CX3CR1 Is Expressed by Human B Lymphocytes and Meditates CX3CL1 Driven Chemotaxis of Tonsil Centrocytes

Background Fractalkine/CX3CL1, a surface chemokine, binds to CX3CR1 expressed by different lymphocyte subsets. Since CX3CL1 has been detected in the germinal centres of secondary lymphoid tissue, in this study we have investigated CX3CR1 expression and function in human naïve, germinal centre and memory B cells isolated from tonsil or peripheral blood. Methodology/Principal Findings We demonstrate unambiguously that highly purified human B cells from tonsil and peripheral blood expressed CX3CR1 at mRNA and protein levels as assessed by quantitative PCR, flow cytometry and competition binding assays. In particular, naïve, germinal centre and memory B cells expressed CX3CR1 but only germinal centre B cells were attracted by soluble CX3CL1 in a transwell assay. CX3CL1 signalling in germinal centre B cells involved PI3K, Erk1/2, p38, and Src phosphorylation, as assessed by Western blot experiments. CX3CR1+ germinal centre B cells were devoid of centroblasts and enriched for centrocytes that migrated to soluble CX3CL1. ELISA assay showed that soluble CX3CL1 was secreted constitutively by follicular dendritic cells and T follicular helper cells, two cell populations homing in the germinal centre light zone as centrocytes. At variance with that observed in humans, soluble CX3CL1 did not attract spleen B cells from wild type mice. OVA immunized CX3CR1−/− or CX3CL1−/− mice showed significantly decreased specific IgG production compared to wild type mice. Conclusion/Significance We propose a model whereby human follicular dendritic cells and T follicular helper cells release in the light zone of germinal centre soluble CX3CL1 that attracts centrocytes. The functional implications of these results warrant further investigation.

Introduction CX 3 C chemokine ligand 1 (CX 3 CL1) is a multidomain molecule consisting of a chemokine domain linked to a transmembrane domain via an extended mucin-rich stalk, and of an intracellular domain [1]. CX 3 CL1, that exists as membrane-anchored and soluble forms, is constitutively expressed in many hematopoietic and non-hematopoietic tissues [2,3,4,5]. CX 3 C chemokine receptor 1 (CX 3 CR1), the exclusive CX 3 CL1 receptor, is a pertuxis toxin (PTX)-sensitive seven-transmembrane G protein-coupled receptor (GPCR) expressed on human NK cells, monocytes, Th1 CD4 + cells, CD8 + T cells and mast cells [6]. Two types of interactions between CX 3 CL1 and CX 3 CR1 have been reported, one occurring between the membrane anchored forms of CX 3 CL1 and CX 3 CR1, the other one between surface CX 3 CR1 + and soluble CX 3 CL1. Membrane-bound CX 3 CL1 induces firm adhesion of leukocytes under static and flow conditions without activating integrins [7], whereas soluble CX 3 CL1 released from the cell surface following proteolytic cleavage [1,3] induces chemotaxis of CX 3 CR1 + cells.
Previous studies failed to detect CX 3 CR1 expression in human B lymphocytes [6,8,9]. Since CX 3 CL1 is expressed in secondary lymphoid follicles [2], we reasoned that this chemokine could be involved in the local B cell trafficking and decided to re-investigate with different approaches the expression of CX 3 CR1 in the major B cell subsets from tonsil and peripheral blood. We show that indeed CX 3 CR1 is expressed by human naïve, memory and germinal center (GC) B cells, that are the only subset attracted by soluble CX 3 CL1 in chemotaxis assays.
The GC is the site where antigen activated naïve B cells migrate, proliferate and undergo class switch recombination, antibody gene diversification and affinity maturation [10]. GC have two distinct zones called dark and light zones based upon histological appearance. B cells in the dark zone, called centroblasts, proliferate and somatically hypermutate antibody variable genes, then move to the light zone where they are selected based on the affinity of the B cell receptor for antigen [10]. Recent studies have delineated two alternative models for B cell trafficking in the GC, i) the cyclic re-entry model, whereby B cells that have migrated from the dark to the light zone and have been selected subsequently return to the dark zone for further proliferation and, ii) the intrazonal recirculation model, whereby most B cells stay in dark or light zones and selection in these areas operates independently. This latter model keeps into account the low frequency of GC B cells recirculating from one zone to the other [11,12,13,14,15].
These mechanisms have been identified in mouse models [11,12,14,15,16], whereas less is known on human GC B cell trafficking [17,18]. This study adds a new piece of information on the latter issue.

Expression and Function of CX 3 CR1 in Human Tonsil B Cells
We first investigated CX 3 CR1 expression in highly purified human B cells from tonsil and peripheral blood by flow cytometry. Ten independent experiments demonstrated that B cells from both sources expressed CX 3 CR1 (tonsil B cells median 46%, range 35-70; median MRFI 5.2, range 4.0-8.0; blood B cells median 52%, range 36-59; median MRFI 6.0, range 4.0-7.0). In spite of a trend to a higher MRFI in blood vs tonsil B cells the difference was not statistically significant.
Representative histograms from tonsil and blood B cell fractions, and control cell lines (THP-1 and Raji, positive and negative controls, respectively) are shown in Fig. 1A. CX 3 CR1 mRNA expression was next evaluated by real time PCR in tonsil and blood B cells and compared with that in THP-1 cells. As shown in Fig. 1B, CX 3 CR1 mRNA was detected in tonsil B cells and, at higher amount, in blood B cells. Expression of CX 3 CR1 mRNA was never detected in the Raji Burkitt lymphoma cell line (Fig. 1B).
Ligand specificity of CX 3 CL1 binding to CX 3 CR1 was demonstrated by competition binding experiments. Displacement curves show that unlabeled CX 3 CL1 competed for binding of 125 I-CX 3 CL1 to tonsil and circulating B cells, as well as to THP-1 cell line (positive control), in a dose-dependent manner (Fig. 1C, left panel), showing up to 70% inhibition with 100 nM unlabeled CX 3 CL1. On the contrary, no competition for 125 I-CX 3 CL1 binding was observed in Raji cell line used as negative control. Finally, 100 nM unlabeled CXCL8, tested as negative control, did not compete for binding of 125 I-CX 3 CL1 to tonsil and circulating B cells, as well as to THP-1 and Raji cell lines (Fig. 1C right panel).
Taken together, the above experiments demonstrated that B cells expressed unambiguously CX 3 CR1 mRNA and protein.
Subsequent experiments were carried out with naïve B cells, memory B cells and B cells with a GC phenotype purified from tonsil as detailed in Material and Method. For the sake of brevity, from now onwards the latter cells will be referred to as GC B cells. Since spontaneous apoptosis of freshly isolated GC B cells was a matter of concern [10], we investigated the proportions of Annexin V + early apoptotic cells at time 0 and in 1 to 4h cultures using five different GC B cell suspensions. As shown in Fig. 2A, GC viability ranged from a maximum of 96% (freshly isolated cells) to a minimum of 76% (after 4 h culture).
In ten different experiments, a half of freshly isolated GC, naïve, and memory tonsil B lymphocytes expressed CX 3 CR1, as assessed by flow cytometry (Fig. 2B). Chemotaxis to rCX 3 CL1 was next investigated in a two hour assay. In five different experiments, dose-dependent chemotaxis of GC B cells was observed (Fig. 2C). 300 and 600 ng/ml rCX 3 CL1 increased significantly chemotaxis of GC B cells compared to medium alone (P = 0.043 for both concentrations) (Fig. 2C). By contrast, naïve and memory B cells were not attracted by rCX 3 CL1, but migrated to CXCL12 tested as control (Fig. 2C). CX 3 CR1 mediated signal transduction is inhibited by pretreatment with pertuxis toxin (PTX) [6]. We pretreated freshly isolated GC B cells with PTX and tested their chemotaxis to rCX 3 CL1. In three different experiments, GC B cell migration was completely abolished by PTX pre-treatment (Fig. 2D).

CX 3 CR1 Mediated Signal Transduction in Tonsil B Cells
In four different experiments, low density CD39 2 tonsil GC B cells [21,22] were incubated with or without rCX 3 CL1 and subjected to Western blot analysis of MAPK phosphorylation using antibodies to phosphorylated (p)-p38, p-Akt, p-Erk1 and p-Erk2. Low levels of phosphorylated proteins were detected in freshly isolated GC B cells (Fig. 3A). Following incubation with rCX 3 CL1, p-p38 expression peaked after 1 min (mean fold increase 2.360,6), p-Akt after 10 min (mean fold increase 2.160,8), p-ERK1 after 1 min (mean fold increase 2.160,4), and p-ERK2 after 1 min (mean fold increase 2.2 fold60.07). The bands corresponding to non-phosphorylated p38, Akt, Erk1 and Erk2 did not change in intensity following cell treatment with rCX 3 CL1 at any time tested (Fig. 3A). In the same cell populations, we studied Src activity by the in vitro kinase assay evaluating the activation of Hck, a member of Src kinases. The two bands of autophosphorylation and enolase phosphorylation peaked after 1 min rCX 3 CL1 treatment (mean fold increase 2.560.8 for the endogenous substrate Hck; 2.260.5 for the exogenous substrate enolase) (Fig. 3A).
In four different experiments, non-GC B cells incubated with soluble rCX 3 CL1 showed constitutive low level expression of p-Src, but not of p-p38, p-Akt or p-ERK 1/2 (Fig. 3B). p-Akt was induced de novo and peaked after 1 min exposure to rCX 3 CL1. p-p38 or p-ERK1/2 were not induced by CX 3 CL1 in non-GC B cells (Fig. 3B). In the latter cells, p-Src peaked after 1 min (mean fold increase 2.561.1 for the endogenous substrate Hck; 1.560. 35 for the exogenous substrate enolase) as assessed by in vitro kinase assay (Fig. 3B).
Next, chemotaxis of GC B cells to CX 3 CL1 was investigated following cell preincubation with the PI3K inhibitor LY294002, Akt inhibitor, the MEK inhibitor PD98059 that targets ERK1/2, the p38 inhibitor SB203580, the Src family inhibitor PP1 or medium. GC B cell chemotaxis was significantly inhibited by the PI3K inhibitor LY294002 (20 mM) (P = 0.009), the MEK inhibitor PD98059 (1 mM) (P = 0.03), the p38 inhibitor SB203580 (1 mM) (P = 0.009) and the Src family inhibitor PP1 (10 mM)) (P = 0.03). In contrast, the Akt inhibitor had no effect (Fig. 3C). These results demonstrate that PI3K and Src family kinases are involved in signal transduction initiated by CX 3 CL1 in human GC B cells.

Characterization of CX 3 CR1 + GC B Cells
In subsequent experiments, low density CD39 2 tonsil GC B cells were characterized immunophenotypically by double staining for CX 3 CR1 and a panel of GC B cell related markers.
Morphometric analysis of cytospins from freshly isolated CX 3 CR1 + and CX 3 CR1 2 GC B cells showed that the latter cells were significantly larger than the former (mean diameter 6 SD = 7.460.97 and 6.060.58 micron, respectively; P = 0.003) (Supplemental Fig. 1A).
Next, CX 3 CR1 + and CX 3 CR1 2 GC B cells were cultured 16 h in the presence or absence of rCD40L and IL-4 and stained with Annexin V. Both cell fractions underwent spontaneous apoptosis and were rescued by incubation with rCD40L (Supplemental Fig. 1B). In contrast, CX 3 CL1 did not affect survival of CD40L stimulated or unstimulated CX 3 CR1 + GC B cells in 16 h cultures.
V H 5(D)J H -m rearrangement sequences were next evaluated in CX 3 CR1 + and CX 3 CR1 2 tonsil GC B cell subsets. IgVH5 gene family rearrangements were chosen because i) IgM was the major isotype expressed within CX 3 CR1 + GC B cells, and ii) the VH5 family is composed of a relatively small number of members (and allelic variants) thus allowing an extremely reliable mutation analysis [23,24].
We compared 104 molecular clones expressed in CX 3 CR1 + GC B cells from 5 subjects to 63 molecular clones expressed in CX 3 CR1 2 GC B lymphocytes from 3 individuals. In addition, 34 molecular clones from CD10 2 , CD27 2 naive B cells of 3 out of the 5 subjects tested above were analyzed as controls. In both CX 3 CR1 + and CX 3 CR1 2 GC B cells the majority of V H sequences showed #1% mutations (#3 nucleotide substitutions). In particular, 67/104 (64.4%) from CX 3 CR1 + GC B cells and 42/ 63 (66.6%) clones from CX 3 CR1 2 GC B lymphocytes displayed ,1% deviation from the V H germline sequence. In V H 5 sequences from CD10 2 , CD27 2 naive B cells, only one clone out of the 34 analyzed showed .3 mutations (Supplemental Table 1).
The number of somatic mutations per clone was similar in CX 3 CR1 + and CX 3 CR1 2 GC B cell subsets. A total of 281 mutations were observed in the 104 VH5 sequences from CX 3 CR1 + GC B cells (2.7 mutations per V H sequence) compared to 189 mutations detected in the 63 clones from CX 3 CR1 2 GC B cells (3.0 mutations per V H sequence) (Fig. 4B). Fifteen nucleotide substitutions were found in the 34 molecular clones from CD10 2 , CD27 2 naïve B cells (0.4 mutations per V H sequence) (Fig. 4B).
Antigen selection, as determined by significant accumulation of R mutations in the complementary determining region (CDR) regions and/or significant preservation of framework (FR) amino acid sequences, was observed in similar proportions of mutated V H sequences both in the CX 3 CR1 + and CX 3 CR1 2 GC B cell subsets (8/35 and 4/19 clones respectively) (Supplemental Table 1).
Clonal relatedness (but not intraclonal diversification) was observed in 2 out of the 3 groups of CX 3 CR1 2 GC B cell molecular clones, whereas it was not detected in any group of  CX 3 CR1 + GC B cell molecular clones. Finally, no clonal relatedness between CX 3 CR1 + and CX 3 CR1 2 GC B cell clones from the same subjects was observed.

Immunophenotypic and Functional Dissection of CX 3 CR1 + GC B Cells
The immunophenotypic profile of CX 3 CR1 + GC B cells indicated that these cells expressed CD23 and CD27 (Fig. 4A), that are detected in mutually exclusive B cell subsets [25,26]. To gain further insight into this issue, CX 3 CR1 + GC B cells were positively selected from low density CD39 2 GC B cells by immunomagnetic beads and double stained with CD23 and CD27 mAbs. Fig. 5A, upper panel, shows that CD23 and CD27 mAbs marked two discrete CX 3 CR1 + GC B cell subsets.
Next, CX 3 CR1 + GC B cells isolated as above were double stained with CD23 or CD27 mAbs in combination with a panel of mAbs to various GC B cell related markers. CX 3 CR1 + GC B cells isolated as above were then subjected to chemotaxis to CX 3 CL1 and migrated cells were characterized by double staining for CD23 and CD27. Fig. 5B shows that migrated cells were comprised of CD27 + cells only.
In order to characterize more extensively the immunophenotype of GC B cells attracted by CX 3 CL1, low density CD39 2 GC B cells were subjected to CX 3 CL1-driven chemotaxis and migrated cells were subsequently stained with a panel of mAbs to GC B cell related markers. As shown in Fig. 5C, most of the GC B cells attracted by CX 3 CL1 had a CD27 + , Bcl-2 + , IgG + , CD77 2 , CD44 2 , CD23 2 immunophenotype, approximately a half expressed IgM and CD95, a minority of cells was ki67 + . These experiments demonstrate that the GC B cells attracted by CX 3 CL1 displayed the immunophenotypic features of centrocytes.

Expression of CX 3 CL1 in Tonsil Cell Populations
CX 3 CL1 expression was investigated in tonsil tissue sections by immunohistochemistry. As shown in Fig. 6A (one representative experiment out of three), CX 3 CL1 was expressed in the GC, the follicular mantle (FM) and the subepithelial area (SE).
Next, GC B cells, non-GC B cells and CD4 + , CXCR5 + T FH cells were tested for production of soluble CX 3 CL1 in 24 h culture supernatants. GC B cells were cultured with or without rCD40L, non-GC B cells with or without anti-Ig antibodies plus rCD40L, and T FH cells with or without PHA plus IL-2 [27]. As shown in Fig. 6B, GC B cells did not release CX 3 CL1 irrespective of the culture conditions, non-GC B cells produced constitutively soluble CX 3 CL1 and such production increased significantly (P = 0.028) following stimulation, while the amounts of CX 3 CL1 released by unstimulated and stimulated follicular helper T cells were similar.
Expression of CX 3 CL1 by FDC was subsequently investigated by immunohistochemical staining of formalin-fixed tonsil sections with CD21 and anti-CX 3 CL1 mAbs. FDC were found to express CX 3 CL1 (Fig. 6C).
We next investigated CX 3 CL1 release in 24 h culture supernatants from two human FDC cell lines, i.e. HK cells [28] and EF, generated in our lab [29]. Both cell lines were incubated without or with TNF or IFN-c, two physiological inducers of CX 3 CL1 production [30]. Fig. 6D shows that, in four different experiments, constitutive release of CX 3 CL1 by HK and EF cells was increased significantly upon stimulation with TNF for the HK cell line (P = 0.028) and with IFN-c for the EF cell line (P = 0.026).
Impaired Production of Specific IgG in CX 3 CR1 2 / 2 and CX 3 CL1 2 / 2 Immunized with OVA First, splenocytes from WT mice were double stained with the pan-B anti-B220 mAb and anti-CX 3 CR1 antiserum. As shown in the representative experiments in Fig. 7A, a large proportion of splenic B cells expressed CX 3 CR1.
Splenocytes from WT mice were next tested for chemotaxis to murine CX 3 CL1 and CXCL12 as control. Migrated B cells were enumerated by flow cytometry using anti-B220 mAb. As shown in Fig. 7B, no migration was detected in response to CX 3 CL1, whereas CXCL12 induced B cell chemotaxis.
These results indicate that impaired IgG production in CX 3 CR1 2 / 2 or CX 3 CL1 2 / 2 mice was related to deficient interaction between surface CX 3 CR1 and CX 3 CL1, but do not allow to establish whether this effect is attributable to direct ( B cell related) or indirect (accessory cell related) mechanisms.

Discussion
Human GC B cells migrate in vitro to anti-Ig antibodies [17] or CXCL12 [18] only following CD40 mAb and IL-4 mediated rescue from apoptosis. Murine GC B cells overexpressing Bcl-2 under control of the Ig m enhancer and therefore resistant to spontaneous apoptosis show different homing properties regulated by chemokine gradients in vivo. In particular, centroblasts express high levels of CXCR4 and home to the dark zone where the CXCR4 ligand CXCL12 is abundantly expressed. Centrocytes expressing high levels of CXCR5 accumulate in the light zone of the GC where the CXCR5 ligand CXCL13 is expressed at elevated concentrations [11]. Thus, current evidence indicates that only GC B cells rescued from apoptosis are enabled to migrate along chemokine gradients.
Here we show unambiguously that human B cells expressed CX 3 CR1 mRNA and protein as assessed by quantitative PCR, flow cytometry and competition binding. The discrepancy between these and previous results is likely related to differences in the antibodies and the techniques for mRNA analysis used [6,8,9]. CX 3 CR1 triggered GC B cell chemotaxis to CX 3 CL1, that was abolished by PTX pre-treatment, as expected from the well known PTX sensitivity of CX 3 CR1 [6]. Signal transduction induced by CX 3 CL1 in GC B cells was similar to that previously reported for other cell types [34,35]. In particular, p-PI3K, p-Erk1/2, p-p38 and p-Akt, as well as p-Src, were found to be up-regulated in CX 3 CL1 treated vs untreated GC B cells. Inhibition of PI3K, Erk1/2, p38 or Src, but not of Akt, suppressed CX 3 CL1 induced GC B cell migration. Thus, the PI3K and Src pathways integrated signals originating from either scaffold, as reported for other GPCR [36,37]. CX 3 CR1 + naïve and memory B cells were not attracted by CX 3 CL1 and showed a limited signalling pattern, i.e. p-Akt and p-Src were up-regulated, whereas p-PI3K, p-p38, and p-ERK1/2 were not. These results indicate that i) signalling and chemotaxis are dissociated in non-GC B cells, ii) lack of p-PI3K activation is consistent with the crucial role of this pathway in GC B cell migration, and iii) CX 3 CL1 driven signalling in non-GC B cells suggests functions of the chemokine alternative to chemotaxis that warrant investigation. CX 3 CR1 + GC B cells were almost devoid of CD77 + centroblasts and enriched for cells expressing the naïve B cell-related markers CD23, Bcl-2, sIgM and sIgD compared to CX 3 CR1counterparts. The proportions of Ki-67 + proliferating cells was similar in CX 3 CR1 + and CX 3 CR1 -GC B cells, challenging the tenet that proliferation in the GC is restricted to centroblasts [13].
Analysis of V H 5(D)J H -m rearrangement sequences demonstrated that the rate of somatic mutations was similar in CX 3 CR1 + and CX 3 CR1 2 GC B cells, a finding consistent with the similar expression of CD27 [25]. Likewise, the proportions of antigen selected CX 3 CR1 + and CX 3 CR1 2 GC B cells was superimposable. Lack of clonal relatedness between these cell populations may depend on the relatively limited size of the samples investigated. CX 3 CR1 + GC B cells contained two major subsets expressing CD23 or CD27. CD23 + cells contained CD44 + , IgM + and Bcl2 + cells, that may represent antigen-activated naïve B cells that have recently accessed the GC [38]. CD27 + cells were enriched for cells expressing IgG, IgM, Bcl2 and CD95 but not CD44. Only the CX 3 CR1 + , CD27 + GC B cell subset was attracted by CX 3 CL1. Further characterization of migrated GC B cells showed that they displayed a predominant CX 3 CR1 + , CD27 + , Bcl-2 + , IgG + , CD77 2 , CD44 2 , CD23 2 immunophenotype, consistent with that of centrocytes. Therefore, these experiments suggest that CX 3 CL1 participates in the control of centrocyte trafficking within the light zone of the GC.
In the GC, T FH cells promote centrocyte differentiation to Igsecreting cells [33], and contact between centrocytes and FDC results into positive selection of centrocytes expressing high affinity B cell receptors for antigen displayed on FDC [10]. Recent studies have suggested that FDC and T FH cells may collaborate in B cell selection. Thus, B cells that have captured, processed and presented high amounts of antigen to T cells would receive help from the latter cells at the expenses of B cells that have captured lower concentrations of antigen [13].
In this study, CX 3 CL1 was abundantly expressed in the GC, as assessed by immunohistochemistry [2]. Such expression was detected not only in the GC, but also in the FM and SE area. These findings are consistent with our unpublished results (Corcione A et al.) showing that a half of GC, naïve and memory B cells express constitutively CX 3 CL1 on the cell surface. CX 3 CL1 was released constitutively by T FH cells and FDC, whose production was increased by stimulation with TNF or IFNc [30], two cytokines produced in tonsils in vivo [39]. In this respect, our preliminary experiments show that T FH cells and FDC express constitutively ADAM10, a metalloprotease involved in constitutive shedding of CX 3 CL1 from the cell surface [40]. Since T FH cells and FDC reside in the light zone of GC, they may control through CX 3 CL1 release local in vivo migration of human CX 3 CR1 + centrocytes. Thus, CX 3 CL1 would cluster together T FH cells, FDC and centrocytes and promote differentiation of the latter cells to Ig-secreting cells.
Mouse B cells, although expressing CX 3 CR1, did not migrate to CX 3 CL1, indicating that surface rather than soluble CX 3 CL1 may be involved in cellular interactions leading to antigen specific IgG production.  . Impaired OVA specific IgG production in CX 3 CR1 2 / 2 or CX 3 CL1 2 / 2 mice. (A) Splenocytes from WT mice double stained with anti-CX 3 CR1 and B220 mAbs were analyzed by flow cytometry. Histograms from two different WT mice are shown. (B) Splenocytes from WT mice were tested for chemotaxis to murine rCX 3 CL1 or rCXCL12 (control). Migrated B cells were enumerated by flow cytometry using anti-B220 mAb. Results are median number of migrated cells, minimum and maximum values from four different experiments. (C) CX 3 CR1 2 / 2 , CX 3 CL1 2 / 2 or WT mice were immunized with OVA and tested at different dilutions for specific IgG production by ELISA. Results were expressed as median optical densities (OD), maximum and minimum values, from ten different experiments. Serum from two WT mice was titrated for OVA IgG antibodies (1: 200, 1:500, 1:1000, 1:10000, 1:50000, 1:100000 dilutions) and used in each tests as standard curve, whereas serum from mice before immunization was used as negative control. doi:10.1371/journal.pone.0008485.g007 In vivo experiments demonstrated that OVA-specific IgG production was significantly impaired in both CX 3 CR1 2 / 2 or CX 3 CL1 2 / 2 mice, in the absence of changes in microarchitecture, organization of lymphoid follicles and lymphocyte subset distribution. Further studies are required to understand what mechanisms underlie defective antibody production and where they operate, i.e. in secondary lymphoid organs or at periphery.
In conclusion we have identified a novel chemokine candidate to regulate human centrocyte trafficking in secondary lymphoid organs. Differences between human and mouse models highlight the difficulties at comparing the results obtained in these systems.

Ethic Statement
For research involving human participants, in our Institute this type of study did not require an ethic statement prior to the use because all tonsil samples are generally discarded after surgery and peripheral blood samples are from healthy donors of Blood Transfusion Center of the Institute. However, an informed consent, performed at the Medical Direction (Legal Medicine) of the G.Gaslini Institute, was asked to children parents or their legal guardians for tonsil samples. As well as, an informed consent was asked to donator for blood samples.  ); unconjugated CD77 and CD39 mAbs from Immunotech, Marseille, France; unconjugated rabbit anti-human CX 3 CR1 (Serotec) that detects an epitope starting from 175 to 189 amino acids; unconjugated CD3, CD56, CD68, CD10, CD27, and anti-human IgM mAbs from DAKO. Rat anti-mouse CD4-PE/Cy5.5, rat anti-mouse CXCR5-PE, rat anti-mouse CD19-FITC, rat anti-mouse CD3-FITC, and hamster anti-mouse CD128-FITC were from BioLegend (San Diego, CA, USA).
Cells were stained with fluorochrome conjugated or unconjugated antibodies followed by secondary reagents. Isotype and fluorochrome matched antibodies were tested as controls. For intracellular staining, cells were fixed, permeabilized, incubated with fluorochrome-conjugated mAbs and analyzed. Cells were run on a FACSCalibur (BD). 10 4 events were acquired and analyzed using the CellQuest software (BD).

Cell Isolation and Culture
Mononuclear cells (MNC) were isolated from tonsils and peripheral blood by Ficoll-Hypaque (Sigma Chemical Company, St. Louis, MO) and depleted of T lymphocytes by rosetting with sheep erythrocytes and of CD56 + and CD68 + cells by immunomagnetic beads. (Milteny Biotec Inc., Auburn, CA). These fractions contained .95% CD19 + B cells.
Tonsil B lymphocytes were incubated with CD10 mAb and separated by immunomagnetic beads into CD10 + GC [41] and CD10 2 non-GC B cells at 4uC to prevent spontaneous apoptosis. The latter cells were further separated into CD27 + memory and CD27 2 naïve B cells by immunomagnetic beads [25,42]. Alternatively, GC B cells were isolated from the low density 30/ 40 fraction of a Percoll gradient (Pharmacia, Uppsala, Sweden) [21,22] followed by depletion of CD39 + B cells. In some experiments, CD39 2 low density B cells, that contained .95% CD10 + GC B lymphocytes, were separated into CX 3 CR1 + and CX 3 CR1 2 cells. Annexin V staining (Bender Systems, Burlingame, CA, USA) was used to detect apoptosis.

Chemotaxis
Chemotaxis was investigated using 5 mm pore-size transwell plates (Costar, Cambridge, MA). Five 610 5 cells were dispensed in the upper chamber, and increasing concentrations of rCX 3 CL1/ Fractalkine (R&D System) or medium were added to the lower chamber. rCXCL12 (R&D System) was tested as control in some experiments at the concentration of 300 ng/ml [18].

Follicular Dendritic Cell Line Isolation and Culture
Two human FDC lines, HK cells [28] and EF cell generated in our lab [29], were cultured 24 h with or without TNF (10 mg/ml, R&D System) or IFN-c (10 mg/ml, R&D System) and supernatants stored at 280uC.

Signal Transduction Assays
Western blot analysis of MAPK phosphorylation and kinase assay of Hck activity (Src family) were investigated in GC and non-GC B cells and performed as reported [43]. Cells were incubated with or without 300 ng/ml rCX 3 CL1 for 1, 10, 20, 40, 60, and 90 min. The pellets were lysated with NP40 buffer. In the Western blot assay, lysates resolved by gel electrophoresis, transferred to nitrocellulose membranes, and incubated with specific anti p-p38, anti p-Akt, and anti p-ERK-1/2 mAbs (all from Santa Cruz Biotechnology). The membranes were stripped with Restore TM Western Blot Stripping Buffer (Pierce, Rockford, IL, USA) and reprobed with mAbs to detect total p38, Akt, and ERK-1/2 (Santa Cruz Biotechnology). After incubation with horse radish peroxidase (HRP) antibodies, blots were developed with Enhanced Chemiluminescence Detection. In kinase assay, lysates (300 mg) were pre-cleared with rabbit serum and Protein A-agarose immunoprecipitated overnight at 4uC with 1 mg Hck mAb (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Each immunoprecipitate sample was washed twice with lysis buffer and twice with kinase buffer (30 mM HEPES, pH 7.5, 10 mM MgCl 2 , 1 mM CaCl 2 , 5 mMCl 2 ). After washing, cells were incubated for 15 minutes at 30uC in 30 ml kinase buffer 10 mM ATP (Sigma), and 5 mCi c 32 P ATP (ICN Biomed) in presence or in absence of 10 mg of acid-denaturated enolase (muscle rabbit enolase, Sigma, St. Louis, MO, USA). The reaction was stopped by the addition of 4X Laemmli sample buffer and boiled for 5 min. Samples were separated by electrophoresis on 12% SDS-polyacrylamide gel. Kinase activity was detected by autoradiography.

Densitometric Analyses
Immage of immunoblots and kinase assay were scanned and quantification was carried out by IAS 2000 Image Analysis program from Deltasistemi (Latina, Italy). For MAPKs phosphorylation, values were normalized to total amounts of p38, Akt, and ERK-1/2. Ratios between values of the most intense band following rCX 3 CL1 treatment and those at time 0 are expressed as mean fold increase6SEM from four independent experiments.

Competition Binding Assay
The competition binding assay was performed according to Wang et al. [44]. Briefly, cells were suspended in binding buffer (PBS pH 7.4 containing 2% BSA) supplemented with 1 nM 125 I-CX 3 CL1 (Perkin Elmer, specific activity 2200/Ci mmol) in the absence or presence of 1, 10 and 100 nM cold CX 3 CL1 for 2 h at 4uC. Non specific binding was determined in parallel through cell incubation with 125 I-CX 3 CL1 in the presence of 100-fold excess unlabeled CX 3 CL1 and subtracted from total binding to yield specific binding. After incubation, medium was removed, cells were washed with binding buffer and radioactivity counted. Percentage of binding inhibition by unlabeled CX 3 CL1, calculated as ratio between cell-bound cpm in the presence of unlabeled ligand and cell-bound cpm in the absence of unlabeled ligand multiplied by 100, was used as a measure for competition between 125 I-labeled and unlabelled CX 3 CL1. Some experiments were performed using 100 nM cold CXCL8 (BioSource International, Flynn, Camarillo, Ca, USA) as negative control.
Amplification and Sequencing of the Ig V H 5(D)J Rearrangements Derived from Tonsil B Cell Subsets DNA-free RNA was extracted using the RNeasy mini kit (Qiagen S.p.A., Milan, Italy). One mg RNA was reverse-transcribed into cDNA using M-MLV reverse transcriptase (RT, Invitrogen S.R.L., Milano, Italy) primed by an oligo dT (16mer) primer. Reactions were carried out in 25 ml volume using 20 pmoles of primer, 200 U of RT at 42uC for 1 h. IgV H 5 rearrangements were determined by amplifying 2 ml of cDNA with primers and PCR conditions previously described [23,24]. Sequences were compared with Ig germline gene database at ImMunoGeneTics H http://imgt.cines.fr [45]. MacVector 6 software (Accelrys, S.Diego, CA, USA) was used for further sequence analyses.
Distribution of mutations among CDR and FR gene segments was evaluated by the Chang-Casali binomial distribution model [46]. A binomial probability model was used to evaluate whether the excess of replacement (R) mutations in CDR or their scarcity in FR was due to chance.
mRNA Analysis by Quantitative RT-PCR RNA extraction and reverse transcription were performed as above. CX 3 CR1-R FW 679 TTTTCTTGGCTTCCTACTCC, CX 3 CR1-R RV 809 CACGATGACCACCAGAAGG primers and a TaqMan MGB probe (59-YAK-TGTGGTTCTTGC-AGGAAAACAGC-BBQ-39) were designed. Quantification of CX 3 CR1 mRNA expression was performed on a Rotor-Gene 3000 instrument (Corbett Life Science, Sydney, Australia). Two independent experiments, with samples run in quadruplicate, were performed. As endogenous control, polymerase II polypeptide A (POLR2A) was co-amplified in the reaction using a FAM labeled probe. The relative expression of CX 3 CR1-R among the samples was calculated using the DDCt method as implemented in the Rotorgene 3000 software (Corbett Life Science). CX 3 CR1 expression was normalized among samples using the POLR2A expression level and was represented as fold of CX 3 CR1 expression relative to the THP-1 cell line, arbitrarily set at 1.

Immunohistochemistry
Immunohistochemical staining of tonsil tissue sections was performed using the Envision System HRP and alkaline phosphatase (AP) mouse (DAKO), as reported [47]. In double staining experiments, sections were incubated in DAKOCytomation Antigen Retrieval Solution for antigen retrieval, and subsequently with CD21 mAb (DAKO). The slides were then incubated with Dako Envision System HRP Mouse and peroxidase activity was detected with DAKO Liquid DAB Substrate Chromogen System. Then, the slides were washed, incubated overnight at 4Cu with anti-CX 3 CL1 mAb (PeproTech, Rocky Hill, NJ, USA) and treated with DAKO Envision System AP mouse (DAKO). AP activity was detected by incubating the sections for 20 min at room temperature with DAKO Liquid Fucsin Substrate Chromogen System (DAKO) and counterstained with Mayer's hematoxylin (Sigma; St. Louis, MO, USA).
Formalin fixed, paraffin embedded mouse tissue sections were incubated for 60 min at room temperature with a rat anti-mouse B220 (clone RA3-6B2, Serotec) or a rat anti-mouse CD3 (Abcam, Cambridge; UK). Sections were subsequently reacted for 30 min at room temperature with HRP conjugated rabbit polyclonal antibodies to rat IgG (Abcam). Controls were sections incubated in the absence of primary antibody or with irrelevant primary antibodies of the same isotype as test mAb.

Statistical Analysis
Data were reported in terms of medians, minimun and maximum values or quartiles. The Mann-Whitney U test was used to compare quantitative variables between two groups of observation. The one-way non parametric ANOVA (Kruskal Wallis test) was used for in vivo experiments to compare parameters in more than two groups of data. The Mann-Whitney U test with the Bonferroni's correction was used as a posterior test. A P value of less than 0.05 was considered statistically significant. Statistical analyses were performed using Graph Pad Prism 3 software and the statistical package ''Statistica 6''. Figure S1 Morphological and functional characterization of CX 3 CR1 + and CX 3 CR1 2 GC B cells Found at: doi:10.1371/journal.pone.0008485.s001 (3.67 MB DOC)