts are believed to play a role in the histological features and tumor growth patterns of colorectal neoplasms (169, 227). For example, it has been suggested that well-developed ISEMF are a more consistent feature of adenomas than of carcinomas, and polypoid carcinomas are more likely to have well-developed ISEMF than nonpolypoid growths (227). ISEMF appear to be responsible for the desmoplastic (fibrotic) reactions seen in many gastrointestinal tumors (198, 228). Because TGF- is a prominent fibrogenic growth factor, a role for this cytokine has been suggested (116). The TGF- could be of either myofibroblast or neoplastic epithelial origin. Furthermore, ISEMF have been proposed to play an important role in tumor metastasis, either through formation of an abnormal basement membrane (15, 129, 165, 166), through their ability to secrete matrix metalloproteinases that allow the breakdown of tumor-containing connective tissue (17, 51), or through their ability to alter cell-to-cell adhesion protein expression (40). Because myofibroblasts themselves are mobile cells, they could conceivably "carry" tumor cells into adjacent tissue, lymphatics, or blood vessels (129). Perhaps most importantly, TGF- secreted by tumor myofibroblasts may directly stimulate tumor cell motility (37, 39, 130).

In addition to its potential to cause desmoplastic reactions in cancer, TGF- has important effects on tumor cell growth. Because TGF- induces apoptotic cell death in many epithelial cells, the finding that the type II receptor (TGF- RII) is mutated or absent in many colon carcinomas suggests that such mutations might play a fundamental role in tumor development (125, 126). These mutations in TGF- RII develop predominantly from microsatellite instability secondary to the DNA mismatch gene repair abnormalities seen in hereditary nonpolyposis colon cancer syndrome (2) and in ulcerative colitis-associated colorectal neoplasms (192) but not necessarily in sporadic colorectal cancer (204). Patients with colorectal cancer may have increased plasma levels of TGF-, and this appears to be of lamina propria (myofibroblast?) origin (204). Because TGF- production and receptors may be of either myofibroblast or epithelial origin, more definition of this area is needed to sort out the role of this growth hormone and its receptors in colorectal cancer.

Perhaps the most important connection between myofibroblasts and colorectal cancer is in the definition of the mechanisms whereby NSAIDs cause regression of polyps in FAP and the mechanism of NSAID chemoprevention of sporadic colon cancer. Aspirin and other NSAIDs clearly are chemopreventive for the development of polyp expression in mouse and human FAP (6, 62, 146, 149) and in sporadic colorectal cancer in humans (63, 64, 177). The mechanism of this antitumorigenic and chemopreventive effect is unclear and much discussed (41, 49, 71, 80, 124, 134, 183, 221, 223). The most recent hypothesis is that NSAIDs cause increased apoptosis of colonic epithelial cells (179, 182, 203) through inhibition of PG synthesis; the resulting elevated arachidonic acid levels in the cell are thought to stimulate the conversion of sphingomyelin to ceramide, a known inducer of apoptosis (22). An alternative hypothesis is that NSAIDs may cause apoptosis of growth-factor-secreting ISEMF, which leads secondarily to decreased epithelial proliferation or increased epithelial apoptosis.

COX-1 is detected in normal colonic epithelial cells (26) as well as colonic cancers (45, 99). In contrast, COX-2 is rarely detectable in normal tissues but is found in >90% of colon tumors. COX-2 expression is present in less than one-half of premalignant colonic polyps (45, 99). COX-1 and COX-2 are also found in ISEMF (77, 122). Therefore, the key question is whether the effect of NSAIDs resides at the level of the neoplastic epithelial cell (223) or in the subepithelial mesenchymal tissue (probably myofibroblasts) (71, 80). Of note, NSAID administration causes apoptosis of embryonic fibroblasts (119). Loss of growth-factor-producing ISEMF, through a similar apoptotic process, could affect tumor epithelial cell proliferation. Studies of the APC model give some insight into this question (164). In the APC knockout mouse, specific COX-2 inhibitors have been shown to suppress the development of intestinal polyps (146), and the site of COX-2 expression in these animals was subepithelial mesenchymal cells, probably myofibroblasts (146). However, adenomas originating in Min mice also express high levels of COX-2, some of which can be localized to the epithelium (222). These observations, combined with the fact that a large percentage of epithelial tumors express COX-2, suggest that, along the progression from polyp to cancer, acquisition of the ability to express COX-2 provides a selective advantage to the cancer cell.

A final area of consideration is the possible role of tissue myofibroblasts in the desmoid tumors, cutaneous cysts, dental abnormalities, osteomas, and retinal pigment epithelial abnormalities often seen in the Gardner's variant of FAP (190). The desmoid tumors are certainly of myofibroblast origin, and myofibroblasts are located in the areas of the other various extraintestinal manifestations [see part I of this review (163)]. A mouse model with these extraintestinal manifestations has been reported (190), and perhaps study of this animal will clarify the role of the myofibroblasts in this syndrome.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION ICC ISEMF FUNCTIONS OF ISEMF CONCLUSIONS REFERENCES

Scientists are just beginning to appreciate the multiple roles of intestinal myofibroblasts, cells that were discovered nearly half a century ago. They play fundamental roles in the growth and development of the intestine, its protection from noxious agents, and its repair after damage. These cells have specific functions in the gastrointestinal tract, including the neuromodulation of intestinal motility and the regulation of intestinal water and electrolyte transport. They also play fundamental roles in intestinal inflammation, but specific functions such as altering the phenotype of lymphocytes and/or the development of oral tolerance remain to be determined. Their exact role in the development of intestinal neoplasms is still unclear, but some effect seems certain. These cells will be exciting targets of investigation over the next several years.