Gut microbiota regulates NKG2D ligand expression on intestinal epithelial cells

Authors

  • Camilla H. F. Hansen,

    Corresponding author
    1. Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Frederiksberg C, Denmark
    • Full correspondence: Dr. Camilla H. F. Hansen, Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Thorvaldsensvej 57, 1870 Frederiksberg C, Denmark

      Fax: +45-35332755

      e-mail: camfriis@sund.ku.dk

      Additional correspondence: Dr. Søren Skov, Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Thorvaldsensvej 57, 1870 Frederiksberg C, Denmark.

      Fax: +45-35332755

      e-mail: sosk@sund.ku.dk

    Search for more papers by this author
  • Thomas L. Holm,

    1. Department of Immunopharmacology, Novo Nordisk A/S, Måløv, Denmark
    Search for more papers by this author
  • Łukasz Krych,

    1. Department of Food Science, Faculty of Science, University of Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author
  • Lars Andresen,

    1. Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author
  • Dennis S. Nielsen,

    1. Department of Food Science, Faculty of Science, University of Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author
  • Ida Rune,

    1. Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author
  • Axel K. Hansen,

    1. Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author
  • Søren Skov

    Corresponding author
    1. Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Frederiksberg C, Denmark
    • Full correspondence: Dr. Camilla H. F. Hansen, Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Thorvaldsensvej 57, 1870 Frederiksberg C, Denmark

      Fax: +45-35332755

      e-mail: camfriis@sund.ku.dk

      Additional correspondence: Dr. Søren Skov, Section of Biomedicine, Department of Veterinary Disease Biology, Faculty of Health and Medical Sciences, University of Copenhagen, Thorvaldsensvej 57, 1870 Frederiksberg C, Denmark.

      Fax: +45-35332755

      e-mail: sosk@sund.ku.dk

    Search for more papers by this author

Abstract

Intestinal epithelial cells (IECs) are one of a few cell types in the body with constitutive surface expression of natural killer group 2 member D (NKG2D) ligands, although the magnitude of ligand expression by IECs varies. Here, we investigated whether the gut microbiota regulates the NKG2D ligand expression on small IECs. Germ-free and ampicillin-treated mice were shown to have a significant increase in NKG2D ligand expression. Interestingly, vancomycin treatment, which propagated the bacterium Akkermansia muciniphila and reduced the level of IFN-γ and IL-15 in the intestine, decreased the NKG2D ligand expression on IECs. In addition, a similar increase in A. muciniphila and a decreased NKG2D ligand expression was seen after feeding with dietary xylooligosaccharides. A pronounced increase in NKG2D ligand expression was furthermore observed in IL-10-deficient mice. In summary, our results suggest that the constitutive levels of NKG2D ligand expression on IECs are regulated by microbial signaling in the gut and further disfavor the intuitive notion that IEC NKG2D ligand expression is caused by low-grade immune reaction against commensal bacteria. It is more likely that constitutively high IEC NKG2D ligand expression is kept in check by an intestinal regulatory immune milieu induced by members of the gut microbiota, for example A. muciniphila.

Introduction

Commensal bacteria are important in maintaining immune tolerance and intestinal epithelial barrier integrity. As such, the commensal microbiota is an integral part of the normal gut. It is tolerated by the mucosal immune system [1], which however may rapidly switch from its suppressive state to become activated upon pathogen engagement [2]. The natural killer group 2 member D (NKG2D)/NKG2D ligand interaction is part of this immunological sensor system that detects malfunctioning. Chronic inflammatory conditions in the gut such as the autoimmune celiac disease and Crohn's disease in humans, and colitis in mice, are associated with increased surface expression of NKG2D ligands on intestinal epithelial cells (IECs) and lamina propria dendritic cells [3-6] which is also observed after infection with certain pathogenic strains of Escherichia coli [7].

NKG2D ligands belong to the nonclassical MHC class I molecules and include MICA, MICB, and ULBP 1–6 proteins in human [8, 9] and the H60a/-b/-c, Rae-1, and Mult1 proteins in mice [10]. Their expression levels are generally low, but are greatly enhanced during specific types of cellular stress, including inflammation and neoplastic transformation [11]. IECs were recognized early on as one of the few cell types in the body with constitutive surface expression of NKG2D ligands [12]; however, the level of NKG2D ligand expression on IECs is not uniform, and higher surface expression has generally been observed in the colon compared with that in the small intestine [13]. The ligands are recognized by the activating NKG2D receptor expressed on NK cells, most human CD8+ T cells and activated CD8+ T cells from mice [11, 14, 15], but the NKG2D receptor can also be expressed by γδ T cells and certain activated CD4+ T cells [16], one example being CD4+ T cells from Crohn's disease patients [3].

The regulation of NKG2D ligand surface expression has been intensely studied. However, a unifying controlling mechanism, if one exists, has not yet been established. It is clear that NKG2D ligand expression is regulated at multiple levels. Heat shock, DNA damage, CMV infection, and exposure to histone deacetylase inhibitors and propionic bacteria induce transcriptional activation of NKG2D ligands in mice and human cells [8, 17-22]. Which of the ligands are induced by a specific stimulus, however, is highly dependent upon the cell type and its activation state. In addition, Nice et al. [23] have shown that the murine Mult1 protein is further regulated at the posttranslational level through ubiquitination-dependent degradation. Several forms of cancer are also recognized for their ability to shed surface NKG2D ligands in soluble forms by proteolytic cleavage [24], and Ashiru et al. [25] recently showed that the most prevalent MICA allele (MICA*008) can be directly shed in exosomes from tumors.

Gene regulatory mechanisms inhibiting the NKG2D/NKG2D ligand system are less elucidated. The transcription factor Stat3 is often over-expressed by tumor cells [26] and has been shown to inhibit the MICA promoter activity in HT29 colon carcinoma cells through direct interaction [27]. It is also widely recognized that TGF-β downregulates the NKG2D expression on both NK and CD8+ T cells [28, 29].

Several studies in recent years have demonstrated that different classes of commensal gut microorganisms (e.g. segmented filamentous bacteria) critically affect mucosal immunity [30, 31]. In addition, altered gut microbiota composition and failure to control immunity against intestinal bacteria has been linked to the development of inflammatory bowel disease [32]. A simultaneous increase in NKG2D ligands on IECs in these patients [3], and the observed attenuation of colitis in mice following inhibition of the NKG2D receptor function suggest a commensal-regulated modification of NKG2D ligands expression that may be involved in the induction of mucosal inflammation during these diseases [4, 33]. The aim of the present study was therefore to investigate the regulation of NKG2D ligand expression on IECs mediated by the gut microbiota.

Results

NKG2D ligand expression on IECs in antibiotic-treated mice

To determine the role of bacterial communities in the gut for NKG2D ligand expression on IECs, we first treated C57BL/6NTac (B6) and BomTac:NMRI (NMRI) mice with two different antibiotics administered via the drinking water. In comparison with samples obtained from the control mice receiving water without antibiotics, NKG2D ligand expression on epithelial cells isolated from the entire small intestine was significantly higher in the ampicillin-treated mice (p < 0.001) (Fig. 1A). Furthermore, NKG2D ligand expression was downregulated to the level seen in the untreated mice after microbiota recolonization 10 weeks posttreatment, which illustrates that the increased NKG2D ligand expression during treatment was due to the lack of a full gut microbiota.

Figure 1.

Gut microbiota regulates NKG2D ligand expression on small intestinal epithelial cells (IECs). (A) C57BL/6NTac (B6) male mice received either pure tap water (untreated) or tap water with 1 g/L ampicillin for 4 weeks. Mice which were recolonized for 10 weeks postampicillin treatment are also illustrated (recolonized). The NKG2D ligand-positive IECs are shown from three serial experiments with nine mice killed and analyzed in each experiment; bars represent means. **p < 0.01, one-way ANOVA test with Tukey's posttest. (B) Vancomycin treatment reduces NKG2D ligand expression on small IECs. C57BL/6NTac (B6) and BomTac:NMRI (NMRI) female mice received either pure tap water or tap water with vancomycin hydrochloride (van; 0.5 g/L) for 4 weeks. The NKG2D ligand-positive IECs are shown from two serial experiments with 16 mice killed first (four per group) and 18 killed in the second experiment; bars represent means. *p < 0.05, unpaired two-tailed t-test. Recombinant human NKp80 Fc chimera, human IgG, and PBS were used as negative controls before staining with antihuman IgG. Stained cells were analyzed by flow cytometry. (C) A representative histogram of staining with recombinant mouse NKG2D/CD314 Fc chimera from one mouse treated with ampicillin (black line), and recombinant mouse NKG2D/CD314 Fc chimera staining of IECs from one untreated mouse (blue line), and the recombinant human NKp80 Fc chimera negative control staining (red line) is shown. Data shown are representative of one experiment performed, but similar staining was done for all flow cytometric experiments.

Interestingly, NKG2D ligand expression on small IECs decreased (p < 0.05) following vancomycin treatment in both C57BL/6 mice and NMRI mice, compared to untreated mice (Fig. 1B), which is in contrast to the results obtained in the ampicillin-treated mice. Similarly, the MFI of this staining was significantly lower for the vancomycin-treated B6 mice compared with that in untreated mice, whereas the vancomycin treatment in NMRI mice and ampicillin treatment did not induce any modification of the surface expression of NKG2D ligands (Table 1).

Table 1. MFI of the NKG2D ligand-positive intestinal epithelial cells
Group (figure)Median25% percentile75% percentile
  1. a

    Analyzed on BD Calibur.

  2. b

    Analyzed on Accuri C6.

  3. c

    Analyzed on BD LSRII.

  4. *p < 0.01, two-tailed Mann–Whitney test was used.

Untreatedb42 19635 36056 989
Ampicillinb47 64040 70761 172
Recolonizedb40 91831 61147 569
Untreated B6a11593125
Vancomycin B6a78*69102
Untreated NMRIa138115155
Vancomycin NMRIa120105140
SPF duodenum SWa11291122
Germfree duodenum SWa10197112
SPF ileum SWa126106143
Germfree ileum SWa9386106
Untreatedb55 15649 62863 608
XOS supplementedb42 174*34 75049 191
Control B6c250624662614
IL-10 KOc261725682721

Rae-1 and H60 gene expression in IECs in antibiotic-treated mice

In order to validate the flow cytometry results by a secondary technique and to investigate the specific nature of the NKG2D ligands, real-time (RT) PCR was performed on RNA extracted from the IECs. It is important to note that posttranscriptional regulation of NKG2D ligands may cause different results between the two methods. Nonetheless, Rae-1 gene expression decreased significantly in the vancomycin-treated mice compared with that in both untreated and ampicillin-treated mice similarly to the flow cytometry results. However, the ampicillin-treated mice showed merely a tendency to increased Rae-1 gene expression compared to the untreated mice. Furthermore, although exhibiting a similar trend as the flow cytometry results, the gene expression level of H60 was not significantly different between the groups (Fig. 2A and B). Similar levels of gene expression between treated and untreated mice were also observed for Mult1 (Fig. 2C). In fact, an almost opposite trend was seen, as the Mult1 gene expression seemed to rather decrease in the ampicillin-treated mice compared with that in untreated mice. These data indicate that only some of the NKG2D ligands, such as Rae-1, can be regulated by the gut microbiota.

Figure 2.

Real-time quantitative analyses of NKG2D ligand expression in intestinal epithelial cells isolated from six ampicillin-treated C57BL/6NTac (B6) mice, six vancomycin-treated B6 mice, and six untreated B6 mice from two serial experiments with nine samples from each experiment. Relative gene expression (fold change) of (A) Rae-1, (B) H60c, and (C) Mult1 analyzed in intestinal epithelial cells isolated from ileum is shown. Data were normalized to Actb and then to the mean ileum control group, which was defined to 1. Data are shown as mean and SEM, *p < 0.05, one-way ANOVA test with Tukey's posttest.

Fecal gut microbiota analysis of antibiotic-treated mice

To confirm the broad antimicrobial effect of ampicillin treatment that we have previously shown [34], denaturing gradient gel electrophoresis (DGGE) analysis was performed on feces samples collected from antibiotic-treated and untreated mice. The DGGE profiles from feces samples collected during treatment with ampicillin showed that most bacterial strains were eliminated by the antibiotic treatment, illustrated by the lack of most bands compared with that in untreated mice (Fig. 3A). In addition, it was observed that the ampicillin-treated mice were recolonized by a complete gut microbiota 10 weeks after treatment had ended (Fig. 3A).

Figure 3.

Gut microbiota analyses of the ampicillin- and vancomycin-treated mice. (A) Feces samples were collected from the untreated and ampicillin-treated C57BL/6NTac (B6) mice during and after recolonization from three serial experiments with nine mice killed and sampled in each experiment. DNA was extracted and amplified by means of PCR, using primers specific to the V3 region of the 16S rRNA gene. Amplicons were thereafter separated by means of denaturing gradient gel electrophoresis (DGGE). The picture shows a segment of the gel illustrating DGGE profiles of four untreated mice, four ampicillin-treated mice, and four mice which have recolonized for 10 weeks after ended ampicillin treatment before sampling (recolonized). (B, C) Feces samples were collected from seven untreated (round symbols) and six vancomycin-treated (diamonds) (B) B6 mice or (C) NMRI mice from two serial experiments. DNA was extracted and amplified by PCR, using primers specific to the V3 region of the 16S rRNA gene. Amplicons were thereafter separated by means of DGGE and the profiles were subjected to cluster analysis (Dice algorithm) followed by principal component analysis as shown. (D) Real-time PCR analysis of feces samples collected from untreated, ampicillin-treated, and vancomycin-treated B6 mice. The relative distribution of Akkermansia muciniphila within all bacteria is shown; bar represents the mean. All feces samples analyzed (ten per group sampled during two serial experiments) were quantified in duplicate.

In a previous study, we demonstrated by pyrosequencing how vancomycin eliminates many major species of both Gram-positive and Gram-negative bacteria [35]. Supportive of this, principal component analysis of DGGE profiles revealed a similar clear separation of the vancomycin-treated and untreated mice (Fig. 3B and C), demonstrating major changes in the gut microbiota composition in feces from vancomycin-treated B6 and NMRI mice compared with those from untreated mice.

In addition, vancomycin treatment was previously shown by us to propagate one single species, the mucus-degrading bacteria Akkermansia muciniphila, which dominated most of the gut microbiota [35]. To confirm this, RT-PCR of feces samples from both ampicillin- and vancomycin-treated mice was performed and we found that only very low proportions of A. muciniphila existed in the untreated and ampicillin-treated mice. However, almost 60% of the gut microbiota in the mice treated with vancomycin was constituted by A. muciniphila, indicating a NKG2D ligand downregulating effect of A. muciniphila (Fig. 3D).

NKG2D ligand expression on IECs in germ-free mice

As ampicillin treatment does not eliminate all bacteria, we needed to further verify that the increased NKG2D expression after ampicillin treatment was actually caused by a broad elimination of most bacteria. Germ-free Swiss Webster (Tac:SW) mice were euthanized and compared with specific pathogen free (SPF) SW mice. On both the duodenal and ileac epithelial cells, NKG2D ligand expression was significantly higher in the germ-free mice compared with that in SPF mice, clearly indicating a suppressive effect of the intestinal microbiota (Fig. 4A).

Figure 4.

Germ-free (GF) mice upregulate NKG2D ligand expression on intestinal epithelial cells (IECs). Duodenal and iliac IECs were isolated from female GF and specific pathogen free (SPF) Swiss Webster mice. Isolation of IECs and staining of NKG2D ligands was as described in Figure 1 before IECs were analyzed by flow cytometry. The data are shown as mean +SEM (six mice per group from two serial experiments). *p < 0.05, **p < 0.01, ***p < 0.001, one-way ANOVA test with Tukey's posttest.

Intestinal cytokine production in antibiotic-treated mice

Selected bacteria may alter the homeostatic state of low-grade inflammation in the gut, and we therefore hypothesized that the microbial changes induced by the antibiotic treatments would modify the intestinal cytokine balance in a way that could relate to the NKG2D ligand expression. Cytokine protein levels were measured by Luminex xMAP technology in the supernatant of homogenized small intestinal tissue samples of antibiotic-treated and untreated mice. Interestingly, the level of the proinflammatory cytokines IFN-γ, IL-17, and IL-15 were downregulated in the mice treated with vancomycin compared to the untreated mice, whereas the ampicillin treatment seemed to only downregulate IL-17 production (Fig. 5). Instead, a significant increase could be observed in IL-15 in the ampicillin-treated mice compared with that in untreated and vancomycin-treated mice (Fig. 5B). All other cytokines (IL-1α, IL-1β, IL-2, IL-5, IL-6, IL-7, IL-9, IL-10, IL-12) measured above detection level were not significantly different between the groups (data not shown).

Figure 5.

Ileum cytokine profile of inflammatory mediators in vancomycin- and ampicillin-treated C57BL/6NTac mice. The levels of (A) IFN-γ, (B) IL-15, and (C) IL-17 were measured by multiplex bead-based Luminex analysis on homeogenate supernatants from 2 cm ileum sampled upon euthanization (two to three samples per group from each of two serial experiments). Bars represent the means. *p < 0.05, ** p < 0.01, one-way ANOVA test with Tukey's posttest. Amp: ampicillin-treated mice; Vanco: vancomycin-treated mice.

NKG2D ligand expression on IECs in IL-10 KO mice

Vancomycin-treated mice have previously been shown to have higher proportions of IL-10-producing Treg cells compared with untreated mice [36], and despite the similar levels of IL-10 found in our antibiotic-treated and untreated mice, we did observe a less proinflammatory state in the small intestine during vancomycin treatment. We therefore hypothesized that low levels of NKG2D ligands in vancomycin-treated mice could be explained by a less proinflammatory milieu in the gut further regulated by the gut microbiota.

To test if a less immune-suppressed intestinal environment could play a role in the potential gut microbiota-mediated suppression of NKG2D ligands on IECs, IL-10 B6 KO mice were compared with wild-type B6 mice as IL-10 is a key immunoregulatory cytokine counteracting the production of several proinflammatory cytokines and which thereby acts as an essential immunosuppressant in the gastrointestinal tract [37]. NKG2D ligand expression on epithelial cells isolated from the entire small intestine was significantly higher (p < 0.001) in IL-10 KO mice compared with B6 mice which indicate an, at least indirect, suppressive role of IL-10 in NKG2D ligand expression (Fig. 6).

Figure 6.

NKG2D ligand expression on small intestinal epithelial cells (IECs) isolated from IL-10 KO mice. IECs were isolated from female C57BL/6NTac and IL-10 KO mice before the onset of clinical sign of colitis. IECs were stained, first with recombinant mouse NKG2D/CD314 Fc chimera and second with polyclonal rabbit antihuman IgG. Recombinant human NKp80 Fc chimera, human IgG, and PBS were used as negative controls before staining with antihuman IgG. Stained cells were analyzed by flow cytometry. Data shown are from one experiment with ten mice per group, bars represent means. ***p < 0.001, unpaired two-tailed t-test.

NKG2D ligand expression on IECs in mice on a diet supplemented with xylooligosaccharides (XOS)

In order to alter the gut microbiota in a less-extreme way, male B6 mice were fed with a diet supplemented with XOS. XOS are a prebiotic candidate that stimulates microbes in the gut, such as bifidobacteria that may have beneficial effects on the host including anti-inflammatory effects on the immune system to proliferate [38]. Thus, XOS feeding induces changes in the gut microbiota without compromising the physiologically normal functions of the gut, as opposed to antibiotic treatment, and may therefore in future treatment strategies be considered as a better opportunity to correct dysbiosis. The NKG2D expression on duodenal IECs in B6 mice fed with XOS diet was found to be significantly lower compared than that in mice fed with standard diet (Fig. 7). In addition, the MFI was also significantly lower (Table 1). It is therefore likely that the gut microbiota profile obtained after XOS feeding suppresses NKG2D ligand expression.

Figure 7.

Dietary xylooligosaccharides (XOS) reduces NKG2D ligand expression on small intestinal epithelial cells (IECs). Male C57BL/6NTac mice received either a standard Altromin C1000 rodent diet (untreated) or an Altromin C1000 diet supplemented with 10% XOS for 5 weeks. Duodenal IECs were isolated by EDTA chelation and subsequently stained first with recombinant mouse NKG2D/CD314 Fc chimera and second with FITC-labeled polyclonal rabbit antihuman IgG. Recombinant human NKp80 Fc chimera, human IgG, and PBS were used as negative controls before staining with antihuman IgG. Stained cells were analyzed by flow cytometry. (A) The data shown are from three serial experiments with six to seven mice killed in each experiment, bars represent the means. *p < 0.05, unpaired two-tailed t-test. (B) Representative flow cytometric dot plots illustrating the percentages of NKG2D ligand-positive IECs are also shown. (C) Real-time PCR analysis of feces samples collected from B6 mice fed control diet (untreated) or diet supplemented with XOS. The relative distribution of A. muciniphila within all bacteria is shown. All feces samples analyzed (from three serial experiments with three samples per group) were quantified in duplicate. *p < 0.05, unpaired two-tailed t-test.

Next, we analyzed the proportions of A. muciniphila in the XOS-fed mice, as we had seen an inverse correlation between this bacteria and the NKG2D ligand expression in the vancomycin-treated mice. Interestingly, this inverse correlation was clearly observed in the XOS-fed mice which also had significantly higher proportions of A. muciniphila in the gut compared with that in the control group (Fig. 7C).

Discussion

Our observations suggest that the gut microbiota strongly influences the expression of NKG2D ligands on small IECs. Germ-free mice lacking a commensal microbiota had an increased surface expression of NKG2D ligands, and a similar result was seen during ampicillin treatment which depleted most of the murine commensal bacteria. The NKG2D ligand expression returned to lower levels seen in the untreated mice after ampicillin treatment ended. Our data further suggest that commensal bacteria actively regulate IEC NKG2D ligand expression, and that IECs constitutively express NKG2D ligands without further stimulation, which is in agreement with the original observation by Groh et al. [12]. It is therefore unlikely that the constitutive NKG2D ligand expression is caused by a low-grade inflammatory activity against the commensal bacteria. The NKG2D ligands were detected using recombinant NKG2D and the specific nature of the NKG2D ligands were investigated by RT-PCR which showed that only Rae-1 had a similar expression pattern as the flow cytometry results, whereas H60c merely showed a tendency toward this. Hence, not all NKG2G ligands on IECs seem to be regulated by the gut microbiota.

We further found a striking downregulation of IEC NKG2D ligand expression in vancomycin-treated mice, which contradicted the findings in ampicillin-treated and germ-free mice. Vancomycin is a well-known anti-Gram-positive antibiotic but also inhibits many Gram-negative Firmicutes species [36], most likely as a result of an ancient evolutionary co-dependency of certain Gram-positive and Gram-negative bacteria. However, we previously observed a manyfold increase of A. muciniphila in feces from vancomycin-treated nonobese diabetic mice which constituted almost 90% of the remaining microbiota [35]. This species has been suggested to possess an anti-inflammatory protective effect against inflammatory bowel disease [39], and recent findings in gnotobiotic mice mono-colonized with A. muciniphila suggest a transcriptional host response upon colonization that involves immune tolerance against commensal gut bacteria [40]. It is thus tempting to speculate that the dominance of this single species in vancomycin-treated mice is linked to the decreased NKG2D ligand expression on IECs, especially as we found high levels of A. muciniphila in the vancomycin-treated mice which corresponded with low levels of NKG2D ligand expression whereas increased expression of A. muciniphila was not observed in the ampicillin-treated mice. We also found that dietary XOS propagated A. muciniphila, and in parallel to the data obtain in the vancomycin-treated mice, XOS feeding also caused a marked reduction in the IEC NKG2D ligand expression. The nature and mechanisms behind this interesting correlation, as well as specifying other microbes that may modulate NKG2D ligands, need further investigation in, for example gnotobiotic mice.

The commensal microbiota can affect NKG2D ligand expression by several different mechanisms, which may not necessarily be mutually exclusive. For instance, the commensal bacteria may establish a regulatory milieu in the intestine, with increased expression of immuno-inhibitory cytokines such as TGF-β and IL-10. In this regard, it is notable that both TGF-β and IL-10 have been shown to downregulate NKG2D ligand surface expression [41, 42]. In agreement with this, IL-10 KO mice were shown to have an increase in IEC NKG2D ligand expression. Similarly, IL-10 has previously been reported to restrict the expression of the human NKG2D ligand MICA in melanoma cells [42]. IL-10 KO mice naturally develop inflammation in the colon from 10 to 12 weeks of age [43]; however, in the present study, the NKG2D ligand expression on small IECs was investigated in the IL-10 KO mice before any development of clinical sign of colitis. Nonetheless, we cannot exclude that NKG2D ligand upregulation is induced by an inflammatory molecule produced in these mice, especially as we in the present study found no alterations in the intestinal IL-10 levels of the antibiotic-treated mice.

In addition, decreased level of IFN-γ and IL-15 in the small intestine was observed in the vancomycin-treated mice similar to the NKG2D ligand expression and IL-15 was furthermore increased in the ampicillin-treated mice as was the NKG2D ligand expression. This is interesting, as IL-15 is known to be directly involved in NKG2D ligand upregulation on IELs during celiac disease [5], and it is thus tempting to speculate that a less proinflammatory state, kept in check by the commensal microbes, actively keeps the NKG2D ligand expression low, although such a scenario needs experimental verification. IL-17 was however downregulated in both ampicillin- and vancomycin-treated mice which suggests that this cytokine is not involved in the regulation of NKG2D ligands on IECs. Instead, both antibiotic treatments most likely eradicated important bacteria, for example segmented filamentous bacteria which can induce IL-17 [31, 44].

The commensal microbiota may also directly express or secrete molecules that affect NKG2D ligand surface expression. We have previously shown that propionate from propionic bacteria is involved in the opposite scenario, as it increases NKG2D ligand expression [17]. Further studies are however needed to establish the mechanisms behind these interesting observations.

It is noteworthy that the level of NKG2D ligand expression was substantially lower in the B6 mice housed in the Novo Nordisk animal facility compared with that in B6 mice housed at the University of Copenhagen. Differences in gut microbiota compositions in the groups of untreated control mice because of the different facility environments, sex, and animal vendors from which the mice were purchased, may explain the observed differences in NKG2D ligand expression. In general, we believe that it is important to take differences in microbiota composition into account, when comparing levels of NKG2D ligands measured by different laboratories. This could, at least partly, explain differences observed in the past.

NKG2D ligand regulation by microbial interaction is supported by a growing body of data. Tieng et al. [7] have shown increased expression of NKG2D ligands on IECs after infection with certain pathogenic strains of E. coli and IECs have also been shown to express NKG2D ligands upon TLR3-dependent poly I:C treatment [45]. Modulation of NKG2D ligand expression is important for control of intestinal integrity and mucosal immune regulation, and in light of our current results, it is also noteworthy that Crohn's disease patients, besides having an abnormal NKG2D ligands expression in the gut [3], also have a severely altered gut microbiota [46]. In summary, we have shown that absence of gut microbiota causes a pronounced increase in NKG2D ligand expression and suggest that the normal immune-suppressed milieu in the gut, regulated by the gut microbiota, actively suppresses NKG2D ligand expression. It therefore seems that the symbiotic microbial inhabitants of the healthy gut play a protective role by downregulating NKG2D ligand expression on IECs, and particularly A. muciniphila may be of potential significance in this process.

Materials and methods

Animals

The experiments were carried out in accordance with the Council of Europe Convention European Treaty Series (ETS) 123 on the Protection of Vertebrate Animals used for Experimental and Other Scientific Purposes, and the Danish Animal Experimentation Act (LBK 1306 from November 23, 2007). The study was approved by the Animal Experimentation Inspectorate, Ministry of Justice, Denmark (License number: 2007–561-1434).

Outbred female SPF BomTac:NMRI, female germ-free and SPF Tac:SW mice, and inbred female and male SPF C57BL/6NTac (B6) were purchased from Taconic (Lille Skensved, Denmark). They were housed in groups of five to six mice per cage at the University of Copenhagen, Frederiksberg, Denmark under SPF conditions.

IL-10-deficient female B6.129P2-IL10tm1Cgn/J mice and control female C57BL/6J (B6) mice were purchased from the Jackson Laboratories (Bar Harbor, ME, USA) in accordance with a license agreement with MCG (Munich, Germany). Both strains were housed at Novo Nordisk A/S in groups of ten mice per cage under SPF conditions. The animal studies were also approved by the Novo Nordisk ethical review committee.

All mice had free access to an Altromin 1324 diet (Brogaarden, Lynge, Denmark) and tap water unless stated otherwise, and health monitoring was conducted according to FELASA guidelines [47]. Germ-free SW mice were euthanized immediately upon arrival in a germ-free cylinder. Ampicillin-treated mice were euthanized at 17 weeks of age. All other mice were euthanized by cervical dislocation at 8–10 weeks of age, including the IL-10 KO mice before clinical onset of colitis. The mice were killed in serial experiments with three to four mice per group at a time.

Antibiotic treatment

C57BL/6NTac and BomTac:NMRI received either vancomycin hydrochloride (0.5 g/L; ThermoFisher Scientific Inc., Waltham, MA, USA) or ampicillin (1 g/L; Ampivet® vet., Boehringer Ingelheim, Copenhagen, Denmark) in the drinking water for 4 weeks. Bottles with water and antibiotics were changed twice weekly for both the treated mice and the untreated mice that received pure tap water. One group of mice was recolonized after ended ampicillin treatment for 10 weeks before they were killed.

Modified diets

From weaning and 5 weeks onwards male C57BL/6NTac mice received an Altromin C1000 rodent diet (Brogaarden) supplemented with 10% XOS (XOS DP 2–6, Active Nutrition, DuPont Nutrition & Health, Kantvik, Finland) at the expense of polysaccharides. A control group received Altromin C1000 rodent diet with no supplements. XOS are nondigestible carbohydrates suggested as a prebiotic candidate.

IEC cell isolation and flow cytometric analysis of NKG2D ligands

Immediately after euthanization intestines were cleaned from residual mesenteric fat, opened longitudinally, washed with cold PBS and cut in 1 cm pieces. The pieces were incubated in 5 mL PBS containing 2 mM EDTA for 20 min at 37°C with agitation (50 rpm). The fragments were subsequently shaken intensively to detach the epithelial cells and passed through a 70 μm cell strainer. Cells were washed twice in ice-cold PBS before staining of the IECs for NKG2D ligands. After 30-min incubation on ice with 4 μg/mL recombinant mouse NKG2D/CD314 Fc chimera (R&D systems, Inc., Minneapolis, MN, USA), or control human NKp80 Fc chimera (R&D systems), or human IgG (Bethyl laboratories Inc., Montgomery, TX, USA) in PBS, or PBS alone all IEC samples were washed twice and stained with FITC-labeled polyclonal rabbit antihuman IgG (Dako, Glostrup, Denmark) at a dilution of 1/100 for 30 min at 4°C. Analysis was performed using an Accuri C6 flowcytometer, BD Calibur or BD LSRII.

RT-PCR analysis of NKG2D ligands

A 0.5 cm part of ileum next to caecum was sampled from antibiotic-treated and untreated mice immediately after euthanization and stored in RNA later at 4°C overnight until frozen in an empty cryo tube at −80°C. RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA, USA) and reverse-transcribed using SuperScript III reverse transcriptase enzyme (Invitrogen). PCR was performed using standard conditions. Rae-1, H60c, and MULT1 primer sequences and the housekeeping gene β-actin primer sequences are given in Table 2. For quantitative RT-PCR analysis, the PCR was performed using Brilliant SYBR Green QPCR Master Mix kit (Stratagene, Santa Clara, CA, USA) and samples were run and analyzed on a Stratagene MX3005P thermocycler in duplicate.

Table 2. RT-PCR primer sequences used for analyzing intestinal NKG2D ligand expression
Rae1pan_F5′-ACCAAAGTGGACACTCACAAGACCA-3′
Rae1pan_R5′-ACCCCTGATTCATCATTAGCTGATCTCCA-3′
H60c-(115)_F5′-CTCTGGGCACTGTCACGCCT-3′
H60c-(369)_R5′-GCAGGACACAGCATAGTGGCAT-3′
Mult1-(130)_F5′-AGCAGCTATGGAGCTGACTGCCA-3′
Mult1-(499)_R5′-AGCCTGCAGAGTGAGGGGCTTT-3′
β-actin_F5′-GCGAGCACAGCTTCTTTGCAGC-3′
β-actin_R5′-GGTTGGCCTTAGGGTTCAGGGG-3′

Bacterial analysis by DGGE

The analyzed samples included feces samples attained aseptically after the mice were euthanized and stored at −80°C. A detailed description on the analysis by DGGE is described in detail elsewhere [48]. Briefly, DNA was extracted using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany), and amplified by means of PCR, using primers specific to the V3 region of the 16S rRNA gene. The amplicons were thereafter separated by means of DGGE on a polyacrylamid gel containing a 30–65% denaturing gradient (100% corresponds to 7 M urea and 40% formamide) and DGGE profiles were analyzed using BioNumerics Version 4.5 (Applied Maths, Sint-Martens-Latem, Belgium) for cluster analysis (Dice similarity coefficient with a band position tolerance and optimization of 1% using the Unweighted Pair Group Method with Arithmetic averages clustering algorithm and principal component analysis).

RT-PCR of fecal A. muciniphila

All feces samples analyzed were quantified in duplicate for the relative abundance of A. muciniphila using the 7500 Fast Real-time PCR System (Applied Biosystems, Foster City, CA, USA). Two primer sets, universal: F_338, R_518 [49] and A. muciniphila specific [50], were used in this assay. In order to ensure an optimal specificity with the A. muciniphila genome, the sequence of the forward primer was modified as follow: F-5′-ACWCCTACGGGWGGCAGCAG-3′.

The reaction mixture (20 μL) composed of 1× SYBR green PCR Master Mix (Applied Biosystems), 1 μL of each, either A. muciniphila-specific primers or 16S rRNA universal primers at a final concentration of 0.25 μM, and 5 μL of template DNA adjusted to approximately equal concentration of 20 ng/μL. The PCR temperature profile was as follows: 95°C for 5 min, followed by 40 cycles of 95°C for 15 s, 60°C for 1 min. The fluorescence acquiring was set in the annealing/extension step. Subsequent to the amplification, a melting curve analysis was performed in order to distinguish putative nonspecific amplification. Serial tenfold dilutions of A. muciniphila pure culture (DSMZ 22959) genomic DNA was used to generate standard curves.

Milliplex xMAP Luminex analysis of intestinal cytokines

A 2 cm part of ileum (starting from caecum) was partitioned lengthwise, washed gently in ice-cold PBS and one half stored in tissue homogenate lysis buffer (Ampliqon, Skovlunde, Denmark) at −80°C after immediately frozen in liquid nitrogen, while the other half was used for flow cytometry analysis. The intestines were homogenized by using a T25 Ultraturrax homeogenizer (IKA, Staufen, Germany) and subsequently centrifuged for 15 min at 10 000 G and 4°C. The supernatant was decanted and centrifugation was repeated twice, the last time in a 5 μm Ultrafree MC-centrifugal filter device (Millipore, Billerica, MA, USA). Samples were kept cold during all steps. Next, the supernatant was analyzed for levels of IFN-γ, TNF-α, IL-1α, IL-1β, IL-2, IL-5, IL-6, IL-7, IL-9, IL-10, IL-12(p40), IL-15, and IL-17 by bead-based Milliplex xMAP Luminex technology (Millipore) in accordance with manufacturer's instructions.

Statistics

Statistical analysis was performed using GraphPad Prism version 5.02 (GraphPad Software, San Diego, CA, USA). Statistical significance was evaluated by the one-way ANOVA test with Tukey's posttest when comparing three or more groups. Unpaired two-tailed t-test or Mann–Whitney test was used when comparing two groups. p values less than 0.05 were considered statistically significant.

Acknowledgments

This work was supported by a grant from “Arkitekt Holger Hjortenberg og hustru Dagmar Hjortenbergs fond” and it was carried out as part of “Center for Applied Laboratory Animal Research” (www.calar.dk).

Conflict of interest

The authors declare no financial or commercial conflict of interest.

Abbreviations
DGGE

denaturing gradient gel electrophoresis

IEC

intestinal epithelial cell

NKG2D

natural killer group 2 member D

SPF

specific pathogen free

SW

Swiss Webster

XOS

xylooligosaccharides

Ancillary