Coeliac Disease: An Update on Facts and Questions Based on the 10th International Symposium on Coeliac Disease

Defined as an enteropathy induced by gluten in genetically predisposed patients, coeliac disease (CD) has raised considerable interest in the past few years with recent epidemiologic data showing a high incidence of the disease, a broadening of the clinical spectrum, and the unravelling of the pathogenesis of the disease. This review focuses on the recent advances and remaining questions that were discussed during the 10th International Symposium on Coeliac Disease.

UPDATE ON CLINICAL AND EPIDEMIOLOGIC DATA

The Coeliac Iceberg

Clinically Overt Coeliac Disease

Recent studies have emphasized the extreme variability of presentation and evolution of CD. In 1996, a prospective study of the Groupe Francophone d'Hepatologie, Gastro-Enterologie et Nutrition Pediatriques (GFGHNP) in France indicated that the diagnosis of CD in children is still mainly made in toddlers 12 to 24 months of age who have moderate symptoms related to malabsorption (anorexia, diarrhea, and growth arrest). The relatively uniform clinical presentation in children contrasts with the wide variability in nature and intensity of clinical symptoms in adults. Iron deficiency anemia is the symptom considered classic in revealing the presence of CD in adults. However, the diagnosis is also suggested by an array of digestive and extradigestive symptoms, including those of irritable bowel syndrome, abnormal liver chemistries (rise in aminotransferase), gynecologic symptoms (sterility, repeated abortions), neurologic/psychiatric symptoms (epilepsy, ataxia, depression, migraine), and symptoms related to associated phenomena particularly autoimmunity (^1-3).

The variable clinical presentation of CD is further emphasized by recent studies in the United States, where the disease has only recently attracted much attention. An epidemiologic study in Olmsted County, Minnesota is interesting because precise medical records kept for the past 50 years were used to determine the presentation of CD in a population almost entirely of Scandinavian origin. The annual incidence of CD has increased from 0.8/100,000 in the 1960s to 9.4/100,000 in 2001 and now matches the incidence of overt disease in European reports. We suspect that the increased incidence is a result of growing awareness of CD and the wider use of CD related screening serology. This study also points out that presenting clinical symptoms have changed during the past 50 years. Diarrhea is now reported less frequently. Patients present with higher body mass indices and even with obesity. Surprisingly, pediatric cases have always been rare in this community. The population incidence increases with age, with a mean age at presentation of approximately 45 years. Life expectancy of patients with CD in this study was not different from that expected in the general population despite the late age at diagnosis/onset (⁴). The differences in pediatric data between in the Minnesota study and similar studies in Europe cannot be accounted for by racial or genetic differences. They might be attributable to distinct infant rearing and feeding practices, as suggested by a recent epidemiologic study in Sweden. An increase in the number of cases in Swedish children younger than 2.5 years and born in 1992-1993 coincided with changes in the diet; the increase subsided in children born after 1996, when gluten was introduced more gradually during prolonged breastfeeding (⁵). Follow-up studies should help to delineate whether this decrease in pediatric cases will result in decreased incidence or delayed onset with milder symptoms.

Silent and Latent Coeliac Disease

Serologic screening of different populations has further unravelled the heterogeneity of the disease. Silent disease characterized by positive serology and villous atrophy with no or very mild symptoms is approximately 7 to 15 times more common than symptomatic CD in adolescents and children, depending on epidemiologic studies. A recent well-controlled study in Finland in students 7 to 16 years old suggests a seroprevalence of between 1:99 (cases confirmed by biopsy) and 1:67 (when all patients with anti-endomysium antibodies and DQ2 positivity were included, with or without confirmation by biopsy) (⁶). These figures are comparable with those obtained in a recent multicenter study in the United States (1:133) (⁷). In France, the prevalence of silent disease has been estimated only in adults. The estimates, which range between 1:280 (Cohort Monica North of France) and 1:500 (Cohort su.vi.max), appear surprisingly low, but it remains unclear whether this lower incidence is related to bias in the screened population or to a true variability in incidence. In this regard, it is interesting that a recent study suggests a prevalence of 1.2% in an English population aged 45 to 75 years (⁸), but this study was based only on the detection of anti-endomysium antibodies without control biopsies.

A recent Finnish study based on two serologic screens in 1994 and 2001 of 3654 subjects aged 7 to 16 years provides further indication of the clinical heterogeneity of CD (⁶). Only 10 of 56 subjects with positive serology in 1994 developed overt symptoms of CD before the second screen in 2001. Twenty-seven subjects with positive serology in 1994 continued to have no symptoms in 2001, despite a flat mucosal biopsy, and 10 cases were symptom free with normal biopsies. Interestingly, two subjects with positive antibodies in 1994 and 2001 and at risk for CD because of HLA DQ2 haplotype had normal mucosa, but both had increased epithelial expression of HLA-DR suggestive of ongoing inflammation and one had very high counts of TcRγδ intra epithelial lymphocytes (IELs) (⁶). These patients might represent cases of latent disease susceptible to evolving into overt CD after several years, as previously described (⁹). Finally, five patients with HLA-DQ2 and positive antibodies in 1994 had negative ones in 2001. Their intestinal biopsies were normal, but all had increased HLA-DR expression, and four of five had markedly increased numbers of TcRγδ+ IELs (⁶). This latter finding might indicate a variation in the natural history of CD in which gluten sensitivity fluctuates with time. This possibility has been raised previously in a subset of patients who had shown no adverse clinical effects of dietary gluten challenge or who had stopped the gluten free diet as teenagers. A fraction of these patients, whose CD had become silent, had normal intestinal mucosal architecture on re-biopsy. Their disease had apparently returned to a state of latency. Similar to the cases with latent disease described by Mäki et al., these patients usually continued to have high counts of TcRγδ IELs in intestinal mucosal biopsies (¹⁰).

The spectrum of CD was further broadened by a recent study from Germany, which suggested that some patients with irritable bowel syndrome, no villous atrophy, and no circulating antibodies might have latent gluten intolerance because they had the HLA-DQ2 or DQ8 haplotypes, had antigliadin and antitransglutaminase antibodies in the duodenal fluid, and showed a dramatic clinical improvement when on a gluten-free diet (¹¹). However, the link with CD in these cases remains to be definitively ascertained.

The variability in presentation led Anne Ferguson to describe CD as the coeliac iceberg, in which overt disease is only the emerging peak. The genetic and environmental factors controlling the expression of CD and passage from latent to overt disease remain unknown. Environmental factors other than gluten with an impact on the local immune response likely are involved. The discovery of the immersed part of the coeliac iceberg has transformed the status of CD, long considered a rare disease, particularly in adults, to that of a health problem not yet fully recognized. Nevertheless, the heterogeneity in the presentation points out that all patients with CD are not alike. The wide clinical spectrum might reflect a large inter-individual level of gluten sensitivity. Thus, one important challenge for the future remains to precisely assess the risk of complications in each subset of patients.

Update on the Risks of CD to Adults

Since the late 1980s, a major focus of epidemiological research has been on the risk of malignancy associated with CD. Patients with CD or dermatitis herpetiformis (DH) have been reported to have a 10-fold increased risk for certain carcinomas of the gastrointestinal tract or a 40-fold increased risk for non-Hodgkin's lymphoma (NHL), respectively (^12,13). In 2001, Corrao et al. (¹⁴) suggested there was 70-fold increased risk for dying of NHL among patients with CD, NHL representing two thirds of all observed cancer cases. Two more recent prospective Italian and European studies assessing the prevalence of CD (overt or latent) in patients with NHL as compared with control subjects indicate that the risk of lymphoma is much less than initially thought. In the Italian study, patients with CD had a 3.1 increased risk of NHL (¹⁵). In the European study, which included 10 countries from May 1998 through April 2001, the prevalence of CD was increased 2.6 fold in 1446 patients with NHL compared with 9659 control subjects. Cases mainly involved clinically overt CD, with silent CD being more common in control subjects than in patients with NHL. However, when only small bowel lymphomas were considered, the prevalence of CD was much higher (odds ratio: 11.8). The increased odds ratio was particularly great when considering enteropathy-type intestinal T lymphomas (EITL) (odds ratio: 28). These results indicate that intestinal T lymphomas and chiefly EITL are highly characteristic of, if not specific for, CD. However, this type of lymphoma is rare, which explains why the overall incidence of lymphoma is still only moderate in CD (L. Mearin et al., unpublished data).

These data are confirmed by a vast retrospective Swedish study of 12,000 hospitalized patients with either CD or DH between 1964 and 1994 with a mean follow-up of 10 years (range, 0 to 32 years) (¹⁶). Overall cancer risk was only modestly increased (1.3 fold, 95% CI = 1.2-1.5) in CD. Several digestive tract carcinomas occurred in excess, including oropharyngeal (Standardized Incidence Ratio [SIR]: 2.3), esophageal (SIR: 4.2) and small intestinal carcinomas (SIR: 10), as well as primary liver cancers (SIR: 2.7), whereas the risk of breast cancer was decreased (SIR: 0.3). The risk of lymphoma was increased by 5.9 (95% CI = 4.3-7.9) in CD. The risk of cancer disappeared after a 10-year follow-up. Accordingly, individuals first hospitalized as children or adolescents had no detectable increased cancer risk during their follow-up evaluation. The declining risk of malignancies with increased duration of follow-up and thus with the length of gluten-free diet confirms the results of a previous study, which indicated that a strict gluten-free diet for 5 years reduced the risk of cancer to that of the general population (¹²). Interestingly, the relative risk of lymphoma decreased with successive periods of diagnosis, from SIR = 12 during 1970 to 1979 to 3.5 during 1990 to 1995. Potential biases include increased diagnosis of silent or mild CD with time, but these observations also might be compatible with a beneficial effect of increasing awareness and adherence to a gluten-free diet.

Another type of lymphomatous complication recently described in CD and not considered in current epidemiologic studies is clonal coeliac refractory sprue. Refractory sprue is a rare disorder in adults defined by villous atrophy resistant to a strict gluten-free diet for more than 6 months. In a national cooperative study, we have identified 58 patients with refractory sprue, 50 of whom had a coeliac-related disorder. Among these 50 patients, 80% had a massive infiltration of their epithelium by abnormal lymphocytes characterized by a distinctive phenotype (lacking surface expression of CD3, T-cell receptors and generally CD8 but containing intracellular CD3ϵ), and by the presence of clonal rearrangements of the T-cell receptor γ chain (^17,18). In contrast with the enteropathy type intestinal T cell lymphoma lymphocytes, IEL remain p53 negative, and do not divide actively in situ. However, their malignant nature is indicated by (1) their ability to progressively invade the whole epithelial length, from the stomach to the colon and, secondarily in a subset of patients, to invade the lamina propria, the blood (¹⁹), and occasionally the skin and chest; (2) the presence of chromosomal abnormalities, including a recurrent trisomy 1q (²⁰); and (3) their evolution into an overt high-grade lymphoma sharing the same clonality in more than 40% of cases. Thus, clonal refractory coeliac sprue (RCS) can be considered a cryptic intraepithelial T-cell lymphoma. Its study points to the profound alteration of intraepithelial lymphocyte homeostasis in CD and confirms that the tendency for lymphoma development in CD primarily arises from the intraepithelial lymphoid compartment [reviewed in Cerf-Bensussan et al. (²¹)]. In our experience, RCS is rare but more common than overt EITL. Its exact prevalence remains to be evaluated.

A second complication of CD is autoimmune disease, which has a prevalence of approximately 20% in adult CD patients. Ventura et al. (²²) suggested an increased frequency of autoimmune disease with increased age at diagnosis and lack of diet therapy, conclusions challenged by more recent studies (^23,24). Thus, the role of the diet in the prevention of autoimmune complications awaits additional studies.

The third and most common complication of CD is osteopenia. As many as 70% of patients with untreated CD have osteopenia, and the prevalence of osteopenia increases with age at diagnosis. A gluten-free diet allows marked improvement within 1 year and total and persistent recovery when started in children (²⁵). In contrast, full recovery is difficult to obtain in adults (²⁶). Osteopenia is less common in patients with silent CD, but prevalence remains 27% in the series presented by Bai and as high as 40% in other series (²⁶). A key unanswered question is the exact functional consequences of osteopenia for the patients. An increased risk of fractures was observed in patients with overt CD in one study (²⁶) but not in a more recent one (²⁷). It remains unclear whether any risk exists in patients with silent disease.

Finally, a high prevalence of silent CD has been reported recently in women with recurrent spontaneous abortions and intrauterine fetal growth retardation, underlining the need to test for CD in these women (²⁸). The effect of the diet on the recovery of fertility does not seem to have been established.

In conclusion, results of these studies indicate that the risk of NHL and particularly of T lymphomas is much weaker than initially estimated. Furthermore, the European study suggests that this risk is particularly weak in silent CD. The high prevalence of autoimmune disease and osteopenia is confirmed, but the functional consequences of osteopenia in silent CD remain unclear, as does the protective effect of gluten free diet against autoimmune diseases. Altogether these data should lead to a reconsideration of the justification for mass screening. An alternative strategy might be screening at-risk individuals (such as families of patients with CD, patients with autoimmune disease, patients with trisomy 21, and patients with unexplained atypical symptoms).

Update on Serology Screening

The new techniques for antibody detection (detection of antigliadin isotypes by the ELISA technique, of anti-endomysium antibodies by immunofluorescence) and their application to mass screening has radically changed the epidemiology of CD by revealing the high incidence of silent CD. The recent identification of the autoantigen recognized by anti-endomysium (alternatively called antireticulin) antibodies as tissue transglutaminase (TG2), a ubiquitous enzyme participating in matrix integrity (²⁹), has allowed the development of ELISA techniques that ultimately could be simpler and more reproducible than immunofluorescence (which requires a trained observer). The first results are promising. A working group of 13 European laboratories funded by ESPGHAN and the European Medical Research Council has started to improve standardization by establishing standard curves on reference sera and protocols for calibration of quality controls. This collaborative approach has shown that the IgA anti-endomysium antibody with a sensitivity of 90%, specificity of 99%, and reproducibility of 93% currently remains the gold standard. The IgA antihuman TG2 is slightly less reliable (sensitivity 93%, specificity 95%, reproducibility 83%), and the anti-IgG and IgA antigliadin antibodies are the least reliable. However, even the IgA anti-endomysium antibodies leave undetected 10% to 15% of CD cases, including those in patients with IgA deficiency (2% of patients with CD). ELISA tests based on the detection of guinea-pig TG2 are responsible for many false-positive results, particularly in at-risk individuals with autoimmune diseases or liver disease and in patients with other inflammatory bowel diseases (³⁰). Finally, IgG and IgM transglutaminase antibodies seem to lack sensitivity and specificity.

Update on the Pathogenesis of Coeliac Disease

Genetics

Coeliac clusters in families, with a sibling relative risk of 20 to 60 and the high concordance between monozygotic twins (75%) point to the strong contribution of genetics to CD. [For a review of the genetics, see Sollid (³¹).] HLA genes have a key contribution. The primary HLA association in most patients with CD is with the DQ2 heterodimer (DQA1*05/DQB1* 02), which can be encoded in cis-position by the DR3-DQ2 haplotype (bringing DQA1*0501 and DQB1*0201) or in trans-position in heterozygotes bearing DR5 and DR7 haplo types (bringing, respectively, DQA1*0501 and DQB1*0202). The DQ α and β chains encoded by the genes present in these different HLA haplotypes differ by only one amino acid with no functional importance. Thus, these chains combine to form the same functional DQ2 molecule on the cell surface. Approximately 10% of patients in Europe are DQ2 negative and possess the DR4 haplotype. The strong bias in favor of the DR4-DQ8 haplotype (DRB1*04-DQA1*0301-DQB1*0302), suggests that DQ8 is the susceptibility factor in these patients, a hypothesis supported by functional data. Patients lacking DQ2 or DQ8 are exceptional, and the lack of either haplotype strongly argues against the diagnosis of CD. Other genes in the HLA region might play a role, but a search for other HLA genes is complicated by the strong linkage disequilibrium within the HLA region. In patients with DR3, an as-yet-unidentified gene in the region telomeric to the HLA-F locus modifies susceptibility to CD and type I diabetes.

Studies in siblings indicate that the HLA region contributes approximately 35% to 40% to familial aggregation, thus leaving a potential important role for non-HLA-linked genes. Genome searches have failed to unambiguously identify any new susceptibility locus. The failure to identify additional genetic loci may be attributable to a minor influence of each potentially associated predisposing gene, to the genes' heterogenous distribution in an outbred human population, or to gene-gene interactions. The region most consistently associated with CD is 5q31-33, a region encompassing many candidate susceptibility genes. None of them have been tested, except the IL12B gene, which was not found to be associated with CD. It is reasonable to assume that some of the unidentified genes control the immune response. Accordingly, several studies of candidate genes suggest the role of the gene encoding CTLA-4 (2q33), a negative regulator of the immune response or of a neighbor gene.

Pathogenesis of the Villous Atrophy

CD is a model for immunologic diseases with a defined trigger. Thus, intestinal inflammation is strictly dependent on dietary exposure to prolamins (Fig. 1). These proteins, characterized by their high content in glutamine (34%) and proline (20%), are present in wheat gluten, where they include gliadins, 250 to 300 amino acid long polypeptides with many variants subdivided into α-, γ-, ω-gliadins, and glutenins divided into high- (650-800 residues) and low- (270-330 residues) molecular-weight molecules. In addition, toxic prolamins are found in barley (hordeins) and rye (secalins). Recent work has provided new insight into the pathogenesis of CD and suggest that there is an interplay between adaptive immunity, characterized by a specific and memory T-cell response to gluten peptides, and innate immunity, involving less specific mechanisms.

Fig. 1

The Adaptive Immune Response to Gluten

L. Sollid and F. Koning have accumulated strong experimental evidence that toxicity of prolamins is largely attributable to their ability to generate a specific T-cell response to gliadin peptides in the intestine of individuals at risk (adaptive immune response). They propound an attractive pathogenic scheme that explains the interaction between the key environmental (gluten) and genetic (HLA-DQ2/8) factors and also provides a role for the autoantigen, tissue transglutaminase (TG2) [reviewed in Sollid (³¹)].

The first key observation was the demonstration that CD4+ T lymphocytes (TL) derived from the lamina propria of patients with CD but not of control subjects, were activated in response to gluten peptides presented by HLA-DQ2/8 molecules expressed at the surface of antigen-presenting cells. This T-cell response was ascribed to the selective capabilities of HLA-DQ2/8 molecules to bind gluten peptides (³²).

It became rapidly clear, however, that binding of gluten-derived peptides to HLA-DQ2/8 required prior deamidation. The pocket of HLA-DQ2/8 molecules, that binds antigenic peptides, contains positively charged basic amino acid residues in several anchor positions and thus has a low avidity for native gliadin peptides, which contain few negatively charged amino acids. Sollid and Koning observed that deamidation of specific glutamine residues into glutamic acid introduced a negative charge, which simultaneously enhanced peptide binding and T-cell stimulation. They took a further step by demonstrating the role of the autoantigen tissue transglutaminase (TG2) in the in situ deamidation of gliadin peptides (^33,34). TG2 is a ubiquitous enzyme that catalyzes both transamidation and deamidation of specific polypeptide-bound glutamines. Substrates of TG2 include endogenous as well as exogenous proteins, including gliadin, for which it has a high avidity. TG2 transamidation activity predominates at neutral pH and in the presence of primary amines. It allows the formation of isopeptide bonds between glutamine and lysine residues in proteins (cross-linking), thereby playing a physiologic role in matrix assemblage and tissue repair. In CD, the cross-linking activity of TG2 is thought to play a role in the appearance of anti-endomysium antibodies. Thus, the formation of complexes in which TG2 is bound to gliadins likely allows the formation of a neoantigen recognized by the immune system. The disappearance of anti-TG2 autoantibodies after gluten exclusion argues for this hypothesis. When pH decreases below 7.3, the transamidation activity of TG2 decreases, while its deamidation rate increases. Expression of TG2, increased in active CD (as in other inflammatory conditions), has been noted in the brush border and in the subepithelial area. It is upregulated at the surface of activated macrophages, induced during the maturation of dendritic cells, where it might participate in the endocytic process. Thus, TG2 might deamidate gluten peptides either in the brush border (where pH is approximately 6.6 in the proximal intestine) or during their endocytosis and processing in the endosomal compartment of lamina propria dendritic cells (where the pH is low 5) [reviewed in Sollid (³¹)].

Study of TG2 specificity has provided additional insight regarding its role in CD. Accessibility of glutamine residues to deamidation by TG2 depends on their position in relation to proline residues. TG2 acts preferentially on the sequence Gln-X-Pro. This observation provides a first explanation for the importance of proline residues in the toxicity of prolamines. Furthermore, this observation suggests that TG2 is involved in the selection of the gluten epitopes recognized by T cells. Algorithms based on optimal glutamine-proline spacing for TG2 deamidation predict numerous immunogenic peptides in toxic prolamins (gliadins, glutenins, hordeins, and secalins), but not in avenins which have low or no toxicity (³⁵). In keeping with these theoretical data, a growing number of T-cell epitopes (of approximately 10 amino acids each) have been identified by functional studies in α and γ gliadins and in glutenins, which cluster in the predicted regions. Their number is more important in children than in adults, in whom the immune response seems to concentrate on a restricted number of so-called immunodominant epitopes (³⁶).

The cascade of events triggered in the lamina propria by gliadin peptide activation of CD4+ T lymphocytes is not entirely elucidated, but interferon γ seems to play a central role. The in situ expression of interferon γ is massively increased in active CD, and its secretion is observed in vitro in response to the stimulation of lamina propria-derived T-cell clones or of organ cultures of patients with CD treated with gluten-free diet. Furthermore, STAT-1, the transcription factor downstream the signaling pathway elicited by interferon γ is activated in situ in patients with active CD, and anti-sens oligonucleotides inhibiting STAT-1 transcription prevent the induction of two genes under the control of interferon γ in organ cultures of treated patients stimulated by gliadin peptides (³⁷). Interferon γ could induce epithelial damage, either via a direct toxic effect on the enterocytes or indirectly by stimulating other cell types in lamina propria, particularly macrophages. Activation of the latter cells can lead to the release of large amounts of compounds deleterious to the mucosa, such as metalloproteases.

The Role of Innate Immunity

The physiopathogenic scheme propounded by L. Sollid and F. Koning has considerably improved our understanding of the pathogenesis of intestinal lesions in CD. However, it leaves a number of questions unanswered. The contribution of innate immunity (the part of the immune system involved in the immediate and nonspecific responses of the host against pathogens) recently has been considered.

a) The first question concerns the role of the peptide 31-49 (or of the shorter peptide 31-43) common to the N-terminus region of the α-gliadins. Several studies using organ culture, as well as one study of biopsy specimens after in vivo instillation of peptides in the duodenum of treated patients (³⁸), pointed to this peptide as one major toxic peptide of gluten based on its ability to induce epithelial lesions and recruitment of IELs. Surprisingly, this peptide did not stimulate intestinal CD4+T cells from patients with CD. Maiuri et al. (³⁹) recently showed that addition of peptide 31-43 to duodenal biopsy specimens in organ cultures activates the intestinal macrophages of patients with CD, inducing macrophage production of the IL-15 cytokine, their expression of COX2, and the activation of the p38 MAP kinase. IL-15 subsequently enhanced the antigen presentation capacity of dendritic cells and thus the adaptive T-cell response against immunostimulatory peptides. Peptide 31-43 also was able to enhance epithelial apoptosis and IEL recruitment via a relay involving IL-15 (³⁹). It remains unclear why this peptide is active only on innate immune cells in patients with CD. It implies that, in patients with CD, this peptide somehow finds a way to bind and specifically activate the latter cells, likely via a receptor to be identified. Another recent study suggests that some gluten-derived peptides might activate innate immunity, but in that study, the activation of macrophages was not specific to patients with CD (⁴⁰).

b) A second question concerns the changes in the homeostasis of IEL that lead to the emergence of lymphoid malignancies in some CD patients.

A massive increase in IELs is a hallmark of the disease not observed in other inflammatory bowel diseases. IEL with a CD8+ phenotype and a T-cell receptor for antigen of the αβ type (TcRαβ) are mainly increased during active CD, suggesting that they may participate together with the CD4+ lamina propria T cells in an adaptive response to gluten. This hypothesis is supported by recent work showing their activation by a gliadin peptide presented by the MHC class I antigen HLA-A2 (⁴¹). However, it is likely that mechanisms related to innate immunity also participate in the activation/expansion of IELs. A second subset of IEL bearing the γδ T-cell receptor is increased in patients with CD. This increase is not restricted to active CD but is detected in patients who have recovered a normal intestinal histologic status after a gluten-free diet and in patients with latent disease without villous atrophy (^9,10). The nature of the antigens recognized by intestinal TcRγδ IEL and their functions remain elusive, but it is generally believed that they act as sentinels allowing immediate response to epithelial damage, so they belong to innate immunity. Their increase might result from epithelial alterations caused by inflammation or from alteration of the nonspecific mechanisms that control their homeostasis. Our results support both hypotheses. Expression of complementary receptors involved in innate immunity or response to epithelial stress are enhanced in IELs or epithelial cells during active CD. Their exact contribution is being studied (^42,42a). However, IL-15, a cytokine that controls the homeostasis of IEL, is massively increased in the intestine of patients with active CD or RCS (⁴³). Our data in RCS indicate that IL-15 plays a key role in the alterations of IEL homeostasis that may ultimately lead to lymphomas. Thus, IL-15 cytokines provides signals mandatory for the survival or expansion of the abnormal clonal IELs. In addition, IL-15 cytokine induces their secretion of interferon γ and their cytotoxicity against epithelial cells, thereby favoring the severe enteropathy characteristic of this condition (⁴³).

Altogether, our results and those of Maiuri et al. point to the important role of IL-15 in the pathogenesis of CD. This cytokine at the interface between innate and adaptive immunity might contribute to the activation of lymphocytes within the intraepithelial compartment and trigger the adaptive CD4+ T-cell response in the lamina propria. The mechanism (or mechanisms) that leads to its overproduction in CD remains unknown.

Alterations in the Digestion and Processing of Gliadin Peptides in CD?

Recent studies by Shan et al. have demonstrated the marked resistance of gliadins to digestion by intraluminal and brush border enzymes. These authors showed that treating a recombinant α-gliadin with a combination of digestive enzymes in conditions mimicking the in vivo situation released a highly resistant 33 amino acid peptide (33-mer) encompassing a cluster of three immunodominant T-cell epitopes. The resistance of this peptide to digestive enzymes was ascribed to its high proline content, as rapid hydrolysis could be obtained in the presence of a bacterial prolyl-endopeptidase. The authors pointed to the lack of enzymes appropriate to cut proline-rich polypeptides in the brush border. They suggested this physiologic defect might favor the access of gliadin peptides, particularly the oligomerized T-cell epitopes present within the 33-mer into the mucosa (^44,45).

Other studies suggested that gliadin might alter the epithelial barrier. Fasano et al. (⁴⁶) reported that epithelial permeability is increased in active CD and ascribed this change to an increased expression of zonulin, a negative regulator of epithelial tight junctions. Furthermore, Fasano and colleagues recently suggested that gliadin might induce zonulin-dependent actin polymerization in normal intestinal epithelial cells from CD with disassembly of tight junctions, thereby promoting its own entrance (⁴⁷). Another recent study suggested that gliadin can induce the apoptosis of intestinal epithelial cell lines, a mechanism that might alter the epithelial barrier (⁴⁸). However, these interesting data are difficult to reconcile with the lack of gliadin toxicity in healthy individuals.

Using the Ussing chamber to study the processing and transport of gliadin peptides in intestinal biopsy specimens of control subjects and patients with CD, we have come to partly divergent conclusions (⁴⁹). First, no major leakage through the epithelial barrier was detected in treated or active CD. Furthermore, the purified peptides tested, including the toxic 31-49 peptide and the 33-mer encompassing three T-cell epitopes, did not exert any detectable toxic effect within the 3 hours of the assay. As also reported by Shan et al., these gliadin peptides were highly resistant to hydrolysis by brush-border enzymes in the mucosal compartment. Our approach did allow us to demonstrate that, in control subjects and in patients with CD who were on a strict gluten-free diet, the gliadin peptides could be efficiently digested into amino acids by the epithelium, with only tiny amounts of intact peptides gaining access to the serosal compartment. The situation was different in active CD. In these patients, the total flux of peptides remained small (no more than 0.3% of the input), a finding which argues against any major paracellular leakage. However, a substantial fraction of the peptides reached the serosal compartment with little or no digestion. Thus, 50% of peptide 31-49 was delivered intact into the serosal compartment, whereas the 33-mer was only partially degraded during transport. Therefore, it is possible that immunostimulatory sequences are released from the 33-mer during its intestinal transport. Altogether, our results indicate that, in active CD, both the peptide 31-49 and immunostimulatory peptides embedded within the large 33-mer have an increased access into the mucosa, through a mechanism that remains to be determined. Thus, they could stimulate innate and adaptive immunity, thereby perpetuating T-cell activation and intestinal inflammation. However, in healthy individuals and in treated patients, the entrance of gliadin peptides appeared minimal because of their efficient digestion during epithelial transport, which argues against the hypothesis of a privileged access of gliadin peptides into the undamaged intestinal mucosa. It is possible that small amounts of gliadin peptides are sufficient to trigger the disease in predisposed patients. Alternatively, the initial triggering of the abnormal immune intestinal response may require a specific alteration of the epithelial barrier. This could be the case early in life, upon first introduction of gluten in young children, because of the immature intestinal barrier or later after infectious intestinal episodes.

Primary and/or Acquired Alterations of Immune Regulatory Mechanism(s) in CD?

Although our understanding of the different immunologic mechanisms involved in the generation of the intestinal inflammatory response has improved greatly, one key question remains unsolved: Why do only a small subset of individuals with the HLA-DQ2/8 haplotype develope CD? In other words, why does tolerance to gluten fail in only a subset of HLA-DQ2/8 patients?

Oral administration of soluble antigen normally induces systemic (and local) hyporesponsiveness. This phenomenon, defined as oral tolerance, has been well studied in animal models. It relies on several complementary immunoregulatory mechanisms, and particularly on the induction of regulatory T cells producing cytokines, such as TGFβ and IL10, which have strong inhibitory effects on proinflammatory immune responses. Strikingly, in mice engineered to express human HLA-DQ8 and CD4 molecules, gluten induced a strong proliferative response of intestinal lymphocytes, but the latter cells release large amounts of TGFβ and no interferon γ. As a consequence, in this model, gluten did not induce any enteropathy (⁵⁰). These data suggest that a defect in the mechanisms maintaining oral tolerance might be present in patients with CD. The high incidence of autoimmune disorders in patients with CD (type I diabetes, thyroiditis) points to a possible general defect in immune regulation, inasmuch as recent studies have underlined similarities between the regulatory T cells that control autoimmunity and intestinal inflammatory responses (⁵¹). Thus, it is possible that the additional genetic risk factors suggested by the familial and genome studies in CD influence the regulation of the immune response. This hypothesis is supported by the linkage of CD with the gene encoding CTLA-4, a negative regulator of the immune response, and with the 5qter chromosomal region, which contains numerous genes involved in the immune response.

Another hypothesis raised by a recent study suggests a possible contribution of the autoantigen TG2. Knock-out mice who lack TG2, show deficient apoptosis and develop autoimmune diseases (⁵²). These defects have been ascribed to the role of TG2 in the conversion of latent inactive TGFβ1 into its active form. TGFβ1 is a potent pro-apoptotic and anti-inflammatory cytokine, which also plays a role in the final differentiation of epithelial cells. As suggested by Halttunen and Mäki (⁵³), anti-TG2 autoantibodies might inhibit the function of TG2 and decrease the amount of active TGFβ available in situ, thereby impairing epithelial cell differentiation and local T-cell regulation. However, the effects of the autoantibodies on TG2 activity remain a matter of controversy. Furthermore, the inhibition of TG2 also might decrease its enhancing effects on peptide recognition.

Finally, it is known that CD can remain latent for many years, indicating that tolerance to gluten can be acquired but may be subsequently broken. The triggering of clinically overt CD upon treatment by interferon α (it is alpha and not gamma at this site) and detection of this pro-inflmmatory cytokine in the intestine of patients with active CD underlines the possible role of viral infections in the dysregulation of the local immune response. Viruses are strong inducers of interferon α (idem alpha and not gamma), which could favor the loss of oral tolerance by inducing the interferon γ pathway and the activation of antigen-presenting cells (⁵⁴). As suggested by Maiuri et al. (³⁹), such a triggering role might also be played by IL-15, which is induced in response to intestinal infections. The role of intestinal infections might not be only to disrupt local immune homeostasis. A nonexclusive hypothesis is that there is a mimicry between peptides derived from infectious agents and gliadin peptides. Mimicry with peptides from an adenovirus was suggested initially by M. Kagnoff (⁵⁵), but later ruled out because of the lack of serologic evidence. Nieuwenhuizen et al. (⁵⁶) recently showed some similarities with peptides from Candida albicans but have not provided any experimental evidence to support the role of this infectious agent.

THE GLUTEN-FREE-DIET AND WHAT'S NEXT?

For 50 years, the gluten-free diet, excluding proteins from wheat, barley, rye, and oats has been the only treatment for CD. The toxicity of oats, a genetically more distant cereal, has always been a matter of discussion. Several studies from Finland and England of a large cohort of patients prospectively followed for more than 1 year on a diet including more than 50 g of oats per day have provided convincing evidence that oats are not deleterious for patients with CD (⁵⁷). Data derived from in vitro functional studies are conflicting, with one recent study arguing against the immunoreactivity to oats of patients with CD (⁵⁸), whereas peptides derived from avenins were found to activate T lymphocytes in a patient clinically and histologically sensitive to oats (⁶¹).

The gluten-free diet is a safe and efficient treatment but adherence is complicated and a true social burden for patients. There is a clear need for alternative therapies. The increasingly precise knowledge of the immunostimulatory peptides has suggested several strategies. Genetic modification of wheat to delete toxic peptides is complicated by the large number of T-cell epitopes and by the complexity of wheat genetics. An alternative strategy might be to develop peptide analogs capable of interfering with HLA-DQ binding and T-cell activation in order to redirect the immune response toward tolerance. However, such a strategy, tested in other models of autoimmune diseases (⁵⁹), awaits experimental testing in CD. In mice, a vaccinal strategy based on the intranasal administration of whole gliadin or of one of its isoform partially inhibited the systemic T-cell response to parenteral challenge by whole gliadin (⁶⁰). However, this approach is not easy to transpose into humans because of the non-negligible risk of enhancing immunization, instead of promoting tolerance.

Another strategy suggested by Maiuri et al. (³⁹) is blockade of signals derived from the cytokine IL-15. This proposal might still be premature in uncomplicated CD in the absence of data concerning the consequence of IL-15 blockade in vivo in humans. However, blocking IL-15 and its signals is an attractive possibility in refractory coeliac sprue, when patients have become unresponsive to the diet, and do not experience a response to conventional anti-inflammatory treatments, and are in a life-threatening condition.

Finally, one interesting suggestion recently was made by Shan et al. (⁴⁵), who suggested using an endoprotease of bacterial origin, to cleave gliadin peptides left uncut by the digestive and brush border enzymes. Our results, which indicate that the defective intraluminal digestion of gliadin peptides is largely compensated by their efficient epithelial processing in treated patients with CD, do not entirely argue against this possibility (⁴⁹). It might be useful to decrease the intraluminal concentration of peptides to further reduce the tiny amounts of gliadin peptides that can escape epithelial processing. This approach, which seems reasonably safe, probably could be easily tested by delivering known amounts of gliadin in capsules with or without the bacterial enzyme.

In conclusion, the improved knowledge of CD pathogenesis allows us to envision several alternatives to the gluten-free diet. Because CD is, in most cases, a benign disease fully cured by a safe treatment, any alternative treatment must meet high standards of efficiency and safety. The creation of good animal models for the investigation of CD would be useful in confirming the new pathogenic hypotheses and testing therapeutic approaches. This important step still encounters the complexity of the interplaying mechanisms. However, there is strong hope that these difficulties will be overcome by the increasing efforts at research justified by the high frequency of the disease. Promoting these efforts is the major purpose of the International Society for Research on Coeliac Disease, which was founded during the 10th International Symposium on Coeliac Disease in Paris.