|Home | About | Journals | Submit | Contact Us | Français|
Thomas L. Rothstein, The Feinstein Institute for Medical Research, Manhasset, NY 11030, USA.
Karen J. Repetny, Millennium Pharmaceuticals, Cambridge, MA 02139, USA.
Upon stimulation of mature B cells, class switch recombination (CSR) can alter the specific immunoglobulin heavy chain constant region that is expressed. In a tissue culture cell line, we previously demonstrated that inhibition of late SV40 factor (LSF) family members enhanced IgM to IgA CSR. Here, isotype specificity of CSR regulation by LSF family members is addressed in primary mouse splenic B cells. First, we demonstrate that LBP-1a is the prevalent family member in B lymphocytes. Second, we demonstrate by ChIP that LBP-1a binds genomic sequences around mouse switch regions (S) in an isotype-specific manner, in accordance with computational predictions: binding is observed to Sμ and Sα, but not to the tested Sγ1, regions. Importantly, binding of LBP-1a is tightly regulated, with occupancy at genomic S regions dramatically decreasing following LPS stimulation. Finally, the consequence of DNA-binding by LBP-1a is determined using bone marrow chimeric mice in which LSF/LBP-1 activity is inhibited in hematopoietic lineages. Upon in vitro stimulation of such primary B-cells, CSR occurs with a higher efficiency to IgA, but not to IgG1. These results are supportive of a model whereby LBP-1a represses CSR in an isotype-specific manner via direct interaction with switch regions involved in the recombination.
Class switch recombination (CSR) involves genomic DNA recombination in mature B cells that alters expression of the immunoglobulin heavy (H) chain constant region. The consequence is diversification of antibody effector function while preserving antigen specificity. The nonhomologous DNA recombination occurs within or near switch (S) regions, which consist of tandemly repeated DNA motifs located upstream of each H chain constant coding region. Antigens, bacterial products, and cytokines induce switching, with each distinct combination of signals generating expression of specific isotype(s) [1–3]. Extracellular signaling drives switching to a particular isotype by activating transcription from a promoter located 5’ of the respective S region. Such germline transcription is required to facilitate accessibility of S region DNA for CSR [4;5]. The extent to which the concurrent, enhanced chromatin accessibility in S regions promotes CSR is unclear .
Although molecular mechanisms for CSR have been extensively studied, little is known about non-transcriptional regulatory pathways, and in particular, means to prevent inappropriate CSR. As the predominant sites of DNA recombination , the S regions themselves provide intriguing regulatory targets. Indeed, several proteins specifically recognize S regions and activate CSR [1;6;7]. Of these, NF-κB/p50 can promote isotype-specific CSR without influencing germline transcription [8;9].
As we demonstrated previously, another class of transcription factors, the Late SV40 Factor (LSF) family members, binds switch region sequences in vitro . In the I.29μ B cell line, which undergoes IgM to IgA isotype switching in vitro, LSF or other family members represses CSR without altering levels of germline transcripts. This provided the first instance of non-transcriptional repression of CSR by a specific DNA-binding protein . The LSF family of transcription factors includes three paralogs, which share a DNA recognition motif, CTGG N6 CTT/GG. The two ubiquitous members of the family, LSF and LBP-1a/b (leader binding protein-1a/b), are 72% identical in aa sequence and may be functionally redundant in many tissues .
Here we report that LBP-1a is the dominant family member expressed in mouse primary B lymphocytes. A high density of binding sites is predicted only in particular S regions. Consistent with these predictions, in resting B lymphocytes, LBP-1a specifically binds to particular S regions (e.g. Sα) but not demonstrably to others (e.g. Sγ1). Notably, LBP-1a is released from bound S regions after LPS stimulation. Finally, inhibition of LBP-1a function leads to enhanced expression of only specific isotypes (IgA but not IgG1) when mouse B lymphocytes are stimulated to undergo CSR in vitro, correlating with binding of LBP-1a to the relevant S region.
In order to distinguish LSF and LBP-1a, anti-peptide antibodies were produced against divergent regions (Fig. 1A). The antibody specificities and sensitivities were tested against recombinant proteins (Fig. 1B) and mammalian cell extracts in which each protein was independently overexpressed (Fig. 1C). In both instances, each antibody was highly specific for its cognate protein. The LBP-1a antibody recognizes multiple protein species from untransfected 293T extracts (Fig. 1C); due to alternative splicing, the cellular LBP-1a gene can encode two major products: LBP-1a and the larger LBP-1b . LBP-1b is expressed in a tissue-specific manner , and is not evident in primary B cells (Fig. 1D).
With these new reagents, we determined the relative abundance of LSF and LBP-1a in primary B-cells. By comparing the reactivity of each antibody against whole cell lysates from mouse splenic B-cells (Fig. 1D, right panels) to its reactivity against a standard curve of purified recombinant His-LSF or His-LBP-1a (Fig.1D, left panels), we concluded that primary mouse B cells contain at least 5-fold more LBP-1a (on the order of 104 molecules/cell) than LSF.
We previously demonstrated in vitro that recombinant LSF, as well as related proteins in mouse splenic cellular extracts, could bind specific sequences within Sμ, Sα, and Sε . S regions can be clustered into two groups: Sμ, Sε, and Sα contain repeated pentameric sequences, whereas Sγ1, Sγ2a, Sγ2b, and Sγ3 contain 49–52 base-pair repeats [2;3]. We computationally analyzed the likelihood of LSF/LBP-1a binding sites within each S region. In some (Sμ, Sε, Sα, and Sγ3), numerous LSF/LBP-1a recognition sites were predicted; in others (Sγ1, Sγ2a, and Sγ2b), exceedingly few (Fig. 2A and Fig. S1A in supporting material). A similar pattern was evident when human S regions were analyzed (Fig. S2).
We tested these predictions by ChIP assays in primary murine splenic B-lymphocytes, focusing on Sμ, which is operative in all CSR, and Sα and Sγ1, which are target switch regions of the two classes of repeats. Binding of LBP-1a, the more abundant LSF family member in B-cells (Fig. 1D), was monitored. Amplification of specific switch regions in the genomic DNA was analyzed by semi-quantitative PCR analysis (see Materials and Methods); the highly repetitive nature of these genomic sequences prevented design of real-time PCR primers. Binding in vivo was robust at both Sμ and Sα (Fig. 2B), where the amplicons were located at least 1.5 kb and 3.5 kb, respectively, downstream of transcriptional regulatory elements (the intronic promoters and the enhancer region), and either at the border of or well within S region sequences. However, as expected given the low density of predicted high affinity binding sites (roughly one per 3500 bp) LBP-1a binding was not detectable around Sγ1 in the region of the amplicon (Fig. 2B). Similarly, LBP-1a occupied Sε and Sγ3, the respective amplicons being at least 2.5 kb and 3.2 kb downstream of the intronic promoters, with minimal or no occupancy of Sγ2a and Sγ2b in the regions probed (Fig. S1B in supporting material). The experimental findings at all heavy chain gene loci are therefore consistent with predictions of LSF/LBP-1a binding sites within these S region sequences.
Functionality of LBP-1a association with S regions in vivo was tested initially by assaying binding upon stimulation of primary B-lymphocytes. In previous studies, inherent DNA-binding activity of LSF/LBP-1a to S sites in mouse splenic extracts, assayed by EMSA, decreased after stimulation to undergo CSR . Using the specific antibody (Fig. 1), levels of LBP-1a were measured in whole cell extracts of purified splenic B-lymphocytes stimulated by LPS for increasing amounts of time (Fig. 2C). Surprisingly, LBP-1a levels (and LSF protein levels; data not shown) substantially increased by 24 h after stimulation. In parallel LPS-stimulated cultures, we directly measured binding of LBP-1a to genomic S regions by ChIP. Occupancy at both Sμ (Fig. 2D) and Sα (Fig. 2E) was readily apparent at 0 and 10 h, but markedly decreased at 24 and 48 h post-stimulation. Thus, although protein levels increased at late time points after stimulation, binding decreased.
The decrease in DNA-binding with LPS stimulation is likely due to direct modification of LBP-1a. The closely related paralog, LSF, is directly targeted by mitogenic signal transduction pathways in multiple cell types, altering LSF DNA-binding potential, in part through modification of S291 [14;15]. LBP-1a contains an analogous serine; we hypothesize that LBP-1a is targeted downstream of cytokine signaling cascades in resting B cells, being modified to reduce binding to S regions.
The kinetics of release of LBP-1a from S region sequences following stimulation of B cells are consistent with LBP-1a inhibiting early events in CSR. Induction of AID mRNA and some germline transcripts occurs by 12 to 24 h post-stimulation [16;17], with Iα sterile transcripts , formation of R-loops , and alterations in histone acetylation by 48 h post-stimulation . The inverse correlation between binding of LBP-1a and activation steps of CSR suggests that LBP-1a is a repressor.
In a B cell line capable of undergoing CSR only from Sμ to Sα in vitro, LSF/LBP-1a did indeed repress CSR . However, the question remained as to whether, in primary B cells, CSR in general would be subject to such inhibition, since all CSR in IgM-expressing B cells involves the Sμ switch region with which LSF/LBP-1a interacts, or if CSR only at certain heavy chain gene loci would be inhibited. To distinguish between these possibilities, we generated bone marrow chimeric mice in which a dominant negative form of LSF (LSFdn) was expressed in donor hematopoietic cells. LSF family members oligomerize with each other and bind DNA as obligate tetramers . Two aa substitutions in the DNA-binding region of LSFdn prevent DNA-binding ; due to oligomerization this is a dominant phenotype that blocks all LSF family members . This decrease in DNA-binding of endogenous LSF family members by LSFdn expression has previously been demonstrated in multiple studies, including in a murine B cell line .
Bone marrow cells from donor mice were transduced in vitro with retrovirus capable of expressing both LSFdn and GFP. Transplantation of these cells into lethally irradiated mice repopulated the recipient hematopoietic system. The resulting transduced, GFP-expressing splenic B cells were isolated and analyzed for CSR upon stimulation in vitro. Two control B cell populations were used. First, non-transduced cells in the donor bone marrow gave rise to GFP-negative splenic B cells in the same mouse, providing an ideal internal control for the GFP-positive, LSFdn-expressing cells (see Fig. S3 in supporting material for demonstration of GFP-negative versus GFP-positive populations). Second, other recipient mice were transplanted with donor bone marrow transduced with parental retrovirus, capable of expressing only GFP, in order to control for the possibility that transduction with any retrovirus would alter the efficiency of CSR.
Expression of LSFdn was first validated in the LSFdn group of mice. Isolated GFP-positive and GFP-negative populations of primary splenic B-cells were harvested and immunoblotted with antibody specific to LSF. Endogenous murine LSF was detected in GFP- cells, whereas at least 50-fold higher levels of exogenous human LSFdn were detected in GFP+ cells from the same mouse (Fig. 3A). Considering relative expression levels of LBP-1a and LSF (Fig. 1D), LSFdn expression was also at least 10-fold higher than that of LBP-1a. Even equimolar concentrations, both in vitro and in vivo, are sufficient to dramatically suppress wild type DNA-binding of LSF family members [21;22].
To determine whether this expression of LSFdn affects the frequency of CSR, purified splenic B-cells from both chimeric LSFdn mice and chimeric control mice were stimulated to undergo CSR in vitro. Unlike results in the I.29μ B cell line, in the absence of stimulation, there was no detectable IgA or IgG1 expression in B cells isolated from LSFdn mice (data not shown). That alleviation of repression by LBP-1a is not sufficient to induce CSR in primary resting B cells in the absence of signaling is consistent with extensive analyses indicating that specific activation signals are absolutely required.
Due to the differences in density of LBP-1a occupancy at switch regions Sα vs. Sγ1 (Fig. 2), and to the relatively high efficiency of switching that can be induced in vitro to these two regions, we compared switching to IgA with that to IgG1 . Cells were gated on the basis of GFP expression and surface levels of IgG1 and IgA expression after stimulation were monitored by flow cytometry (see Fig. S3 in supporting material for example). In initial experiments, comparison of the degree of CSR in BALB/cByJ B cells to that in the chimeric B cells indicated that CSR in the chimeric cells was less robust (approximately one fourth the levels of CSR), suggesting that bone marrow transplantation compromises the degree of CSR achievable in primary B cells in culture. We note that the depression in frequency of CSR in chimeric mouse B cells unfortunately precluded investigation of the inherently less efficient switching in vitro to isotypes other than IgA and IgG1.
By normalizing the percentage of transduced cells undergoing CSR to that of nontransduced cells in the same B cell population from the same mouse, we could most accurately determine the effect of LSFdn expression on CSR. The ratios for each individual mouse are plotted for expression of IgA and IgG1 (Fig. 3B and C, respectively), compiled from multiple independent experiments. These experiments included activation of B cells using multiple protocols; the effect of LSFdn on IgA CSR induced either by LPS, IL-5 and TGF-β1 (left panel) or by LPS and IL- 4 (right panel) are shown in Fig. 3B. Importantly, for chimeric control mice, splenic B cells switched expression from IgM to either IgG1 or IgA at the same frequencies, irrespective of whether or not they expressed GFP, resulting in a ratio for each isotype of 1.0. For the chimeric LSFdn-expressing mice, expression of IgG1 was also unaltered in the transduced cells (ratio of 0.94, Fig. 3C). In stark contrast, the percentage of LSFdn-expressing cells presenting surface IgA was significantly elevated (1.8-fold, P<0.005) compared to their non-transduced counterparts (Fig. 3B, left panel). A distinct cytokine cocktail appeared to result in a similar derepression of IgA expression by LSFdn (1.6-fold, P=0.067; Fig. 3B, right panel). These data demonstrate that LSF family members do not downregulate CSR in general, but only to certain isotypes (e.g. IgA).
How LBP-1a reduces CSR requires further investigation. Effects of LBP-1a, or other LSF family members, on CSR could potentially be either indirect or direct. Indirect effects might result, for instance, from LBP-1a affecting the degree of cell cycling. However, we view this possibility as unlikely for the following reasons. First, in a B cell line capable of undergoing CSR, expression of LSFdn, although similarly affecting CSR , did not detectably alter cell growth properties (data not shown). Second, since induction of IgG1 and of IgA in primary B cells both require multiple cell divisions [23;24], it is difficult to conceive of how indirect effects of LBP-1a would impact switching to one, but not to the other. Our combined findings that LBP-1a binds Sμ and Sα, but apparently not Sγ1, and that inhibition of LSF/LBP-1a increases levels of immunoglobulin class switching to IgA, but not to IgG1, instead suggest that LBP-1a inhibits CSR directly, as a consequence of LBP-1a binding S region sequences upstream of the respective heavy chain coding region.
Direct regulation of CSR by LBP-1a could result from either transcriptional or non-transcriptional mechanisms. Regulation of germline transcription by LBP-1a is unlikely, given that LSFdn did not alter the levels of germline transcripts in a B cell line in which it regulated CSR , and that the demonstrated binding of LBP-1a at the heavy chain loci is distant from the transcriptional regulatory regions. Alternative mechanisms for regulation of CSR by LBP-1a include, although are not limited to, inhibition of S-S synapsis, counteracting AID deamination or strand breakage, or reduction in chromatin accessibility. With regards to the latter, activation of CSR correlates with histone acetylation in S regions [5;25]. We note that LBP-1a may have the capability of inhibiting chromatin accessibility, in that the highly similar LSF interacts with multiple inhibitory chromatin modifying factors including histone deacetylases, Sin3A corepressor and the polycomb protein RING .
The roles of specific switch region sequences in regulation of CSR remain to be fully elucidated, especially since the specific primary sequences of S regions appear unessential for the basic mechanism of CSR . We hypothesize that specific repeat sequences mediate differential regulation of CSR, by providing specific binding sites for isotype-specific activators (e.g. NF-κB), and isotype-specific repressors (e.g. LBP-1a) to finetune appropriate CSR in response to the complex signaling to B-lymphocytes in vivo.
6X His-tagged LSF and LBP-1a were isolated using Ni-NTA nickel charged resin from bacteria transformed with pQE30LSF and pQE30LBP-1a, respectively (see supporting material for plasmid constructions). The concentrations of each recombinant protein were determined against a standard curve of purified BSA by SDS-PAGE, staining with Commassie-G250, densitometry, and ImageQuant analysis (version 1.2).
Peptides corresponding to aa 317–332 of LBP-1a and 314–328 of LSF (bolded, Fig. 1A) were coupled to keyhole limpet hemocyanin. Antisera were produced in rabbits (Covance Research Products Inc.). Antibodies were purified using a HiTrap protein G column followed by affinity purification against peptide immobilized to SulfoLink coupling gel.
Purified B cells (5–20 × 106) were lysed in 2% SDS, 50 mM Tris HCl, pH 8.0, 0.1 mM DTT, 10% glycerol, subjected to SDS-PAGE, and immunoblotted with affinity-purified antibodies, followed by anti-β-actin (Sigma). 293T cells were maintained in DMEM plus 10% fetal bovine serum. Cells were transfected using lipofectamine (Invitrogen) and a total of 8 µg DNA, including pCXLBP-1a and pCXLBP-1c (LSF)  as indicated, per 10 cm dish. Nuclear extracts were prepared as described previously  prior to immunoblotting.
Male BALB/cByJ mice (6–8 weeks) from The Jackson Laboratory were handled in accordance with the National Institutes of Health and institutional guidelines. Five days prior to bone marrow harvest from femurs, donor mice were injected i.p. with 5 mg of 5-fluorouracil. Marrow was subjected to density separation using Lympholyte-M and cultured in DMEM supplemented with 10% fetal bovine serum, 2 mM L-glutamine, antibiotics, 100 ng/ml SCF, 100 ng/ml flt-3 ligand and 50 ng/ml Thrombopoetin (R and D Systems) for 24 h prior to retroviral transduction.
Retroviral packaging cells were maintained in DMEM plus 10% fetal bovine serum. Following transfection with pCL-Eco  plus either MIGW  or MIGW-LSFdn (see supporting material for construction) using Fugene (Roche), viral supernatants were collected. Bone marrow cells were transduced 3 times for 24 h each with 80% viral supernatant, 20% bone marrow medium and 12 µg/ml polybrene.
Recipient mice were irradiated with 900 rads of gamma rays; 3 h later bone marrow from 1 donor (0.5 × 106 cells) was injected into the tail vein of one recipient mouse.
Splenic B cells were isolated 8–10 wks after bone marrow transplantation, using negative selection . Briefly, splenocytes were depleted of T cells using anti-Thy1.2 and rabbit complement. B-cells were isolated by density gradient centrifugation and cultured in RPMI plus 10% fetal bovine serum, 2 mM L-glutamine, 50 µM 2-ME, and antibiotics. The B cells were phenotyped as containing fewer than 1% T cells (B220-CD3+) or macrophages (B220-CD14+). For chimeric mice, cells were sorted by FACS into GFP-expressing and non-GFP-expressing populations. The gate for non-GFP expressing cells was set at minimal fluorescence, to eliminate contamination with cells expressing GFP at low levels. The gate for GFP-expressing cells was set above levels of maximal background fluorescence from BALB/cByJ splenocytes.
LSF/LBP-1 binding sites were predicted with POSSUM (http://zlab.bu.edu/~mfrith/possum; score threshold 7, residue abundance range 1000, pseudocount 1) using the LSF binding matrix (accession number M00947) from TRANSFAC and mouse S region accession numbers Sμ: J00440-J00442, Sα: D11468, Sγ1: D78344.
ChIP assays were performed according to the Upstate Cell Signaling Solutions protocol. Briefly, 20 × 106 isolated, splenic B-cells were crosslinked in 1% formaldehyde for 10 min at 37°C, and quenched with 125 mM glycine. Cell lysates were sonicated to shear chromatin to around 500 bp; DNA was extracted from a fraction of the lysate for input controls. Precleared lysates were immunoprecipitated using 100 µl preimmune or immune rabbit serum or 5.0 µg normal rabbit IgG (Southern Biotech). Input and immunoprecipitated DNA were assayed for switch region sequences by PCR. Through amplification of increasing amounts of input DNA (e.g. Fig. 2D, 2E), a range of signals could be ascertained that was monotonically increasing with respect to the amount of input. In this manner, saturation of signals was avoided, and the results are semi-quantitative. Primer sequences included: Sμ as previously described ; Sα: AGCTTACCTAGACCGGGCAGAC and ACCTCAACTAAGTCCATCTTAGGCC; Sγ1: TATGATGGAAAGAGGGTAGCATT and CTGGGCTGGTCTGTCAACTCCTT.
Isolated splenic B-cells were stimulated either 4 or 7 days with 25 µg/ml LPS and 25 ng/ml IL- 4 to assay IgG1 or IgA expression or with 25 µg/ml LPS, 25 ng/ml IL-5 and 1 ng/ml TGF-β1 to assay IgA expression. Surface IgG1 and IgA were measured by flow cytometry, using biotinylated IgG1 or IgA antibodies (1:500 dilution) and allophycocyanin-conjugated streptavidin (1:2000 dilution) in the presence of 2.4G2 anti-FcR antibody (BD Biosciences). The GFP-positive gate was set above the fluorescence levels of non-GFP-transduced B cells.. IgA- or IgG-positive gates were set above fluorescence levels of B-cells from chimeric mice stained either with biotinylated isotype control antibody (553923-BD Biosciences) followed by allophycocyanin-conjugated streptavidin, or allophycocyanin-conjugated streptavidin alone. Analyses were performed using FlowJo software. In calculating percentage of cells undergoing CSR, the percentage of IgA+ (or IgG1+) cells in the experimental sample were first adjusted by the percentage of IgA+ (or IgG1+) cells in the parallel isotype control analysis of the same cells. Samples in which the isotype control percentages were high relative to the experimental percentages (e.g. LSFdn mice, Fig. S3B in supporting material) were removed from consideration.
This work was supported by NIH R01 CA081157 to U.H., and NIH R01 AI29690 and NIH P01 AI60896 to T.L.R. K.J.R. was supported in part by T32 HL07501. We thank Quan Zhu for pQE30LSF, Robert Roeder for LBP-1a and LSF expression constructs, Luk Van Parijs for retroviral constructs, and Chun-Yan Bai and Essi Vulli for technical assistance.
Conflicts of Interest: The authors declare no financial or commercial conflict of interest.