Lamina propria lymphocytes
Infant food allergies are increasing, and many breast-fed infants now sensitize to maternally-ingested antigens. As low-dose oral tolerance requires generation of suppressor lymphocytes producing TGF-β1 (Th3 cells), we studied these cells in duodenal biopsies after diagnostic endoscopy. Spontaneous production of Th1, Th2 and Th3 cytokines by duodenal lymphocytes was studied using flow cytometry in 20 children with no eventual clinico-pathological diagnosis (controls), 30 children with multiple food allergy, nine with celiac disease and six with inflammatory enteropathies. Immunohistochemistry and in situ hybridization were used to localize TGF-β1 protein and mRNA in matched biopsies. We found no significant Th1/Th2 skewing amongst mucosal lymphocytes in allergic children compared to controls, although celiac and inflammatory enteropathy patients showed increased Th1 responses. By contrast, the allergic children showed reduction of TGF-β1+ lymphocytes in both epithelial and lamina propria compartments. Reduction of TGF-β1 expression was also seen in mononuclear cells and epithelium in food allergy by immunohistochemistry and in situ hybridization. The dominant mucosal abnormality in food allergic children was, thus, not Th2 deviation but impaired generation of Th3 cells. As generation of these cells requires innate immune response to enteric bacteria, we suggest that changing infectious exposures may inhibit primary establishment of basic oral tolerance mechanisms.
The spectrum of food allergy in young children of the developed world has changed substantially, with marked increase in both IgE-mediated and non-IgE-mediated allergies 1, 2. Increasing numbers of developed-world infants now sensitize to multiple antigens in early life while exclusively breast-fed, a phenomenon rare a generation ago and still almost unknown in less privileged countries, presenting with either immediate reactions (urticaria, wheezing, vomiting) or delayed hypersensitivity (eczema, colitis, enteropathy or dysmotility) 1–4. Sensitization is often associated with evidence of immune dysregulation, including low IgA and IgG subclass deficiency 1, 4, 5.
Socio-economic improvement leads to rapid increase in childhood allergy at a national level 4, 6, suggesting that early environmental exposures may be as important as genetic predisposition in causing sensitization. Probiotic administration to at-risk newborns, indeed, reduced later eczema 7. The concept has thus arisen that the central difference between children in the allergy-free developing world and the allergy-driven developed world is that regulatory lymphocytes are not properly established unless the infant receives sufficient infectious challenge 8. While this represents a fundamental shift from a straightforward Th1/Th2 paradigm for allergic sensitization, there are so far few confirmatory reports of such reduction in regulatory lymphocyte responses.
The development of systemic tolerance in neonates is prevented if antigen-presenting cells are activated, or antigen presented at unusually low dose 9, 10. By contrast, in the induction of enteric tolerance, the bacterial flora plays an essential role, and animals maintained germ-free show impaired tolerance mechanisms and have substantially increased energy requirement for growth 11, 12. Tolerance to dietary antigen in mice is mediated by specific regulatory lymphocyte populations, notably, TGF-β1-producing T cells (Th3/TC3 cells), which are critically important in tolerance for antigens ingested in low doses 4, 13–15. These cells are generated in mucosal lymphoid tissue in an activation-dependent manner, and mediate "bystander tolerance" within the gut by inhibiting activation of all surrounding lymphocytes 13–15. This activation-dependent response to low doses may be more difficult to establish in infancy than high dose tolerance, which is mediated by T cell anergy, as neonatal lymphocytes are relatively difficult to activate 10. Indeed, oral administration of low-dose myelin basic protein sensitized neonatal rats, while similar amounts induced protective tolerance in adults 16.
The input from intestinal bacteria in the generation of tolerogenic lymphocytes is mediated through production of prostaglandin E2 by innate immune cells 17. Prevention of such an inflammatory response to the gut flora, thus, prevents oral tolerance 17. By contrast, TGF-β responses are up-regulated by mucosal inflammation 18. Thus, it has been postulated that the recent increase in food allergies may related to impaired early-life generation of TGF-β-producing lymphocytes, due to recent changes in the gut flora from the time of first colonization 4, 19. We have examined TGF-β responses by mucosal T cells in infants and children with food allergies to determine whether these children may show impaired development of a basic oral tolerance mechanism.
2.1 Mucosal lymphocyte populations
Yields harvested from biopsies from individual patients ranged from 2×106–3×106 cells from the lamina propria and 3×105–6×105 from the epithelium, with approximately 30–40% lymphocytes (viability >90%). Spontaneous cytokine production was seen in all patient groups (Fig. 1, 2). We did not detect significant differences in total cell numbers between groups within the lamina propria. The food allergic children showed significant reduction of the percentage of total intraepithelial lymphocytes (IEL) compared to controls, for both CD8+ and the minority CD4+ populations (mean ± SE, 17.4±1.8 vs. 25.9±2.7 and 8.8±0.9 vs. 12.9±1.3% respectively, p<0.05). However percentages of lamina propria CD4 and CD8 cells were unchanged (13.4±1.3 vs. 11.7±1.5 and 12.4±1.5 vs. 10.3±1.8 respectively, p>0.1). Percentage CD8+ IEL and CD4+ and CD8+ LPL were increased in coeliac disease compared to controls and food allergy (38.2±5.6, p<0.01 , 27.6±3.1, p<0.001 and 25.2±3.8., p<0.05 respectively). There was substantial variation amongst the inflammatory group, with non-significant increase in LPL (CD4 15.4±4.8 %, CD8 13.9±3.6) and reduced CD8+ IEL (14.9±1.9, p<0.05).
2.2 Cytokine production
The normal controls showed higher percentages amongst CD4+ cells of both Th1 and Th2 cytokine expression than in CD8+ cells (10–25% vs. 2–12%, Fig. 2). Broadly, similar results were seen for the Th1 cytokines IL-2 and IFN-γ. These percentages did not vary significantly with age in any of the groups studied. Representative 4-quadrant plots are shown in Fig. 1. We did not find evidence of substantial deviation away from Th1 or towards Th2 responses in the multiple food allergy group in comparison to the normal controls, although the percentage of IL-2+CD4+ LPL was reduced (Fig. 2A–C). In celiac disease the percentage of Th1 cytokine-producing cells was non-significantly increased, but overall numbers of lymphocytes were substantially higher, while an increased Th1 and Th2 cytokine-producing cell percentage was seen in inflammatory controls, but with large scatter.
A contrasting pattern was seen when spontaneous TGF-β1 production was studied. As for other cytokines, the controls showed a relatively higher percentage of TGF-β+ cells amongst CD4+ than CD8+ cells, but the children with multiple food allergy showed a reduced percentage of TGF-β+ T cells in each population studied, reaching significance for both CD8+ and CD4+ IEL and CD8+ LPL (Figs. 1, 2D). It was notable that TGF-β-producing lymphocytes were also relatively low in celiac disease in comparison to IFN-γ (Fig. 2D). There was no relationship between TGF-β+ lymphocyte numbers and the child's age in any of the groups.
2.3 Mucosal distribution of TGF-β1 protein and mRNA
The normal controls generally showed strong expression of TGF-β1 protein and mRNA within the intestinal epithelium, as previously reported 18, 20. Protein expression was maximal towards the villous tip, whereas mRNA was strongly expressed in the crypts. The mean mucosal density of TGF-β1 mononuclear cells in both normal and inflammatory controls was similar to previous reports 18, and was relatively lower than for IFN-γ+ cells (Fig. 3). Numbers of both IFN-γ+ and TGF-β1+ cells increased in inflammatory enteropathy, whereas in celiac disease only IFN-γ+ cells showed increase. The children with multiple food allergy showed no reduction in the density of IFN-γ+ mononuclear cells, providing further evidence that mucosal Th1 responses are not significantly reduced. By contrast, they showed significant reduction in the density of TGF-β1+ mononuclear cells (Fig. 3, 4). In situ hybridization demonstrated a similar reduction in the density of TGF-β1 mRNA+ cells in the food allergy group, which was also seen in celiac disease (Fig. 3, 4). The intensity of epithelial expression of both TGF-β1 protein and mRNA was markedly reduced in the food allergy group. In celiac disease there was reduced epithelial expression of TGF-β1 protein, as previously reported 20, while there was relative preservation of mRNA expression.
We have found consistent evidence, by flow cytometry, immunohistochemistry and in situ hybridization, to suggest that the dominant immunological abnormality in the small bowel of children with multiple food allergy may be a failure to establish normal numbers of TGF-β-producing regulatory cells. This was seen in children with either immediate or delayed reactions. Our findings of reduced TGF-β1 expression by mucosal lymphocytes in children with food allergy concord with recent reports by two other groups: reduction of both TGF-β1 and its type 1 receptor was demonstrated immunohistochemically in the mucosa of infants with food protein-induced enterocolitis 21, while milk-specific T cell lines derived from the mucosa of food allergic infants showed a bias towards Th2 cytokine production, notably IL-13, but did not produce TGF-β 22.
The technique of flow cytometry appears particularly well suited for the assessment of spontaneous cytokine production by mucosal lymphocytes, but provides challenges due to the relatively small number of cells that can be obtained from pediatric mucosal biopsies. This prevented more extensive testing of specific lymphocyte reactivities to individual dietary antigens in these freshly studied cells, and determination whether these TGF-β-producing lymphocytes could produce other cytokines in addition. It is important to note that this technique, in common with ELISPOT, provides a percentage figure rather than giving absolute numbers of cytokine-producing lymphocytes within the mucosa. In addition, the sensitivity of the technique means that cells with low-level intracellular cytokine production may be detected, leading to overall percentages higher than using other techniques. With these reservations, we suggest that these findings support recent contention that allergic sensitization may be due to inadequate regulatory responses rather than Th1/Th2 imbalance 8. Within the gut, reduction of regulatory lymphocyte numbers is likely to lead to a lack of "bystander tolerance", and thus, a propensity for multiple sensitizations 13–15. These findings suggest that sensitization of exclusively breast-fed infants may, indeed, occur because of impaired TGF-β-mediated low-dose oral tolerance 19. We also noted decreased expression of TGF-β1 protein and mRNA by the epithelium in both the children with multiple food allergy and in celiac disease 20. As epithelial antigen presentation is important in inducing tolerance 23, the mechanisms inducing this change warrant further study. Our findings suggest that this fast-increasing form of childhood allergy represents a primary failure to establish oral tolerance, rather than the loss of previously acquired tolerance, that hitherto characterized childhood food allergies 1.
The increase in mucosal TGF-β1+ cell numbers that is induced by inflammation 18 may at least partly explain the reduction in allergic sensitization afforded by upbringing in the developing world, or in circumstances of increased pathogen challenge 4, 24. The density of duodenal TGF-β1+ cells in rural Gambian infants is, indeed, up to ten times higher than in UK controls 25. Clearly, there is important genetic predisposition to mucosal sensitization 4, and recent evidence suggests that reduced proinflammatory responses to bacterial lipopolysaccharide via NF-κB is important in both human and murine inflammatory Bowel disease (IBD), where tolerance is lost to enteric bacteria rather than dietary antigen 26, 27.
We speculate that such genetically reduced NF-κB responses to enteric bacteria may give early-life survival advantage in the pathogen-challenged environment faced by most pre-20th Century or modern developing world infants. However, as the early bacterial exposures of infants reduce with socio-economic advance, similar hypo-responsiveness to bacterial components may then be inadequate for the induction of tolerogenic lymphocytes. Neonatal administration of probiotics may, thus, prevent sensitization of potentially atopic children 7 by inducing sufficient NF-κB responses in the gut epithelium and macrophages to allow effective generation of regulatory lymphocyte populations 28. As TGF-β-producing lymphocytes mediate low-dose but not high-dose oral tolerance, the apparent paradox that early supplementation of potentially atopic infants with cow's milk-derived formulas could actually protect against allergy 29 may be explained by delivery of sufficient antigen to invoke high-dose tolerance when low-dose mechanisms have become ineffective.
TGF-β-producing lymphocytes are also recognized to be important in preventing allergen-induced bronchospasm, and may, thus, play an important regulatory role in allergic asthma 30, which develops later in many former food-allergic infants in the so-called "atopic march". Little is known of the development of lung regulatory lymphocytes in childhood, but it is notable that a history of enteric rather than respiratory infections protects against the development of asthma 24. This suggests that similar mechanisms may also apply in the establishment of pulmonary tolerance, and thus, these multiply-sensitized infants may provide early evidence of a direct link between early infectious exposures and allergic sensitizations which is potentially amenable to specific therapeutic input. It will be important to determine whether other known regulatory cell populations, including IL-10-producing T regulator-1 cells and anergic/suppressive CD4+CD25+ cells, are similarly reduced in children with dietary allergies.
The inherent antigenicity of an individual dietary component may have become less important in privileged countries than the dosage and timing of its administration to infants and young children, with genetic predisposition to atopy and early infectious exposures the critical determinants of tolerance or sensitization. Optimization of preventative strategies for pediatric food allergy should include study of the regulatory lymphocyte response to immunomodulators such as probiotic organisms.
4 Materials and methods
4.1 Clinical details
We studied prospectively 65 children, referred in 1998–1999 to the pediatric food allergy and inflammatory bowel disease clinics at the Royal Free Hospital, a British Society of Allergy and Clinical Immunology-recognized tertiary referral center, for investigation of possible gastrointestinal disease. All required diagnostic small intestinal biopsy on clinical grounds and were endoscoped under general anesthesia. Ethical approval for extra biopsies was obtained from our local Research Ethics Committee and written informed consent was obtained from the parents in all cases. The exclusion criteria were simply based on sufficient biopsy material for analysis and availability of flow cytometric facilities on the day of endoscopy. Additional biopsy specimens were obtained from the third part of the duodenum and rapidly processed within 1 h.
The patients were grouped into four categories on the basis of findings upon investigation and subsequent response to dietary exclusions: (a) non-disease controls, (b) multiple food allergy, (c) celiac disease, (d) inflammatory controls. Clinical details of the patient groups are given in Table 1. For the non-disease controls (n=20), no final clinico-pathological diagnosis was made, and there was no evidence of food allergy. The children with multiple food allergy (n=30) were referred from our tertiary Paediatric Food Allergy clinic, with a diagnosis of allergy to at least two antigens, most commonly cow's milk, soya or egg, including those showing either immediate or delayed hypersensitive responses (Table 1). Allshowed clinical response on exclusion, with recurrence of symptoms on supervised open challenge or double-blind placebo-controlled challenge. Of the 30 children, eight suffered immediate reactions within 1 h, such as skin rash or lip swelling, as well as delayed reactions including eczema, diarrhea, vomiting after ingestion of dietary antigen, while 22 showed delayed reactions only (Table 1). Diagnosis was confirmed on the basis of clinical response on exclusion with recurrent symptoms on either open challenge (n=23) or placebo-controlled food challenge (n=7). Skin prick tests were positive for cow's milk and at least one further antigen in 10 patients, seven of whom manifested early reactions. Clinical details of several cases are reviewed in moredepth elsewhere 5. Biopsies were taken while the children were symptomatic, and further dietary exclusions were initiated on the basis of histological findings. Medications at the time of biopsy included ranitidine, cisapride, domperidone, gaviscon and/or omeprazole in nine patients. Histological findings were normal in only seven, while esophagitis was seen in 20 children and small intestinal enteropathy in 23. The enteropathy was generally mild, with villous blunting and increased lamina propria mononuclear cell infiltration. The mean age of onset of symptoms was 13 weeks (SE=2.4 weeks, range <1–40), and began while exclusively breastfed in 18/30. Intolerance to hydrolysate formulas occurred in 14/30 infants, who required an amino-acid formula. The children with active celiac disease had positive serology and showed crypt hyperplastic villous atrophy, while the inflammatory controls (n=6) all showed evidence of active duodenal inflammation (Table 1).
|Group||n||Clinical manifestations||Age (years) (± SE)|
|Non-disease controls||20||Investigated for reflux or enteropathy. Histology within normal limits and no gastroenterological diagnosis made.||2.9 (± 0.6)|
|Multiple food allergy||30||All required exclusion of 2 or more foods, making clinical response to exclusion and relapsing on challenge. Exclusion required for cow's milk (30/30), soya (22/30), egg (18/30), wheat (15/30), fish (3/30), peanut (4/30 − others never given peanut), sesame (3/30). Clinical details given in Sect. 4.||1.8 (± 0.4)|
|Investigations showed median IgA 0.3 g/l (upper quartile 0.7), IgE 15 kIU/1 (upper quartile 68, 2 patients > 1,000 kIU/1). Histological oesophagitis in 20/30, enteropathy in 23/30, normal appearances in 7/30.|
|Celiac disease||9||Active disease on gluten-containing diet. Diagnosed by ESPGHAN criteria||1.8 (± 0.2)|
|Inflammatory controls||6||Autoimmune enteropathy (2 cases), Crohn's disease (3 cases), chronic granulomatous disease (1 case).||3.4 (± 1.2)|
4.2 Flow cytometric assessment
Spontaneous cytokine production by mucosal lymphocytes was assessed in IEL and LPL using flow cytometric analysis. We studied cytokine production by CD4 (Th) cells and CD8 (cytotoxic T cells, Tc) cells, including Th1/Tc1 cytokines (IL-2, IFN-γ), Th2/Tc2 cytokines (IL-4) and Th3/Tc3 cytokines (TGF-β1). Single cell suspensions suitable for flow cytometry (using a 4-channel Coulter Epics-XL-MCL ) were prepared in two stages. Firstly the epithelial layer was removed using calcium-free Hanks' balanced salts solution. The remaining lamina propria tissue was digested with collagenase 4 mg/ml (Sigma, GB) for 2–3 h. Following washing steps, a single cell suspension prepared from each compartment and maintained at 4°C until fluorochrome staining (within 1 h). Conventional intracellular cytokine analysis was performed 32, using FITC-conjugated anti-IFN-γ, IL-2 and IL-4 (R & D Systems, GB) and PE-conjugated anti-TGF-β1 (human-specific, PE-conjugated clone TB21, IgG1, kindly provided by Serotec, GB), with cells phenotyped using antibodies recognizing CD4 and CD8 (Beckman-Coulter, GB). Clone TB21 reacts with either monomeric or dimeric TGF-β1 under reducing and nonreducing conditions, and has been validated for flow cytometric analysis using the TGF-β1-expressing colorectal cell line ATT CRL-2159 in comparison to the similar, but non-TGF-β1-expressing line ATT CCL-218 33. Caco-2 cells were used as TGF-β-producing cell lines for neutralization studies using recombinant TGF-β1 (Serotec), with dose-dependent inhibition of staining seen from 50 pg-5 ng/ml, and in activated PBL, where 5 ng/ml recombinant protein largely abrogated staining in both CD4 and CD8 cells (Fig. 1).
Cells were fixed with 4% paraformaldehyde and permeabilized in 1% saponin. In view of the small numbers of lymphocytes obtained from pediatric biopsies, an internal control using unstimulated PBL from each patient was employed to determine the parameters for positive cytokine staining. Isotype control reagents (mouse IgG1-FITC, ECD, PC5 and IgG2a-PE from Coulter-Immunotech, GB) were initially used to determine optimum fluorescence quadrant markers and to define nonspecific fluorescence in gut-derived and peripheral blood lymphocytes in 18 patients studied with this technique (not shown). No significant differences were found compared to the unstimulated PBL controls. Following acquisition of the samples by flow cytometry, data was analyzed using Winlist Version 4.0. A tight scatter region was first drawn around the major lymphocyte population, gating on the basis of forward and side scatter, as previously reported 30, 31. On a lymphocyte-gated plot of CD4 vs. CD8, regions were set to include all CD4+bright and CD8+bright cells. A minimum of 1,000 events within each gate was required for further analysis. One-dimensional histogram plots of the gated CD4+ and CD8+ cells were then generated, with the lower limit for distinguishing cytokine-positive cells being set for each antibody from the matched PBMC unstimulated cell sample. The percentage of total CD4+ and CD8+ cells within each compartment was determined, and then the percentage of cytokine-producing cells amongst these CD4+ and CD8+ populations. In initial studies, annexin-V and 7-amino actinomycin D-staining had identified apoptosis in 5–15% of LPL (not shown). Depletion of dead cells using annexin-V-coated magnetic beads prior to labeling did not, however, significantly alter the percentage of spontaneous cytokine-producing cells, and these studies were performed without such depletion.
4.3 Immunohistochemistry and in situ hybridization
Assessment of the mucosal density of TGF-β1 mononuclear cells was made using immunohistochemistry and in situ hybridization, performed on formalin-fixed biopsy samples taken during the same procedure. Antibodies used were monoclonal anti TGF-β1 (1/10, Genzyme, USA) and rabbit polyclonal anti IFN-γ (1/75, R & D). Biotin/avidin peroxidase immunohistochemistry (Vectastain Elite®, Vector Laboratories) was used, with bound antibody visualized with diaminobenzidine and counterstaining with Mayer's hemalum. Staining was first optimized on formalin-fixed sections of tonsil, spleen, normal and inflamed intestine, and non-primary and isotype-matched antibody controls were used throughout. TGF-β1 mRNA was localized using in situ hybridization. Biotin-labeled TGFβ1-probe cocktails, recognizing exons 6, 7A and 7B (Genebank accession number X02812, R & D Systems), were applied in hybridization buffer after conventional preparation (200 ng/ml), and the sections hybridized overnight at 37°C. After hybridization, sections were fixed in 0.4% paraformaldehyde in PBS and then rinsed in diethyl-pyrocarbonate (DEPC)-water, decreasing concentrations of SSC and PBS/Triton. Staining was visualized using peroxidase immunohistochemistry (Vectastain), following incubation with anti-biotin monoclonal antibody (Dako). Specificity of staining was assessed using sense probe, no probe, RNase-pretreated tissue and positive control tissue (tonsil and IBD).
4.4 Quantitation of mucosal staining density
The density of cytokine-immunoreactive and mRNA-positive mononuclear cells within the lamina propria was assessed by computerized image analysis (Leica, GB) with manual identification of positively stained cells within a defined area, which excluded all crypt epithelium. Only cells of clearly mononuclear morphology were counted and cells of obvious macrophage or fibroblast origin excluded.
4.5 Statistical analysis
Data were analyzed using Statgraphics Plus® software. As some data was not distributed normally, we compared lymphocyte percentages and mucosal density values in the disease groups (food allergy, celiac disease and inflammatory enteropathy) with the normal controls using the Kruskal-Wallis test, with p<0.05 being considered significant.
This work was supported by the UK Food Standards Authority (formerly Ministry of Agriculture, Fishery and Foods) FS30 Food Intolerance Program. We are grateful to Alison Lang, Josephine Garvey and Rita Shergill-Bonnar for expert dietary assessment. We thank our colleagues in the Department of Renal Medicine for allowing us generous use of flow cytometric facilities, and Dr Susan Davies and colleagues for histopathological assessment of esophageal and duodenal biopsies.