Report

Dietary antigens limit mucosal immunity by inducing regulatory T cells in the small intestine

See allHide authors and affiliations

Science  19 Feb 2016:
Vol. 351, Issue 6275, pp. 858-863
DOI: 10.1126/science.aac5560

Keeping immune cells quiet on a diet

Over thousands of years, our immune systems has evolved to distinguish self from foreign, perpetrating attacks on microbes but not ourselves. Given this, why do we fail to mount an immune response against most of the food we eat? Kim et al. compared normal mice, mice lacking microbes, and mice lacking microbes that were fed an elemental diet devoid of dietary antigens (see the Perspective by Kuhn and Weiner). Dietary antigens normally induced a population of suppressive immune cells called regulatory T cells in the small intestine. The cells were distinct from regulatory T cells induced by microbial antigens and prevented strong reactions against food.

Science, this issue p. 858; see also p. 810

Abstract

Dietary antigens are normally rendered nonimmunogenic through a poorly understood “oral tolerance” mechanism that involves immunosuppressive regulatory T (Treg) cells, especially Treg cells induced from conventional T cells in the periphery (pTreg cells). Although orally introducing nominal protein antigens is known to induce such pTreg cells, whether a typical diet induces a population of pTreg cells under normal conditions thus far has been unknown. By using germ-free mice raised and bred on an elemental diet devoid of dietary antigens, we demonstrated that under normal conditions, the vast majority of the small intestinal pTreg cells are induced by dietary antigens from solid foods. Moreover, these pTreg cells have a limited life span, are distinguishable from microbiota-induced pTreg cells, and repress underlying strong immunity to ingested protein antigens.

The ingestion of certain foods can trigger an immune reaction, ranging from mild allergy to anaphylaxis, in a fraction of the population. How dietary antigens (Ags) are normally rendered nonimmunogenic through oral tolerance is poorly understood, but it is known to require the participation of regulatory T (Treg) cells expressing the transcription factor Foxp3 (1, 2). Two types of Foxp3+ CD4+ Treg cells exist: the thymic Treg (tTreg) cells that develop from hematopoietic progenitors in the thymus, and the peripheral Treg (pTreg) cells that develop extrathymically from conventional T cells (36).

pTreg cells are abundant in the intestine but not in the secondary lymphoid tissues (7), implying that pTreg cells develop in response to enteric Ags derived from the commensal microbiota and/or food. In the colon, the intestinal microbes induce the development of pTreg cells, and these cells are depleted in germ-free (GF) mice (79). However, GF mice possess normal numbers of pTreg cells in the small intestine (7), the origin of which has yet to be documented. Oral administration of a nominal protein Ag under experimental conditions can induce a fraction of Ag-specific CD4+ T cells to differentiate into pTreg cells (10), but whether pTreg cells are generated in response to a typical diet thus far has been unknown. To address this, we studied the effect of depleting dietary Ags by raising mice on a chemically defined elemental diet devoid of macromolecules. We derived such Ag-free (AF) mice by producing offspring from GF mice that were weaned onto and subsequently raised on the elemental Ag-free diet (table S1) (11).

Young adult AF C57BL/6 (B6) mice appeared healthy, were similar in size and weight, and possessed comparable serum biochemical and metabolic markers, except for moderately low levels of vitamin A and D, albeit within the acceptable ±100% range relative to those of specific pathogen-free (SPF) and GF B6 mice (fig. S1). As previously reported (12, 13), AF mice had lower serum immunoglobulins, normal-sized spleens, and lower lymphocyte counts in the small intestinal lamina propria (siLP), but not in the colonic lamina propria (cLP), relative to SPF mice (fig. S2). The hypocellularity in the siLP was particularly pronounced for CD4+ T cells because of the depletion of memory-phenotype (CD44hi CD62Llo) cells, including Tbet+ cells (Fig. 1, A and B, and fig. S4); CD4+ T cell counts in the cLP were normal, but memory-phenotype CD4+ T cells were significantly decreased, as in GF mice (figs. S3 and S4). These phenotypes suggest that local activation of CD4+ T cells in the small intestine is driven mainly by dietary Ags, whereas in the colon it is induced by the microbiota.

Fig. 1 Depletion of macromolecules from the diet precludes the development of pTreg cells in the small intestine.

Shown are comparisons of T cell populations from the siLP and cLP harvested from aged-matched young adult (6- to 10-week-old) SPF, GF, and AF B6 mice. (A) CD4+ T cells from the siLP, expressed as percentages of lymphocytes (left) and total numbers (right). (B) Representative fluorescence-activated cell sorting (FACS) plots of CD44 versus CD62L (numbers in the boxes indicate percentages of cells in the gate), with a graph of the percentages of CD44hi CD62Llo cells among the siLP CD4+ T cells. (C) Representative FACS plots of CD4 versus Foxp3 on gated CD4+ T cells from the spleen (SPL), siLP, and cLP. (D) Foxp3+ CD4+ T cells from the siLP are expressed as percentages of lymphocytes (left) and total numbers (right). (E) Representative histograms of Nrp-1 expression, with a graph of the percentages of Nrp-1lo cells among gated Foxp3+ CD4+ T cells from the SPL, mesenteric lymph node (mLN), siLP, and cLP. At least four independent experiments had similar results. Graphs in (A), (B), and (D) show pooled data points from four independent experiments. P values were determined by one-way analysis of variance (ANOVA) with the Bonferroni post-test. Error bars show SEM. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Foxp3+ CD4+ Treg cell counts in adult AF B6 mice were also normal in the periphery but much reduced in the intestine compared with those of SPF mice. In the cLP, AF mice resembled GF mice and possessed about half the percentages and total numbers of Treg cells relative to SPF mice (Fig. 1C and fig. S5B). In the siLP of AF mice, despite higher percentages, total numbers of Treg cells were about one-fifth those of GF and SPF mice (Fig. 1, C and D). To distinguish between tTreg and pTreg cell subsets, the expression of neuropilin-1 (Nrp-1) was analyzed. As reported previously (7), pTreg cells expressing little Nrp-1 (Nrp-1lo ) were largely absent in the spleen and lymph nodes but made up 50 to 70% of Treg cells in the siLP and cLP in SPF mice; in GF mice, pTreg cells were rare in the cLP but prominent in the siLP (Fig. 1E). In contrast, in AF mice, Nrp-1lo pTreg cells were depleted in both the siLP and cLP (Fig. 1E). Intestinal Nrp-1lo pTreg cells in SPF mice, but not the few remaining in AF mice, also expressed higher levels of CTLA-4 and the cytokine interleukin (IL)–10 than Nrp-1hi tTreg cells did (figs. S6 and S7). IL-10+ Foxp3 CD4+ cells, which were presumably type 1 regulatory T (Tr1) cells, were present in low numbers in the gut of SPF but not of GF and AF mice, and Tr1 cells expressed lower levels of IL-10 than did Foxp3+ Treg cells (fig. S7, A and B). Together, these results indicate that dietary Ags induce the development of most of pTreg cells in the siLP under normal conditions, and pTreg cells appear more essential than Tr1 or tTreg cells in regulating immune responses to dietary Ags.

To assess the relative contributions of milk versus solid-food Ags in inducing the development of intestinal pTreg cells, neonatal and adult mice were examined. pTreg cells in the siLP were sparse before weaning in AF, GF, and SPF B6 mice but became prominent shortly after weaning these mice onto a normal chow diet (Fig. 2A and fig. S8). Furthermore, the deprivation of dietary Ags, accomplished by weaning neonatal GF mice onto the Ag-free diet or onto an “amino acid” diet that was devoid of proteins, prevented efficient development of pTreg cells in the siLP (Fig. 2, B and C); in addition, unlike regular AF mice, these mice had normal levels of serum vitamin A, which promotes the generation of pTreg cells through its metabolite retinoic acid (fig. S9, A and B) (14, 15). Moreover, the administration of retinoic acid to adult AF mice did not result in the emergence of siLP pTreg cells (fig. S9C). Hence, the majority of pTreg cells in the siLP develop in response to Ags derived from proteins in a solid-food diet. As expected, siLP pTreg cells underwent rapid turnover and had a short life span when deprived of food Ags. Thus, when adult (8-week-old) GF mice were placed on the Ag-free diet, the fraction of siLP pTreg cells decreased considerably (by ~40%) after ~4 weeks (Fig. 2D). These results indicate that pTreg cells in the siLP are continuously generated and replaced in response to dietary Ags, with a half-life of 4 to 6 weeks.

Fig. 2 The bulk of small intestinal pTreg cells develop when solid food is introduced and have a relatively short life span.

(A to C) Analysis of Nrp-1 expression on gated Foxp3+ CD4+ T cells from the siLP and cLP of neonatal (3-week-old) GF and AF B6 mice, before and after weaning onto a specific diet. Shown are representative histograms of Nrp-1 expression, with graphs of the percentages of Nrp-1lo cells on gated Foxp3+ CD4+ T cells (A) from the siLP of neonatal GF and AF mice before and 4 and 8 weeks after weaning onto a sterile chow diet (CD), (B) from the indicated tissues of neonatal GF mice 2 and 4 weeks after weaning onto the Ag-free or sterile chow diet, and (C) from the indicated tissues of GF mice 3 weeks after weaning onto a sterile chow diet or an amino acid diet devoid of proteins. (D) Adult (8-week-old) GF mice were fed with the Ag-free diet (AFD) for 2 and 4 weeks, and Nrp-1 expression on siLP and cLP Foxp3+ CD4+ T cells was examined. Representative histograms of Nrp-1 expression (left) and graphs of percentages and numbers (right) of Nrp-1lo Foxp3+ CD4+ T cells are shown. Graphs in (A), (B), and (D) show combined data points from two independent experiments. The graph in (C) is representative of two independent experiments. P values were determined by two-tailed unpaired Student’s t tests [(B) and (C)] and one-way ANOVA with the Bonferroni post-test (D). Error bars show SEM. *P < 0.05,;**P < 0.01; ***P < 0.001.

In addition to dietary Ags, commensal microbial Ags presumably contribute to the generation of pTreg cells in the siLP. To determine whether such pTreg cells can be selectively identified, we examined the expression of RAR (retinoic acid receptor)–related orphan receptor gamma t (RORγt), which is expressed by lymphocytes that interact with microbes (16). As previously reported (17), in SPF B6 mice, RORγt+ Treg cells were present in the siLP and cLP, mostly as Nrp-1lo pTreg cells (Fig. 3, A and B). In contrast, RORγt+ Treg cells were almost undetectable in GF and AF B6 mice (Fig. 3, A and B); RORγt+ CD4+ conventional T cells were similarly absent (fig. S10, A and B). The gut RORγt+ Treg cells rapidly appeared in GF mice after conventionalization (fig. S10, C and D). Moreover, treatment of SPF mice with a cocktail of antibiotics for 4 weeks from the time of weaning led to a depletion of RORγt+ pTreg cells in the siLP and cLP, without affecting the generation of RORγt pTreg cells (Fig. 3C). In addition, weaning SPF mice onto the Ag-free diet or an amino acid diet caused a severe reduction of Nrp-1lo pTreg cells in the siLP, mostly due to a depletion of RORγt pTreg cells (Fig. 3D and fig. S11). The bacterial load, but not the composition, in the feces of SPF mice that were weaned onto the Ag-free diet was relatively normal (fig. S12). Collectively, these findings add to recent reports (18, 19) by showing that whereas RORγt+ pTreg cells are induced by commensal microbiota, RORγt pTreg cells are driven by dietary Ags.

Fig. 3 Under normal conditions, a subset of small intestinal pTreg cells is induced to develop by the commensal microbiota and expresses RORγt.

Shown are analyses of RORγt expression on gated Foxp3+ CD4+ T cells from the siLP and cLP of young adult SPF, GF, and AF B6 mice [(A) and (B)] or on SPF mice after treatment with antibiotics or after weaning onto an Ag-free diet [(C) and (D)]. (A) Representative histograms of RORγt expression, with a graph of the percentages of RORγt+ cells among gated Foxp3+ CD4+ T cells from the SPL, mLN, siLP, and cLP of SFP, GF, and AF mice. (B) Representative FACS plot of Nrp-1 versus RORγt on gated Foxp3+ CD4+ T cells from the indicated tissues and mice. (C and D) Representative FACS plots of Nrp-1 versus RORγt, with graphs of the percentages of Nrp-1lo (left) and RORγt+ cells (right) among Foxp3+ CD4+ T cells in the siLP and cLP of recently weaned SFP mice; the mice were given water with or without broad-spectrum antibiotics (ABX) (C) or a chow diet or Ag-free diet (D) for 4 weeks. The graph in (A) is representative of at least two independent experiments, and graphs in (C) and (D) show combined data points from two independent experiments. P values were determined by one-way ANOVA with the Bonferroni post-test in (A) and by two-tailed unpaired Student’s t tests or t tests with Welch’s correction in case of unequal variance in (C) and (D). Error bars show SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

We next examined the dendritic cell (DC) subsets in the intestines of AF mice. Unexpectedly, the siLP of AF mice contained a ~40% lower representation of the CD103+ CD11b+ subset that promotes pTreg cell development (14, 20) and about three times the proportion of the CD103+ CD11b subset, relative to SPF and GF mice, resembling conventional splenic DCs (fig. S13, A to C) (21). However, such a skewed DC composition only partially explains the paucity of pTreg cells in the intestines of AF mice. This is because a normal DC subset composition was found in the mesenteric lymph nodes of AF mice, which are the induction sites of pTreg cell development; moreover, purified CD103+ CD11b+ DCs from the siLP of AF mice were able to efficiently stimulate T cells (fig. S14, A and B) and induce up-regulation of Foxp3 in OT-II cells under in vitro conditions (fig. S14, C and D). Furthermore, weaning neonatal AF or GF mice, which also possess the skewed DC subsets (fig. S13D), onto a chow diet led to the normalization of DC subsets within 1 week (fig. S13, D and E) and efficient development of pTreg cells within 2 weeks in the siLP (Fig. 2A). Lastly, siLP pTreg cells from neonatal AF mice weaned onto a chow diet exhibited ~50% demethylation at the CNS2 regions of the Foxp3 locus (fig. S15), indicative of stable Foxp3 expression, as previously reported (7, 22).

Despite the deficiency in siLP pTreg cells, B6 SPF or GF neonatal mice do not develop food allergies when they are weaned onto a chow diet. This could be due to various anti-inflammatory proteins in milk (23), such as immunoglobulins, transforming growth factor (TGF)–β, osteopontin, and lactadherin, which could ameliorate the allergic response. We therefore examined the effect of feeding adult B6 AF mice a sterile chow diet; this failed to induce any gross sign of intestinal pathology. However, the initial local immune response to dietary Ags could be vigorous because of the absence of pTreg cells and the increased numbers of siLP CD103+ CD11b DCs. Hence, we adoptively transferred ovalbumin (OVA)–specific naïve OT-II CD4+ T cells into adult AF mice and then gave them OVA orally. Examination on day 7 revealed that OT-II cells in the siLP of AF mice expanded ~50 and 7 times as much as in control SPF and GF mice, respectively (Fig. 4A). Moreover, whereas the majority (~60%) of expanded OT-II cells up-regulated Foxp3 in SPF and GF mice, only ~30% of OT-II cells up-regulated Foxp3 in AF mice (Fig. 4B).

Fig. 4 Conditions in which pTreg cells are depleted allow strong immune responses to dietary antigens and an increased susceptibility to intestinal allergy.

In (A) to (D), adult SPF, GF, and AF B6 mice (N = 4 per group) were adoptively transferred with Cell Trace Violet–labeled naïve Thy1+ OT-II cells and fed with OVA for 7 days; donor OT-II cells from the siLP were then analyzed along with serum IgE specific to OVA. (A) OT-II cells in representative FACS plots of all lymphocytes, with graphs of percentages (left) and total numbers (right) of OT-II cells. (B) Foxp3 up-regulation in OT-II cells in representative FACS plots of gated OT-II cells, with graphs of percentages (left) and total numbers (right) of Foxp3+ OT-II cells. (C) Tbet up-regulation in OT-II cells in representative FACS plots of gated OT-II cells, with graphs of percentages (left) and total numbers (right) of Tbet+ Foxp3 OT-II cells. (D) Graphs of serum OVA-specific IgE (left) and percentages of GATA3+ Foxp3 OT-II cells (right). (E) Neonatal SPF BALB/c mice were weaned onto a normal chow diet or an amino acid diet (AAD), immunized with OVA plus alum, rested for 2 weeks, gavaged with OVA on an every-other-day basis, and observed for diarrhea. Graphs of percentages of mice with diarrhea (left) and diarrhea scores (right) (N = 5 per group). In the graphs in (A) to (C), the numbers in parentheses denote the fold of cell expansion above that found in SPF hosts. Graphs in (A), (B), and (E) are representative of at least two independent experiments, and graphs in (D) show pooled data points from at least two independent experiments. P values were determined by one-way [(A) to (D)] or two-way (E) ANOVA with the Bonferroni post-test. Error bars show SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

The expanded OT-II cells in AF mice, some of which could be due to homeostatic proliferation, readily differentiated into Tbet+ T helper 1 (Th1) cells, leading to the generation of ~400 or ~14 times as many Th1 cells as in SPF or GF mice, respectively; a similar trend, at smaller magnitudes, applied to interferon-γ+ OT-II cells (Fig. 4C and fig. S16A). Moreover, the relative ratio of Treg to Th1 OT-II cell generation in AF mice was one-twentieth that in SPF mice (fig. S16B). Unlike the Th1 response, Th2 and Th17 responses in AF hosts were comparable to that in control SPF hosts. Hence, whereas GF hosts had increased Th2 responses to OVA, AF hosts possessed only background levels of OVA-specific and total serum immunoglobulin E (IgE), a minimal number of GATA3+ OT-II cells, and a similarly moderate percentage of RORγt+ OT-II cells as that observed in control SPF hosts (Fig. 4D and fig. S16, C to F). As expected, Foxp3+ OT-II cells in AF hosts were mostly Nrp-1lo pTreg cells and expressed higher levels of CTLA-4 and IL-10 than host Treg cells, which were mostly tTreg cells (fig. S17, A to C). Unexpectedly, the majority of Foxp3+ OT-II cells were RORγt+ (fig. S16, E and F), indicating that dietary proteins can induce the development of RORγt+ pTreg cells from a fraction of the T cell repertoire in the absence of the commensal microbiota. Nonetheless, the major characteristics of the OT-II response observed in the siLP of AF hosts were also evident in the mesenteric lymph nodes, but not in the spleen, indicating that the pronounced OVA-driven OT-II cell response in AF mice is restricted to the gut-associated lymphoid tissues. The massive OT-II response in AF hosts also was not due to more efficient Ag presentation of OVA related to a lack of competition from other proteins or low vitamin A levels, because similar T cell expansion occurred in a group of AF mice that were fed with a sterilized chow diet just before introducing OVA or injected with retinoic acid during OVA feeding (fig. S18). Overall, these findings indicate that siLP pTreg but not tTreg cells are essential to suppressing a default strong immune response to newly introduced dietary Ags.

Lastly, based on the OT-II response observed above, we tested whether food allergies can be induced in an experimental model. Hence, neonatal SPF BALB/c mice were weaned onto an amino acid diet (because B6 mice are generally resistant to food allergies) and immunized with OVA. Two weeks later, repeated gavage with OVA led to a higher incidence and severity of diarrhea, with a trend toward higher OVA-specific serum IgE, in the amino acid diet–fed mice than in control mice (Fig. 4E and fig. S19). Hence, the lack of pTreg cells led to an increased susceptibility to OVA-induced intestinal allergy.

Our work demonstrates that under normal physiological conditions, macromolecules from the diet induce the bulk of pTreg cell development in the siLP, but not in the cLP. Most siLP pTreg cells develop after weaning onto solid food, presumably reflecting the limited antigenic complexity of milk, an immature state of the immune system in neonatal mice, and/or the presence of components exclusively found in solid food that boost pTreg cell development. The expansion of siLP pTreg cells after weaning appears to be essential to suppressing a default strong immune response to dietary Ags, supporting the view that pTreg cells are important for controlling mucosal inflammatory and allergic responses (6, 10, 24). The data could also explain why children suffer from a higher incidence of food allergies than adults do and why childhood allergies spontaneously dissipate with time (25). Nevertheless, despite their lack of siLP pTreg cells, most neonates do not exhibit food allergies. Such tolerance might reflect multiple mechanisms, including efficient pTreg cell induction by the high levels of TGF-β in milk (23), compensatory suppression by tTreg cells, and/or rapid generation of pTreg cells induced by commensal microbial Ags. For the last of these, it is known that germ-free mice spontaneously produce large amounts of IgE and that this can be prevented by colonization with a complex commensal microbiota (26, 27). Hence, the presence of both diet- and microbe-induced populations of pTreg cells may be required for complete tolerance to food Ags.

Supplementary Materials

www.sciencemag.org/content/351/6275/858/suppl/DC1

Materials and Methods

Figs. S1 to S19

Table S1

References (2832)

REFERENCES AND NOTES

  1. Acknowledgments: We thank A. Macpherson, K. McCoy, and D. Artis for generously providing various strains of germ-free mice to start our colony; T. K. Kim, J. W. Seo, M. O. Lee, H. J. Woo, H. J. Jung, H.J. Ko, J. Kirundi, S. Sakaguchi, and N. Ohkura for technical support; and J. Sprent, S. H. Im, M. H. Jang, D. Rudra, and J. H. Cho for discussions. The data from this study are tabulated in the main paper and in the supplementary materials. This work was supported by project IBS-R005-D1 of the Institute for Basic Science, Korean Ministry of Science, Information/Communication Technology and Future Planning.
View Abstract

Navigate This Article