To identify changes in the regulation of B cell receptor (BCR) signals during the development of human B cells, we generated genome-wide gene expression profiles using the serial analysis of gene expression (SAGE) technique for CD34+ hematopoietic stem cells (HSCs), pre-B cells, naive, germinal center (GC), and memory B cells. Comparing these SAGE profiles, genes encoding positive regulators of BCR signaling were expressed at consistently lower levels in naive B cells than in all other B cell subsets. Conversely, a large group of inhibitory signaling molecules, mostly belonging to the immunoglobulin superfamily (IgSF), were specifically or predominantly expressed in naive B cells. The quantitative differences observed by SAGE were corroborated by semiquantitative reverse transcription–polymerase chain reaction (RT-PCR) and flow cytometry. In a functional assay, we show that down-regulation of inhibitory IgSF receptors and increased responsiveness to BCR stimulation in memory as compared with naive B cells at least partly results from interleukin (IL)-4 receptor signaling. Conversely, activation or impairment of the inhibitory IgSF receptor LIRB1 affected BCR-dependent Ca2+ mobilization only in naive but not memory B cells. Thus, LIRB1 and IL-4 may represent components of two nonoverlapping gene expression programs in naive and memory B cells, respectively: in naive B cells, a large group of inhibitory IgSF receptors can elevate the BCR signaling threshold to prevent these cells from premature activation and clonal expansion before GC-dependent affinity maturation. In memory B cells, facilitated responsiveness upon reencounter of the immunizing antigen may result from amplification of BCR signals at virtually all levels of signal transduction.
Signal transduction pathways initiated through the B cell receptor (BCR)* determine the fate of B cells within a context of BCR-affinity to antigen, expression levels of stimulatory or inhibitory coreceptors and the differentiation stage of B cells (1). Whereas BCR engagement by self-antigen in immature bone marrow B cells induces receptor editing, deletion, or inactivation (anergy), BCR cross-linking in mature B cells initiates a signaling cascade that ultimately confers positive selection, proliferation, and differentiation.
Early B cell differentiation is defined by a sequence of Ig gene rearrangements determining the configuration of the (pre)-BCR. The recombination machinery first targets DH and JH gene segments at the pro-B cell stage followed by VH-DHJH gene rearrangement in pre-B cells, which subsequently express a pre-BCR, composed of Ig heavy and surrogate light chains, on their cell surface. Expression of Vpre-B and λ-5 surrogate light chain genes precedes the expression of Vκ-Jκ or Vλ-Jλ light chain gene rearrangements at the immature B cell stage, at which, for the first time, a BCR is expressed and the cells enter the peripheral blood as naive B cells. The specificity and structure of the BCR is further modified by somatic hypermutation and class switch recombination during the affinity maturation process within germinal centers (GCs), in which GC B cells are destined to die by apoptosis unless they are rescued through BCR-dependent survival signals upon antigen cross-linking. GC B cells expressing a BCR of high affinity to their cognate antigen subsequently differentiate into memory B cells or antibody-secreting plasma cells. Notably, signaling through the BCR not only determines the fate of a B cell at developmental checkpoints within the bone marrow and in GCs. Also mature B cells depend on the presence of a functional BCR, which continuously delivers a “maintenance” signal (2).
The BCR signaling cascade is initiated by rapid phosphorylation of tyrosine residues in the immunoreceptor tyrosine-based activation motifs (ITAMs) of the BCR coreceptors Igα and Igβ upon antigen cross-linking. The proximal signal transduction mainly involves three different protein tyrosine kinase (PTK) activities including src-family PTKs, SYK, and BTK (3). In addition to kinases, several phosphatases (SHP1, SHP2, CD45, SHIP) and linker proteins (BLNK, GRB2, SHC, NCK) also regulate BCR signal transduction. BLNK was recently shown to act as a scaffolding protein, which mediates the interaction between SYK and the downstream signaling molecules VAV (4) and PLCγ. The latter can hydrolyse PIP2 to IP3 and diacylglycerol, which increases the levels of free cytoplasmic Ca2+ and result in subsequent protein kinase C (PKC) and mitogen-activated protein (MAP) kinase activation, which ultimately initiates functional B cell responses including proliferation, isotype switching and antibody secretion. However, the downstream propagation of activation signals may be attenuated by inhibitory receptor molecules bearing one or more immunoreceptor tyrosine-based inhibitory motifs (ITIMs). Many of these inhibitory receptors belong to the Ig superfamily (IgSF), which includes surface molecules such as CD5, CD22, and FcγRII/ CD32 (5).
To identify changes in the regulation of BCR-dependent activation signals at checkpoints during normal human B cell development, we analyzed and compared genome-wide gene expression profiles from human bone marrow hematopoietic stem cell (HSC), bone marrow pre-B cells, naive B cells, GC B cells, and memory B cells. These gene expression profiles were generated using the serial analysis of gene expression (SAGE) technique, which allows for the genome-wide quantitative analysis of any expressed mRNA in a given cell population (6).
Materials And Methods
Isolation of Human Hematopoietic Stem Cells, Pre-B Cells, and Mature B Cell Subsets.
HSCs and pre-B cells were purified from bone marrow and umbilical cord blood. Purification of bone marrow CD34+ HSC was described (7). Cord blood HSC were isolated using anti-CD34 immunomagnetic beads (Miltenyi Biotec). For enrichment of pre-B cells, mononuclear cells were isolated from four bone marrow samples (Poietics) and from 28 umbilical cord blood samples (according to the principle of informed consent) by Ficoll density gradient centrifugation. T cells and myeloid cells were depleted using anti-CD3 and anti-CD15 immunomagnetic beads (Dynal). Among the remaining cells, immature CD10lowCD19+CD20+ B cells and CD138+ plasma cells were depleted using an anti-CD20 IgG1 antibody (BD Biosciences) together with anti-IgG1 beads and anti-CD138 beads (Miltenyi Biotec), respectively (8). Thereafter, pre-B cells were enriched using anti-CD19 immunomagnetic multisort-beads (Miltenyi Biotec). The beads were released from the CD19+ cells enzymatically. The purified cells were subsequently labeled by a mouse anti-CD10 IgG1 antibody (CALLA; BD Biosciences) and separated using anti–mouse IgG1 beads (Miltenyi Biotec). IgD+CD19+CD27− naive B cells and CD19+CD27+ memory B cells were isolated from peripheral blood using anti-CD19 and anti-CD27 immunomagnetic beads (Miltenyi Biotec) as described (9) and from seven tonsilectomy specimens. For enrichment of tonsillar memory B cells, CD27lowCD38+ GC B cells were depleted using an anti-CD38 PE antibody (BD Biosciences) together with anti-PE microbeads (Miltenyi Biotec). Tonsillar CD77+ GC B cells were isolated as described previously (9) using a rat anti-CD77 IgM antibody (BD Biosciences) together with a mouse anti–rat IgM IgG1 antibody (Serotec) and anti–mouse IgG1 microbeads (Miltenyi Biotec). Only cell purifications of a purity >90% were included in the SAGE analysis.
Verification of B Cell Subset Purification.
The identity of the purified B cell subsets was verified genotypically and phenotypically. The genotype of purified preB cells and mature B cell subsets (naive and memory B cells) was assessed by PCR amplification of rearranged VH1 genes from genomic DNA and subsequent cloning and sequencing of the PCR products as described (10). In addition, the phenotype of the purified HSC, pre-B cells, naive B cells, GC B cells, and memory B cells was assessed by semiquantitative RT-PCR at the mRNA level using IgH Cμ-, IgH Cγ1-, Ig Cκ-, Ig Cλ-, VpreB1-, λ5-, and β2M-specific primers (Integrated DNA Technologies). To distinguish “mature” from germline Cγ1 transcripts, primers were chosen for a fragment between the JH and the hinge region of the constant region (11). The phenotype of purified preB cells was also verified by flow cytometry using CD10- and CD19-specific antibodies for pre-B cells, CD20-, and CD27-specific antibodies for naive and memory B cells, CD38- and CD77-specific antibodies for GC B cells, respectively (FITC- and PE-conjugated antibodies from BD Biosciences).
cDNA-synthesis, SAGE analysis, cloning, and sequencing of SAGE concatemers was performed according to Velculescu et al. (6). The UniGene reference database (March 2001) was obtained at http://www.sagenet.org/SAGEDatabases/unigene.htm. A total of 306,000 SAGE tags were collected for the five SAGE profiles. 106,000 tags were analyzed from the HSC library, 110,000 for pre-B cells, and each ∼30,000 tags for naive, GC, and memory B cells. All SAGE libraries were normalized to 100,000 tags.
Controls for the Accuracy of SAGE Library Construction.
RNA-degradation and incomplete digestion of 3′cDNAs by the so-called tagging enzyme NlaIII may interfere with the quantitative representation of expressed genes in SAGE libraries. RNA degradation within 3′ regions would result in the underrepresentation of genes, whose last NlaIII recognition site (CATG) is particularly far from the poly(dA) tail. The average distance between the last CATG site and the poly(dA) tail is ∼250 bp (6). To search for a potential bias against SAGE tags derived from 5′ sequences, we selected 10 housekeeping genes, whose extreme 3′ CATG site was more than 450 bp distant from the poly(dA) tail. We compared the tag counts for these genes with the tag counts in 55 published SAGE libraries (at http://www.ncbi.nlm.nih.gov/SAGE/index.cgi?cmd=tagsearch). In the SAGE libraries described here, tags for these genes were found at frequencies close to the average of 55 reference SAGE libraries, which argues against bias of quantitative representation introduced by RNA degradation.
Incomplete NlaIII digestion would result in the generation of SAGE tags that are aberrantly derived from upstream CATG sites instead of the extreme 3′ CATG site. To address this issue, we amplified cDNA fragments of the GAPDH and EF1 genes for the SAGE libraries for naive and memory B cells. Primers were chosen so that cleavage by NlaIII would result in the loss of the 5′ primer binding site for GAPDH but not for EF1. Consistent with high efficiency of NlaIII digestion, either no or only very small amounts of PCR product were obtained for GAPDH while amplification of EF1 fragments yielded abundant amplification products (unpublished data).
Selection of BCR-related Signaling Molecules.
In a comprehensive search for positive and negative BCR-related signaling molecules in PubMed, UniGene (http://www.ncbi.nlm.nih.gov/UniGene/) and OMIM (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM), we collected 211 genes, for which a role in positive (129) or negative (82) regulation of BCR-dependent signals was shown. Based on their UniGene-ID, 148 (97 positive and 51 inhibitory signaling molecules) of these genes could be retrieved from at least one of the five SAGE libraries.
Verification of Quantitative Accuracy of SAGE Data.
To corroborate quantitative differences in gene expression among the five populations as determined by SAGE, semiquantitative RT-PCR analysis was performed for a set of 41 selected genes. For verification of SAGE data, unrelated cDNAs from HSC and pre-B cells purified from umbilical cord blood (see above), naive, and memory B cells from peripheral blood were used as template. To specifically address quantitative differences in the expression of positive and inhibitory BCR signaling molecules in naive and memory B cells, we performed semiquantitative RT-PCR verification for 22 genes implicated in propagation and 19 genes involved in attenuation of BCR signals, respectively. RT-PCR was done for 18, 20, 24, 28, 30, or 32 cycles as indicated and normalized for the COX6B gene encoding the cytochrome c oxidase subunit VI and the β2M gene encoding β2-microglobulin.
Effect of mRNA Content per Cell on the Interpretation of SAGE Data.
For all SAGE profiles, expression levels of genes are given as tags per 100,000. This is based on the assumption that the relative contribution of a gene to the transcriptome of a cell population accurately reflects its expression level. However, an alternative viewpoint would be that gene expression levels should be given as absolute number of transcripts per cell. It is critical in the latter but not in the former case that the overall amounts of mRNA are similar in the cell populations compared by SAGE and semiquantitative RT-PCR. For instance, B lymphoblasts such as centroblasts within GCs may double their size and mRNA content as compared with their resting precursors (12). Thus, up-regulation of positive regulators of BCR signals in blastoid GC as compared with resting naive B cells as observed by SAGE (see Fig. 2, A and B) should be even higher if individual cells are analyzed. On the other hand, the quantitative difference of the expression of inhibitory BCR signaling molecules between individual resting naive or blastoid GC B cells would be lower than suggested by SAGE. As for size and mRNA content of memory versus naive B cells no pertinent information was available, total RNA and mRNA was isolated from 106 naive and 106 memory B cells. OD measurements did not reveal a significant difference between the two populations. In addition, we repeated semiquantitative RT-PCR for five genes for 200 and 1,000 cells, which were sorted into reaction tubes by a FACStar™ 440 cell sorter (BD Biosciences). RNA isolation, cDNA synthesis and PCR amplification was done in two independent experiments. The RT-PCR analysis was performed in 30 (for 200 cells) and 26 (for 1,000 cells) cycles for the BAM32, BLNK, SIgLec5, LIRB1, and COX6B genes (not shown). The quantitative proportions were similar as in the RT-PCR analysis, which was normalized for equal mRNA amounts (see Fig. 3). Thus, mRNA contents in naive versus memory B cells appear to be similar and do not significantly interfere with quantitative accuracy of SAGE- and RT-PCR data.
Clustering Analysis of SAGE Data.
For graphic representation of SAGE data, SAGE tags derived from known transcripts were arranged in functional clusters including positive and inhibitory BCR signaling molecules. The SAGE data were sorted based on the ratio of their frequency in memory B cells and naive B cells. For transformation and graphic representation of SAGE tag counts, the Cluster and Treeview software was used (kindly provided by Dr. Michael B. Eisen, Berkeley, CA, at http://rana.lbl.gov/).
Assessment of Expression of BCR Signaling-related Surface Molecules.
To verify differential expression of molecules related to BCR-signaling in naive and memory B cells at the protein level, we analyzed a panel of 10 positive and 10 negative regulators of BCR signaling by flow cytometry. Therefore, naive and memory B cells were purified from peripheral blood from 12 donors. Anti-CD5 PE, anti-CD19 PE, anti-CD21 PE, anti-CD22 FITC, anti-CD24 FITC, anti-CD31 FITC, anti-CD32/ FcRγII FITC, anti-CD40 PE, anti-CD66 FITC, anti-CD72 FITC, anti-CD74 FITC, anti-CD79a/ Igα PE, anti-LIRB1 FITC, anti-CD100 FITC, anti-CD124/ IL-4R PE, anti-CD130/gp130 PE, anti-CD150/SLAM PE, biotinylated anti-CD153/CD30 ligand, anti-PD1 PE, and streptavidin PE were from BD Biosciences. Goat anti-BCMA IgG and donkey anti–goat IgG FITC are from Santa Cruz Biotechnology, Inc.
IL-4–dependent Regulation of BCR Signaling-associated Genes in Naive and Memory B Cells.
Naive and memory B cells were purified from peripheral blood as described above and cultured at 37°C either in RPMI medium (including 10% fetal calf serum) alone, or with 1 ng/ml recombinant human IL-4 (Genzyme) or 50 μg of a neutralizing anti–IL-4Rα antibody (Genzyme)/ml, which was added after 8 h preincubation with IL-4. The cells were cultured at a density of 5 × 105 cells/100 μl/well in 96-well plates. After 48 h, the cells were subjected to RNA isolation and subsequent semiquantitative RT-PCR analysis for COX6, BLK, BTK, BLNK, SYK, LIRB1, LIRB2, LIRB5, SIgLec5, SIgLec8, CD66, CSK, SHIP, and SHP1 at 28 cycles and for LAIR1 at 32 cycles.
Effect of LIRB1 Signaling on BCR-dependent Ca2+ Mobilization in Naive and Memory B Cells.
To address directly how LIRB1 (an inhibitory IgSF receptor found prominently expressed on naive B cells) can affect responsiveness of the BCR to antigen in naive and memory B cells, BCR-dependent Ca2+ mobilization was studied. To this end, peripheral blood naive and memory B cells were purified from four healthy donors as described above and cultured in medium, which had been conditioned for 24 h by PBMCs at a density of 106 PBMCs/ml. PBMCs were stimulated with 1 μg LPS/ml to induce secretion of soluble MHC class I molecules (13), which act as natural ligand of LIRB1 (14). To study the effect of LIRB1 on BCR signals, naive and memory B cells were cultured for 24 h in the presence or absence of an antagonistic (clone HP-F1; reference 14) or agonistic (clone GVI/ 75; BD Biosciences; reference 15) LIRB1 antibody cross-linked by goat anti–mouse IgG serum (Jackson ImmunoResearch Laboratories). HP-F1 was a gift from Dr. Miguel López-Botet, Universitat Pompeu Fabra, Barcelona, Spain. After the preincubation, cells were washed and stained with Fluo-3 dye (Calbiochem) for 30 min. Changes of cytosolic Ca2+ were measured by laser scans using confocal microscopy (16). After 30 s of measurement, anti-human IgM F(ab′)2 and anti–human IgG + IgM F(ab′)2 fragments (Jackson ImmunoResearch Laboratories) were added to naive and memory B cells, respectively. Cytosolic Ca2+ concentrations were calculated as described previously (17). As a negative control, purified B cell populations were also treated with an anti-CD3 antibody (BD Biosciences), which induces Ca2+ mobilization in T but not B cells. For statistical analysis, area under curve values were calculated and compared using Fisher's exact test. P < 0.05 was considered statistically significant.
Effect of IL-4R Signaling on BCR-dependent Ca2+ Mobilization in Memory B Cells.
As treatment of naive B cells with IL-4 had no effect on the expression of genes related to BCR signaling (see above; see Fig. 5), modulation of BCR signals by IL-4 was studied in memory B cells only. Memory B cells from four healthy donors were purified and cultured in supernatant from LPS-stimulated PBMCs for 24 h in the presence or absence of human recombinant IL-4 or an inhibitory anti–IL-4Rα antibody (Genzyme). Changes of cytosolic Ca2+ concentrations upon BCR engagement were measured and analyzed as described above.
Results And Discussion
Verification of the Identity of Purified B Cell Subsets.
CD34+ HSCs (reference 7) and CD10+ CD19+ pre-B cells (Fig. 1 B) were purified from human bone marrow, CD19+CD27− naive and CD19+CD27+ memory B cells were purified from peripheral blood and CD38+CD77+ GC B cells from tonsils and analyzed by flow cytometry (Fig. 1 B). The identity of the purified subsets was further supported by the analysis of their genotype (Table I). As opposed to memory B cells, Ig VH1-gene rearrangements amplified from pre-B cells and naive B cells were virtually devoid of somatic mutations. Consistent with ongoing selection for the expression of a functional Ig heavy chain in bone marrow pre-B cells, nonproductive VH1-gene rearrangements (either due to loss of reading frame or rearrangement of a pseudogene of the VH1 family) were overrepresented in the isolated pre-B cell population (Table I). Prior to the SAGE analysis, specific fragments of the Cμ-, Cγ1-, Cκ, and VpreB genes were amplified from pre-B cells, naive, and memory B cells by RT-PCR, which was normalized for β2-microglobulin (Fig. 1 A). As expected, expression of VpreB is confined to pre-B cells, which, in turn, lack expression of Igκ light chains and IgG1 heavy chains. Cγ1 transcripts were detected in memory but not pre-GC B cells (Fig. 1 A). The identity of the purified B cell subsets was also confirmed retrospectively by the expression pattern of a set of subset-specific markers within the SAGE-libraries (Table II).
Verification of Quantitative Accuracy of SAGE Profiles.
In the analysis of SAGE profiles for CD34+ HSC, pre-B, naive, GC, and memory B cells, we identified a particular gene expression pattern, which involves positive and negative regulatory BCR signaling molecules (see references in Fig. 2). To corroborate the quantitative differences in the expression of BCR signaling molecules as observed by SAGE, semiquantitative RT-PCR was performed for 22 positive and 19 negative regulatory BCR signaling molecules (Fig. 3). For all 41 genes tested, the amounts of the amplification product mirrored the SAGE tag counts in the libraries for naive and memory B cells. Moreover, the expression of 10 costimulatory and 10 inhibitory surface molecules implicated in the propagation or attenuation of BCR-dependent signals was analyzed at the protein level by flow cytometry (Fig. 4). For all 20 surface molecules, the FACS® data correlated with SAGE tag counts, indicating that the large majority of SAGE data are indeed reproducible by alternative methods.
Opposing Regulation of BCR Signaling in Naive and Memory B Cells.
In a comprehensive search throughout the five SAGE profiles and the UniGene database, we identified 148 molecules involved in positive (97 genes) and negative (51 genes) regulation of BCR signaling present in at least one SAGE library. Whereas most of the genes that positively regulate BCR-dependent activation signals are expressed at high levels in pre-B, GC, and memory B cells, this was not the case for naive B cells (Fig. 2, A and B, and Figs. 3 and 4). In many cases, expression of positive BCR signaling molecules was either missing in the SAGE library for naive B cells or reduced to expression levels as in CD34+ HSC. Conversely, inhibitory molecules were expressed either exclusively or predominantly in naive B cells (Figs. 2 C, 3, and 4).
Concomitant down-regulation of mediators of BCR-related activation together with increased expression of inhibitory molecules in naive B cells (Fig. 2, A–C) suggests that an elevated signaling threshold prevents naive B cells from being inappropriately activated upon antigen encounter. There are, however, some exceptions to this seemingly uniform picture: for instance, IKAROS, which appears to reduce BCR-dependent B cell activation (18) is expressed at higher levels in pre-B cells and memory B cells as compared with naive B cells. Also, SAGE tags matching the LYN gene, which is critical for ITIM-dependent negative signaling, and the IκBα gene, encoding an inhibitor of nuclear factor (NF)-κB, were found most frequently in pre-B cells, while expression levels in naive, GC, and memory B cells are similar. Many of the signaling molecules included in this study are also involved in signaling pathways that are not related to the BCR. This applies in particular to downstream kinases (e.g., p85α, JAK3, p110, LCK, FAK, IKK, AKT, HPK1, p115, PKCμ, PKCβ, PKA; Fig. 2) and receptors that are not specific for the B lineage, whose intra-cellular signals may converge with those of the BCR (SLAM, PDGFRα, CD38, CD27, FcRH1, HRH1, CD86, NGFR, CD36, CD74, CD66, CD31, LAIR1, CD5, CD33, LLIR, LIRB1, LIRB2, LIRB5, PIRβ; Fig. 2).
However, mouse mutants for some of these genes exhibit a particular B cell phenotype, and many receptor molecules are involved in either ITAM- or ITIM-dependent signaling, which allows a prediction of their role in either amplification or attenuation of BCR-dependent activation signals. The five SAGE libraries also identified a number of components of the proximal BCR-signaling complex, which have an unambiguous function in (pre)-B cells. These “classical” BCR signaling molecules include BLK, BLNK, BTK, Igα, Igβ, SYK, and CD19, whose expression levels are consistently lower in naive B cells as compared with other B cell subsets (Fig. 2 A and B, and Figs. 3 and 4). To the contrary, well characterized inhibitors of BCR-induced activation signals including the protein tyrosine phosphatases (PTPs) SHP1 and SHIP (19), the PTK CSK, the PTK-associated ubiquitin ligase CBL, and the ITIM-bearing surface receptors CD22, CD32 and CD72 are expressed at high levels in naive B cells but not in pre-B, GC, or memory B cells (Figs. 2 C, 3, and 4).
Facilitated responsiveness of memory B cells to BCR engagement was suggested from earlier findings, namely the expression of the activating coreceptors CD21 (20) and CD27 (21) together with increased expression levels of the costimulatory molecules CD80 and CD86 (22) and in vitro experimentation, which demonstrated that memory cells have higher propensity to undergo activation-induced terminal differentiation than naive B cells (23). To date, a genome-wide analysis of BCR signaling molecules in pre- and (post-) GC B cells is missing. As shown in Fig. 2, A and B, sensitization of memory B cells to BCR-dependent activation signals is not only related to upregulation of surface molecules such as CD21, CD27, CD80, and CD86 but involves virtually all levels of intracellular signal transduction including transmembrane receptors, kinases, linker molecules, phosphatases, Ca2+ channels and transcription factors.
In particular, it was not known that BCR signals in naive B cells can be specifically silenced by the prominent expression of ITIM-bearing molecules belonging to the emerging group of inhibitory immunoglobulin superfamily (IgSF) receptors (Figs. 2 C, 3, and 4). In the following, we describe the expression pattern of inhibitory and stimulatory BCR signaling molecules in naive and memory B cells more in detail, dividing these molecules into functional groups.
Regulation of the Expression of Constituents of the BCR in Human B Cell Subsets.
In mature B cells, the BCR is composed of surface Ig, the coreceptors CD19, CD21, and CD81 and the Igα- and Igβ-signaling chains (1). In naive B cells, mRNA levels of BCR-related molecules are either moderately (Igβ, CD19, CD81; Fig. 2 B) or substantially (Igα, CD21; Fig. 2 A) lower than in other B cell subsets.
Also, expression levels of Ig genes are lower in naive B cells as compared with other B cell subsets: in the SAGE library for naive B cells, we found 143 tags matching to the Cμ gene (pre-B cells: 826; GC B cells: 329; memory B cells: 450), 29 tags for Vκ- and Cκ-genes (pre-B cells: 5; GC B cells: 54; memory B cells: 101) and 199 tags for Cλ-genes (pre-B cells: 10; GC B cells: 447; memory B cells: 1,337). Instead of Igκ- and Igλ- light chain genes, pre-B cells mostly express VpreB (287 tags) and λ-5 (1,975 tags) surrogate light chains. Higher levels of Igκ gene expression in memory than in naive B cells is in agreement with a previous study, in which Igκ mRNA levels were 3- to 11-fold higher in memory as compared with naive B cells (24). As expected, SAGE tags matching to Cγ genes are frequent in memory (1,929 tags) and GC B cells (217 tags), but virtually missing in the libraries for naive (13 tags) and pre-B cells (16 tags).
TNF Receptors May Act as Costimulatory Molecules in GC and Memory B Cells.
The genes downregulated in naive B cells include members of the TNF receptor superfamily: in contrast to GC and memory B cells, CD40 (TNFRSF5), CD27 (TNFRSF7), and NGFR (TNFRSF16) are either missing or expressed only at low levels in naive B cells (Fig. 2, A and B). While CD40-engagement is known to cooperate with IL-4R- and BCR-dependent signals during the GC reaction (25), ligation of the “memory-specific” receptor CD27 (21) by CD70 increases responsiveness of memory B cells to BCR signals and induces plasma cell differentiation. Also, three recently identified members of the TNF receptor superfamily are differentially expressed in naive B cells compared with GC- and memory B cells: BAFF-R (TNFRSF13B) and BCMA (B cell maturation antigen; TNFRSF17) are receptors of the B cell activation factor BAFF and are expressed in GC and memory B but not naive B cells (Fig. 2, A and B, and Fig. 3). Unlike BCMA (26), BAFF-R is critical for the development of marginal zone (i.e., mostly memory) B cells and the T cell–dependent B cell response (27). Although not essential for B cell activation, BCMA was shown to induce NF-κB–activation and p38 MAP kinase-dependent proliferation in synergism with the BCR (28). A third receptor for BAFF, termed TACI, however, is, although at low level, only expressed in naive B cells (Figs. 2 C and 3) and acts as a negative regulator of BCR-dependent B cell activation (29). CD30 ligand (TNFSF8), an inhibitor of B cell costimulation and class switch recombination (CSR; reference 30), is expressed in naive B cells but virtually missing in GC- and memory B cells Figs. 2 C, 3, and 4). In this regard, it is notable that SWAP70, which promotes both B cell activation and CSR (31), is expressed reciprocally with CD30 ligand in naive and (post) GC B cells (Fig. 2 B).
Specific Expression of Inhibitory Ig Superfamily Members in Naive B Cells.
Many of the inhibitory receptors, which we find expressed at high levels in naive B cells belong to the Ig superfamily (IgSF). The inhibitory IgSF molecules are predominantly or exclusively expressed by naive B cells (Fig. 2 C) and typically carry one or more ITIMs within their cytoplasmic tail. Negative regulatory IgSF molecules specifically or predominantly expressed in naive B cells include SIgLec5, SIgLec6, and SIgLec8, members of the sialic acid binding Ig-like lectin-like family, and LIRB1, LIRB2, and LIRB5, which belong to the B group of leukocyte Ig-like receptors (collectively termed CD85, Figs. 3 and 4). Also, ITIM-bearing IgSF molecules, the lectin-like immunoreceptor LLIR, the paired Ig-like receptor PIRβ, as well as CD22, CD31, CD32/FcRγII, the biliary glycoprotein CD66 are expressed by naive B cells and, if at all, only at reduced levels in GC and memory B cells (Figs. 2 C, 3, and 4). This also applies to the newly identified IgSF molecules PD1 (32), G6B (33), and CMRF35H (34). Also, ITIM-bearing but a member of the C-type lectin family, the CD72 molecule is highly expressed in naive B cells (Figs. 2 C, 3, and 4). On the other hand, its antagonistic ligand CD100 can relieve CD72-mediated inhibition of BCR-signals and is predominantly expressed in memory B cells (Figs. 2 B, 3, and 4). Within the IgSF, a group of Fc receptor homologues was recently identified, which comprises positive and negative regulatory coreceptors based on whether they harbor ITAMs or ITIMs within their cytoplasmic tail (35). Like many other inhibitory IgSF receptors, the ITIM-bearing FcRH2/SPAP1 (35) was only found in naive B cells (Fig. 2 C), whereas its ITAM-carrying homologue, FcRH1 was expressed in GC and memory but not naive B cells (Fig. 2 A).
Cytokine Receptor Signaling in Naive and Memory B Cells.
The B cell–homing chemokine receptor BLR1, also termed CXCR5, can cooperate with the BCR by stimulation of Ca2+ influx (36) and is stronger expressed in GC and memory B cells as compared with their naive precursors (Fig. 2 B). Naive B cells also differ from GC B cells and memory B cells in that they lack expression of the signal transducer for IL-6 (CD130 or gp130) and the receptor for IL-4 (CD124). gp130 has been implicated in post-GC development of B cells, as mice expressing a dominant negative form of gp130 exhibit a marked reduction of Ig production (37). Engagement of IL-4R can augment activation signals through the BCR (for a review, see reference 38; see below).
Regulation of BCR-Downstream Linker Molecules and Kinases in Human B Cell Subsets.
A large group of activating linker molecules and PTKs contributes to propagation of activation-stimuli within the distal BCR-signaling cascade. Without exception, activating linker molecules identified in the SAGE libraries, including BLNK, BAM32, GRB2, SOS1, SHC1, GAB1, GAB2, BRDG1, NCK, and BANK, were up-regulated in GC and memory as compared with naive B cells (Fig. 2, A and B). In contrast, inhibitory linker molecules including p62DOK (39) and the SRC-like adaptor proteins SLAP1 and SLAP2 (40) inhibit BCR downstream signals, predominantly in naive B cells (Fig. 2 C).
Activating kinase molecules, as far as identified in the SAGE libraries, are expressed at higher levels in (post) GC B cells as compared with their naive precursors (Fig. 2, A and B). PI-3 kinases p85α and p110 have in common that they may activate AKT/PKB (41) and BTK (42), while activated BTK can induce degradation of the NF-κB inhibitor IκBα by phosphorylation through IKK (43). Only moderately up-regulated in GC and memory B cells, the MAP kinase HPK1 (44) and the PTK PYK2 (45) can augment BCR signals, the former through interaction with BLNK, the latter with BRDG1 (Fig. 2 B). Interestingly, PYK2-deficient mice lack splenic marginal zones, which are thought to be mainly composed of memory B cells (45). Not specific for the B cell lineage, but expressed in GC and memory B cells (Fig. 2 A), the receptor of platelet derived growth factor PDGFRα can act as a stimulatory PTK and promote proliferation (46).
Among the inhibitory kinases, the CD45-antagonist CSK (47) and the PKCs β (48) and μ (49) acting as inhibitors of BTK and SYK, respectively, are expressed at higher levels in naive B cells as compared with GC and memory B cells (Fig. 2 C). However, expression levels of LYN, a key mediator of ITIM-dependent negative signaling is expressed at similar levels throughout all B cell subsets. Specific inhibition of the tyrosine kinase BTK represents another level of BCR-signal attenuation, at which the IBTK (50) and SAB/SH3BP-5 (51) molecules are acting, mostly in naive B cells (Fig. 2 C).
Regulation of PTPs during B Cell Development.
The PTP CD45 represents a critical component of the proximal BCR-signaling cascade (52). Together with its antagonist, CSK, CD45 sets the threshold for antigen receptor signals in B and T cells by reversing CSK-mediated phosphorylation of negative regulatory tyrosine-based motifs in src-family PTKs (53). As opposed to its antagonist CSK, the CD45 gene is higher expressed in memory B cells, GC B cells, and pre-B cells as compared with naive B cells (Fig. 2, A and C). From the five SAGE libraries, multiple SAGE tags matching to the CD45 gene were retrieved, which most likely reflects the expression of multiple splice variants of the CD45 gene. Consistent with the notion of BCR-signal inhibition in naive B cells, the inhibitory PTPs SHP1, SHIP, and the recently identified PTP-PEST molecule (54) are expressed at high levels in naive but not GC or memory B cells (Fig. 2 C). Another PTP, encoded by the PTEN gene, inhibits the activation of BTK in B cells (38), is frequently inactivated by somatic mutation in advanced malignancies resulting in uninhibited PTK activity and expressed at higher levels in naive than (post) GC B cells (Fig. 2 C).
Role of Transcription Factors in the Regulation of BCR Signals.
The genes up-regulated in memory and GC B cells include classical transcriptional activators of Ig genes including OCT2, OBF1, NF-κB1, and NF-κB2 (Fig. 2, A and B). These genes are involved in autocatalytic loops initiated from the BCR through transcriptional activation of Igα (by OCT2; reference 55), synergism with BTK (by OBF1; reference 56), and amplification of BCR-dependent anti-apoptotic signals and proliferation-stimuli (by NF-κB; reference 43). BCR engagement and subsequent NF-κB activation are linked by the protooncogene BCL10 (57), which is expressed at high levels in pre-B and memory but not naive B cells (Fig. 2 B). In GC and memory B cells, OCT2 was found up-regulated together with its transcriptional target gene CD36 (Fig. 2 B). Stage-specific expression of CD36 could be meaningful because based on the dependence of CD36 expression on transcriptional activation by OCT2, it was speculated that OCT2 could regulate B cell differentiation through CD36 (58). The members of the ETS family of transcription factors, FLI-1 (59), PU.1, and SPI-B (60) have recently been demonstrated as critical components of BCR-dependent activation-signaling pathways. FLI-1, PU.1, and SPI-B are expressed at high or intermediate levels in GC and memory B cells but are missing in the SAGE library for naive B cells (Fig. 2 A).
Relief of BCR Signal Inhibition by Engagement of IL-4R in (Post) GC B Cells.
Recent data indicate that enhancement of BCR-dependent B cell activation by IL-4 at least in part results from the relief of inhibitory signals through negative regulatory receptor molecules (61). In agreement with these findings, we observe specific expression of ITIM-bearing inhibitory IgSF receptors on naive B cells (Figs. 2 C and 3), which also lack expression of IL-4R (Figs. 2 B, 3, and 4). Conversely, the SAGE libraries for GC B cells and memory B cells show a correlation between the expression of IL-4R and the absence of ITIM-bearing and other inhibitory receptors (Fig. 2 C). IL-4–dependent changes of gene expression are largely mediated by STAT6 (38), which we find expressed at higher levels in GC (17 tags) and memory B cells (33 tags) than in naive B cells (4 tags).
Thus, antigen-encounter during the GC reaction, up-regulation of IL-4R, and initiation of IL-4–dependent signals might result in a far-reaching phenotypic change, namely transcriptional silencing of a large group of inhibitory receptors that are abundantly expressed in naive B cells. Whether and to which extent the inverse regulation of responsiveness to BCR cross-linking in naive and memory B cells results from differential IL-4R signaling was tested in a cell culture experiment (Fig. 5). Naive and memory B cells were purified from peripheral blood and cultured either in medium alone, or with human recombinant IL-4. In another set of experiments, memory B cells were preincubated with IL-4 and subsequently treated with a neutralizing anti–IL-4Rα antibody. After 48 h, the cells were subjected to RNA isolation and subsequent semiquantitative RT-PCR analysis for mRNA expression of the positive regulatory BCR signaling molecules BLK, BTK, BLNK, and SYK, their inhibitors CSK, SHIP, and SHP1 and negative regulatory IgSF receptors including LIRB1, LIRB2, LIRB5, SIgLec5, SIgLec8, LAIR1, and CD66 (Fig. 5).
In naive B cells, presence or absence of IL-4 did not affect the expression of positive or negative regulatory BCR signaling molecules. In peripheral blood memory B cells, however, complete deprivation from IL-4 in cell culture medium for 48 h resulted in a concomitant decrease of mRNA levels of BLK, BTK, BLNK, and SYK with markedly increased expression levels of inhibitory IgSF receptors (Fig. 5) as compared with expression levels in ex vivo analyzed peripheral blood memory B cells (for comparison, see amplification products in Fig. 3, at 32 cycles for LAIR1 and 28 cycles for the other genes studied here). Loss of positive and gain of negative regulatory BCR signaling molecules upon withdrawal of IL-4 in peripheral blood memory B cells indicates that gene regulation through IL-4R requires continuous presence of its ligand. However, serum levels of IL-4 are low in healthy individuals (i.e., <0.1 pg/ml; reference 38), which suggests that peripheral blood memory B cells are able to respond also in the presence of low concentrations of IL-4. It is indeed conceivable, that memory B cells have acquired higher responsiveness to IL-4 (e.g., by up-regulation of IL-4R) during maturation within GCs, in which IL-4–producing TH2-cells are highly concentrated (62). Treatment of memory B cells with IL-4 at high concentrations induced a slight increase of mRNA levels of BLK, BTK, BLNK, and SYK but a marked reduction of mRNA levels of the inhibitory IgSF receptors LIRB1, LIRB2, LIRB5, SIgLec5, SIgLec8, LAIR1, and CD66. IL-4 treatment did not affect expression levels of other inhibitory BCR signaling molecules in memory B cells including CSK, SHIP, and SHP1, which are expressed at constitutively lower levels in memory than in naive B cells (Fig. 5). Consistent with low-level expression of IL-4R in naive B cells (Figs. 2 A, 3, and 4), IL-4 can induce transcriptional repression of inhibitory IgSF receptors in memory but not naive B cells. Inhibition of IL-4R–dependent signals by a neutralizing antibody, which was added after preincubation with high concentrations of IL-4, only slightly reduced mRNA levels of the positive mediators of BCR signaling BLK, BTK, BLNK, and SYK and had no effect on expression levels of their inhibitors CSK, SHIP, and SHP1. However, inhibitory IgSF receptors including LIRB1, LIRB2, LIRB5, SIgLec5, SIgLec8, LAIR1, and CD66 were markedly up-regulated in memory B cells in the presence of the neutralizing anti–IL-4R antibody and reached similar expression levels as in naive B cells. Thus, down-regulation of inhibitory IgSF receptors in memory B cells largely depends on IL-4R engagement.
Clustering of IgSF Genes Overexpressed in Naive B Cells to a Region on Chromosome 19.
As expression levels of other negative regulatory BCR signaling molecules such as CSK, SHIP, and SHP1 remain stable in the presence or absence of IL-4R signaling, transcriptional repression induced by IL-4 in memory B cells seems to be specific for inhibitory IgSF receptors. In this regard, it is notable that the genes coding for the ITIM-bearing IgSF receptors studied here and also many inhibitory killer cell Ig-like receptors are arranged in a cluster within a 2.9 Mbp region on chromosome 19 (19q13.2–q13.4; see Fig. 5). This cluster of genes also includes the inhibitory IgSF receptors CD22, GP6 and CD33, which are highly expressed in naive but not GC and memory B cells (Figs. 2 C, 3, and 4). Therefore, it is tempting to speculate that this cluster on chromosome 19 harbors a large number of IgSF genes sharing a common mechanism of transcriptional regulation in that they can be silenced by IL-4R–dependent signals.
Differential responsiveness to IL-4 in naive and memory B cells suggests that human B cells may acquire sensitivity to IL-4 during the GC reaction, while IL-4R–dependent signaling itself facilitates transduction of signals initiated from the BCR. Further studies are needed to identify other mediators of the relief of the “inhibition-phenotype” in human B cells after antigen-encounter in the GC.
Regulation of BCR Responsiveness by IL-4R Signals in Memory B Cells.
Having shown that IL-4 can down-regulate inhibitory IgSF receptors in memory but not naive B cells (Fig. 5), we studied the effect of IL-4R signals on BCR-dependent Ca2+ mobilization in memory B cells. After preincubation of memory B cells in PBMC-conditioned medium in the presence or absence of human recombinant IL-4 or an inhibitory anti–IL-4Rα antibody, memory B cells were challenged with anti–human IgG + IgM F(ab′)2 fragments and Ca2+ flux was measured. Addition of IL-4 beyond physiological concentrations had no effect on BCR responsiveness of memory B cells (Fig. 6 C). However, inhibition of IL-4Rα resulted in a decrease of the peak size and a rapid decline of the calcium signal, which suggests that integrity of BCR responsiveness in memory B cells requires signals through the IL-4R (Fig. 6 C). This was expected, as inhibition of IL-4R signaling resulted in reexpression of inhibitory IgSF molecules in memory B cells (Fig. 5). The failure of supraphysiological IL-4 concentrations to further augment BCR-dependent Ca2+ mobilization in memory B cells suggests that already low IL-4 concentrations (e.g., as in human serum) may be sufficient to maintain full BCR responsiveness to antigen.
Regulation of BCR Responsiveness by the IgSF Receptor LIRB1 in Naive and Memory B Cells.
As IL-4 can repress inhibitory IgSF receptors in memory but not naive B cells, we next investigated the direct consequences of IgSF receptor signaling on the responsiveness of the BCR in both B cell subsets. As an example for inhibitory IgSF receptors, we chose LIRB1, which is most prominently expressed in naive B cells (45 tags; Figs. 2 C, 3, and 4) but missing in the SAGE profiles for GC and memory B cells (Figs. 2 C, 3, and 4). To determine how LIRB1 can modify BCR responsiveness, changes of cytoplasmic Ca2+ concentration in response to BCR engagement were measured. To this end, naive and memory B cells were preincubated in supernatants conditioned by LPS-stimulated PBMCs in the presence or absence of agonistic (clone GVI/75 cross-linked by goat anti–mouse serum) or antagonistic (clone HP-F1 noncross-linked) antibodies against LIRB1. After LPS treatment, PBMCs release high levels of soluble MHC class I molecules, which can act as natural ligand for LIRB1 (13, 14). After preincubation, naive and memory B cells were stimulated with anti–human IgM F(ab′)2 and anti–human IgG + IgM F(ab′)2 fragments, respectively, and changes of intracellular Ca2+ concentrations in response to BCR engagement were measured by confocal microscopy. Whereas treatment of naive and memory B cells with an anti-CD3 antibody had no effect on intracellular Ca2+ levels (not shown), stimulation with anti–human IgM F(ab′)2 and anti–human IgG + IgM F(ab′)2 fragments increased the concentration of cytoplasmic Ca2+ in naive and memory B cells (Fig. 6). While stimulation (GAM-cross-linked GVI/75 antibody) and inhibition (noncross-linked HP-F1 antibody) of LIRB1 significantly affected the peak amplitude and duration of the calcium signal in naive B cells (Fig. 6 A), agonistic and antagonistic antibodies had no significant effect on BCR-dependent Ca2+ mobilization in memory B cells (Fig. 6 B). As expected, the overall signal intensity of BCR engagement was significantly higher in memory (peak [Ca2+]i concentration 220 nmol/L) than in naive B cells (peak [Ca2+]i concentration 140 nmol/L). Also, steady-state levels of cytoplasmic Ca2+ were higher in memory (∼100 nmol/L) as compared with naive B cells (∼50 nmol/L). Interestingly, occupancy of LIRB1 by the inhibitory antibody raised Ca2+ mobilization in response to BCR engagement to similar levels as in memory B cells (Fig. 6, A and B). That inhibition of LIRB1 alone could nearly restore BCR-responsiveness in naive B cells in vitro does not rule out an important contribution of other inhibitory IgSF receptors identified in this study to silencing of BCR signals in vivo. This is supported by a number of animal models, in which inactivation of a single gene similarly resulted in profound changes of BCR responsiveness (63–66). That other inhibitory IgSF receptors, which were found specifically expressed in naive B cells by SAGE, RT-PCR and flow cytometry (Figs. 2 C, 3, and 4), have a similar effect on BCR signals in naive B cells appears likely but remains to be established. Conversely, engagement of LIRB1 by the agonistic cross-linked antibody further suppressed and shortened the calcium signal in naive B cells initiated by BCR ligation (Fig. 6 A). None of these effects could be seen in memory B cells (Fig. 6 B), which is consistent with absence of LIRB1 expression in memory B cells (Fig. 2 C, 3, and 4).
We conclude that IL-4 is essential to down-regulate inhibitory IgSF receptors in antigen-experienced B cells. As inhibitory IgSF receptors can indeed silence BCR signals, their repression by IL-4 is required to maintain an intense signaling capacity and responsiveness of the BCR to antigen.
Unlike naive or memory B cells, pre- and GC B cells exhibit a particular propensity to apoptosis in that they enter a readily initiated apoptosis program unless they are rescued by survival signals from the (pre)-BCR. Intensified activation signaling in the surviving pre and GC B cells can, therefore, be an effect of selection. Neither naive nor memory B cells are directly involved in selection processes. In the case of naive and memory B cells, the distinct gene expression pattern of activating and inhibitory BCR signaling molecules is probably largely B cell autonomous and marks a constitutive difference between the two subsets.
In naive B cells, attenuation of BCR signals (e.g., by inhibitory IgSF receptors) may prevent these cells from premature activation upon antigen encounter or represent a safeguard against expansion of autoreactive clones. Conversely, amplification of BCR signals may contribute to facilitated responsiveness to BCR engagement and longevity of memory B cells (67). Sustained expression of activating signaling molecules initiated from the BCR seems to be a characteristic of memory B cells as the expression levels of activating BCR signaling molecules in memory B cells also tend to be higher than those in pre-B or GC B cells (Fig. 2, A and B). High expression levels of costimulatory molecules may reflect activation in pre-B and GC B cells and sensitization and increased responsiveness to antigen in memory B cells.
A constitutively active signaling machinery in memory B cells may not only increase responsiveness to but also confer independence from BCR cross-linking by antigen. Indeed, the capacity to persist independently from the immunizing antigen was thought to define “true immunological memory” (68) as opposed to survival within a chronic immune response, in which the immunizing antigen can be retained and presented for extended periods of time by follicular dendritic cells within GCs. To ensure persistence of B cell memory, perpetuation of a B cell–autonomous “maintenance signal” would be required in the former but not in the latter case. Recent data suggest that the continuous presence of the immunizing antigen is indeed dispensable for the survival of memory B cells provided that a functional BCR is expressed on the cell surface to initiate the “maintenance signal” (68). The concept that the BCR in memory B cells has distinct intrinsic signaling properties is further supported by recent data on a burst-enhancing role of the membrane spanning region of IgG in memory B cells (69). Earlier findings suggested that, unlike IgM (three cytoplasmic residues), the cytoplasmic tail of IgG1 (28 residues) as such can contribute to signal transduction through the BCR, presumably involving a tyrosine-based motif within Cγ1 (70).
Although both naive and memory B cells critically depend on continuous survival signals from a functional BCR (2), we show in a genome-wide gene expression analysis that BCR-dependent signals in naive and memory B cells differ fundamentally from each other. Conceivably, these differences may owe to some extent to IL-4–dependent ablation of negative regulatory signals from inhibitory receptor molecules. However, IL-4R signals, although required for transcriptional suppression of inhibitory IgSF receptors and for integrity of BCR-dependent Ca2+ mobilization in memory B cells, had no effect on their naive precursors. Therefore, the mechanism establishing restraint of BCR signals in naive B cells and its relief upon antigen-encounter within the GC remains elusive and awaits further investigation.
We would like to thank Stefanie Jauch for excellent technical assistance and Dr. Miguel López-Botet, Universitat Pompeu Fabra, Barcelona, Spain, for kindly providing us with the HP-F1 antibody.
H. Wang is supported by the DAAD Bioscience Program, M. Müschen is supported by the Cancer Research Institute, New York, NY, through the Tumor Immunology Program. This work was supported by the Deutsche Forschungsgemeinschaft through the Emmy-Noether-Programm (to M. Müschen; MU1616/2-1 and MU1616/2-2), Harold and Lelia Y. Mathers Foundation (S.M. Wang and J.D. Rowley), SFB 502 and a Heisenberg award to R. Küppers, by the Köln Fortune Program/Faculty of Medicine, University of Cologne, and the German José Carreras Leukemia Foundation (grant to M. Müschen).
Abbreviations used in this paper: BCR, B cell receptor; GC, germinal center; IgSF, Ig superfamily; ITAM, immunoreceptor tyrosine-based activation motif; HSC, hematopoietic stem/progenitor cell; ITIM, immunoreceptor tyrosine-based inhibitory motif; MAP, mitogen-activated protein; PKC, protein kinase C; PTK, protein tyrosine kinase; PTP, protein tyrosine phosphatase; SAGE, serial analysis of gene expression.