Monocyte-like and mature macrophages produce CXCL13 (B cell–attracting chemokine 1) in inflammatory lesions with lymphoid neogenesis

Rheumatoid arthritis (RA) and ulcerative colitis (UC) are chronic inflammatory diseases associated with de novo formation of irregular T- and B-cell aggregates in the synovium¹  and large bowel mucosa,²  respectively. Such local development of lymphoid tissue is thought to contribute to the pathology of chronic inflammation.^3,4 

Lymphoid tissue organization is orchestrated by a subset of the chemokine family, termed homeostatic or lymphoid chemokines because of their constitutive expression in secondary lymphoid tissues.⁵  One of these lymphoid chemokines, CXCL13, also called B cell–attracting chemokine 1 (BCA-1)⁶  or B-lymphocyte chemoattractant (BLC),⁷  and its receptor CXCR5, also called Burkitt lymphoma receptor-1 (BLR1), are required for normal development of secondary lymphoid organs in mice.^8,9  CXCL13 and CXCR5 have, in addition, been detected in various chronic human inflammatory diseases where lymphoid neogenesis occurs, such as Helicobacter pylori gastritis,¹⁰  RA,^1,11  Sjögren syndrome,^12-14  and UC¹⁵  as well as in gastric¹⁰  and primary central nervous system lymphoma.¹⁶ 

Follicular dendritic cells (FDCs) are generally believed to be the main source of CXCL13 in normal^6,7,17  as well as inflamed lymphoid tissue.^1,10,11  In addition, vascular expression of the CXCL13 protein has been reported,^{1,15,16,18,19}  and CXCL13 and CXCR5 have been shown to play an important role in B-cell homing to murine Peyer patches.²⁰  However, no expression analysis of CXCL13 mRNA in human lymphoid tissue^10,15,16,18  has directly shown that CXCL13 is actually produced by FDCs or endothelial cells. Importantly, available data do not exclude the possibility that murine FDCs in fact bind CXCL13 rather than produce it.^7,17  Also of note, CXCL13 protein is undetectable in human FDC-like cells stimulated in vitro.²¹  We have shown earlier that another lymphoid chemokine, CCL19, is present and functions in endothelial cells that are not its source.²² 

In our recent human study of normal and aberrant gut-associated lymphoid tissue, we found that CXCL13 was mainly associated with extracellular fibrils and only minimally with cells displaying the traditional FDC phenotype.¹⁵  This prompted us to explore more extensively the source of this chemokine in human lymphoid neogenesis represented by RA synovium and UC large bowel mucosa. Furthermore, we investigated the potential mechanistic role monocyte-derived cells could play in CXCL13 secretion. Our data documented for the first time that human monocytes/macrophages are a potent inducible source of CXCL13 and, in fact, appear to be the main producers of this chemokine in inflammatory lesions where lymphoid neogenesis occurs. This suggested a role of recently extravasated macrophages in the formation of such ectopic follicles.

Clinicopathologic data and patient treatment schedules are provided in Table 1. Cryosections (8 μm) were cut serially from various human tissue samples, including inflammatory lesions from patients with RA (n = 4) and UC (n = 5). RA synovium specimens were obtained from patients who underwent orthopedic surgery at Diakonhjemmet Hospital (Oslo, Norway) or Rikshospitalet University Hospital. UC bowel specimens came from patients who underwent colectomy at Rikshospitalet. Surgery was performed because of therapy-resistant inflammatory bowel disease. UC specimens were selected histologically on the basis of a large number of irregular lymphoid aggregates.¹⁵  Duodenal biopsies from adult patients with celiac disease (n = 6) were obtained before enrollment in a diet study. Control material consisted of palatine tonsils (n = 3), normal colon (n = 3), and normal Peyer patches (n = 3) obtained from individuals undergoing tonsillectomy or colectomy (due to long-lasting chronic obstipation), respectively. All procedures involving patient material were approved by the Regional Committee for Medical Research Ethics (Health Region South, Oslo, Norway).

Primary mouse monoclonal antibodies (mAbs) with the following human specificities were used for immunofluorescence staining: CXCL13 (53610.11, IgG1; R&D Systems, Oxon, United Kingdom), CD2 (IgG2a; kindly provided by Dr G. Gaudernack, Oslo, Norway), CD4 (72-5A4, IgG2a; Diatec, Oslo, Norway), CD11c (IgG1; kindly provided by Dr K. Pulford, Oxford, United Kingdom and SHCL-3, IgG2b; Becton Dickinson, San Jose, CA), CD14 (anti–human Leu-M3, IgG2b; Becton Dickinson), CD20 (L26, IgG2a; Dako, Glostrup, Denmark), CD33 (My9, IgG2b; Coulter Immunology, Hialeah, FL), and CD68 (EBM11, IgG1; and PG-M1, IgG3; Dako). In addition, we used the following rabbit immune reagents: anti-CD3 (IgG; Dako), anti–von Willebrand factor (IgG; Dako), and antifibronectin (IgG; Dako).

Cryosections were subjected to 3-color immunostaining with mAbs or rabbit immune reagents followed by appropriate second- and third-step reagents as previously described.¹⁵  Briefly, acetone-fixed sections were incubated with a mixture of primary immunoreagents for 1 hour at room temperature. The mAb to CXCL13 was used at a concentration of 10 μg IgG/mL in combination with different reagents detecting leukocyte markers, vessel structures, or extracellular matrix proteins (see “Primary antibodies”). This combination was followed by incubation with the appropriate mixture of reagents and conjugates. When a secondary biotinylated antibody was used, a final incubation step with fluorescent streptavidin was included. All secondary reagents were diluted in 10% normal human serum to block Fc receptor binding and possible cross-reacting antibodies to IgG present in the tissue sections. Reaction sites of anti-CXCL13 were visualized with a subclass-specific secondary reagent (goat anti–mouse IgG1; 1/2000; Southern Biotechnology, Birmingham, AL). Incubation with irrelevant isotypeand concentration-matched mAbs or preimmune rabbit serum followed by the appropriate secondary reagents served as negative controls to assess nonspecific staining. To exclude that costaining was due to macrophage autofluorescence or cross-reactivity of the mAb to CXCL13 with other chemokines, specificity was also confirmed by blocking of the anti-CXCL13 mAb with recombinant CXCL13 (R&D Systems) at an antibody-chemokine molar ratio of 1:2.5. Stained sections were mounted in poly(vinylalcohol) (P8136; Sigma Aldrich, Steinheim, Germany), pH 8.7. Fluorescence microscopy was performed with a Nikon microscope (EcLipse E800, Nikon, Tokyo, Japan) equipped with a Ploem-type beam-splitting device. Images were captured with an F-view digital camera controlled by Analysis 3.2 software (Analysis Soft Imaging System; GmbH, Münster, Germany) and transferred to Adobe Photoshop (Adobe, San Jose, CA).

A 436–base pair (bp) digoxigenin (DIG)–labeled riboprobe was generated from the coding region of cDNA for human CXCL13 with the DIG RNA labeling kit according to the manufacturer's directions (Boehringer Mannheim, Mannheim, Germany). Hybridization and detection of the hybridized probe were performed as previously described.¹⁵  Briefly, after 10 minutes fixation with 4% paraformaldehyde (in phosphate-buffered saline [PBS]) and subsequent washing, the sections were hybridized overnight at 59°C with 250 ng/mL riboprobe in hybridization solution followed by high-stringency washes. The hybridized probe was detected with horseradish peroxidase (HRP)–conjugated rabbit anti-DIG (Dako), followed by 2 steps of biotin-tyramide deposition (Gen Point kit; Dako). Finally, signals were detected with alkaline phosphatase (AP)–conjugated rabbit antibiotin (Dako), followed by the AP substrate Fast Red (Ventana Medical Systems, Tucson, AZ), and the sections were counterstained with hematoxylin.

Human monocytes were isolated from healthy blood donors by IsopachFicoll (Lymphoprep; Nycomed, Oslo, Norway) gradient centrifugation, followed by negative isolation with the Monocyte Negative Isolation Kit (Dynal, Oslo, Norway) according to the manufacturer's directions. Monocyte purity was assessed by flow cytometry and was routinely 85% to 90%. Cells were plated at 7 × 10⁵ cells/well in 24-well plates with RPMI 1640 (BioWhittaker, Walkersville, ML) containing 10% fetal calf serum and either macrophage colony-stimulating factor (M-CSF; 50 ng/mL; R&D Systems) or granulocyte-macrophage colony-stimulating factor (GM-CSF; 100 ng/mL; Peprotech, London, United Kingdom) and interleukin 4 (IL-4; 25 ng/mL; R&D Systems) for 3 or 6 days before stimulation with lipopolysaccharide (LPS; from Escherichia coli serotype 026:B6; Sigma-Aldrich) at various concentrations for different time periods as detailed in the text.

Total RNA was isolated from cultured monocytes by means of RNAwiz (Ambion, Austin, TX) according to the manufacturer's protocol, and reverse transcription-polymerase chain reaction (RT-PCR) was performed as described.¹⁵  Primers used were (5′ to 3′) BCA-1: TGCTAATGAGCCTGGAC, AGGGATAAGGGAAGAATG and glyceraldehyde-3-phosphate dehydrogenase (GAPDH): AAATCCCATCACCATCTTCC, CATGAGTCCTTCCACGATACC. Annealing temperature for both primer pairs was 60°C, and 35 cycles of PCR amplification were performed. The PCR products were separated by electrophoresis in 1.6% (wt/vol) agarose gel and stained with ethidium bromide. The PCR product was confirmed by sequencing.

Mouse mAb to human CXCL13 (53610.11; IgG1) was used as capture antibody and biotinylated goat IgG anti–human CXCL13 served as detection antibody (both from R&D Systems). Streptavidin-conjugated AP (Southern Biotechnology) was used to detect the biotinylated antibody and was visualized with diethanolamine. Absorption was measured at 405 nm. Recombinant human CXCL13 protein (R&D Systems) was used for standard curves. The detection limit was 200 pg/mL.

In situ hybridization on tissue sections from RA and UC lesions showed that CXCL13 mRNA expression was not restricted to organized lymphoid structures but also occurred in areas with diffuse mononuclear cell infiltration and in small clusters of such cells (Figure 1A-B). Immunofluorescence staining for CXCL13 on parallel sections revealed that the CXCL13-expressing cells were CD11c⁺ and located within or near small collections of B cells (Figure 1C and data not shown). Evaluation of serial sections at different levels confirmed that CD11c⁺ CXCL13-expressing cells were not restricted to the 3-dimensional structure of organized lymphoid follicles. CXCL13-expressing cells were detected outside of follicles in all RA and UC specimens. However, these extrafollicular cells were most abundant in specimens characterized by high numbers of organized lymphoid structures or dense clusters of mononuclear cells; the figures are representative of such specimens. CXCL13-expressing cells were not convincingly detected in celiac disease biopsies.

We also showed that scattered CXCL13-expressing cells in germinal centers of organized lymphoid tissue (Figure 1D-E and Carlsen et al¹⁵ ) and inflammatory B-cell aggregates (not shown) were CD11c⁺. This strongly suggested that they were identical to the previously described germinal center dendritic cells (GCDCs) reported to stimulate T cells in this lymphoid compartment.²⁴ 

To further investigate the in situ phenotype of all CD11c⁺ cells in germinal centers, we costained for CD11c and the macrophage marker CD68 in sections from UC (Figure 2A) and RA (not shown) lesions as well as tonsils (Figure 2B) and Peyer patches (not shown). Surprisingly, all CD11c⁺ cells located in germinal centers expressed CD68 (Figure 2A-B). This was confirmed by application of different combinations of mAbs (not shown). As expected, numerous CD68⁺CD11c^– macrophages were detected outside of follicles in UC and RA (Figure 2A), whereas some CD68⁺CD11c^– macrophages together with CD11c⁺ CD68^– dendritic cells were seen in the T-cell areas of lymphoid tissue such as Peyer patches and tonsils (Figure 2B). The largest and most prominent CXCL13-expressing cells were found within the germinal centers. These cells were evenly distributed in this compartment (Figure 2C), corresponding to the previously described distribution of GCDCs, whereas macrophages described in germinal centers, the so-called tingible body macrophages, are located mainly in the dark zone.²⁴ 

Paired immunostaining for the macrophage marker CD68 in UC and RA lesions showed that CXCL13-expressing cells, both within germinal centers and scattered in the T-cell areas, as well as within or near small collections of B cells, also expressed variable levels of CD68 (Figure 2D and data not shown). Far from all CD68⁺ cells were positive for CXCL13, but costaining for the extracellular matrix protein fibronectin showed that this chemokine was either expressed by CD68⁺ cells or associated with extracellular fibrils (Figure 3A-C). Thus, our finding suggests that tissue macrophages constitute the major source of CXCL13 in these inflammatory lesions, and the secreted chemokine appeared to be subsequently deposited on fibrillar structures. Numerous CD68⁺ tissue macrophages were also present in celiac disease lesions, but none of these deemed to be positive for CXCL13 (Figure 3D).

To further investigate the phenotype of the CXCL13-expressing macrophages, we costained for CD14, which is a marker for recently recruited macrophages in inflammatory bowel disease lesions.²⁵  Scattered CD14⁺ cells that expressed distinctly CXCL13 were detected outside of the lymphoid aggregates in inflamed tissue (Figure 3E), which suggested that such immature macrophages could give rise to the progeny of cells that constitute the major tissue source of this chemokine. Notably, CD68⁺CXCL13⁺ macrophages were also found at such sites where neither FDCs nor high endothelial venules (HEVs) could be detected. However, CXCL13-expressing macrophages were also abundantly found in organized lymphoid structures of inflammatory lesions where FDCs and HEVs could be identified.

The possibility exists that macrophages rapidly take up CXCL13 from another cellular source, resulting in the immunohistologic CD68⁺CXCL13⁺ phenotype. We therefore tested whether stimulated monocytes are able to produce CXCL13. Freshly isolated human monocytes from healthy donors did not contain detectable mRNA for CXCL13. However, 4 hours after stimulation with endotoxin (LPS) from E coli, such a message was detected in the monocytes (Figure 4A). We also matured monocytes for 3 days to differentiate them toward macrophages (with M-CSF; Becker et al²⁶ ) or dendritic cells (DCs; with IL-4 and GM-CSF; Sallusto and Lanzavecchia²⁷ ) prior to LPS stimulation. CXCL13 mRNA was not constitutively expressed in these cell populations as likewise observed for immature DCs.²⁸  Nevertheless, after LPS stimulation, an increased level of CXCL13 mRNA was detected in the macrophages matured in vitro compared with monocytes; in contrast, the level in the DCs matured in vitro tended to be decreased (Figure 4A). CXCL13 protein was also detected by immunofluorescence staining of monocytes (Figure 4B) and macrophages matured in vitro following LPS stimulation (Figure 4C).

Association of CXCL13 with fibronectin-bearing extracellular fibrils is a striking feature of both normal lymphoid tissue and inflammatory aggregates (Figure 3A-C and Carlsen et al¹⁵ ). We therefore addressed the question of whether monocytes and macrophages can secrete CXCL13, which subsequently becomes positioned on extracellular matrix components. Supernatants of cultured monocytes were analyzed with a sandwich enzyme-linked immunosorbent assay (ELISA) specific for CXCL13. During 3 days of culture with LPS, increasing amounts of CXCL13 protein were detected, but with large individual differences (Figure 5A). Adding M-CSF at the start of culture did not alter the secretion of CXCL13 protein (data not shown). Dose-response experiments showed that in 2 of 3 individual cultures, CXCL13 protein was secreted by monocytes in response to LPS concentrations down to 0.1 ng/mL (Figure 5B). Increasing the LPS concentrations up to 1 μg/mL enhanced the secretion only slightly. However, macrophages generated in vitro secreted substantially elevated amounts of CXCL13 protein in response to increasing LPS concentrations (Figure 5C). When monocytes were cultured with M-CSF for 6 days prior to stimulation, the dose-response relationship with LPS was further elevated (Figure 5D). Conversely, when monocytes were differentiated toward DCs by incubation with IL-4 and GM-CSF for 3 days prior to LPS stimulation (100 ng/mL), no CXCL13 could be detected in the culture supernatant (data not shown).

We show here that human monocytes/macrophages are a potent source of CXCL13 and, in fact, appear to be its main source in inflammatory lesions of RA and UC where lymphoid neogenesis occurs. This novel finding was supported by the presence of both CXCL13 protein and mRNA in CD68⁺cells, whereas the protein alone was shown to be associated with extracellular fibrils. The presence of CD68⁺ CXCL13-expressing macrophages also at inflammatory sites where neither FDCs nor HEVs could be detected accorded well with an earlier finding that CXCL13 and lymphotoxin β are necessary, but not sufficient, for the occurrence of FDCs in RA.¹  It furthermore agreed with murine studies in which ectopic expression of CXCL13²⁹  or other lymphoid chemokines³⁰  was associated with lymphoid neogenesis, including development of HEVs, although failing to generate convincing FDCs.

The causative role of CXCL13 in inflammation-associated lymphoid neogenesis is difficult to address directly in human studies. However, in a study of 64 patients with RA¹  CXCL13 mRNA at varying levels was detected in synovial tissue extracts. The largest quantities of CXCL13 mRNA were detected in synovial tissues that contained germinal centers, determined to be 15- to 30-fold more than in synovial tissues with T-cell/B-cell aggregates or diffuse lymphocytic infiltrates. This finding suggested that CXCL13 was necessary, but not sufficient, for a synovial germinal center reaction. Although the source of CXCL13 was not identified, CXCL13 protein was abundantly present in germinal centers but also dispersed throughout synovial tissue without germinal centers. We did not detect CXCL13 in biopsies from celiac disease, a chronic lesion with great immunologic activity but without lymphoid neogenesis. This discrepancy could, however, be a matter of sensitivity and inflammatory intensity because active UC is characterized by more severe mucosal inflammation than active celiac disease. Nevertheless, our findings suggested that a certain level of CXCL13 is necessary for the development of lymphoid follicles in inflamed tissue.

FDCs have been previously considered to be the main source of CXCL13, both in normal and aberrant lymphoid tissue. However, expression analyses of CXCL13 mRNA in human lymphoid tissue^10,15,18  have not supported the notion that CXCL13 is actually produced by FDCs, and no CXCL13 protein could be detected after in vitro stimulation of isolated human FDC-like cells.²¹  Although vascular localization of CXCL13 has been detected in human and murine lymphoid tissue^15,18,19  as well as in inflamed human tissue^1,13,15  and malignant lymphoma,¹⁶  in situ hybridization for mRNA did not support the possibility that endothelial cells do synthesize CXCL13.^12,15,16,18  This has led to the notion that the chemokine can be taken up in endothelial cells by transcytosis.^13,16  Indeed, we have shown this earlier for another lymphoid chemokine, CCL19, which is present and functions in endothelial cells although being produced by extravascular cells.²²  In a study of RA synovial tissue, CXCL13 expression was likewise detected in synovial fibroblasts despite the fact that endothelial cells were the most frequent cell type positive for CXCL13.¹  Our finding of CXCL13 associated with extracellular fibrils could possibly explain its apparent expression by synovial fibroblast as shown by immunohistochemistry.¹  However, the possibility remains that synovial fibroblasts might represent an additional source of CXCL13 in RA because these cells cannot be distinguished among scattered cells expressing CXCL13 mRNA.

In a recent study of primary lymphoma within the central nervous system, in situ hybridization verified the expression of CXCL13 mRNA by malignant B cells,¹⁶  and another study showed that malignant B cells obtained from follicular lymphomas could secrete CXCL13.³¹  Notably, malignant B cells³¹  and activated normal CD11c⁺ blood DCs²⁸  are the only human cells previously reported to secrete CXCL13 protein. We show here that monocytes, macrophages, and immature DCs can be induced to express CXCL13 mRNA, whereas there is no constitutive expression in these cells. Interestingly, only activated monocytes and macrophages were found to secrete CXCL13 protein, as opposed to the immature DCs. Our negative mRNA data for unstimulated cells are in line with a study showing constitutive expression of CXCL13 mRNA in day 6 monocyte-derived DCs, whereas such expression was not detected in the day 3 immature DCs.²⁸  Interestingly, our in vitro stimulation data showed that LPS titration only slightly influences CXCL13 secretion in monocytes, whereas in macrophages it is highly dependent on the LPS level. Thus, at the lowest LPS concentrations (0.1 ng/mL) only monocytes but not macrophages secreted detectable CXCL13 protein, whereas at 1 ng/mL LPS the two cell types secreted comparable amounts and the macrophages continued to show elevated secretion when the concentration of LPS was increased.

Importantly, our data clearly show that activation is required for CXCL13 protein secretion to occur from both monocytes and macrophages. This is well in keeping with the presence of recently extravasated macrophages in inflamed tissue. Because the purity of negative selection was not 100% in our monocyte cultures, a small population of contaminating cells could not be completely ruled out as an in vitro source of CXCL13. However, this possibility would not have explained the striking effect of culture with M-CSF as opposed to culture with IL-4 and GM-CSF prior to LPS stimulation. As shown by immunostaining, neither in vitro cultured monocytes nor macrophages appeared to produce uniformly CXCL13, which agrees with the relatively low level of CXCL13 secretion obtained. This heterogeneity could possibly reflect different activation stages or regulatory mechanisms, and accorded with our in situ data, where far from all monocytes/macrophages in RA and UC lesions were shown to express CXCL13.

CD4⁺CD11c⁺CD3^– GCDCs were originally distinguished from tingible body macrophages (which likewise are CD4⁺CD11c⁺CD3^–) in tonsillar germinal centers by a flow cytometric procedure in which large cells with bright autofluorescence were gated out.²⁴  GCDCs isolated from tonsils by this procedure have been reported to express CXCL13 mRNA.²⁸  Our finding of strong CD68 expression in all germinal center CD11c⁺ cells suggested a common macrophage precursor for GCDCs and tingible body macrophages. Indeed, the latter cell type has previously been distinguished from GCDCs only by its intense autofluorescence and nonspecific esterase staining.²⁴  Our data clearly implied that GCDCs and tingible body macrophages present in germinal centers represent different maturational or activation stages derived from the monocyte lineage. Importantly, all GCDCs did not express the same high level of CXCL13 as shown in Figures 1D-E and 2C, and such variability might reflect different activation stages of the actual germinal center. In our previous study of gut-associated lymphoid tissue,¹⁵  CXCL13 mRNA was detected in most, but not all, germinal centers by in situ hybridization, whereas CXCL13 mRNA always occurred at the periphery of the follicles. Together, these findings point to the possibility that GCDCs do not constitutively express CXCL13, but need a certain level of activation to produce this chemokine.

Myeloid DCs have been identified as a source of aberrant CXCL13 expression in murine lupus, and CXCL13 was suggested to play a pivotal role in the pathogenesis of this mouse model of autoimmune disease.³²  In that study, CD11c⁺CD11b⁺ DCs were identified as the main source of this chemokine,³²  which accords with our finding of CD11c⁺ GCDCs expressing CXCL13 in human chronic inflammatory lesions. Peritoneal macrophages have also been identified as a source of CXCL13 in normal mice.³³  Interestingly, a recent study of mouse spleen showed that macrophages control the retention and trafficking of B lymphocytes in the marginal zones of this organ.³⁴  However, there are no reports of CXCL13-producing tissue macrophages in murine disease models. Thus, our data are the first evidence for CXCL13 production by macrophages in inflammatory lesions. Our identification of scattered CD14⁺ and CXCL13-expressing cells outside of the lymphoid aggregates in inflamed tissue strongly suggests that extravasated monocytes may give rise to the progeny of tissue cells producing this chemokine. Importantly, these cells might also be the source of CXCL13 present in vascular endothelium^1,15  and associated with other cell types as well.¹ 

In conclusion, our study documents that human monocytes/macrophages are potent inducible producers of CXCL13 and appear to represent the main source of this chemokine in inflammatory lesions where lymphoid neogenesis occurs. Within germinal centers, these producers acquire the previously defined GCDC phenotype.²⁴  Also notably, we identified CXCL13-expressing monocytes/macrophages at inflammatory sites even in the absence of FDCs or HEVs. Taken together, these results suggest that induction of CXCL13-producing cells of the monocyte linage is an early event in human lymphoid neogenesis associated with chronic inflammation. Therefore, such cells may represent a link between innate and specific immune responses in the pathogenesis of UC and RA. This points to circulating monocytes as candidates for targeted therapy of inflammatory disease in the future.

The skillful technical assistance of Aaste Aursjø and Kathrine Hagelsteen is gratefully acknowledged.