|Home | About | Journals | Submit | Contact Us | Français|
Ectopic or tertiary lymphoid tissues (TLTs) are often induced at sites of chronic inflammation. They typically contain various hematopoietic cell types, high endothelial venules, and follicular dendritic cells; and are organized in lymph node–like structures. Although fibroblastic stromal cells may play a role in TLT induction and persistence, they have remained poorly defined. Herein, we report that TLTs arising during inflammation in mice and humans in a variety of tissues (eg, pancreas, kidney, liver, and salivary gland) contain stromal cell networks consisting of podoplanin+ T-zone fibroblastic reticular cells (TRCs), distinct from follicular dendritic cells. Similar to lymph nodes, TRCs were present throughout T-cell–rich areas and had dendritic cells associated with them. They expressed lymphotoxin (LT) β receptor (LTβR), produced CCL21, and formed a functional conduit system. In rat insulin promoter–CXCL13–transgenic pancreas, the maintenance of TRC networks and conduits was partially dependent on LTβR and on lymphoid tissue inducer cells expressing LTβR ligands. In conclusion, TRCs and conduits are hallmarks of secondary lymphoid organs and of well-developed TLTs, in both mice and humans, and are likely to act as important scaffold and organizer cells of the T-cell–rich zone.
Acute inflammation has evolved to facilitate effective elimination of pathogens and tumors. It is normally transient and turned off when the causative stimulus has been eliminated. Occasionally, the inflammation is sustained and becomes persistent.1 This complex pathophysiological process is characterized by the infiltration of target organs, with lymphocytes organizing themselves into distinct T- and B-cell–rich zones. Because these infiltrates show considerable morphological, cellular, and molecular similarities to secondary lymphoid organs (SLOs), especially lymph nodes (LNs), they are referred to as tertiary lymphoid tissues (TLTs). There is increasing evidence that these TLTs, observed within a wide range of inflamed tissues, are immunologically active and have the potential to cause severe tissue damage.2–4
Specialized stromal cell types present within inflamed tissues are thought to be critical for the switch from acute to chronic inflammation and for disease persistence.1 In particular, high endothelial venules (HEVs), follicular dendritic cells (FDCs), and fibroblasts are hallmarks of many TLT types.2,3 However, the presence and role of T-zone fibroblastic reticular cells (TRCs or T-FRC) in the formation and persistence of TLTs has not yet been carefully examined. In murine SLOs, FDCs and TRCs form a dense three-dimensional (3D) network throughout the B and T zones, respectively.5–7 The TRCs serve as a scaffold for lymphocyte migration within the T zone.8 They produce the CCR7 ligands, CCL19 and CCL21, required for T-cell attraction and motility and dendritic cell attraction.9–11 They also regulate the survival of naïve T cells by producing IL-7.10 Finally, they produce the extracellular matrix (ECM) scaffold that forms a network of microvessels, called conduits.7,12,13 In LNs, these conduits drain fluid and small molecules from the subcapsular sinus to HEVs.13,14 Thus far, no single marker unambiguously identifies TRCs in murine SLOs. The TRCs can be identified by their T-zone localization, reticular morphological features, and surface expression of podoplanin (also referred to as gp38); and the absence of endothelial (CD31 or LYVE-1), FDC (CD21/35), and hematopoietic (CD45) markers.6,10,12 In humans, podoplanin has been widely used as a marker for lymphatic endothelial cells15; recently, it was used for the identification of FDC16,17 and cancer-associated fibroblasts.18,19
Reticular networks of podoplanin+ and BP-3+ stromal cells have also been observed in murine TLTs, such as pancreatic infiltrates of rat insulin promoter (RIP)–CCL19, RIP-CCL21, and RIP-CXCL13 animals,20,21 and recently at sites of acute inflammation and tumors.22 By using the antibody ER-TR7, similar networks have been visualized in other murine inflammation models.20,23–25 However, recent evidence10,12,13 suggests that ER-TR7 stains a conduit component, rather than TRCs. In human tissues, such as LNs and arthritic joints, the presence of reticular cells in the T zone expressing α smooth muscle actin (αSMA), CCL19, and CCL21 has recently been reported.26 However, a detailed characterization of stromal cells of the T zone has been lacking for murine TLTs and human SLOs and TLTs.
The formation of TLTs in adult mice is thought to be governed by the same factors that regulate LN and Peyer's patch (PP) development during embryonic and postnatal life.2,3 Members of the tumor necrosis factor–lymphotoxin (LT) family and lymphoid tissue chemokines (ie, CXCL13, CCL19, and CCL21) are central to this process. Lymphoid tissue inducer (LTi) cells express cell surface LTαβ and trigger the LTβ receptor (LTβR) present on fibroblastic stromal cells, thereby starting extensive hematopoietic-stromal cell cross talk.27,28 This induces the expression of CCL19, CCL21, and CXCL13 and adhesion molecules by these stromal cells. The chemokines then attract more LTi cells expressing the chemokine receptors CCR7 and CXCR5 that bind the ligands CCL19/CCL21 and CXCL13, respectively. Finally, the development of HEVs brings in many more lymphocytes, leading to large T- and B-cell compartments. Consistent with this model, the expression of LTα, LTαβ, tumor necrosis factor-α, CXCL13, CCL19, or CCL21 under the RIP induces the formation of extensive TLT structures next to pancreatic islets.2,3,20,21,29–32 The observation that TLT structures require LTαβ signaling for their induction and maintenance suggested that they also depend on cross talk between LTαβ+ hematopoietic cells and LTβR+ stromal cells.33 The LTαβ+ LTi cells have been shown to be present in adult spleen.34–36 The observation of LTi-like cells in pancreatic islets of neonatal RIP-CXCL13 mice37 suggested that they may play a role in TLT development. The role of LTi cells in this process is controversial,25,38 as is the identity of the LTβR-expressing stromal cells.4,28
Herein, we performed a detailed characterization of stromal cell subsets in murine and human TLTs and in human SLOs. We provide several lines of evidence suggesting that TRC-like cells exist in human SLOs and in infiltrates of two chronic immune-mediated inflammatory diseases in humans [ie, primary biliary cirrhosis (PBC)39 and Sjögren's syndrome].40 Similar TRC networks were associated with the T-cell–rich infiltrate of several murine models of chronic inflammation.21,30 Their phenotype closely resembled their counterpart in SLOs. We further dissected the cells and signals required for RIP-CXCL13–driven formation of ectopic TRCs and provide evidence that their maintenance is dependent on LTαβ-expressing LTi cells.
The following mouse strains were used (aged 3 to 6 months): C57BL/6 (B6; Janvier, Le Chenest-Saint-Isle, France) RIP-LTα29 (provided by Mathias Heikenwälder, University Hospital Zurich, Zurich, Switzerland; originally from Nancy Ruddle, Yale University, New Haven, CT), nonobese diabetic (NOD) (Jackson, Bar Harbor, ME), RIP-CXCL13,21 retinoid-related orphan receptor (ROR)γ−/−,41 (provided by Dan Littman, New York University, New York), and T cell receptor (TCR) βδ−/− (Jackson). Some adult (aged 4 months) RIP-CXCL13–transgenic animals were injected i.p. twice per week for 22 days with 100 μg of soluble murine LTβR or control human LFA3 (CD58) fused to the Fc portion of human IgG1, as previously described42 (provided by Jeff Browning, Biogen Idec, Boston, MA). All mice were maintained under pathogen-free conditions. All mouse experiments were authorized by the Swiss Federal Veterinary Office.
Tonsils (n = 5) were removed from patients undergoing routine tonsillectomy, and mesenteric LNs (n = 1) and livers were removed from patients during liver transplantation (normal liver, n = 1; PBC liver, n = 4; two contained well-defined TLT structures at several levels). Sjögren's syndrome salivary gland tissue (n = 10; two contained well-defined TLT structures at several levels) was obtained for biopsy. All human tissue samples were frozen in liquid nitrogen. Ethical approval for the use of tissue samples taken during this study was obtained from the Birmingham and local Research Ethics Committee (REC 2002/088, LREC 5735, and 06/Q2706/66).
Acetone-fixed 5-μm sections of snap-frozen tissues were prepared and stained, as previously described,43 with the antibodies listed in Table 1. Slides were mounted in 2.4% w/v 1,4-diazabicyclo(2,2,2)octane (Aldrich, Gillingham, England) in 10% v/v glycerol (Fisons Scientific, Loughborough, UK) in PBS (pH, 8.6) and analyzed by confocal microscopy (LSM 510; Carl Zeiss Ltd, Welwyn Garden City, England).
To visualize functional conduits, Texas Red–labeled 10-kDa dextran (Invitrogen AG, Basel, Switzerland) was injected i.v. and mice were sacrificed 5 minutes later. The pancreas was removed, fixed in four percent paraformaldehyde at 4°C, and saturated in 30% sucrose for 3 hours at 4°C before embedding in Optimal Cutting Compound (OCT, Tissue-Tek; Sakura), followed by freezing in an ethanol dry ice bath. For all other experiments, dissected tissues were embedded in OCT without prior fixation. Immunofluorescence stainings were performed on cryostat sections (8 μm) with the antibodies listed in Table 1, as previously described.10 For podoplanin and CCL21, primary antibodies were revealed using biotinylated secondary antibodies and streptavidin–horseradish peroxidase, followed by tyramide signal amplification (kit 22; Invitrogen), according to the manufacturer's instructions; however, a borate buffer (0.1 mol/L in PBS, pH 8.5) was used for tyramide dilution. Images were acquired on a microscope (Axioplan) with AxioCam MRm (Zeiss) or on a different microscope (DM IRE2) with laser-scanning confocal head TCS SP2 AOBS (acoustic optical beam splitter; Leica, Nunningen, Switzerland).
Isolated peripheral LNs and pancreas were fixed overnight at 4°C in freshly prepared one percent paraformaldehyde (in PBS), washed, and embedded in four percent low-gelling agarose (Sigma) in PBS; 100- to 200-μm sections were cut using a vibratome (Microm HM 650V). Sections were blocked with one percent bovine serum albumin for 1 hour and stained for at least 3 hours or overnight with the antibodies listed in Table 1. Subsequently, sections were washed extensively in PBS and embedded using Mowiol 4-88 (Merck, Darmstadt, Germany). Images were taken with an upright microscope (Axio Imager with ApoTome; Zeiss). The 3D image reconstructions were made with Axio Vision software version 4.8.1 (Zeiss).
Peripheral LNs, spleen, or pancreas was dissected from euthanized mice. The capsule of the LNs was opened with 26-gauge syringe needles; the spleen and pancreas were cut into small pieces. The LN, spleen, and pancreas pieces were then digested with gentle stirring for 20 minutes at 37°C in 2 mL of RPMI 1640 medium containing 1 mg/mL collagenase IV (Worthington, Lakewood, NJ), 40 μg/mL DNase I (Roche, Basel, Switzerland), and two percent (v/v) FBS. Collagenase D, 1 mg/mL (Roche), was added; and digestion continued for another 20 minutes. Every 10 minutes, the suspension was gently pipetted to break up remaining aggregates until no visible fragments remained. During the last pipetting, EDTA was added to a final concentration of 5 mmol/L to further reduce cell aggregates. Cells were then passed through a 40-μm mesh, washed twice, and resuspended in RPMI 1640 medium containing two percent (v/v) FBS and 5 μg/mL DNase I at a maximum concentration of 2 × 107 cells/mL. The cell suspension was then underlayed with 2.5 mL of mixed Lympholyte M (Cedarlane, Burlington, VT) at 25°C and centrifuged at 750 × g for 20 minutes. The cells at the gradient interphase were collected and used for flow cytometry.
Stainings were performed with 1 × 105 to 1.5 × 106 cells in 96-well V-bottom plates (Falcon; Milian AG, Geneva, Switzerland). Cells were blocked with one percent normal mouse serum and stained with the antibodies listed in Table 1 in 25 μL of PBS containing two percent FBS, 2 mmol/L EDTA, and 0.1% NaN3; the cells were washed twice with 200 μL of this buffer after each step. Unconjugated primary antibodies were detected with the secondary reagents listed in Table 1. Data were acquired on a flow cytometer (FACSCanto; BD Biosciences, Basel, Switzerland) and analyzed with Flowjo software version 9.1 (TreeStar, Ashland, OR). To detect surface expression of LTβR, cells were pretreated with FcR blocking antibody (2.4.G2; BD Biosciences) and 0.5% normal mouse and rat serum. The LTβR-Fc (provided by J. Browning) and the FN14-Fc (provided by P. Schneider) (both human IgG1) were added and detected using a biotinylated goat anti-human IgG (Jackson ImmunoResearch, Newmarket, UK) pretreated for 30 minutes with four percent normal mouse and rat serum. Finally, streptavidin-phycoerythrin (eBioscience, San Diego, CA) was added, along with other surface markers.
Mice were anesthetized and perfused with PBS, followed by four percent paraformaldehyde in PBS. The pancreas was then removed and further fixed in four percent paraformaldehyde overnight and saturated in 30% sucrose for 3 hours at 4°C before embedding in OCT, followed by freezing in an ethanol–dry ice bath. Frozen sections (8 μm) were then treated as previously described9 and hybridized overnight at 60°C with a CCL21-antisense digoxigenin-labeled riboprobe in hybridization solution. After washing at high stringency, sections were incubated with sheep anti-digoxigenin (Roche), followed by alkaline phosphatase–coupled donkey anti-sheep antibody (Jackson ImmunoResearch Laboratories) and developed with Nitroblue-tetrazoliumchloride (NBT, Bio-Rad) and 5-Bromo-4-chloro-3-indolyl phospate (BCIP, Sigma).
Statistical significance was determined using an unpaired two-tailed Student's t-test. A Levene's test for homogeneity of variances was used to check for equal variance. An unpaired two-tailed Student's t-test for unequal variance was used. P < 0.05 indicated significance.
To determine whether human SLOs and TLTs develop a TRC network similar to murine SLOs,10,12 we analyzed histological sections for the presence of podoplanin+ reticular cells. Three different antibodies to podoplanin comparably labeled the reticular networks of the T zone in both human LNs and tonsils (Figure 1, see Supplemental Figure S1 at http://ajp.amjpathol.org), while labeling most strongly FDCs and lymphatic vessels. The podoplanin+ cells of the T zone could be distinguished from FDCs through their localization among T cells and the absence of CD21 (Figure 1) and from lymphatic vessels by their reticular morphological features and absence of the endothelial marker CD31 and the lymphatic marker Lyve-1 (see Supplemental Figure S2 at http://ajp.amjpathol.org). Ectopic lymphoid infiltrates occurring in chronically inflamed nonlymphoid tissues, such as PBC liver or Sjögren's syndrome salivary gland, showed similar reticular networks of podoplanin+ CD31−Lyve1− cells in some, but not all, CD3+ T-cell areas (Figure 1, see Supplemental Figures S2 and S3 at http://ajp.amjpathol.org). Similar to LNs and tonsils, podoplanin- and CD21-coexpressing FDCs were observed as dense clusters inside B-cell rich zones. However, in noninflamed liver, podoplanin-expressing cells were virtually absent. These results suggest that TRC-like cells are present in human SLOs and can develop at sites of chronic inflammation.
To test whether these TRC-like cells observed in human tissues share other properties with TRCs characterized in murine LNs and spleen, we further defined these cells histologically. In T zones of LNs and tonsils, podoplanin+ cells formed a dense reticular network that associated with ECM molecules (ie, collagen-1, laminin, and fibronectin) (Figure 2, data not shown) in structures that resemble, in their composition and morphological features, the conduits found in murine SLOs.13 Podoplanin+ cells colocalized with CCL21 protein and were in close contact with CD11c-positive dendritic cells (Figure 2). Similarly, at sites of chronic inflammation, such as PBC liver and Sjögren's syndrome salivary gland, but not in healthy liver, ECM fibers were observed in association with podoplanin+ cells in T cell–rich infiltrates (Figure 2). To a variable extent, TRC-like cells displayed some detectable CCL21 protein and were associated with DCs. Together, these data indicate a high similarity of the stromal compartments of the T zone between murine SLOs and human lymphoid tissues.
To study the development and properties of TLT-associated TRCs in greater detail, we tested several mouse models displaying lymphocytic tissue infiltration for the presence of these stromal cells: NOD mice that are a model of autoimmune diabetes mellitus and develop mostly smaller pancreatic lymphoid infiltrates44 and RIP-LTα transgenic and RIP-CXCL13 mice that form large infiltrates of mostly naïve T and B cells in kidney and pancreas, respectively, without causing overt disease.21,29,30 Similar to murine LNs, reticular podoplanin+ cell networks were present in the T- but not the B-cell–rich areas in TLTs of all three mouse models. These fibroblastic cells showed a surface phenotype distinct from podoplanin− CD31high blood vessels and podoplanin+/− lyve1+ lymphatic vessels associated with the infiltrate, in addition to their difference in cell morphological features and organization (Figure 3, A and B; see Supplemental Figure S4 at http://ajp.amjpathol.org). Although large and intermediate infiltrates with well-developed T-cell areas were always associated with a reticular podoplanin+ cell network, smaller infiltrates without a T zone were often lacking it (not shown). A higher magnification showed that the podoplanin+ stromal cells also expressed desmin and surrounded conduit-like structures recognized by the ER-TR7 antibody, similar to LNs (Figure 3C). Contrary to a previous report,21 we occasionally found CD35+ FDCs inside the B-cell follicles of large RIP-CXCL13 infiltrates. However, the podoplanin+ cell network in the T zone was always CD35− (Figure 3D), suggesting the presence of two distinct stromal cell networks. Therefore, TLTs in all three tissue infiltration models displayed reticular podoplanin+ cell networks with a close resemblance to the TRC network of the LN T zone.
To further compare the phenotype of podoplanin+ CD31− stromal cells found in TLTs with their counterparts in LNs,10,12 a panel of markers was tested on pancreatic infiltrates of RIP-CXCL13–transgenic mice, which, among the three mouse models, displayed the highest frequency of large infiltrates and lymphoid tissue structures. Confocal microscopic images showed that many TRC markers, including BP-3, vascular cell adhesion molecule-1 (VCAM-1), platelet-derived growth factor receptor (PDGFR)α/β, and LTβR, colocalized with podoplanin staining (Figure 4A). Comparable to LNs,10 most CCL21 protein was localized in conduit-like structures that were enwrapped by podoplanin+ stromal cells (Figure 4B). By using in situ hybridization analysis, we detected Ccl21 mRNA expression on HEVs and in a reticular pattern in the T-cell–rich zone of pancreatic infiltrates, indicating that the CCL21 protein is locally produced (Figure 4C). To determine whether TLTs also form a conduit system, as found in SLOs,13,14 ectopic pancreatic infiltrates of RIP-CXCL13 transgenic (tg) mice were labeled with antibodies to podoplanin and various ECM molecules. A 3D reconstruction of confocal microscopic images, obtained from thick vibratome sections, confirmed that TLTs develop 3D networks of podoplanin+ cells wrapping around laminin+ matrix structures, similar to LNs (Figure 4D). These structures consisted of a cellular layer expressing podoplanin, a basement membrane containing fibronectin and laminin, and a core containing collagen-1 (Figure 4E). Pancreatic infiltrates in prediabetic NOD mice displayed matrix structures with the same composition (Figure 4E). The functionality of the conduits present in RIP-CXCL13 tg+ pancreata was demonstrated by the rapid transport of i.v. injected fluorescent tracer (dextran–Texas Red) that highlights the ER-TR7+ reticular network within minutes after injection (Figure 4F). Similar to LNs, major histocompatibility complex class II+ DCs were associated on the outside of the podoplanin+ cells, possibly retrieving antigen from the conduit lumen13 (Figure 4G). In summary, functional conduits develop in T-cell–rich areas of TLTs and are tightly associated with TRCs, the likely producers of these ECM molecules. Because of the high phenotypic and functional resemblance of TRC-like cells in TLTs with their counterpart in SLOs, we will refer to them as TRCs as well.
The formation and maintenance of several lymphoid structures in CXCL13-transgenic pancreata is known to be LTα and LTβR dependent, including HEVs, BP-3+, and CCL21+ stromal cells.21 To determine whether LTβR signaling was also involved in the maintenance of podoplanin+ TRC networks and conduits in RIP-CXCL13 tg+ mice, they were treated for 22 days with LTβR-Fc fusion protein, leading to the disappearance of most large infiltrates (see Supplemental Figure S5A at http://ajp.amjpathol.org).21 The remaining infiltrates showed small T-cell–rich zones that correlated with a marked reduction in the extent of the reticular TRC network (podoplanin+ CD31−Lyve1−) and of CD31+ blood vessels and Lyve1+ lymphatic vessels (Figure 5, A and B). Surprisingly, staining with antibodies against podoplanin, fibronectin, laminin, and collagen-1 revealed that the overall structure of the small remaining TRC network and conduits was not altered in LTβR-Fc–treated mice (Figure 5C, see Supplemental Figure S5B at http://ajp.amjpathol.org). In some infiltrates, the expression of the chemokine CCL21 was still detectable after LTβR-Fc treatment but always in a much reduced area (Figure 5D), confirming an earlier report21 and correlating with the smaller podoplanin+ TRC network. Therefore, the maintenance of functional and well-developed T zones in RIP-CXCL13 tg+ mice, including TRC networks and their associated conduits, is partially dependent on continuing LTβR signaling.
To test if LTi cells could play a role as the LTαβ source in the formation of TLTs and LTβR+ TRC networks in particular, we crossed RIP-CXCL13 tg+ mice to mice deficient in the transcription factor RORγ, which lack LTi cells and, consequently, all LNs and PPs.41 We confirmed by flow cytometry that few CD4+ CD3− CD45+ LTi cells expressing IL-7Rα and LTβR ligands can be found in the LNs and spleen of adult wild-type mice (Figure 6A), as previously reported.35,36,45 Consistent with the RORγ requirement of LTi cell development, cells with this phenotype were completely absent from spleens of adult RORγ−/− mice (Figure 6A), confirming an earlier report35 on RORγ−/− bone marrow chimeras. The CD4+ CD3− cells expressing IL-7Rα and LTβR ligands were also detected in the pancreas of adult RIP-CXCL13–transgenic mice, but not in nontransgenic littermates or transgenic mice on an RORγ-deficient background (Figure 6B, data not shown). In the three mouse strains, we did not obtain strong evidence for CD4− LTi cells because most CD4− CD3− cells were negative for IL-7Rα and LTβR ligands.46 Strikingly, in RIP-CXCL13–transgenic pancreas, the LTi frequency was in a similar range as in the SLOs of adult mice (Figure 6, A and B). Histological analysis showed that most IL-7Rα+ CD3− cells localized to perifollicular regions in the pancreatic infiltrates, similar to adult LNs and spleen (Figure 6C).45 The phenotype and localization of these cells in adult tissues suggest they are LTi cells. This finding raises the possibility of cross talk between these LTi cells and LTβR+ TRCs found in the same zone.
To test whether LTi cells play a role in the formation and maintenance of TLTs, we performed a histological analysis of pancreatic infiltrates in RORγ-deficient RIP-CXCL13 tg+ mice. Hematoxylin-stained sections revealed a striking deficiency in the formation of large infiltrates in these mice (Figure 6D). Only large infiltrates (more than 300 cells) have previously displayed clearly defined T- and B-cell–rich zones, along with extensive BP-3 and CCL21 expression.21 Therefore, we analyzed the composition and organization of infiltrates in RORγ-deficient RIP-CXCL13 mice using immunofluorescence microscopy. The small and few medium-sized infiltrates in these mice consisted of B cells and a few T cells that were not well segregated (Figure 6E). Distinct podoplanin+ TRC networks were lacking within the infiltrates while they were present within the spleen (data not shown). Podoplanin was only detected on the islet capsule and lymphatic vessels, similar to wild-type pancreata. The TRC-associated conduits were mostly absent, based on stainings for fibronectin, collagen I, laminin, ER-TR7, and CCL21 (Figure 6E, see Supplemental Figure S6 at http://ajp.amjpathol.org). Interestingly, PNAd- and MAdCAM1-positive HEVs were also strongly reduced in the infiltrates (see Supplemental Figure S6 at http://ajp.amjpathol.org). To exclude that the failure to develop TLTs in RORγ−/− mice was because of altered expression of LTαβ on B cells, we compared wild-type with RORγ−/− B cells by flow cytometric analysis. Yet, no difference in LTβR-Fc staining could be detected (see Supplemental Figure S7 at http://ajp.amjpathol.org). In addition to lacking LTi cells, RORγ−/− mice have strongly reduced numbers of mature T cells41 and lack type 17 helper T cell and natural killer p46+ cell subsets.47 To exclude a role for T cells in TLT formation, TCRβδ-deficient RIP-CXCL13 tg+ mice were analyzed. These mice showed normal development of large infiltrates containing mainly B cells. Interestingly, extensive podoplanin+ cell networks and fibronectin+ conduits developed in central parts of the B-cell–dominated infiltrates (Figure 6E, data not shown). In addition, PNAd+ and MAdCAM+ HEVs also developed normally (see Supplemental Figure S6 at http://ajp.amjpathol.org). These data indicate that the presence of LTi and/or possibly natural killer p46+, but not T cells, is required for the development of large LN-like structures in adult RIP-CXCL13 pancreas, including TRC networks, conduits, and HEVs.
Stromal cells may play a key role in the switch leading from acute resolving to chronic persistent inflammation.1 Therefore, we performed a comprehensive analysis of reticular stromal cells arising in various types of TLTs in human disease and murine models of inflammation in which TLTs form. We found that the T-cell–rich region of TLTs contains a TRC-like stromal cell type similar to LNs. It is distinct from FDCs in B zones. Therefore, larger TLTs closely replicate the stromal cell organization of SLOs that is critical for T- and B-zone compartmentalization.
In TLTs of NOD, RIP-LTα, and RIP-CXCL13 transgenic mice, extensive T-zone stromal cell networks were identified based on the following surface phenotype: podoplanin+ CD31− CD35−. These cells also expressed other TRC markers, including LTβR and its potential downstream targets (ie, BP-3 and VCAM-1).10,12 The coexpression of PDGFRα, PDGFRβ, and desmin is consistent with a mesenchymal origin for these cells and suggests that they may be functional myofibroblasts similar to their counterparts in LNs and wound healing.10 In support of this notion, we observed low αSMA expression in these cells (data not shown). The cell morphological features were characterized by long and fine processes typical for fibroblastic reticular cells and distinct from endothelial vessels. They wrapped around ECM structures comparable in composition and organization to the conduits in SLOs,7,12–14 extending previous studies20,23–25,48,49 showing reticulin and ER-TR7 staining in TLTs. We demonstrate herein that these conduits contain CCL21 and are functional because they can transport small-molecular-weight molecules. The elucidation of their precise function for TLT function will need further work. Thus far, functional conduits had only been described for LNs, spleen, and thymus.14,50,51 Similar to the two latter organs, it seems plausible that conduits in TLTs start next to blood vessels and drain the exudate, including blood-borne molecules and blood serum. In conclusion, the phenotype, organization, and compartmentalization of TRC-like cells, found in larger murine TLTs, are indistinguishable from their counterparts in SLOs.
One of the key functions of TRCs in SLOs is the secretion of CCL19 and CCL21.9,10 During the past few years, expression levels of CCL19 and CCL21 have been reported for various murine TLTs2,3,49 and occasionally in tumors.22,52 The CCL21 protein staining in murine TLTs was associated with HEVs and lymphatic vessels.2,3 Reticular CCL21 protein staining was previously reported in RIP-CXCL13 tg+ infiltrates.21 Herein, we show that most reticular CCL21 protein staining localizes to the conduits, suggesting active chemokine transport. Because CCL21 transcripts are also found in a reticular pattern throughout the T zone, it is likely that the podoplanin+ stromal cells forming a 3D network are the major local CCL21 source mediating the attraction of CCR7+ T cells and DCs. Indeed, DCs, presumably CCR7+, were often observed immediately adjacent to the podoplanin+ cells, suggesting active adhesion of DCs to this network, similar to LNs.8,53 T cells localized within the TRC network, suggesting that TRC-derived CCL21 promotes T-cell attraction and local motility that is thought to be critical for efficient selection of the rare antigen-specific T cells in SLOs.54 Therefore, we propose that TRC-like cells in TLTs reproduce both the structure and function of T zones in SLOs. They form a functional scaffold that brings together T cells with antigen-bearing DCs, thereby driving the local adaptive immune response, in addition to its potential role in angiogenesis of blood and lymphatic vessels. This notion is supported by a recent molecular analysis22 of TRCs from LNs and sites of inflammation and tumors, where TRCs were a source of transcripts for IL-7 and CCL19/21 and of angiogenic factors.
The second reticular stromal cell type present in SLOs, FDCs, has been previously described in many types of murine TLTs using CD35 and FDC-M1 as FDC-specific markers.2,3 In RIP-CXCL13 tg+ mice, only large infiltrates contained CD35+ FDC networks that localized among B-cell–rich areas and were not in contact with T cells, DCs, and HEVs. They were podoplaninlow and PDGFRα/β− and did not associate with CCL21 protein and conduits. Therefore, FDCs and TRCs found in TLTs are two separate reticular cell types with distinct anatomical localization (B- or T-cell zones) and function. In addition to providing structural support, we propose that these two networks and their chemokine expression profiles provide the underlying organizing principle of B- and T-zone segregation observed in many pathological TLTs,2,3 similar to their physiological counterparts in SLOs.55 How they develop and influence the local regulation of immune responses during disease remains to be shown.
Although FDCs have been well characterized in human SLOs and TLTs, little is known about the stromal cells of the T zone.2,3 Their presence in SLOs was described in several morphological studies56,57 showing a reticular network in T zones. Recently, a network of αSMA+ cells colocalizing with CCL19 and CCL21 expression was conserved in the SLOs and TLTs of patients with rheumatoid arthritis.26 Herein, we extend these findings by showing that these cells in SLOs and TLTs resemble each other, based on their surface phenotype (podoplanin+ CD21−) and function (ECM production and conduits). Podoplanin+ CD21− TRCs had long cell processes, formed a 3D reticular network, localized to T-cell–rich zones, and associated with complex ECM structures resembling conduits and with CCL21 protein and CD11c+ DCs. When cultured in vitro, these podoplanin+ cells exhibited typical fibroblast morphological features and showed high expression of ECM molecules and variable levels of αSMA (our unpublished data), similar to their murine counterparts12 (our unpublished data). Within the same TLT, we observed heterogeneity in the amount of CCL21 and ECM associated with the TRCs that could indicate different degrees of TRC differentiation or activation. The ECM staining, similar to that observed in LN T zones, has previously been described in the samples of patients with multiple sclerosis, suggesting that conduits and possibly TRC-like cells may be present in many more sites of chronic inflammation.58 In summary, TRC-like cells are a frequent component of ectopic infiltrates associated with human disease and closely resemble their SLO counterparts. They are likely to contribute, together with FDCs, to the overall organization and possible pathogenic function of human TLTs.
Given the frequent presence of TRCs at sites of chronic inflammation, understanding their development becomes an important issue. In this study, we show that the maintenance of the complex TRC network with its conduit structures within infiltrates of RIP-CXCL13 transgenic mice is partially dependent on LTβR signals, reminiscent of previous observations with HEVs.21,33 Other unidentified receptors appear to contribute to the formation of small and medium-sized infiltrates containing small TRC networks, with possible candidates being tumor necrosis factor-R1 and CD30.12,59 We propose that LTβR signals may be critical for the development and/or maintenance of larger and more differentiated stromal cell compartments, as observed in large infiltrates. The LTβR signaling can be triggered by two ligands (ie, LTαβ and LIGHT).33,60 Because the formation of pancreatic infiltrates in RIP-CXCL13 transgenic mice was largely dependent on LTα21 and the effects of LIGHT deficiency on lymphoid organ development are minor compared with LTαβ,60 we suggest that LTαβ is likely to be the dominant LTβR ligand for TLT formation. Signaling via LTβR on stromal cells is thought to be the key pathway inducing the expression of CCL19 and CCL21 in TRCs and of CXCL13 in FDCs. Because these chemokines, in turn, attract lymphocytes and further induce their LTαβ expression,55,60 this stromal-hematopoietic cross talk may lead to a positive feedback loop, allowing the further growth and organization of infiltrates, analogous to the process of embryonic LN/PP development. Blocking of LTβR signaling in RIP-CXCL13 tg+ mice likely interrupts this feedback loop and leads to a decreased size of the TRC network and conduit system. Although this may contribute to the reduction in TLT cellularity, the disappearance of well-differentiated HEVs, expressing integrin ligands, is likely to play an important role as well.33 In contrast to LNs, neutralization of LTβR signaling in adult mice partially disrupts stromal cell networks in splenic T and B zones, including CCL21 and CXCL13 expression.33,55 Therefore, stromal cells in TLTs resemble those in spleen because both depend on a continuous dialogue with LTαβ+ hematopoietic cells. In several other inflammatory models, such as RIP-CCL21 tg+, NOD, experimental autoimmune encephalitis, apolipoprotein E−/−, and influenza-infected mice, blocking LTβR leads to a strong reduction of TLT size and organization.33,49,61 Because the associated pathological features were reduced by this treatment, this pathway holds promise as a future drug target.33,62
The LTα1β2 signal required for splenic white pulp development is given by B cells and possibly LTi cells, whereas LTi cells are sufficient to provide this signal during early LN and PP development.27,28,34,63 Several recent studies have suggested that CD4+ CD3− LTi cells can persist into adulthood. Within the adult spleen, LTi-like cells may have organized T cell B cell segregation,34 repaired damaged stromal cell networks,35 and improved memory formation.45 Preliminary evidence37,38,64 suggested a role for LTi cells in CXCL13- or IL-7–induced TLT formation. We detected a population of CD4+ CD3− cells expressing IL-7Rα and LTαβ in the spleen and LNs of adult mice and within pancreatic infiltrates of adult RIP-CXCL13 tg+ mice. They share this phenotype with neonatal LTi cells and were absent from the spleen and pancreas of RORγ-deficient mice. Surprisingly, the absence of LTi cells reduced the size and organization of the infiltrate to a similar extent as the lack of LTα.21 The absence of B cells led to even smaller infiltrates, possibly because B cells are the main cell population attracted by CXCL13.21 The other possibility is that both LTi and B cells provide LTαβ signals to stromal cell precursors, possibly in a sequential manner. The appearance of LTi cells in RIP-CXCL13+ pancreata before birth was reported, with B cells colonizing shortly thereafter.37 Distinct lymphocyte accumulations become visible only 1 to 3 weeks after birth (our unpublished data), indicating that they develop postnatally, along with nasal-associated lymphoid tissue and PPs; therefore, they may be considered ectopic LNs. Because both cell types are still present in the infiltrates of adult mice, both may contribute to LTαβ-dependent TLT maintenance. The colocalization of LTαβ+ LTi cells with LTβR+ TRCs would be consistent with such an interaction that may lead to chemokine expression, as reported in other systems.34,35
The CD4+ CD3− cells were also detected in ectopic lymphoid infiltrates of adult apolipoprotein E−/−65 and villin-CXCL13 transgenic mice64; they could play a similar role in these mice. Ectopic LNs and PPs, developing in IL-7–overexpressing mice, were RORγ and LTα dependent, implying LTi cell involvement.38 In contrast, the lack of LTi cells [due to deficiency of inhibitor of DNA-binding (Id) 2] or LTα did not influence the formation of TLTs in a model in which CCL21 expression was expressed in the thyroid.25 This discrepancy may stem from the potential heterogeneity among LTi cells, with some being RORγ and others being Id2 dependent.4 Alternatively, the onset of tissue infiltration may differ, with lymphocytes infiltrating early after birth in RIP-CXCL13+ pancreata and possibly later in CCL21-overexpressing thyroid. Interestingly, RIP-CCL21 tg+ pancreata also contain LTi cells shortly after birth (our unpublished data), suggesting that both CXCL13 and CCL21 can attract LTi cells. Yet another difference may be the cell type providing lymphoid tissue–inducing signals (LTαβ or other) because B, T, or natural killer cells could replace LTi cells in their inductive function, especially when they are activated.60 In many autoimmune diseases, sufficient LTαβ is likely provided by activated lymphocytes.4 In line with this notion, we have not been able to reproducibly detect LTi cells in NOD pancreas (data not shown), possibly because of the relatively low expression of LTi attractants, such as CXCL13 and CCL21, at the inflammatory site.20 More studies4 using different murine TLT models are required to clarify whether LTi cells are a frequent feature of the development of ectopic lymphoid structures. Similar studies47 should be performed on human tissues because human LTi-like cells have also been recently described. Given that podoplanin+ TRC can evolve at sites of acute inflammation in the absence of LTαβ signals, the pathways leading to TRC formation are likely to be redundant, thereby increasing the robustness of the system.22
In summary, we demonstrate that the similarities between SLOs and ectopic lymphoid tissues (TLTs) extend further than previously appreciated. Lymphocytes, FDCs, and HEVs are characteristic features, as are the TRC network and the conduit system. The markers used in this study to identify TRCs should facilitate further examination of these cells in inflammatory diseases, both in mice and humans. In particular, podoplanin as a surface marker will allow isolation of these cells ex vivo and investigation of them further, both phenotypically and functionally. Finally, because TRCs are the main producers of the chemokines (ie, CCL19 and CCL21), TRCs might play an important role in the perpetuation of chronic inflammation and could represent an interesting target for therapy.
We thank Karin Hirsch for expert technical help; Mathias Heikenwälder for critical reading; Jeff Browning and Pascal Schneider for Fc-fusion proteins; Max D. Cooper, Willem van Ewijk, Beat Imhof, and Shin-ichi Nishikawa for antibodies or hybridomas; and Hans Acha-Orbea, Mathias Heikenwälder, and Dan Littman for mice.
Supported by the Peter Hans Hofschneider Foundation (M.S.), Arthritis Research UK and Autocure (European Union) (C.D.B.), and the Swiss National Science Foundation (grant number PPOOA-68805 and PPOOA-116896 to S.A.L.).
Author contributions: A.L. performed flow cytometry experiments, immunofluorescence microscopy on mouse tissues, and wrote the manuscript; D.L.H. performed the experiments on human tissues; S.F. performed immunofluorescence on mouse tissues; M.R.B. performed in situ hybridization; D.H.A. provided liver biopsy specimens; J.G.C. had the initial idea to cross RORγ−/− with RIP-CXCL13 mice when S.A.L. was a postdoctoral student with J.G.C.; M.S. performed immunofluorescence on vibratome sections and 3D image reconstruction; C.D.B directed the experiments on human tissues; and S.A.L. designed and directed the study; contributed to experiments, and wrote the manuscript. All authors critically read the manuscript.
The authors declare no competing financial or personal interests.
Comparison of three commercially available antibodies to human podoplanin tested on human tonsil sections (green) along with CD21-identifying B-zone stromal cells (FDC). All three antibodies to podoplanin show a comparable staining pattern.
Immunofluorescence staining of sections from either tonsils or salivary glands with Sjögren's syndrome, with the indicated antibodies. The three upper images of each organ represent stainings of the same section. A region selected from the third image is displayed at a higher magnification in the three lower images. Podoplanin+lyve1+ lymphatic vessels can be clearly distinguished from the more prevalent podoplanin+lyve1− reticular fibroblasts.
Additional examples of infiltrates in PBC liver (A) or salivary glands with Sjögren's syndrome (B) showing podoplanin+ CD21− T-zone stromal cell networks colocalizing with CD3+ T cells and podoplanin+ CD21+ stromal cell networks typical for B zones.
Immunofluorescence staining of a pancreas section from a prediabetic NOD mouse. Consecutive sections were stained for insulin-producing B cells (red) and infiltrating CD4+ cells (green) and for CD31 vessels (green; mainly HEV) and podoplanin+ FRC networks (red) colocalizing with infiltrating T cells.
Maintenance of the ectopic TRC network and conduit system is LTβR dependent. Adult RIP-CXCL13 mice (aged 4 months) were treated twice per week for 22 days with 100 μg of soluble LTβR-Fc or control LFA-3-Fc fusion proteins. A: Pancreata were sectioned and analyzed for the frequency and size of islet-associated mononuclear infiltrates using hematoxylin staining. Results were derived from 100 to 200 islets in each of the three transgenic mice per group (aged 7 to 12 weeks). Frequencies of small (5 to 30 cells/infiltrate in a section), medium (31 to 300 cells), and large (greater than 300 cells) infiltrates and total infiltrates. B: Confocal images of 8-μm-thick pancreatic sections stained with the conduit markers laminin (red) and collagen-1 (green). Scale bar = 10 μm. Data are representative of three experiments.
Immunofluorescence staining of a pancreas section from RIP-CXCL13 mice bred onto the RORγ+/+, RORγ−/−, or TCRβδ−/− background. The section stained for infiltrating CD3+ T and B220+ B cells is identical to the one displayed in Figure 6E and is shown to help localize the T- and B-cell–rich zones analyzed in consecutive sections for the stromal cell markers gp38/fibronectin and the HEV markers PNAd/MAdCAM. The RIP-CXCL13 on an RORγ−/− background develop few gp38+fibronectin+ structures and PNAd+ or MAdCAM+ vessels within the lymphocyte-rich infiltrate. Although some of the largest infiltrates are shown for each group, the small infiltrates (greater than 30 infiltrating cells) are most frequently observed on the RORγ−/− background; these rarely contain the stromal cell structures previously described.
The LTβR-Fc staining on B cells is unaltered in RORγ-deficient mice. Flow cytometric analysis of surface LTαβ expression on B cells in a single-cell suspension prepared from adult spleen of RORγ−/− or RORγ+/+ mice and labeled with antibodies to CD19, CD69, CD21, CD8α, and LTβR-Fc or control FN14-Fc fusion proteins. Histograms are pregated for naïve B cells (CD19+ CD69−) and show LTβR-Fc staining on RORγ+/+ (black line) and RORγ−/− (dashed line) B cells and control FN14-Fc fusion protein staining on RORγ+/+ B cells (gray shading). Data are representative of two experiments.