Altered mucin expression in the gastrointestinal tract: a review

Transcription

1 J.Cell.Mol.Med. Vol 5, No 3, 2001 pp Frontiers in Gastroenterology Altered mucin expression in the gastrointestinal tract: a review J. R. Jass a *, M. D. Walsh b a Department of Pathology, University of Queensland, Queensland, Australia b Royal Brisbane Hospital Research Foundation, Queensland, Australia Received: July 30, 2001; Accepted: September 12, 2001 Introduction Carbohydrate component Mechanisms underlying normal variation - Blood group substances - Sialic acid - Transitional mucosa Mechanisms in disorders of growth and differentiation - Sialic acid alterations - Altered assembly of carbohydrate side-chains - Altered expression of core proteins - Metaplasia Disease specific alterations in mucin core protein expression - Ulcerative colitis - Ileo-anal pouch - Ulcer associated cell lineage - Helicobacter pylori gastritis - Intestinal metaplasia of gastric mucosa - Barrett's oesophagus - Gastric neoplasia - Colorectal polyps and cancer Conclusion Abstract Early studies of changes in mucin expression in disorders of the gastrointestinal tract focused on alterations in the carbohydrate chain. This review briefly considers the various mechanisms by which such alterations may come about: (a) normal variation, (b) sialic acid alterations, (c) defective assembly of carbohydrate side-chains, (d) changed expression of core proteins and (e) epithelial metaplasia. The availability of monoclonal antibodies to mucin core proteins adds a new dimension to mucin histochemistry. It is now possible to offer explanations for traditional mucin histochemical findings on the basis of lineage-specific patterns of mucin core protein expression. Changes in core protein expression are described in inflammatory, metaplastic and neoplastic disorders of the gastrointestinal tract. The possibility that mucin change could be important in the aetiology of some diseases such as ulcerative colitis and H. pylori gastritis is considered. It is more probable, however, that changes in mucin expression are secondary to reprogramming of cellular differentiation and altered cell turnover. As such they may serve as markers to explain pathogenesis and provide novel diagnostic and prognostic information. Keywords: mucin gastrointestinal tract histochemistry differentiation metaplasia neoplasia polyps sialic acid Helicobacter pylori *Correspondence to: Prof. Jeremy R. JASS Department of Pathology, University of Queensland School of Medicine, Herston Road, Queensland 4006, Australia. Tel.: , Fax: j.jass@mailbox.uq.edu.au

2 Introduction In this review, observational data relating to the distribution of mucin core proteins in health and disease are integrated with earlier literature on biochemical and histochemical alterations of the carbohydrate side-chains. Consideration is given to general mechanisms underlying: (1) variation of mucin expression in the normal gastrointestinal tract mucosa, and (2) disorders of growth and differentiation. These mechanisms are addressed in specific conditions or lesions of the gastrointestinal tract commencing with inflammatory conditions exemplified by chronic inflammatory bowel disease, the ileo-anal pouch, the ulcer associated cell lineage and Helicobacter pylori gastritis. The possibility that mucin alterations might serve as primary defects underlying inflammatory disease aetiology is discussed. Nevertheless, different pathogenetic pathways are likely to reflect different underlying aetiologies and this concept will be expanded with respect to gastrointestinal metaplasia and neoplasia and illustrated specifically with the examples of intestinal metaplasia of gastric mucosa, Barrett s oesophagus, gastric neoplasia and colorectal polyps and cancer. Carbohydrate component The carbohydrate component is arranged in oligosaccharide chains of varying length and degree of branching, and the chains are attached to the protein backbone by covalent linkage. Oligosaccharide chains containing 2 to 12 monosaccharides have been demonstrated in goblet Fig. 1 Examples of carbohydrate chain structures in epithelial mucins of the gastrointestinal tract. The longer chain (top) comprises core structure 3, a repetitive backbone and peripheral blood group A. The shorter chain (below) is sialosyl Tn (STn) formed through a2,6 sialylation of Tn, the precursor of core structure 3. SA: sialic acid, Fuc: L- fucose, Gal: D-galactose, GlcNAc: N-acetyl-D-glucosamine, GalNAc: N-acetyl-D-galactosamine. 328

3 J.Cell.Mol.Med. Vol 5, No 3, 2001 cell (colonic) mucin [1-3]. During the course of biosynthesis, monosaccharides are added in stepwise fashion by specific glycosyl transferases. Compositional analysis of intestinal goblet cell mucin has demonstrated the presence of five saccharides: the hexosamines N-acetyl-Dgalactosamine (GalNAc) and N-acetyl-Dglucosamine (GlcNAc); the hexose D-galactose (Gal); smaller amounts of 6-deoxyhexose L-fucose (Fuc); and a variety of neuraminic (sialic) acid derivatives. Sulphation of certain sugars contributes further to the acidity [4]. Examples of long and short oligosaccharide chains are shown in Fig. 1. The oligosaccharide chains can be considered to have three structurally distinct domains: a core region incorporating the linkage to the protein, a backbone, and a peripheral region [5,6]. Among the four best-established core structure, core 3 structure is well represented within gastrointestinal mucin: GlcNAc 3GalNAc-Oserine/threonine [1,2]. The precursor of core structure 3 is the Tn antigen (GalNAc-O-R). The presence of core 1 structure (Gal 1 3GalNAc-O- R) (T antigen) within either normal or cancerous mucin is generally accepted [7] but may be an erroneous view. The abnormal expression of T antigen within colorectal cancer was demonstrated by means of the lectin peanut agglutinin (PNA) [8-10] and by immunohistochemistry [11]. However, apart from the fact that core 3 structure predominates in structural analyses of human colonic mucin [1,2], PNA is not in fact specific for T antigen, binding also to the backbone disaccharide sequence (Gal 1 3/4GlcNAc-R) [5,6]. Furthermore, highly specific monoclonal antibodies to the T antigen fail to localize this structure in either normal or malignant tissues of the colorectum [12]. The backbone of the oligosaccharide chain is made up of one or more disaccharide repeats (Gal 1 3/4GlcNAc-). There are two types of backbone depending on whether the linkage is 1 3 (type 1) or 1 4 (type 2). This variable also determines the type of peripheral blood group substance (see below). The mucins of normal intestinal goblet cells are constituted predominantly but not exclusively of type 1 oligosaccharide chains (hence type 1 blood group substances). Increased amounts of type 2 chain are found in cancer mucin [13,14]. Mechanisms underlying normal variation Much of the variation in mucin structure in the normal gastrointestinal tract occurs at the level of carbohydrate side-chain structures than through modified expression of core mucin proteins. Blood group substances The terminal component of the oligosaccharide chain comprises a blood group substance. The type (1 or 2) is determined by the structure of the backbone of the side-chain. Type 2 chain dominates in secretory mucin. The variability of the terminal component is under genetic control centring on the Lewis (Le) gene or FUT3 and the secretor (Se) gene or FUT2 [15]. The Se (FUT2) gene is carried by about 75% of Caucasians [5] and converts the precursor structure Lec to H substance. H substance is in turn converted to blood group substances A, B and Leb according to genetic constitution (Leb is expressed by all secretors). Synthesis of Lea and SLea does not depend on the presence of the Se (FUT2) gene. The Le (FUT3) and Se (FUT2) genes compete with each other and with other transferases involved with chain elongation or branching as well as with sialotransferases (Fig. 2). During fetal life there is expression of blood group substances by secretory mucin throughout the gastrointestinal tract. Subsequently, A, B, H and Leb are restricted to the foregut and midgut (in secretors) while expression is lost in the distal large intestine [16]. Expression of Lea is not influenced by secretor status and occurs throughout the gut. SLea occurs in sites of sialylation, namely small and large intestine. A third fucosyltransferase gene (H or FUT1 is required for synthesis of type 1 chains. The type 1 homologues of Lea, Leb, and SLea are Lex (hapten X, CD15, SSEA), Ley and SLex respectively. Sialylation precedes the final and terminating step of fucosylation in the biosynthesis of blood group substances Slea and Slex (Fig. 1) [17]. Type 1 blood group substances show little expression within the normal gastrointestinal tract. Trace amounts are coexpressed with MUC1 in the basal compartment of colonic crypts [18]. In addition, FUT1 and therefore type 1 chain biosynthesis occurs in normal stomach [19]. 329

4 Fig. 2 Synthetic pathways for the formation of type 1 Lewis structures within gastrointestinal mucin. Sialic acid Sialic acid differs from other sugars contributing to the oligosaccharide chain in having an acidic carboxy group and a three carbon sidearm existing in a number of variant forms. Sialic acid is largely limited to the secretory mucins of the small and large intestine where it is added either to short core structures (STn), to the oligosaccharide backbone, or to terminal blood group substances (see above). Sialic acid variants are of two major types, with or without O-acetyl substituents. O- acetyl groups may be substituted within the three carbon sidearm (C7,8,9) or at C4 [20]. O-acetyl substitution renders sialic acid resistant to neuraminidase digestion. O-acetyl sialic acid is found in the secretory mucin of large intestine, whereas small intestinal goblet cell mucin is mainly non-o-acetylated. The presence or absence of O-acetyl groups influences antigenic determinants. For example sialylated structures incorporating non-o-acetyl sialic acid are recognised by monoclonal antibodies AM-3 [21], TKH2 (STn) [22] and a panel of antibodies to small intestinal mucin antigen (SIMA) [23]. Structures incorporating O-acetyl sialic acid are recognised by antibodies PR3A5 [24], 3NM [25,26] and MMM-17 [27]. The simplest technique for demonstrating small intestinal-type or non-o-acetylated sialic acid is mild periodic acid Schiff (mpas). Interestingly, in about 9% of Caucasians [28] and a higher proportion of Asians [29], colorectal goblet cells are positive with mpas. The same subjects express the tumour-associated structures Slex, Slea and STn within colorectal goblet cells [22,30]. The secretion of non-o-acetyl sialic acid in the colorectum is likely to be explained by genetic variation with respect to the gene coding for O- acetyl transferase (OAT). Nine percent of the Caucasian population will be homozygous for inactive variants of the OAT gene (OAT-/OAT-). From the Hardy-Weinburg equilibrium, 42% of the population is predicted to be heterozygous (OAT+/OAT-). Intestinal crypts are clonal units and inactivation of OAT+ by mutation or loss in a 330

5 J.Cell.Mol.Med. Vol 5, No 3, 2001 crypt stem cell of a heterozygous subject will result in cryptal conversion to OAT-/OAT-. Monocryptal expression of non-o-acetyl sialic acid may then be visualised by mpas staining or expression of STn by immunohistochemistry [31]. This can be demonstrated in about 42% of subjects, as predicted by the Hardy-Weinburg equilibrium. Transitional mucosa The transitional mucosa adjacent to colorectal cancer is often thickened and formed of elongated and branched crypts lined by a tall epithelium. Secretory mucins show loss of sulphation and/or increased sialylation [32] and reduced O- acetylation of sialic acid [33]. The latter may relate to the demonstration of neuraminidase-sensitive small intestinal mucin antigen (SIMA) [23,34]. Despite these and other changes, transitional mucosa is generally regarded as a reactive process, since similar structural and functional alterations occur adjacent to metastases and in colorectal mucosa undergoing to prolapse [35]. Aberrant expression of MUC5AC has been reported in mucosa in the vicinity of villous adenomas of the colorectum [36]. The possibility that this could represent a pre-malignant field change could not be confirmed, however [37]. Mechanisms in disorders of growth and differentiation Instances in which mucins appear to be altered or expressed differently in the context of disorders of growth and differentiation are numerous and open to several interpretations. This is an area of research that is influenced by several decades of investigation of the carbohydrate side-chains of epithelial mucins through a variety of techniques including traditional histochemistry, lectin histochemistry, and immunohistochemistry. A weakness underlying much of this work has been a tendency to ascribe unwarranted clinical and scientific significance to the changes without considering the underlying mechanisms adequately. Loss or gain of a staining pattern may be explained by a surprisingly large number of unrelated mechanisms. These will now be outlined. Sialic acid alterations As noted above, sialic acid occurs in sidearm and peripheral locations of the carbohydrate chain of small intestinal and colorectal goblet cell mucins. O-acetyl substituents at C4 and C7,8,9 characterise colorectal sialic acid and explain resistance to digestion by neuraminidase (sialidase). Loss of O- acetyl substituents may be genetically determined, occur spontaneously in normal and diseased colorectal tissues and may be achieved as a histochemical procedure through the interposition of a step of saponification. Loss of O-acetyl substituents has two major effects: (1) increases sensitivity of sialic acid to neuraminidase digestion and (2) results in altered antigenicity. Many antibodies reactive with sialylated carbohydrate structures, for example STn, SLex, and SLea, recognise only those structures in which sialic acid is non-o-acetylated. In the past, reports focusing upon colorectal neoplasms have commented upon the aberrant expression of tumour-associated markers including STn, SLex and SLea [38-40]. In fact, these structures are expressed in normal colorectal mucins, albeit antigenically silenced through the masking influence of O-acetyl substituents (see above). Spontaneous loss of O- acetyl groups may be explained either by inactivation of O-acetyl transferase (for example by genetic mutation or loss), or to metaplasia (see below) to the small intestinal phenotype (since small intestinal sialic acid is non-o-acetylated). Altered assembly of carbohydrate side-chains Abnormal extension, shortening or branching of carbohydrate side-chains may be explained by a failure to achieve the orderly transcriptional activation and orchestration of glycosyl transferases. Shorter and perhaps sparser carbohydrate side-chains are found in the goblet cell mucins of the lower crypt. Arrested maturation at this level of crypt differentiation would account for the demonstration of hypoglycosylated mucins within neoplasms. An example is provided by the increased proportion of short chain structures such as STn [22]. Midgut metaplasia (see below) affecting neoplasms of the distal large bowel (derived from hindgut) would account for the neoexpression of blood group substances (normally 331

6 expressed by midgut epithelium) consistent with the blood group and secretor status of the affected subject. Increased expression of specific glycosyltransferases has been associated with aberrant blood group expression in gastrointestinal and other malignancies. For example, sialotransferase- 3O (ST3O) is associated with increased expression of SLea [41], fucosyltransferase-4 (FUT4) with Ley [41] and fucosyltransferase-6 (FUT6) with SLex [42]. Altered expression of core proteins If the types of carbohydrate chain vary with the underlying species of core protein, then carbohydrate changes may be secondary to altered core protein expression. In the normal colon a similar distribution of MUC1, the type 1 blood group substances Lex and Ley and the core structure T antigen may be observed within the apical membrane of immature crypt base cells [18]. In colorectal neoplasms, increased expression of MUC1 is accompanied by similarly enhanced and similarly distributed expression of Lex, Ley and T antigen [18]. The same principle is likely to apply to other recently described mucin core proteins. Mechanisms for altered tissue expression of mucin core proteins may occur at the transcriptional, translational or post-translational (e.g. reduced glycosylation) levels. Lineage preference within neoplasms serves as a major but rarely acknowledged determinant of mucin expression. The majority of colorectal cancers show loss of goblet cell mucin, an observation linked to loss of differentiation in a generic sense. Such cancers contain luminal secretions that are eosinophilic with haematoxylin and eosin and positive with diastase periodic acid Schiff staining [18]. A large component of this material is MUC1. Secretory mucin, appearing in haematoxylin and eosin sections as pale, basophilic and wispy material, is only abundant in so-called mucinous cancers (accounting for 10% of all colorectal cancers). Lesser amounts of secretory mucin may be observed in a further 20% of colorectal cancers. In mucinous areas, the malignant epithelium will include goblet cells or columnar cells with apical mucin droplets. MUC2, the principal component of intestinal secretory mucin, can be demonstrated in the cytoplasm of these cells [18]. The presence or absence of secretory mucin in gastrointestinal cancers has as much to do with the lineages represented within the neoplasm as with the grade of differentiation. In fact, the grade or extent of differentiation of a gastrointestinal cancer is determined in diagnostic practice more by glandular architecture and cytology rather than functional characteristics. A further explanation for core protein alteration is metaplasia (see below). Mucin core proteins discussed in this review are summarised in Table 1. Metaplasia Metaplasia is an acquired change affecting adult tissues in which a mature tissue type is replaced by another, generally in a background of chronic inflammation. Metaplasia is distinguished from congenital heterotopia in which there is a developmental basis for the abnormally located tissue. The mechanism underlying metaplasia is not known, but is likely to implicate genes involved in developmental programming. Regardless of the mechanism, the net effect is the switching off of genes expressed in the original tissue type and the switching on of genes that are expressed in the metaplastic tissue. Common examples of metaplasia in the gastrointestinal tract are intestinal metaplasia in the stomach, columnar cell metaplasia in the lower oesophagus (Barrett s oesophagus), and intestinal metaplasia in pancreatic and biliary duct systems including gall bladder. Metaplasia also occurs with benign and malignant neoplasms. Intestinalisation has long been recognised within gastric neoplasms [43]. The observation may be explained either through the origin of the neoplasm in a background of intestinal metaplasia or because metaplasia and neoplasia develop as concurrent processes within nonmetaplastic epithelium. The literature includes well-documented examples of small intestinal metaplasia in colorectal lesions [23,44]. However, since the structural and functional differences between small and large intestine are not as obvious as the differences between stomach and intestine or oesophagus and intestine, small intestinal 332

7 J.Cell.Mol.Med. Vol 5, No 3, 2001 Table 1 Principal mucin core proteins referred to in review. metaplasia has been largely overlooked within the colon. Nevertheless, small intestinal and even gastric metaplasia is a likely mechanism for many of the mucin changes occurring in disorders of growth and differentiation of the colorectum. Disease specific alterations in mucin core protein expression Ulcerative colitis The aetiology of chronic inflammatory bowel disease (ulcerative colitis and Crohn s disease) is still poorly understood. Both ulcerative colitis and Crohn s disease may be conceived as exaggerated immunological reactions triggered by environmental factors in genetically predisposed individuals. The diffuse mucosal inflammation characterising ulcerative colitis is associated with the passage of abundant blood stained mucus per rectum. At the tissue level there is a reduction in the amount of mucus within the goblet cell theca. It is likely that inflammatory mediators are favouring secretion over storage of the mature product. Measurement of MUC2 and mrna levels in ulcerative colitis is problematical since inflamed tissues are oedematous and heavily infiltrated by cells that do not express MUC2. This may explain why some studies show no increase in MUC2 mrna levels [45] when others show increased expression of MUC2 mrna by in situ hybridisation and MUC2 apomucin by immunohistochemistry [46]. Given the facts of increased synthesis of MUC2, accelerated secretion of the mature product and increased epithelial turnover, it would be surprising if there were not, in addition to core protein alterations, evidence of defective glycosylation accompanied by a decrease in sulphation. Indeed, defective glycosylation has been proposed as a principal explanation for the marked increase in MUC2 reactivity using VNTR anti-muc2 antibodies [47]. The question remains, however, as to whether a primary defect in epithelial mucin could be a factor in the aetiology of ulcerative colitis. A defect in the mucosal barrier could predispose the underlying epithelium to injury mediated by luminal pathogens. Podolsky and Isselbacher [48] described a selective reduction of mucin subclass IV defined by discontinuous gradient anion-exchange gel chromatography. Although this finding could not be confirmed [49], subclass IV deficiency was also noted in the unaffected twins of subjects with ulcerative colitis, implying a primary defect with a genetic basis. The observation has not been confirmed or explained by a biochemical mechanism. 333

8 O-acetylation of sialic acid is under genetic control and, as noted above, sialic acid lacking O- acetyl substituents is sensitive to degradation by bacterially derived neuraminidase. Nine percent of the Caucasian population is homozygous for an inactive form of O-acetyl transferase and in these individuals goblet cell mucins throughout the colorectum lack O-acetylated sialic acid. Nevertheless, this phenotype was not overrepresented in subjects with ulcerative colitis, nor was it seen more frequently in colitics with severe disease requiring treatment by surgery [50]. Despite the negative or unconfirmed findings to date, the possibility that a primary or inherited alteration in mucus could underlie the early pathogenesis of inflammatory bowel disease is intriguing and may yet receive support through studies of recently defined mucin core proteins. Linkage to 7q22, the locus of MUC3, MUC11 and MUC12, has been described [51]. Analysis of polymorphisms of variable number tandem repeats (VNTRs) within MUC3 has indicated that rare alleles of this gene may confer susceptibility to ulcerative colitis [52]. Reduced O-acetylation of sialic acid may also be acquired in ulcerative colitis. This occurs within patches of regenerative or hyperplastic mucosa showing epithelial serration reminiscent of the hyperplastic polyp of the colorectum [50]. As noted above, loss of O-acetyl substituents explains increased neuraminidase sensitivity (hence increased neuraminidase-peanut agglutinin binding) and unmasks the sialylated carbohydrate structures SLea, SLex and STn. Hyperplastic patches showing these changes were more frequent in colitics with cancer than in age and sex matched colitics without cancer [50]. It has been suggested that expression of STn in non-dysplastic mucosa is a biomarker for cancer risk [53]. It is likely that epithelial hyperplasia rather than the mucin change per se serves as the more fundamental marker for cancer risk. Aberrent expression of gastric mucins MUC5AC and MUC6 in non-dysplastic mucosa has been associated with dysplasia and cancer in ulcerative colitis [54]. Again, it is likely that these changes will co-segregate with patches of serrated hyperplasia since the same alterations are observed in hyperplastic polyps of the colorectum (see below). Ileo-anal pouch Construction of an ileo-anal pouch or neorectum is a form restorative proctocolectomy offered to subjects with severe ulcerative colitis or familial adenomatous polyposis (FAP). Within a year of the procedure there is a switch from a small intestinal to a colonic mucin pattern. The principal change is to the carbohydrate component as evidenced by increased sulphation and O-acetylation [55,56]. Pouchitis, which occurs mainly in subjects with ulcerative colitis, has been viewed as a return of the disease in a segment of ileum that has undergone an adaptive colonisation. However, complete adaptation to a colonic mucosa is not seen, even within pouches established for over five years. No changes in mrna levels of MUC1, MUC2, MUC3 or MUC4 are seen but there is immunohistochemical reduction of MUC1 and MUC3 compared with normal ileum [57]. Ulcer associated cell lineage The ulcer associated cell lineage (UACL) is a glandular structure found at the borders of chronic ulcers occurring in the gastrointestinal tract [58,59]. The structure has been characterised in most detail in the ileal ulcers of Crohn s disease in which the change is also known as pseudopyloric or pyloric metaplasia. The UACL evolves as a bud from the base of intestinal crypts. This matures to form a coiled acinus with a duct leading to the mucosal surface. The mature structure resembles a gastric glandular unit in which the superficial crypt-like (duct) component includes a proliferative zone generating cells which migrate upwards to populate the epithelial surface while the deep component closely resembles a Brunner s acinus [60]. The UACL expresses peptides implicated in the repair of damaged tissues. These include epidermal growth factor and members of the trefoil factor family (TFF) [61]. Trefoil peptides are potent motogens involved in the coordination of epithelial migration following ulceration [62]. The genes for the TFF peptides TFF1 (ps2), TFF2 (hsp) and TFF3 are clustered together on chromosome 21q22.3 [63]. The peptides are expressed in a site specific distribution within the normal gastrointestinal tract: TFF1 (ps2) in the foveolar and surface epithelium of the stomach, TFF2 (hsp) 334

9 J.Cell.Mol.Med. Vol 5, No 3, 2001 in the distal stomach and Brunner s acini in the duodenum [64] and TFF3 in goblet cells of the small and large intestine [65]. Trefoil peptides are localised within the mucinous granules of mucin secreting cells [60]. Interactions between trefoil peptides and mucins have been described. TFF1 interacts with the von Willebrand factor C cysteine rich domains that are present in the poorly glysolyted non-vntr regions of the gel forming mucins [66]. The UACL expresses three gel-forming mucins: MUC5AC, MUC6 and MUC5B. MUC2 expression is lost [67]. MUC5AC is expressed in distal ducts and surface cells and is co-expressed with TFF1. MUC6 is expressed by acini and proximal ducts and is co-expressed with TFF2. MUC5B is normally secreted by mucous glands of the respiratory tract, salivary gland acini, gall bladder, submucosal glands and ducts of the oesophagus and crypt base goblet cells in the colorectum [68]. Weak and sporadic expression also occurs in the mucous neck cells of the gastric body mucosa [67]. In the UACL, MUC5B is expressed mainly within acini and proximal ducts and to a lesser extent in distal ducts and surface cells. MUC5B therefore shows an overlapping pattern, co-localising with both TFF1 and TFF2. Although MUC2 is lost, TFF3 expression occurs sporadically throughout the UACL with strongest immunoreactivity in surface cells. The overall distribution fits with the type A lectin binding pattern representing MUC5AC and the type B lectin binding pattern representing MUC6, evidence that carbohydrate and mucin core protein expression may, at least in some instances, be closely linked [69]. The UACL is therefore unique among gastrointestinal mucosal phenotypes in expressing three gel forming mucins and all three known trefoil peptides within close proximity. It also appears that specific patterns of mucin and trefoil peptide co-expression are maintained in metaplastic and early neoplastic lesions throughout the gastrointestinal tract. This would require the preservation of coordinated regulation between chromosome 11p15.5 and 21q22.3 regions [67]. The theme of coordinated gel forming mucin expression has been noted in ulcerative colitis and will be presented in subsequent accounts of intestinal metaplasia of gastric mucosa, Barrett s oesophagus and serrated polyps of the colorectum. Helicobacter pylori gastritis In subjects with peptic ulcer both the thickness and hydrophobicity of the mucous gel layer are known to be reduced [70]. Since H. pylori is the most important cause of peptic ulcer disease, it is evident that these changes must be attributable to infection by this organism, either directly or indirectly. The changes could be due to decreased synthesis, increased degradation or an alteration in mucin type or structure. Degradation of mucus by proteases or sulphatases has not been demonstrated [71]. Urease production increases ph by generating ammonia. This leads to partial destabilisation of the mucus gel layer but the barrier thickness is not compromised [72]. Further evidence against H. pylori mediated degradation is the finding that gastric mucins from subjects with and without infection show no differences in viscous properties [73]. Attention has therefore focused on changes in the synthesis of gastric mucins in subjects with H. pylori gastritis. MUC5AC expression was decreased whereas MUC6 expression was increased within the surface epithelium. Additionally, carbohydrate structures expressing Lex and showing paradoxical ConA staining were expressed by surface mucous cells [74]. These effects were reversed following eradication of H. pylori. Cell culture experiments indicate that H. pylori inhibits the synthesis of mucin core structures MUC5AC and MUC1 but does not influence glycosylation, degradation or secretion of mucin [75]. It is likely that H. pylori within the mucous gel layer are non-pathogenic. Adhesion of the organism to receptors in the apical surface of gastric epithelium triggers the induction of proinflammatory chemokines such as IL-8 [76]. The presence of MUC5AC in the mucous gel layer and MUC1 in the glycocalyx of surface epithelium may impede the adherence of H. pylori. A vicious cycle may therefore be set up in which H. pylori downregulates biosynthesis of MUC5AC and MUC1, this in turn facilitates the adhesion of H. pylori to surface epithelium, and mucin synthesis is further compromised [75]. Intestinal metaplasia of gastric mucosa Metaplasia or the replacement of indigenous tissue by a well differentiated tissue normally found in a 335

10 different anatomical location is an acquired change brought about by longstanding inflammation. The change is assumed to be an adaptive process and to provide an improved defence against the underlying injurious factor or factors. This is well evidenced by intestinal metaplasia of gastric mucosa that develops in subjects with chronic infection by Helicobacter pylori. H. pylori is generally not apparent in foci of intestinal metaplasia. Moreover, well developed islands of intestinal metaplasia contain fewer lamina propria inflammatory cells than adjacent non-intestinalised mucosa[77]. Whether the change is brought about by genetic or epigenetic mechanisms is unknown. Mutations have been demonstrated in intestinal metaplasia but may not necessarily be causative. Intestinal metaplasia of gastric mucosa may be morphologically, ultrastructurally and funtionally indistinguishable from normal small intestinal mucosa, though with relatively poorly formed villi. The cellular constituents include goblet cells secreting non-sulphated acid mucins (sialomucins), absorptive columnar cells with a well developed brush border expressing digestive enzymes and Paneth cells within the crypt base. Variant forms occur also. In these columnar cells are mucus secreting and lack a brush border and brush border enzymes while Paneth cells are infrequent or absent. The resulting tissue has no normal counterpart and has therefore been described as incomplete intestinal metaplasia to distinguish it from the complete form (type I) that is virtually identical to small intestine. Incomplete intestinal metaplasia has in turn been classified into two subtypes. In the first (type II intestinal metaplasia) the columnar cells secrete mainly neutral mucins like normal gastric foveolar cells. The overall appearance may therefore be described as the presence of scattered intestinal goblet cells within gastric crypts or a gastric and intestinal hybrid epithelium. The columnar cells may not necessarily show an exact likeness to gastric columnar cells. There may be relatively little mucus in the apical theca and alcian blue/pas staining may show a degree of acidity. The immature crypt base columnar cells in normal intestinal epithelium secrete small amounts of mainly neutral mucin that is lost as the cells differentiate into absorptive enterocytes. Therefore type II intestinal metaplasia could be conceived as an immature form of small intestinal metaplasia with arrested enterocytic maturation. Regardless of these considerations, type II intestinal metaplasia is regarded as a transitional or transient form that will ultimately be replaced by complete or type I intestinal metaplasia [78,79]. Type II intestinal metaplasia occurs in bile reflux gastritis as well as H. pylori gastritis [78]. A second incomplete subtype is characterised by the secretion of sulphated acid mucin (type III intestinal metaplasia). This subtype has been termed colonic metaplasia by some though the crypts are not morphologically identical to normal colonic crypts. Additionally, sulphomucin is not restricted to the colonic mucosa but is also found in biliary epithelium, pancreatic ducts and submucous oesophageal glands. Furthermore, type II and III intestinal metaplasia are indistinguishable morphologically. Mucin core protein expression patterns clarify the relationship between the three types of intestinal metaplasia. Type I intestinal metaplasia displays little or no expression of MUC1 or the gastric mucins MUC5AC or MUC6 whereas intestinal mucins MUC2 and MUC4 are conspicuous in goblet cells [19,79]. This pattern reinforces the close resemblance of type I intestinal metaplasia to normal small intestinal mucosa. Types II and III intestinal metaplasia share an identical pattern. Gastric mucin core proteins MUC1 and MUC5AC are expressed in both goblet cells and columnar cells. MUC6 expression occurs in lower crypt and glandular epithelium but not to the extent observed in normal gastric mucosa (Fig. 3). MUC2 and MUC4 are expressed in goblet cells [19,79]. These findings do not support the suggestion that type III intestinal metaplasia is colonic, but rather reinforce the close identity of types II and III apart from a single difference in glycosylation, namely the presence of sulphated sugars in type III intestinal metaplasia. With regard to the sequential evolution of the various subtypes, type II, as noted above, may represent an unstable form of intestinal metaplasia that subsequently transforms into complete IM (type I). Alternatively, type II may acquire stability and persist or it may transform and stabilise as type III. The factors determining these divergent pathways are unknown and there is no agreed explanation for the association between type III 336

11 J.Cell.Mol.Med. Vol 5, No 3, 2001 intestinal metaplasia and gastric cancer [80-82]. It is possible that type III intestinal metaplasia is a marker for chronic and persistent gastric injury. This is supported by the fact that type III intestinal metaplasia is found in stomachs with very extensive chronic gastritis [83]. Additionally, sulphomucin may be particularly important in providing cytoprotection. It is interesting that Helicobacter pylori can adhere to areas of incomplete but not complete intestinal metaplasia of gastric mucosa [84]. It is possible that type III intestinal metaplasia provides simultaneous protection from a dual assault by Helicobacter pylori and refluxing bile. Barrett s oesophagus This term refers to an acquired change characterised by replacement of the squamous epithelium of the lower oesophagus by a columnar epithelium and occurring in subjects with gastroesophageal reflux disease. The mucosa represents a complex mixture of cell types and architectural patterns found in the stomach and small intestine [85]. While the various types of epithelium lining Barrett s oesophagus may not be arranged in definite zonal distribution as thought originally [86], it is nevertheless possible to observe discrete areas of gastric metaplasia and intestinal metaplasia. Gastric metaplasia (also termed cardiac type) comprises surface and foveolar epithelium lined by columnar mucous cells and cardiac or pyloric-type mucous glands. Chief and parietal cells may be present but are rarely conspicuous. Intestinal metaplasia is typically incomplete, comprising intestinal goblet cells and gastric foveolar type columnar mucous cells (also termed specialised type epithelium). Absorptive cells and Paneth cells are inconspicuous. Intestinalised Barrett s epithelium is more likely to show a villous architecture and is recognised as the precursor of dysplasia and adenocarcinoma [87]. Although some restrict the diagnosis of Barrett s oesophagus to cases with intestinal metaplasia [88], gastric or cardiac metaplasia is an abnormal finding and a marker for gastroesophageal reflux disease [89]. Barrett s mucosa is recognised in haematoxylin and eosin stained sections not only by the presence of goblet cells but by distinctive architectural changes including villosity and crypt architectural Fig. 3 Expression of MUC6 (CLH5, Research Diagnostics Inc., Flanders, NJ, U.S.A. [79,115]) within mucous neck cells of gastric mucosa. abnormalities (loss of parallelism, tortuosity, branching and varying degrees of atrophy). Squamous islands and ducts of submucosal eosophageal glands (when present) confirm the site of biopsied columnar epithelium. The lamina propria may be replaced by fibromuscular tisue and splitting of the muscularis mucosae into two layers is often observed in surgical specimens [90]. Barrett s mucosa is therefore more than a simple epithelial change dominated by goblet cell metaplasia. Goblet cells in gastroesophageal biopsies are not pathognomonic for the condition. These cells may be seen in intestinal metaplasia of the cardia associated with Helicobacter pylori gastritis. In this situation, however, intestinal metaplasia is likely to be complete [91]. 337

12 A B C 338

13 J.Cell.Mol.Med. Vol 5, No 3, 2001 D E Fig. 4 (A) Barrett's oesophagus (haematoxylin and eosin) with expression of (B) MUC2 (4F1, [134]), (C) MUC4 (M4.275, [135]), (D) MUC5AC (M1, NeoMarkers, Fremont, CA, U.S.A. [122]) and (E) MUC6 (CLH5, Research Diagnostics Inc., Flanders, NJ, U.S.A. [79,115]). There is conspicuous expression of both MUC2 and MUC5AC by crypt and surface epithelium whereas MUC6 is secreted mainly by deeper mucous glands. Studies of mucin gene expression in Barrett s oesophagus reveal the expression of gastric mucins MUC5AC and MUC6 as well as intestinal mucins MUC2, MUC3 and MUC4 by intestinalised epithelium [92-94] (Fig. 4 A to E). These findings confirm the incomplete nature of intestinalisation in Barrett s oesophagus. A similar mucin pattern is observed in incomplete intestinal metaplasia of gastric mucosa. However, whereas MUC5AC appears to be restricted to columnar cells in Barrett s oesophagus [92], it is expressed by both goblet cells and columnar cells in incomplete intestinal metaplasia of the stomach. As in incomplete intestinal metaplasia of gastric mucosa, subtypes of Barrett s epithelium secreting sulphomucin have been associated with neoplastic progression [95,96]. This finding occurs with high frequency in Barrett s epithelium and has not translated into a practical marker of increased risk. 339

14 It is unclear whether Barrett s mucosa develops as an upward extension of a metaplastic gastric mucosa or represents a transformation of stem cells within the ducts of oesophageal mucous glands [97]. The presence of MUC5B in oesophageal glands but not in Barrett s oesophagus fits the first hypothesis [92]. The finding of cytokeratin (CK) 7 within both Barrett s epithelium and oesophageal glands but not to the same extent in intestinal metaplasia of gastric mucosa supports the second hypothesis [98]. The possibility that reflux injury in the lower oesophagus could be caused by defective oesophageal mucin production has been considered [99]. The main secretory mucin MUC5B is produced by the submucous oesophageal glands but membrane bound mucins MUC1 and MUC4 are expressed by oesophageal squamous epithelium and may have an important cytoprotective role [92]. The mixed gastric and intestinal phenotype characterising intestinalised mucosa could represent an adaptation to cell injury mediated by the combination of gastric acid and bile [92]. A cytoprotective role is supported by the demonstration of expression of trefoil peptides TFF1 and TFF2 by Barrett s epithelium [93,100]. Progression of Barrett s epithelium to dysplasia and malignancy is usually accompanied by downregulation of secretory mucins MUC2, MUC5AC and MUC6, similar to cancer of stomach and colon [94]. Membrane bound mucins MUC1 and MUC4 may show upregulation, as in the other regions of the gastrointestinal tract, whereas the membrane bound mucin MUC3 is downregulated [92]. Epigenetic mechanisms such as methylation may account for the early alterations of phenotype in Barrett s oesophagus [101], whereas mutation and loss of genes will underlie subsequent neoplastic transformation. Gastric neoplasia The study of gastric neoplasia presents a number of difficulties: (1) Advanced cancers are morphologically heterogeneous, (2) Different types of classification are in use, (3) Histogenesis is poorly understood. These points may be contrasted with the relatively uniform histology of colorectal cancer and the conceptually straightforward origin of colorectal cancer from a pre-existing adenoma. To simplify matters, many classifications of gastric recognise two main types. Differentiated cancers comprise glands and/or papillary structures. Undifferentiated (poorly differentiated) cancers lack such structures but may occur in two distinct forms. In the first, cells are discohesive or isolated and may contain intracytoplasmic mucin (signet ring cells). In the second, cells form solid sheets or masses. However, there is a sizeable mixed category [102]. The classification of Laurén into intestinal and diffuse and combining both histology and pattern of growth shows partial overlap with the differentiated/undifferentiated classification [43]. Again there will be cases that cannot be easily accommodated, for example cancers formed of glands showing diffuse spread, undifferentiated cancers with a highly circumscribed growth pattern and mixed pattern types. Mucinous cancers may be either differentiated or undifferentiated (signet-ring cell). In addition to classifications that are primarily morphologically based are those that reflect histogenetic pathways. The intestinal-type cancer of Laurén is more likely to be associated with intestinal metaplasia of gastric mucosa than the diffuse type. It would be simple and satisfying if intestinal-type cancers arose within intestinal metaplasia (making them exact counterparts of cancers arising in the intestinal tract) and undifferentiated or diffuse-type cancers developed within non-intestinalised gastric mucosa. However, the truth is not this straightforward. As noted above, intestinal metaplasia is often a hybrid lesion combining both gastric and intestinal phenotypes. The incomplete form of intestinal metaplasia appears to have a more selective association with gastric cancer than complete IM which closely resembles normal small intestinal mucosa. Furthermore, the demonstration of intestinal phenotypes within a gastric cancer does not prove an origin within intestinal metaplasia. Intestinal differentiation could occur subsequent to the establishment of a neoplastic clone within nonintestinalised gastric mucosa [103,104]. New insights into the histogenesis of gastric cancer have been achieved with the aid of lineagespecific probes to mucin core proteins. Early gastric cancers (limited to mucosa or submucosa) are particularly informative in bridging the steps between earlier mucosal changes and advanced 340

15 J.Cell.Mol.Med. Vol 5, No 3, 2001 gastric cancer. Other precancerous lesions are more problematical. Most examples of gastric dysplasia (documented in literature from the West) are accommodated (in the Japanese literature) by three alternative diagnoses: (1) regenerative change, (2) adenoma and (3) well differentiated mucosal adenocarcinoma [105,106]. In a study of early gastric cancer, Kabashima et al. employed MUC2 as a marker of intestinal goblet cell differentiation and MUC5AC as a marker of the gastric foveolar phenotype [107]. They also used the more traditional technique of paradoxical concanavalin A (ConA) as a marker of pyloric gland differentiation and CD10 (CALLA) as a small intestinal marker. Differentiation of early gastric cancer and background intestinal metaplasia was classified according to particular combinations of these markers. Complete (small) intestinal differentiation was diagnosed when CD10 was positive, MUC5AC was negative and MUC2 (not invariably) was positive. Incomplete intestinal differentiation implied a combination of intestinal (MUC2) and gastric (MUC5AC) differentiation, with or without CD10 expression. Gastric differentiation implied absence of intestinal marker expression. Intestinal metaplasia in the background mucosa was classified as complete or incomplete using similar criteria. The spectrum of complete intestinal, incomplete intestinal or gastric differentiation varied according to whether early gastric cancers were multiple or solitary. Multiple early gastric cancers were more likely to show gastric differentiation. Overall, 52% of early gastric cancers showed incomplete intestinal differentiation, 39% showed gastric differentiation while only 8% showed complete intestinal differentiation. Intestinalisation of background mucosa was found in 81% of cases but was significantly more likely to be incomplete in subjects with multiple as opposed to solitary early gastric cancers. Wang et al. (2000) also demonstrated gastric mucinous phenotypes in the majority of early gastric cancers, regardless of whether these were typed as intestinal or diffuse. By contrast, adenomas had a mainly pure intestinal phenotype [108]. Endoh et al (2000) describe a form of early or mucosal carcinoma that mimics normal gastric foveolar epithelium morphologically and in terms of mucin expression. This type of pure gastric cancer often showed DNA microsatellite instability and hypermethylation of the promoter region of the DNA mismatch repair gene [109]. Gastric cancers showing exclusively intestinal differentiation are uncommon. Endoh et al highlighted an early cancer showing small intestinal metaplasia evidenced by expression of CD10 and MUC2 and closely mimicking complete intestinal metaplasia of gastric mucosa [110]. These workers recognise two uncommon and extreme subtypes showing either exclusive foveolar or exclusive small intestinal differentiation and distinguish these from the more usual gastric cancers showing a mixed phenotype [111]. A similar approach applied to the study of advanced gastric cancer was employed by Machado et al [112]. They utilised MUC6 as a marker of gastric mucous neck cells and pyloric glands in place of ConA and included trefoil peptides in their panel of gastric phenotypes. The study by Machado et al. illustrates the high frequency of gastric and incomplete gastric differentiation in advanced gastric cancers, though a non-gastric phenotype was more frequent in intestinal-type than diffuse cancer. Expression of the intestinal mucin MUC2 was relatively infrequent. However MUC2 expression is known to be lost in many colorectal cancers [113] and lack of MUC2 expression therefore does not exclude a background origin within intestinalised mucosa. Although MUC2 expression undoubtedly occurs in intestinal-type gastric cancer [112,114], most intestinal cancers also express gastric mucins. Additionally, diffuse cancers may express MUC2, either alone or in combination with gastric mucin. Marked MUC2 expression is a feature of mucinous carcinoma independent of intestinal or diffuse type [115]. The preceding observations expose the limitations of classical carbohydrate histochemistry. Acid mucins characterise intestinal goblet cells and are conspicuous in intestinal-type gastric cancer whereas neutral mucins predominate in normal gastric mucosa and diffuse-type gastric cancer. However, intestinal and diffuse cancers express both gastric and intestinal core mucin peptides. Incomplete intestinal metaplasia expresses mucins found in normal gastric mucosa (MUC1, MUC5AC, MUC6) and intestinal mucosa (MUC2, MUC4) [19,79] and shows additional changes that are not present in normal gastric 341

16 epithelium but are characteristic of gastric cancer. Of the principal subtypes of intestinal metaplasia, the complete type establishes a normal pattern of small intestinal differentiation and is regarded as having little malignant potential. By contrast, incomplete IM (particularly a variant secreting sulphomucin) has been associated with a higher risk of malignant change. The demonstration of a mixed gastric and intestinal phenotype in the majority of gastric cancers is consistent with a precancerous role for incomplete intestinal metaplasia. The use of probes to core mucin peptides shows that the concept of intestinal-type gastric cancer is flawed at the histogenetic level. Nevertheless, it is unlikely that the popular intestinal versus diffuse classification of Laurén will disappear rapidly [43]. The preceding insights also reinforce the view that incomplete intestinal differentiation of gastric epithelium is a biomarker for mucosal instability. While incomplete intestinal metaplasia may be of limited value as a risk factor for the individual patient, an understanding of the underlying molecular mechanisms and related lifestyle factors may translate into novel strategies for gastric cancer prevention. Colorectal polyps and cancer The two most frequent types of colorectal polyp are the hyperplastic polyp and adenoma. The latter is regarded as the more important of the two, serving as the main precursor of colorectal cancer. Nevertheless, only a small proportion of adenomas, perhaps 5%, will transform into cancer. Adenomas are classified by size, grade of epithelial dysplasia and architecture (tubular, tubulovillous and villous). Adenomas with the highest potential for malignant change are large, show high grade epithelial dysplasia and include a villous component. Dysplasia is assessed on the extent of cytological atypia, glandular abnormality and level of differentiation. Serrated adenomas are a distinct subset with architectural and cytological features reminiscent of hyperplastic polyp. These include a serrated or saw-tooth crypt configuration, cytoplasmic eosinophilia and a relatively low nuclear:cytoplasmic ratio. There is now evidence that hyperplastic polyp and serrated adenoma are histogenetically related with mixed polyps providing a bridge between the two [116]. Studies of mucin core protein expression in colorectal polyps and cancers have yielded interesting but sometimes discrepant results. MUC1 expression is higher with increasing grades of epithelial dysplasia in tubular adenomas whereas secretory mucin MUC2 is reduced [117]. This combination of changes can be at least in part explained on the basis of maturation arrest as evidenced by failure of the normal switch from proliferation to differentiation and maturation. MUC1 expression is most obvious within the proliferative zone in normal mucosa whereas mature goblet cells secreting MUC2 are mainly represented in the maturation zone. Adenomas may therefore be considered to recapitulate the structure and function of the normal crypt base. An alternative explanation for the domination of MUC1 in adenomas is lineage preference in favour of columnar cells over goblet cells culminating in the near absence of goblet cells in adenomas showing high grade dysplasia [18]. Markedly increased MUC2 expression may be observed in villous adenomas [118] and in mucinous carcinomas [113,118]. This observation suggests that villous adenoma and mucinous carcinoma do not originate within tubular adenomas but represent an independent neoplastic pathway. This suggestion is reinforced by the demonstration of aberrant expression of gastric mucin MUC5AC in villous adenomas [37,119]. Some describe MUC5AC mrna and expression of MUC5AC protein in colorectal tubular adenomas [120], but others, using antibodies to the C-terminal region of the molecule, find little or absent MUC5AC expression in tubular adenomas [119,121]. Gastric mucin MUC6 has also been described in colorectal neoplasms [120]. Of particular interest is the markedly increased expression of both MUC2 and MUC5AC in hyperplastic polyps and serrated adenomas [119,121] (Fig. 5A to C). The M1 antigens [122] show a similar distribution and are considered to be derived from the MUC5AC gene. We have also observed trace expression of gastric mucin MUC6 in the crypt base region of hyperplastic polyps (unpublished observations). Given accumulating evidence favouring serrated polyps as the usual precursors of sporadic colorectal cancers showing DNA microsatellite instability-high (MSI-H) [123] it is interesting that most examples of the latter 342

17 J.Cell.Mol.Med. Vol 5, No 3, 2001 Fig. 5 (A) Serrated adenoma of colorectum (haematoxylin and eosin) with increased expression of both (B) MUC2 (4F1, [134]) and (C) MUC5AC (M1, NeoMarkers, Fremont, CA, U.S.A., [79,115]). show increased expression of both MUC2 and MUC5AC regardless of whether they have been typed as mucinous or non-mucinous carcinomas [124]. Preferential goblet cell differentiation may be due to dysregulation of the TGFbeta signalling pathway. Excess of TGFbeta depresses columnar cell proliferation but goblet cells have an abnormally glycosylated TGF RIII receptor and may therefore be resistant to growth downregulation [125]. A histogenetic relationship between hyperplastic polyps and tubulovillous or villous adenoma [126] may account for the sharing of the mucinous phenotype MUC2+/MUC5AC+ by these lesions. The preceding findings challenge the conventional typing of colorectal neoplasms and invite a more lineage-based approach. Antibodies to 343