Deficiency of Indoleamine 2,3-Dioxygenase Enhances Commensal-Induced Antibody Responses and Protects against Citrobacter rodentium-Induced Colitis

Abstract

Indoleamine 2,3-dioxygenase (IDO) is a negative regulator of lymphocyte responses that is expressed predominantly in macrophages and dendritic cells. We detected it at high levels in the small intestine and mesenteric lymph node of young adult mice, suggesting a role in intestinal immunity. Consistent with this idea, we found that IDO-deficient mice had elevated baseline levels of immunoglobulin A (IgA) and IgG in the serum and increased IgA in intestinal secretions. These abnormalities were corrected by a course of broad-spectrum oral antibiotics started at weaning, indicating that they were dependent on the intestinal microbiota. Kynurenine and picolinic acid, two IDO-generated metabolites of tryptophan, were able to inhibit lipopolysaccharide-induced antibody production by splenocytes in vitro, and kynurenine also induced B-cell apoptosis, findings that provide an explanation for the elevated Ig levels in animals lacking IDO. The intestinal secretions of IDO-deficient mice had elevated levels of IgA antibodies that cross-reacted with the gram-negative enteric bacterial pathogen Citrobacter rodentium. In keeping with the functional importance of this natural secretory IgA, the mutant animals were more resistant to intestinal colonization by Citrobacter, developed lower levels of serum Citrobacter-specific IgM and IgG antibodies following oral infection, and had significantly attenuated Citrobacter-induced colitis. Our observations point to an important role for IDO in the regulation of immunity to the gut commensal microbiota that has a significant impact on the response to intestinal pathogens.

Indoleamine 2,3-dioxygenase (IDO) is an intracellular enzyme that catalyzes the initial rate-limiting step in the catabolism of tryptophan via the kynurenine pathway (21). It is expressed in a number of tissues, predominantly in dendritic cells and macrophages, and is up-regulated by immune and inflammatory stimuli. IDO-mediated depletion of tryptophan from the local microenvironment inhibits the proliferation of T cells, NK cells, and possibly B cells (1, 13, 22, 29, 40). The response to tryptophan deprivation in T cells has been shown recently to involve the activation of the GCN2 kinase, a key component of a stress response signaling pathway that can lead to cell cycle arrest or alterations in T-cell differentiation and function (11, 28, 36). The cytotoxicity of tryptophan catabolites such as kynurenine, picolinic acid, and quinolinic acid also contributes to the effects of IDO (13, 35, 40). Because of its ability to inhibit lymphocyte activation and expansion in various ways, IDO is generally considered to be immunosuppressive and anti-inflammatory in function. Indeed, immune-mediated pathology is exacerbated by inhibition of IDO in several experimental models and attenuated by increased expression of the enzyme (1, 4, 14, 15, 23, 42).

In our earlier experiments, we showed that expression of IDO in the murine gut increases significantly with age via a gamma interferon-dependent mechanism (34). Furthermore, levels of IDO in the adult intestine were found to be markedly reduced in mice raised under germfree conditions, suggesting that commensal microorganisms are involved in the normal age-dependent increase in IDO expression. These findings, together with the known immunomodulating functions of the enzyme, suggest that IDO might play a role in regulating mucosal immunity to the intestinal microbiota. We examine this possibility in the present work using IDO-deficient mice. We demonstrate that IDO has significant effects on the B-cell response to commensal bacteria and that these effects alter the outcome of infection with the enteric pathogen Citrobacter rodentium.

MATERIALS AND METHODS

Animals.

Wild-type (WT) C57BL/6 mice were originally obtained from the Jackson Laboratory. The C57BL/6 IDO knockout (KO) mice have been described earlier (22). Colonies of both sets of mice were bred at the Massachusetts General Hospital animal facility and housed under identical specific-pathogen-free conditions. All animal experiments were approved by the institutional Subcommittee on Research Animal Care.

Quantitative RT-PCR for IDO expression.

Total RNA was prepared from segments of the gut as well as the mesenteric lymph node and spleen. After reverse transcription (RT), the cDNA was amplified in the presence of SYBR green (Bio-Rad) by use of conditions and IDO-specific primers described in detail previously (34). The amplification was carried out and monitored in an Opticon 2 DNA engine (MJ Research). The relative expression of IDO was calculated by the 2^−ΔΔCT method with normalization to GAPDH, using the mean of the normalized spleen IDO threshold cycle values as the basis for comparison in the analysis of the gut-associated tissues.

IDO immunohistochemistry.

Five-micrometer frozen sections of tissue were stained with a rat anti-mouse IDO monoclonal antibody (BioLegend) followed by a fluorescein-conjugated goat anti-rat immunoglobulin G (IgG) antibody (Zymed) according to protocols provided by the manufacturers.

Collection of serum and intestinal washes.

Blood was collected from the tail vein into serum separator tubes (Becton-Dickinson) and centrifuged briefly to obtain serum. Aliquots were stored at −20°C. For intestinal washes, the entire small intestine was excised at necropsy and flushed with 3 ml of phosphate-buffered saline containing protease inhibitors. Insoluble material was removed by centrifugation and the supernatant stored in aliquots at −20°C until use.

Estimation of Ig levels.

Enzyme-linked immunosorbent assay (ELISA) kits for IgA, IgG, and IgM (Bethyl Laboratories) were used to quantify the corresponding Igs in serum and intestinal wash samples according to the manufacturer's recommendations. Standard curves were generated using purified Igs run in parallel with the samples. For estimation of Citrobacter-specific antibodies, the ELISA plates were coated with 50 μg/ml of a sonicate of C. rodentium prepared as previously described (8), incubated with serum, and developed with the appropriate anti-Ig secondary antibodies. To identify Citrobacter-specific IgA in intestinal secretions, we incubated the bacteria overnight at 4°C with intestinal washes from WT or KO mice and then washed and detected bound IgA with a fluorescein-tagged anti-mouse IgA antibody (Southern Biotech) before carrying out flow cytometry. For specificity controls, we carried out the same experiment using either intestinal wash that had been depleted of IgA by immunoprecipitation with an anti-IgA antibody (Zymed) and protein G-Sepharose or an irrelevant purified mouse IgA (Southern Biotech).

Antibiotic treatment.

Mice were given ampicillin at 1 g/liter, metronidazole at 1 g/liter, neomycin at 1 g/liter, and vancomycin at 0.5 g/liter in their drinking water for 4 weeks as described previously (33), starting immediately upon weaning at the age of 3 weeks. All antibiotics were obtained from Sigma.

Effect of tryptophan metabolites on B-cell antibody production.

After lysis of erythrocytes, single-cell suspensions of total splenocytes were resuspended in complete medium at 5 × 10⁵ cells per 200 μl. Lipopolysaccharide (LPS) was added to a final concentration of 10 μg/ml and the cultures were incubated for 3 days, with or without the addition of different amounts of kynurenine and picolinic acid (both from Sigma). Cell supernatants were collected and analyzed for Ig levels by ELISA. B-cell death was measured at the end of the culture period by staining aliquots of the cells with fluorochrome-conjugated anti-B220 and annexin V (Pharmingen) and carrying out flow cytometry analysis on a Becton-Dickinson FACScan using CellQuest software.

Citrobacter rodentium infection.

Groups of 6- to 8-week-old WT and KO mice were infected orally with 5 × 10⁸ CFU of C. rodentium strain DBS100 (ATCC 51459) by use of a 21-gauge ball-tipped feeding needle. Body weights and clinical status were recorded daily. Euthanasia and necropsy were performed 12 to 14 days after infection.

Assessment of intestinal inflammation.

At necropsy, portions of the colon were formalin fixed and processed for hematoxylin-eosin staining and histological evaluation. An investigator who was unaware of the genotype of the tissues examined the stained colon sections and evaluated the severity of inflammation using an established scoring system that was described in detail earlier (8). Total colonic RNA was isolated and quantitative RT-PCR carried out using tumor necrosis factor alpha (TNF-α) and GAPDH primers as described previously (34).

Stool cultures.

Stool was collected aseptically, weighed, and homogenized in sterile phosphate-buffered saline. Serial dilutions of the homogenates were plated on MacConkey agar and incubated overnight at 37°C to determine the number of bacteria per mg stool. In some experiments, mucosal colonization by Citrobacter was evaluated by homogenizing fragments of washed colon in sterile 1% Triton X-100 and plating serial dilutions of the homogenates on MacConkey agar. Colony numbers were normalized to the protein concentrations of the homogenates.

Statistical analysis.

The Student t test or nonparametric analysis with the Mann-Whitney test was used, as indicated in the figures, to compare data from different groups. A P value of <0.05 was considered significant.

RESULTS

Elevated serum and intestinal Ig levels in IDO-deficient mice.

We used quantitative RT-PCR to examine IDO expression in different segments of the intestine of young adult C57BL/6 mice as well as associated lymphoid tissue. As shown in Fig. 1A, the highest expression (relative to the spleen) was found in the small intestine and mesenteric lymph node, with lower levels in the cecum and colon. We also analyzed expression of IDO protein in the small intestine and mesenteric lymph node by staining tissue sections with an IDO-specific antibody. As shown in Fig. 1B, bright staining was detected in cells of the intestinal lamina propria as well as the extrafollicular region of the lymph node. No staining was detected when the anti-IDO antibody was used on tissue from IDO KO mice, confirming its specificity (data not shown). Although the level of IDO expression in the colon was low in unmanipulated mice, we found that it was up-regulated following infection with the gram-negative bacterial enteropathogen Citrobacter rodentium (Fig. 1C).

FIG. 1.

IDO expression in the gastrointestinal tract. (A) Total RNA prepared from small intestine (SI), cecum (Ce), colon (Co), mesenteric lymph node (MLN), and spleen (Sp) of adult WT mice was subjected to quantitative RT-PCR with IDO- and GAPDH-specific primers. IDO expression in the various tissues (normalized to GAPDH) is shown relative to that in the spleen. Means and standard errors of expression levels from three or four animals are shown for each tissue. (B) Sections of WT small intestine and mesenteric lymph node immunostained to detect IDO protein expression, viewed with the 10× or 40× objective. (C) IDO mRNA levels in WT colon under control conditions or 12 days after infection with Citrobacter. IDO levels were determined as described for panel A and expressed relative to the control colon levels. Means and standard errors of expression levels for three animals per group are shown.

The basal pattern of IDO mRNA and protein expression revealed by these studies, together with our earlier experiments showing that IDO levels in the gut are dependent on the gut microbiota (34), suggests that the enzyme might have a role in regulating lymphocyte responses to commensal microorganisms. One component of such responses is the induction, in the Peyer's patches and mesenteric lymph node, of antibodies that have specificity for broadly conserved microbial components. These antibodies, mainly of the IgA isotype, help to limit translocation of commensals across the intestinal epithelium (37, 38). A significant fraction of so-called “natural” or “innate” antibodies found in the sera of unmanipulated animals is in fact produced in response to the microbiota of the gut (19, 37). We reasoned, therefore, that if IDO is involved in regulating lymphocyte responses to intestinal commensals, we might see abnormalities of Ig levels in the sera of IDO-deficient mice. Accordingly, we analyzed serum Igs of uninfected young adult (6 to 12 weeks of age) WT and IDO KO mice. We found that although the levels of IgM were similar in the two groups of animals, there was a significant elevation of both IgG and IgA in the IDO-deficient mice (Fig. 2, left). Much of the IgA made in the gut-associated lymphoid tissue (GALT) is ultimately translocated across the intestinal epithelium into the lumen as secretory IgA. Therefore, we also estimated Ig levels in small intestinal secretions and found a significant increase in IgA in the IDO-deficient mice (Fig. 2, right).

FIG. 2.

Serum and intestinal IgA and IgG levels are significantly elevated in IDO-deficient mice. Serum (left) or intestinal (right) Ig concentrations in unmanipulated WT and IDO KO mice were estimated by ELISA at 6 to 12 weeks of age. *, P = 0.0003; **, P = 0.011; ***, P = 0.0037 (Mann-Whitney U test; n = 8 to 14 mice in each group).

Role of the gut microbiota in the abnormal elevation of Igs in the IDO-deficient mice.

To determine if the abnormal increase in IgA and IgG was related to the influence of IDO on commensal-induced responses, we treated a group of IDO-deficient mice for 4 weeks with a broad-spectrum oral antibiotic regimen, starting immediately upon their weaning at the age of 3 weeks. Untreated littermates served as controls. Stool cultures performed at 3 weeks after the start of treatment indicated a >4-log reduction in bacterial counts when the cultures were grown on three different nutrient media under both aerobic and anaerobic conditions, similar to results obtained by others (33) (data not shown). The animals were sacrificed and analyzed 4 weeks after starting antibiotics. Although the Ig levels of these KO mice were generally lower than those analyzed as described for Fig. 2 (probably a reflection of their younger age), we found that antibiotic treatment significantly reduced serum IgA and IgG levels even further (Fig. 3). This observation is consistent with the idea that the abnormal elevation of these Igs in the IDO KO mice is driven by the commensal flora of the gut.

FIG. 3.

Oral antibiotic treatment reduces serum IgA and IgG levels in IDO-deficient mice. Groups of 3-week-old WT and IDO KO mice were treated for 4 weeks with a mixture of oral antibiotics (Abx) as described in the text. Control animals (C) were left untreated. Serum and intestinal Igs were estimated by ELISA at 7 weeks of age. *, P = 0.0079; **, P = 0.0159 (Mann-Whitney U test; n = 5 mice in each group).

Taken together, our observations demonstrating the commensal-dependent expression of IDO in the gut (34), the elevation in baseline serum IgA and IgG in the IDO-deficient mice (Fig. 2), and the correction of these abnormalities by oral antibiotic treatment (Fig. 3) suggest that IDO is involved in a negative-feedback mechanism that limits B-lymphocyte responses to intestinal commensal microorganisms. Somewhat unexpectedly, intestinal IgA was not significantly reduced by the oral antibiotic treatment (Fig. 3). We do not have a definitive explanation for this finding, but it could be related to changes in transepithelial transport or intralumenal stability of IgA in the commensal-depleted animals.

Inhibition of B-lymphocyte responses by IDO-generated tryptophan metabolites.

We went on to explore the mechanism by which IDO inhibits B-cell responses by examining the effects of IDO-generated metabolites of tryptophan on LPS-induced Ig production in vitro. As shown in Fig. 4, two such metabolites, kynurenine and picolinic acid, inhibited IgM secretion by both WT and IDO KO splenocytes. This inhibition is likely to involve the induction of cell death, since we found that kynurenine produced a dose-dependent increase in the proportion of B220⁺ B cells that expressed the apoptotic marker annexin V (Fig. 5). These findings are in keeping with an earlier study showing that IDO-generated tryptophan metabolites induced the death of B cells, T cells, and NK cells in vitro (40). Although high doses of kynurenine and picolinic acid are required to inhibit IgM production when they are used individually, a combination of the two has similar effects even at lower concentrations (Fig. 4), suggesting that our observations are likely to be relevant to the in vivo situation, where multiple tryptophan metabolites are generated together and can act in concert. Indeed, Terness et al. have shown that by using a mixture of five different metabolites, B-cell death is induced by concentrations of the individual compounds as low as 8 to 32 μM, approximating those that occur in vivo during IDO activation (40). These findings suggest that the mechanism by which IDO inhibits B-cell responses involves the cytotoxic effects of IDO-generated tryptophan metabolites, although the possibility of contributions from tryptophan depletion cannot be excluded.

FIG. 4.

Inhibition of LPS-induced IgM production by tryptophan metabolites. Splenocyte cultures from WT or KO mice were stimulated with 10 μg/ml LPS, with or without the addition of kynurenine or picolinic acid or a mixture of the two at the concentrations indicated. Supernatant IgM was estimated by ELISA. Combined results of three separate experiments, showing means and standard errors, expressed as a percentage of IgM produced by LPS-stimulated cultures without added tryptophan metabolites.

FIG. 5.

Induction of B-cell apoptosis by kynurenine. Splenocyte cultures from WT or IDO KO mice were stimulated with LPS as described for Fig. 4, with or without kynurenine added at the indicated concentrations. The cells were then stained with fluorochrome-conjugated annexin V and anti-B220 and subjected to flow cytometry analysis. Numbers in the upper right quadrant of each density plot indicate B220⁺ annexin V⁺ cells as a percentage of the total lymphoid population. FITC, fluorescein isothiocyanate.

Attenuation of Citrobacter-induced colitis in IDO-deficient mice.

A proportion of the antibodies induced by the intestinal microbiota is reactive against determinants conserved among broad groups of bacteria and provides a first line of defense against pathogens (19, 32, 37). The importance of such antibodies in mucosal protection was demonstrated recently by experiments showing that mice lacking the ability to translocate innate secretory IgA across the intestinal epithelium were more susceptible to oral Salmonella enterica serovar Typhimurium infection than were WT animals (43). To determine whether the elevated levels of IgA and IgG in the IDO-deficient mice might enhance their resistance to enteric pathogens, we first asked if these antibodies cross-reacted with Citrobacter rodentium, a gram-negative rodent pathogen similar to human enteropathogenic Escherichia coli (27). The use of Citrobacter seemed particularly appropriate for our experiments, since humoral immunity has been shown to play an important role in protection against this organism (5, 18, 44).

In initial experiments with a Citrobacter-specific ELISA, we detected either no or very low levels of antibodies reactive with this organism in the sera and intestinal washes of WT and IDO-deficient mice (data not shown). However, when we used a flow cytometry-based assay, we readily identified Citrobacter-reactive IgA antibodies in intestinal washes, with consistently higher levels of fluorescence when washes from the KO animals were used (Fig. 6, left and middle). We confirmed the specificity of this assay by demonstrating that no fluorescence above background was detected when intestinal wash immunodepleted of IgA was used (Fig. 6, right). Furthermore, no fluorescence above background was detected when a purified irrelevant mouse IgA was used (data not shown), making it unlikely that the Citrobacter was binding to IgA via interactions with the Fc region of the antibody.

FIG. 6.

Intestinal secretory IgA cross-reacts with Citrobacter. Intestinal washes from WT or IDO KO mice were incubated with Citrobacter and bound IgA was detected with a fluorescein-tagged anti-IgA antibody. KO wash immunodepleted of IgA was used as a control.

We then infected WT and IDO KO mice orally with Citrobacter. There were appreciable differences in the responses of the two groups of animals. By 5 or 6 days after infection, the WT mice had decreased activity, ruffled fur, and hunched posture, whereas the IDO-deficient mice appeared well throughout the course of the experiment. At the time of sacrifice on day 12, most of the WT mice had lost a significant amount of weight, whereas the mutant animals had little or no weight loss (WT, 89.72% ± 8.19% of starting body weight; KO, 101.83% ± 4.57% of starting body weight; P = 0.005 by Mann-Whitney test). Examination of hematoxylin-eosin-stained sections of the colon from the infected animals demonstrated that edema, cellular infiltration, and epithelial damage were all reduced in the IDO-deficient mice (Fig. 7A). This visual impression was substantiated by formal histopathological scoring of multiple sections by an investigator who was unaware of the genotype of the tissues, which revealed that the KO colons had significantly lower scores than the WT (Fig. 7B). In keeping with the histopathology, colonic TNF-α expression was also significantly less in the KO mice (Fig. 7C). Thus, based on all these parameters, the intestinal inflammatory response to Citrobacter was significantly attenuated by the deficiency of IDO. In keeping with the notion that the increased levels of natural secretory IgA contributed to this attenuation, we found that intestinal colonization by the pathogen, as indicated by fecal CFU at 7 days postinfection, was significantly reduced in the IDO-deficient mice (Fig. 8A). The reduced Citrobacter colonization of the KO intestine was also noted at the time of sacrifice at 12 days postinfection [WT mice, (5.05 ± 2.9) × 10⁵ CFU/mg stool, (3.82 ± 5.07) × 10⁴ CFU/mg colonic tissue; KO mice, (0.80 ± 0.93) × 10⁵ CFU/mg stool, (0.70 ± 0.42) × 10⁴ CFU/mg colonic tissue] although the difference from colonization in the WT intestine was more variable and did not reach statistical significance. Furthermore, serum Citrobacter-specific IgM and IgG levels were also significantly reduced in the infected mutant animals, indicating a lower level of stimulation of systemic humoral immunity by the pathogen (Fig. 8B). Citrobacter-specific IgA was not detected in the sera of either WT or KO mice (data not shown).

FIG. 7.

Citrobacter-induced enterocolitis is attenuated in IDO-deficient mice. (A) Hematoxylin-eosin-stained sections of colon from WT or KO mice at 12 days after Citrobacter infection visualized with a 10× or 20× objective. (B) Histopathology scores of infected WT or KO colons. *, P = 0.016 (Mann-Whitney U test; n = 5 mice of each genotype). (C) Means and standard errors of colon TNF-α transcript levels (normalized to GAPDH and expressed relative to the WT control) in WT and IDO KO mice 12 days after infection with Citrobacter. *, P = 0.002 (Student's t test; n = 4 or 5 mice of each genotype).

FIG. 8.

(A) Intestinal colonization by Citrobacter is reduced in IDO-deficient mice. Fecal Citrobacter numbers in WT and IDO KO mice 7 days after infection. *, P = 0.0003 (Mann-Whitney U test; n = 10 stool samples from six mice of each genotype). (B) Citrobacter-specific IgM and IgG are reduced in infected IDO-deficient mice. Serum Citrobacter-specific Igs were estimated by ELISA in uninfected mice (Con) or in Citrobacter-infected mice (CB) 12 days after infection. *, P = 0.029; **, P = 0.016 (Mann-Whitney U test; n = 4 mice of each genotype).

DISCUSSION

Our observations have revealed a new aspect of IDO function, i.e., the control of B-lymphocyte responses to the commensal microbiota of the gut. Commensals colonize the gut soon after birth, after which they increase in number and complexity until they reach a steady state that is characteristic of both the host species and the individual (19, 32). A small number of these bacteria penetrate the epithelium and are carried by dendritic cells to the mesenteric lymph node, where they activate lymphocyte responses (20). One outcome of these interactions is the induction of Ig class switching in B cells, mainly to IgA, as well as the up-regulation of molecules involved in the homing of lymphocytes to the gut, properties that are conferred in part by retinoic acid and reactive nitrogen intermediates produced by GALT dendritic cells (25, 26, 41). The activated B cells migrate from the mesenteric lymph node to the intestinal lamina propria, and the Ig that they secrete is transcytosed across the epithelium by the polymeric Ig receptor (17). The secretory Ig binds to luminal bacteria and helps to reduce, although not completely eliminate, further translocation of these organisms across the epithelial monolayer (19, 38). B cells in the GALT are thus under conditions of continuous stimulation by commensals. This stimulation can be T dependent or T independent, involves both antigen-specific and nonspecific mechanisms (acting via the B-cell antigen receptor and pattern recognition receptors, respectively), and contributes to the pool of natural antibodies, a fraction of which is reactive with commensal microbial antigens (3, 7, 16, 19, 39). Such chronic, polyclonal B-cell activation must be tightly regulated in order that it does not get out of control. The data that we have presented suggest that IDO, expressed by dendritic cells and/or macrophages in the GALT, is one component of this regulatory mechanism. To our knowledge, this is the first demonstration of an in vivo, “physiological” function for IDO, although it is consistent with numerous earlier studies that have used pharmacological inhibitors or overexpression approaches to demonstrate the inhibitory effect of this enzyme on lymphocyte proliferation (1, 13, 22, 28, 29, 40).

The effect of IDO on commensal-driven antibody production may be the result of a direct influence on B-cell proliferation or an indirect consequence of IDO-dependent alterations in the T-cell response. Commensal-driven production of IgA in the GALT can be either T dependent or T independent (19), so both mechanisms may operate. The existing information in the literature on the influence of IDO on B-cell responses is contradictory, with some studies showing that IDO-generated metabolites can directly inhibit B-cell proliferation, while others indicate no such effects (13, 15, 40). The in vitro studies we have reported (Fig. 4 and 5) are consistent with the idea that IDO can directly influence B-cell responses via the generation of cytotoxic tryptophan metabolites. However, further work will be required to clarify this issue, particularly examination of B-cell responses to the commensal flora in IDO- and T-cell-double-deficient mice.

One apparent benefit of the increase in natural antibody that occurs in the IDO-deficient mice is enhanced resistance to the colitis caused by infection with Citrobacter rodentium (Fig. 7 and 8). This rodent pathogen usually colonizes the surface of the intestinal epithelium and causes a transient inflammatory response that resolves as the development of adaptive immunity leads to clearance of the organism (27). Both Th1- and T-dependent B-cell responses are involved in elimination of the infection (5, 6, 18). Although earlier work has suggested a less significant function for IgA in protection against this pathogen, IgA-deficient mice did have a slight delay in clearing the infection compared to what was seen for WT animals (18). Our observations support the idea that natural secretory IgA provides resistance to intestinal colonization by Citrobacter, similar to recent findings with Salmonella (43). In addition, IgA can have direct effects on intestinal inflammation that may contribute to the attenuated Citrobacter-induced colitis in the IDO-deficient animals. These anti-inflammatory effects include the ability to neutralize LPS within epithelial cells and the delivery of inhibitory signals to cells of the innate immune system (9, 12, 30, 31). The elevated serum levels of natural IgG in the IDO-deficient mice may also help to reduce the severity of the Citrobacter-induced intestinal inflammation, since this Ig has been shown to play an important role in resistance to the infection, possibly by protecting the intestinal epithelium against abnormal penetration by the pathogen (5, 18, 44).

At first glance, it seems counterintuitive that mice lacking IDO, a negative regulator of lymphocyte responses, have abnormally low levels of Citrobacter-specific serum IgM and IgG following infection (Fig. 8B). However, this unexpected result can be explained in the context of our observation that a major outcome of IDO deficiency appears to be exaggerated commensal-induced antibody production, which is consistent with the established lymphocyte-inhibitory function of the enzyme. The elevated level of natural secretory IgA that is one manifestation of the exaggerated B-cell response can contribute in two ways to the abnormally low serum levels of Citrobacter-specific IgM and IgG in the KO mice—one, by inhibiting intestinal colonization by the pathogen, and two, by reducing priming of systemic immune responses. It will be of considerable interest to compare changes in Citrobacter-specific B- and T-cell responses in the WT and IDO KO mice over longer periods of infection to elucidate the role of IDO-dependent effects on pathogen clearance and resolution of the colitis. It is also important to mention that definitive confirmation of the role of elevated natural antibodies in the relative resistance of the IDO-deficient mice to Citrobacter infection will require experiments with double KO mice that lack both IDO and either B cells or IgA. Until the results of such studies are available, we cannot exclude the possibility that the attenuation of Citrobacter-induced colitis in the IDO KO mice may reflect B-cell-independent effects of IDO.

Gurtner et al. have reported that pharmacological inhibition of IDO leads to exacerbation of the colitis induced by administration of trinitrobenzene sulfonic acid (TNBS) (14), a finding that appears to be contradictory to the observations that we describe here. However, there are significant differences between their experimental model and ours that could account for the discrepant results. First, Gurtner et al. instituted 1-methyl tryptophan-mediated inhibition of IDO in 6-week-old WT mice for a period of 10 days. This pharmacological approach may have been too late and too brief to have had any effects on the response to commensal microorganisms, unlike congenital deficiency of IDO. Second, TNBS-induced colitis is initiated by chemically induced epithelial damage and the subsequent sensitization and activation of T cells reactive to components of the commensal flora and/or haptenated colonic proteins. It is unlikely that natural antibodies would have significant protective effects in such circumstances. In contrast, Citrobacter-induced colitis depends on colonization of the epithelial surface by the pathogen, a process that is likely to be susceptible to antibody-mediated inhibition. Finally, as discussed in more detail below, a second IDO gene, IDO2, has been identified recently (2, 24). Interestingly, 1-methyl tryptophan preferentially targets IDO2 activity (24), raising the possibility that the results obtained by Gurtner et al. reflect the effects of inhibiting this enzyme rather than IDO. Clearly, it will be important to determine what influence IDO deficiency has on TNBS-induced colitis. These studies are planned as part of further characterization of the IDO KO phenotype.

In addition to its influence on lymphocytes, IDO has been reported to inhibit the growth of some microorganisms by virtue of its ability to deplete tryptophan in the microenvironment (10). Although such antimicrobial actions could not directly explain the decrease in Citrobacter colonization in the IDO-deficient mice, it could have indirect effects via changes in the commensal microbiota. We have examined the composition of the fecal flora in WT and IDO KO mice by using quantitative PCR to amplify 16S rRNA genes of several groups of commensal organisms and were unable to find significant differences, at least at this level of analysis (unpublished data). Therefore, we think it is unlikely that IDO-dependent changes in the commensal flora play a role in altering the outcome of Citrobacter infection in the KO animals.

Very recently, two groups of investigators have independently described the existence of an IDO-like gene in both humans and mice (2, 24). This gene, which has been named IDO2, has a more restricted distribution than IDO but was found to be expressed in dendritic cells and to be up-regulated by gamma interferon (24). The biological functions of IDO2 have not been delineated, but the overlap between its properties and those of IDO raises the possibility that it may compensate for IDO deficiency in some situations. Nevertheless, the clear immunological phenotype that we have described for the IDO KO mice indicates that at least some functions of the two enzymes are nonredundant. Further work will be required to tease apart the contributions of each to the regulation of T- and B-lymphocyte responses.

In conclusion, the observations reported here indicate that IDO has a previously unrecognized function in regulating commensal-induced IgA and IgG production and that alterations in this function can have a significant influence on the response to infectious agents such as Citrobacter. Pharmacological targeting of IDO activity may represent a new approach to manipulating intestinal immunity and controlling the pathology caused by enteric pathogens.