Luminal Bacteria Recruit CD103⁺ Dendritic Cells into the Intestinal Epithelium to Sample Bacterial Antigens for Presentation

SUMMARY

CD103⁺ dendritic cells (DCs) carry bacteria from the small intestine and can present antigens to T cells. Yet they have not been recorded sampling luminal bacteria or presenting bacterial antigens in mesentery lymph nodes. We used 2-photon microscopy in live Cx3cr1^+/gfp × Cd11c-YFP mice to study these processes. At steady state, sparse CD103⁺ DCs occupied the epithelium. They patrolled among enterocytes while extending dendrites toward the lumen, likely using tight-junction proteins to penetrate the epithelium. Challenge with Salmonella triggered chemokine- and toll-like receptor (TLR)-dependent recruitment of additional DCs from the lamina propria (LP). The DCs efficiently phagocytosed the bacteria using intraepithelial dendrites. Noninvasive bacteria were similarly sampled. In contrast, CD103⁺ DCs sampled soluble luminal antigen inefficiently. In mice harboring CD103⁺ DCs, antigen-specific CD8 T cells were subsequently activated in MLNs. Intestinal CD103⁺ DCs are therefore equipped with unique mechanisms to independently complete the processes of uptake, transportation, and presentation of bacterial antigens.

INTRODUCTION

To initiate adaptive immunity against enteric pathogens and to maintain tolerance toward commensal flora and food antigens (Ags), dendritic cells (DCs) acting as professional antigen-presenting cells (APCs) must survey the intestinal contents and deliver intestinal Ag to mesenteric lymph nodes (MLNs) (Coombes and Powrie, 2008; Rescigno, 2010; Varol et al., 2010; Scott et al., 2011).

A seminal study by Rescigno et al. (2001) has demonstrated that CD11c⁺ cells, believed to be DCs, send transepithelial dendrites (TEDs) from the lamina propria (LP) that penetrate through tight junctions and capture Salmonella from the lumen. It since has been recognized that the LP contains two developmentally-distinct populations of CD11c⁺ mononuclear phagocytes: CD11c^hi CD103⁺ CD11b⁺ CX₃CR1⁻ cells and CD11c^int CD103⁻ CD11b⁺ CX₃CR1⁺ cells (Bogunovic et al., 2009; Schulz et al., 2009; Varol et al., 2009). Of these, CD103⁺ cells are considered true DCs (Jaensson et al., 2008; Bogunovic et al., 2009; Schulz et al., 2009), whereas CX₃CR1⁺ cells are now commonly thought to be resident macrophages (Persson et al., 2010; Varol et al., 2010).

CD103⁺ DCs seem more adept than CX₃CR1⁺ macrophages at coordinating adaptive immunity against gut Ags: They express the chemokine receptor CCR7, allowing them to migrate to MLNs (Jaensson et al., 2008; Bogunovic et al., 2009; Schulz et al., 2009) carrying Ags from gut-encountered bacteria (Bogunovic et al., 2009). At the MLN, they excel at imprinting T cells for gut homing (Annacker et al., 2005; Johansson-Lindbom et al., 2005), and, depending whether the context is inflammatory or tolerogenic, either inducing effector T cells (Johansson-Lindbom et al., 2005; Laffont et al., 2010; Semmrich et al., 2012) or regulatory T cells (Annacker et al., 2005; Sun et al., 2007; Coombes and Powrie, 2008; Scott et al., 2011).

Paradoxically, CX₃CR1⁺ macrophages, rather than the CD103⁺ DCs, are the cells that have been consistently observed sampling the intestinal luminal content by extending TEDs (Rescigno et al., 2001; Niess et al., 2005; Chieppa et al., 2006; Vallon-Eberhard et al., 2006; Varol et al., 2009). Moreover, TLR-ligands (Varol et al., 2009) and contact with microbes (Vallon-Eberhard et al., 2006), including Salmonella, have been shown to induce TED formation (Chieppa et al., 2006) and do so in a CX₃CR1-dependent manner (Niess et al., 2005; Vallon-Eberhard et al., 2006).

Unlike CX₃CR1⁺ macrophages, CD103⁺ DCs have not been observed extending TEDs, and it remains unclear how they gain access to luminal bacteria. DCs may directly capture only those pathogens that actively invade the epithelium (Niess et al., 2005) or enter through villus M cells (Jang et al., 2004; Vallon-Eberhard et al., 2006). Alternatively, it has been speculated that CD103⁺ DCs acquire bacterial Ags from CX₃CR1⁺ macrophages (Rescigno, 2010; Scott et al., 2011) or extend their own protrusions toward the gut lumen (Chieppa et al., 2006; Rescigno, 2010). If the latter suggestion is true, some CD103⁺ DCs would have to be located close to the epithelium or inside it, rather than remain confined to the deep LP (Johansson-Lindbom et al., 2005; Schulz et al., 2009).

For uptake of soluble Ags, the molecular mechanisms may be entirely different. CX₃CR1⁺ macrophages have been shown to out-perform CD103⁺ DCs at such uptake and to do so independently of TEDs (Schulz et al., 2009). In contrast, a recent study (McDole et al., 2012), found CD103⁺ DCs to be superior, relying on interactions with Ag-channeling goblet cells.

In this study, we used intravital 2-photon microscopy in fluorescent reporter mice to study the dynamic response of CD103⁺ DCs to Salmonella challenge and the cellular behavior that underlies sampling of bacteria and soluble Ag. We found that bacterial challenge recruits CD103⁺ DCs from the LP into the epithelium, in which they crawl laterally while sending dendrites into the intestinal lumen. Luminal bacteria are captured using these dendrites and their antigens are subsequently presented in the MLNs of mice harboring CD103⁺ DCs.

RESULTS

At Steady State, Sparse CD11c⁺CX3CR1⁻ Cells Patrol the Intestinal Epithelium

To study dynamic cell behaviors in the small intestines, we developed a protocol for 2-photon microscopy of the luminal aspect of the ileum in live mice. Our procedure minimized peristaltic movement in the externalized tissue, preserved blood flow, and maintained the integrity of the epithelial surface (see Movie S1 available online). Using this protocol, we could observe both the epithelium and the LP for up to 4 hr.

To distinguish between mononuclear phagocytes in vivo, we crossed the Cd11c-YFP (Lindquist et al., 2004) and Cx3cr1^gfp (Jung et al., 2000) mouse strains. In this dual reporter strain, CD103⁺ DCs are expected to be YFP⁺ and GFP⁻ and appear yellow, whereas CX₃CR1⁺ macrophages should be YFP⁺ and GFP⁺ and appear light-colored.

As expected, 2-photon microscopy recorded highly-elaborate cells which constituted most of the fluorescent cells in the LP and were absent from the epithelium (Figure 1). These cells, consistent with earlier descriptions of CX₃CR1⁺ macrophages (Niess et al., 2005; Chieppa et al., 2006; Varol et al., 2009) expressed both CX₃CR1-GFP and CD11c-YFP and their exact hue depended on the ratio of these two fluorescent proteins (Figure 1A). CX₃CR1⁺ macrophages showed little lateral displacement, but projected highly-dynamic dendrites within the LP and occasionally in between epithelial cells (Movie S2).

Figure 1. A Population of Motile CX₃CR1-GFP⁻CD11c-YFP^int Cells Occupies the Intestinal Epithelium.

Two-photon microscopy in the ileum of a live Cx₃cr1^+/gfp × Cd11c-YFP mice.

(A) A cross-section through a villus; the LP is populated by sparse GFP⁻YFP^int cells and abundant GFP⁺YFP^hi cells, whose exact hue depends on the ratio of GFP to YFP (scale bar represents 25 μM).

(B) Examining reconstructed data from three axes highlights a GFP⁻YFP^int cell lodged within the epithelial layer.

(C) A cross-section through the apical epithelium of several adjacent villi shows 3 GFP⁻YFP^int cells occupying the space between epithelial cells (scale bar represents 25 μM).

(D) A GFP⁻YFP^intcell (arrow) fluoresces in the yellow channel but not the blue channel. Solid and dashed lines outline the luminal surface and basement membrane, respectively. (A–D, data are representative of at least 20 independent experiments).

(E) In 2-photon microscopy, large motile amoeboid CX₃CR1-GFP⁻ cells showed on average half the YFP intensity as CX₃CR1⁻GFP⁺ cells (horizontal bars denote averages, p < 0.0001, n = 65, five independent experiments).

(F) A YFP^int cell was followed in the epithelium for 9 min. The cell meandered between enterocytes, whose inferred positions are indicated by white polygons, without displacing them (scale bar represents 30 μM).

(G) Based on intravital 2-photon microscopy, the displacement rate of CD103⁺ DCs was significantly (p < 0.0001) higher than of X₃CR1⁺ macrophages imaged in the same villi (n = 27, three independent experiments). In vitro live-cell microscopy of CD103⁺ DCs (H and J) and CX₃CR1⁺ macrophages (I and K) sorted using flow cytometry from the LP revealed their intrinsic motility patterns (scale bars represent 30 μM).

(H) CD103⁺ DCs displayed amoeboid morphology.

(I) CX₃CR1⁺ macrophages adhered to the plate showing stellate morphology.

(J) CD103⁺ DCs crawled vigorously, using lamellipodia at the leading edge and trailing a distinct CD103⁻rich uropod; tracks follow 3 cells for 30 min.

(K) Similarly tracked CX₃CR1⁺ macrophages displayed little lateral displacement.

(L) Image analysis demonstrate that, in vitro, CD103⁺ DCs crawled four times faster than CX₃CR1⁺ macrophages (n = 50, p < 0.000.1, horizontal bars denote averages, representative of three independent experiments).

A separate cell population was CD11c-YFP^int and CX₃CR1-GFP⁻ (Figure 1). These large cells displayed amoeboid morphology (Figures 1A and 1B); most resided in the LP (Movie S2), but some, estimated at around 1–2 per villus, occupied the epithelium itself (Figures 1B and 1C; Movie S3). When compared to macrophages in the LP, these cells displayed less than half of the YFP intensity and could therefore easily go unnoticed (Figures 1D and 1E).

YFP^intGFP⁻ cells in the epithelium were motile and meandered, encircling the lateral aspects of epithelial cells, suggesting that they were located above the basal membrane (Figure 1F; Movie S3). Three-dimensional movement analysis recorded a significantly higher displacement rate for these cells over CX₃CR1⁺ macrophages (Figure 1G). Based on their morphology, motility, and fluorescence profile, we suspected that these cells were CD103⁺ DCs.

Isolated CD103⁺ DCs Show Intrinsic Motility

To test our hypothesis, we first examined the morphology and motility of CD103⁺ DCs in vitro. By using flow cytometry, we sorted CD103⁺ mononuclear phagocytes from the ileum and visualized their behavior (Figures 1H–1L; Movie S4). CD103⁺ DCs displayed amoeboid morphology (Figures 1H and 1J) while spontaneously and vigorously crawling at an average speed of 11 μm/min (Figures 1J and 1L). They extended lamellipodia at the leading edge and trailed distinct CD103⁺ uropods behind them (Figure 1J). In contrast, most CX₃CR1⁺ macrophages adhered to the plate, displaying stellate morphology and dynamic protrusions but showing little lateral movement (Figures 1I, 1K, and 1L).

CD11c-YFP^int Cells Are CD103⁺ DCs

We proceeded to establish the identity of CD11c-YFP^int cells as CD103⁺ DCs.

First, we excised the small intestine and rapidly fixed it to characterize CD11c-YFP^int cells using whole-mount fluorescent immunohistology. Staining with antilaminin indicated that some YFP⁺ cells resided directly above the basement membrane among enterocytes (Figure 2A), exhibited intermediate YFP intensity (Figure 2A), and adopted the amoeboid morphology previously recorded using 2-photon microscopy (Figures 2B and 2C). In Cd11c-YFP × Cx3cr1^gfp mice, YFP⁺ cells in the epithelium never expressed GFP (Figure 2C). YFP^int cells expressed CD11c (Figures 2D, 2E, and 2G) and CD103 (Figures 2F–2H), but not CD4 or CD8 (Figures 2A–2C and 2E and 2F), indicating that they were not intraepithelial lymphocytes (IELs). This was further demonstrated by the presence of YFP^int CD11c⁺ cells in the epithelium of Rag1^−/− mice (Figure 2I), which lack intraepithelial T cells.

Figure 2. Immunohistology of the Epithelium Identifies CD11c-YFP^int Cells as CD103⁺ DCs.

Samples of the small intestine from untreated mice were rapidly excised and fixed for immunohistological analysis of whole-mounted tissues. Single Z-planes are shown (scale bars represent 10 μM).

(A) In Cd11c-YFP mice, not only CD4⁺CD8⁺ IELs but also some CD11c-YFP⁺ cells (arrow) were located directly above the laminin⁺ basement membrane among enterocytes. Such cells expressed less YFP than the highly-branched cells confined to the LP.

(B) Within the epithelial layer, amoeboid YFP⁺ cells (arrows) could be seen alongside IELs.

(C) A CX₃CR1-GFP⁻CD11c-YFP⁺ cell (in yellow) in the epithelium exhibited an elaborate shape and no T cell markers

(D) YFP^int cells (arrows), located beneath and among enterocytes expressed membranal CD11c.

(E) A YFP^int cell (arrow) expresses CD11c, but not CD4 or CD8, whereas adjacent IELs do.

(F) CD103 is expressed by the large YFP⁺ cells (arrow), which do not express CD4 or CD8.

(G) CD103 and CD11c are coexpressed by a large YFP⁺ cell (arrow) in the epithelium.

(H) CX₃CR1⁻GFP⁻CD11c-YFP^int cells, marked with membranal CD103⁺, can be located in the epithelium.

(I) YFP⁺ CD11c⁺ cells were also observed in the epithelium of Rag1^−/− mice (arrows). All micrographs are representative of at least six independent samples.

Second, we examined CD11c-YFP^int cells using flow cytometry. We harvested the small intestine of Cd11c-YFP mice, discarding the Peyer’s patches (PPs), and gently separated the LP from the epithelium. As expected, MHCII⁺ CD11b⁺ mononuclear phagocytes included two major populations: CD11c^hiCD103⁺ DCs and CD11c^intCD103⁻ cells (Figures 3A and 3B; Figure S1A). In Cx3cr1^+/gfp mice, almost all the latter cells were GFP⁺ (Figures 3A and 3B), so they are hereafter referred to as CX₃CR1⁺ macrophages. Although they expressed more CD11c, CD103⁺ DCs in the LP, epithelium (Figures 3A and 3B) and MLNs (Figure 3C) displayed lower YFP intensity than CX₃CR1⁺ macrophages, corresponding with 2-photon microscopy and immunohistological results.

Figure 3. CX₃CR1-GFP⁻CD11c-YFP^int Cells Are CD103⁺ DCs Responsive to GM-CSF.

(A and B) Following separation of the ileal tissue to the LP (A) and intraepithelial compartment (B), live MHC-II⁺ CD11c⁺ cells were analyzed by flow cytometry as detailed in Figure S1. Data are representative of at least five independent experiments. (A) In the LP, CX₃CR1⁺ macrophages outnumbered CD103⁺ DCs 3- to 4-fold (left). Even though CD103⁺ DCs expressed more membranal CD11c (center left), they showed less YFP fluorescence (center right). (B) In the epithelium, CD103⁺ DCs dominated, although contaminant CX₃CR1⁺ cells from the LP were also detectable. A similar relation between YFP and CD11c was observed.

(C) MHC-II⁺CD11b⁺CD11c⁺CD103⁺ DCs from the epithelium showed the same YFP intensity as did CD103⁺ DCs in the LP and MLN.

(D–I) GM-CSF specifically expanded GFP⁻YFP^int cells. Using 2-photon microscopy, we compared untreated Cx3cr1^+/gfp × Cd11c-YFP mice (D) with mice inoculated with a GM-CSF-secreting tumor (E). Arrows point to DCs in the epithelium (scale bar represents 50 μM). (F) GM-CSF expanded GFP⁻YFP^int cells 6-fold (p < 0.0001) in both the LP and epithelium, while GFP⁺YFP^hi cells were not affected noticeably (16 villi from two experiments, error bars indicate SEM for total numbers). (G) Flow cytometry analysis confirmed the proliferation of CD103⁺ DCs in the epithelium. Columns depict the percentages of CD103⁺ DCs out of hematopoietic cells based on data pooled from three experiments. (H) Whole-mount immunohistology confirmed that GFP⁻YFP^int cells in GM-CSF-treated mice expressed membranal CD103 (scale bars represent 10 μM). (I) Serial Z sections through a villus in a GM-CSF-treated mouse show multiple CX₃CR1-GFP⁻ CD11c-YFP^int cells (arrows), which did not express CD4 or CD8 markers (scale bars represent 30 μM). Images are representative of two experiments.

As noticed before (Bogunovic et al., 2009; Schulz et al., 2009), in the LP, CX₃CR1⁺ macrophages outnumbered CD103⁺ DCs 3- to 4-fold (Figure 3A; Figure S3A). However, in the epithelium, the opposite was true, and CD103⁺ DCs typically outnumbered CX₃CR1⁺ cells (Figure 3B; Figure S3A). Detailed analysis indicated that CX₃CR1⁺ macrophages found in epithelial preparations were a contamination from the LP, because these cells crossed the basal membrane during cell separation (Figure S1B; Movie S5) and gradually accumulated in the epithelial cell fraction (Figure S1C).

To examine the relationship between CD103⁺ DCs located in the epithelium and those in the LP, we compared the activation markers major histocompatibility complex class II (MHC-II), CD40, CD80, and CD86, and the linage markers CD115 (MCSF-R), GR-1, and CD135 (Flt3). The two populations were indistinguishable (Figure S1D) indicating that they did not represent distinct linages of DCs but belong to the same cell pool shared between the two locations.

Third, because CD11c is expressed not only by myeloid cells but also by some lymphoid cells, such as IELs (Huleatt and Lefrançois, 1995) and plasma cells (Hebel et al., 2006), we verified that we had not mistaken these cells for CD11c-YFP^int DCs. For this aim we further characterized the location, brightness, and behavior of CD11c⁺ lymphoid cells.

Flow cytometry analysis of Cd11c-YFP mice revealed that about 1% of their IELs were YFP^lo (Figure S2A), showing, on average, half the YFP intensity of CD11c-YFP^int DCs (Figure S2B). Most of these cells were too dim to identify using 2-photon microscopy (Lindquist et al., 2004), but because IELs dominated the epithelial leukocyte population, the brightest ones could indeed be imaged and appeared smaller and rounder than YFP^int DCs (Figures S2C and S2D). Fluorescent immunohistology failed to detect CD4⁺ or CD8⁺ IELs that expressed YFP, confirming their low fluorescence (Figure 2; Figure S2E).

Consistent with previous studies (Hebel et al., 2006), around 20% of intestinal immunoglobulin A (IgA)-producing plasma cells were YFP^hi (Figure S2F). These cells, though, were immobile ovoid cells that resided deep in the LP (Figure S2G). No YFP was expressed by other immune cells, including neutrophils, B cells, plasmacytoid DCs, and eosinophils.

Finally, we examined the response of CD11c-YFP^int cells to granulocyte macrophage colony-stimulating factor (GM-CSF), which promotes proliferation of intestinal CD103⁺ CD11b⁺ DCs (Bogunovic et al., 2009; Schulz et al., 2009). Two-photon microscopy of mice transplanted with a tumor secreting GM-CSF (Mach et al., 2000) revealed that dynamic GFP⁻YFP^int cells expanded in both the LP and epithelium (Figures 3D–3F; Movie S6). Flow cytometry corroborated this finding (Figure 3G) while immunohistology confirmed that expanded GFP⁻YFP^int cells expressed CD103 (Figure 3H) but not T cell markers (Figure 3I).

Taken together, data from immunohistology, 2-photon microscopy, and flow cytometry strongly suggest that low numbers of CD103⁺ DCs reside at steady state not only in the LP, but also in the epithelium, and that they correspond to GFP⁻ YFP^int cells in Cx3cr1^+/gfp × Cd11c-YFP mice. We shall henceforth refer to these cells as CD103⁺ DCs.

Bacterial Challenge Recruits CD103⁺ DCs to the Epithelium

In correspondence with previous research (Rescigno et al., 2001; Niess et al., 2005; Chieppa et al., 2006; Vallon-Eberhard et al., 2006; Arques et al., 2009; Bogunovic et al., 2009; Uematsu and Akira, 2009; Varol et al., 2009), we used Salmonella typhimurium to study the response of CD103⁺ DCs to luminal bacteria. We applied χ4550 GFP⁺ Salmonella (Yrlid et al., 2001) to the luminal surface. As shown by 2-photon microscopy, within 30 min CD103⁺ DCs were mobilized from the LP, accumulated in the epithelium, and started crawling laterally within it (Figures 4A and 4B; Movie S7). DC recruitment was confirmed by immunohistological analysis (Figure 4C). We used flow cytometry to quantify DC recruitment. Injection of Salmonella into the ligated ileum (Rescigno et al., 2001) increased the frequency of CD103⁺ DCs in the epithelium (Figure 4D), raising their absolute numbers from ~2,500 to ~22,000 cells (Figure 4E). DCs were likely recruited from the LP, which experienced a simultaneous decrease in DC frequency. CX₃CR1⁺ macrophages, in contrast, did not mobilize.

Figure 4. Luminal Salmonella Recruit CD103⁺ DCs to the Epithelium.

(A and B) Two-photon microscopy of the epithelial layer in several adjacent villi exposed to Salmonella. (A) Some intraepithelial YFP^int CD103⁺ DCs were already visible at time 0, but within 30 min (B) they were joined by several other DCs that had migrated up into the epithelium (scale bar represents 50 μM).

(C) Immunohistology following Salmonella challenge captured several CD103⁺ DCs (arrows) within the epithelium. Image is representative of three experiments.

(D and E) By flow cytometry, Salmonella challenge increased the percentage of CD103⁺ DCs in the epithelium (p < 0.0001) and concomitantly decreased it in the LP (p = 0.003, data pooled from four experiments; percentages of DCs were calculated out of total live cells). (E) This translated to a marked increase in the total number of CD103⁺ DCs in the epithelium (p < 0.0001).

(F) We examined the ileal epithelium of WT mice, mice deficient in MyD88 and Ticam1 (TRIF) and chimeric mice whose hematopoietic cells are either MyD88⁻/⁻ × Ticam1⁻/⁻ or WT. Myd88⁻/⁻ × Ticam1⁻/⁻ mice had significantly lower percentages of intraepithelial CD103⁺ DCs (p = 0.012) and only in WT mice did Salmonella challenge significantly recruit CD103⁺ DCs, as indicated by ANOVA (p < 0.001, n = 27, data were pooled from two experiments and percentages calculated out of total live cells).

(G) Overnight treatment of mice with PTx significantly reduced the recruitment of CD103⁺ DCs to the epithelium (significant interaction at p < 0.001, n = 18, data pooled from two experiments). Error bars indicate SEM. Columns depict the percentages of CD103⁺ DCs out of total live cells.

Recruitment of CD103⁺ DCs Likely Depends on TLRs and Chemokines

To better understand molecular cues required for DC recruitment, we investigated signaling pathways downstream of chemokine and Toll-like receptors (TLRs).

We first tested whether TLRs could mediate DC recruitment. To that aim, we examined the intestines of Myd88^−/− × Ticam1^−/− double-deficient mice whose TLRs as well as inter-leukin-1 (IL-1) and IL-18 receptors cannot signal (Yamamoto et al., 2003). Although the steady-state frequency of CD103⁺ DCs in the LP of Myd88^−/− × Ticam1^−/− mice was normal (~1.4% of total cells), it was greatly reduced in the epithelium (Figure 4F). Moreover, Salmonella challenge in Myd88^−/− × Ticam1^−/− mice failed to recruit more CD103⁺ DCs from the LP to the epithelium (Figure 4F). To test which cells require TLR signaling for DC recruitment, we produced chimeric mice harboring (wild-type) WT nonhematopoietic cells and Myd88^−/− × Ticam1^−/− hematopoietic cells, as well as the reciprocal chimeras. In both groups, Salmonella failed to recruit CD103⁺ DCs (Figure 4F), likely indicating that both DCs and enterocytes need intact signaling cascades for recruitment.

Next we asked whether CD103⁺ DC recruitment is chemokine-dependent: We systemically treated mice overnight with pertussis toxin (PTx), known to block G protein-mediated (Gαi) signaling and shown to inhibit chemokine-induced migration of leukocytes in vivo (Huang et al., 2007; Fooksman et al., 2010). Although PTx did not reduce the steady-state frequency of CD103⁺ DCs in the epithelium (Figure 4G) and LP (Figure S3A), it significantly reduced the efficiency of DC recruitment by Salmonella (Figure 4G).

Taken together, these data establish that DC recruitment to the epithelium is an active chemokine-driven process that likely requires signaling downstream of MyD88 or TRIF in both DCs and epithelial cells.

A chemokine which might attract CD103⁺ DCs into the epithelium is CCL20, which is secreted by inflamed intestinal epithelial cells (Izadpanah et al., 2001), acts through the CCR6 receptor, and has been implicated in attracting DCs to the epithelium (Anjuère et al., 2004; Le Borgne et al., 2006; Cruickshank et al., 2009). Salmonella infection was shown to induce CCL20 secretion and encourage TED formation (Chieppa et al., 2006). Our analysis indicates that CD103⁺ DCs express some membranal CCR6 (Figure S3B), but neither CCL20 blockade (Figure S3C) nor deficiency in CCR6 in hematopoietic cells (Figure S3D) prevented their recruitment to the epithelium. Similarly, Cd103^−/− mice, which lack the αE integrin, had normal numbers of CD11c^hi DCs in the epithelium at baseline and following Salmonella challenge (Figure S3E), indicating that this integrin, although possibly involved in interactions with enterocytes, is not strictly required for entry into the intra-epithelial space.

CD103⁺ DCs in the Epithelium Actively Sample Luminal Salmonella

We next assessed whether, following migration to the epithelium, CD103⁺ DCs sample bacteria from the intestinal lumen.

We first assessed the intrinsic phagocytic capacity of CD103⁺ DCs in comparison to CX₃CR1⁺ macrophages, which are believed to be phagocytic (Rivollier et al., 2012). The DCs, isolated using flow cytometry and coincubated with Salmonella in vitro, proved as phagocytic as macrophages along varying bacterial concentrations (Figure 5A), indicating that they can sample these bacteria directly and efficiently.

Figure 5. Salmonella Uptake by CD103⁺ DCs Is Active and Chemokine-Dependent.

(A) We incubated 2 × 10⁶ cells containing all subsets of mononuclear phagocytes from the LP with increasing numbers of GFP⁺ Salmonella. CD103⁺ DCs were as efficient as CX₃CR1⁺ macrophages at ingesting Salmonella (representative of two experiments).

(B) After applying Salmonella to the ileal lumen, 2-photon microscopy could visualize GFP⁺ bacteria accumulating, within 30 min, inside intraepithelial CD103⁺ DCs (maximum intensity projection, left). Individual Z-planes (depths indicated) show that bacteria were internalized.

(C–E) Ileal loops were injected with GFP⁺ Salmonella. In vivo bacterial sampling by intestinal CD103⁺ DCs was assessed by flow cytometry. Error bars indicate SEM (C) CD103⁺ DCs in the epithelium were the first to capture Salmonella: by 2 hr, 2.5% the DCs in this compartment became GFP⁺, whereas no ingestion was detected in the LP. At 5 hr, bacteria further accumulated in intra-epithelial DCs but were also detected in DCs in the LP. (D) Intestines were injected with the invasive Salmonella strain χ4550 or with the noninvasive strain SB161. Both strains were sampled as efficiently by CD103⁺ DCs isolated from the epithelium at 2 hr (n = 13, data pooled from two experiments). (E) Overnight treatment of mice with PTx almost ablated Salmonella uptake at 2 hr (p < 0.0001, n = 18, two combined experiments).

(F) As CD103⁺ DCs crawl above the basement membrane, they extend dendrites through the epithelium (arrow on left). These extensions are thicker than the TEDs extended by CX₃CR1⁺ macrophages (arrow on right). Data are representative of at least ten independent experiments.

(G) In several instances, the dendrites of CD103⁺ DCs could be seen engulfing Salmonella (two bacteria circled), and retracting them toward their soma (scale bar represents 20 μM).

To establish whether CD103⁺ DCs can utilize this capacity in vivo, we imaged the intestine following challenge with GFP⁺ Salmonella. Observing the luminal aspect of the live ileum 1–2 hr after challenge, we could clearly visualize CD103⁺ DCs replete with intracellular bacteria (Figure 5B).

To quantify this uptake, we injected a ligated section of the ileum with GFP⁺ Salmonella and assessed phagocytosis in the LP and epithelium using flow cytometry (Figure 5C; Figure S4A). Salmonella first appeared in CD103⁺ DCs in the epithelium—by 2 hr, 2.4% of CD103⁺ DCs in this compartment contained bacteria, whereas ingestion in the LP remained negligible. At 5 hr, Salmonella⁺ DCs were also detected in the LP. ImageStream analysis indicated that the bacteria had been internalized (Figure S4B). To ensure that DCs did not pick up unwashed bacteria during cell isolation, we conducted it in the presence of gentamicin (Bowe and Heffron, 1994)—the efficiency of bacterial sampling was not reduced (Figure S4C).

Salmonella Uptake by CD103⁺ DCs Is an Active Chemokine-Dependent Process

The χ4550 Salmonella strain we had used (Yrlid et al., 2001), although attenuated, can attach to the epithelium and invade it (Schödel et al., 1994), potentially engaging CD103⁺ DCs. We therefore assessed whether CD103⁺ DCs in the epithelium can sample noninvasive SB161 Salmonella from the lumen. This strain can attach to the epithelium but cannot actively cross it due to a nonfunctional TTSS-1 gene (Hapfelmeier et al., 2005). CD103⁺ DCs sampled this strain as efficiently as they sampled the invasive χ4550 strain (Figure 5D). Correspondingly, intraepithelial CD103⁺ DCs could sample inert beads (Figure S4D) and were observed accumulating them and traveling back to the LP (Figure S4E).

Because the bacteria did not have to invade the epithelium to be taken up by CD103⁺ DCs, we hypothesized that bacterial uptake depended on chemokine-directed motility. In agreement with that, overnight treatment with PTx not only reduced CD103⁺ DCs recruitment to the epithelium (see above) but also severely impaired the ability of those DCs that had reached the epithelium to take up bacteria (Figure 5E).

Intraepithelial DCs Extend Dendrites into the Lumen and Capture Bacteria

While patrolling the epithelium, most CD103⁺ DCs could be visualized sending dendrites between epithelial cells into the lumen (Figure 5F; Movies S6 and S8). These dendrites differed morphologically from the previously-described TEDs extended by CX₃CR1⁺ macrophages (Niess et al., 2005). On the basis of morphometric analysis of 15 DCs, the dendrites had a larger diameter (~4.5 μm versus ~1.8 μm), narrowed toward their tips, and did not terminate in globular structures. On average, each such CD103⁺ DC extended 6.5 dendrites per hr, retracting them after ~3.5 min. The presence of luminal Salmonella did not stimulate more formation of such dendrites.

TED formation by CX₃CR1⁺ macrophages did not feature in our intravital preparations as prominently as in previous histological studies (Rescigno et al., 2001; Niess et al., 2005; Vallon-Eberhard et al., 2006; Varol et al., 2009). Although we could occasionally see such protrusions probing between epithelial cells (Movie S8), they were relatively rare, typically did not reach the lumen, and never terminated in globular structures. We could only observe such globular structures in explanted tissue, especially following removal of the epithelium (Figure S1B; Movie S5). Based on morphological considerations, these globular structures may indicate apoptosis (Brokaw et al., 1998).

To retrieve Ags located beyond an epithelial barrier, the dendrites of DCs must breach the tight junctions that seal the epithelial barrier at the luminal surface (Rescigno et al., 2001). To do that, they may be expressing tight junction proteins, as reported for lung CD103⁺DCs (Sung et al., 2006) and epidermal Langerhans cells (Zimmerli and Hauser, 2007). Indeed, real-time PCR analysis revealed that CD103⁺ DCs (as well as CX₃CR1⁺ macrophages) express the tight junction proteins Claudin-4 and ZO-2 (Figures S4F and S4G).

We proceeded to inspect the cellular mechanism of bacterial sampling directly. CD103⁺ DCs could be seen sending dendrites across the epithelial cell layer, and engulfing one or two Salmonella bacteria in the lumen (Figure 5G). Strikingly, the dendrites of CD103⁺ DCs rapidly retracted toward the soma after contacting Salmonella, bringing the captured bacteria with them (Movie S9).

Intraepithelial CD103⁺ DCs Capture Soluble Ag from the Lumen

To check whether the intraepithelial position of CD103⁺ DCs could also facilitate the capture of soluble proteins from the intestinal lumen, we injected an ileal section with fluorescently-labeled ovalbumin prior to flow cytometry analysis. CD103⁺ DCs in the epithelium were four times more likely to ingest OVA-Alexa-594 Ag than their counterparts in the LP (Figure 6A). However, in accordance with previous results (Schulz et al., 2009), CX3CR1⁺ macrophages in the LP took up soluble OVA more efficiently than CD103⁺ DCs—perhaps due to differential expression of scavenger receptors.

Figure 6. CD103⁺ DCs in the Epithelium Sample Soluble Ag.

(A) The ligated ileum was injected with Alexa-549-OVA and harvested at 2 hr to assess Ag uptake by MHCII⁺CD11c⁺ mononuclear phagocytes in the LP and epithelium using flow cytometry. Although CD103⁺ DCs in the LP failed to ingest OVA (top left), significantly more (p < 0.002) DCs in the epithelium could ingest it (top right). CX₃CR1⁺ macrophages in the LP (bottom) were significantly more efficient at ingesting OVA. Data are representative of three independent experiments.

(B and C) Two-photon microscopy of a Cx₃cr1-GFP ×Cd11c-YFP mouse 20 min after Alexa-549-OVA was applied to the mucosa (scale bar represents 20 μM). (B) CD103⁺ DCs in the epithelium extend dendrites and accumulate Ag inclusions (arrows). (C) Most CX₃CR1⁺ macrophages accumulate OVA in vacuoles. Ag uptake was not confined to macrophages contacting the epithelium. Ovoid CD11c-YFP^hi plasma cells are also visible. In the insets, OVA⁺ vesicles propagate along macrophages TEDs.

(D) OVA⁺ vacuoles are engulfed by macrophage membranes. The scale bar represents 30 μM; images are representative of at least five independent experiments.

To understand whether CD103⁺ DCs can use their dendrites to sample luminal soluble Ags, we imaged the uptake of OVA-Alexa-594 injected into the lumen. Following injection, rare CD103⁺ DCs accumulated the Ag inside small endocytic vesicles (Figure 6B) while actively crawling among epithelial cells in the apical villi. Several of these cells could be seen extending dendrites toward the lumen (Figure 6B; Movie S10). No such Ag-containing cells could be located in the LP.

Unlike CD103⁺ DCs, which rarely took up OVA, the majority of CX₃CR1⁺ macrophages ingested the Ag, corroborating our FACs results. OVA quickly concentrated in large cytoplasmic vacuoles inside macrophages (Figure 6C; Movie S10). Some cells could be visualized shuttling the Ag through delicate dendrites toward the soma (Figure 6C; Movie S10, inset), or engulfing large OVA inclusions (Figure 6D).

In accordance with a recent report (McDole et al., 2012), we could observe epithelial cells, morphologically consistent with goblet cells, containing OVA Ag. OVA-containing CX₃CR1⁺ macrophages, however, were dispersed throughout the LP, not necessarily in contact with the epithelium. So, although goblet cells are possible gateways for entry of soluble Ags from the lumen, we have no evidence for direct transfer of these Ags to mononuclear phagocytes in general and to CD103⁺ DCs in particular.

CD103⁺ DCs from the Epithelium Can Activate T Cells

Finally, we examined the potential of CD103⁺ DCs as APCs. Coculture of intraepithelial CD103⁺ DCs pulsed with ovalbumin protein with OT-I CD8⁺ T cells, resulted in T cell proliferation (Figure 7A) and in simultaneous upregulation of the gut-tropic α4β7 integrin (Figure 7B) showing that these cells can potentially participate in cross-presentation and imprinting. In addition, following coculture with Salmonella (Figure 7C), CD103⁺ DCs isolated from the epithelium, but not CX₃CR1⁺ macrophages, upregulated the CCR7 receptor, required for migration toward lymphatics (Jang et al., 2006) and into them (Tal et al., 2011). To study the involvement of CD103⁺ DCs in cross-presentation in vivo, we compared the activation of transferred OT-I T cells in the MLNs of WT mice and Flt3^−/− mice (Mackarehtschian et al., 1995) which have CX₃CR1⁺ macrophages but not CD103⁺ DCs in the LP (Bogunovic et al., 2009; Varol et al., 2009). Forty hours after OVA-GFP Salmonella had been introduced into the gut, OT-I T cells transferred into WT mice become activated, as evidenced by significantly upregulated CD69 compared to controls. In contrast, in Flt3^−/− mice, no such increase occurred (Figures 7D and 7E).

Figure 7. CD103⁺ DCs from the Epithelium Are Capable APCs, and Their Absence Impairs T Cell Activation.

(A and B) CD103⁺ DCs, either pulsed with OVA or used as controls were cocultured with CSFE-labeled OT-I CD8⁺ T cells. At 3 days, 54% of T cells have proliferated (A) and upregulated the gut-tropic α4β7 integrin (B). (A) and (B) are representative of three independent experiments.

(C) Following in vitro exposure to Salmonella for 5 hr, both CD103⁺ DCs and CX₃CR1⁺ macrophages upregulated MHC-II but only CD103⁺ DCs also upregulated CCR7 (is representative of two independent experiments).

(D) WT and Flt3⁻/⁻ mice were transferred i.v. with OT-I cells and inoculated with WT or OVA⁺ Salmonella in the gut. Analysis of CD69 expression in MLNs 40 hr later showed Ag-specific activation of OT-I T cells in WT (p < 0.001) but not Flt3⁻/⁻ (p = 0.84, n = 8) mice compared to the response to WT Salmonella, as quantified in (E). Error bars represent SEM.

DISCUSSION

This study observed CD103⁺ DCs as they patrol the epithelium of the small intestine and capture pathogenic bacteria attaching to its surface. We used 2-photon microscopy in live Cx3cr1^+/gfp × Cd11c-YFP mice to locate a population of GFP⁺ YFP^int cells that occupied the LP and the epithelium. The small epithelial population consisted of amoeboid cells that crawled, at relatively high speeds, on top of the basal membrane among enterocytes. Their identity as CD103⁺ DCs was corroborated based on immunohistology and their response to GM-CSF. Flow cytometry data, although it may over-represent CD103⁺ DC numbers due to contamination from the LP, also supported this identification. The findings demonstrated that YFP^int cells were neither CD11c⁺ lymphocytes nor contaminant cells from the LP. Phenotypically, intraepithelial CD103⁺ DCs were indistinguishable from those in the LP, indicating that they come from a common cellular pool spanning the two locations. Salmonella challenge recruited many more CD103⁺ DCs to the epithelium but only when both hematopoietic cells and nonhematopoietic cells had functional TRIF and MyD88; likewise, treatment with PTx blocked DC recruitment. Bacteria first appeared inside DCs in the epithelium, where they could be visualized using 2-photon microscopy, and only later in the LP. Sampling was inefficient in mice treated with PTx. The DCs expressed the tight-junction proteins Claudin-4 and ZO-2 and sent thick dendrites through epithelial tight junctions. These dendrites then engulfed Salmonella bacteria in the lumen and rapidly retracted toward the soma. CD103⁺ DCs located in the epithelium, but not in the LP, could also capture soluble protein from the lumen, but did so less efficiently than CX₃CR1⁺ macrophages. As indications that they can act as APCs, following maturation, intraepithelial CD103⁺ DCs upregulated CCR7, induced proliferation of Ag-specific CD8⁺ T cells and imprinted them for gut homing. Correspondingly, intraluminal Salmonella activated Ag-specific T cells in the MLNs of WT mice, but not Flt3^−/− mice, which lack intestinal CD103⁺ DCs.

These observations suggest a likely scenario for the interaction between CD103⁺ DCs and intraluminal bacteria: At steady state, most CD103⁺ DCs inhabit the LP, but sparse cells patrol the epithelium, maintained there by TLR signaling from the intestinal flora. The cells glide on the basement membrane below the tight junction layer. When bacteria attach to the luminal surface, CD103⁺ DCs capture them in a two-step process. First, enterocytes and sentinel DCs sense bacteria-derived TLR ligands and secrete chemokines to rapidly recruit more LP CD103⁺ DCs to the epithelium. Second, while patrolling this space, the newly-arrived DCs extend dendrites toward the lumen, relying on tight junction proteins to penetrate the tight junctions. The dendrites extend in response to chemokines and rapidly engulf bacteria. CD103⁺ DCs then upregulate CCR7, return to the LP, and emigrate through lymphatics to the draining MLNs, where they present the captured Ags to T cells.

Our findings place CD103⁺ DCs, armed with suitable cellular behavior, at an advantageous position to efficiently sample luminal Ags. For this task, they require assistance from neither CX₃CR1⁺ macrophages nor villous M cells. We thus suggest that CD103⁺ DCs can independently carry out the entire process of Ag uptake, transportation, and presentation.

Although we have not followed intraepithelial CD103⁺CD11b⁺ DCs as they migrate through lymphatics, their role in delivering pathogenic Salmonella to the MLNs has been demonstrated before (Bogunovic et al., 2009). Using a similar approach, we further showed that Flt3^−/− mice lacking CD103⁺ DCs show a poor CD8 T cell response in the MLNs against antigen from Salmonella. CD103⁺ DCs, therefore, may out-perform CX₃CR1⁺ macrophages not only in pathogen delivery, but also in inducing adaptive immune response against acute enteric infections. However, under chronic inflammation (Zigmond et al., 2012), or dysbiosis (Diehl et al., 2013), other mononuclear phagocytes make take over this role.

Whether CD103⁺ DCs would lose their intrinsic tolerogenic nature during Salmonella infection to become potent inflammatory APCs is still an open question. Nonetheless, complementary pathways that involve bacterial transcytosis through M cells in PPs (Jones et al., 1994) and isolated lymphoid follicles (Halle et al., 2007) are probably the ones to deliver most of the luminal Salmonella into DCs and likely contributed to the in vivo T cell activation we have observed. Transfer of particulate bacteria through goblet cells is less likely, because these cells failed to transfer large proteins and beads into the lumen (McDole et al., 2012).

Unlike the lung, where CD103⁺ DCs maintain close association with the basolateral side of bronchial epithelial cells (Sung et al., 2006), previous studies of the gut have concluded that CD103⁺ DCs are located deeper in the LP than CX₃CR1 macrophages (Johansson-Lindbom et al., 2005; Schulz et al., 2009). Others, though, reported that CD103⁺ DCs amount to a third of the CD11c⁺ cells in the epithelium (Bogunovic et al., 2009) and could locate them in histological sections of the jejunum (McDonald et al., 2012). MHC-II⁺ DCs could also be identified in the epithelium of the rat colon and jejunum (Maric et al., 1996). In agreement with these latter studies, we show that, even at steady state, CD103⁺ DCs occupy the epithelium and crawl among enterocytes. Electron microscopy of the basement membrane of the intestinal villi has revealed pores through which lymphocytes and macrophages sent protrusions or trafficked (Takahashi-Iwanaga et al., 1999). CD103⁺ DCs may thus take advantage of these pores to migrate from the LP to the epithelium and back.

Although several groups have imaged the small intestine of Cd11c-YFP mice (Schulz et al., 2009; McDole et al., 2012), they have not identified this YFP^int DC population within the epithelium. The cells may have been missed because, at 1–2 cells per villus, they are relatively rare under steady-state conditions, and because they are dim, expressing ~2.5 times less YFP than nearby CX₃CR1⁺ macrophages.

A recent study by McDole et al. (2012) used Cx3cr1^+/gfp × Cd11c-YFP mice and equated CD103⁺ DCs with a population of YFP^hi GFP^lo cells imaged in the LP. These cells appeared to form the majority of YFP⁺ cells in the LP, and sampled luminal Ag more efficiently. This finding, however, is at odds with previous literature showing that CD103⁺ DCs are outnumbered 3 to 1 by CX₃CR1⁺ macrophages in the LP (Bogunovic et al., 2009; Schulz et al., 2009; Kinnebrew et al., 2012), and are poor at OVA uptake (Schulz et al., 2009). Here we have directly shown that YFP^hi cells in these reporter mice are actually CX₃CR1⁺ macrophages.

Surprisingly, our imaging and flow cytometry findings show that although they express the highest amounts of membranal CD11c, CD103⁺ DCs in the LP, epithelium and MLNs show intermediate YFP intensity. Perhaps the longer turnover of CX₃CR1⁺ macrophages (Jaensson et al., 2008; Schulz et al., 2009) allows them to accumulate more YFP. Alternatively, the CD103⁺ DCs may require additional transcriptional control elements to strongly express YFP. Similarly, intestinal plasmacytoid DCs in Cd11c-YFP mice do not express any YFP, even though they express some CD11c (Nakano et al., 2001).

The presence of CD103⁺ DCs in the epithelium, rather than only in the LP, could solve two puzzles: First, in vitro studies suggest that DCs are conditioned by epithelial cells to express RALDH2 (Iliev et al., 2009), and so produce abundant retinoic acid (Iwata et al., 2004), a molecule critical for shaping the homing patterns and cell fate of intestinal T lymphocytes (Mucida et al., 2007; Sun et al., 2007; Coombes and Powrie, 2008). The intimate DC-epithelial contacts that we have demonstrated would allow direct conditioning of CD103⁺ DCs, as recently suggested (McDonald et al., 2012). Second, epithelial location would explain why intestinal DCs express their hallmark membranal marker—CD103. This molecule, also known as the αE integrin subunit, couples with the β7 chain to form the αEβ7 integrin (Parker et al., 1992). In this capacity it binds E-cadherin (Cepek et al., 1994), an adhesion molecule found on the basolateral membranes of epithelial cells (Boller et al., 1985; Hermiston and Gordon, 1995). CD103 expression, together with the tight junction proteins Claudin-4 and ZO-2, could assist DCs in navigating among enterocytes and sending dendrites through their connecting junctions. Indeed, a role for CD103 in dendrite formation has been established in epidermal T cells (Schlickum et al., 2008). Our studies in Cd103^−/− mice, though, did not find this molecule to be essential for DC recruitment to the epithelium.

Previously, Salmonella was shown to trigger the formation of TEDs by mononuclear macrophages; the cells themselves remained in the LP (Rescigno et al., 2001; Niess et al., 2005; Chieppa et al., 2006). Here in contrast, the bacteria recruited the entire cell bodies of CD103⁺ DCs into the epithelium. The processes sent by these DCs are “intraepithelial,” rather than “transepithelial” dendrites. Although the latter may exist, true TEDs sent by CD103⁺ DCs are unlikely contributors to bacterial sampling. Indeed, within 2 hr of Salmonella injection, LP DCs phagocytosed negligible numbers of bacteria (Figure 5C; Figure S7A). This suggests that, as a rule, to gain access to luminal bacteria, DCs must first perform “step 1”—recruitment—before going through “step 2”—sampling.

Correspondingly, mononuclear phagocytes have been recorded in the epithelium before: CX₃CR1⁺ cells migrated unidirectionally into the lumen of the small intestine following infection with flagellated Salmonella (Arques et al., 2009); DCs were shown to populate the colonic epithelium following infection with Trichuris muris (Cruickshank et al., 2009); most pertinently, DCs whose marker profile is consistent with CD103⁺ DCs were found in the epithelium of the small intestine after oral administration of cholera toxin (Anjuère et al., 2004).

It remains to be seen which specific TLR ligands, expressed by Salmonella or released by it, recruit CD103⁺ DCs into the epithelium. These DCs were shown to express TLR5 and TLR9, but not TLR2, TLR3, TLR4, or TLR7 (Uematsu et al., 2006, 2008; Fujimoto et al., 2011; Kinnebrew et al., 2012). Notably, TLR5 deficiency impaired the transport of Salmonella from the intestinal tract to MLNs (Uematsu et al., 2006), raising the possibility that migratory DCs might require TLR5 to sample Salmonella. The epithelial cells of the small intestine express a wider variety of TLRs (Abreu, 2010), including TLR1 through TLR9 and TLR11, and the engagement of these receptors may trigger the recruitment of CD103⁺ DCs, much like LPS triggers the extension of TEDs in the distal ileum (Chieppa et al., 2006).

This division of labor may explain our finding using bone marrow transfer: To recruit CD103⁺ DCs to the epithelium, both hematopoietic cells (likely DCs) and nonhematopoietic cells (likely enterocytes) had to possess intact MyD88 and TRIF signaling. This finding suggests a cascade of events in which cells exposed to the luminal content (e.g., enterocytes or Paneth cells) recognize the bacteria through TLRs and induce local inflammation through cytokines, such as IL-1 and IL-18, whose receptors are MyD88-depndent (Adachi et al., 1998).

The end-result of Ag presentation by CD103⁺ DCs, be it tolerogenic (Coombes et al., 2007; Sun et al., 2007) or immunogenic (Laffont et al., 2010), depends on the context of Ag uptake (Laffont et al., 2010; Scott et al., 2011; Semmrich et al., 2012). Ultimately, manipulating the access of CD103⁺ DCs to bacterial Ag, as we crudely did here, could relieve inflammatory bowel disease or boost oral vaccination. Better understanding of the molecular mechanisms that allow these DCs to sample luminal Ags and means to specifically target these cells in vivo would be needed before this becomes a viable option.

EXPERIMENTAL PROCEDURES

Mice

Strains used were as follows: C57BL/6 Cd45.2, Cd11c-YFP, Cx3cr1^gfp/+, Actin-Cfp, MyD88^−/− × Ticam1^−/−, Ccr6^−/−, Cd103^−/−, OT-I, and Flt3^−/−. Bone marrow chimeras were produced by irradiation at 950 rad and transfer of 3 × 10⁶ bone marrow cells. Mice were maintained in specific pathogen-free conditions and handled according to protocols approved by the Weizmann Institute Animal Care Committee.

GM-CSF Treatment

Mice were injected subcutaneously (s.c.) with 4 × 10⁶ B16 tumor cells expressing GM-CSF (Mach et al., 2000) and analyzed 2 weeks later.

Isolation of Intestinal Cells

The ileum was resected and separated from PPs. Cells were released from the epithelium by rotation for 15 min at 37°C with 5 mM EDTA. LP cells were released after further 45 min incubation with collagenase-D and DNase.

Bacterial Strains

The following strains of Salmonella typhimurium were used: χ4550 expressing or not expressing OVA-GFP (Yrlid et al., 2001) and noninvasive SB161 (Hapfelmeier et al., 2005).

In Vitro Phagocytosis Assay

We isolated 2 × 10⁶ cells from the LP, cocultured them with GFP⁺ Salmonella for 1 hr at 37°C, treated them with 500 μg/ml gentamicin for 60 min, and washed them for flow cytometric analysis.

Analysis of T Cell Proliferation

CD103⁺ DCs from the epithelial fraction were sorted, incubated with 100 μg ovalbumin, and 20 μg/ml Poly(I:C) for 2 hr at 37°C, washed extensively, and incubated with CFSE-labeled OT-I cells in round bottom 96-well plates at a 1:5 ratio (10⁴ DCs with 5 × 10⁴ T cells). CFSE dilution was quantified 3 days later.

Immunohistology

For whole-mount preparations, the intestinal tissue was fixed with 10% paraformaldehyde (PFA) and immunostained according to standard procedure. For CD103 staining (Semmrich et al., 2012), mice were injected intraperitoneally with 10 μg of anti-CD103 PE. For IgA staining, the tissue was fixed with 2% PFA and cryosectioned.

In Vivo Uptake Assays

A ligate ileal loop was injected with 3–5 × 10⁹ GFP-expressing Salmonella 3 × 10⁹ beads or 100 μg of OVA-Alexa-594. Cells isolated from ileal segments were analyzed by flow cytometry 2–5 hr later. Some mice were intravenously (i.v.) injected with 400 μg/kg PTx 14–17 hr before Salmonella challenge (Huang et al., 2007; Fooksman et al., 2010).

Intravital 2-Photon Microscopy

Anesthetized mice were laparotomized. The ileum was externalized, immobilized, and cut open longitudinally using a cautery to be imaged using an Ultima 2-photon microscope (Prairie Technologies).