Intricate regulation of tolerance to the intestinal commensal microbiota acquired at birth is critical. We hypothesized that epithelial cell tolerance toward early gram-positive and gram-negative colonizing bacteria is established immediately after birth, as has previously been shown for endotoxin. Gene expression in the intestine of mouse pups born to dams that were either colonized with a conventional microbiota or monocolonized (Lactobacillus acidophilus or Eschericia coli) or germ free was examined on day 1 and day 6 after birth. Intestinal epithelial cells from all groups of pups were stimulated ex vivo with L. acidophilus and E. coli to assess tolerance establishment. Intestine from pups exposed to a conventional microbiota displayed lower expression of Ccl2, Ccl3, Cxcl1, Cxcl2, and Tslp than germ-free mice, whereas genes encoding proteins in Toll-like receptor signaling pathways and cytokines were upregulated. When comparing pups on day 1 and day 6 after birth, a specific change in gene expression pattern was evident in all groups of mice. Tolerance to ex vivo stimulation with E. coli was only established in conventional animals. Colonization of the intestine was reflected in the spleen displaying downregulation of Cxcl2 compared with germ-free animals on day 1 after birth. Colonization reduced the expression of genes involved in antigen presentation in the intestine-draining mesenteric lymph nodes, but not in the popliteal lymph nodes, as evidenced by gene expression on day 23 after birth. We propose that microbial detection systems in the intestine are upregulated by colonization with a diverse microbiota, whereas expression of proinflammatory chemokines is reduced to avoid excess recruitment of immune cells to the maturing intestine.
- mucosal immunology
- epithelial cells
the intestinal microbiota represents a large load of foreign antigen that must be tolerated by the immune system. Remarkably, this microbiota plays a crucial role in immunogenesis, with specific microbial constituents currently being assigned different roles (7, 17). The establishment of tolerance to microbial stimuli after birth must therefore be accompanied by appropriate signals necessary for the layout of a functioning immune system. Despite progress in mapping the metagenome of the intestinal system, conflicting reports exist on the bacteria necessary to shape the immune system and maintain tolerance. Segmented filamentous bacteria (7) and purified microbe-associated molecular pattern molecules such as a Bacteroides fragilis polysaccharide (17) have been shown to be sufficient to establish normal numbers of T cell subsets in the intestine and in the spleen, respectively. In contrast, a full conventional microbiota is required for induction of oral tolerance to food antigens (24), whereas little is known about which signals are required for establishment of tolerance to the microbiota itself.
Colonization of adult germ-free animals rapidly infers maturation of the compromised immune system, locally in the intestine as well as systemically (10, 17). In the natural setting, where colonization occurs at birth, this colonization-induced maturation is accompanied by age-dependent maturation of the intestine and lymphoid tissues. Intestinal epithelial cells (IECs) are sentinels of the bacterial microbiota of the gut and communicate with the immune system via secretion of cytokines and chemokines. As IEC chemokines recruit myeloid and lymphoid cells to the gut, they are crucial in directing the early maturation of both the local and systemic immune system (9). IECs exposed to microbe-associated molecular patterns become insensitive to further bacterial stimuli (14, 32), indicating a critical mechanism in maintaining noninflammatory conditions in the gut. Part of the response of epithelial cell lines to TLR2 and TLR4 ligands is transfer of Toll-like receptors (TLRs) from the apical to the cytosolic compartment, causing tolerance to subsequent ligand challenges (2, 23). In vivo, the intracellular TLR4 receptors are responsive, and the major LPS-tolerizing mechanisms in neonate intestinal epithelial cells have been shown to be microRNA-induced ubiquitination and degradation of IRAK-1 (5, 14) and LPS-dependent inhibition of p38 MAPK via NF-κB regulated MAPK phosphatase-1 (32). Certain IEC-secreted factors, such as cathelin-related antimicrobial peptide (19), are not regulated by colonizing bacteria, but levels secreted evolve with age. Hence, although it is well established that microbiota-enterocyte interactions play a key role in maturation of the gut immune system, few studies have addressed the chemokine and cytokine response of IECs to stimulation at time of birth and the consequences for immune maturation (14).
Lactobacilli and Escherichia coli strains are as facultative anaerobic microorganisms successful early colonizers of the sterile gastrointestinal tract (27). Single strains of these bacteria are insufficient to establish tolerance to oral antigens (a process involving both the innate and the adaptive parts of the immune system) in adult animals (24). We hypothesized that in contrast, to prevent excessive inflammation in the immature intestine, exposure to either of these bacteria (originating from the monocolonized mothers) would render IECs from the pups nonresponsive to subsequent ex vivo stimulation by the same bacteria. We used the strains Lactobacillus acidophilus NCFM and E. coli Nissle, which are potent modulators of enterocyte responses in vitro (34). To characterize the initial response to these intestinal bacteria, we measured the expression of cytokine, chemokine, and pattern recognition signaling genes. We demonstrate that only conventional colonization induces IEC tolerance to subsequent ex vivo stimulation, that this tolerance induction was concomitant with increased intestinal TLR2 expression, with reduced expression of chemokines and thymic stromal lymphopoietin (TSLP) in the intestine and in the spleen, and that this was followed by reduced expression of genes involved in antigen presentation in mesenteric lymph nodes (MLN).
MATERIALS AND METHODS
Germ-free and conventional Swiss Webster mice were purchased from Taconic (Lille Skensved, Denmark) and kept in germ-free isolators or housed under specific pathogen-free conditions, respectively. Absence of bacteria in the germ-free mice was confirmed by cultivation of fecal samples. Six sets of two germ-free females and one male were housed together until plugs were observed. Similarly, two breeding sets were set up for conventionally colonized mice. Monocolonization of female mice with E. coli Nissle and L. acidophilus NCFM was performed 7 days after mating by applying 5 × 108 CFU/ml in 0.5 ml PBS suspension orally and 0.5 ml to the abdominal skin. Four litters spontaneously delivered from the four dams in each group [conventional, germ-free, and monocolonized (E. coli Nissle or L. acidophilus NCFM)] were used for the experiment. On the morning of the night when pups were born, postnatal day 1 (PND1), and on postnatal day 6 (PND6), four pups per litter were euthanized. Spleen and a segment of the distal ileum (3 cm from cecum and up) from two pups per litter were dissected and frozen in RNAlater (Qiagen, Hilden, Germany). The small intestines from two other pups were used for determination of gene expression in isolated epithelial cells. IECs were isolated as described below for ex vivo studies, immediately pelleted. and resuspended in RNAlater for storage at −80°. LPS content in stomachs from PND6 pups was determined by the Pyrochrome kit (Associates of Cape Cod, East Falmouth, MA). From two conventional and two germ-free pups from separate litters, MLN, and popliteal lymph nodes (PLN) were dissected at postnatal day 23 (PND23) and frozen in RNAlater. The mouse experiment was performed under a license to Department of Microbiology, National Food Institute, from the Danish Council for Animal Experimentation (Dyreforsøgstilsynet).
Preparation of bacterial suspensions.
L. acidophilus NCFM was grown anaerobically in de Man, Rogosa, and Sharpe broth (MRS, Merck, Darmstadt, Germany) and E. coli Nissle aerobically in Luria-Bertani broth (LB, Merck) overnight at 37°C. The cultures were harvested, washed two times in sterile phosphate-buffered saline (PBS, Lonza, Basel, Switzerland), resuspended in PBS and frozen at −80°C. For use in ex vivo experiments, bacteria were killed by a 40-min UV exposure prior to freezing. The endotoxin concentration in L. acidophilus NCFM preparations were <0.10 EU/ml measured by limulus amoebocyte assay (Associates of Cape Cod).
Isolation of epithelial cells for ex vivo stimulation.
At PND6, epithelial cells were isolated for ex vivo stimulation studies from small intestines of two to three pups per litter. The small intestines were placed in Hanks' buffered saline (HBSS, Lonza), opened longitudinally, and cut in small pieces. The epithelial cells were detached from the underlying tissue by incubation in fresh HBSS containing 2 mM EDTA at 37°C for 10 min. Residual tissue was removed by use of a 70-μm filter. Cells were washed in cold PBS and resuspended in culture medium [RPMI 1640 supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM l-glutamine, and 10% (vol/vol) heat-inactivated FCS; all from Lonza]. Cells were seeded in 48-well tissue culture plates (Nunc, Roskilde, Denmark) at 4×105 cells per 500 μl per well. To each well was added either 50 μl of culture medium (unstimulated), L. acidophilus NCFM, or E. coli Nissle suspensions, to a final concentration of 30 μg/ml. The IECs were stimulated for 2 h at 37°C in 5% CO2 and subsequently frozen in RNAlater (Qiagen). The purity of IECs was assessed by staining for the leukocyte marker CD45 (phycoerythrin-labeled rat anti-mouse CD45 purchased from Abcam, Cambridge, UK) by flow cytometry. IECs contained 1.3 ± 0.4% CD45+ cells compared with 0.5 ± 0.3% CD45+ cells in IECs stained with an isotype control antibody (IgG2b). Viability and cell numbers of IECs were determined by propidium iodide exclusion using a NucleoCounter (Chemometec, Allerød, Denmark). IECs were >70% viable after isolation.
RNA isolation and amplification.
Cell samples were spun at 3,000 g, 5 min at 4°C to remove RNAlater. Tissue was removed from RNAlater and homogenized by use of a rotor stator in RLT buffer (Qiagen). RNA from cell pellet and tissue homogenate was extracted by using the RNeasy Mini Kit from Qiagen following the supplier's protocol. The quantity and purity of extracted RNA was evaluated by Nanodrop spectroscopy (Wilmington, DE). cDNA was produced from ∼500 ng total RNA by using High-Capacity cDNA Reverse Transcriptase Kit (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions.
Gene expression analysis by real-time PCR.
A custom TaqMan Low Density Array with 24 TaqMan Gene Expression Assays (Applied Biosystems, Table 1) was used for gene expression analysis of terminal ileum and intestinal epithelial cells of mouse pups born to dams either conventionally colonized, monocolonized with L. acidophilus or E. coli, or germ free, on PND1 and PND6. An inventoried TaqMan Mouse Immune Array (Part Number 4367786, Applied Biosystems) containing 90 TaqMan Gene Expression Assays of genes known to have implications on the immune response was used for expression analysis of lymph nodes on PND23. The 24 genes on the custom arrays (Table 1) were chosen on the basis of in vitro studies (34) and preliminary comparisons of germ-free and colonized mice by using the Immune Array. PCR was performed as previously described (34). Briefly, to each cDNA sample (50 ng RNA in 50 μl) was added 50 μl TaqMan Universal PCR Master Mix (Applied Biosystems). Arrays were run in standard mode by using the 7900HT Fast Real-time PCR system (Applied Biosystems). Single gene expression of Cxcl2, Il6, Il10, Tnf, and Actb was analyzed in IECs and Cxcl2 and Actb in spleen. For each sample, 2 μl cDNA (3 ng/μl) was amplified in duplicates by using universal fast thermal cycling parameters (Applied Biosystems) and TaqMan Fast Universal PCR Master (Applied Biosystems) in a total reaction volume of 10 μl. Fold changes in gene expression were calculated by the comparative cycle threshold (CT) method. The expression of target genes was normalized to a reference gene [ΔCT = CT(target) − CT(reference)]. We compared Actb and 18S rRNA, which gave comparable results, and chose to use Actb as reference gene. For tissue samples and isolated IEC, the average gene expression of germ-free PND1 mice was used as calibrator. For ex vivo stimulated IECs, unstimulated germ-free cells were used as calibrator. Fold change in gene expression was calculated as 2−ΔΔCT where ΔΔCT = ΔCT(sample) − ΔCT(calibrator).
GraphPad Prism version 5.01 (GraphPad Software, San Diego, CA) was used to perform two-way ANOVA on all four treatment groups with Bonferroni posttest, except for data in Figs. 4 and 5, which were analyzed by Student's t-test. Although fold increase is plotted in gene expression experiments, statistical analysis was performed on ΔCT values [ΔCT = CT(target) − CT(reference)] since these are assumed normally distributed as opposed to the fold change values. Principal component analysis (PCA) was performed on −(ΔCT) values using Latentix 2.00 (Latent5 Aps, Copenhagen, Denmark). In PCA plots, similar location (top/bottom or left/right in the plots) of a gene symbol and a sample marker therefore indicates low ΔCT values, i.e., high relative expression of that particular gene in the sample. Values were normalized by subtracting the mean and dividing by the standard deviation (autoscaling) prior to modeling. Il6 was excluded from all models because its expression was not detected in 25% of the samples.
The presence of a conventional microbiota, but not L. acidophilus or E. coli, affects intestinal expression of cytokines, chemokines, and signaling molecules in neonatal mice.
Microorganisms are involved in the recruitment of immune cells to the gut. A too vigorous influx and activation of neutrophils in the newly colonized gut may, however, be hazardous, leading to acute inflammation and tissue necrosis in the epithelium. Mechanisms that initially help avoiding too rigorous responses are needed to establish tolerance to gut microbes. Moreover, in this period the gut epithelium develops rapidly, in particular in the presence of microorganisms (18). Accordingly, we investigated the effect of colonization on day 1 and day 6 postnatally on gene expression of a number of chemokines and other genes involved in recruitment and activation of immune cells as well as genes involved in microbial recognition and response. A conventional microbiota caused significant alterations in intestinal gene expression of a subset of the measured cytokine and chemokine genes on PND1 and PND6, whereas monoassociation with E. coli or L. acidophilus did not, compared with pups remaining germ-free (Fig. 1A, monoassociated groups not shown when not significantly different from germ-free). On PND1, Cxcl2 and Tslp displayed a lower expression in conventionally colonized pups than in germ-free pups. In contrast, on PND6, Tslp expression was higher in conventionally colonized pups than in germ-free pups. The genes encoding CCL19 [a chemoattractant for lymphocytes and dendritic cells (11)] and TNF-α were upregulated in conventionally colonized intestines compared with the sterile intestines, but only on PND6. Monocolonization with L. acidophilus was the only treatment causing a lower expression of Il7 compared with germ-free conditions. Gene expression of most of the assayed receptor genes and signaling molecules was only significantly altered by a conventional microbiota (Fig. 1B, monoassociated groups not shown) and, except for Tlr2, only on PND1. Only expression of Clec7 and Stat4 displayed a different pattern, being downregulated by E. coli monocolonization, compared with no colonization. Intestinal gene expression of Tlr2, Tlr4, Ptprc, Ikbkb, Nfkb1, Nfkb2, Ly96, and Irak1 was increased by conventionalization compared with germ-free mouse pups. Expression of Ccl19, Tslp, Tnf, Tlr2/4, Ptprc, Ikbkb, Nfkb1/2, Ly96, and Irak1 was furthermore increased with age, probably due to growth of epithelial tissue after birth.
When the individual variation is large relative to the average colonization-induced differences, it can be useful to employ PCA for interpretation of data. In this way, we investigated correlations between all genes measured in the individual samples on PND1. In the PCA score plot (Fig. 2A) single animals are indicated as colored “sample markers.” The variables measured (in this case expression of genes) are indicated as gene symbols in the corresponding loading plot. If a sample marker is located in a area of the score plot that corresponds to a gene symbol in the loading plot, this sample has a high expression of that particular gene. In our analysis, samples from conventionally colonized mice clustered to the right, similarly to most receptor and cytokine genes (Fig. 2A). This indicates that these samples were very similar and displayed a high expression of especially Il10, Stat4, Nfkb1/2, Clec7a, Ccl19, Tnf, and Tlr2 compared with samples from non- and monocolonized animals. In contrast, the majority of the samples from germ-free mice were found in the lower part of the plot, similarly to Tslp, Ccl2, Ccl3, Cxcl1, and Cxcl2 gene symbols, indicating a high expression of this group of genes in germ-free pups. Samples from monoassociated animals were distributed in a pattern resembling samples from germ-free mice. The same grouping of genes and samples was evident when samples of isolated IECs were analyzed (data not shown). When intestinal samples from both PND1 and PND6 were included in a PCA model, PND1 sample markers were found segregating to the left and PND6 samples in the upper right of the plot (Fig. 2B), showing that the expression of Ccl19, Tlr2/4, Tollip, Nfkb2, and most other genes increased with age in all groups. Again, Tslp, Ccl2, Ccl3, Cxcl1, and Cxcl2 clustered separately from the majority of genes, indicating a different age-dependent induction of these genes (lower expression with higher age). Again, the same pattern was seen when isolated IEC were analyzed (data not shown). Taken together, our data demonstrate that in particular genes encoding chemokines involved in recruitment of neutrophils and monocytes were downregulated by colonization with a conventional microbiota immediately after birth (PND1), followed by a significant upregulation of CCL19, which are involved in recruitment of dendritic cells and lymphocytes, key players in the adaptive immune response, and of TSLP, which modulates dendritic cells to become inducers of tolerance (31) at PND6.
Only a conventional microbiota induces tolerance toward restimulation.
To elucidate how neonatal exposure to single bacteria or a full microbiota influences IEC responsiveness, we studied the expression of four cytokines and chemokines in IECs isolated from differentially colonized mice at PND6 and stimulated ex vivo with E. coli or L. acidophilus (Fig. 3). Ex vivo exposure to E. coli significantly increased the expression of Cxcl2, Tnf, Il6, and Il10 in IEC from germ-free mice (Fig. 2A). In contrast, in unstimulated IECs from E. coli monoassociated mice, gene transcription was at the same high level as in E. coli-stimulated cells (Fig. 3A). This could be interpreted as induction of endotoxin tolerance; however, it probably rather indicates a high baseline gene transcription unmodifiable by bacteria added ex vivo. Endotoxin present in the gastrointestinal tract of the E. coli-associated mouse pups may cause this high baseline activation. Consistent with this, we measured the endotoxin content in stomachs of mice at PND6 and found high levels in stomachs of E. coli-exposed animals (1,942 ± 302 EU/stomach, n = 3), low levels in conventional mice (2.6 ± 1.9 EU/stomach), and no detectable endotoxin in stomachs of germ-free animals or L. acidophilus-associated animals.
IECs isolated from conventionally and L. acidophilus-associated pups responded similarly to cells from germ-free animals to ex vivo L. acidophilus stimulation with increased Il10 expression and upregulated all cytokine and chemokine genes in response to E. coli stimulation (Fig. 3B). Expression of Cxcl2 was significantly blunted in response to E. coli stimulation in IECs from conventional mice (∼3-fold increase with stimulation) compared with IECs from germ-free mice (∼6-fold increase with stimulation), suggesting that endotoxin tolerance was induced upon acquisition of a normal microbiota. We have previously measured the concentration of secreted CXCL2 18 h after stimulation of epithelial cells by ELISA and could confirm a close relationship between the gene expression and the secreted protein (34). IL-6 and TNF-α induction by E. coli was not significantly blunted in cells from conventionally colonized animals. Enterocytes from pups of mice monocolonized with L. acidophilus responded just as potently to L. acidophilus and E. coli stimulation as IECs from germ-free mice, suggesting that no tolerance to gram-positive or gram-negative bacteria was induced in this group. The data even indicate that L. acidophilus colonization induced hyperresponsiveness toward endotoxin stimulation, since E. coli-induced Il6 expression was higher in cells from L. acidophilus-associated pups than in germ-free animals.
Cxcl2 expression is lower in the spleen of conventional and E. coli-associated mice on the day of birth compared with germ-free animals.
The immature gut is believed to lack integrity, and we speculated that IECs in a period shortly after birth may be bypassed by bacteria translocating to other tissues and that the early response to intestinal colonization may therefore not be limited to IECs. Consequently, we assayed the expression of Cxcl2 in the spleens from germ-free, monocolonized, and conventionally colonized mouse pups on PND1, since neutrophils are recruited to the spleen during pathogen translocation in mice (21), probably because of upregulation of Cxcl2. In addition, Cxcl2 expression can serve as a measure of NF-κB activation (14). Surprisingly, no activation reminiscent of inflammation in the spleen was observed. Instead, expression of Cxcl2 was decreased in the spleen of mouse pups born to mothers harboring a conventional microbiota compared with germ-free pups (Fig. 4A), and, in contrast to the ileal response (Fig. 4B), monoassociation with E. coli also reduced Cxcl2 expression in the spleen. Although purely speculative, our data may indicate that bacterial translocation takes place immediately after birth and that mechanisms exist that help tolerating the presence of bacteria in the spleen.
MLNs of conventionally colonized mice display reduced gene expression of genes involved in antigen presentation compared with PLNs.
Immune cell tolerance to the intestinal microbiota is believed to occur through regulatory priming of lamina propria and MLN dendritic cells by TSLP produced by epithelial cells (25) and through provision of retinoic acid to CD103+ dendritic cells that induce regulatory T cells (9). It is not known how intestinal microbes themselves participate in maintaining tolerance, although there is some evidence of bacteria inducing TSLP (35), and bacteria may regulate the retinoic acid available for Treg priming. Assessing expression of our panel of genes in intestinal tissue and spleen at weaning (3 wk of age) revealed no differences between germ-free and conventionally colonized animals (data not shown). We therefore aimed to detect whether conventional colonization induced a distinctive tolerogenic environment in MLN. We dissected MLN and PLN from germ-free and conventionally colonized animals on day 23 and screened for expression of 90 immune system-related genes. Of these, Ctla4, Cd40, Cd80, Smad7, and Selp were selectively downregulated in MLN in conventionally colonized compared with germ-free mice (Fig. 5A) and were unchanged by colonization in PLN. These are all genes involved in antigen-specific responses, whose downregulation perhaps contributes to the ability to establish oral tolerance. H2-Ea and Stat6 were highly upregulated in both MLN and PLN in conventional mice compared with germ-free mice (Fig. 5B), showing that the gut microbiota influences the expression of specific genes systemically.
The purpose of this study was to assess whether monocolonization with gram-negative or gram-positive bacteria at birth induces the same pattern of gene expression in IECs as a conventional microbiota, and whether IECs of colonized mice become nonresponsive to the bacterial colonizers.
A conventional microbiota is known to promote tolerance not only to commensal bacteria, but also locally and systemically to food antigens. We observed that early exposure to single strains of E. coli and L. acidophilus stimulated IECs modestly, whereas a conventional microbiota induced a shift in expression of cytokine and TLR-signaling related genes that set conventional animals apart from germ-free and monoassociated animals both on PND1 and PND6. However, expression of cytokine and TLR-signaling genes was to some extent increased in all animal groups with increased time after birth (Fig. 2B), indicating that colonization with a conventional microbiota accelerated the development of the intestine and further matured IECs compared with germ-free conditions.
Lotz et al. (14) have reported that although isolated primary cells from late gestational fetuses respond readily to LPS stimulation by producing KC (CXCL1) and MIP-2 (CXCL2), the LPS response in IECs from conventionally colonized newborn mice is blunted after a surge of chemokine production only hours after birth, suggesting acquisition of postnatal LPS tolerance. This is the only study to date addressing LPS tolerance in neonate primary cells, and it implies that engagement of TLR4 with LPS changes the IEC responsiveness just as published with several immortalized cell lines (1, 4), although through degradation of IRAK-1 (14). We wanted to assess whether tolerance can also be induced to intact gram-negative and gram-positive bacteria in newborn mice. A conventional microbiota induced E. coli (i.e., LPS) tolerance, measured as a lower Cxcl2 gene expression response in conventional IECs compared with germ-free IECs. This difference was not significant for Il6 or Tnf expression, perhaps because CXCL2 is a major chemokine secreted by IEC and its induction is therefore more sensitive to regulation, but the significance of this discrepancy is unknown. No tolerance was induced in gnotobiotic mice only colonized by L. acidophilus. On the contrary, cells from this group of mice responded more strongly than the germ-free group by Il6 expression, which may indicate that monocolonization with certain bacteria, including lactobacilli, may halt the development of intestinal tolerance in some respects. In IECs from E. coli monoassociated pups, cytokine and chemokine gene expression was increased to a level nonmodifiable by ex vivo stimulation. This high nonstimulated ex vivo response may reflect an increased sensitivity to culturing after high-dosage LPS exposure in vivo or a high load of LPS present in the isolated IECs “overcoming” tolerance and is in line with the findings that only acquiring a conventional microbiota substantially modulates intestinal cytokine gene expression (Figs. 1 and 2). Although our previous work has demonstrated the importance of assessing gene expression over time in in vitro stimulated cells, because of scarcity of isolated IECs we only measured the ex vivo gene expression after 2 h of stimulation (33, 34). However, when the response of cells from differently colonized mice stimulated with the same bacteria differs, this may indicate differences between the cells, e.g., in expression of pattern recognition receptors.
The differences in gene expression patterns between conventionally and E. coli-associated animals exclude the possibility that LPS alone can regulate the intestinal immune maturation guided by IEC chemokines. Importantly, different parts of the immune system have been shown to require stimulation from specific microbes to attain full function. This is true for CD4+ Th cells, which are induced in the spleen by a Bacteroides zwitterionic polysaccharide (17), and specifically for CD4+ Th17 cells, which require the presence of segmented filamentous bacteria to reach normal numbers (7, 10). In the ileum, segmented filamentous bacteria (present in a conventional microbiota) changed the expression of more genes than lactobacilli or bifidobacteria alone (29), which corresponds to our observations.
Increased expression of Tlr4 and especially Tlr2 was found in mice harboring a conventional microbiota compared with germ-free mice. An upregulation of TLR2 in vivo has been demonstrated to be strongly involved in epithelial repair and integrity of the gut barrier (3, 12), and an intact intestine is believed to be of major importance for the regulation of local and systemic immunity. Hence, establishment of integrity of the epithelial barrier may be one of the major achievements of the first-arriving members of the conventional microbiota. Upregulation of TLR4 has also been reported in newly colonized adult animals (7). Downstream signaling molecules in the TLR signaling cascade were slightly upregulated, perhaps indicating a more rapid growth of the gut epithelium in the conventionalized mice. Also the genes encoding TNF-α, CCL19, and other genes involved in recruitment of cells of the adaptive immune system were upregulated at PND6 in mice with a conventional microbiota compared with germ-free mice (Fig. 1).
IL-7 is a cytokine involved in B and T cell development and may thus play an important role in the maturation of the gut immune system. Whether the lower expression in L. acidophilus-associated animals has any significance cannot be concluded alone from the presented data.
The group composed of Ccl2, Ccl3, Cxcl1, Cxcl2, and Tslp were seemingly regulated independently of all other genes (Fig. 2, A and B). Our data on the expression of Tslp suggest that the amount of TSLP drops shortly after birth and then increases again and that the microbiota speeds up this maturation process. TSLP plays an important role in maintaining a tolerant state in the gut and promotes Th2 responses (31). One possible explanation for the drop in Tslp expression early after birth could be that it delays antibody responses against the beneficial microbiota. This is, however, purely speculative.
Downregulation of the four chemokines may indicate that in particular recruitment of cells of the innate immune system should not be too vigorous during the very early colonization. This may be an important first step in establishing tolerance toward the gut microbiota. Overall, our analysis indicates that intestinal transcriptional regulation of TSLP and the major neutrophil and macrophage attractant chemokines is fundamentally different from transcriptional regulation of the assayed cytokines, receptors, and transcription factors in the early postnatal period. This dichotomy may indicate that mechanisms specialized to cope with the dramatic change from fetus to individual organisms exist, to allow the presence of the establishing microbiota.
We and others (14, 34) have previously demonstrated that Cxcl2 expression is lower in IECs in late gestational fetuses than on PND1. However, unlike Lotz et al. (14), we did not observe a higher Cxcl2 expression on PND 1 compared with PND6 selectively in conventional animals, but in all groups (Fig. 2B), and Cxcl2 was downregulated both with increasing age and by colonization. The higher chemokine expression in young, germ-free animals may indicate that proinflammatory chemokines are necessary to recruit resident phagocytic cells in the fetal state both in the spleen and in the mucosa but are downregulated by the microbiota after birth. This may be due to a postnatal degradation of IRAK-1 protein as shown by Lotz et al., and our finding that Irak1 expression increased with age indicates the importance of a constant expression of this gene for a rapid change in response during, e.g., virus infection. Perhaps representing an important prenatally recruited population, resident chemokine (C-X3-C motif) receptor 1-positive intestinal macrophages have been shown to be crucial for the induction of tolerance to food antigens (8, 14).
The increased CD45 expression with colonization and age may indicate recruitment of intraepithelial lymphocytes. On PND6, <1% of isolated IECs was CD45 positive (data not shown). Because isolated IECs (data not shown) display the same gene expression pattern as the intact intestine tissue (Fig. 2, A and B), our data mainly describe how colonization impacts on cells in the epithelial lining, although we cannot exclude that early colonization also affects the lamina propria immune cell compartment (20).
Stat4 and Clec7 were the only genes expressed to a lower degree in E. coli-exposed intestine than in the germ-free intestine on PND1. The Dectin-1 encoding gene Clec7 is also downregulated by E. coli in IECs in vitro (34), probably excluding an effect of E. coli colonization on the number of dendritic cells in the intestinal mucosa [Dectin-1 may be used as a dendritic cell marker in some tissues but it cannot differentiate dendritic cell from IECs (26, 34)].
Downregulation of Cxcl2 expression in the spleen of conventionally colonized animals already within the first 8 h after birth accompanied the reduced expression of this chemokine in the intestine. This finding suggests that the systemic immune system participates in tolerance induction to the commensal microbiota, preventing excessive inflammation very early in life. An overview of the most important findings is shown in Fig. 6.
At 3 wk of age, when a more complex microbiota is establishing itself because of consumption of solid food (28), the gene expression of cytokines, chemokines, and signaling molecules in ileal tissue and spleen did not differ between conventional and germ-free animals (data not shown). Hereby our study underlines that, after PND1, the microbiota primarily speeds up the maturation process during the first weeks of life and that differences in gene expression in germ-free animals colonized at a certain age does not necessarily reflect the normal development of the symbiotic relationship between host and microbiota. In contrast, gene expression in the intestinal-draining MLNs from conventionalized animals differed from that of the distant PLN cells at weaning. The tolerogenic nature of MLNs, which continuously sample the commensal microbes arriving as “cargo” of intestinal dendritic cells (16), was reflected in the downregulation of Cd40, Cd80 (T cell costimulatory molecules), and Smad7 (inhibitor of immune regulatory TGF-β signaling). Ctla4 was also selectively downregulated in MLNs, perhaps indicating a reduced need for T cells to evade stimulatory signals when local antigen-presenting cells display an anti-inflammatory phenotype. Finally, Selp was downregulated in MLNs by microbial colonization. P-selectin, encoded by Selp, recruits neutrophils to inflammatory sites, a process that may require regulation for the immune system not to overreact to the commensal microbiota. The selective decrease in expression of antigen-presenting relevant genes in MLN may reflect that MLN cells should rarely mount a strong immune response to commensal microorganisms, whereas the presence of bacteria in PLN would indicate infection.
Strikingly, histocompatibility 2, class II antigen E alpha (H2-Ea, the mouse homologue of HLA-DR) expression was 40-fold higher in MLN and 800-fold higher in PLN in conventional mice compared with germ-free mice at weaning. The number of MHC-II molecules has previously been shown to increase in the kidneys of recently colonized germ-free mice (6), but such a strong systemic increase of MHC-II expression with colonization has not been reported before.
STAT6 is the primary signaling molecule responding to ligation of the IL-4 and IL-13 receptor, and its activation is required for Th2 polarization (22). Stat6 expression was increased more than 500-fold in both MLN and PLN of colonized mice, compared with germ-free pups that expressed hardly any Stat6. Downregulation of Stat6 was found in the ileum of adult conventionalized mice by Gaboriau-Routhiau et al. (7), but high Stat6 expression in lymph nodes has not previously been correlated with the presence of a microbiota. On the contrary, IL-4 responses have been shown to be conserved under germ-free conditions (30), so the role of more Stat6 transcripts is unknown. Of note, STAT6 has been shown to play a key role in the regulation of inflammation through NF-κB and to be involved in the protection against sepsis (13). Interestingly, IL-4 and IL-13, both signaling through STAT6, are the only cytokines able to limit IEC chemokine secretion (15). Thus the upregulation of STAT6 in MLN and PLN may constitute an important regulatory mechanism to avoid vigorous responses to commensal microorganisms. In addition, H2Ea (MHC-II) and especially Stat6 expression may serve as markers for early systemic immune maturation enhanced by colonization.
A major limitation of the present study is that we did not confirm differences in mRNA levels on the protein level. Accordingly, we cannot conclude that upregulated genes actually result in functional proteins. However, we have previously demonstrated that Cxcl2 gene expression is reflected in secreted protein in in vitro stimulated germ-free IECs (34). As regards the genes encoding for CCL2, CCL3, CXCL1, CXCL2, and TSLP exhibiting a downregulated expression in conventionalized mice compared with germ-free, we may expect that microorganisms are involved in a specific suppression of the transcription of these genes. The suppression of Cxcl2 expression took place both in the gut as well as in the spleen, but whether the suppression in the spleen was indirectly mediated or a consequence of microorganisms present in the spleen due to translocation from the gut remains to be established. Reduced protein expression of this important chemokine, as a consequence of conventional colonization, also remains to be confirmed in future studies, in vivo in the intestine and spleen as well as ex vivo in bacteria-stimulated IECs. Likewise, investigation of an extended number of cytokines may add to the understanding of the postnatal events in the intestine. In conclusion, a complex microbiota, but not single bacterial strains, reduced the gene expression of proinflammatory chemokines in vivo and induced tolerance toward E. coli stimulation ex vivo. Conventional colonization was also required for a MLN-specific tolerogenic gene expression profile to develop. Overall, our data support the importance of a diverse microbiota for mucosal as well as systemic immune tolerance development.
This work was supported through the FoodDTU collaboration, funded by an internal grant from the Technical University of Denmark.
No conflicts of interest, financial or otherwise, are declared by the author(s).
The authors thank Heidi Letting, Camilla S. Dall, Lisbeth B. Rosholm, Christina Jakobsen, and Anne Ørngren, whose expert technical assistance is greatly appreciated.
Present address of L. N. Fink and L. H. Zeuthen: Novo Nordisk, Novo Allé, 2880 Bagsværd, Denmark.
- Copyright © 2012 the American Physiological Society