GM-CSF promoted DC recruitment and survival governs the intestinal mucosal response to enteric attaching-and-effacing bacterial pathogens

Summary

The attaching-and-effacing (A/E) lesion-causing enteric pathogen, Citrobacter rodentium, was used to probe the importance of GM-CSF in mucosal protection against enteric bacterial infection. C. rodentium infection increased GM-CSF production and CD11c⁺ dendritic cells (DC) in the colon of wild-type mice. After infection, mice lacking GM-CSF (GM-CSF^-/-) had significantly fewer mucosal CD11c⁺ DC, greater bacterial burden, increased mucosal inflammation and systemic spread of infection, decreased antibody responses to C. rodentium, and delayed pathogen clearance. The defective mucosal response to infection in GM-CSF^-/- mice was rescued by GM-CSF and mimicked by CD11c⁺ DC depletion in infected wild-type mice, indicating that CD11c⁺ DC are major targets of GM-CSF in the intestinal mucosa in vivo. Diminished mucosal DC numbers in infected GM-CSF^-/- mice reflected decreased DC survival and recruitment within the colon mucosa. The latter was related to a failure to upregulate epithelial cell production of the DC chemoattractant chemokine, CCL22, in the absence of GM-CSF.

Introduction

Granulocyte-macrophage colony-stimulating factor (GM-CSF) stimulates myeloid cell development and maturation (Hamilton and Anderson, 2004), and promotes dendritic cell (DC) differentiation and survival in vitro. It is used in cell culture to generate DC from blood and bone marrow precursors (Inaba et al., 1992; Markowicz and Engleman, 1990). However, mice deficient in GM-CSF (GM-CSF^-/-) have normal numbers of bone marrow DC progenitors and mature DCs in spleen, thymus and lymph nodes, which suggests a redundant role for GM-CSF in steady-state DC maintenance (Dranoff et al., 1994; Stanley et al., 1994). GM-CSF^-/- mice have defective alveolar macrophage function and develop a spontaneous pulmonary alveolar proteinosis-like phenotype (Paine et al., 2000; Stanley et al., 1994). Consistent with this, GM-CSF has been reported to activate alveolar macrophages that can protect the host from respiratory pathogens (Dranoff et al., 1994; LeVine et al., 1999; Shibata et al., 2001), and GM-CSF^-/- mice are more susceptible to several respiratory pathogens (Deepe et al., 1999; Gonzalez-Juarrero et al., 2005; LeVine et al., 1999; Paine et al., 2000). However, the role of GM-CSF in regulating host defenses in other mucosal organs remains poorly understood. In particular, whether GM-CSF has a physiological role in the intestinal tract, the most important entry portal for microbial infections in mammals, with respect to antimicrobial defense or DC functions is not known.

Enterohemorrhagic Escherichia coli (EHEC) and enteropathogenic E. coli (EPEC) are important causes of severe diarrhea and death worldwide (Kaper et al., 2004). Such pathogens, termed A/E pathogens, induce characteristic attaching-and-effacing (A/E) lesions in the intestinal epithelium, which are important for establishing infection in the host. Citrobacter rodentium is a naturally occurring mouse pathogen that is widely used to model infections with A/E pathogens in humans (Borenshtein et al., 2008). In wild-type mice, C. rodentium colonizes the apical surface of colon epithelial cells, effaces the epithelial cell microvilli, but does not invade deeper layers of the colon mucosa or spread systemically. Infection is characterized by an inflammatory cell infiltrate in the colon lamina propria and hyperplasia of the colonic crypts (Eckmann, 2006; Maaser et al., 2004; Mundy et al., 2005). We previously reported that CRAMP, an epithelial cell antimicrobial protein belonging to the cathelicidin family, is important in determining early colonization of the host with C. rodentium (Iimura et al., 2005), whereas CD4⁺ T cells, B cells and IgG antibodies to C. rodentium are important in controlling infection in the later periods and are required for ultimate pathogen clearance(Bry and Brenner, 2004; Maaser et al., 2004; Simmons et al., 2003). Furthermore, several cytokine knockout mice (e.g. interferon-γ, tumor necrosis factor-α, IL-6, and either p19 or IL-12p40) have delayed clearance of C. rodentium infection (Dann et al., 2008; Goncalves et al., 2001; Mangan et al., 2006; Simmons et al., 2002).

We used the A/E pathogen, C. rodentium, to probe the functional importance and cellular targets of GM-CSF in the intestinal mucosa. We show that GM-CSF produced in the intestinal mucosa in vivo acts in a non-redundant manner to enhance host protection to an A/E pathogen through mechanisms that involve DC and epithelial cells.

Results

Colonic GM-CSF induction after C. rodentium infection

To probe the in vivo functions of GM-CSF in mucosal defense, WT B6 mice were infected with the A/E pathogen, C. rodentium. GM-CSF production after infection increased significantly in the colon (Fig. 1A,B), but not in spleen or mesenteric lymph nodes (MLN), nor were serum GM-CSF levels elevated (data not shown). GM-CSF protein was primarily observed in CD11c⁺ DCs and F4/80⁺ macrophages in the infected colon mucosa (Fig. 1C). To validate that GM-CSF immunostaining reflected GM-CSF production by lamina propria leukocytes, rather than GM-CSF uptake by lamina propria leukocytes that might express GM-CSF receptors, GM-CSF mRNA was assessed in lamina propria leukocytes and epithelial cells isolated from C. rodentium-infected (2 weeks) and uninfected mice. GM-CSF mRNA expression was upregulated by 4-5-fold in lamina propria leukocytes (p<0.05, n=3), but was not upregulated in colon epithelial cells after infection. Furthermore, lamina propria leukocytes had ~400-fold higher GM-CSF mRNA levels than colon epithelial cells. Thus, increased mucosal GM-CSF production after infection largely reflects de novo production by DCs and F4/80+ cells in the lamina propria. Together, these findings demonstrate that mucosal GM-CSF production is increased after infection and suggest that the cytokine may have a role in mucosal host defense to this pathogen.

Figure 1. GM-CSF induction in the colon of C. rodentium-infected mice.

(A) Colon from B6 mice left uninfected or infected with C. rodentium for 2 weeks was immunostained with anti-GM-CSF antibody. Infected mice manifest marked colonic crypt hyperplasia. Original magnification, x 100. (B) GM-CSF in colon tissue was determined by ELISA at the indicated times after infection. Data are mean ± SEM (n = 5 mice at each time point). ND = none detected. *; p < 0.05 vs. uninfected control. (C). Top row. Colon from B6 mice infected with C. rodentium for 2 weeks was immunostained for GM-CSF (left panel, red) and CD11c (middle panel, green). Right panel is a merged image of the left and middle panels and also includes nuclear staining. Bottom row. Colon was stained for GM-CSF (left panel, red) and F4/80 (middle panel, green). Right panel is a merged image including nuclear staining. Merged images were created using Adobe Photoshop software. Arrows indicate GM-CSF producing cells. Original magnification, x 400.

Increased susceptibility of GM-CSF^-/- mice to C. rodentium infection

To test the importance of GM-CSF during C. rodentium infection, we employed gene-targeted mice deficient in GM-CSF. Bacterial colonization was comparable early after infection (4 days), indicating that GM-CSF^-/- mice had no apparent defect in innate antibacterial defense. Consistent with this conclusion, GM-CSF deficiency did not impact expression of the epithelial cell-produced antimicrobial peptide mCRAMP (data not shown), which is critical in early defense against C. rodentium (Iimura et al., 2005). However, at one week after infection, GM-CSF^-/- mice had significantly increased mucosal colonization with C. rodentium compared to WT B6 controls (Fig. 2A), significantly greater fecal counts of C. rodentium (Fig. 2B), and significantly greater systemic infection in spleen and MLN (Fig. 2C). In parallel, GM-CSF^-/- mice had lost significantly more body weight after 2 weeks than WT B6 mice (97.8 ± 1.5% vs. 103.2 ± 1.1% of pre-infection weight, respectively; p<0.05), underlining the overall clinical impact of GM-CSF deficiency on the course of the infection. B6 mice had cleared infection by 3 weeks, whereas clearance did not occur until 4 weeks after infection in GM-CSF^-/- mice (Fig. 2B).

Figure 2. C. rodentium infection in B6 and GM-CSF^-/- mice.

(A) One week after C. rodentium infection, colon sections from B6 and GM-CSF^-/- mice were co-stained for C. rodentium (red) and F-actin (green). Co-localization of C. rodentium and F-actin is shown in yellow. Original magnification, x 100. (B) Fecal counts (CFU) of C. rodentium were determined at the indicated times after oral gavage of 5 × 10⁸ C. rodentium to B6 and GM-CSF^-/- mice. Data are mean ± SEM of 3 independent experiments (n = 10-17 mice for each data point). Dotted line indicates detection limit of the CFU assay. (C) C. rodentium in spleen and MLN of B6 and GM-CSF^-/- mice were determined 2 weeks after infection. Data are shown as CFU from individual mice and bars represent the geometric means (n = 15-16). Dotted line indicates the detection limit of the CFU assay. (D) IgG and IgM serum antibodies to C. rodentium in B6 and GM-CSF^-/- mice were assayed by ELISA at the indicated times after infection. Data are mean ± SEM of optical density (n = 10-15 mice time point). (E) B6 and GM-CSF^-/- mice were left uninfected or infected with C. rodentium for up to 4 weeks. Sections of colon obtained at the indicated times were stained with H&E. Original magnification, x 100. Scale bar is 200 μm. (F) Colonic crypt depth in C. rodentium infected B6 and GM-CSF^-/- mice was assessed at the indicated times after infection. Data are mean ± SEM (n = 5-10 mice per group at each time point). (G) MPO in colon tissue was assayed by ELISA. Data are mean ± SEM (n = 5-10 mice per group at each time point). (H) TNFα, KC and MIP-2α mRNA transcripts in colon tissue of B6 and GM-CSF^-/- mice were assayed by real time RT-PCR (n = 4-5 mice per group at each time point). * p < 0.05 relative to B6 mice at indicated time points.

Serum titers of IgM and IgG anti-C. rodentium antibodies were significantly lower in infected GM-CSF^-/- than WT B6 mice (Fig. 2D). Further, after C. rodentium infection, GM-CSF^-/- mice had increased and more persistent colonic crypt hyperplasia (Fig. 2E,F), significantly higher levels of mucosal MPO (Fig. 2G), and increased expression of the proinflammatory cytokines TNF-α, KC, and MIP-2α (Fig. 2H). No significant differences between WT and GM-CSF^-/- mice were found in the expression of IFN-γ, IL-12p40, IL-23p19, IL-10, IL-17, IL-6, IL-4 or IL-1β during the 3 weeks after infection (data not shown). Taken together, these data indicate GM-CSF has an important role in controlling the magnitude of mucosal and systemic bacterial infection, the mucosal proinflammatory cytokine response, and the adaptive immune response important for clearance of C. rodentium.

Decreased CD11c⁺ DC numbers in C. rodentium infected GM-CSF^-/- mice

To begin to define the mechanisms by which GM-CSF contributes to mucosal host defense, we examined mucosal DC numbers as DC precursors from bone marrow in vitro are targets of GM-CSF, and bone marrow-derived DCs cultured in GM-CSF and IL-4 have an inflammatory phenotype possibly relevant to innate antimicrobial host defense and inflammation in vivo (Serbina et al., 2003; Xu et al., 2007). Uninfected WT B6 and B6 GM-CSF^-/- mice had similar numbers and distribution of CD11c⁺ DC in the colon (Fig. 3A,B), which is consistent with prior studies suggesting that GM-CSF is not required for steady-state DC maintenance in lymphoid organs (Kingston et al., 2009; Vremec et al., 1997).

Figure 3. Decreased CD11c⁺ DC in C. rodentium infected GM-CSF^-/- mice.

(A) CD11c⁺ cells in the colon of B6 and GM-CSF^-/- mice infected for 2 weeks with C. rodentium or left uninfected. CD11c⁺ cells are shown in red and phalloidin staining of F-actin is in green. Arrows indicate CD11c⁺ cells underlying surface epithelial cells. Original magnification, x 200. Similar results were obtained from 3 week-infected mice. (B) Flow cytometric analysis of cells isolated from the colon lamina propria of B6 and GM-CSF^-/- mice 2 weeks after infection or left uninfected. Cells were stained with PE-conjugated CD11c and FITC-conjugated I-A^b. Results of a representative experiment are shown. Similar results were obtained in 3 repeated experiments. (C) Flow cytometric analysis of CD11c⁺ DCs isolated from the colon of B6 and GM-CSF^-/- mice infected for 2 weeks. DCs were stained with PE-conjugated CD103 and FITC-conjugated CD11b. Results of a representative experiment are shown. Similar results were obtained in 2 repeated experiments. (D) DCs were isolated from 2 week infected colon as in (C) and surface marker expression of the activation markers CD40, CD80, CD86 and the expression of the chemokine receptors CCR6 and CCR7 on DCs was assessed by flow cytometry. (E) Spleen and MLN cells from B6 and GM-CSF^-/- mice obtained 2 weeks after infection were characterized as in (B). (F,G) C. rodentium-infected colon from B6 and GM-CSF^-/- mice was stained 2 weeks after infection for F4/80 (red, F) and Gr-1 (red, G). F-actin was stained with Alexa-Fluoro 488 phalloidin (green). Original magnification, x 200.

After C. rodentium infection of WT B6 mice, CD11c⁺ DC were markedly increased in the lamina propria surrounding the colon crypts and in the area immediately underlying surface epithelial cells (Fig. 3A). In contrast, significantly fewer CD11c⁺ DCs were present in the colon mucosa of infected GM-CSF^-/- mice (Fig. 3A,B). Despite the absolute numeric difference, the relative proportions of the two major lamina propria DC subpopulations (i.e., MHC class II⁺, CD11c⁺,CD11b⁺,CD103⁺ and MHC class II⁺, CD11c⁺, CD11b⁺, CD103^-) (Bogunovic et al., 2009), the expression of key activation markers (CD40, CD80, CD86), and the expression of the migration-associated chemokine receptors, CCR7 and CCR6, on CD11c⁺ DC did not differ between infected WT B6 and GM-CSF^-/- mice (Fig 3.C,D). However, compared to CD11c⁺ DC isolated from the colon of infected WT B6 mice, those from GM-CSF^-/- mice manifested a significant decrease in the uptake of FITC-dextran, suggesting an abnormality in endocytotic activity (Fig. S1). No significant increase of CD11c⁺ DC was observed in the spleen or MLNs after infection of WT B6 and GM-CSF^-/- mice (Fig. 3E).

In contrast to CD11c⁺ DC, F4/80⁺ macrophages and Gr-1⁺ leukocytes were markedly increased in the colon mucosa of infected GM-CSF^-/- compared to WT B6 mice (Fig. 3F,G). These results indicate that GM-CSF has an important role in determining the number and localization of CD11c⁺ DC under inflammatory conditions, and the magnitude of the inflammatory response in the colon during C. rodentium infection.

GM-CSF administration rescues the abnormal host response in GM-CSF^-/- mice

To exclude the possibility that GM-CSF deficiency had caused irreversible developmental abnormalities in the colon, we administered mGM-CSF daily to C. rodentium infected GM-CSF^-/- mice and examined the effects on DC numbers and mucosal defense. GM-CSF treatment significantly increased CD11c⁺ DCs in the colon (Fig. 4A) to levels approaching those seen in infected WT B6 mice (see Fig. 3B). Moreover, cytokine treatment partially corrected the attenuated IgG anti-C. rodentium antibody response of GM-CSF^-/- mice (Fig. 4B) and significantly decreased systemic C. rodentium infection in the MLN and spleen (Fig. 4C). Thus, the lack of DC in the mucosa of infected GM-CSF^-/- mice was not due to a developmental DC defect.

Figure 4. Rescue of the abnormal phenotype in C. rodentium infected GM-CSF^-/- mice by GM-CSF administration.

(A) GM-CSF^-/- mice infected with 5 × 10⁸ C. rodentium were injected with PBS or GM-CSF (5 μg/mouse) i.p. daily for 2 weeks, after which lamina propria cells from the colon were isolated and analyzed by flow cytometry as in Fig. 2B. Similar results were obtained in 2 repeated experiments. (B) IgG antibodies to C. rodentium from PBS or GM-CSF-treated C. rodentium-infected GM-CSF^-/- mice were assayed by ELISA. Data are mean ± SEM (n = 12 mice per group). (C) C. rodentium in the spleen and MLN of PBS or GM-CSF treated GM-CSF^-/- mice were determined by CFU assay 2 weeks after infection as in Fig. 2. Data are shown as CFU from individual mice (n =11) and the bar indicates the geometric mean. Dotted line indicates the detection limit of the CFU assay. * p < 0.05 relative to PBS-treated mice at the indicated time points.

Functional importance of CD11c⁺ DCs in the host response to C. rodentium

Because GM-CSF deficiency resulted in diminished mucosal DC numbers after infection, we reasoned that the DC loss may be responsible for the compromise in host defense. To determine the functional importance of CD11c⁺ DCs, CD11c⁺ DCs were depleted immediately after C. rodentium infection using a DT-mediated strategy (Abe et al., 2007; Berndt et al., 2007; Jung et al., 2002; Zaft et al., 2005). We used bone marrow (BM) chimeras generated by reconstituting irradiated WT B6 mice with CD11c-DTR BM for these studies as those mice tolerate repeated DT injections without adverse side effects (Varol et al., 2009; Zaft et al., 2005). Repeated injections of C. rodentium infected B6 DTR-BM chimera mice every third day with DT over a 15-day period significantly decreased CD11c⁺ DCs in the colon, spleen and MLN (Fig. S2A,B). Compared to DT injected WT B6 controls, DT-injected B6 DTR-BM chimera mice had more severe colonic mucosal injury and inflammation characterized by multiple deep mucosal ulcerations after infection (Fig. 5A), significantly higher levels of mucosal MPO (Fig. 5B), increased numbers of C. rodentium in the feces and spleen (Fig. 5C,D), and lower serum titers of anti-C. rodentium IgG antibody (Fig. 5E). This defective host response in DC-depleted mice after C. rodentium infection resembled that in infected GM-CSF^-/- mice (Fig. 2). DT injection had no impact on mucosal histology, MPO levels in the colon, the magnitude of C. rodentium colonization (as assessed by fecal CFU), or IgG anti-C. rodentium antibody titers after infection of control B6 BM chimera mice or WT B6 mice (Fig. S3C-F).

Figure 5. C. rodentium-induced colitis in CD11c⁺ DC depleted mice.

(A) H&E staining of the colon from 2 week-infected B6 and B6 DTR-BM chimera mice treated with DT. Original magnification, x 100. Scale bar is 200 μm. (B) MPO in colon tissue from 2 week infected mice treated with DT was quantified by ELISA. Data are mean ± SEM (n = 10). (C) Fecal CFU of C. rodentium were determined at 1 and 2 weeks after oral gavage of 5 × 10⁸ C. rodentium to B6 and B6 DTR-BM chimera mice treated with DT. Data are mean ± SEM (n = 9 mice per group). (D) C. rodentium CFU in the spleens of DT treated mice were determined 2 weeks after infection. Data are shown from individual mice. Bar represents the geometric mean (n = 9). Dotted line indicates the detection limit of the CFU assay. (E) Serum anti-C. rodentium IgG at the indicated times after infection with C. rodentium in B6 and B6 DTR-BM chimera mice treated with DT was determined by ELISA. Data are mean ± SEM (n = 5-10 mice for each group at each time point). * p < 0.05 relative to B6 mice.

GM-CSF deficiency results in increased apoptosis of CD11c⁺ cells in the colon mucosa of infected mice

To address mechanisms responsible for decreased numbers of CD11c⁺ DCs in C. rodentium infected GM-CSF^-/- compared to WT B6 mice, we asked whether proliferation or survival of mucosal DCs was influenced by GM-CSF deficiency. Infected B6 and GM-CSF^-/- mice did not differ significantly in the number and distribution of proliferating cells as shown by BrdU staining of the intestinal mucosa (Fig. 6A), with most proliferating cells being epithelial cells. However, the numbers of apoptotic cells were significantly increased in the colon of C. rodentium infected GM-CSF^-/- mice compared to B6 mice (Fig. 6B). Most TUNEL-positive cells in infected GM-CSF^-/- mice were located in the lamina propria and not the epithelium (Fig. 6C). Colon sections analyzed for CD11c⁺ DCs by immunostaining and TUNEL+ cells confirmed a significant increase in TUNEL-positive DCs in infected GM-CSF^-/- mice (Fig. 6D). Further, the percentage of TUNEL+ DCs was significantly greater among CD11c⁺ cells isolated from the lamina propria of 2-week infected GM-CSF^-/- compared to B6 mice (Fig. 6E). Thus, GM-CSF appeared to be more important for mucosal DC survival than DC proliferation after C. rodentium infection.

Figure 6. DC proliferation and apoptosis in C. rodentium infected mice.

(A) BrdU staining (red) in colon sections of C. rodentium-infected (2 weeks) B6 and GM-CSF^-/- mice. Nuclei were stained with Hoechst 33258 (blue). Original magnification, x 200. (B) TUNEL staining (green, examples of apoptotic cells shown by arrows) of colon sections from 2 week-infected B6 and GM-CSF^-/- mice. F-actin (red) was stained with Alexa-Fluoro 568-conjugated phalloidin. Original magnification, x 200. (C) Colon sections from 2 week-infected GM-CSF^-/- mice. TUNEL staining (green), F-actin staining (red), nuclear staining (blue, Hoechst 33258 dye) and merged image. Arrows indicate TUNEL-positive cells in the lamina propria and arrowheads indicate TUNEL-positive epithelial cells, respectively. Original magnification, x 200 (D) Colon sections from 2 week-infected GM-CSF^-/- mice. TUNEL staining (green) and CD11c staining (red). Merged image and magnified image of the inset (right panel) are also shown. Arrows indicate TUNEL and CD11c double positive cells. Original magnification x 200. (E) TUNEL staining of CD11c⁺ DCs isolated from the colon of 2 week-infected B6 and GM-CSF^-/- mice. The percentage of TUNEL-positive cells in 3 different 400 x visual fields of each cytospin slide was calculated and is shown as mean ± SD. Results of a representative experiment are shown. Similar results were obtained in 2 repeated experiments. *p<0.05 relative to B6 mice.

GM-CSF dependent CCL22 production by intestinal epithelial cells determines CD11c⁺ DC recruitment and localization

Chemokines have a major role in DC recruitment and localization in peripheral tissues. We tested the possibility that increased GM-CSF in the colon of infected mice regulated the expression of mucosal DC chemoattractants. The chemokine CCL22 (also termed macrophage-derived chemokine, MDC) was significantly increased at the mRNA and protein level in the colon of C. rodentium infected WT B6 but not in GM-CSF^-/- mice (Fig. 7A,B). Moreover, CCL22 was mostly produced by colon epithelial cells after infection (Fig. 7C,D, Fig S3). In parallel, CD11c⁺ DCs were prominently localized in regions immediately beneath surface colon epithelial cells with upregulated epithelial CCL22 expression in infected WT B6 mice (Fig. 7E, see also Fig. 3A). By comparison, in uninfected B6 mice, which have little if any CCL22 expression, DCs were few and localized mainly near the base of the crypts (see Fig. 3A). The loss of CCL22 expression in the absence of GM-CSF was selective, since expression of another epithelial cell-derived DC chemoattractant chemokine, CCL20 (MIP-3α), was significantly increased in GM-CSF^-/- compared to WT B6 mice after C. rodentium infection (Fig. 7A), whereas the DC chemoattractant, CCL8 (MCP-2), was expressed at similar levels in infected WT B6 and GM-CSF^-/- mice (Fig. 7A).

Figure 7. GM-CSF upregulates epithelial cell expression of the DC chemoattractant CCL22.

(A) mRNA transcripts for the three DC chemoattractant chemokines, CCL22, CCL8 and CCL20, were assayed by real time RT-PCR in the colon of B6 and GM-CSF^-/- mice left uninfected or infected with C. rodentium (n = 4-5 mice per group at each time point). (B) CCL22 protein was assayed in colon extracts from B6 and GM-CSF^-/- mice. Data are mean ± SEM (n = 5 mice per group at each time point). For (A, B) * p < 0.05 relative to B6 mice at the indicated time points. (C) Colon sections from B6 mice left uninfected and B6 and GM-CSF^-/- mice infected for 3 weeks with C. rodentium were stained for CCL22 (red). Original magnification, x 200. (D) Colon sections from B6 mice infected for 3 weeks were stained for CCL22 (red) and β-catenin (green) as an epithelial cell marker. Nuclei were stained with Hoechst 33258 (blue). Specificity controls using control antibody or anti-CCL22 absorbed with CCL22 peptide showed no immunostaining (see Fig. S3). (E) Colon sections from B6 mice infected for 3 weeks were stained for CCL22 (red) and CD11c (green). Original magnification, x 400. (F) WT B6 mice were infected with C. rodentium for 2 weeks. On days 9, 11 and 13 after infection mice were injected i.p. with 20 μg IgG goat anti-CCL22 or control goat IgG.. Colon sections from anti-CCL22 and control goat IgG treated mice were stained for CD11c (red) and F-actin (green). Original magnification, x 200. (G) Cells isolated from colon lamina propria, spleen, and MLN of anti-CCL22 and control goat IgG treated mice were characterized by flow cytometry. Similar results were obtained in 2 independent experiments.

To directly address the importance of CCL22 in DC recruitment and localization during infection, C. rodentium infected B6 mice were injected with anti-CCL22 neutralizing antibody or control IgG. The mucosa of CCL22 antibody-treated B6 mice infected with C. rodentium lacked focal accumulations of DC in the subepithelial regions underlying surface epithelial cells (Fig. 7F), and numbers of CD11c⁺ cells were significantly decreased in the lamina propria, but not in the spleen or MLN (Fig. 7G). Consistent with this, at 2 weeks after infection after infection, a significant 5-fold increase in colon colonization with C. rodentium (p<0.05, n=3) was observed in B6 mice injected with anti-CCL22 relative to those injected with isotype control antibody. These results indicate a key role for GM-CSF, acting in concert with CCL22, in the recruitment and retention of DCs in the mucosa upon C. rodentium infection.

Discussion

A/E lesion-inducing enteric pathogens are significant causes of diarrhea and life-threatening disease in humans. Like enteropathogenic and enterohemorrhagic E. coli in humans, the murine A/E pathogen, C. rodentium, interacts with and signals via the intestinal epithelium, but does not invade the mucosa of WT mice to a significant degree, raising fundamental questions about the nature of effective defense against minimally invasive mucosal pathogens. We show here that GM-CSF produced in the intestinal mucosa is an important mediator of host defense against C. rodentium, since mice deficient in GM-CSF manifested increased mucosal colonization with the bacteria, more severe and prolonged colonic mucosal inflammation, delayed bacterial clearance, and increased systemic infection.

CD11c⁺ dendritic cells, intestinal epithelial cells, and interactions between products of those cells were key for GM-CSF-mediated host defense. GM-CSF, alone and in combination with IL-4, stimulates the generation of DCs from bone marrow or peripheral blood mononuclear cells in culture, but its physiological role in the generation and survival of DCs in the intestinal mucosa in vivo and in host protection from enteric pathogens was not known. We considered that decreased numbers of DCs in the mucosa of infected GM-CSF^-/- mice might reflect a developmental defect or the lack of differentiation of a population of DCs that populate the colon lamina propria, as GM-CSF has been reported in two recent studies to be variably important in the preferential development/differentiation of the lamina propria CD11b⁺CD103⁺ DCs (Bogunovic et al., 2009) or alternatively CD11b⁺ but mainly CD103^- DCs (Varol et al., 2009). Alternatively, a defect might exist in recruitment of the major DC subpopulations to the lamina propria. However, we found no evidence in support of those possibilities in our infection model, since the proportion of the predominant CD11b⁺CD103⁺ DCs relative to CD11b⁺CD103^- cells in the lamina propria of infected WT B6 and GM-CSF^-/- mice did not differ significantly. We also detected no difference between WT B6 and GM-CSF^-/-mice in the relative number of DCs expressing CCR7, a receptor which is preferentially expressed on CD11b⁺CD103⁺ DC in the lamina propria (Bogunovic et al., 2009) and controls DC migration to draining lymph nodes (Forster et al., 1999; Johansson-Lindbom et al., 2005; Ohl et al., 2004). Moreover, supplementing GM-CSF deficient mice with GM-CSF restored DCs in the colon to levels approaching those in WT B6 mice, indicating GM-CSF deficiency does not cause an irreversible developmental defect in DCs. Rather, we show that GM-CSF is important for providing survival signals to intestinal DCs, since their numbers and viability were decreased after infection in the absence of GM-CSF, most likely due to increased apoptosis. Consistent with CD11c⁺ DC having a central role in GM-CSF dependent mucosal defense, depletion of CD11c⁺ DC in infected WT mice recapitulated the major abnormalities in infected GM-CSF^-/- mice. These results provide a paradigm for the role of GM-CSF during infection with A/E pathogens, although we recognize that additional or other mechanisms may come into play during infection with enteroinvasive pathogens that use different strategies for invasion and evading host defense mechanisms.

We show that the C-C chemokine, CCL22, produced by intestinal epithelial cells after infection in a GM-GSF dependent manner, has a major role in the recruitment of CD11c⁺ DCs and their localization to the subepithelial region of surface epithelial cells, as, neutralization of CCL22 in WT mice abrogated the subepithelial recruitment of CD11c⁺ DC. These results provide an important physiological correlate to the in vitro observation that CCL22 is chemotactic for monocyte-derived and bone marrow DCs (Bonecchi et al., 1998; Vecchi et al., 1999). CCL22 also chemoattracts CD4⁺ T cells, NK cells, and monocytes in vitro, although we did not note decreases in these cells in GM-CSF deficient mice. Furthermore, the observed predominant epithelial production of CCL22 is consistent with the reported inducible production of the chemokine in human intestinal epithelial cell lines (Berin et al., 2001). Other cell types, including monocyte-derived DCs, macrophages, pro B cells, and T cells (Godiska et al., 1997; Mantovani et al., 2000), can produce CCL22, yet we did not observe CCL22 production in those cells despite their abundance in the intestinal mucosa after infection. In cultured epithelial cell lines, IFNγ and TNFα upregulate CCL22 production (Berin et al., 2001), yet induction of these cytokines was not impacted by GM-CSF deficiency, suggesting that in the absence of GM-CSF those cytokines were not sufficient to upregulate epithelial CCL22 production. However, whether CCL22 production after infection in vivo is mediated by a direct or indirect activity of GM-CSF on epithelial cells is not known. In contrast to CCL22, another DC chemoattractant, CCL20, important for DC infiltration and mucosal host defense during parasitic infection with Trichuris muris (Cruickshank et al., 2009), was not required for DC recruitment or localization in C. rodentium infected WT or GM-CSF^-/- mice (Y. Hirata et al., data not shown),

GM-CSF had a non-redundant role in host defense against C. rodentium that affected several phases of the host response to infection, including the extent of surface colonization, the mucosal inflammatory response, and the later adaptive immune response required for clearing infection. GM-CSF contrasts with several other cytokines and antimicrobial peptides that can contribute to mucosal defense to C. rodentium (Dann et al., 2008; Goncalves et al., 2001; Iimura et al., 2005; Mundy et al., 2005; Simmons et al., 2002; Spahn et al., 2004; Zheng et al., 2008). For example, GM-CSF^-/- mice, but not mice deficient in IL-6 (Dann et al., 2008), had an increased burden of infection by 1 week (but not very early at 4 days) after infection and decreased antibacterial antibodies essential for bacterial clearance (Maaser et al., 2004). Despite markedly increased expression of IFNγ and TNFα, which play a role in bacterial clearance (Goncalves et al., 2001; Simmons et al., 2002), mice deficient in GM-CSF^-/- mice had delayed clearance of the pathogen, indicating that GM-CSF has unique immune defense functions that cannott be compensated for by these two cytokines.

The intestinal mucosa of GM-CSF^-/- mice was histologically normal in the absence of infection. However, the inflammatory response in infected GM-CSF^-/- mice compared to WT mice was markedly more severe and characterized by increased expression of proinflammatory cytokines, neutrophil and monocyte/macrophage chemoattractants, and an accompanying infiltration of F4/80⁺ macrophages and Gr-1⁺ neutrophils. Nonetheless, the increased inflammatory response did not compensate for the indispensable host protective role mediated by GM-CSF in the intestinal tract. GM-CSF also has important functions in host defense against pulmonary infection with Group B Streptococcus, Pseudomonas aeruginosa and Pneumocystis carinii (Ballinger et al., 2006; LeVine et al., 1999; Paine et al., 2000). However, in contrast to enteric infection with C. rodentium, where CD11c⁺ DC primarily mediate GM-CSF dependent host defense, the functions of GM-CSF in defense against those pulmonary pathogens appear mainly dependent on the activity of alveolar macrophages (Ballinger et al., 2006; LeVine et al., 1999; Paine et al., 2000).

GM-CSF augments the survival of neutrophils and monocyte/macrophage, and the production of proinflammatory chemokines and mediators by these cells (Hamilton, 2008; Hamilton and Anderson, 2004). Administration of GM-CSF in vivo provokes inflammation and exacerbates autoimmune diseases, such as inflammatory arthritis and experimental autoimmune encephalitis (EAE) in mice and rheumatoid arthritis in humans (Bischof et al., 2000; Campbell et al., 1997; Hamilton, 2008; Marusic et al., 2002). Conversely, mice deficient in GM-CSF^-/- were protected from collagen-induced arthritis (Campbell et al., 1998; McQualter et al., 2001) and EAE. In Crohn’s disease, a chronic inflammatory intestinal disease, two modes of therapy known to decrease GM-CSF production, anti-TNFα antibodies and glucocorticoids, are widely used to attenuate inflammation. However, contrary and counterintuitive to this, GM-CSF administration to Crohn’s disease patients resulted in clinical improvement and less mucosal inflammation (Dieckgraefe and Korzenik, 2002; Korzenik et al., 2005). The present data on host mucosal response to an A/E pathogen may have relevance for understanding the abnormal host response to the enteric microbial flora in Crohn’s disease, which is thought to be a significant factor in the pathogenesis of the disease and in mouse models of chronic intestinal inflammation (Salzman and Bevins, 2008; Wirtz and Neurath, 2007). Thus, our studies indirectly suggest that DC functions in response to microbial stimuli may be deficient in patients with Crohn’s disease, perhaps explaining the increase in chronic infiltration with acute and chronic inflammatory cells and the beneficial effects of exogenous administration of GM-CSF in this disease.

Experimental Procedures

Mice

Wild-type (WT) C57BL/6J (B6) mice and B6.FVB-Tg(Itgax-DTR/EGFP)57Lan/J mice were from The Jackson Laboratory. GM-CSF^-/- mice, backcrossed for more than 10 generations to B6 mice, were kindly provided by Dr. B. Trapnell (Children’s Hospital Medical Center, Cincinnati, Ohio), and maintained in the University of California, San Diego, animal care facilities. All animal studies were approved by the UCSD Institutional Animal Care and Use Committee.

Bone marrow chimera mice and dendritic cell depletion

To generate bone marrow (BM) chimeric mice, WT B6 recipients were exposed to a single lethal dose of 9 Gy total body irradiation from a ¹³⁷Cesium source and, on the following day, were injected i.v. with 2-5×10⁶ BM cells from WT B6 or B6.FVB-Tg (Itgax-DTR/EGFP)57Lan/J mice (hereafter referred to as B6 WT-BM chimera and B6 DTR-BM chimera, respectively)(Zaft et al., 2005). Mice were rested for 8 weeks before use. For DC depletion, C. rodentium infected WT B6, B6 WT-BM chimera, and B6 DTR-BM chimera mice were injected i.p. with 100 ng diphtheria toxin (Sigma-Aldrich) immediately after infection, and every third day thereafter during the 15 day experimental period.

Infections and CFU assays

C. rodentium was grown overnight in Luria-Bertani broth at 37°C. Mice (8 to 12 weeks) were infected with 5 × 10⁸ C. rodentium by oral gavage (Eckmann, 2006; Maaser et al., 2004). We found that this inoculum uniformly infects WT B6 mice, yet is not lethal, with pathogen clearance occurring by 3 weeks post infection. Bacterial numbers in feces and homogenized tissues were determined by plating serial dilutions onto MacConkey agar. The detection limit of the CFU assay was 10³ colonies per g of feces and 10 colonies per organ. Samples without detectable C. rodentium colonies were assigned a log₁₀ value equivalent to half of the detection limit of the CFU assay (Maaser et al., 2004).

Antibodies and Reagents

The following monoclonal antibodies were used: rat anti-mouse F4/80 (biotinylated-, and FITC-conjugated, IgG2b), rat anti-mouse GM-CSF (IgG2a), and rat anti-mouse CCR7 (FITC-conjugated, IgG2b) from Serotec. Rat anti-mouse Gr-1 (biotinylated-, and FITC- conjugated, IgG2b), hamster anti-mouse CD11c (biotinylated-, FITC-, and PE-conjugated, IgG1), and murine anti-mouse I-A^b (FITC-conjugated, IgG2a) from BD Biosciences. Hamster anti-mouse CD40 (FITC-conjugated, IgM), hamster anti-mouse CD80 (FITC-conjugated, IgG), rat anti-mouse CD86 (FITC-conjugated, IgG2a), and rat anti-mouse CD16/32 (IgG2a) from eBiosciences. Rat anti-mouse CCR6 (FITC-conjugated, IgG2a) and rat anti-mouse GM-CSF (IgG2a) from R&D systems. The following polyclonal antibodies were used: goat anti-mouse and rat CCL22 (IgG) from Santa Cruz for immunostaining experiments, goat anti-mouse CCL22 (IgG) from R&D systems for in vivo neutralization experiments, rabbit anti-FITC (Alexa 488-conjugated, IgG) from Molecular Probes. Control goat IgG was from R&D Systems. Pegylated mouse GM-CSF (mGM-CSF) was kindly provided by Dr. E. Croze, Berlex Bioscience.

Histologic analysis and immunohistochemistry

Colons were processed as Swiss rolls (Maaser et al., 2004). Tissues were fixed in 10% formalin, embedded in paraffin, and 5 μm sections were stained with hematoxylin and eosin (H&E). Colonic crypt depth was determined using imaging software (PictureFrame, Optronics). Well-oriented crypts, at least 10 crypts apart, were measured throughout the distal colon and the 3 highest values obtained were used to calculate maximal crypt depth (Maaser et al., 2004).

For immunohistochemistry, tissues were frozen in OCT compound and 5 μm sections were fixed in acetone for 5 min, washed and blocked with PBS/2% BSA, followed by avidin biotin blocking (Avidin Biotin blocking kit, DAKO). Sections were incubated with primary antibody overnight. Antibodies and dilutions used in immunohistochemistry were anti-GM-CSF (1:25), biotinylated anti-CD11c (1:400), biotinylated anti-F4/80 (1:50), biotinylated anti-Gr-1 (1:400), and anti-CCL22 antibody (1:400). Antibodies conjugated with biotin (dilution 1:100) were detected using Cy3-labeled streptavidin (Jackson Laboratories). F-actin and nuclei were stained with Alexa 488-conjugated phalloidin and Hoechst 33342, respectively. For co-staining, sections stained as above were blocked with 2% BSA and 10% rat, or goat serum, after which FITC-conjugated anti-CD11c (1:400), FITC-conjugated anti-F4/80 (1:100), or FITC-conjugated anti-Gr-1 (1:400) were added for 2 h, followed by incubation with Alexa 488-conjugated anti-FITC antibody (1:500). For BrdU staining, mice were injected i.p. with BrdU (1 mg) 2 h before sacrifice, and colon tissues were stained using a BrdU In-Situ staining kit (BD Biosciences). TUNEL staining on colon tissue and cytospin preparations used an ApoAlert DNA Fragmentation kit (Clontech Laboratories). Images from fluorescence microscopy were photographed using imaging software (PictureFrame). Digital images were processed using Adobe Photoshop 7.0 (Adobe Systems), with the same settings being applied for test and control antibodies.

RNA extraction, RT-PCR, Protein extraction, ELISA

Whole colon tissues were washed in PBS, cut into small pieces, and homogenized in lysis buffer. RNA was extracted and DNA was eliminated using an RNA extraction kit and DNase (Qiagen). Quantitative realtime RT-PCR was performed as described before (Hirata et al., 2007) using primer sequences as described before for TNF-α (Chae et al., 2006), MIP-2α, (Iimura et al., 2005), IFN-γ and IL-12 p40, (Wheeler et al., 2006), IL-1β (Schroppel et al., 2005), IL-17 (Lubberts et al., 2005), GM-CSF (Sakai et al., 1999), KC, CCL8, CCL20 and CCL22 (Lean et al., 2002). For protein extraction, small pieces of colon were homogenized in lysis buffer (50 mM Tris-Cl, pH 8.0, 1% NP-40, 150 mM NaCl, 100 mg/ml leupeptin, 1 mM PMSF, 5 mM Na₃VO₄) and centrifuged for 15 min at 15,000 rpm. GM-CSF, CCL22 (R&D systems), and myeloperoxidase (MPO, Hycult Biotechnology) in supernatants were determined by ELISA according to the manufacturers’ instructions. Assay sensitivities were <10 pg/ml for GM-CSF and CCL22, and 1 ng/ml for MPO. ELISA results were adjusted for tissue protein concentration determined by Bradford assay (Bio-Rad).

IgG and IgM anti-C. rodentium antibody

C. rodentium extract (5 mg/ml) in coating buffer (60 mM Na₂CO₃, 40 mM NaHCO₃, pH 9.6) was added to 96-well plates (Nunc) and incubated overnight at 4°C. Plates were washed and blocked with PBS containing 5% nonfat dry milk and 0.5% Tween 20 for 1 h. After washing, serial dilutions of mouse serum were added to wells and incubated for 2 h at RT. Wells were washed and optimal dilutions of peroxidase-conjugated anti mouse IgG or IgM was added for 1 h. Bound peroxidase was visualized with tetramethylbenzidine. Reactions were stopped with sulfuric acid and read using a microplate reader.

Colon lamina propria, spleen and MLN leukocytes

Colons were cut into small pieces and washed in ice cold PBS with 1 mM DTT. To remove epithelium, colons were incubated for 60 min in HBSS containing 5 mM EDTA, 10 mM Hepes, and 0.05 mM DTT at 37°C with shaking. After discarding supernatants containing epithelial cells, tissues were digested in RPMI containing 10% FBS, 200 U/ml of collagenase VIII (Sigma), and 50 μg/ml of DNase I (Roche) with a magnetic stirrer for 60 min. Digested tissues were passed through cell strainers, cells were washed with PBS and further purified by centrifugation on a Percoll gradient (70%/40%) for 20 min at 850 g. Cells at the gradient interphase were used as lamina propria leukocytes. For DC isolation, CD11c⁺ cells were enriched by positive selection using MACS Microbeads (Miltenyi Biotec), according to the manufacturer’s instruction. Preparations contained >90% CD11c⁺ cells by flow cytometry. Spleen cut into small pieces, and MLN were digested in HBSS containing 10% FBS, 25 μg/ml of DNase I, 1.5 mg/ml of Dispase (Roche Diagnostics), and 15 mM Hepes for 30 min. RBC in spleen preparations were lysed, MLN and spleen cells were washed and used for flow cytometry.

Flow cytometry

Isolated cells were incubated with 0.5 μg of anti-mouse CD16/32 for 30 min at 4°C to block Fc receptors. After blocking, surface markers were detected using FITC- or PE- conjugated anti-CD11c and, anti-I-A^b with appropriate control antibodies. For characterization of MACS-purified lamina propria DCs, FITC-conjugated CD11b, CD40, CD86, CCR6 CCR7 and PE-conjugated CD103 with appropriate control antibodies were used. Samples were analyzed by FACScan (BD Biosciences) using CellQuest software (BD Biosciences).

Endocytosis assay

DCs isolated from colon tissues (1 × 10⁵/sample) were incubated with 1 mg/ml FITC-dextran (40 kDa; Sigma) for 1 h at 37° C or at 4° C (control). After incubation, cells were washed 3 times in ice cold PBS containing 5% FBS, and analyzed by flow cytometry for mean fluorescence intensity (MFI). Endocytic activity was expressed as experimental MFI –control MFI.

Statistical analysis

Differences between samples were evaluated with the Mann-Whitney rank sum test or t test, as appropriate. p values < 0.05 were considered significant.