Establishment of Lymphotoxin β Receptor Signaling-Dependent Cell Lines with Follicular Dendritic Cell Phenotypes from Mouse Lymph Nodes1

Follicular dendritic cells (FDCs) have been shown to play a crucial role in the positive selection of high-affinity B cells that are generated by somatic hypermutation in germinal center (GC). Because of technical difficulties in preparing and maintaining pure FDCs, a role for FDCs in this complicated process has not been fully elucidated. In this study, we established a cell line designated as pFL that retained major FDC phenotypes from a three-dimensional culture of mouse lymph node cells. pFL cells proliferated slowly in response to an agonistic anti-lymphotoxin β receptor mAb and TNF-α. A more rapidly growing clone, named FL-Y, with similar requirements for growth was isolated from a long-term culture of pFL. Analysis of surface markers in these two cell lines by immunostaining, flow cytometry, and DNA microarray revealed the expression of genes, including those of CD21, FcγRIIB, lymphotoxin β receptor, ICAM-1, VCAM-1, IL-6, and C4, which have been shown to be characteristic of FDCs. In addition, B cell-activating factor was expressed in these two cell lines. At the pFL or FL-Y:B cell ratio of 1:100, the cell lines markedly sustained B cell survival and Ab production during 2 wk of culture, while most B cells collapsed within 1 wk in the absence of the FDC-like cells. Interestingly, expression of typical GC markers, Fas and GL-7, was notably augmented in B cells that were cocultured with Th cells on these two cell lines. Thus, pFL and FL-Y cells may be useful for providing insight into the functional role for FDCs in GC.

G erminal centers (GCs) 3 that are formed after immunization in B cell follicles in the secondary lymphoid tissues play a pivotal role in generating high-affinity Abs and memory B cells (1)(2)(3)(4). In this study, Ag-stimulated B cells differentiate into centroblasts and diversify their BCRs by somatic hypermutation of Ig genes. This is followed by positive selection of B cells that acquired mutated high-affinity BCR, thus resulting in an increase in the affinity of secreted Abs for an inducing Ag, a process termed affinity maturation (2, 4 -7). It has been reported that the network of follicular dendritic cells (FDCs) present in GCs plays a critical role in positively selecting high-affinity B cells (8 -12). A variety of genetic defects leading to blockade of FDC development has been shown to impair the spatial segregation of T and B cell areas in lymphoid tissues and the formation of GCs after immunization (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23). For instance, mice whose lymphotoxin (LT) ␣ gene or LT␤R gene was disrupted did not develop normal FDC network, nor formed GCs, thus resulting in impaired isotype switching and affinity maturation (13)(14)(15)(16)(17)24). FDCs have been shown essential for proper integration of the GC architecture, which may provide a microenvironment for efficient processing of GC B cells (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24).
It has been widely believed that the criterion for the positive selection of GC B cells is the binding avidity to immune complexes (ICs) that are trapped on the surface of FDCs (25)(26)(27). However, controversies have been raised recently as to whether ICs retained on FDCs are essential as a ligand for the positive selection of GC B cells, because affinity maturation still occurred in the mutant mice that do not produce soluble Abs, and thus deposit no significant ICs on FDCs (28,29). Many issues have remained to be elucidated: as for molecular mechanisms underlying how FDCs can determine the life or death of GC B cells. These ambiguities are mainly due to the difficulty in isolating and maintaining pure FDCs reproducibly to carry out in vitro experiments. FDC-like human cell lines, including HK, have been established to overcome these problems (30 -33). Although these cell lines have been shown to retain a variety of FDC functions, some respects of them, including expression of surface markers, were different from those of primary FDCs. In previous in vitro experiments using primary FDCs that were isolated from murine lymphoid tissues, high FDC:B cell ratios, such as 1:3, have been used to observe augmenting effects of FDCs on the differentiation of B cells (25)(26)(27)34). These experimental conditions might not reproduce in vivo situations accurately because it has been pointed out that the ratio may be much lower (1:100 or lower) in GCs (29).
To investigate the functional role for FDCs in eliciting GC reaction under more physiological conditions, we planned to generate a novel cell line that retains major FDC phenotypes from mouse lymphoid tissues. For this purpose, we took advantage of three-dimensional culture of mouse lymph node (LN) cells that were included in collagen gel matrix, and succeeded in growing a FDC-like cell line named pFL. In addition, a more stably growing clone designated as FL-Y was isolated from a long-term culture of pFL. In this study, we characterized pFL and FL-Y cells, focusing on their ability to support survival and differentiation of B cells in vitro. Results suggest that these two cell lines provide useful means to gain insight into biological functions of FDCs.

Experimental animals
Male BALB/c mice were purchased from Charles River Laboratories. Quasimonoclonal mice (H-2 b/b , 17.2.25V H /J H Ϫ , Ϫ / Ϫ ), whose major B cells bear L chains exclusively, and are specific for the 4-hydroxy-3nitrophenylacetyl (NP) hapten (35), were provided by M. Cascalho (Department of Surgery and Immunology, Mayo Medical School, Rochester, MN). (Quasimonoclonal ϫ BALB/c)F 1 (QCF 1 ) mice (H-2 b/d , 17.2.25V H / germline, ϩ / Ϫ ) were generated in our laboratory. Mice at 8 -12 wk of age were used throughout present experiments. All mice were treated in accordance with the guidelines approved by the Committee of Laboratory Animal Care, Okayama University.

Establishment of a FDC-like cell line, pFL, from mouse LN
BALB/c mice were immunized with 20 l of saline containing 20 g of TNP-KLH and 0.4 mg of alum in each hind footpad. Popliteal LNs were collected on day 7 after immunization, and fragmented into pieces with the diameter of ϳ1 mm. The LN fragments were embedded in the collagen gel matrix, as follows. The collagen gel solution was made by mixing 8 vol of ice-cold Cellmatrix I-A, 1 vol of 10ϫ MEM, and 1 vol of 0.08 N NaOH solution containing 200 mM HEPES. LN fragments prepared from three LNs were mixed with 0.15 ml of DMEM/RPMI and 0.3 ml of the collagen gel solution, and plated onto the center of each 35-mm well in a 6-well plate (Nalge Nunc International). To solidify the gel, the plate was kept at 37°C for 30 min under humidified atmosphere of 95% air and 5% CO 2 , followed by overlaying with 4 ml of DMEM/RPMI containing 20% (v/v) HS and 50 g/ml TNP-KLH. The LN fragments included in the gel matrix were cultured for 2 wk with changing one-half of the culture medium every 3-4 days. Then gel discs were removed from the wells, and digested with 0.5 mg/ml collagenase dissolved in 10% HS-containing DMEM/RPMI (HS-DMEM/RPMI) at 37°C for 10 min. The released LN fragments were subsequently treated with 50 g/ml DNase I in the same medium for 10 min at 37°C, followed by gently pipetting. These treatments led to the release of dissociated cells from the LN fragments. The released cell preparation contained adherent cells with dendrites that formed cluster with B220 ϩ cells, which were purified by panning with anti-B220-conjugated magnetic beads. Starting from 24 LNs, 5ϳ7 ϫ 10 4 clustered adherent cells were obtained at this stage. The cells were suspended at 1.25 ϫ 10 4 cells/ml in HS-DMEM/RPMI containing 1 ϫ 10 Ϫ5 M 2-ME, plated into a 48-well plate at 0.4 ml/well, and cultured with 10 g/ml TNP-KLH for 1 wk. Then the cells were stimulated with 4 ϫ 10 5 /well mitomycin C-treated LN cells (referred to as feeder cells, hereafter) in 0.4 ml of HS-DMEM/ RPMI containing 5 ng/ml each TNF-␣, GM-CSF, IL-4, and TGF-␤1 and 10 g/ml TNP-KLH. Feeder cells were prepared by treating LN cells (2 ϫ 10 7 cells/ml) from TNP-KLH-primed BALB/c mice with 25 g/ml mitomycin C for 30 min at 37°C. TGF-␤1 was usually added, as it has been reported to inhibit Fas-mediated apoptosis in a human FDC-like cell line (38). IL-4 was sometimes replaced by 10% (v/v) Con A-sup. The medium in each well containing feeder cells, indicated cytokines, and the Ag was renewed every 7 days. After 5ϳ6 passages, the adherent cells increased to the total number of ϳ10 6 (a 14ϳ17-fold increase). The resultant cell line was designated as pFL. The established pFL cells were routinely expanded in HS-DMEM/RPMI supplemented with feeder cells and cytokines indicated above, and could be stored frozen for Ͼ6 mo.

Isolation of a FDC-like clone from cultured pFL cells
To isolate a pFL-derived clone that can grow more rapidly, pFL cells were plated at 3 ϫ 10 4 cells/well into 24 wells in a 48-well plate. The cells were maintained in 10% (v/v) FCS-containing DMEM/RPMI (FCS-DMEM/ RPMI) in the presence of 5 ng/ml TNF-␣ without supplementing feeder cells and other cytokines. Two colonies developed after 30ϳ50 days of the culture. One of them that retained a variety of FDC phenotypes and grew rapidly in response to LT␤R stimulation was isolated, and designated as FL-Y. FL-Y cells were routinely maintained in the same culture medium as that used for cloning.

Culture of murine lymphocytes on pFL or FL-Y cells
pFL or FL-Y cells were seeded into a 48-well plate at 4 ϫ 10 3 cells/well and incubated for 24 h in 0.4 ml of FCS-DMEM/RPMI containing 5 ng/ml each TNF-␣, TGF-␤1, and GM-CSF and 10% Con A-sup. After removing the medium, 0.4 ml of the culture medium containing 1 ϫ 10 6 /ml QCF 1 B cells with or without 1 ϫ 10 3 /ml OVA-specific Th clone, DO11.10 (39), was added to each well that was coated with adherent pFL or FL-Y cells. The cells were cultured for 7ϳ13 days in the presence of 0.01ϳ1.0 g/ml NP-OVA. Thus, the FDC-like cell:B cell is 1:100 in these cultures. B cells were prepared by depleting T cells using anti-Thy-1.2-conjugated magnetic beads from nonadherent popliteal LN cells of QCF 1 mice that had been immunized with 20 g of NP-CGG and 0.4 mg of alum in each hind footpad. At indicated time points, lymphocytes were harvested and examined for the viability and GC marker expression in the B cells. Culture supernatants were assayed for the level of secreted anti-NP Abs. In some experiments, TNP-KLH-primed LN cells from BALB/c mice were cultured on pFL or FL-Y cells with or without 10 g/ml TNP-KLH in the same fashion.
Proliferation of QCF 1 B cells was assessed using the CFSE-labeled cells, as described (40). QCF 1 B cells were washed and suspended in PBS at 10 7 cells/ml, and labeled with 5 M CFSE at 37°C for 10 min. Excess CFSE was quenched with FCS and washed twice with MEM. The labeled cells were then cultured on pFL cells with DO.11.10 and 0.1 g/ml NP-OVA for 4 days, as described above. CFSE contents of cultured QCF 1 B cells were estimated by flow cytometry.
In some experiments, isolated GC B cells were cultured on pFL cells. LN cells from TNP-KLH-primed BALB/c mice were collected on day 7 after immunization. GC B cells were purified from the LN cells by sorting GL-7 ϩ B220 ϩ cells with FACS Aria (BD Biosciences). More than 95% of the resultant cells possessed GC marker. GC B cells thus obtained (1 ϫ 10 5 cells) were cultured in triplicate with or without pFL (8 ϫ 10 3 cells) in 48-well plates in the presence of 1 g/ml anti-CD40 and 10 ng/ml IL-4. Cultured cells were harvested and enumerated on day 4. to trap the Abs onto the solid phase (27,41). IgM and IgG Abs bound to the plates were assayed with HRP-conjugated goat IgG Abs to each class. Bound HRP activity was measured using 5 mM ABTS/5 mM H 2 O 2 .

Phenotypic analysis of pFL and FL-Y cells
pFL and FL-Y cells were seeded on Lab-Tek chamber slides (Nalge Nunc International) at 3ϳ5 ϫ 10 3 /well and incubated for 24 h. The slides were washed with PBS and fixed in 3% paraformaldehyde. Endogenous peroxidase activity was blocked by immersing the slides in 0.1% H 2 O 2 . After blocking with 1% BSA, the cells were labeled, respectively, with biotinylated anti-B220, anti-CD21, anti-C4, F4/80, or 2.4G2. After washing, the slides were reacted with streptavidin-HRP. When anti-C4 was used as the first Ab, the slide was labeled with biotinylated rabbit anti-rat IgG F(abЈ) 2 in combination with streptavidin-HRP. Bound HRP activity was visualized using 3-amino-9-ethyl-calbazole.
Surface markers on FL-Y cells were analyzed by flow cytometry using FACSCalibur and CellQuest software (BD Biosciences), as described previously (37). FL-Y cells were harvested from culture plates by washing with 0.02% EDTA-containing PBS (pH 8.0).

IC binding to pFL cells
ICs were prepared by incubating 200 g/ml NP-BSA-biotin with 500 g/ml mouse anti-NP IgG1 mAb (G1-5) overnight at 4°C. To detect IC deposition on pFL, the cells adherent to Lab-Tek chamber slides were treated with the 10-fold diluted IC solution for 4 h, washed, and fixed with 3% paraformaldehyde. After blocking endogenous HRP activity with 0.1% H 2 O 2 , bound ICs were labeled with streptavidin-HRP and visualized using 3-amino-9-ethyl-calbazole.

Microarray analysis
Total RNA samples were prepared from each 5 ϫ 10 5 cells of bone marrow-derived DCs (BMDCs), pFL, and FL-Y using TRIzol reagent (Invitro-gen Life Technologies), according to the manufacturer's indication. BMDCs were obtained from the culture of murine BM cells that were depleted of T and B cells and erythrocytes in the presence of each 10 ng/ml GM-CSF and IL-4, as described (42). cDNA synthesis, aRNA amplification, biotinylation, and fragmentation were performed with a One-Cycle Target Labeling Kit (Affymetrix). A total of 20 g of labeled samples was added to the hybridization mixture, and hybridized with Mouse Genome 430 2.0 GeneChips (Affymetrix) at 45°C for 16 h, as described in the manufacturer's instructions. Washing and streptavidin-PE staining were conducted using a GeneChip Fluidics Station (Affymetrix). Subsequently, the chips were scanned using a GeneChip Scanner 3000 (Affymetrix). At least two biological replicates per chip were analyzed. The intensity for each probe set was calculated using the MAS5 method of the GCOS software package (Affymetrix) at the default setting. Per chip normalization was performed using a median correction program in the GeneSpring software package (Agilent). The data of probe sets were excluded when the values were judged as absent by the GCOS program in at least one sample in the replicates. Functional annotation of genes on the Mouse Genome 430 2.0 array was performed using information from the NetAffyx analysis center (͗www.affymetrix.com/analysis/index.affyx͘). The raw data of the array analyses are available from the Gene Expression Omnibus of the National Center for Biotechnology Information (accession numbers of the data are GSM101418, GSM101419, GSM101420, GSM101421, GSM101422, and GSM101423).

RT-PCR and quantitative real-time PCR analyses
Total RNA samples were extracted from FL-Y cells with TRIzol, as described above. Each cDNA was generated using Superscript II (Invitrogen Life Technologies) reverse transcriptase and oligo(dT) nucleotides. The resulting cDNA was used in quantitative real-time PCR using iQ 5 Real Time PCR Detection System and iQ SYBR Green Supermix (Bio-Rad), as FIGURE 1. Establishment of an FDC-like cell line, pFL, from three-dimensional culture of mouse LN fragments. A, Scheme for three-dimensional culture of LN fragments. Popliteal LNs from TNP-KLH-immunized mice were fragmented and included in the collagen gel matrix. After culture in HS-containing medium for 14 days, the LN cells were released from the gel matrix by treating with collagenase/DNase I. B, Microscopic observation of adherent cells clustered with lymphocytes after being released from the gel matrix. C, Requirements for growth of the adherent cells that were harvested from the gel matrix. Adherent cells clustered with B cells were purified from the cells that were released from the three-dimensional culture by panning with anti-B220-conjugated magnetic beads. The purified cells were stimulated with or without feeder cells and TNP-KLH in the presence of indicated cytokines every 7 days. Where indicated, following cytokines were added: 4, IL-4; G, GM-CSF; T, TNF-␣; A, all of these three cytokines. Number of viable adherent cells was counted on day 17 of the culture. Each column represents the mean of duplicate wells. Five to six passages of the adherent cell culture under the optimal condition (see Materials and Methods) led to the establishment of a FDClike cell line, pFL. D, Adherent cell cultures were conducted in the same fashion as in C, except that feeder cells and TNP-KLH were replaced by an agonistic anti-LT␤R mAb (L) and/or TNF-␣ (T). Each column represents the mean Ϯ SD from triplicate wells. Significantly different from the control at p ϭ 0.023 ‫)ء(‬ and p ϭ 0.002 ‫.)ءء(‬ E, LT␣ and LT␤ expression in feeder cells. Feeder cells that were stimulated with TNP-KLH for 0 or 20 h were examined for the gene expression by RT-PCR. NIH 3T3 cells were used as a negative control. described by the manufacturer's protocol. PCR primers used for these analyses are as follows: LT␤R, 5Ј-GCCCCTGTGACATTGTGCT-3Ј and 5Ј-GGCAGAGTACAAGCCGCTC-3Ј; Fc␥RIIB, 5Ј-ATGGGAATCCTGC CGTTCCTA-3Ј and 5Ј-CCCAGCAGCAAGATTTAGCAC-3Ј; BAFF, 5Ј-CAGCGACACGCCGACTATAC-3Ј and 5Ј-TAGCCTGTTTGCCTCAC CAC-3Ј; LT␣, 5Ј-GTGCCTTTCTCCGACATGG-3Ј and 5Ј-GGTAGATGG GAGTGGGAATGG-3Ј; and LT␤, 5Ј-TCCAATGCTTCCAGGAATCT AGC-3Ј and 5Ј-GATCTGGTGTAGAATCC GCAG-3Ј (PrimerBank ͗http://pga.mgh.harvard.edu/primerbank͘). All real-time PCR were performed in triplicate.
LT␣ and LT␤ gene expression in feeder cells (TNP-KLH-primed LN cells) was analyzed by RT-PCR. LT␣, LT␤, and ␤-actin cDNA were amplified by AmpliTaq Gold DNA polymerase (Applied Biosystems) using specific primer pairs described above. PCR was conducted, as follows: denaturation at 94°C for 1 min, annealing at 58°C for 30 s, and extension at 72°C for 1 min. Amplified cDNA was visualized on polyacrylamide gel and stained with SYBR Green I Nucleic Acid Gel stain (Cambrex Bio Science Rockland).

Statistical analysis
Student's t test was used to compare results obtained from each triplicate experiment.

Establishment of a FDC-like cell line, pFL, from three-dimensional culture of mouse LN fragments
To establish a cell line with FDC phenotypes from murine lymphoid tissues, we took advantage of three-dimensional culture of LN fragments in collagen gel matrix using culture medium containing HS. The culture in collagen gel matrix may help maintain the LN architecture that will provide a microenvironment sustaining optimal growth and differentiation of constituent cells (43)(44)(45)(46). In addition, the use of HS instead of FCS has been reported to be beneficial in avoiding the outgrowth of fibroblasts that will otherwise overwhelm the growth of desired cells (46). Thus, we embedded popliteal LN fragments prepared from TNP-KLH-immunized mice in collagen gel matrix, and cultured in the medium containing HS and TNP-KLH ( Fig. 1A; for details, see Materials and Methods). After 2 wk, the cells included in the gel were released by the treatment with collagenase and DNase I, and further cultured with TNP-KLH for 5-7 days. Microscopic observation of the culture revealed the occurrence of many adherent cells that formed cluster with lymphocytes (Fig. 1B). Such clustered adherent cells were not obtained from the LN fragments cultured in the absence of the gel matrix.
When the cells released from the gel matrix were restimulated with TNP-KLH, B cells were found to proliferate actively concomitant with a slight increase in the number of B cell-binding adherent cells (data not shown). Because FDCs have been shown to develop in the B cell follicle in response to LT␤Rand TNFRmediated signals (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24), we hypothesized that if the adherent cells were FDC-like, the cells might be maintained and hopefully expanded in the culture when they were stimulated with LT␣1␤2 and TNF-␣. Thus, we tested whether the adherent cells recovered from the gel culture could proliferate when the cells were cultured with mitomycin C-treated TNP-KLH-primed LN cells (feeder cells) as a source of LT␣1␤2 and other growth factors. To investigate optimal conditions for growing FDC-like lines, the adherent cells clustered with B cells were first purified by panning with anti-B220-conjugated magnetic beads, and cultured with feeder cells in the presence of various cytokines. The purified adherent cells grew poorly in the absence of feeder cells. The growth was enhanced slightly by the addition of feeder cells (columns 1 and 2 in Fig. 1C), and further stimulated in the presence of TNP-KLH, the priming Ag of feeder cells (column 3 in Fig. 1C). The cell growth was further enhanced by the addition of IL-4, TNF-␣, or the combination of IL-4, GM-CSF, and TNF-␣ in the presence of both feeder cells and TNP-KLH (columns 4ϳ7 in Fig. 1C), but not in their absence (data not shown). Although GM-CSF alone did not significantly improve the cell growth, the cells developed more dendrites and formed clusters with lymphocytes more efficiently in its presence (data not shown). In cultured human FDCs, it has been reported that GM-CSF was found to augment the cell viability and ICAM expression (31), suggesting beneficial effects of this cytokine. Thus, we usually added GM-CSF to the culture. Interestingly, an agonistic anti-LT␤R mAb plus TNF-␣ partially replaced the effects of feeder cells (Fig. 1D). In addition, it was confirmed by RT-PCR that feeder cells expressed both LT␣ and LT␤ (Fig.  1E). Thus, it was suggested that feeder cells were effective, at least FIGURE 2. Phenotypic analysis of pFL cells by immunostaining. pFL cells were seeded on chamber slide and incubated for 24 h. A, Adherent cells were fixed, stained with each specific Ab, and finally visualized by bound HRP activity (red). Staining with F4/80 was the same as negative control (data not shown here). B, Binding of ICs on pFL cells. pFL cells on slides were incubated with none, NP-BSA-biotin alone, or ICs that were formed by mixing NP-BSA-biotin and mouse anti-NP IgG1. After washing, bound NP-BSA-biotin was visualized using streptavidin-HRP. The cells reacted with NP-BSA-biotin alone were shown as negative control. C, pFL or NIH 3T3 cells that adhered on chamber slides were incubated with TNP-KLH-primed LN cells and TNP-KLH for 5 days. After removing nonadherent cells, the slides were labeled with biotinylated anti-B220. Bound B220 ϩ cells (red) were visualized using streptavidin-HRP.
in part, as an LT␣1␤2 source (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24). Because the cell growth was enhanced most reproducibly by the combination of IL-4, GM-CSF, and TNF-␣ in the presence of feeder cells and TNP-KLH, the culture was continued with these supplements being replaced every 7 days. Starting from 24 popliteal LNs, 5ϳ7 ϫ 10 4 purified adherent cells were obtained after the three-dimensional culture, FIGURE 3. Enhanced viability of B cells that were cultured on pFL cells. QCF 1 B cells (1 ϫ 10 6 /ml) were cultured with or without 1 ϫ 10 3 /ml DO11.10 Th cells in 0.4 ml of culture medium in the presence or absence of 4 ϫ 10 3 pFL cells that had been adhered to each well. NP-OVA was added at indicated concentrations to each well and cultured for 13 days. On days 7 and 13, viability of B cells in each harvested sample was estimated by flow cytometry and microscopic observation. As almost no viable cells were found on day 13 in the pFL-free culture, the result was not shown. A, Analysis of the proportion of B cells after lymphocyte cultures in the presence or absence of pFL cells. Proportion of B cells in each sample was assessed by flow cytometry on the day indicated. A typical experiment in which B cells were cultured with Th cells and 0.1 g/ml NP-OVA is shown. Percentage of B cells that were gated in R3 was shown below each diagram. B, Effects of pFL cells and NP-OVA concentrations on the viability of B cells that were cultured with or without Th cells. Cells were harvested on days 7 and 13 of the culture, and examined for the viability of total cells and B cells, respectively. The results were presented as the mean (Expt. I) or the mean Ϯ SD (Expt. II) from triplicate wells, and expressed as percentage of initially added cells: f, total viable cells; Ⅺ, viable B cells. In Expt. II, ‫ء‬ and ‫ءء‬ indicate significant differences from each corresponding -pFL control at p Ͻ 0.05 and p Ͻ 0.01, respectively. The data are the representatives of four independent experiments. C, Effects of pFL cells on the proliferation of B cells. CFSE-labeled QCF 1 B cells were cultured with DO11.10 Th cells and 0.1 g/ml NP-OVA in the presence or absence of pFL for 4 days, as described above, followed by analysis with flow cytometry. B220 ϩ cells gated in R3 were estimated as B cells that entered the cell cycle.
which were expanded to ϳ10 6 cells after 5ϳ6 passages. The resulting cell line was designated as pFL, which could be stored frozen for Ͼ6 mo.

Analysis of phenotypic marker expression on pFL cells
pFL cells thus obtained were examined for the expression of surface markers to investigate whether the cell line retains characteristic phenotypes of FDCs. Immunostaining with each specific Ab revealed that pFL cells were positive for CD21, Fc␥RII/III, and C4, which also have been recognized as FDC-M2 (36), all of which have been reported to be expressed on primary FDCs ( Fig.  2A) (11,47). pFL cells were, however, negative for FDC-M1, a typical FDC marker (data not shown). More than 95% of cultured pFL cells were positive for CD21 and Fc␥RII/III, thus confirming the purity of the cell line. In contrast, the cells were negative for F4/80, which is a marker of the macrophage lineage (48).
NP-BSA was found to deposit on pFL cells when it was incubated with the cells in the presence of an anti-NP IgG1 mAb, but not in its absence, suggesting the ability of the cells to bind ICs (Fig. 2B). In addition, pFL cells formed clusters with B cells when TNP-KLH-primed LN cells were cultured on pFL cells (Fig. 2C). Collectively, these results strongly suggest that pFL cells retain intrinsic properties of FDCs.

Augmentation of cell viability, GC marker expression, and Ab production in B cells by pFL cells
FDCs have been reported to support the viability of GC B cells that are otherwise apt to undergo apoptosis in this microenvironment (9,11). Thus, we investigated whether pFL cells can support the viability of B cells during culture. To examine this, we used B cells from QCF 1 mice that bear the knockin 17.2.25 V H gene that is known to constitute NP-specific BCR when the encoded H chains associate with L chains and some L chains (49). Thus, 30 -40% of total B cells in the NP-CGG-primed LN cells showed specificity for NP, as assessed by the binding of NP-BSA (data not shown). In this experiment, B and T cells were cultured at low density (see Materials and Methods) to minimize supportive effects of autocrine trophic factors. When B cells from NP-CGGprimed QCF 1 mice were cultured with OVA-specific Th clone, DO11.10 in the presence of pFL cells, the proportion and the absolute number of viable B cells after 7 days' culture increased in the presence of a cognate Ag, NP-OVA, in a dose-dependent fashion (Fig. 3, A and B). Although data were not shown, B cell viability was not enhanced with 0.01-0.1 g/ml OVA instead of TNP-OVA, thus suggesting that cognate interaction between NPspecific B cells and the OVA-specific Th cells is responsible for the enhancement. Even in the absence of Th cells, the B cell viability was enhanced by pFL cells, but did not depend on the concentration of TNP-OVA. Thus, it is suggested that pFL cells can support B cells in an Ag-independent fashion in the absence of Th cells, and that BCR engagement alone does not contribute to the increase of B cell viability on pFL cells. In contrast, the viability of B cells was dramatically reduced in the pFL-free culture (Expts. I and II in Fig. 3B). In the presence of pFL, more B cells were found to enter the cell cycle than in its absence, as assessed by using CFSE-labeled B cells (Fig. 3C). As the proportion of B cells in the recovered cells from the pFL-containing culture was much higher than that from the pFL-free culture (Fig. 3, A and B), it is likely that pFL cells preferentially support the B cell viability. Even on day 13 of the culture, 20 -25% of the initial number of B cells were maintained in the culture containing Th and pFL cells (Fig. 3B), while no viable B cells were found in pFL-free cultures on day 13 (data not shown).
Consistent with the long-term maintenance of B cell viability by pFL cells, it was found that Ag-specific IgM and IgG production in QCF 1 B cells that were cocultured with DO11.10 Th cells was induced, and sustained from days 7 to 13 of the culture in the presence of pFL cells, while the Ab response in the absence of pFL cells was Ͻ1/15 of that elicited in their presence (Fig. 4). It should be noted that the B cell-supporting activities were clearly observed at the pFL:B cell ratio of 1:100 that may mimic the situation in GC, thus suggesting the strong competence of this cell line.
Furthermore, pFL cells markedly augmented expression of GC markers, GL-7, and Fas in B cells that were cultured with Th cell and Ag. On day 0 of the culture, no GL-7 ϩ Fas ϩ B cells were observed, while a significant proportion of B cells became positive for the GC markers after culture for 7 days in the presence of pFL cells, but not in their absence (Fig. 5A). It was found that the number of GL-7 ϩ Fas ϩ B cells markedly increased during culture strictly depending on pFL cells and the dose of NP-OVA (Fig. 5B). In addition, purified GC B cells that were stimulated with anti-CD40 plus IL-4 increased in number by ϳ4-fold after culture for 4 days only in the presence of pFL cells (Fig. 5C). GC B cells did not increase when anti-CD40 and IL-4 were not added to the culture (data not shown). Thus, pFL cells are considered to be potent in supporting GC B cell proliferation. The Ab response (Fig. 4) and GC marker expression (Fig. 5) were not induced in the absence of Th cells or NP-OVA (data not shown). Taken together, results show that pFL cells retain characteristic functions of FDCs that may be responsible for promoting GC reaction.
Isolation of a more rapidly growing clone from pFL cells As described above, we established culture conditions to prepare a FDC-like cell line, pFL, from the primary culture of LN fragments. Although the cell line thus induced is valuable to investigate FDC functions in vitro, the preparation procedure is a time-consuming and laborious process. Thus, we next tried to isolate a pFL-derived clone that can be proliferated more stably and rapidly. When pFL cells were maintained in the culture medium supplemented with TNF-␣ alone, two colonies were found to grow after 30 -50 days, one of which was named FL-Y (Fig. 6A). FL-Y cells grew slowly in the presence of TNF-␣ alone, but not in its absence. Interestingly, more rapid growth was induced in the presence of an agonistic anti-LT␤R mAb, which was further accelerated when TNF-␣ was supplemented in addition to anti-LT␤R (Fig. 6B). The decline of the growth curve in the anti-LT␤R/TNF-␣-stimulated cells on day 8 was due to overgrowth of the cells. If appropriately FIGURE 4. Long-term maintenance of Ab production by pFL cells. Culture supernatants were collected on days 7 and 13 from the coculture of QCF 1 B cells and DO11.10 cells that were conducted in Fig. 3. NP-specific IgM (Ⅺ) and IgG (f) in each sample were assayed as anti-NIP Abs (see Materials and Methods). Each column represents the mean Ϯ SD from triplicate wells. Significantly different from each control at p Ͻ 0.05 ‫)ء(‬ and p Ͻ 0.01 ‫.)ءء(‬ diluted, the cells could be maintained in the culture for more than several months. Although signals mediated by TNFR and/or LT␤R have been shown to be required for maturation and maintenance of FDCs in vivo (13-24), FDC-like cell lines whose growth is dependent on these stimuli have not been described. In addition, flow cytometric analysis revealed that FL-Y cells are positive for LT␤R, VCAM-1, Fc␥RII/III, CD21, and BAFF (Fig. 6C). Quantitative real-time RT-PCR confirmed the expression of LT␤R, Fc␥RIIb, and BAFF in FL-Y cells, which was apparently marginal in flow cytometric analysis (Fig. 6D), thus indicating that FL-Y cells retained major phenotypic markers of FDCs. In addition, it was found that BAFF and Fc␥RIIb transcripts were increased in response to LT␤R stimulation in FL-Y cells. BAFF and Fc␥RIIb, but not LT␤R, were expressed at higher levels in pFL cells than FL-Y cells (Fig. 6D). In contrast, FL-Y cells as well as pFL cells were negative for FDC-M1 (Fig. 6C). It is possible that FDC-M1 expression was unstable after the cells were transferred to in vitro

Characterization of gene expression profile in pFL and FL-Y cells by microarray analysis
To confirm the character of pFL and FL-Y cells as FDC-like lines, gene expression profile of these cell lines was examined in comparison with that of BMDCs by microarray analysis. In the analysis of FL-Y cells, unstimulated and anti-LT␤R-stimulated cell preparations were used. pFL and FL-Y cells were shown to share almost the same gene expression pattern, thus indicating that FL-Y was a direct descendant of pFL (Fig. 7A). However, several genes, including C4 and Cxcl13, were expressed at lower levels in FL-Y cells than pFL cells. Because the gene expression profile of BM-DCs was distinct from that of pFL or FL-Y lines, it was clearly shown that our cell lines do not belong to the DC lineage. In addition to analyses of cell surface markers (Figs. 2 and 6), the microarray analysis further supported that the cell lines were closely related to FDCs. For instance, consistent with previous reports (8,50), expression levels of class II Ags (H-2Aa and H-2Ea) were high in BMDCs, but much lower in pFL and FL-Y cells (Fig. 7, A and C). In contrast, among the genes expressed at higher levels in our FDC-like lines compared with BMDCs (listed in Fig. 7B), those marked with asterisks have been shown to be expressed in FDC preparations described in previous reports from other laboratories (51)(52)(53). These included the genes of CXCL12, CXCL13, procollagen type Ia, biglycan, VCAM-1, C4, CXCL1, connective tissue growth factor, insulin-like growth factor-binding protein 3, secreted acidic cysteine-rich glycoprotein, and prostacyclin synthase, thus further confirming the FDC-like characters of pFL and FL-Y cells. FDCs present in GCs have been shown to express Fc␥RIIb by immunohistochemical analysis (26). We revealed that pFL and FL-Y cells were reactive with 2.4G2 that recognizes both Fc␥RII and Fc␥RIII ( Figs. 2A and 6C). It was confirmed by microarray analysis and real-time RT-PCR that the Fc␥r2b gene was expressed in these two cell lines (Figs. 6D and  7A).

B cell-supporting activity of FL-Y cells
FL-Y cells were found to support B cells in terms of sustaining viability, GC marker induction, and differentiation to CD138 ϩ plasma cells as efficiently as the parental line, pFL (Fig. 8A). As assessed by flow cytometry, real-time RT-PCR, and DNA microarray analysis, both pFL and FL-Y cell lines were shown to express BAFF (Fig. 6, C and D, and Fig. 7, A and B). Immunohistochemical analysis of human tonsil sections has revealed recently BAFF expression in FDCs within GCs (54), but, to our knowledge, this is the first report describing that murine FDC-like cells express BAFF. Because enhancement of B cell viability by FL-Y cells during culture was significantly reduced by the addition of neutralizing anti-BAFF Abs, but not of negative control Abs (Fig. 8B), BAFF may be responsible, at least in part, for supporting B cells by FL-Y. Collectively, FL-Y cells may be useful as a FDC-like line to examine biological role for FDCs in GC reaction at cellular and molecular levels.

Discussion
Normal development of FDCs has been shown to be essential for the establishment of proper segregation of T/B cell areas in secondary lymphoid organs and the formation of GCs within the B cell follicle (8 -23). Although accumulating evidence suggests a pivotal role of FDCs in B cell differentiation and selection in GCs, molecular mechanisms underlying these complicated processes have not been fully elucidated. This is because FDCs have been difficult to examine due to problems in isolating cells to significant homogeneity and in maintaining or proliferating them in vitro. Thus, it is of great importance to develop a procedure for isolating primary FDCs more easily or to establish cell lines that retain major FDC functions as sufficiently as possible.
Primary FDCs prepared from lymphoid organs in irradiated mice have been used in many in vitro experiments reported to date (25)(26)(27)34). B cell viability, proliferation, and Ab production have been reported to be markedly enhanced when B cells were cocultured with the FDC preparations at high FDC:B cell ratio, such as 1:2ϳ1:5, which may be, however, too high compared with the ratio in GCs (1:100 or lower) (29). This may be because active FDCs were rare in these FDC preparations, or primary FDCs were not sufficiently functional under the culture conditions used. In contrast, a human cell line, HK, has been established from tonsil, and widely used as a FDC-like cell line (30). HK cells retain some properties characteristic of FDCs, including LT␣1␤2-induced upregulation of cell adhesion molecules (ICAM-1 and VCAM-1) and cytokines (IL-6, IL-15) (55), TNF-induced NF-B activation (56), and supporting GC B cell survival and proliferation (38,(55)(56)(57)(58)(59)(60)(61), while it has been reported that HK cells lost a representative surface marker of FDCs, CD21, shortly after isolation from human tonsil (30).
However, murine FDC-like cell lines that retain major physiological functions of FDCs have not been established. In the present work, we succeeded in establishing culture conditions to obtain FDC-like cell lines from mouse LN. The use of three-dimensional primary culture of mouse LN fragments in HS-containing culture medium is considered to be critical for successful isolation of the cell lines. It has been reported that a mouse stromal cell line that can support long-term hemopoiesis was established from mouse spleen fragments by the same procedure (46). pFL cells could be reproducibly established from the three-dimensional culture, and maintained in the culture without changing the major phenotypes for at least 3-4 mo. In addition, the cells could be stored frozen.
Furthermore, we isolated a more rapidly growing clone, FL-Y, from a long-term culture of pFL cells. Microarray analysis revealed that pFL and FL-Y cells expressed a panel of genes that have been reported to be expressed in some previous FDC preparations (51)(52)(53), thus confirming the FDC-like characters of these cell lines. pFL and FL-Y cells were positive for CD21 (Figs. 2 and 6). However, CD19 that is associated with CD21 in B cells was found to be negative in these cells in our microarray analysis (data not shown). Interestingly, FL-Y cells can be continuously expanded in the TNF-␣-supplemented culture (Fig. 6B). More rapid growth was induced by the addition of an agonistic anti-LT␤R or feeder cells. Although LT␣1␤2 and/or TNF-␣ signals have been shown to be critical for the maintenance of FDC network in vivo (13)(14)(15)(16), to our knowledge, this is the first report that showed the effectiveness of these stimuli in proliferating FDC lineage cells. Thus, it is likely that LT and TNF signals may be involved not only in the maintenance, but also in the proliferation of FDCs in vivo during development of FDC network.
It has been shown that LT␤Rand TNFR-mediated signals control the maintenance and gene expression in FDCs. For instance, IC-bearing, FDC-M1-positive cells rapidly disappeared from the B cell follicle in the mouse spleen after injection of LT␤R-Ig fusion protein that blocks LT␤R-mediated signal (13). In human FDClike cell lines, LT␣1␤2or TNF-␣-induced up-regulation of VCAM-1, ICAM-1, IL-6, and IL-15 has been observed (55). We also observed in the microarray analysis that expression of several genes, including, for instance, BAFF, Fc␥RIIB, and ICAM-1, was increased in response to anti-LT␤R stimulation in FL-Y cells (Fig.  7A), thus further confirming the responsiveness of this cell line to LT␤R-mediated signals.
Our FDC-like lines, pFL and FL-Y, showed strong supporting effects on the viability and Ab production of B cells when lymphocytes were cocultured with these FDC-like cells that were added to the culture at 1/100 of total B cells, thus suggesting that our cell lines retain more sufficient functions than primary FDCs described in previous reports (25)(26)(27)34). pFL and FL-Y cells preferentially augmented B cell maintenance and proliferation (Figs. 3 and 8A), while T cell proliferation was rather suppressed under the same culture conditions (our unpublished data). Although FDCs in human tonsil GCs have been shown to express BAFF as assessed by a specific mAb (54), BAFF expression in murine counterparts has not been reported. In the present work, we found that both pFL and FL-Y cells expressed BAFF (Figs. 6C and  7, A and B). Enhancement of B cell viability by pFL and FL-Y cells may be, at least in part, due to the expression of BAFF in these cells because the effect was significantly neutralized by anti-BAFF Abs (Fig. 8B). It has been proposed that soluble trophic factors released from FDCs, instead of ICs trapped on FDCs, might be more important in the selection process of high-affinity B cells (29). Our results imply that FDC-derived BAFF plays a role in the differentiation of B cells in GCs.
Another interesting observation is that expression of GL-7 and Fas that are characteristic markers in GC B cells was markedly enhanced in B cells that were cultured in the presence of our FDClike cell lines (Figs. 5 and 8). Thus, these FDC-like cell lines may be useful to examine functional roles of FDCs in GC reaction.