Lymphotoxin- Receptor-Dependent Genes in Lymph Node and Follicular Dendritic Cell Transcriptomes
http://www.100md.com
免疫学杂志 2005年第9期
Abstract
Affinity maturation and Ab class switches occur in lymphoid germinal centers (GCs), in which differentiation and maintenance depend on lymphotoxin (LT) signaling and include differentiation of follicular dendritic cells (FDCs). The events leading to FDC and GC maturation are poorly defined. Using several approaches of functional genomics, we enumerated transcripts affected in mice by suppressing LT receptor (LTR) signaling and/or overrepresented in FDC-enriched GC isolates. Protein expression analysis of 3 of 12 genes both enriched in FDCs and down-regulated by LTR signaling suppression validated them as FDC markers. Functional analysis of one of these three, clusterin, suggests a role as an FDC-derived trophic factor for GC B cells. Hence, the set of genes presented in this study includes markers emanating from LTR signaling and transcripts relevant to GC and FDC function.
Introduction
Signaling through the lymphotoxin (LT) 3 receptor (LTR) and the TNFR1 is crucial for organogenesis and maintenance of the structural integrity of secondary lymphoid organs (1, 2). These organs provide the compartmentalized microenvironment essential for mounting efficient humoral immune responses.
Mice deficient for LTR, TNFR1, or their ligands suffer from complex and partially overlapping pathologic phenotypes of the lymphoreticular system. TNF- mice and TNFR1-deficient (TNFR1–/–) mice develop lymph nodes, but TNFR1–/– mice show hypoplastic Peyer’s patches (3, 4, 5). LT-deficient (LT–/–) and LT-deficient (LT–/–) mice lack Peyer’s patches and most lymph nodes, except for mesenteric and cervical lymph nodes (6, 7). LTR–/– mice display the severest phenotype: they lack Peyer’s patches and all lymph nodes (1).
All these mutant mice develop a spleen, but show varying degrees of disturbance in white pulp compartmentalization, as well as impaired formation of germinal center (GC) and follicular dendritic cell (FDC) networks after challenge with antigenic substances. These deficits are reflected in reduced isotype switching (LT–/–, LT–/–, LTR–/–, TNF-–/–, and TNFR1–/–) (1, 3, 4, 7, 8), impaired affinity maturation (LTR–/–) (1), and compromised B cell memory (LT–/–, LTR–/–, and TNFR1–/–) (1, 8, 9). FDCs may represent a key player in promoting and regulating these events occurring during the GC reaction. In vitro experiments have shown that FDCs and FDC-conditioned medium exert chemotaxis on B cells and CD4+ T cells (10). FDCs can promote survival and proliferation of GC B cells (11, 12), and coculture with FDCs can enhance Ig secretion by B cells (13).
Given the complex nature of GCs, the identification of molecules that act downstream of LT and TNF signaling and that may control these phenomena has proven difficult. Also, the specific molecular contribution of FDCs to these processes has been elusive, as it has proven technically impossible to isolate mature FDCs to purity without the concomitant loss of markers that authenticate their mature state (FDC-M1) and functionality (CD21/CD35 and FDC-M2).
To gain insights into FDC-associated functions controlled by LTR signaling, many of which may contribute to the GC reaction, we pursued two complementary strategies aimed at defining the transcriptional profile associated with the presence of GC and FDC networks. First, we used high-density oligonucleotide microarrays to identify transcripts affected in vivo by administration of a soluble LTR-Ig, which blocks LTR signaling. Second, we screened for transcripts overrepresented in cell preparations enriched for FDCs relative to non-FDC splenic cell types by both microarrays and suppression subtractive hybridization. As might be expected, some of the transcripts were found to overlap in both populations. This allowed us to identify LTR signaling-dependent genes preferentially expressed in GCs in association with FDCs.
Materials and Methods
Enrichment of FDC clusters
FDCs were prepared by adapting a procedure of Thielen et al. (14). Eight- to 10-wk-old female C57BL/6 mice were injected with 200 μg of OVA in alum i.p. to increase the amount of FDCs present in secondary lymphoid tissues. Spleens were harvested 3 days after injection and cut in small pieces using a tissue chopper. The spleen pieces were digested two times for 20 min at 37°C in RPMI 1640 medium containing 1 mg/ml collagenase A (Boehringer Mannheim), 0.5 mg/ml dispase (type II; Boehringer Mannheim), 0.04 mg/ml DNase (type I; Boehringer Mannheim), and 0.4% BSA (Sigma-Aldrich). The supernatants of the two digestions were pooled and the cells were collected by centrifugation. The pelleted cells were resuspended in PBS containing 0.4% BSA and layered over FCS for four repeated sedimentations at 1 g at 4°C. The cells contained in the supernatant of the first sedimentation were collected and remaining FDCs were removed using FDC-M1 coupled magnetic beads (Dynal Biotech). The resulting cell fraction was frozen and used as the FDC-depleted sample for RNA isolation. To remove contaminating macrophages, the FDC-enriched cell clusters obtained after the four repetitive sedimentations were incubated in a plastic culture dish for 60 min with RPMI 1640 containing 10% FCS at 37°C, 5% CO2. Nonadherent cell clusters were frozen and used as the FDC-enriched cell fraction for immunofluorescence staining and RNA isolation.
Treatment of C57BL/6 mice with soluble LTR-Ig
Ten-wk-old female C57BL/6 mice were injected with 100 μg of soluble LTR-Ig i.v. As a control, identical mice were injected with the same volume of carrier only (PBS). The experimental mice were sacrificed 1, 2, 3, 27, and 35 days after injection; the control mice were sacrificed 3 days after administration of PBS. Mesenteric lymph nodes and spleens were harvested for further processing.
Preparation of labeled cRNA and microarray hybridization
Total RNA was extracted from all samples (FDC-enriched or -depleted cell fractions and mesenteric lymph nodes) using TRIzol (Invitrogen Life Technologies) according to the manufacturer’s instructions. RNA was subjected to a clean up step using RNeasy mini kit (Qiagen). RNA quality was assessed by agarose gel electrophoresis or Agilent 2100 Bioanalyzer. Processing of FDC-enriched and depleted cell fractions used: 15 μg of total RNA was subjected to cDNA synthesis using a cDNA Synthesis kit (cat. no. 11917-010; Invitrogen Life Technologies) and primer 5'-GGCCAGTGAATTGTAATACG ACTCACTATAGGGAGGCGG(dT)24–3'. Biotin-labeled cRNA was synthesized, using Enzo BioArray HighYield RNA transcript labeling kit (T7) (Enzo Biochem), cleaned with the RNeasy mini kit, quantified by spectrophotometry, and quality was assessed again by agarose gel electrophoresis or Agilent 2100 Bioanalyzer. Typical yields ranged from 60 to 90 μg of labeled cRNA. Labeled cRNA (15 μg) was fragmented in 40 mM Tris-acetate, 100 mM KOAc, 30 mM MgOAc, pH 8.1, at 95°C for 35 min. After fragmentation an aliquot of one representative of the replicates of each sample was hybridized to Affymetrix Test 3 arrays to determine the quality of the probes, reflected by the 3' to 5' ratio, and to test all buffers. The 3' to 5' ratio of the cRNAs used was always <2. The samples were then hybridized to Affymetrix MGU74vA2, Bv2, and Cv2 chips. All hybridizations were conducted for 16 h at 45°C at 60 rpm. After hybridization chips were washed and conjugated with streptavidin-PE according to the manufacturer’s instructions on the Affymetrix GeneChip Fluidics Station 450 with and scanned using the Affymetrix GS Scanner 2500 in conjunction with Affymetrix Microarray Suite 5.0 software. Processing of mesenteric lymph nodes samples showed: equal amounts of high quality total RNA from the individual mesenteric lymph node preparations were combined to obtain three individual pools of 15 μg of RNA per time point. Complementary DNA and RNA synthesis were conducted as described.
Labeled cRNA was hybridized to Affymetrix MOE430A and B chips.
Microarray data analysis
Intensity values of all chips were normalized with dChip software (www.dchip.org) (15) and, applying the model-based expression analysis algorithm (PM-only model), the expression values were calculated.
FDC-enriched (three biological replicates per chip set) and FDC-depleted (two biological replicates per chip set) cell samples were compared applying a 2-fold change (using the lower 90% confidence boundary of fold change) as the lower limit. In the case of the MGU74Bv2 chip set, two instead of three biological replicates for the FDC-enriched samples were included in the analysis due to contamination of one of the replicates. Only those genes that were flagged present in all FDC-enriched samples were considered. To increase the stringency a 4-fold change including the same criteria as mentioned was applied as the lower limit.
Significantly modulated genes by soluble LTR-Ig treatment were determined by significance analysis of microarrays (SAM) (16). A total of 100 random permutations of the data set were applied to calculate background changes and the false discovery rates used in SAM. Thresholds for significant changes were set to 0.38285 for analysis of MOE430A chips (lower cutoff 0.90741) and to 0.20248 for MOE430B chips (lower cutoff 0.90741). The random number seed was set to 1234567. Of the significantly changed genes, those considered for further analysis were flagged in two of three replicates of at least one sampling time point. Unsupervised hierarchical clustering of genes and correlation analysis was performed using the algorithms provided by the dChip software. For the correlation analysis of the genes represented on the MOE430A chips, cluster A was divided into two smaller subclusters exhibiting expression minima at days 3 or 27 postinjection. The average expression pattern of the two subclusters was used to calculate the correlation coefficients. For genes represented on the MOE430B chips the entire cluster A of this chip set was used for calculation of the correlation coefficients. Functional annotation of genes was performed using information provided by Affymetrix NetAffx (www.affymetrix.com/analysis/index.affx), LocusLink (www.ncbi.nlm.nih.gov/), and PubMed (www.ncbi.nlm.nih.gov/entrez/query.fcgi) databases. All microarray data sets are available at www.ncbi.nlm.nih.gov/geo. Accession numbers are GSE2123 (FDC-E vs FDC-D) and GSE2124 (soluble LTR-Ig-treated mesenteric lymph nodes).
Immunohistochemistry of cytospin FDC clusters
FDC clusters were isolated and cytospun onto glass slides. Slides were fixed with acetone on ice, air dried and stored frozen at –80°C until used for staining. For immunostaining slides were air dried, refixed with acetone, and redried. After rehydration in PBS, blocking was achieved with 2% goat serum in PBS. Then clusters were stained with FDC-M1 (1/50; BD Pharmingen) and FITC goat F(ab')2 anti-rat IgG (H+L) (1/200; BioSource International) and after thorough washing were fixed with 4% paraformaldehyde in PBS. Staining was then continued with CD4-Bio (1/100; R&D Systems), CD8-Bio (1/100; R&D Systems), CD68-Bio (1/200; Inotech), CD11c-Bio (1/100; BD Pharmingen), and streptavidin-Alexa 594 (1/500; Molecular Probes). All Abs were diluted in 2% goat serum in PBS. Photographs were taken with a Leica SP2 confocal laser scanning microscope.
Immunohistochemistry on cryosections
FDC-M1, CD21/CD35, peanut agglutinin (PNA), and B220 stainings for light microscopy were performed as previously described (17). For double staining with milk fat globule-epidermal growth factor 8 (MFG-E8) and prion protein (PrP), 5 μM cryosections were blocked with 1% goat serum and 0.5% BSA in PBS. Washes were performed with PBS and incubations with primary and secondary Abs with 1% goat serum and 0.5% BSA in PBS at room temperature. Abs used were: anti-MFG-E8 (1/1000; kind gift from Dr. S. Nagata, Osaka University Medical School, Osaka, Japan), anti-PrP XN (1/1000), B220 (1/400; BD Pharmingen), F4-80 (1/50; Inotech), FDC-M1 (1/50; BD Pharmingen), FDC-M2 (1/50; Immunokontakt), Alexa Fluor 594 goat anti-rat IgG (H+L) (1/100; Molecular Probes), FITC F(ab')2 goat anti-hamster IgG (1/100; Inotech), and Alexa 546 goat anti-rabbit IgG (H+L) (1/500; Molecular Probes).
For double staining with clusterin, endogenous peroxidase was quenched with iView inhibitor (Ventana). Then staining of clusterin was performed with the Tyramide Signal Amplification fluorescein system (TSA; PerkinElmer), using anti-clusterin Ab (1/500; Santa Cruz Biotechnology) as a primary and HRP donkey anti-goat IgG (1/200; Jackson ImmunoResearch Laboratories) as a secondary Ab according to the manufacturer’s instructions. Double staining with FDC-M1, B220, and F4-80 were performed as described. Photographs were taken with an Axiovert 200 M microscope (Zeiss).
Quantitative real-time PCR
Snap frozen mesenteric lymph nodes or spleens were homogenized in TRIzol reagent (Invitrogen Life Technologies) with a T18 Basic Disperser (Ika Works), and total RNA was extracted according to the manufacturer’s instructions. Before cDNA synthesis, residual genomic DNA was removed by the DNA-free kit (Ambion). Total RNA (5 μg) was converted into cDNA using the first strand cDNA synthesis kit (Amersham Pharmacia Biotech). Successful cDNA synthesis and contamination of total RNA with genomic DNA was tested by PCR with primers specific for Actb. Quantitative real-time PCR was performed using the SYBR Green PCR Master Mix (Applied Biosystems) on an ABI PRISM 7700 Sequence Detector (PerkinElmer). The following primer pairs were used: clusterin Clu, 5'-CCAGTTCCCAGACGTTGATT-3' and 5'-AGCAGGGATGAGGTGTTGAG-3'; cochlear Coch, 5'-CGTGCAAGGGGATCTAATGT-3' and 5'-GCTTACCTATCCACTTGAATGC-3'; ectonucleotide pyrophosphatase/phosphodiesterase 2 Enpp2, 5'-TGGCTTACGTGACATTGAGG-3' and 5'-AAATCCAAACCGGTGAGATG-3'; glycoprotein GpM6b, 5'-AAGAGCTGCACGGTGAGTTT-3' and 5'-GCACAAGCCACAATGAACAGG-3'; glycodelin A Gda, 5'-GTTGTGGGCCTAGTCCATGT-3' and 5'-GATACCGAGGGGAAGGATGT-3'; lysozyme Lyzs, 5'-TCCCTTGTCAGTCAGCACAG-3' and 5'-CTGTGCCCTCAGAAACCTTC-3'; Serpina1a, 5'-TCCAGATCCATATCCCCAGA-3' and 5'-AGGAACGGCTTCAAAGACTG-3'; glycosylation-dependent cell adhesion molecule 1 Glycam1, 5'-GCATTGATGGGCTCAGATTT-3' and 5'-ACTTCAACCCCAGGAAAACC-3'; RIKEN cDNA C030033F14Rik, 5'-AAAGTAAACACCGAAGGGACA-3' and 5'-TCGGGAAAGCAACAATCAAT3'; Mfge8, 5'-CCCTTCTCTCAGGCATTCTG-3' and 5'-AACCTGTCAACCACCCAGAG-3'; prion protein Prnp, 5'-GCTGGCCCTCTTTGTGACTA-3' and 5'-CTGGGCTTGTTCCACTGATT-3'; Cxcl13, 5'-TCGTGCCAAATGGTTACAAA-3' and 5'-ACAAGGATGTGGGTTGGGTA-3'; actin Actb, 5'-GACGGCCAGGTCATCACTAT-3' and 5'-ACATCTGCTGGAAGGTGGAC-3'.
Isolation of GC B cells
Eight-wk-old C57BL/6 mice were immunized as described for FDC enrichment, and spleens were isolated. RBC were lysed with 155 mM NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA. The spleen cells were then incubated with FITC anti-IgD Ab (1/100; BD Pharmingen) and GC B cells were thereafter negatively selected with magnetic beads directed against FITC, CD43, Thy1.2, Mac-1, and CD11c (all Miltenyi Biotec) using an autoMACS sorter (Miltenyi Biotec). Alternatively, GC B cells were positively selected by incubation with FITC anti-GL7 Ab (1/100; BD Pharmingen) and subsequent autoMACS sorting with anti-FITC magnetic beads. Purity of the GC B cells was assessed by FACS: 1 x 105 cells taken before and after MACS were resuspended in FACS buffer (20 mM EDTA, 2% FCS, 0.05% NaN3) and stained with the Abs B220-allophycocyanin (1/100; BD Pharmingen) and IgMb-PE (1/100; BD Pharmingen). FACS was performed on a four-color FACSCalibur flow cytometer (BD Biosciences). Purity of B220+IgMb+IgD– negatively selected GC B cells was >86% and purity of B220+GL7+ positively selected cells was >93% after MACS.
Cell viability assay
MACS purified GC B cells were cultured in IMDM medium (Sigma-Aldrich) supplemented with 1% penicillin/streptomycin (Invitrogen Life Technologies), 1.5% FCS (Sigma-Aldrich), and 2 mM L-glutamine (Invitrogen Life Technologies) in 96-well plates at a density of 1.5 x 105 cells/ml. Untreated cells were cultured in medium only, to treated cells the following reagents were added: anti-CD40 mAb (FGK 45.5, 10 μg/ml), human clusterin (50 μg/ml) purified as previously described (18), lysophosphatidic acid (LPA, 1 μM; Sigma-Aldrich), and human 1-antitrypsin (1 mg/ml; Sigma-Aldrich). Boiled human clusterin was incubated at 95°C for 15 min. Viability of the cells was assessed at the time points indicated by trypan blue exclusion in triplicates.
Construction of subtracted cDNA libraries using suppression subtractive hybridization (SSH) technique
FDC clusters and FDC-depleted cells isolated from 200 C57BL/6 mice as described were resuspended in buffer RLT of the RNeasy Midi kit (Qiagen) and homogenized with a T18 Basic Disperser (Ika Works). Total RNA was extracted according to the manufacturer’s instructions. Poly(A+) mRNA was isolated from the total RNA using Oligotex mRNA spin columns (Qiagen) following the manufacturer’s instructions. Total RNA was DNase treated using the DNA-free kit (Ambion).
Poly(A+) mRNA (0.3 μg) from FDC-enriched cells was used as the "tester" or the population whose up-regulated transcripts were to be identified, and an equal amount of poly(A+) mRNA from FDC-depleted cells served as the "driver" or the population whose transcripts served as a reference for cDNA subtraction (forward SSH library). The construction of the forward and reverse libraries was performed as per the SSH procedure using the PCR select cDNA subtraction kit (BD Clontech). Subtracted target cDNAs were ligated with the pGEM-T easy plasmid vector (Promega).
Inverted Northern blot
Plasmid (300 ng) was digested with 10 U EcoRI for 3 h at 37°C in 97-well plate in a GeneAmp PCR System 9700 PCR machine (Applied Biosystems). The digested product was divided into two halves and electrophoresed through a 1% agarose gel in duplicates. The DNA was then transferred to a Hybond-N+ membrane (Amersham Pharmacia Biotech). For hybridization 50 ng of FDC-E and FDC-D cDNA were labeled with [-32P]dCTP using random hexamer primers (Stratagene). After hybridization the membranes were exposed to BAS-MS imaging plates (FUJIFILM) overnight and the plates were scanned using a BAS-1800 II PhosphorImager (FUJIFILM). The signals were quantified with AIDA 2.41 image analyzer (Raytest). Hybridization signals obtained for each clone were normalized to the background signal originating from the vector in the same lane. Normalized signals for each clone were then compared between the two hybridizations.
In situ hybridization
Plasmids containing the differentially expressed cDNAs were linearized by restriction digest with NotI or SalI and isolated by gel purification (QIAquick gel purification kit). Digoxigenin (DIG)-labeled antisense and sense riboprobes were synthesized using the SP6/T7 DIG labeling kit (Roche). Thick cryosections (5 μM) of spleens were directly fixed with 4% paraformaldehyde in PBS on ice for 30 min and washed in PBS. Slides were then pretreated with 0.1 M HCl for 5 min and after washing acetylated with acetic anhydride for 5 min. After further washing with PBS the sections were dehydrated with ethanol (EtOH, 50, 70, 80, and 95%) and air-dried.
After prehybridization at 37°C for 3 h with prehybridization solution (50% formamide, 2.5x Denhardt’s, 50 mM EDTA, 50 mM Tris-HCl pH 7.6, 0.25 mg/ml yeast total RNA, 20 mM NaCl), the slides were incubated with denatured probe diluted in hybridization solution (50% formamide, 10% dextransulfate, 0.33 M NaCl, 20 mM Tris-HCl pH 7.5, 1 mM EDTA, 1x Denhardt’s, 0.5 mg/ml yeast tRNA) at 45–55°C in a humidified chamber overnight. The slides were then washed three times for 30 min each at the temperature used for hybridization with 0.2x SSC, 0.1x SSC, 50% formamide, 0.2x SSC, and then at room temperature with 0.2x SSC and 0.1 M Tris-HCl pH 7.5, 150 mM NaCl. Blocking was performed with blocking reagent (Roche) according to the manufacturer’s instructions. DIG labels were then detected with anti-DIG Fab (Roche) at a dilution of 1/100 and a color reaction with nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate.
Results
Transient blockade of LTR signaling in mesenteric lymph nodes by soluble LTR-Ig
LTR signaling can be transiently inhibited in adult mice by administration of the soluble decoy receptor, soluble LTR-Ig (19, 20, 21). We elected to analyze the transcriptional profile of mesenteric lymph nodes associated with the blockade of LTR signaling. We chose mesenteric lymph nodes because we expected that, based on their histologic structure, the contribution of GC and FDC to the total tissue transcriptome would be far greater than in spleen.
Although the effects of soluble LTR-Ig treatment on FDCs in splenic follicles have been previously documented (19), the consequences of this treatment for mesenteric lymph nodes have not been described. We first defined the time course of events occurring after soluble LTR-Ig injection on FDCs in mesenteric lymph nodes. A single i.v. dose of 100 μg of soluble LTR-Ig was administered to C57/BL6 mice, and general mesenteric lymph node follicular microarchitecture was analyzed by immunohistochemistry at defined time points following treatment. In contrast to carrier-only injections (PBS), FDC markers FDC-M1 and CD21/CD35 completely disappeared from the lymph follicles of mesenteric lymph nodes within 3 days. However, both markers reappeared 27–35 days later (Fig. 1). PNA staining of GC B cells disappeared with slightly delayed kinetics relative to the FDCs, as residual PNA positivity was still detectable at day 3 postinjection. Reappearance of PNA staining 27 and 35 days after administration of soluble LTR-Ig occurred in parallel with the FDC-marker FDC-M1 (Fig. 1). Treatment with soluble LTR-Ig had no morphologically discernible effects on the localization of non-FDC cells or on the expression of non-FDC markers within mesenteric lymph nodes during the period of analysis, as assessed by staining with the B cell marker B220 (Fig. 1), the macrophage marker F4-80, the dendritic cell marker CD11c, and the T cell markers CD4 and CD8 (data not shown).
FIGURE 1. Immunohistochemical analysis of the temporal effects of soluble LTR-Ig treatment on mesenteric lymph nodes. Stains were performed on cryostat sections of mesenteric lymph nodes harvested form C57BL/6 mice at the indicated time points post administration of soluble LTR-Ig or carrier only (PBS). Sections were counterstained by hematoxylin. Treatment with soluble LTR-Ig leads to complete disappearance of FDC markers (FDC-M1, CD21) and reactive GCs (PNA) form the follicular area within 3 days. Reformation of FDC networks and PNA-positive GC can be detected at days 27–35 postinjection. Localization and expression of markers specific for B cells (B220) were not affected by the treatment. Scale bar represents 200 μm.
Transcriptional patterns in mesenteric lymph nodes after transient soluble LTR-Ig-mediated FDC depletion
To identify gene expression patterns resulting in mesenteric lymph nodes from transient depletion of FDCs, we profiled the transcriptome of mesenteric lymph nodes harvested 1, 2, 3, 27, and 35 days after injection of soluble LTR-Ig or PBS. Three biological replicates per time point were hybridized to Affymetrix oligonucleotide microarray sets MOE430A and MOE430B. Genes whose expression was significantly changed by the blockade of LTR signaling were identified by SAM (16). When highly stringent thresholds for significant changes were set, SAM identified 235 probe sets with a false discovery rate of 0.35% for chip A and 145 probe sets with a false discovery rate of 3.11% for chip B.
Unsupervised hierarchical clustering (dCHIP software) (15) with these genes revealed four major expression patterns (Fig. 2A) Two of these, clusters A and B, contained genes reflecting short response latency to the blocking of the LTR signaling pathway. These transcripts showed an early change (decrease or increase) in expression with a peak of maximal change at day 2, 3, or 27, and returned to near-baseline levels at day 35 (Fig. 2A). The remaining two clusters, C and D, showed longer response latency with little change during the first two days of the treatment. Only at day 3 did the expression levels of the genes within these clusters start to be affected, with a maximal change at day 35 relative to control treatment.
FIGURE 2. Four predominant expression patterns are induced in mesenteric lymph node by soluble LTR-Ig treatment, which represent separate functional categories. A, Unsupervised hierarchical clustering of genes significantly changed (as detected by SAM) during the sampled time points represented on MOE430A and MOE430B chips. The normalized expression values of each gene were standardized to have a mean of 0 and a SD of 1. Each expression measurement displayed represents an average of three biological replicates. Rows represent the standardized expression levels during the time course of the individual probe sets and columns represent sampling time points. Red reflects expression levels above mean expression of a probe set across all time points, green reflects expression lower than the mean, black reflects mean expression. Gene clusters (A, B, C, and D) with similar expression patterns in response to soluble LTR-Ig treatment are indicated. The color-scale indicates the color code applied to display standardized expression values between –3.0 and 3.0. B, The genes of each of the four clusters (A, B, C, and D) were annotated to the functional categories: (I) membrane-bound surface molecules; (II) chemokines, cytokines, immunomodulatory and other secreted molecules; (III) immune effector molecules and apoptosis related; (IV) signal transduction related; (V) cytoskeleton related; (VI) cell cycle and stress; (VII) DNA binding, transcription, and translation; (VIII) metabolism; (IX) transport; (X) ECM components; (XI) other functions; and (XII) unknown functions. The percentage of the total genes represented by each functional category is indicated for the four clusters identified by hierarchical clustering.
The rapid response to the treatment of the genes contained in clusters A and B suggested that LTR signaling may transcriptionally regulate some of these genes. Hence they may reflect biological functions that are directly controlled by this pathway. In contrast, the longer response latency of the genes in clusters C and D could reflect secondary and/or indirect effects of the treatment. To further assess these questions we functionally annotated the known genes identified by SAM. For those genes with a known function, the four predominant clusters showed a bias toward separate functional categories (Fig. 2B). Genes of the clusters A and B were primarily annotated to the functional categories: (I) membrane-bound surface molecules; (II) chemokines, cytokines, immunomodulatory and other secreted molecules; (III) immune effector and apoptosis-related molecules; and (IV) signal transduction. Genes of the clusters C and D fell mostly into the categories (VII) DNA binding, transcription, translation; (VIII) metabolism; and (IX) transport (Fig. 2B). In cluster A the genes related to cell-cell adhesion (category I, 30.4%) and chemotaxis (category II, 13.0%) were particularly striking. Cluster B was enriched in surface receptors (category I, 25.0%) and cytokines and chemokines (33.3%). Cluster C showed an enrichment of genes encoding ribosomal subunits, proteasome subunits, or ubiquitin-conjugating enzymes, transcription factors, and proteins involved in energy metabolism, representing a total of 23.3% of all genes in this cluster. The genes of cluster D reflected similar functional categories as cluster C, but the majority (63.0%) could not be functionally annotated (Fig. 2B). These results suggested that clusters A and B might contain genes coding for specific biological functions attributable to LTR signaling, whereas the changes in expression of genes in clusters C and D reflected more general changes in basic cellular activity, possibly reflecting secondary effects downstream of the blockade of LTR signaling.
A collection of transcripts predominantly expressed by FDCs
As the expression pattern of cluster A correlated well with the kinetics of disappearance and reappearance of PNA+ GC and FDC networks from soluble LTR-Ig-treated mesenteric lymph node cryosections (Fig. 1), we analyzed this cluster in more detail. The highly stringent filtering criteria applied in SAM yielded an initial set of 21 genes for cluster A. To avoid excluding potentially valid candidates from further analysis, we chose to perform a correlation analysis using the dCHIP software. We calculated correlation coefficients for the association of each gene expression pattern with the average expression pattern of cluster A. We then selected genes with a correlation coefficient >0.9 and a greater than 0.85-fold down-regulation (p < 0.05) at the time of minimal expression relative to control treatment. We identified a total of 80 unique transcripts whose expression patterns correlated with cluster A, 42 of which could be annotated to known genes. Eleven of these genes are known to be expressed by stromal cells within secondary lymphoid organs, and most of them have been shown to be expressed by FDCs or to form part of subcellular structures, such as desmosomes that have been assigned to FDCs by electron microscopy (22). These include cell adhesion molecules (Coch, Dsc3, Dsg2, Glycam1, and Madcam1), chemokines (Cxcl13 and Ccl19), and cell surface molecules (BstI, Prnp, Cr2, and Plxnb1). Genes typically associated with the macrophage/monocyte lineage were also affected (Lyzs, Mfge8, and Faah), which is consistent with the expression of LTR on macrophages (23).
Selection of a set of GC/FDC-associated genes
To complement the first approach, and to identify transcripts that are overrepresented in cells associated with FDC networks in GCs, we purified FDC cell clusters and characterized their transcriptome relative to FDC-depleted spleen cells. Immunohistochemical analysis showed that the FDC clusters basically represented small functional units of GC light zone. Double staining with FDC-M1, B220, CD4, CD68, and CD11c revealed the presence of B cells, Th cells, macrophages, and dendritic cells adherent to FDCs in these clusters (Fig. 3A).
FIGURE 3. Immunofluorescence characterization of FDC-enriched cell clusters and functional annotation of genes overrepresented in such preparations. A, FDC-enriched cell clusters were cytospun onto glass slides and stained with FITC-labeled Abs directed against FDCs (FDC-M1) and costained with Alexa 594-coupled Abs for the presence of B cells (B220), Th cells (CD4), dendritic cells (CD11c), or macrophages (CD68). Most of the cells, adherent to FDCs, were B cells followed by Th cells. Only very few dendritic cells could be detected. This indicated that the major cell types taking part during a GC reaction are present in isolated FDC clusters. Scale bar represents 40 μm. B, The genes >4-fold overrepresented in FDC-enriched cell clusters were annotated to the 12 categories described in Fig 3. The percentage of the total genes represented by each functional category is indicated.
Labeled cRNA from FDC-enriched cell clusters and FDC-depleted spleen cells was hybridized to the entire chip sets (A, B, and C) of Affymetrix MGU74v2 chips. Using 2-fold changes as the lower limit and further filtering by selecting those probe sets with high significance (lower 90% confidence boundary of fold change), we found that 960 probe sets were significantly overrepresented in the FDC-enriched vs the FDC-depleted specimen. To reduce background noise, we increased the stringency of our criteria by selecting only genes with a >4-fold change for functional annotation. This filter was passed by 478 probe sets.
The majority of annotated genes encoded membrane proteins (category I, 18.2%) and secreted proteins, such as immunomodulatory cytokines and chemokines (category II, 7.1%; Fig. 3B). Among the membrane proteins, the cell adhesion molecules ICAM-1, VCAM-1, MCAM, GlyCAM-1, and E-selectin were overexpressed in FDC-enriched cell clusters. Jagged-1 and Delta-1, two ligands of the Notch-2 receptor, were overrepresented in the FDC associated over the FDC-depleted cell fraction. Within the class of secreted proteins (category II) we identified six chemokines that attract lymphocytes and macrophages (CCL2, CXCL12, and CXCL13), T cells and NK cells (CXCL10), and neutrophils (CXCL1 and CXCL7). Of the cytokines that were overrepresented in FDC-enriched cell clusters, IL-1 and IL-6 have been shown to act on B cells. Three other cytokines exert their effects mainly on the stromal compartment (platelet-derived growth factor (PDGF)A, connective tissue growth factor (CTGF), bone morphogenetic protein (BMP)-2). They can induce proliferation, migration or differentiation of mesenchymal cells and lead to production of extracellular matrix (ECM) components (24, 25, 26). In accordance with the up-regulation of these cytokines, there was a clear overrepresentation of genes encoding ECM components in the FDC clusters vs the remainder of spleen cells, such as Col1a1, Col1a2, Col3a1, Col4a1, Col4a2, Col5a2, Col14a1, Col18a1, Tnxb, Coch, Bgn, and Mglap (category X, 6.3%).
GC/FDC-associated genes identified by SSH
In a further experiment we used SSH to identify transcripts associated with GC and FDC networks. SSH is PCR-based and nonbiased: it allowed us to generate a library of cDNAs overrepresented in FDC-enriched vs FDC-depleted cell clusters without prior knowledge of their sequence. The FDC-enriched cDNA library was screened for FDC/GC-associated transcripts by inverted Northern blot. Seventeen clones from the library were found to be overexpressed in GC and FDC networks >2-fold and were sequenced (Fig. 4A). Six of 16 genes (Clu, Cyr61, Igfbp3, Mfge8, Serpina1, and Sparc) were also identified as overexpressed by the filters set in the corresponding microarray analysis (Fig. 4A). In situ hybridization with two of the candidates, Igfbp3 and Mfge8, revealed that they were both expressed in the follicular areas of spleen (Fig. 4B). Although Mfge8 expression appeared confined to the follicles, Igfbp3 showed additional intense labeling of marginal zone cells. These data confirm the concept that novel genes preferentially expressed in association with FDCs can be identified by comparing FDC-enriched and FDC-depleted splenic cell fractions.
FIGURE 4. Identification of genes overexpressed in association with FDCs by SSH. A, The 17 genes identified by SSH with their corresponding proteins are listed. Genes identified by both SSH and microarray analysis are shown in gray. LocusLink numbers and the fold overexpression in FDC-enriched (FDC-E) compared with FDC-depleted (FDC-D) specimens as determined by inverted Northern blot are indicated. B, In situ hybridization of two representative SSH candidates. Spleen cryosections from C57BL/6 mice were hybridized with DIG-labeled antisense riboprobes for Mfge8 and Igfbp3. To control for hybridization specificity sense riboprobes were incubated in parallel. The images are representative of two independent experiments.
A core set of LTR responsive transcripts in FDC-enriched cell clusters
By combining the two sample preparation approaches we selected a core set of genes that were not only associated with the presence of FDCs in enriched cell clusters but were also dependent on signaling through LTR (Fig. 5A). The intersection of the 80 genes associated with FDC disappearance upon blockade of LTR signaling and the 960 genes enriched in FDC cell clusters yielded a core set of 12 genes (Fig. 5B), three of which were also identified by SSH (Clu, Mfge8, and Serpina1). Two of the proteins encoded by the 12 genes are related to cell adhesion (GlyCAM-1 and MFG-E8), three have enzymatic activity (ENPP2, GDA, and LYSZ), one has chemotactic activity (CXCL13), one is related to apoptosis (clusterin), one to proteolysis (serpin a1a), one is an ECM component (cochlin), and for three the functions are not clear at the time (prion protein, glycoprotein m6b, RIKEN C030033F14Rik).
FIGURE 5. Identification and confirmation of genes blocked by soluble LTR-Ig treatment and overexpressed in FDC clusters. A, Venn diagram illustrating the approach used to identify genes blocked by soluble LTR-Ig treatment and overexpressed in FDC clusters. Twelve core set genes were found in common between 80 genes correlating with expression pattern A in mesenteric lymph nodes upon soluble LTR-Ig treatment (correlate with cluster A) and the 960 genes overrepresented in FDC-enriched cell clusters (FDC-E). B, The 12 core set genes with their corresponding proteins are listed. NCBI accession numbers are indicated in column 4 (footnote 1). Maximal fold down-regulation relative to control treatment as detected by microarray analysis is indicated in column 5 (footnote 2). Fold overrepresentation in FDC-enriched cell clusters (FDC-E) relative to FDC-depleted (FDC-D) spleen cells are indicated. C, Down-regulation of the 12 core set genes by soluble LTR-Ig treatment and enrichment in FDC-cell clusters was confirmed by quantitative real-time PCR. Soluble LTR-Ig treatment shown as values of fold-reduction relative to control treatment (PBS) at the day of maximal effect (soluble LTR-Ig treatment). FDC enrichment shown as values expressed as fold-enrichment in FDC-enriched cell clusters relative to non-FDC spleen cells. The experiment was done in triplicate and SEM is indicated with error bars.
We verified the down-regulation of the core set genes by blockade of LTR signaling and the overrepresentation of these genes in FDC cell clusters by quantitative real-time PCR. All 12 genes were consistently down-regulated in soluble LTR-Ig-treated relative to control treated mesenteric lymph nodes and were up-regulated in purified FDC clusters relative to non-FDC associated spleen cells (Fig. 5C). The fold down-regulation (between 2- and 20-fold) and fold overexpression (between 1.5- and 283-fold) measured by quantitative PCR reproduced the microarray experiments, although the magnitude of changes was generally underestimated by the DNA chips.
Immunohistochemical validation of transcripts identified by microarray analysis
To correlate the transcriptional analysis with protein expression, we performed immunohistochemistry of spleen cryosections of mice treated with PBS or soluble LTR-Ig with fluorescent-labeled Abs directed against PrPC, which has been previously identified as being expressed by FDCs (21, 27) and two of the novel candidate proteins. Two-color immunostaining with Abs directed against FDCs (FDC-M1), B cells (B220), and macrophages (F4-80) showed that clusterin, MFG-E8, and PrPC were expressed in the B cell area of splenic follicles (Fig. 6). Whereas clusterin immunoreactivity was also found at low levels in the marginal zone and red pulp (Fig. 6A), most of the MFG-E8 and PrPC signal was confined to the follicular area (Fig. 6, B and C). High power images of double-stained spleen cryosections with each of the Abs directed against the candidate proteins and FDC-M1 or FDC-M2 showed a high degree of colocalization of all three candidate proteins with these two markers on cells with morphology typical of FDCs localized on FDCs (Fig. 7). In the case of MFG-E8 staining was also observed on round cells lacking the typical shape of FDCs in addition to FDCs (Fig. 7, C and D). These cells may represent tingible-body macrophages. As Hanayama et al. (28) have previously shown by costaining of MFG-E8 with CD68, these non-FDC cells are most likely tingible-body macrophages (Fig. 7C).
FIGURE 6. Analysis of the localization and expression of a subset of the core candidates by immunofluorescence double staining. To determine whether the candidate proteins were localized in splenic GCs (right) and localized in FDCs (left) in a LTR-dependent fashion, cryostat sections of spleens from control-treated mice (PBS) or mice treated with soluble LTR-Ig for 2 days, were stained with Abs or antisera directed against clusterin (A), MFG-E8 (B) and PrPC (C) (all green) and Abs to FDCs (FDC-M1), B cells (B220), and macrophages (F4-80), respectively (red). Reactivity against the candidate proteins colocalized mainly with FDC networks in control-treated mice, whereas it was mostly absent 2 days after administration of soluble LTR-Ig (left). Costaining with B cell and macrophage markers demonstrated that expression of the candidate proteins was confined to the follicular B cell areas and that treatment with soluble LTR-Ig did not lead to down-regulation of the B220 and F4-80 epitopes (right). Exposure times of images with the same immunostainings were kept constant to ensure comparability of the images. The images are representative of two independent experiments. Scale bars represent 200 μm.
FIGURE 7. High power images depicting colocalization of clusterin, MFG-E8, and PrPC with FDC markers FDC-M1 and FDC-M2. Cryostat sections of spleens were costained with Abs or antisera directed against clusterin (A and B), MFG-E8 (C and D), and PrPC (E and F) (all green) and the FDC markers FDC-M1 (A, C, and E) or FDC-M2 (B, D, and F) (red). Original magnification of the images is x200.
We then tested whether blockage of LTR-signaling by soluble LTR-Ig affected protein expression levels of the three candidates. Costaining of spleen cryosections taken from mice treated with soluble LTR-Ig for 2 days showed that protein expression of clusterin, MFG-E8 and PrPC and localization within the GC were all critically dependent on intact LTR signaling (Fig. 6).
Expression of the core set genes in mice lacking GC and FDC networks
Having defined a core set of genes whose expression was associated with functional GC and FDC networks, we assessed whether expression of these genes is uniquely down-regulated by pharmacologic suppression of LTR signaling, or whether they would also be down-regulated in various mouse models with genetic disturbances of these structures. We therefore measured the expression levels of each of the 12 genes by quantitative real-time PCR in spleens of LTR–/–, TNFR1–/–, and TNFR2–/– mice, and compared them with expression in wild-type mice (C57BL/6).
In LTR–/– mice, eight of 12 genes were significantly down-regulated (Fig. 8A). Particularly strong was the down-regulation of Cxcl13 (97.0-fold), Clu (8.1-fold), Enpp2 (15.3-fold), Mfge8 (22.8-fold), and Serpina1a (34.7-fold). Mice deficient in TNFR1 signaling showed a similar down-regulation in seven of the eight genes down-regulated in LTR–/– mice (Fig. 8A). However, the fold-reduction was not as pronounced (Cxcl13, 3.3-fold; Clu, 2.4-fold; Enpp2, 2.4-fold; Mfge8, 2.4-fold; and Serpina1a, 9.4-fold). This is consistent with the observation that the lymphoid architecture of TNFR1–/– mice is less dramatically impaired than that of LTR–/– mice. TNFR2–/– mice do not have any impairment in formation of GC and FDC networks (29, 30). Accordingly, TNFR2–/– mice showed marginal reduction in three of the 12 core set genes and no significant change in the remaining nine genes (Fig. 8A).
FIGURE 8. Expression of the core set genes correlates with presence of functional GC. A, Real-time quantitative PCR was performed on cDNAs derived from spleens of C57BL/6, LTR–/–, TNFR1–/–, and TNFR2–/– mice. Fold down-regulation was calculated relative to C57BL/6 mice after normalization to the -actin signal. Error bars indicate SEM and are representative of three biological replicates. Purified murine GC B cells were cultured in the presence or absence of anti-CD40 Ab FGK45.5 (10 μg/ml) (B), purified human serum clusterin (50 μg/ml) (C), boiled human clusterin (50 μg/ml) (D), LPA (1 μM) (E), or human 1-antitrypsin (1 mg/ml) (F). Viability was assessed by trypan blue exclusion at the start of the cultures and after 1, 2, and 3 days. The experiment was performed in triplicate plus SEM. SD is indicated by error bars.
Clusterin acts as a survival factor for GC B cells in vitro
Experiments with FDC cultures have suggested that FDCs provide Ag-independent support for B cell follicles and GC reactions (10, 11, 12, 13). We therefore investigated whether clusterin, human 1-antitrypsin (the human homologue of Serpina1a), and LPA, which is produced by ENPP2 (31), could support survival of GC B cells in vitro. In two independent experiments GC B cells were isolated from Ova-immunized C57BL/6 mice by positive or negative (data not shown) magnetic bead sorting. Viability of GC B cells placed in culture under low serum conditions (1.5% FCS) was increased from 36.8 to 60.4% after 24 h, from 18.9 to 42.4% after 48 h, and from 9.9 to 25.6% after 72 h and in 35 to 57% after 24 h, from 8 to 29% after 48 h, and from 2 to 19% after 72 h of culture with purified human clusterin relative to untreated controls (Fig. 8B). Treatment with anti-CD40 Ab, a potent stimulator preventing GC B cell apoptosis (32, 33), increased survival to a similar degree as did clusterin (Fig. 8A) with 59.6% viability at 24 h, 48.1% at 48 h, and 33.9% at 72 h. To exclude the possibility that contaminating endotoxins may have contributed to the survival effect observed in the clusterin treated samples, we included boiled clusterin as a control. This treatment abrogated the GC B cell survival effect of clusterin (Fig. 8D) (54% viability at 24 h, 30% at 48 h, and 20% at 72 h). Addition of human 1-antitrypsin had little effect during the first 48 h of culture but increased viability of GC B cells to 20% at 72 h of culture (Fig. 7D). Viability of cultured cells was not increased by addition of LPA or 1-antitrypsin to the medium (Fig. 8, E and F), which speaks against a survival function of these molecules for LPA in the GC reaction. Both, positive and negative sorting (data not shown) of GC B cells yielded identical results and it is therefore unlikely that the survival results from activation of the cells during isolation.
Discussion
Although much is known about the developmental and morphogenetic events governed by LTR signaling, the LTR-dependent molecular cascades that control GC and FDC network function are ill defined. To circumvent the inherent difficulties in analyzing these structures, we used two complementary approaches to characterize the transcriptional profiles associated with GC and FDC networks and to identify genes that may functionally contribute to the events occurring therein.
In a first approach, we took advantage of the transient depletion of GCs and FDCs that occurs upon interference with the LT signaling axis by administration of soluble LTR-Ig. We reasoned that any gene that might experience down-regulation under these conditions would represent at least a candidate marker for GCs and FDCs, and that some of them might be involved in GC or FDC function.
Transient blockade of the LTR pathway led to the identification of a total of 80 differentially expressed transcripts that exhibited rapid and reversible changes in gene expression after administration of soluble LTR-Ig. These closely followed the sequence of histologic disappearance and reappearance of FDCs, GCs, and their markers from mesenteric lymph nodes and may be directly regulated by LTR signaling. The proteins encoded by these genes were predominantly assigned to four key categories (I-IV). This expands previous reports that LTR signaling controls proinflammatory molecules, chemokines, and cytokines (34). The fact that 25% of these genes are normally associated with the stromal compartment of secondary lymphoid organs emphasizes the importance of this compartment in executing functions controlled by LTR signaling.
Our second approach consisted of analyzing cell clusters highly enriched for FDCs. These clusters represented small functional units of GCs and allowed us to define a set of genes overexpressed in such structures. Particularly remarkable was the identification of genes of the functional categories (I) surface receptors and cell adhesion molecules, (II) chemokines and cytokines, and (X) ECM components.
We have identified a specific set of chemokines that are expressed in association with FDCs. These are able to attract the typical cell types of hemopoietic origin, which cluster during a GC reaction in FDC networks, such as B cells (CXCL13) (35), Th cells (CXCL10, CXCL12) (36, 37), and macrophages/monocytes (CCL2, CXCL12) (37, 38). Their high level expression within FDC networks explains the chemotactic activities for B cells and Th cells previously identified in FDC conditioned medium (10).
The overrepresentation of several cell adhesion molecules in FDC cell clusters reflects the intensive intercellular contacts occurring between FDCs and cognate cells of hemopoietic origin. ICAM-1 and VCAM-1 are both highly expressed on FDCs (39), and similarly E-selectin has been detected on FDCs present in inflamed thyroid glands of patients with Hashimoto’s thyroiditis (40). Expression of GlyCAM-1 and MCAM by FDCs had not been appreciated yet. Ablation of VCAM-1 expression had identified a crucial role for this molecule in mounting of efficient secondary IgG1 Ab responses (41). GlyCAM-1 has been shown to stimulate adherence of lymphocytes to ICAM-1 expressing high endothelial venules via activation of L-selectin (42). Local synthesis of GlyCAM-1 suggests that FDCs may actively stimulate B cells, via activation of L-selectins, to attach to ICAM-1 molecules present on FDCs. Additional experiments will need to be devised to clarify the contribution to FDC function of the additional cell adhesion molecules identified in this study.
In accordance with elevated expression of cytokines exerting their effects mainly on the stromal compartment (PDGFA, CTGF, BMP-2), we detected a marked increase in transcripts encoding several types of collagens and other ECM proteins. It is likely that their increased deposition in the vicinity of FDCs facilitates migration and adherence of immune cells into this area, for example, via interactions with integrins on these cells. They may also have proliferative effects, as was shown for biglycan, which stimulates proliferation of pre-B cells by bone marrow stromal cells (43); and may enhance clustering of B lymphocytes in this area, as is the case for tenascin in an LFA-1-dependent fashion (44).
By intersection of the two analyses we selected a core set of genes that is not only tightly associated with FDCs and is highly expressed within GC, but is also dependent on signaling through the LTR. We demonstrated the validity of this approach by protein expression analysis of PrPC, clusterin, and MFG-E8.
Expression analysis in spleens of LTR–/–, TNFR1–/–, and TNFR2–/– mice uncovered that most core set genes are down-regulated whenever induction of GC structures and FDC network formation is defective (LTR–/– and TNFR1–/– mice) (1, 2, 30), but are normally expressed when normal induction is possible, as in wild-type or TNFR2–/– mice (29). Expression of three of the core set genes (Cxcl13, Mfge8, and Clu) has been found to be diminished also in LT–/– mice (45), supporting the concept that expression of these genes is linked to the presence of functional GC and FDC networks. The fact that seven of eight genes down-regulated in LTR–/– spleens were also down-regulated in TNFR1–/– spleens suggests that the two pathways exert partially redundant functions. Given that the two receptors share downstream signaling adaptors, such as NF-B (46), it is not surprising that an overlapping set of genes is down-regulated in mice lacking the respective receptors.
Our core set of genes indicates that LTR signaling participates at several stages in the GC reaction by controlling gene expression associated with FDC networks: secretion of chemokines may attract subsets of activated cells of hemopoietic origin into the follicular B cell areas. For example, Ag-specific activated B cells are attracted by CXCL13 (35). FDCs and stromal cells present in this area provide a specific extracellular environment by secretion of particular ECM components and presentation of cell adhesion molecules, which are permissive for migration, adherence, interaction and survival of the infiltrating cells. LTR signaling may foster adherence by enhancing expression of molecules such as GlyCAM-1 and cochlin. Activated B cells that have migrated into the GC area differentiate into centroblasts and expand rapidly in an Ag-independent fashion to undergo somatic hypermutation and Ab class switch. 1-Antitrypsin, the human homologue to Serpina1a, can act as a costimulus for IgE and IgG4 synthesis by human B cells (47). We therefore speculate that LTR signaling might influence switching to specific Ig classes. Most of the random hypermutations are detrimental to Ag binding, and the majority of the emerging centrocytes undergo rapid apoptosis. LTR signaling seems to control the MFG-E8-dependent removal of such apoptosed cells by tingible-body macrophages (28). High-affinity centrocytes are positively selected and receive survival signals from FDCs and Th cells. By incubating GC B cells with purified clusterin we show that clusterin can prolong survival of such B cells with an efficiency comparable to a CD40 agonistic Ab. It has been demonstrated that clusterin binds to the Fc and Fab regions of Igs by a multivalent mechanism (18). It is therefore conceivable that clusterin produced by FDCs in high local concentrations may cross-link BCR molecules present on the surface of GC B cells and may thus promote their survival in an Ag-independent manner. Additional experiments devised at characterizing possible signals elicited by clusterin downstream of the BCR will be important to clarify the mode of action of clusterin.
Attempts by others to identify characteristic molecules expressed by human FDCs led to the identification of 8D6 (48). The murine homologue of this novel molecule was not part of the core set candidates in our analysis. 8D6 was not found enriched in FDC preparations relative to FDC-depleted spleen specimens. This may be due to expression of 8D6 on cell types other than FDCs present in secondary lymphoid organs. Expression of 8D6 in mesenteric lymph nodes was also not dependent on intact LTR signaling (48).
GC structures are not only involved in physiological humoral immune responses, but are exploited by various pathogens, which have evolved to thrive within them. HIV, for example, is mainly localized to FDCs of secondary lymphoid organs throughout the preclinical phase of infection (49). Most notably, prions accumulate in lymphoid organs early after peripheral infection in most instances of spongiform encephalopathies, including scrapie (50) and variant (51) as well as sporadic (52) Creutzfeldt-Jakob disease. Whenever the localization of prions within the lymphoreticular system was studied, they were traced to FDCs.
It is not implausible to expect that pharmacologic ablation of FDCs may be beneficial in these diseases, as it would remove the sanctuaries of pathogen persistence. If the regimen is initiated early after exposure, depletion of FDCs by soluble LTR can largely protect mice from the clinical consequences of exposure to infectious scrapie prions (21). The set of molecules identified in the present work may therefore encompass potential therapeutic target genes, whose manipulations in the conditions described above may prove clinically beneficial.
Acknowledgments
We thank Rita Moos for excellent technical assistance and Bernhard Odermatt for discussion.
Disclosures
The authors have no financial conflict of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported by grants of the Bundesamt für Bildung und Wissenschaft (EU) and the Swiss National Foundation, the U.S. National Prion Research Program, and the NCCR on Neural Plasticity and Repair (to A.A.), and by the Verein zur F?rderung des Akademischen Nachwuchses and the Stiftung für Biomedizinische Forschung (to C.H.) and Functional Genomics Centre, Zürich.
2 Address correspondence and reprint requests to Drs. Adriano Aguzzi or Gino Miele, Institute of Neuropathology, University Hospital of Zürich, Schmelzbergstrasse 12, CH-8091 Zürich, Switzerland. E-mail address: adriano@pathol.unizh.ch or gino.miele@usz.ch
3 Abbreviations used in this paper: LT, lymphotoxin; SAM, significance analysis of microarray; PNA, peanut agglutinin; SSH, suppression subtractive hybridization; DIG, digoxigenin; GC, germinal center; FDC, follicular dendritic cell; ECM, extracellular matrix; LPA, lysophosphatidic acid; MFG, milk fat globule.
Received for publication October 27, 2004. Accepted for publication February 14, 2005.
References
Futterer, A., K. Mink, A. Luz, M. H. Kosco-Vilbois, K. Pfeffer. 1998. The lymphotoxin receptor controls organogenesis and affinity maturation in peripheral lymphoid tissues. Immunity 9:59.
Pasparakis, M., L. Alexopoulou, M. Grell, K. Pfizenmaier, H. Bluethmann, G. Kollias. 1997. Peyer’s patch organogenesis is intact yet formation of B lymphocyte follicles is defective in peripheral lymphoid organs of mice deficient for tumor necrosis factor and its 55-kDa receptor. Proc. Natl. Acad. Sci. USA 94:6319.
Pfeffer, K., T. Matsuyama, T. M. Kundig, A. Wakeham, K. Kishihara, A. Shahinian, K. Wiegmann, P. S. Ohashi, M. Kronke, T. W. Mak. 1993. Mice deficient for the 55 kd tumor necrosis factor receptor are resistant to endotoxic shock, yet succumb to L. monocytogenes infection. Cell 73:457.
Pasparakis, M., L. Alexopoulou, V. Episkopou, G. Kollias. 1996. Immune and inflammatory responses in TNF -deficient mice: a critical requirement for TNF in the formation of primary B cell follicles, follicular dendritic cell networks and germinal centers, and in the maturation of the humoral immune response. J. Exp. Med. 184:1397.
Korner, H., M. Cook, D. S. Riminton, F. A. Lemckert, R. M. Hoek, B. Ledermann, F. Kontgen, B. Fazekas de St Groth, J. D. Sedgwick. 1997. Distinct roles for lymphotoxin- and tumor necrosis factor in organogenesis and spatial organization of lymphoid tissue. Eur. J. Immunol. 27:2600.
De Togni, P., J. Goellner, N. H. Ruddle, P. R. Streeter, A. Fick, S. Mariathasan, S. C. Smith, R. Carlson, L. P. Shornick, J. Strauss-Schoenberger, et al 1994. Abnormal development of peripheral lymphoid organs in mice deficient in lymphotoxin. Science 264:703.
Koni, P. A., R. Sacca, P. Lawton, J. L. Browning, N. H. Ruddle, R. A. Flavell. 1997. Distinct roles in lymphoid organogenesis for lymphotoxins and revealed in lymphotoxin -deficient mice. Immunity 6:491.
Fu, Y. X., H. Molina, M. Matsumoto, G. Huang, J. Min, D. D. Chaplin. 1997. Lymphotoxin- (LT) supports development of splenic follicular structure that is required for IgG responses. J. Exp. Med. 185:2111.
Karrer, U., C. Lopez-Macias, A. Oxenius, B. Odermatt, M. F. Bachmann, U. Kalinke, H. Bluethmann, H. Hengartner, R. M. Zinkernagel. 2000. Antiviral B cell memory in the absence of mature follicular dendritic cell networks and classical germinal centers in TNFR1–/– mice. J. Immunol. 164:768
Bouzahzah, F., N. Antoine, L. Simar, E. Heinen. 1996. Chemotaxis-promoting and adhesion properties of human tonsillar follicular dendritic cell clusters. Res. Immunol. 147:165.
van Eijk, M., C. de Groot. 1999. Germinal center B cell apoptosis requires both caspase and cathepsin activity. J. Immunol. 163:2478.
Li, L., Y. S. Choi. 2002. Follicular dendritic cell-signaling molecules required for proliferation and differentiation of GC-B cells. Semin. Immunol. 14:259
Wu, J., D. Qin, G. F. Burton, A. K. Szakal, J. G. Tew. 1996. Follicular dendritic cell-derived antigen and accessory activity in initiation of memory IgG responses in vitro. J. Immunol. 157:3404.
Thielen, C., N. Antoine, F. Melot, J. Y. Cesbron, E. Heinen, R. Tsunoda. 2001. Human FDC express PrPc in vivo and in vitro. Dev. Immunol. 8:259.
Li, C., W. H. Wong. 2001. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc. Natl. Acad. Sci. USA 98:31.
Tusher, V. G., R. Tibshirani, G. Chu. 2001. Significance analysis of microarrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. USA 98:5116
Fehr, T., R. C. Rickert, B. Odermatt, J. Roes, K. Rajewsky, H. Hengartner, R. M. Zinkernagel. 1998. Antiviral protection and germinal center formation, but impaired B cell memory in the absence of CD19. J. Exp. Med. 188:145
Wilson, M. R., S. B. Easterbrook-Smith. 1992. Clusterin binds by a multivalent mechanism to the Fc and Fab regions of IgG. Biochim. Biophys. Acta 1159:319.
Mackay, F., G. R. Majeau, P. Lawton, P. S. Hochman, J. L. Browning. 1997. Lymphotoxin but not tumor necrosis factor functions to maintain splenic architecture and humoral responsiveness in adult mice. Eur. J. Immunol. 27:2033.(Christoph Huber, Caroline)
Affinity maturation and Ab class switches occur in lymphoid germinal centers (GCs), in which differentiation and maintenance depend on lymphotoxin (LT) signaling and include differentiation of follicular dendritic cells (FDCs). The events leading to FDC and GC maturation are poorly defined. Using several approaches of functional genomics, we enumerated transcripts affected in mice by suppressing LT receptor (LTR) signaling and/or overrepresented in FDC-enriched GC isolates. Protein expression analysis of 3 of 12 genes both enriched in FDCs and down-regulated by LTR signaling suppression validated them as FDC markers. Functional analysis of one of these three, clusterin, suggests a role as an FDC-derived trophic factor for GC B cells. Hence, the set of genes presented in this study includes markers emanating from LTR signaling and transcripts relevant to GC and FDC function.
Introduction
Signaling through the lymphotoxin (LT) 3 receptor (LTR) and the TNFR1 is crucial for organogenesis and maintenance of the structural integrity of secondary lymphoid organs (1, 2). These organs provide the compartmentalized microenvironment essential for mounting efficient humoral immune responses.
Mice deficient for LTR, TNFR1, or their ligands suffer from complex and partially overlapping pathologic phenotypes of the lymphoreticular system. TNF- mice and TNFR1-deficient (TNFR1–/–) mice develop lymph nodes, but TNFR1–/– mice show hypoplastic Peyer’s patches (3, 4, 5). LT-deficient (LT–/–) and LT-deficient (LT–/–) mice lack Peyer’s patches and most lymph nodes, except for mesenteric and cervical lymph nodes (6, 7). LTR–/– mice display the severest phenotype: they lack Peyer’s patches and all lymph nodes (1).
All these mutant mice develop a spleen, but show varying degrees of disturbance in white pulp compartmentalization, as well as impaired formation of germinal center (GC) and follicular dendritic cell (FDC) networks after challenge with antigenic substances. These deficits are reflected in reduced isotype switching (LT–/–, LT–/–, LTR–/–, TNF-–/–, and TNFR1–/–) (1, 3, 4, 7, 8), impaired affinity maturation (LTR–/–) (1), and compromised B cell memory (LT–/–, LTR–/–, and TNFR1–/–) (1, 8, 9). FDCs may represent a key player in promoting and regulating these events occurring during the GC reaction. In vitro experiments have shown that FDCs and FDC-conditioned medium exert chemotaxis on B cells and CD4+ T cells (10). FDCs can promote survival and proliferation of GC B cells (11, 12), and coculture with FDCs can enhance Ig secretion by B cells (13).
Given the complex nature of GCs, the identification of molecules that act downstream of LT and TNF signaling and that may control these phenomena has proven difficult. Also, the specific molecular contribution of FDCs to these processes has been elusive, as it has proven technically impossible to isolate mature FDCs to purity without the concomitant loss of markers that authenticate their mature state (FDC-M1) and functionality (CD21/CD35 and FDC-M2).
To gain insights into FDC-associated functions controlled by LTR signaling, many of which may contribute to the GC reaction, we pursued two complementary strategies aimed at defining the transcriptional profile associated with the presence of GC and FDC networks. First, we used high-density oligonucleotide microarrays to identify transcripts affected in vivo by administration of a soluble LTR-Ig, which blocks LTR signaling. Second, we screened for transcripts overrepresented in cell preparations enriched for FDCs relative to non-FDC splenic cell types by both microarrays and suppression subtractive hybridization. As might be expected, some of the transcripts were found to overlap in both populations. This allowed us to identify LTR signaling-dependent genes preferentially expressed in GCs in association with FDCs.
Materials and Methods
Enrichment of FDC clusters
FDCs were prepared by adapting a procedure of Thielen et al. (14). Eight- to 10-wk-old female C57BL/6 mice were injected with 200 μg of OVA in alum i.p. to increase the amount of FDCs present in secondary lymphoid tissues. Spleens were harvested 3 days after injection and cut in small pieces using a tissue chopper. The spleen pieces were digested two times for 20 min at 37°C in RPMI 1640 medium containing 1 mg/ml collagenase A (Boehringer Mannheim), 0.5 mg/ml dispase (type II; Boehringer Mannheim), 0.04 mg/ml DNase (type I; Boehringer Mannheim), and 0.4% BSA (Sigma-Aldrich). The supernatants of the two digestions were pooled and the cells were collected by centrifugation. The pelleted cells were resuspended in PBS containing 0.4% BSA and layered over FCS for four repeated sedimentations at 1 g at 4°C. The cells contained in the supernatant of the first sedimentation were collected and remaining FDCs were removed using FDC-M1 coupled magnetic beads (Dynal Biotech). The resulting cell fraction was frozen and used as the FDC-depleted sample for RNA isolation. To remove contaminating macrophages, the FDC-enriched cell clusters obtained after the four repetitive sedimentations were incubated in a plastic culture dish for 60 min with RPMI 1640 containing 10% FCS at 37°C, 5% CO2. Nonadherent cell clusters were frozen and used as the FDC-enriched cell fraction for immunofluorescence staining and RNA isolation.
Treatment of C57BL/6 mice with soluble LTR-Ig
Ten-wk-old female C57BL/6 mice were injected with 100 μg of soluble LTR-Ig i.v. As a control, identical mice were injected with the same volume of carrier only (PBS). The experimental mice were sacrificed 1, 2, 3, 27, and 35 days after injection; the control mice were sacrificed 3 days after administration of PBS. Mesenteric lymph nodes and spleens were harvested for further processing.
Preparation of labeled cRNA and microarray hybridization
Total RNA was extracted from all samples (FDC-enriched or -depleted cell fractions and mesenteric lymph nodes) using TRIzol (Invitrogen Life Technologies) according to the manufacturer’s instructions. RNA was subjected to a clean up step using RNeasy mini kit (Qiagen). RNA quality was assessed by agarose gel electrophoresis or Agilent 2100 Bioanalyzer. Processing of FDC-enriched and depleted cell fractions used: 15 μg of total RNA was subjected to cDNA synthesis using a cDNA Synthesis kit (cat. no. 11917-010; Invitrogen Life Technologies) and primer 5'-GGCCAGTGAATTGTAATACG ACTCACTATAGGGAGGCGG(dT)24–3'. Biotin-labeled cRNA was synthesized, using Enzo BioArray HighYield RNA transcript labeling kit (T7) (Enzo Biochem), cleaned with the RNeasy mini kit, quantified by spectrophotometry, and quality was assessed again by agarose gel electrophoresis or Agilent 2100 Bioanalyzer. Typical yields ranged from 60 to 90 μg of labeled cRNA. Labeled cRNA (15 μg) was fragmented in 40 mM Tris-acetate, 100 mM KOAc, 30 mM MgOAc, pH 8.1, at 95°C for 35 min. After fragmentation an aliquot of one representative of the replicates of each sample was hybridized to Affymetrix Test 3 arrays to determine the quality of the probes, reflected by the 3' to 5' ratio, and to test all buffers. The 3' to 5' ratio of the cRNAs used was always <2. The samples were then hybridized to Affymetrix MGU74vA2, Bv2, and Cv2 chips. All hybridizations were conducted for 16 h at 45°C at 60 rpm. After hybridization chips were washed and conjugated with streptavidin-PE according to the manufacturer’s instructions on the Affymetrix GeneChip Fluidics Station 450 with and scanned using the Affymetrix GS Scanner 2500 in conjunction with Affymetrix Microarray Suite 5.0 software. Processing of mesenteric lymph nodes samples showed: equal amounts of high quality total RNA from the individual mesenteric lymph node preparations were combined to obtain three individual pools of 15 μg of RNA per time point. Complementary DNA and RNA synthesis were conducted as described.
Labeled cRNA was hybridized to Affymetrix MOE430A and B chips.
Microarray data analysis
Intensity values of all chips were normalized with dChip software (www.dchip.org) (15) and, applying the model-based expression analysis algorithm (PM-only model), the expression values were calculated.
FDC-enriched (three biological replicates per chip set) and FDC-depleted (two biological replicates per chip set) cell samples were compared applying a 2-fold change (using the lower 90% confidence boundary of fold change) as the lower limit. In the case of the MGU74Bv2 chip set, two instead of three biological replicates for the FDC-enriched samples were included in the analysis due to contamination of one of the replicates. Only those genes that were flagged present in all FDC-enriched samples were considered. To increase the stringency a 4-fold change including the same criteria as mentioned was applied as the lower limit.
Significantly modulated genes by soluble LTR-Ig treatment were determined by significance analysis of microarrays (SAM) (16). A total of 100 random permutations of the data set were applied to calculate background changes and the false discovery rates used in SAM. Thresholds for significant changes were set to 0.38285 for analysis of MOE430A chips (lower cutoff 0.90741) and to 0.20248 for MOE430B chips (lower cutoff 0.90741). The random number seed was set to 1234567. Of the significantly changed genes, those considered for further analysis were flagged in two of three replicates of at least one sampling time point. Unsupervised hierarchical clustering of genes and correlation analysis was performed using the algorithms provided by the dChip software. For the correlation analysis of the genes represented on the MOE430A chips, cluster A was divided into two smaller subclusters exhibiting expression minima at days 3 or 27 postinjection. The average expression pattern of the two subclusters was used to calculate the correlation coefficients. For genes represented on the MOE430B chips the entire cluster A of this chip set was used for calculation of the correlation coefficients. Functional annotation of genes was performed using information provided by Affymetrix NetAffx (www.affymetrix.com/analysis/index.affx), LocusLink (www.ncbi.nlm.nih.gov/), and PubMed (www.ncbi.nlm.nih.gov/entrez/query.fcgi) databases. All microarray data sets are available at www.ncbi.nlm.nih.gov/geo. Accession numbers are GSE2123 (FDC-E vs FDC-D) and GSE2124 (soluble LTR-Ig-treated mesenteric lymph nodes).
Immunohistochemistry of cytospin FDC clusters
FDC clusters were isolated and cytospun onto glass slides. Slides were fixed with acetone on ice, air dried and stored frozen at –80°C until used for staining. For immunostaining slides were air dried, refixed with acetone, and redried. After rehydration in PBS, blocking was achieved with 2% goat serum in PBS. Then clusters were stained with FDC-M1 (1/50; BD Pharmingen) and FITC goat F(ab')2 anti-rat IgG (H+L) (1/200; BioSource International) and after thorough washing were fixed with 4% paraformaldehyde in PBS. Staining was then continued with CD4-Bio (1/100; R&D Systems), CD8-Bio (1/100; R&D Systems), CD68-Bio (1/200; Inotech), CD11c-Bio (1/100; BD Pharmingen), and streptavidin-Alexa 594 (1/500; Molecular Probes). All Abs were diluted in 2% goat serum in PBS. Photographs were taken with a Leica SP2 confocal laser scanning microscope.
Immunohistochemistry on cryosections
FDC-M1, CD21/CD35, peanut agglutinin (PNA), and B220 stainings for light microscopy were performed as previously described (17). For double staining with milk fat globule-epidermal growth factor 8 (MFG-E8) and prion protein (PrP), 5 μM cryosections were blocked with 1% goat serum and 0.5% BSA in PBS. Washes were performed with PBS and incubations with primary and secondary Abs with 1% goat serum and 0.5% BSA in PBS at room temperature. Abs used were: anti-MFG-E8 (1/1000; kind gift from Dr. S. Nagata, Osaka University Medical School, Osaka, Japan), anti-PrP XN (1/1000), B220 (1/400; BD Pharmingen), F4-80 (1/50; Inotech), FDC-M1 (1/50; BD Pharmingen), FDC-M2 (1/50; Immunokontakt), Alexa Fluor 594 goat anti-rat IgG (H+L) (1/100; Molecular Probes), FITC F(ab')2 goat anti-hamster IgG (1/100; Inotech), and Alexa 546 goat anti-rabbit IgG (H+L) (1/500; Molecular Probes).
For double staining with clusterin, endogenous peroxidase was quenched with iView inhibitor (Ventana). Then staining of clusterin was performed with the Tyramide Signal Amplification fluorescein system (TSA; PerkinElmer), using anti-clusterin Ab (1/500; Santa Cruz Biotechnology) as a primary and HRP donkey anti-goat IgG (1/200; Jackson ImmunoResearch Laboratories) as a secondary Ab according to the manufacturer’s instructions. Double staining with FDC-M1, B220, and F4-80 were performed as described. Photographs were taken with an Axiovert 200 M microscope (Zeiss).
Quantitative real-time PCR
Snap frozen mesenteric lymph nodes or spleens were homogenized in TRIzol reagent (Invitrogen Life Technologies) with a T18 Basic Disperser (Ika Works), and total RNA was extracted according to the manufacturer’s instructions. Before cDNA synthesis, residual genomic DNA was removed by the DNA-free kit (Ambion). Total RNA (5 μg) was converted into cDNA using the first strand cDNA synthesis kit (Amersham Pharmacia Biotech). Successful cDNA synthesis and contamination of total RNA with genomic DNA was tested by PCR with primers specific for Actb. Quantitative real-time PCR was performed using the SYBR Green PCR Master Mix (Applied Biosystems) on an ABI PRISM 7700 Sequence Detector (PerkinElmer). The following primer pairs were used: clusterin Clu, 5'-CCAGTTCCCAGACGTTGATT-3' and 5'-AGCAGGGATGAGGTGTTGAG-3'; cochlear Coch, 5'-CGTGCAAGGGGATCTAATGT-3' and 5'-GCTTACCTATCCACTTGAATGC-3'; ectonucleotide pyrophosphatase/phosphodiesterase 2 Enpp2, 5'-TGGCTTACGTGACATTGAGG-3' and 5'-AAATCCAAACCGGTGAGATG-3'; glycoprotein GpM6b, 5'-AAGAGCTGCACGGTGAGTTT-3' and 5'-GCACAAGCCACAATGAACAGG-3'; glycodelin A Gda, 5'-GTTGTGGGCCTAGTCCATGT-3' and 5'-GATACCGAGGGGAAGGATGT-3'; lysozyme Lyzs, 5'-TCCCTTGTCAGTCAGCACAG-3' and 5'-CTGTGCCCTCAGAAACCTTC-3'; Serpina1a, 5'-TCCAGATCCATATCCCCAGA-3' and 5'-AGGAACGGCTTCAAAGACTG-3'; glycosylation-dependent cell adhesion molecule 1 Glycam1, 5'-GCATTGATGGGCTCAGATTT-3' and 5'-ACTTCAACCCCAGGAAAACC-3'; RIKEN cDNA C030033F14Rik, 5'-AAAGTAAACACCGAAGGGACA-3' and 5'-TCGGGAAAGCAACAATCAAT3'; Mfge8, 5'-CCCTTCTCTCAGGCATTCTG-3' and 5'-AACCTGTCAACCACCCAGAG-3'; prion protein Prnp, 5'-GCTGGCCCTCTTTGTGACTA-3' and 5'-CTGGGCTTGTTCCACTGATT-3'; Cxcl13, 5'-TCGTGCCAAATGGTTACAAA-3' and 5'-ACAAGGATGTGGGTTGGGTA-3'; actin Actb, 5'-GACGGCCAGGTCATCACTAT-3' and 5'-ACATCTGCTGGAAGGTGGAC-3'.
Isolation of GC B cells
Eight-wk-old C57BL/6 mice were immunized as described for FDC enrichment, and spleens were isolated. RBC were lysed with 155 mM NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA. The spleen cells were then incubated with FITC anti-IgD Ab (1/100; BD Pharmingen) and GC B cells were thereafter negatively selected with magnetic beads directed against FITC, CD43, Thy1.2, Mac-1, and CD11c (all Miltenyi Biotec) using an autoMACS sorter (Miltenyi Biotec). Alternatively, GC B cells were positively selected by incubation with FITC anti-GL7 Ab (1/100; BD Pharmingen) and subsequent autoMACS sorting with anti-FITC magnetic beads. Purity of the GC B cells was assessed by FACS: 1 x 105 cells taken before and after MACS were resuspended in FACS buffer (20 mM EDTA, 2% FCS, 0.05% NaN3) and stained with the Abs B220-allophycocyanin (1/100; BD Pharmingen) and IgMb-PE (1/100; BD Pharmingen). FACS was performed on a four-color FACSCalibur flow cytometer (BD Biosciences). Purity of B220+IgMb+IgD– negatively selected GC B cells was >86% and purity of B220+GL7+ positively selected cells was >93% after MACS.
Cell viability assay
MACS purified GC B cells were cultured in IMDM medium (Sigma-Aldrich) supplemented with 1% penicillin/streptomycin (Invitrogen Life Technologies), 1.5% FCS (Sigma-Aldrich), and 2 mM L-glutamine (Invitrogen Life Technologies) in 96-well plates at a density of 1.5 x 105 cells/ml. Untreated cells were cultured in medium only, to treated cells the following reagents were added: anti-CD40 mAb (FGK 45.5, 10 μg/ml), human clusterin (50 μg/ml) purified as previously described (18), lysophosphatidic acid (LPA, 1 μM; Sigma-Aldrich), and human 1-antitrypsin (1 mg/ml; Sigma-Aldrich). Boiled human clusterin was incubated at 95°C for 15 min. Viability of the cells was assessed at the time points indicated by trypan blue exclusion in triplicates.
Construction of subtracted cDNA libraries using suppression subtractive hybridization (SSH) technique
FDC clusters and FDC-depleted cells isolated from 200 C57BL/6 mice as described were resuspended in buffer RLT of the RNeasy Midi kit (Qiagen) and homogenized with a T18 Basic Disperser (Ika Works). Total RNA was extracted according to the manufacturer’s instructions. Poly(A+) mRNA was isolated from the total RNA using Oligotex mRNA spin columns (Qiagen) following the manufacturer’s instructions. Total RNA was DNase treated using the DNA-free kit (Ambion).
Poly(A+) mRNA (0.3 μg) from FDC-enriched cells was used as the "tester" or the population whose up-regulated transcripts were to be identified, and an equal amount of poly(A+) mRNA from FDC-depleted cells served as the "driver" or the population whose transcripts served as a reference for cDNA subtraction (forward SSH library). The construction of the forward and reverse libraries was performed as per the SSH procedure using the PCR select cDNA subtraction kit (BD Clontech). Subtracted target cDNAs were ligated with the pGEM-T easy plasmid vector (Promega).
Inverted Northern blot
Plasmid (300 ng) was digested with 10 U EcoRI for 3 h at 37°C in 97-well plate in a GeneAmp PCR System 9700 PCR machine (Applied Biosystems). The digested product was divided into two halves and electrophoresed through a 1% agarose gel in duplicates. The DNA was then transferred to a Hybond-N+ membrane (Amersham Pharmacia Biotech). For hybridization 50 ng of FDC-E and FDC-D cDNA were labeled with [-32P]dCTP using random hexamer primers (Stratagene). After hybridization the membranes were exposed to BAS-MS imaging plates (FUJIFILM) overnight and the plates were scanned using a BAS-1800 II PhosphorImager (FUJIFILM). The signals were quantified with AIDA 2.41 image analyzer (Raytest). Hybridization signals obtained for each clone were normalized to the background signal originating from the vector in the same lane. Normalized signals for each clone were then compared between the two hybridizations.
In situ hybridization
Plasmids containing the differentially expressed cDNAs were linearized by restriction digest with NotI or SalI and isolated by gel purification (QIAquick gel purification kit). Digoxigenin (DIG)-labeled antisense and sense riboprobes were synthesized using the SP6/T7 DIG labeling kit (Roche). Thick cryosections (5 μM) of spleens were directly fixed with 4% paraformaldehyde in PBS on ice for 30 min and washed in PBS. Slides were then pretreated with 0.1 M HCl for 5 min and after washing acetylated with acetic anhydride for 5 min. After further washing with PBS the sections were dehydrated with ethanol (EtOH, 50, 70, 80, and 95%) and air-dried.
After prehybridization at 37°C for 3 h with prehybridization solution (50% formamide, 2.5x Denhardt’s, 50 mM EDTA, 50 mM Tris-HCl pH 7.6, 0.25 mg/ml yeast total RNA, 20 mM NaCl), the slides were incubated with denatured probe diluted in hybridization solution (50% formamide, 10% dextransulfate, 0.33 M NaCl, 20 mM Tris-HCl pH 7.5, 1 mM EDTA, 1x Denhardt’s, 0.5 mg/ml yeast tRNA) at 45–55°C in a humidified chamber overnight. The slides were then washed three times for 30 min each at the temperature used for hybridization with 0.2x SSC, 0.1x SSC, 50% formamide, 0.2x SSC, and then at room temperature with 0.2x SSC and 0.1 M Tris-HCl pH 7.5, 150 mM NaCl. Blocking was performed with blocking reagent (Roche) according to the manufacturer’s instructions. DIG labels were then detected with anti-DIG Fab (Roche) at a dilution of 1/100 and a color reaction with nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate.
Results
Transient blockade of LTR signaling in mesenteric lymph nodes by soluble LTR-Ig
LTR signaling can be transiently inhibited in adult mice by administration of the soluble decoy receptor, soluble LTR-Ig (19, 20, 21). We elected to analyze the transcriptional profile of mesenteric lymph nodes associated with the blockade of LTR signaling. We chose mesenteric lymph nodes because we expected that, based on their histologic structure, the contribution of GC and FDC to the total tissue transcriptome would be far greater than in spleen.
Although the effects of soluble LTR-Ig treatment on FDCs in splenic follicles have been previously documented (19), the consequences of this treatment for mesenteric lymph nodes have not been described. We first defined the time course of events occurring after soluble LTR-Ig injection on FDCs in mesenteric lymph nodes. A single i.v. dose of 100 μg of soluble LTR-Ig was administered to C57/BL6 mice, and general mesenteric lymph node follicular microarchitecture was analyzed by immunohistochemistry at defined time points following treatment. In contrast to carrier-only injections (PBS), FDC markers FDC-M1 and CD21/CD35 completely disappeared from the lymph follicles of mesenteric lymph nodes within 3 days. However, both markers reappeared 27–35 days later (Fig. 1). PNA staining of GC B cells disappeared with slightly delayed kinetics relative to the FDCs, as residual PNA positivity was still detectable at day 3 postinjection. Reappearance of PNA staining 27 and 35 days after administration of soluble LTR-Ig occurred in parallel with the FDC-marker FDC-M1 (Fig. 1). Treatment with soluble LTR-Ig had no morphologically discernible effects on the localization of non-FDC cells or on the expression of non-FDC markers within mesenteric lymph nodes during the period of analysis, as assessed by staining with the B cell marker B220 (Fig. 1), the macrophage marker F4-80, the dendritic cell marker CD11c, and the T cell markers CD4 and CD8 (data not shown).
FIGURE 1. Immunohistochemical analysis of the temporal effects of soluble LTR-Ig treatment on mesenteric lymph nodes. Stains were performed on cryostat sections of mesenteric lymph nodes harvested form C57BL/6 mice at the indicated time points post administration of soluble LTR-Ig or carrier only (PBS). Sections were counterstained by hematoxylin. Treatment with soluble LTR-Ig leads to complete disappearance of FDC markers (FDC-M1, CD21) and reactive GCs (PNA) form the follicular area within 3 days. Reformation of FDC networks and PNA-positive GC can be detected at days 27–35 postinjection. Localization and expression of markers specific for B cells (B220) were not affected by the treatment. Scale bar represents 200 μm.
Transcriptional patterns in mesenteric lymph nodes after transient soluble LTR-Ig-mediated FDC depletion
To identify gene expression patterns resulting in mesenteric lymph nodes from transient depletion of FDCs, we profiled the transcriptome of mesenteric lymph nodes harvested 1, 2, 3, 27, and 35 days after injection of soluble LTR-Ig or PBS. Three biological replicates per time point were hybridized to Affymetrix oligonucleotide microarray sets MOE430A and MOE430B. Genes whose expression was significantly changed by the blockade of LTR signaling were identified by SAM (16). When highly stringent thresholds for significant changes were set, SAM identified 235 probe sets with a false discovery rate of 0.35% for chip A and 145 probe sets with a false discovery rate of 3.11% for chip B.
Unsupervised hierarchical clustering (dCHIP software) (15) with these genes revealed four major expression patterns (Fig. 2A) Two of these, clusters A and B, contained genes reflecting short response latency to the blocking of the LTR signaling pathway. These transcripts showed an early change (decrease or increase) in expression with a peak of maximal change at day 2, 3, or 27, and returned to near-baseline levels at day 35 (Fig. 2A). The remaining two clusters, C and D, showed longer response latency with little change during the first two days of the treatment. Only at day 3 did the expression levels of the genes within these clusters start to be affected, with a maximal change at day 35 relative to control treatment.
FIGURE 2. Four predominant expression patterns are induced in mesenteric lymph node by soluble LTR-Ig treatment, which represent separate functional categories. A, Unsupervised hierarchical clustering of genes significantly changed (as detected by SAM) during the sampled time points represented on MOE430A and MOE430B chips. The normalized expression values of each gene were standardized to have a mean of 0 and a SD of 1. Each expression measurement displayed represents an average of three biological replicates. Rows represent the standardized expression levels during the time course of the individual probe sets and columns represent sampling time points. Red reflects expression levels above mean expression of a probe set across all time points, green reflects expression lower than the mean, black reflects mean expression. Gene clusters (A, B, C, and D) with similar expression patterns in response to soluble LTR-Ig treatment are indicated. The color-scale indicates the color code applied to display standardized expression values between –3.0 and 3.0. B, The genes of each of the four clusters (A, B, C, and D) were annotated to the functional categories: (I) membrane-bound surface molecules; (II) chemokines, cytokines, immunomodulatory and other secreted molecules; (III) immune effector molecules and apoptosis related; (IV) signal transduction related; (V) cytoskeleton related; (VI) cell cycle and stress; (VII) DNA binding, transcription, and translation; (VIII) metabolism; (IX) transport; (X) ECM components; (XI) other functions; and (XII) unknown functions. The percentage of the total genes represented by each functional category is indicated for the four clusters identified by hierarchical clustering.
The rapid response to the treatment of the genes contained in clusters A and B suggested that LTR signaling may transcriptionally regulate some of these genes. Hence they may reflect biological functions that are directly controlled by this pathway. In contrast, the longer response latency of the genes in clusters C and D could reflect secondary and/or indirect effects of the treatment. To further assess these questions we functionally annotated the known genes identified by SAM. For those genes with a known function, the four predominant clusters showed a bias toward separate functional categories (Fig. 2B). Genes of the clusters A and B were primarily annotated to the functional categories: (I) membrane-bound surface molecules; (II) chemokines, cytokines, immunomodulatory and other secreted molecules; (III) immune effector and apoptosis-related molecules; and (IV) signal transduction. Genes of the clusters C and D fell mostly into the categories (VII) DNA binding, transcription, translation; (VIII) metabolism; and (IX) transport (Fig. 2B). In cluster A the genes related to cell-cell adhesion (category I, 30.4%) and chemotaxis (category II, 13.0%) were particularly striking. Cluster B was enriched in surface receptors (category I, 25.0%) and cytokines and chemokines (33.3%). Cluster C showed an enrichment of genes encoding ribosomal subunits, proteasome subunits, or ubiquitin-conjugating enzymes, transcription factors, and proteins involved in energy metabolism, representing a total of 23.3% of all genes in this cluster. The genes of cluster D reflected similar functional categories as cluster C, but the majority (63.0%) could not be functionally annotated (Fig. 2B). These results suggested that clusters A and B might contain genes coding for specific biological functions attributable to LTR signaling, whereas the changes in expression of genes in clusters C and D reflected more general changes in basic cellular activity, possibly reflecting secondary effects downstream of the blockade of LTR signaling.
A collection of transcripts predominantly expressed by FDCs
As the expression pattern of cluster A correlated well with the kinetics of disappearance and reappearance of PNA+ GC and FDC networks from soluble LTR-Ig-treated mesenteric lymph node cryosections (Fig. 1), we analyzed this cluster in more detail. The highly stringent filtering criteria applied in SAM yielded an initial set of 21 genes for cluster A. To avoid excluding potentially valid candidates from further analysis, we chose to perform a correlation analysis using the dCHIP software. We calculated correlation coefficients for the association of each gene expression pattern with the average expression pattern of cluster A. We then selected genes with a correlation coefficient >0.9 and a greater than 0.85-fold down-regulation (p < 0.05) at the time of minimal expression relative to control treatment. We identified a total of 80 unique transcripts whose expression patterns correlated with cluster A, 42 of which could be annotated to known genes. Eleven of these genes are known to be expressed by stromal cells within secondary lymphoid organs, and most of them have been shown to be expressed by FDCs or to form part of subcellular structures, such as desmosomes that have been assigned to FDCs by electron microscopy (22). These include cell adhesion molecules (Coch, Dsc3, Dsg2, Glycam1, and Madcam1), chemokines (Cxcl13 and Ccl19), and cell surface molecules (BstI, Prnp, Cr2, and Plxnb1). Genes typically associated with the macrophage/monocyte lineage were also affected (Lyzs, Mfge8, and Faah), which is consistent with the expression of LTR on macrophages (23).
Selection of a set of GC/FDC-associated genes
To complement the first approach, and to identify transcripts that are overrepresented in cells associated with FDC networks in GCs, we purified FDC cell clusters and characterized their transcriptome relative to FDC-depleted spleen cells. Immunohistochemical analysis showed that the FDC clusters basically represented small functional units of GC light zone. Double staining with FDC-M1, B220, CD4, CD68, and CD11c revealed the presence of B cells, Th cells, macrophages, and dendritic cells adherent to FDCs in these clusters (Fig. 3A).
FIGURE 3. Immunofluorescence characterization of FDC-enriched cell clusters and functional annotation of genes overrepresented in such preparations. A, FDC-enriched cell clusters were cytospun onto glass slides and stained with FITC-labeled Abs directed against FDCs (FDC-M1) and costained with Alexa 594-coupled Abs for the presence of B cells (B220), Th cells (CD4), dendritic cells (CD11c), or macrophages (CD68). Most of the cells, adherent to FDCs, were B cells followed by Th cells. Only very few dendritic cells could be detected. This indicated that the major cell types taking part during a GC reaction are present in isolated FDC clusters. Scale bar represents 40 μm. B, The genes >4-fold overrepresented in FDC-enriched cell clusters were annotated to the 12 categories described in Fig 3. The percentage of the total genes represented by each functional category is indicated.
Labeled cRNA from FDC-enriched cell clusters and FDC-depleted spleen cells was hybridized to the entire chip sets (A, B, and C) of Affymetrix MGU74v2 chips. Using 2-fold changes as the lower limit and further filtering by selecting those probe sets with high significance (lower 90% confidence boundary of fold change), we found that 960 probe sets were significantly overrepresented in the FDC-enriched vs the FDC-depleted specimen. To reduce background noise, we increased the stringency of our criteria by selecting only genes with a >4-fold change for functional annotation. This filter was passed by 478 probe sets.
The majority of annotated genes encoded membrane proteins (category I, 18.2%) and secreted proteins, such as immunomodulatory cytokines and chemokines (category II, 7.1%; Fig. 3B). Among the membrane proteins, the cell adhesion molecules ICAM-1, VCAM-1, MCAM, GlyCAM-1, and E-selectin were overexpressed in FDC-enriched cell clusters. Jagged-1 and Delta-1, two ligands of the Notch-2 receptor, were overrepresented in the FDC associated over the FDC-depleted cell fraction. Within the class of secreted proteins (category II) we identified six chemokines that attract lymphocytes and macrophages (CCL2, CXCL12, and CXCL13), T cells and NK cells (CXCL10), and neutrophils (CXCL1 and CXCL7). Of the cytokines that were overrepresented in FDC-enriched cell clusters, IL-1 and IL-6 have been shown to act on B cells. Three other cytokines exert their effects mainly on the stromal compartment (platelet-derived growth factor (PDGF)A, connective tissue growth factor (CTGF), bone morphogenetic protein (BMP)-2). They can induce proliferation, migration or differentiation of mesenchymal cells and lead to production of extracellular matrix (ECM) components (24, 25, 26). In accordance with the up-regulation of these cytokines, there was a clear overrepresentation of genes encoding ECM components in the FDC clusters vs the remainder of spleen cells, such as Col1a1, Col1a2, Col3a1, Col4a1, Col4a2, Col5a2, Col14a1, Col18a1, Tnxb, Coch, Bgn, and Mglap (category X, 6.3%).
GC/FDC-associated genes identified by SSH
In a further experiment we used SSH to identify transcripts associated with GC and FDC networks. SSH is PCR-based and nonbiased: it allowed us to generate a library of cDNAs overrepresented in FDC-enriched vs FDC-depleted cell clusters without prior knowledge of their sequence. The FDC-enriched cDNA library was screened for FDC/GC-associated transcripts by inverted Northern blot. Seventeen clones from the library were found to be overexpressed in GC and FDC networks >2-fold and were sequenced (Fig. 4A). Six of 16 genes (Clu, Cyr61, Igfbp3, Mfge8, Serpina1, and Sparc) were also identified as overexpressed by the filters set in the corresponding microarray analysis (Fig. 4A). In situ hybridization with two of the candidates, Igfbp3 and Mfge8, revealed that they were both expressed in the follicular areas of spleen (Fig. 4B). Although Mfge8 expression appeared confined to the follicles, Igfbp3 showed additional intense labeling of marginal zone cells. These data confirm the concept that novel genes preferentially expressed in association with FDCs can be identified by comparing FDC-enriched and FDC-depleted splenic cell fractions.
FIGURE 4. Identification of genes overexpressed in association with FDCs by SSH. A, The 17 genes identified by SSH with their corresponding proteins are listed. Genes identified by both SSH and microarray analysis are shown in gray. LocusLink numbers and the fold overexpression in FDC-enriched (FDC-E) compared with FDC-depleted (FDC-D) specimens as determined by inverted Northern blot are indicated. B, In situ hybridization of two representative SSH candidates. Spleen cryosections from C57BL/6 mice were hybridized with DIG-labeled antisense riboprobes for Mfge8 and Igfbp3. To control for hybridization specificity sense riboprobes were incubated in parallel. The images are representative of two independent experiments.
A core set of LTR responsive transcripts in FDC-enriched cell clusters
By combining the two sample preparation approaches we selected a core set of genes that were not only associated with the presence of FDCs in enriched cell clusters but were also dependent on signaling through LTR (Fig. 5A). The intersection of the 80 genes associated with FDC disappearance upon blockade of LTR signaling and the 960 genes enriched in FDC cell clusters yielded a core set of 12 genes (Fig. 5B), three of which were also identified by SSH (Clu, Mfge8, and Serpina1). Two of the proteins encoded by the 12 genes are related to cell adhesion (GlyCAM-1 and MFG-E8), three have enzymatic activity (ENPP2, GDA, and LYSZ), one has chemotactic activity (CXCL13), one is related to apoptosis (clusterin), one to proteolysis (serpin a1a), one is an ECM component (cochlin), and for three the functions are not clear at the time (prion protein, glycoprotein m6b, RIKEN C030033F14Rik).
FIGURE 5. Identification and confirmation of genes blocked by soluble LTR-Ig treatment and overexpressed in FDC clusters. A, Venn diagram illustrating the approach used to identify genes blocked by soluble LTR-Ig treatment and overexpressed in FDC clusters. Twelve core set genes were found in common between 80 genes correlating with expression pattern A in mesenteric lymph nodes upon soluble LTR-Ig treatment (correlate with cluster A) and the 960 genes overrepresented in FDC-enriched cell clusters (FDC-E). B, The 12 core set genes with their corresponding proteins are listed. NCBI accession numbers are indicated in column 4 (footnote 1). Maximal fold down-regulation relative to control treatment as detected by microarray analysis is indicated in column 5 (footnote 2). Fold overrepresentation in FDC-enriched cell clusters (FDC-E) relative to FDC-depleted (FDC-D) spleen cells are indicated. C, Down-regulation of the 12 core set genes by soluble LTR-Ig treatment and enrichment in FDC-cell clusters was confirmed by quantitative real-time PCR. Soluble LTR-Ig treatment shown as values of fold-reduction relative to control treatment (PBS) at the day of maximal effect (soluble LTR-Ig treatment). FDC enrichment shown as values expressed as fold-enrichment in FDC-enriched cell clusters relative to non-FDC spleen cells. The experiment was done in triplicate and SEM is indicated with error bars.
We verified the down-regulation of the core set genes by blockade of LTR signaling and the overrepresentation of these genes in FDC cell clusters by quantitative real-time PCR. All 12 genes were consistently down-regulated in soluble LTR-Ig-treated relative to control treated mesenteric lymph nodes and were up-regulated in purified FDC clusters relative to non-FDC associated spleen cells (Fig. 5C). The fold down-regulation (between 2- and 20-fold) and fold overexpression (between 1.5- and 283-fold) measured by quantitative PCR reproduced the microarray experiments, although the magnitude of changes was generally underestimated by the DNA chips.
Immunohistochemical validation of transcripts identified by microarray analysis
To correlate the transcriptional analysis with protein expression, we performed immunohistochemistry of spleen cryosections of mice treated with PBS or soluble LTR-Ig with fluorescent-labeled Abs directed against PrPC, which has been previously identified as being expressed by FDCs (21, 27) and two of the novel candidate proteins. Two-color immunostaining with Abs directed against FDCs (FDC-M1), B cells (B220), and macrophages (F4-80) showed that clusterin, MFG-E8, and PrPC were expressed in the B cell area of splenic follicles (Fig. 6). Whereas clusterin immunoreactivity was also found at low levels in the marginal zone and red pulp (Fig. 6A), most of the MFG-E8 and PrPC signal was confined to the follicular area (Fig. 6, B and C). High power images of double-stained spleen cryosections with each of the Abs directed against the candidate proteins and FDC-M1 or FDC-M2 showed a high degree of colocalization of all three candidate proteins with these two markers on cells with morphology typical of FDCs localized on FDCs (Fig. 7). In the case of MFG-E8 staining was also observed on round cells lacking the typical shape of FDCs in addition to FDCs (Fig. 7, C and D). These cells may represent tingible-body macrophages. As Hanayama et al. (28) have previously shown by costaining of MFG-E8 with CD68, these non-FDC cells are most likely tingible-body macrophages (Fig. 7C).
FIGURE 6. Analysis of the localization and expression of a subset of the core candidates by immunofluorescence double staining. To determine whether the candidate proteins were localized in splenic GCs (right) and localized in FDCs (left) in a LTR-dependent fashion, cryostat sections of spleens from control-treated mice (PBS) or mice treated with soluble LTR-Ig for 2 days, were stained with Abs or antisera directed against clusterin (A), MFG-E8 (B) and PrPC (C) (all green) and Abs to FDCs (FDC-M1), B cells (B220), and macrophages (F4-80), respectively (red). Reactivity against the candidate proteins colocalized mainly with FDC networks in control-treated mice, whereas it was mostly absent 2 days after administration of soluble LTR-Ig (left). Costaining with B cell and macrophage markers demonstrated that expression of the candidate proteins was confined to the follicular B cell areas and that treatment with soluble LTR-Ig did not lead to down-regulation of the B220 and F4-80 epitopes (right). Exposure times of images with the same immunostainings were kept constant to ensure comparability of the images. The images are representative of two independent experiments. Scale bars represent 200 μm.
FIGURE 7. High power images depicting colocalization of clusterin, MFG-E8, and PrPC with FDC markers FDC-M1 and FDC-M2. Cryostat sections of spleens were costained with Abs or antisera directed against clusterin (A and B), MFG-E8 (C and D), and PrPC (E and F) (all green) and the FDC markers FDC-M1 (A, C, and E) or FDC-M2 (B, D, and F) (red). Original magnification of the images is x200.
We then tested whether blockage of LTR-signaling by soluble LTR-Ig affected protein expression levels of the three candidates. Costaining of spleen cryosections taken from mice treated with soluble LTR-Ig for 2 days showed that protein expression of clusterin, MFG-E8 and PrPC and localization within the GC were all critically dependent on intact LTR signaling (Fig. 6).
Expression of the core set genes in mice lacking GC and FDC networks
Having defined a core set of genes whose expression was associated with functional GC and FDC networks, we assessed whether expression of these genes is uniquely down-regulated by pharmacologic suppression of LTR signaling, or whether they would also be down-regulated in various mouse models with genetic disturbances of these structures. We therefore measured the expression levels of each of the 12 genes by quantitative real-time PCR in spleens of LTR–/–, TNFR1–/–, and TNFR2–/– mice, and compared them with expression in wild-type mice (C57BL/6).
In LTR–/– mice, eight of 12 genes were significantly down-regulated (Fig. 8A). Particularly strong was the down-regulation of Cxcl13 (97.0-fold), Clu (8.1-fold), Enpp2 (15.3-fold), Mfge8 (22.8-fold), and Serpina1a (34.7-fold). Mice deficient in TNFR1 signaling showed a similar down-regulation in seven of the eight genes down-regulated in LTR–/– mice (Fig. 8A). However, the fold-reduction was not as pronounced (Cxcl13, 3.3-fold; Clu, 2.4-fold; Enpp2, 2.4-fold; Mfge8, 2.4-fold; and Serpina1a, 9.4-fold). This is consistent with the observation that the lymphoid architecture of TNFR1–/– mice is less dramatically impaired than that of LTR–/– mice. TNFR2–/– mice do not have any impairment in formation of GC and FDC networks (29, 30). Accordingly, TNFR2–/– mice showed marginal reduction in three of the 12 core set genes and no significant change in the remaining nine genes (Fig. 8A).
FIGURE 8. Expression of the core set genes correlates with presence of functional GC. A, Real-time quantitative PCR was performed on cDNAs derived from spleens of C57BL/6, LTR–/–, TNFR1–/–, and TNFR2–/– mice. Fold down-regulation was calculated relative to C57BL/6 mice after normalization to the -actin signal. Error bars indicate SEM and are representative of three biological replicates. Purified murine GC B cells were cultured in the presence or absence of anti-CD40 Ab FGK45.5 (10 μg/ml) (B), purified human serum clusterin (50 μg/ml) (C), boiled human clusterin (50 μg/ml) (D), LPA (1 μM) (E), or human 1-antitrypsin (1 mg/ml) (F). Viability was assessed by trypan blue exclusion at the start of the cultures and after 1, 2, and 3 days. The experiment was performed in triplicate plus SEM. SD is indicated by error bars.
Clusterin acts as a survival factor for GC B cells in vitro
Experiments with FDC cultures have suggested that FDCs provide Ag-independent support for B cell follicles and GC reactions (10, 11, 12, 13). We therefore investigated whether clusterin, human 1-antitrypsin (the human homologue of Serpina1a), and LPA, which is produced by ENPP2 (31), could support survival of GC B cells in vitro. In two independent experiments GC B cells were isolated from Ova-immunized C57BL/6 mice by positive or negative (data not shown) magnetic bead sorting. Viability of GC B cells placed in culture under low serum conditions (1.5% FCS) was increased from 36.8 to 60.4% after 24 h, from 18.9 to 42.4% after 48 h, and from 9.9 to 25.6% after 72 h and in 35 to 57% after 24 h, from 8 to 29% after 48 h, and from 2 to 19% after 72 h of culture with purified human clusterin relative to untreated controls (Fig. 8B). Treatment with anti-CD40 Ab, a potent stimulator preventing GC B cell apoptosis (32, 33), increased survival to a similar degree as did clusterin (Fig. 8A) with 59.6% viability at 24 h, 48.1% at 48 h, and 33.9% at 72 h. To exclude the possibility that contaminating endotoxins may have contributed to the survival effect observed in the clusterin treated samples, we included boiled clusterin as a control. This treatment abrogated the GC B cell survival effect of clusterin (Fig. 8D) (54% viability at 24 h, 30% at 48 h, and 20% at 72 h). Addition of human 1-antitrypsin had little effect during the first 48 h of culture but increased viability of GC B cells to 20% at 72 h of culture (Fig. 7D). Viability of cultured cells was not increased by addition of LPA or 1-antitrypsin to the medium (Fig. 8, E and F), which speaks against a survival function of these molecules for LPA in the GC reaction. Both, positive and negative sorting (data not shown) of GC B cells yielded identical results and it is therefore unlikely that the survival results from activation of the cells during isolation.
Discussion
Although much is known about the developmental and morphogenetic events governed by LTR signaling, the LTR-dependent molecular cascades that control GC and FDC network function are ill defined. To circumvent the inherent difficulties in analyzing these structures, we used two complementary approaches to characterize the transcriptional profiles associated with GC and FDC networks and to identify genes that may functionally contribute to the events occurring therein.
In a first approach, we took advantage of the transient depletion of GCs and FDCs that occurs upon interference with the LT signaling axis by administration of soluble LTR-Ig. We reasoned that any gene that might experience down-regulation under these conditions would represent at least a candidate marker for GCs and FDCs, and that some of them might be involved in GC or FDC function.
Transient blockade of the LTR pathway led to the identification of a total of 80 differentially expressed transcripts that exhibited rapid and reversible changes in gene expression after administration of soluble LTR-Ig. These closely followed the sequence of histologic disappearance and reappearance of FDCs, GCs, and their markers from mesenteric lymph nodes and may be directly regulated by LTR signaling. The proteins encoded by these genes were predominantly assigned to four key categories (I-IV). This expands previous reports that LTR signaling controls proinflammatory molecules, chemokines, and cytokines (34). The fact that 25% of these genes are normally associated with the stromal compartment of secondary lymphoid organs emphasizes the importance of this compartment in executing functions controlled by LTR signaling.
Our second approach consisted of analyzing cell clusters highly enriched for FDCs. These clusters represented small functional units of GCs and allowed us to define a set of genes overexpressed in such structures. Particularly remarkable was the identification of genes of the functional categories (I) surface receptors and cell adhesion molecules, (II) chemokines and cytokines, and (X) ECM components.
We have identified a specific set of chemokines that are expressed in association with FDCs. These are able to attract the typical cell types of hemopoietic origin, which cluster during a GC reaction in FDC networks, such as B cells (CXCL13) (35), Th cells (CXCL10, CXCL12) (36, 37), and macrophages/monocytes (CCL2, CXCL12) (37, 38). Their high level expression within FDC networks explains the chemotactic activities for B cells and Th cells previously identified in FDC conditioned medium (10).
The overrepresentation of several cell adhesion molecules in FDC cell clusters reflects the intensive intercellular contacts occurring between FDCs and cognate cells of hemopoietic origin. ICAM-1 and VCAM-1 are both highly expressed on FDCs (39), and similarly E-selectin has been detected on FDCs present in inflamed thyroid glands of patients with Hashimoto’s thyroiditis (40). Expression of GlyCAM-1 and MCAM by FDCs had not been appreciated yet. Ablation of VCAM-1 expression had identified a crucial role for this molecule in mounting of efficient secondary IgG1 Ab responses (41). GlyCAM-1 has been shown to stimulate adherence of lymphocytes to ICAM-1 expressing high endothelial venules via activation of L-selectin (42). Local synthesis of GlyCAM-1 suggests that FDCs may actively stimulate B cells, via activation of L-selectins, to attach to ICAM-1 molecules present on FDCs. Additional experiments will need to be devised to clarify the contribution to FDC function of the additional cell adhesion molecules identified in this study.
In accordance with elevated expression of cytokines exerting their effects mainly on the stromal compartment (PDGFA, CTGF, BMP-2), we detected a marked increase in transcripts encoding several types of collagens and other ECM proteins. It is likely that their increased deposition in the vicinity of FDCs facilitates migration and adherence of immune cells into this area, for example, via interactions with integrins on these cells. They may also have proliferative effects, as was shown for biglycan, which stimulates proliferation of pre-B cells by bone marrow stromal cells (43); and may enhance clustering of B lymphocytes in this area, as is the case for tenascin in an LFA-1-dependent fashion (44).
By intersection of the two analyses we selected a core set of genes that is not only tightly associated with FDCs and is highly expressed within GC, but is also dependent on signaling through the LTR. We demonstrated the validity of this approach by protein expression analysis of PrPC, clusterin, and MFG-E8.
Expression analysis in spleens of LTR–/–, TNFR1–/–, and TNFR2–/– mice uncovered that most core set genes are down-regulated whenever induction of GC structures and FDC network formation is defective (LTR–/– and TNFR1–/– mice) (1, 2, 30), but are normally expressed when normal induction is possible, as in wild-type or TNFR2–/– mice (29). Expression of three of the core set genes (Cxcl13, Mfge8, and Clu) has been found to be diminished also in LT–/– mice (45), supporting the concept that expression of these genes is linked to the presence of functional GC and FDC networks. The fact that seven of eight genes down-regulated in LTR–/– spleens were also down-regulated in TNFR1–/– spleens suggests that the two pathways exert partially redundant functions. Given that the two receptors share downstream signaling adaptors, such as NF-B (46), it is not surprising that an overlapping set of genes is down-regulated in mice lacking the respective receptors.
Our core set of genes indicates that LTR signaling participates at several stages in the GC reaction by controlling gene expression associated with FDC networks: secretion of chemokines may attract subsets of activated cells of hemopoietic origin into the follicular B cell areas. For example, Ag-specific activated B cells are attracted by CXCL13 (35). FDCs and stromal cells present in this area provide a specific extracellular environment by secretion of particular ECM components and presentation of cell adhesion molecules, which are permissive for migration, adherence, interaction and survival of the infiltrating cells. LTR signaling may foster adherence by enhancing expression of molecules such as GlyCAM-1 and cochlin. Activated B cells that have migrated into the GC area differentiate into centroblasts and expand rapidly in an Ag-independent fashion to undergo somatic hypermutation and Ab class switch. 1-Antitrypsin, the human homologue to Serpina1a, can act as a costimulus for IgE and IgG4 synthesis by human B cells (47). We therefore speculate that LTR signaling might influence switching to specific Ig classes. Most of the random hypermutations are detrimental to Ag binding, and the majority of the emerging centrocytes undergo rapid apoptosis. LTR signaling seems to control the MFG-E8-dependent removal of such apoptosed cells by tingible-body macrophages (28). High-affinity centrocytes are positively selected and receive survival signals from FDCs and Th cells. By incubating GC B cells with purified clusterin we show that clusterin can prolong survival of such B cells with an efficiency comparable to a CD40 agonistic Ab. It has been demonstrated that clusterin binds to the Fc and Fab regions of Igs by a multivalent mechanism (18). It is therefore conceivable that clusterin produced by FDCs in high local concentrations may cross-link BCR molecules present on the surface of GC B cells and may thus promote their survival in an Ag-independent manner. Additional experiments devised at characterizing possible signals elicited by clusterin downstream of the BCR will be important to clarify the mode of action of clusterin.
Attempts by others to identify characteristic molecules expressed by human FDCs led to the identification of 8D6 (48). The murine homologue of this novel molecule was not part of the core set candidates in our analysis. 8D6 was not found enriched in FDC preparations relative to FDC-depleted spleen specimens. This may be due to expression of 8D6 on cell types other than FDCs present in secondary lymphoid organs. Expression of 8D6 in mesenteric lymph nodes was also not dependent on intact LTR signaling (48).
GC structures are not only involved in physiological humoral immune responses, but are exploited by various pathogens, which have evolved to thrive within them. HIV, for example, is mainly localized to FDCs of secondary lymphoid organs throughout the preclinical phase of infection (49). Most notably, prions accumulate in lymphoid organs early after peripheral infection in most instances of spongiform encephalopathies, including scrapie (50) and variant (51) as well as sporadic (52) Creutzfeldt-Jakob disease. Whenever the localization of prions within the lymphoreticular system was studied, they were traced to FDCs.
It is not implausible to expect that pharmacologic ablation of FDCs may be beneficial in these diseases, as it would remove the sanctuaries of pathogen persistence. If the regimen is initiated early after exposure, depletion of FDCs by soluble LTR can largely protect mice from the clinical consequences of exposure to infectious scrapie prions (21). The set of molecules identified in the present work may therefore encompass potential therapeutic target genes, whose manipulations in the conditions described above may prove clinically beneficial.
Acknowledgments
We thank Rita Moos for excellent technical assistance and Bernhard Odermatt for discussion.
Disclosures
The authors have no financial conflict of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported by grants of the Bundesamt für Bildung und Wissenschaft (EU) and the Swiss National Foundation, the U.S. National Prion Research Program, and the NCCR on Neural Plasticity and Repair (to A.A.), and by the Verein zur F?rderung des Akademischen Nachwuchses and the Stiftung für Biomedizinische Forschung (to C.H.) and Functional Genomics Centre, Zürich.
2 Address correspondence and reprint requests to Drs. Adriano Aguzzi or Gino Miele, Institute of Neuropathology, University Hospital of Zürich, Schmelzbergstrasse 12, CH-8091 Zürich, Switzerland. E-mail address: adriano@pathol.unizh.ch or gino.miele@usz.ch
3 Abbreviations used in this paper: LT, lymphotoxin; SAM, significance analysis of microarray; PNA, peanut agglutinin; SSH, suppression subtractive hybridization; DIG, digoxigenin; GC, germinal center; FDC, follicular dendritic cell; ECM, extracellular matrix; LPA, lysophosphatidic acid; MFG, milk fat globule.
Received for publication October 27, 2004. Accepted for publication February 14, 2005.
References
Futterer, A., K. Mink, A. Luz, M. H. Kosco-Vilbois, K. Pfeffer. 1998. The lymphotoxin receptor controls organogenesis and affinity maturation in peripheral lymphoid tissues. Immunity 9:59.
Pasparakis, M., L. Alexopoulou, M. Grell, K. Pfizenmaier, H. Bluethmann, G. Kollias. 1997. Peyer’s patch organogenesis is intact yet formation of B lymphocyte follicles is defective in peripheral lymphoid organs of mice deficient for tumor necrosis factor and its 55-kDa receptor. Proc. Natl. Acad. Sci. USA 94:6319.
Pfeffer, K., T. Matsuyama, T. M. Kundig, A. Wakeham, K. Kishihara, A. Shahinian, K. Wiegmann, P. S. Ohashi, M. Kronke, T. W. Mak. 1993. Mice deficient for the 55 kd tumor necrosis factor receptor are resistant to endotoxic shock, yet succumb to L. monocytogenes infection. Cell 73:457.
Pasparakis, M., L. Alexopoulou, V. Episkopou, G. Kollias. 1996. Immune and inflammatory responses in TNF -deficient mice: a critical requirement for TNF in the formation of primary B cell follicles, follicular dendritic cell networks and germinal centers, and in the maturation of the humoral immune response. J. Exp. Med. 184:1397.
Korner, H., M. Cook, D. S. Riminton, F. A. Lemckert, R. M. Hoek, B. Ledermann, F. Kontgen, B. Fazekas de St Groth, J. D. Sedgwick. 1997. Distinct roles for lymphotoxin- and tumor necrosis factor in organogenesis and spatial organization of lymphoid tissue. Eur. J. Immunol. 27:2600.
De Togni, P., J. Goellner, N. H. Ruddle, P. R. Streeter, A. Fick, S. Mariathasan, S. C. Smith, R. Carlson, L. P. Shornick, J. Strauss-Schoenberger, et al 1994. Abnormal development of peripheral lymphoid organs in mice deficient in lymphotoxin. Science 264:703.
Koni, P. A., R. Sacca, P. Lawton, J. L. Browning, N. H. Ruddle, R. A. Flavell. 1997. Distinct roles in lymphoid organogenesis for lymphotoxins and revealed in lymphotoxin -deficient mice. Immunity 6:491.
Fu, Y. X., H. Molina, M. Matsumoto, G. Huang, J. Min, D. D. Chaplin. 1997. Lymphotoxin- (LT) supports development of splenic follicular structure that is required for IgG responses. J. Exp. Med. 185:2111.
Karrer, U., C. Lopez-Macias, A. Oxenius, B. Odermatt, M. F. Bachmann, U. Kalinke, H. Bluethmann, H. Hengartner, R. M. Zinkernagel. 2000. Antiviral B cell memory in the absence of mature follicular dendritic cell networks and classical germinal centers in TNFR1–/– mice. J. Immunol. 164:768
Bouzahzah, F., N. Antoine, L. Simar, E. Heinen. 1996. Chemotaxis-promoting and adhesion properties of human tonsillar follicular dendritic cell clusters. Res. Immunol. 147:165.
van Eijk, M., C. de Groot. 1999. Germinal center B cell apoptosis requires both caspase and cathepsin activity. J. Immunol. 163:2478.
Li, L., Y. S. Choi. 2002. Follicular dendritic cell-signaling molecules required for proliferation and differentiation of GC-B cells. Semin. Immunol. 14:259
Wu, J., D. Qin, G. F. Burton, A. K. Szakal, J. G. Tew. 1996. Follicular dendritic cell-derived antigen and accessory activity in initiation of memory IgG responses in vitro. J. Immunol. 157:3404.
Thielen, C., N. Antoine, F. Melot, J. Y. Cesbron, E. Heinen, R. Tsunoda. 2001. Human FDC express PrPc in vivo and in vitro. Dev. Immunol. 8:259.
Li, C., W. H. Wong. 2001. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc. Natl. Acad. Sci. USA 98:31.
Tusher, V. G., R. Tibshirani, G. Chu. 2001. Significance analysis of microarrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. USA 98:5116
Fehr, T., R. C. Rickert, B. Odermatt, J. Roes, K. Rajewsky, H. Hengartner, R. M. Zinkernagel. 1998. Antiviral protection and germinal center formation, but impaired B cell memory in the absence of CD19. J. Exp. Med. 188:145
Wilson, M. R., S. B. Easterbrook-Smith. 1992. Clusterin binds by a multivalent mechanism to the Fc and Fab regions of IgG. Biochim. Biophys. Acta 1159:319.
Mackay, F., G. R. Majeau, P. Lawton, P. S. Hochman, J. L. Browning. 1997. Lymphotoxin but not tumor necrosis factor functions to maintain splenic architecture and humoral responsiveness in adult mice. Eur. J. Immunol. 27:2033.(Christoph Huber, Caroline)