A chemokine-driven positive feedback loop organizes lymphoid follicles

 Lymphoid follicles are B-cell-rich compartments of lymphoid organs that function as sites of B-cell antigen encounter and differentiation. CXC chemokine receptor-5 (CXCR5) is required for B-cell migration to splenic follicles 1 , but the requirements for homing to B-cell areas in lymph nodes remain to be defined. Here we show that lymph nodes contain two types of B-cell-rich compartment: follicles containing follicular dendritic cells, and areas lacking such cells. Using gene-targeted mice, we establish that B-lymphocyte chemoattractant (BLC/BCA1) 2 , 3 and its receptor, CXCR5, are needed for B-cell homing to follicles in lymph nodes as well as in spleen. We also find that BLC is required for the development of most lymph nodes and Peyer's patches. In addition to mediating chemoattraction, BLC induces B cells to upregulate membrane lymphotoxin 1 2, a cytokine that promotes follicular dendritic cell development and BLC expression 4 , 5 , establishing a positive feedback loop that is likely to be important in follicle development and homeostasis. In germinal centres the feedback loop is overridden, with B-cell lymphotoxin 1 2 expression being induced by a mechanism independent of BLC.

The BLC gene was inactivated by deletion of a portion of exon 2 which encodes amino acids 27�60, including 3 of the 4 conserved cysteine residues (Fig. 1a ). Mice homozygous for the targeted locus (Fig. 1b ) lacked detectable BLC messenger RNA ( Fig. 1c ) and protein (Fig. 1d ). Anatomical analysis revealed that BLC-deficient mice had severe but incompletely penetrant defects in development of peripheral lymphoid organs. Most mice lacked inguinal, iliac, sacral, brachial and axillary lymph nodes, among others ( Table 1 ); however, most of these lymph nodes were found at varying low frequencies, and several other lymph nodes developed normally, with all animals possessing a full set of mesenteric lymph nodes ( Table 1 ). The lymphoid patch of the caecum was also absent and the number of Peyer's patches was severely reduced (Table 1 ). Those Peyer's patches that were found were typically smaller and lacked the characteristic multi-domed structure of wild-type Peyer's patches. Previous characterization of CXCR5-deficient mice established that they typically lack inguinal lymph nodes and have a less severe deficiency in Peyer's patches than do BLC-deficient mice 1 . Further analysis of animals on a 129 or 129 B6 mixed background indicated that they often lack additional peripheral lymph nodes (Table 1 ). These findings establish a broad requirement for the BLC�CXCR5 ligand�receptor pair in one of the earliest steps in the development of Peyer's patches and most lymph nodes.

To test whether BLC is needed for organization of B cells in lymphoid follicles, sections of lymph nodes and spleen from BLC -/- animals were analysed for B-cell distribution. In all these organs, B cells failed to organize in polarized follicular clusters, and instead appeared as a ring of cells at the perimeter of T-cell areas (Fig. 2a ). The boundary between B-cell-rich areas and T zones was often poorly demarcated, with increased numbers of B cells and T cells in reciprocal areas ( Fig. 2a , b ). In addition, staining spleen sections for immunoglobulin (Ig)M and IgD revealed a thickened ring of IgM hi IgD lo marginal zone B cells (Fig. 2a , top). The segregation of B cells between this area and the inner ring of B cells was also disrupted, with increased numbers of IgM lo IgD hi B cells located in the outer marginal zone (Fig. 2a , top).

A key property of B-cell follicles is the presence of immune-complex-presenting follicular dendritic cells (FDCs) 6 . Staining for complement receptor-1 (CR1), which is highly expressed on FDCs and marginal zone B cells, revealed an absence of primary follicle FDCs in BLC-deficient spleen and lymph nodes (Fig. 2b ). Previous studies in CXCR5-deficient mice established a requirement for this receptor in B-cell follicle formation in spleen and Peyer's patches, but did not indicate that the receptor has a role in lymph nodes 1 . During our analysis of BLC expression in lymph nodes, however, we observed two types of B-cell-rich zones: polarized follicles that are BLC positive and contain FDCs but few T cells; and areas that lack staining for BLC or FDC markers, contain substantial numbers of T cells, and lack a polarized morphology (Fig. 2c , see also e, f). BLC staining was detected on a subset of CR1 + FDCs and on adjacent CR1 - follicular stromal cells ( Fig. 2d ). Analysis of lymph nodes from CXCR5 -/- animals for these two types of B-cell areas revealed that they lack FDC-containing primary follicles (data not shown). In transfer experiments, CXCR5 +/+ B cells localized in FDC-containing follicles and FDC-deficient B-cell areas of wild-type lymph nodes (Fig. 2e ). By contrast, CXCR5 -/- B cells failed to enter FDC-containing follicles (Fig. 2f ). CXCR5 -/- B cells also did not migrate into follicles in the spleen and Peyer's patches, as previously observed 1 ; however, CXCR5 -/- B cells did localize in the FDC-deficient B-cell-rich areas in lymph nodes (Fig. 2f ). B-cell homing to lymph node follicles therefore requires BLC and CXCR5, but neither molecule is necessary for migration to FDC-deficient B-cell-rich regions of secondary lymphoid tissues.

A lack of FDCs and organized follicles, as well as impaired development of several lymphoid organs, are phenotypes shared between BLC- and lymphotoxin (LT)-deficient mice 4 . Cell transfer experiments showed that B cells are an essential source of LT 1 2 for the development of FDCs, but it was not clear whether LT 1 2 was expressed by naive B cells 7-9 . Although B cells express LT constitutively 10-12 , surface expression of this subunit requires LT , and LT expression has only been reported after exposure to activating stimuli, such as CD40L or interleukion (IL)-4 (ref. 11 ), or treatment with endotoxin or phorbol esters 13 . Development of primary follicles is not associated with B-cell activation, however, as it occurs in germ-free mice 14 , Ig-transgenic mice 15 , T-cell-deficient mice 16 , CD40-deficient mice 17 and IL4-deficient mice 18 . To investigate LT 1 2 expression on naive B cells, we used a soluble form of the LT 1 2 receptor, LT R-Ig, to stain freshly isolated cells (Fig. 3a�g ). Notably, a significant fraction of total B cells from wild-type spleen showed low but detectable LT R-Ig binding compared with LT - or LT - deficient spleen cells (Fig. 3c , d , g ). An even larger fraction of lymph node B cells stained with LT R-Ig (Fig. 3e , f , g ), whereas B cells in the blood (Fig. 3a , b , g ) and bone marrow (data not shown) were mostly negative. Similar levels of LT 1 2 were observed on B cells in Ig-transgenic animals (Fig. 3g ), showing that the expression was not induced by antigen. We then tested LT 1 2 expression on B cells in BLC-deficient animals (Fig. 3h ). LT 1 2 expression on BLC -/- peripheral blood B cells was not different from wild-type controls, but expression on B cells from spleen and lymph nodes was significantly reduced (Fig. 3h ), establishing that BLC is required for normal LT 1 2 expression on naive B cells.

To determine whether recirculating B cells upregulated LT 1 2 expression as they moved into BLC-expressing lymphoid organs, we isolated cells from the blood of Ly5.1 + donor mice and transferred them into congenic Ly5.2 + recipients. At 0.5 and 6 h after transfer, recipient splenocytes were collected and stained with anti-Ly5.1 to detect the transferred cells, and with LT R-Ig (Fig. 3i ). At the early time point, when most of the transferred B cells in the spleen are in the non-lymphoid area (data not shown), few of the cells expressed LT 1 2 (Fig. 3i ). Six hours after transfer, however, when many of the transferred cells are in BLC + follicular areas (data not shown), LT 1 2 expression was readily detectable (Fig. 3i ). When B cells were taken from spleen and transferred intravenously to syngeneic recipients, LT R-Ig staining of cells in recipient blood and spleen soon after transfer had diminished to levels similar to endogenous blood B cells (Fig. 3i ). The mechanism for this downregulation is not yet understood, but it suggests that as B cells leave a lymphoid organ and enter the blood they rapidly lose surface LT 1 2 expression. In practical terms, this finding allowed us to use cells from spleen for further transfer experiments to test the mechanism of LT 1 2 upregulation on cells entering lymphoid organs. Within 6 h of transfer, untreated spleen B cells expressed amounts of LT 1 2 equal to endogenous B cells (Fig. 3i ). By contrast, cells pretreated with pertussis toxin (PTX), an inhibitor of signalling by G i-coupled receptors including all known chemokine receptors, failed to upregulate LT 1 2 expression after transfer (Fig. 3i ). Furthermore, minimal LT 1 2 upregulation occurred when B cells from wild-type mice were transferred to BLC-deficient recipients (Fig. 3j , k ). Notably, when B cells from BLC-deficient donors were transferred to wild-type recipients an exaggerated induction of LT 1 2 occurred (Fig. 3l , m ), perhaps because B cells that develop in BLC-deficient animals are hypersensitive to the chemokine. Together, these findings provide strong support for the hypothesis that recirculating B cells upregulate LT 1 2 as they migrate to B-cell areas in response to BLC.

To test whether BLC directly induces LT 1 2 expression, we incubated naive B cells in vitro with recombinant BLC (Fig. 4 ). BLC induced a dose-sensitive upregulation of LT 1 2 on cultured B cells (Fig. 4a ). The induction was inhibited by pre-incubation of the cells with PTX (Fig. 4b ). Three other chemokines with chemotactic activity on naive B cells were also tested: the broadly expressed CXCR4 ligand, SDF1; and SLC and ELC, two related CC chemokines made in the T-cell area that are ligands for CCR7 (ref. 19 ). Each of these chemokines caused detectable increases in LT 1 2 expression on B cells, although in all cases the maximal induction was lower than the induction by BLC (Fig. 4c ; and data not shown). Although it is not yet clear whether these chemokines can induce LT 1 2 expression on naive B cells in vivo , it seems possible that they contribute to the remaining expression in BLC -/- mice (Fig. 3h ). BLC-mediated induction of LT 1 2 was sensitive to actinomycin D, cycloheximide and wortmannin (Fig. 4d ), providing evidence that BLC may signal through phosphatidylinositol-3-OH kinase to induce lymphotoxin transcription.

Despite low expression of LT 1 2 on naive B cells and the absence of primary follicle FDCs in BLC-deficient mice, germinal centres formed in lymph nodes and spleen after immunization with a T-cell-dependent antigen (Fig. 5a ). These germinal centres were misplaced and usually smaller than those found in wild-type controls, yet they contained CR1 + FDC networks (Fig. 5a ). Similar observations for spleen germinal centre formation have been made in CXCR5-deficient mice 20 . Germinal centre FDCs, like those of primary follicles, require LT 1 2 for their development and maintenance 4 , 21 . Staining with LT R-Ig revealed high LT 1 2 expression on GL7 + germinal centre B cells in wild-type mice (Fig. 5b ). In contrast to our observations for naive B cells, LT 1 2 expression on germinal centre B cells was not dependent on BLC ( Fig. 5b ). We conclude that the requirement for BLC to induce LT 1 2 is overridden in germinal centres, most likely by T-cell-derived signals. CD40 signalling is necessary for germinal centre formation, and might provide the required signal for inducing LT 1 2. In agreement with this possibility, CD40-mediated induction of LT 1 2 is independent of CXCR5 (Fig. 5c ).

In summary, our findings indicate that BLC has dual roles in follicular compartmentalization of B cells: mediating B-cell attraction, and inducing increased LT 1 2 expression on the recruited cells. LT 1 2 then engages LT Rs on non-haematopoietic stromal cells, promoting maturation of FDCs and leading to increased expression of BLC 4 , 5 . FDC maturation is regulated by signals in addition to LT 1 2, including signals from tumour necrosis factor/tumour necrosis factor receptor-1 (ref. 4 ), and it remains to be determined whether these molecules function in the same or parallel pathways. In addition to showing the expression of LT 1 2 on naive B cells, we have established the existence of a positive feedback loop between BLC and LT 1 2. This feedback loop is likely to be critical in follicle development, causing the low amounts of BLC that are expressed independently of LT 1 2 (ref. 5 ) to become upregulated as B cells are recruited. When the number of B cells entering an adult lymphoid organ increases, for example during an immune response, the feedback loop may help ensure that the follicular compartment can expand and accommodate the increased numbers of B cells. BLC-mediated B-cell recruitment and LT 1 2 induction may also be involved in Peyer's patch organogenesis, because, in addition to requiring BLC and LT 1 2, these structures depend strongly on B cells 22 . In support of this, ectopic expression of BLC induces B-cell-dependent and LT 1 2-dependent lymphoid neogenesis 23 .

A very early step in development of Peyer's patches and lymph nodes is the local accumulation of CD3 - IL7R + cells that express LT 1 2 (refs 24 , 25 ,26 ). As some of these cells express CXCR5 (refs 25 , 26 ), we propose that BLC functions in lymph node development by recruiting CXCR5 + CD3 - IL7R + cells and inducing them to upregulate LT 1 2. Finally, the finding that germinal centre B cells express LT 1 2 and that this expression is independent of BLC, together with the detection of CR1 + FDCs in germinal centres of BLC -/- mice, reveals that the pathway controlling FDC development in germinal centres is distinct from the BLC-mediated feedback loop operating in primary follicles.

Methods Generation of BLC -/- mice A 10-kilobase (kb) Eco RI fragment containing the blc gene was isolated from a 129 mouse genomic DNA library (Genome Systems). A targeting vector was constructed in which nucleotides 18�116 of the second exon of blc were replaced with an in-frame stop codon followed by a Mengo virus internal ribosomal entry site, the gene encoding enhanced green fluorescent protein and, in reverse orientation, a lox P-flanked neomycin resistance gene ( neo r ). The linearized construct was electroporated into JM1 129 mouse ES cells (a gift from N. Killeen). G418 (200 µg ml -1 ) resistant colonies were screened by Southern hybridization of Eco RI-digested genomic DNA to a flanking 1.0-kb probe. Targeted clones were injected into C57BL/6 (B6) blastocysts. Chimaeric males were mated to B6 females, and germline transmission of the targeted allele was detected by Southern blot and by PCR using the primers: 5'-CGTCTATGTTCTTTGTCCAATGGG-3' (sense); 5'-CTTACACAACTTCAGTTTTGGGGC-3' (antisense; wild-type); and 5'-ACCTTGTATTCCTTTGTCGAGAGG-3' (antisense; mutant). Homozygous mutant mice were born at Mendelian ratios, were fertile and appeared healthy. Most mice used in this study were mixed 129 and B6 strain background carrying the original targeted allele, although the phenotype of these mice was indistinguishable from mice homozygous for a neo r -deleted allele. Northern blot analysis was done as described5 using probes specific for blc exon 2 or EF-1 . BLC protein was detected in heparin-sepharose precipitates of 1% NP40 spleen homogenates by western blotting with polyclonal goat anti-mouse BLC (R&D Systems). B6 wild-type mice and some of the LT -/- mice (on a mixed 129/B6 background 27 ) were from Jackson Labs. Other LT -/- mice and the LT -/- mice were from a B6 colony 5 , 28 . Ig-transgenic mice were of the MD4 line 15 . To induce germinal centres, mice were injected intraperitoneally with 100 µg of TNP- or NP-chicken -globulin (Biosearch Technologies) diluted in 200 µl PBS and mixed with the Ribi Adjuvant System (Ribi Immunochem Research).