Activation of Plasmacytoid Dendritic Cells in Colon-Draining Lymph Nodes during Citrobacter rodentium Infection Involves Pathogen-Sensing and Inflammatory Pathways Distinct from Conventional Dendritic Cells

Conventional dendritic cells (cDCs) are sentinels of microbe- and danger-associated signals in their local environment and migrate from tissue to draining lymph nodes (LNs) via the lymphatic system. In the gut, the CD103⁺ DCs are the most predominant population of cDCs. They populate small intestinal lamina propria and migrate to mesenteric LNs (MLN) (1) to induce regulatory and antimicrobial immune responses to food-borne Ags and commensal microbiota (2, 3). The capacity of CD103⁺ migratory DCs to imprint responding CD4⁺ T cells with gut-homing receptors CCR9 and integrin α4β7 (4) and to synthetize retinoic acid have placed this DC subset in a central role in inducing tolerance to food-borne Ags and commensal bacteria (5, 6). In parallel with their ability to induce regulatory Foxp3⁺ regulatory T cells under steady-state conditions, their importance in promoting antimicrobial Th17 helper T cell responses under inflammatory conditions is well documented (7).

Plasmacytoid DCs (pDCs) differ from cDCs in many aspects (reviewed in Ref. 8). They are mainly found in blood and secondary lymphoid tissues (9, 10), but are relatively sparse in peripheral tissues and mostly home in secondary lymphoid tissues directly from circulating blood. Compared to other peripheral tissues, pDC are numerous in intestinal lamina propria (11) of the small intestine. The ability of pDCs to capture Ags and present Ags to T cells appear to be inferior to cDCs, but their ability to sense environmental cues and elaborate mediators of immune response nevertheless points toward an important role in T cell activation and differentiation (reviewed in Ref. 12). pDCs express their unique set of TLR receptors, and, in response to foreign nucleic acids, circulating pDCs produce large quantities of IFN-α. However, in the gut, pDC appear to promote Th17 cells (13) for antimicrobial defense and, during steady-state conditions, to induce regulatory T cells (14). Despite these important aspects, the roles of pDC in intestinal immunity are less well characterized than those of CD103⁺ DCs (15).

Most of the knowledge on DCs and their roles in intestinal immunity is related to the small intestine. Although colonic mucosa and lamina propria immune cells have been the focus of much research, less attention has been paid to the lymphatic drainage of the colon. Carter and Collins (16) used ink injections to draw a sketch on the lymph drainage from different parts of the intestine already in 1974 and proposed that lymph from colon drains to LNs distinct from the main MLN. More recently, Van den Broeck et al. (17) illustrated lymph drainage from the intestine using modern imaging techniques and confirmed the distinction between lymphatics of the small intestine and colon. Despite this, no data were available to evaluate if immune responses in the small intestine and colon involve anatomically distinct lymphoid structures and cell types. This problem was tackled quite recently by Houston et al. (18), who used photoconvertible reporter techniques and multicolor flow cytometry to document DC accumulation from the colon and small intestine into their respective draining LNs. They identified three novel colon-draining LNs distinct from the main MLN draining small intestine. The unique phenotypes of their CD103⁺ DC subsets and T cells and the inability of CD103⁺DCs in these LNs to prime T cells in response to small-intestinal and oral Ags indicated that these LNs and their DC segregate from the immune surveillance of the small intestine (18).

Although the role of colon was long thought to focus mainly on absorption of excess water from luminal contents, recent advances in understanding the importance of colon and its microbiota in host metabolism and maintenance of immune homeostasis has created novel interest in studying the role of colon immune system in health and disease. Under an independent initiative, we set an endeavor to identify the LNs to which lymph drains from the colon to evaluate the roles of cDCs and pDCs in these LNs during microbial infection of the large intestine. Our findings corroborate the existence of three distinct LNs draining the colon, but also identify a role for each one of these in draining the proximal, transverse, or descending section of the colon. We show that microbial challenge of the colon with Citrobacter rodentium infection leads to selective activation of especially pDC in these and identify unique pathogen-sensing and activation pathways in pDC with potential importance in mounting a protective Th1 response (19, 20) to this effacing-attaching pathogen.

C57BL/6NCrl mice used in this study were bred and maintained in Turku University’s central animal facility. All animals were kept under specific pathogen-free housing conditions and under filter lids. Food and water were available ad libitum, and there was a 12-h/12-h light/dark cycle. All experiments were approved by the national board of animal experimentation in Finland under license ESAVI/6082/04.10.07/2014.

Mice were anesthetized using 150 mg/kg ketamine (Ketalar 50 mg/ml; Pfizer) and 10 mg/kg rompun (Rompun vet 20 mg/ml; Bayer) mixture i.p. A small incision was made into the skin, and peritoneal membrane and colon were gently pulled out. FITC-labeled dextran (FITC-Dx; 70 kDa; 80 mg/ml; Sigma-Aldrich) or Alexa Fluor 488–labeled OVA (OVA-488, 1 mg/ml; Life Technologies) was injected into colon wall intraserosally. Separate injections were made into ascending, transverse, and descending parts of the colon. The incision was closed, and mice were kept anesthetized on a warm blanket for 40 min before sacrifice. When analyzing lymph drainage after FITC-Dx, a tissue block including cecum, colon, mesenteric LNs, pancreatic LNs, and pancreas was prepared. The block was glued onto a Petri dish and submerged in cold PBS. FITC signal was analyzed under stereomicroscope (SteREO Lumar V12; Carl Zeiss). To analyze Ag uptake by DCs, LNs were collected and DC populations labeled as described below. Alexa Fluor 488 signal analyzed as mean fluorescence intensity by FACS.

Mice were fed with 1 to 2 × 10⁹ CFU C. rodentium in Luria-Bertani medium or sterile Luria-Bertani (mock) by mouth and sacrificed 1, 3, 5, or 8 d postgavage. Individual LNs were collected and analyzed for DC activation. Shortly, LNs were collected in RPMI 1640 medium and digested with collagenase (10 U/ml, 10 min, 37°C). Collagenase was inactivated by adding 3 ml DMEM with 10% FCS and 1% nonessential amino acid. Digested tissue was pressed through a tight metal mesh and cell suspension filtered through a 77-μm nylon membrane. Unspecific binding was blocked by adding 5% rat serum. In flow cytometer analysis, live single cells were gated and further analyzed as described below. pDCs were identified as pDC Ag-1 (PDCA-1; PE-conjugated clone 927; BioLegend) and B220 (allophycocyanin-conjugated clone RA3-6B2; BioLegend) double-positive cells. Lamina propria–derived DCs were identified as CD11c (PE-conjugated clone N418; BioLegend) and CD103 (allophycocyanin-conjugated clone 2E7; BioLegend) double-positive cells. Both populations were further characterized for expression of CD11c (FITC-conjugated clone N418; BioLegend), CD11b (FITC-conjugated clone M1/70; BD Pharmingen), F4/80 (allophycocyanin-conjugated clone BM8; BioLegend), Siglec-H (PerCP/Cy5.5-conjugated clone 551; BioLegend), Gr1 (Alexa 488–conjugated clone RB6-8C5; eBioscience), and CD8a (FITC-conjugated, catalog number 22150083; ImmunoTools). Activation of these populations were evaluated by abundance of CD80 (FITC-conjugated clone 16-10A1; eBioscience) and CD86 (FITC-conjugated clone GL1; eBioscience). T-lymphocytes were analyzed for IFN-γ and IL-17A production. To analyze cytokine production, harvested cells were stimulated for 4 h in DMEM medium supplemented with 10% FCS, 25 mmol HEPES, 1% nonessential amino acid, and Cell Activation mixture (catalog number 423304; BioLegend). Lymphocytes were stained for CD4 (FITC-conjugated clone RM4-5; BioLegend) and CD8 (allophycocyanin-conjugated clone 53-6.7; BioLegend). Intracellular staining was performed using Transcription Factor buffer set (catalog number 562725; BD Biosciences) according to the manufacturer’s instructions and PE-labeled IFN-γ (clone XMG1.2; BioLegend) or IL-17A (clone TC11-18H10.1; BioLegend) Ab. Stained cells were analyzed with Accuri C6 flow cytometer and Accuri C6 software (BD Biosciences).

To evaluate LN enlargement in response to C. rodentium infection, the size of the largest colon-draining MLN (coMLN) and small intestine–draining MLN (siMLN) was measured at day 3 using an electronic digital caliber. For coMLN, both length and width was measured, and for siMLN, only the thickness of the proximal end, closest to the coMLN, was measured. The volume of coMLN was calculated using the formula for ellipsoid volume (4/3πabc, assuming the ellipsoid shape to be symmetrical, e.g., b = c).

C. rodentium (strain ICC168, from Prof. Fiona Powrie, University of Oxford, Oxford, U.K.) was grown overnight in Luria broth supplemented with nalidixic acid (1 mg/l). Overnight growth was concentrated 10 times by centrifugation. Concentrated bacterial solution was fed to mice by mouth in 200 μl. Bacterial counts were measured by plate assays: serial dilutions were made from feeding solution and plated on Luria agar plates with 50 μg/ml nalidixic acid overnight. Colonies were counted to quantify the fed bacteria.

To analyze infection-derived changes in cytokines and transcription factors, the colon was collected into RNALater (Qiagen, Germantown, MD). Total RNA was extracted with bead-based PowerLyzer RNA isolation kit (MoBio, Carlsbad, CA). cDNA was synthetized with Maxima reverse transcriptase and random hexamer primers (Thermo Fisher, Waltham, MA). For all primer and probe details, see Supplemental Table I. Relative cytokine expression was determined by real-time quantitative PCR (Lightcycler 480; Roche, Basel, Switzerland) using LightCycler 480 Probes Master or LightCycler SYBR Green I Master mastermixes (Roche). All cytokine signals were normalized to β-actin expression and shown as arbitrary relative expression values (1/2^{[threshold cycle(target) − threshold cycle(reference)]}).

Mice were treated with C. rodentium or sterile medium as described above and sacrificed at day 3 postinfection. Three independent experiments were performed, and for each data point, coMLN and siMLN were collected and pooled from four animals. Cell suspensions were made and pDCs stained as described above. To be able to sort both populations from the same tube, CD103⁺ DCs were stained with FITC-conjugated CD11c and PerCP/Cy5.5-conjugated CD103 (clones N418 and 2E7, respectively; BioLegend). Both populations were sorted simultaneously into the lysis buffer using FACSAria II Cell Sorter (BD Biosciences). The cell yields for CD103⁺ cells were 14,584–117,457 cells and for pDC cells 1,158–7,020 cells. GenElute LPA (catalog number 5-6575; Sigma-Aldrich) RNA carrier was added (1 μl) into the lysis buffer prior to RNA extraction from pDC and CD103⁺ DC cells. Total RNA was extracted using RNAeasy Plus Micro kit (catalog number 74034; Qiagen) according to the manufacturer’s instructions. The quantities of the RNA were measured with Nanodrop, and quality was assured with Agilent 2100 Bioanalyzer and Pico RNA Assay (Agilent Technologies) when a sufficient amount of material was available.

Libraries for RNA sequencing were prepared from 10 ng totalRNA with Smart-Seq v4 Ultra Low input RNA kit (catalog numbers 634896 and 634895; Clontech) according to the kit manual. The cDNAs were quantified with ThermoFisher Qubit dsHS assay, and quality was assured with Agilent 2100 Bioanalyzer DNA High Sensitivity assay (Agilent Technologies). From each sample, 150 ng cDNA was amplified with Illumina Nextera XT DNA library kit (catalog number FC-131-1024; Illumina) according to the kit manual (part 15031942 Rev D). Libraries were quantified with ThermoFisher Qubit dsDNA HS assay (Thermo Fisher), and quality was assured with Agilent Bioanalyzer 2100 DNA High Sensitivity Assay (Agilent Technologies). The clonal cluster amplification was carried out with Illumina cBOT System (Illumina). The amplified libraries were run with Illumina HiSeq2500 Next-Generation Sequencing platform (Illumina) in a single lane of a paired-end rapid run with read length of 100 bp. The run generated a total of 1.65 × 10⁸ raw reads passing filter with a Q30% score of 92.5%.

The quality control of raw sequencing reads was performed with FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/), and adapters and low quality bases were trimmed by Trimmomatic (21). The trimmed reads were then aligned to the mouse reference genome NCBIM37/mm9 (Ensembl release 67) using Tophat2 (22). Htseq-count (23) summarized read counts for each gene. The R/Bioconductor package edgeR (24) was used to identify differentially expressed genes. Genes with a false discovery rate (FDR) <0.05 and an absolute fold change >1.5 were considered significantly differentially expressed. Statistical analyses for quantitative PCR and FACS data were done using GraphPad Prism 5 software (GraphPad, La Jolla, CA). Statistical significance was determined using two-tailed Student t test or one-way ANOVA with Dunnett post hoc test. The p values <0.05 were considered statistically significant.

FITC-Dx was injected subserosally into three different areas of the colon corresponding to the proximal, middle, or distal part of the colon, and the passage of FITC-Dx into afferent lymphatics was visualized using a low-magnification UV stereomicroscope. We identified three individual coMLNs (Fig. 1A, 1B). The ascending colon and right colic flexure drained to a separate LN near cecum adjacent to the MLN. The transverse colon drained to a LN located close to the other end of the siMLNs and between the string of MLNs and the colon, and the descending colon drained to a small LN in the mesenterium in close proximity to the pancreas (Fig. 1). All three LNs have their efferent lymph veins flowing to the same large lymph vein in the mesenterium. As these three LNs appeared to collect afferent lymph specifically from the colon and be distinct from the main part of the MLN draining the small intestine, these three LNs will be collectively referred to as coMLN in distinction from the siMLN.

Macroscopic examination of coMLN and siMLN showed that the size of coMLN, but not siMLN, was significantly increased already 3 d after oral inoculation of the pathogen (Fig. 2). In line with earlier reports (25–27), cytokines, transcription factors related to activation of macrophages, innate lymphoid cells, T cells, and stress indicators of epithelial cells became upregulated in colon tissue only later, ∼8 d postinfection (Supplemental Fig. 1). This implicates the coMLNs in immune recognition of the pathogen already before prominent inflammation in the gut.

We studied two different DC populations, cDCs and pDCs in the LNs, which we identified as coMLN. These DC populations were defined by their expression of either PDCA-1 and B220 (pDCs) or CD11c and CD103 (cDCs) (Fig. 3). Additional markers expressed by coMLN-derived pDCs were consistent with known pDC markers (Fig. 3E–G). cDC in coMLN were defined as CD11c^high/CD103⁺ DC, containing both migratory and resident DC populations (28, 29). Interestingly, although CD11b⁺ expression was detected on a subpopulation of CD103⁺ DCs in siMLN (24.01 ± 3.61%), CD11b⁺CD103⁺ DCs were rare in coMLN (3.65 ± 3.98%, p = 0.0009). The majority of CD103⁺ DCs in coMLN expressed CD8a (70.27 ± 12.91%) (Fig. 3H–J). CD11b⁻/CD8a⁺ cells are the dominant subpopulation of CD103⁺ DCs in the coMLN (44.10 ± 6.97%, p < 0.0001) when compared with CD11b⁺/CD8a⁺ and CD11b⁺/CD8a⁻ fractions.

To confirm lymph drainage from different parts of the colon to the abovementioned LNs, we determined Ag uptake by DCs in coMLN after injection of OVA–Alexa 488 into the colon wall. Both the CD103⁺ DC and the PDCA-1⁺/B220⁺ DC were evaluated. Ag uptake was determined 40 min after injection and therefore mostly represents uptake of soluble material arriving in the LN with the lymph (30). Both DC populations in coMLN displayed Alexa 488 fluorescence (Fig. 4), indicating Ag uptake by DCs in the same LNs as the ones to which afferent lymph vessels from the colon ended (as illustrated by microscopy of the mesenterium, see Fig. 1). Interestingly, pDCs were significantly more efficient in taking up OVA–Alexa 488 than the CD103⁺ DCs, indicating that they sample Ags directly and more efficiently than CD103⁺ DCs, which are present in coMLN at time of Ag arrival.

To study the kinetics of DC activation in coMLN, we determined CD80 and CD86 expression on DCs in coMLN and siMLN 1, 3, 5, and 8 d after C. rodentium inoculation. Indeed, both pDC and CD103⁺ DC upregulated costimulatory ligand expression in response to C. rodentium infection. CD103⁺ DCs showed upregulation of CD80 (but not CD86) already at day 1 postinfection, and this occurred concomitantly in coMLN and siMLN (Fig. 5). pDCs upregulated both CD80 and CD86, and importantly, this occurred selectively in coMLN by day 3 postinfection. Only later at day 5, CD80 and CD86 upregulation on pDCs was detected also in siMLN (Fig. 5).

In order to study if DC in coMLN differ from their counterparts in siMLN, total RNA from sorted pDCs and CD103⁺ DCs was sequenced and analyzed for gene expression both from coMLN and siMLN. RALDH (Aldh1a2) mRNA expression was lower in coMLN-derived CD103⁺ DCs compared with CD103⁺ DCs in siMLN, consistent with the findings of Houston et al. (18). This finding is in line with the notion that CD103⁺ DCs in coMLN and siMLN are not identical and that T cells are likely not imprinted with CCR9 expression in coMLN. Differences in the expression of cytokine and chemokine genes were also detected between CD103⁺ DC in coMLN and siMLN, and the same was true for pDC. In total, 416 (p value for individual gene <0.05) and 698 (FDR < 0.05) genes showed differential expression in CD103⁺ DC and pDC, respectively, when comparing either subset in coMLN and siMLN (Supplemental Table II). Comparison of coMLN-derived pDC with previously documented phenotypes of pDC revealed many markers shared between these, whereas significant differences were observed with some markers (Supplemental Table III). CoMLN-derived pDC expressed the common pDC markers PDCA-1 (BST2), B220, Siglec-H, and Ly6C, as well as pattern recognition receptor TLR7. The most striking difference was that pDCs from coMLNs and siMLN did not express IFN-α nor did they express TLR9. Expression of many other markers earlier documented in pDC were expressed also in coMLN-derived pDC, including chemokine receptors and IFN regulatory factors (IRFs). Among these, the coMLN-derived pDCs expressed IRF1 and -3 not reported to be expressed in pDCs from other tissues.

In order to confirm that pDC are activated only in coMLN during early C. rodentium infection and to study the effects of infection on both the pDC and the cDC subsets in more detail, RNA was sequenced from both DC subsets after C. rodentium infection, and gene expression was compared between infected and steady-state conditions. A total of 13,945 different transcripts were identified. Not surprisingly, pDCs and CD103⁺ cDCs expressed clearly different transcriptomes even under steady-state conditions (Fig. 6A, 6B, Supplemental Fig. 2). To examine the effect of C. rodentium infection on gene expression, we filtered the dataset by selecting only those transcripts that had minimum 1.5-fold change over mock infection and FDR <0.05. This resulted in 750 and 220 differentially expressed genes (DEGs) in pDCs in coMLN and siMLN, respectively (Fig. 6C, for all of the DEGs, see Supplemental Table IV). Among these were several immunologically relevant genes involved in IL-1, IL-6, and TLR signaling (Table I). Surprisingly, for cDCs, only two DEGs (Atp1b2 8.1-fold upregulation and Trav15-2-dv6-2 5.7-fold downregulation) fulfilled the same cutoff criteria in coMLN, and no genes were identified in siMLN fulfilling the same criteria for DEGs.

C. rodentium infection affected gene expression in pDCs clearly more profoundly in coMLN than in siMLN (Fig. 6C). Although pDCs in coMLN reacted to C. rodentium by differentially regulating 3.4 times as many genes than pDCs in siMLN (i.e., 750 versus 220 genes), only 22 upregulated genes and only 6 downregulated genes were shared among the coMLN and siMLN pDCs. In order to find any potential DEGs in CD103⁺ cDCs, we then employed less stringent criteria for these. Employing less stringent criteria for DEGs, CD103⁺ DCs up- or downregulated 147 and 54 genes in coMLN and siMLN, respectively. These included few genes directly relevant for antimicrobial immunity, namely ILs Il1b (coMLN only), Il17f, and Il22, which showed 2.7–6.9-fold upregulation after C. rodentium infection (Supplemental Fig. 3, Supplemental Table V). This suggests that C. rodentium infection may reflect some unique effects also on cDCs in coMLN, although the relatively loose criteria employed increases the risk of inadvertent differences. Comparison between pDC and CD103⁺ DC transcripts showed clear differences between the C-type lectin and MHC gene expression (Supplemental Table VI). Six different Clec genes were identified to be differentially expressed in the two DC populations. Clec11a, -4a, 4b, and -12a were more abundant in pDCs, whereas Clec4d and -4e were more abundant in CD103⁺ DCs. Furthermore, CD8α and CD8β expression showed decreasing trend (1.7- and 3-fold, respectively) after C. rodentium inoculation, indicating loss of tolerogenic pDCs in favor of immunogenic pDCs (31). However, a comparison of regression curves of gene expression between infected and mock-infected mice showed that on global gene expression level, C. rodentium infection affected gene expression in pDCs only (Fig. 6E, 6F).

In order to evaluate where the T cell response to colon-derived pathogen is initiated, we determined IFN-γ and IL-17A production in individual T cells after C. rodentium infection in coMLN and siMLN (Fig. 7). IFN-γ–producing CD4⁺ cells were significantly and specifically expanded in the coMLN, when compared with either mock-infected animals or the siMLN compartment (Fig. 7D). Only slight IFN-γ elevation, which was comparable to non–gut-associated brachial LN, was observed in the siMLN when compared with mock infected animals (Fig. 7D). IFN-γ–producing CD8⁺ cells were also expanded in response to C. rodentium infection, and this occurred significantly in both coMLN and siMLN, but not in brachial LN (Fig. 7E). In mock-infected mice, the fraction of IFN-γ–producing cells was observed to be at the same level among all LNs, remaining <2% in CD4⁺ cells and ranging between 5 and 13% in CD8⁺ cells. IL-17–producing CD4⁺ cells were present at considerable numbers but not significantly expanded in any LN postinfection (Fig. 7F).

Recently, intestinal monocytes and macrophages were found to be required for Th1 polarization of Th cells in response to C. rodentium (19), but the site where naive T cells are primed and initially polarized toward Th1 polarization after C. rodentium infection has hitherto not been addressed. In this study, we describe three small LNs in the mesenterium, which drain lymph selectively from the large intestine and identify them as the site in which DC and Th1 helper T cells are activated in response to C. rodentium infection. Although cDCs showed phenotypic signs of activation earlier than pDCs, transcriptional profiling identified many more genes becoming upregulated in pDCs as compared with gut-associated (CD103⁺) cDCs. Genes upregulated in pDCs included genes involved in pathogen-sensing and in adaptive immune responses, suggesting an important role not only for CD103⁺ cDCs (19), but also for pDCs in priming protective immunity during C. rodentium infection.

It was suggested already in 1974 by Carter and Collins (16), and documented more convincingly by Van den Broeck et al. in 2006 (17), that lymphatics from different parts of the intestine drain into different parts of the cluster of MLN in the mouse. However, only very recently, Houston et al. (18) documented for the first time, to our knowledge, the existence of three LNs draining lymph specifically from the colon and showed evidence of CD103⁺ cDC migration from colon into these LNs. Whereas the standard method to detect lymphatic drainage to a given LN has been to inject ink such as Evan blue into the tissue of interest, we used subserosal injections of FITC-Dx into the colon wall and dissected thereafter the whole gut and mesenterium en bloc for fluorescence microscopy. This allowed the simultaneous imaging of the structure of small afferent lymphatic vessels initiating from the gut wall and identification of the collecting LN. Three separate LNs were identified draining different parts of the colon. The three separate LNs, termed in this study as coMLNs, included one at the end of siMLNs (the main MLN) closer to cecum and two at the opposite side of the central siMLNs.

As the first measure to evaluate the role of coMLNs as sentinels of infective colitis, we followed CD103⁺ cDC and pDC activation in local LNs after oral inoculation of C. rodentium. Together with the visual finding of enlargement of coMLNs, we verified that pDCs become activated specifically in coMLNs after C. rodentium inoculation. Interestingly, CD103⁺ cDC in LNs appeared to become activated earlier than pDC, but this occurred both in coMLN and siMLN with no significant changes to the transcriptome. Also, while characterizing the CD103⁺ cDC population, we observed the lack of CD11b⁺ subpopulation in the coMLN. This finding is in line with recently published results by Veenbergen et al. (32), who studied the caudal and ileac LNs (ILNs), which drain the very distal part of the colon and rectum, and reported that ILNs lack the CD11b⁺ migratory CD103⁺ DCs. Although the ILNs are separate from the coMLNs described in this study and were not analyzed in this paper, current results indicate the same phenomenon also in the coMLNs. To further evaluate the role of CD103⁺ cDC and pDC in coMLN, we tested their uptake of soluble Ag after subserosal injections of Alexa 488–labeled OVA into the colon wall. Unexpectedly, Alexa 488 signal was much higher in pDCs than in migratory CD103⁺ cDCs. pDCs have been reported to capture OVA and hen egg lysozyme both in vitro and in vivo also earlier (33–35). Consistent with earlier reports (36, 37), coMLN-derived pDCs expressed lower levels of costimulatory ligands and MHC class II molecules than CD103⁺ cDCs. Nevertheless, pDCs are considered capable of processing and presenting Ags to initiate T cell activation (12). The levels of nonclassical MHC molecules differed between pDCs and CD103⁺ cDCs in coMLN. This suggests that these specialized pDCs in coMLN may have nonredundant roles in presentation of various microbial Ags to T cells, at least if they appear in the LN in a soluble form. This could occur particularly under conditions in which epithelial permeability and lymph flow are increased such as during colon inflammation. In summary, our findings of the activation kinetics of CD103⁺ cDCs and pDCs and the capture of soluble Ag after short-term lymph drainage are consistent with the idea, that migratory CD103⁺ cDCs are the first DCs to sample the bypassing and attaching C. rodentium in both small intestine and the colon. Only once the pathogen attaches firmly to colon wall and infects the mucosal layers, it gains access to coMLNs and can be taken up by resident pDCs therein.

Although rapid production of type I IFNs in response to viral nucleic acids is often regarded as the most important function of pDCs, pDCs in the small intestinal Peyer’s patches appear not to secrete IFN-α in response to TLR9 stimulation (38). Ab crosslinking studies have indicated that crosslinking pDC receptors such as BDCA-2, Siglec-H, and DC inhibitory receptor 2 strongly inhibit type I IFN production in human pDC (39–41). Our RNAseq results indicated a high-level expression of Siglec-H and CLEC4A, the mouse homolog of DC inhibitory receptor 2, in coMLN-derived pDCs, suggesting that ligand binding to these could occur during C. rodentium infection and inhibit type I IFN production after pDC activation. Of note, coMLN-derived pDCs also lacked expression of TLR9, which mediates CpG-induced type I IFN production. Also, the use of transcription factors differed in coMLN pDCs, as a wider variety of IRFs were expressed in coMLN-derived pDCs in comparison with pDCs studied earlier. Together with CD103⁺ DC subpopulation composition in the coMLN, these LNs align more closely with the Peyer’s patches than with the siMLNs (28, 38). These data suggest specialization of coMLN-derived pDCs to detect microbial signals different from other pDCs and their adaptation to modulate gut-associated immune responses.

Along with the detailed analyses of Houston et al. (18), our findings conclusively show that immune surveillance of the small and large intestine is operated by distinct LNs and cells dedicated to these structures. The finding that pDCs upregulated expression of proinflammatory cytokine genes as well as genes involved in pathogen sensing underscores the importance of pDCs in coMLNs as active sentinels of inflammation in the colon. C-type lectins recognize a variety of microbes from fungi and viruses to bacteria and may deliver either inhibitory or activating signals to immune cells expressing them (42). In the case of C. rodentium infection, we hypothesize that lectin receptors on pDCs such as CLEC4a/e and Siglec-H may be involved either directly or indirectly in the recognition and processing of microbes by pDCs and, together with other pathogen-sensing motifs, participate in activating or regulating the immune response against C. rodentium in the coMLNs described in this study.

We thank the Cell Imaging Core for guidance with cell sorting and imaging systems.

The authors have no financial conflicts of interest.