Autoreactive B Cells Discriminate CpG-rich and CpG-poor DNA and This Response is Modulated by IFN-α

Abstract

Autoreactive B cells are activated by DNA, chromatin, or chromatin-containing immune complexes (ICs)⁶ through a mechanism dependent on dual engagement of the BCR and TLR9. We examined the contribution of endogenous DNA sequence elements to this process. DNA sequence can determine both recognition by the BCR and by TLR9. DNA fragments containing CpG islands, a natural source of unmethylated CpG dinucleotides, promote the activation of DNA-reactive B cells derived from BCR transgenic mice as well as DNA-reactive B cells present in the normal repertoire. ICs containing these CpG island fragments are potent ligands for AM14 IgG2a reactive B cells. By contrast, ICs containing total mammalian DNA, or DNA fragments lacking immunostimulatory motifs, fail to induce B cell proliferation, indicating that BCR-crosslinking alone is insufficient to activate low affinity autoreactive B cells. Importantly, priming B cells with IFN-α lowers the BCR activation threshold and relaxes the selectivity for CpG-containing DNA. Together, our findings underscore the importance of endogenous CpG-containing DNAs in the TLR9-dependent activation of autoreactive B cells and further identify an important mechanism through which IFN-α can contribute to the pathogenesis of systemic lupus erythematosus (SLE).

Introduction

Systemic lupus erythematosus (SLE) and other autoimmune diseases are characterized by the development of autoantibodies directed against a limited subset of nucleic acid-containing antigens including DNA, chromatin, and ribonucleoproteins (1). Defects in the clearance of apoptotic material have been associated with the development of anti-nuclear antibodies and autoimmune disease (2). However, the mechanism leading to the production of DNA-reactive autoantibodies is difficult to explain since mammalian DNA is a poor immunogen compared to microbial DNA (3-5). We have previously shown that ICs containing mammalian DNA can very effectively activate IgGautoreactive B cells through a mechanism dependent on engagement of the BCR and the intracellular pattern-recognition receptor TLR9 (6). TLR9 was originally identified as a sensor for microbial DNA, through its recognition of unmethylated CpG motifs found at a high frequency in microbial DNA (7). By contrast, mammalian DNA has a low GC-content, is depleted for CpG dinucleotides, and is highly methylated (8).

The requirement for CpG dinucleotides in immunostimulatory DNAs was first demonstrated in studies examining synthetic phosphodiester linked oligonucleotides (PDODNs) (9, 10). Subsequent studies examining short synthetic phosphorothioate oligonucleotides (PS-ODNs) led to the identification of PuPuCGPyPy as the optimal motif for effective engagement of mouse TLR9 (11). However a series of recent studies have questioned how well these PS-stabilized CpG motifs reflect authentic microbial and/or endogenous ligands. For example, when used at exceedingly high concentrations, phosphodiester-linked (PD) non-CpG ODNs can have stimulatory activity (12-14). Moreover, total mammalian DNA was reported to effectively activate a TLR9 fusion protein expressed on the surface of transfected HEK 293 cells (15), and total mammalian DNA complexed with the anti-microbial peptide LL37 was found to stimulate plasmacytoid dendritic cells (16). In addition, PD-ODNs were recently reported to activate dendritic cells through a sequence independent, backbone-dependent mechanism (17). Nevertheless, the importance of mammalian DNA CpG content in the activation of TLR9, and in particular in the activation of autoreactive B cells, remains unresolved; either the relative activities of CpG-rich and non-CpG rich mammalian DNA have not been accurately compared or experimental systems are used that depend on the delivery of DNA by the addition of a 3′-poly G tail to force aggregation (18) or by artificial delivery to early endosomes with the transfection reagent DOTAP (14).

These data are inconsistent with our own observations, which focused on the more physiologically relevant uptake of autoantigen-containing ICs by either the BCR or FcγRs (19, 20). By using dsDNA fragments approximately 600 bp in length that either did or did not incorporate canonical CpG motifs, we clearly demonstrated a critical role for unmethylated CpG motifs in the activation of autoreactive B cells, and found that total mammalian DNA had only weak activity (19). Although high-dose non-CpG ligands may have the capacity to trigger TLR9 in certain experimental systems, studies involving receptor-mediated uptake are the most relevant to the study of autoimmune disease, as they best recapitulate the route through which self-DNA normally accesses TLR9. In addition, apoptotic nucleosomal DNA larger than 200 bp is hypothesized to be the self-DNA ligand (6, 19), and DNA in this size range is, by itself, taken up inefficiently by B cells compared to short ODNs. Importantly, concentrations of DNA at or below those found in the sera of patients with autoimmune disease (50-250 ng/ml (21)) can activate B cells if taken up via the BCR. In contrast, many of the experimental systems discussed above use concentrations of DNAs 100-1000 fold higher, possibly obscuring the requirements for TLR9 recognition during the course of autoimmune disease.

Given the dependence for CpG-rich DNA and the inability of total mammalian DNA to effectively activate TLR9 in our model system, we have now asked if specific elements of mammalian DNA, namely CpG islands, can preferentially activate B cells through a TLR9-dependent mechanism. CpG islands are regions of the genome that are GC-rich (50-80% GC), not depleted for CpG dinucleotides relative to the rest of the genome, and generally unmethylated (22). CpG islands therefore constitute an endogenous source of CpG-rich DNA, even though the frequency of canonical CpG motifs is relatively low. Our data demonstrate that CpG island dsDNA fragments, in contrast to CpG-poor dsDNA fragments, are remarkably potent TLR9 ligands. In addition, we found that the BCR of DNA-reactive B cells selectively binds to certain DNA sequences influencing the ability of these fragments to traffic to TLR9-containing compartments. IFN-α expression is elevated in patients with SLE and is thought to contribute to both disease initiation and progression (23). Importantly, IFN-α enhances signaling through the BCR (24) and the ensuing antibody response (25, 26). Because the activation of autoreactive B cells depends on signals emanating from both the BCR and TLR9, we also examined the effect of IFN-α on the response of CpG-rich and CpG-poor DNA ICs. Notably, we found that IFN-α priming relaxed the stringency of autoreactive B cells for CpG-rich DNA.

Materials and Methods

Mice

AM14 RF+ mice were obtained from crosses between MRL AM14 H chain transgenic (Tg) and BALB/c Vκ8 L chain Tg mice, or BALB/c AM14R H chain knock-in (KI) and BALB/c Vκ8R L chain KI mice. AM14 KI mice, in which the AM14 heavy chain was inserted into the endogenous IgH locus, will be described in detail elsewhere (S. Christensen and M. Shlomchik, manuscript in preparation). AM14/TLR9-deficient mice were obtained as previously described (27). B6 3H9R H chain KI mice and Vκ8R L chain KI mice were generously provided by Dr. Martin Weigert. 3H9R/Vκ8R mice were obtained from crosses between 3H9R and Vκ8R mice. Anti-HEL BCR Tg mice were obtained from Jackson Lab (C57BL/6-Tg(IghelMD4)4Ccg/J, JAX® mice) and intercrossed to RAG2-deficient mice (B6.129S6-Rag2^tm1Fwa N12, Taconic). Mice were breed and maintained in accordance with the regulations of the American Association for the Accreditation of Laboratory Animal Care. All mouse studies were approved by the BUSM Institutional Animal Care and Use Committee.

DNA fragments

An aliquot of a mouse CpG island library (28) (Geneservice, UK) was expanded and plated to isolate individual colonies, or plasmid DNA was made from the intact library. DNA fragments were made by PCR using primers 3558-EcoRI-F and 3559-EcoRI-R and conditions previously described (28). Genbank accession numbers for individual CpG island library clones are included in Table 1. Mouse DNA was prepared from spleen, and E.coli DNA from DH5α using Qiagen DNeasy® Blood & Tissue Kit, and digested with DdeI. Sumo DNA fragments were made by PCR amplification of a cDNA fragment corresponding to nucleotides 774 – 1392 of the 3′ non-coding region of human SUMO1 mRNA (accession # NM_003352) using primers Sumo-2-EcoRI-F and Sumo-2-BamHIR. Senp1 DNA fragments were made by PCR amplification of a cDNA fragment corresponding to nucleotides 3429 – 3985 of the 3′ non-coding region of human SENP1 mRNA (accession # NM_014554) using and Senp1-2-EcoRI-F and Senp1-2-BamHI-R. The CpG-depleted pCpG fragment was made by PCR of pCpG-mcs (Invivogen) using primers pCpG-EcoRI-F and pCpG-BamHI-R. PCR conditions were: 1 μM each primer, 40 cycles of 94°C 30 s, 57°C 45 s, and 72°C 1 m. CGneg, CG50, and TNP-labeled DNAs were prepared as described (19). PCR amplified mouse genomic DNA was made as previously described (29) except primers were modified to contain EcoRI sites (EcoRIBig). Biotin end-labeled CGneg, CG50, mouse, and E.coli DNA were made by filling-in 5′ overhangs from restriction digestion with Klenow(exo-) fragment in the presence of biotin-16-2′-deoxy-uridine-5′-triphosphate. DNA fragments containing increasing amounts of biotin were made by substituting biotin-16-2′-deoxy-uridine-5′-triphosphate (Roche) at the indicated percentage for dTTP in the PCR reaction. Biotin labeling was confirmed by incubating 50 ng of DNA with 2 μg of 1D4 antibody and running on a 1% agarose gel. Primers and enzymes were removed from all DNAs using the DNA Clean & Concentrator-25^TM kit (Zymoresearch). pCpG fragment was purified by gel isolation using QIAquick Gel Extraction Kit. All DNAs were confirmed as endotoxin-free by Limulus Amebocyte Lysate assay (Cambrex). PCR primer sequences are as follows: 3558-EcoRI-F, 5′-ACGGAATTCGGCCGCCTGCAGGTCGACCATAA-3′; 3559-EcoRI-R, 5′-ACGGAATTCAACGCGTTGGGAGCTCTCCCATAA-3′; Sumo-2-EcoRIF, 5′-GCCTGAATTCGACTTTCCAATTGGCCCTGATGTTCTAGC-3′; Sumo-2-BamHI-R, 5′-GCCTGGATCCCAGAAGGCACTTCAGTAACTTTCAGTGC-3′; Senp1-2-EcoRI-F, 5′-GCCTGAATTCCCAAATTCCAGCACACAGAGATCCCAGCC-3′; Senp1-2-BamHI-R, 5′-GCCTGGATCCGGGAGGACATGTAGTTGCTGGAGTGG-3′; pCpG-EcoRI-F; 5′-GCATGAATTCTGGAGCCAAGTACACTG-3′; pCpG-BamHI-R; 5′-GCATGGATCCAGTACACCACATCACTT-3′; EcoRI-Big; 5′-ACGGATTCTGAGCTGCCTGATGCTGGATC-3′.

Table 1.

CpG island clones

Clone	Size (bp)	% GC	CpG	Observed/ Expected CpG	Optimal CpGa	Sequence element	Accession #
20	121	46	1	0.159	1	LINE1	EU730888
17	468	56	26	0.714	2	CpG islandb	EU730885
2	199	48	4	0.354	0	Genomic	EU730870
10	353	39	3	0.228	1	LINE1	EU730878
15	587	57	34	0.720	3	CpG island	EU730883
6	298	54	18	0.852	0	CpG island	EU730874
21	545	60	40	0.820	1	CpG island	EU730889
12	453	53	19	0.609	0	CpG island	EU730880
3	586	42	17	0.713	0	Mitochondrial	EU730871
22	224	53	17	1.094	0	CpG island	EU730890
9	305	58	26	1.024	2	CpG island	EU730877
19	302	62	14	0.485	2	CpG island	EU730887
24	321	48	13	0.716	1	CpG island	EU730892
16	791	50	30	0.609	2	CpG island	EU730884
14	597	50	38	1.036	1	CpG island	EU730882
13	387	36	12	1.041	1	Satellite	EU730881
18	456	45	27	1.184	0	Unknown	EU730886
7	167	56	17	1.332	1	CpG island	EU730875
4	515	60	47	1.017	2	CpG island	EU730872
1	287	45	24	1.733	1	Unknown	EU730869
23	574	53	40	1.043	0	CpG island	EU730891
8	573	58	41	0.858	2	CpG island	EU730876
11	573	59	42	0.863	2	CpG island	EU730879
5c	573	59	42	0.858	2	CpG island	EU730873

^aOptimal CpG – PuPuCGPyPy

^bDefined as GC content ≥ 50% and ratio of observed to expected CpGs greater than 0.5, or by localization in sequence on GenBank

^c8, 11, and 5 differ only by a few nucleotides each

Antibodies

The monoclonal antibody Hy1.2 (anti-TNP) has been described previously (6). 3H9R/Vκ8R antibody was prepared by stimulating 3H9R/Vκ8R B cells with ssDNA from the CpG island library and fusing B cells to the mouse myeloma fusion partner NSO-bcl-2 (30). Antibody derived from a hybridoma pool was purified by ammonium sulfate precipitation of ascites fluid. Anti-DNA antibody PA4 (31) and anti-chromatin antibody PL2-3 (32) were provided by Dr. Mark Monestier (Temple Univ. School of Medicine). Anti-biotin antibody 1D4 was made by immunizing a Balb/c mouse with hemocyanin-biotin and fusing spleen cells to NSO-bcl-2 (30). All IgG2a antibodies were purified on protein G sepharose.

B cell activation

B cells were positively selected from spleen cell suspensions using anti-B220 microbeads (Miltenyi Biotech) and cultured as described previously (6). DNA was made single-stranded for addition to 3H9R/Vκ8R B cells by incubation at 95°C for 10 m, followed by dilution into ice-cold media. F(ab')₂ goat anti-mouse IgM (Jackson Immunoresearch) was used at 15 μg/ml, PS-ODN 1826 (33) was used at 1 μg/ml, LPS (Sigma) was used at 10 μg/ml, and CLO97 (Invivogen) was used at 100 ng/ml. Protein ICs were composed of 5 μg/ml 1D4 and 300 ng/ml biotinylated OVA. In some experiments, cells were primed with IFN-α (PBL InterferonSource, 1000 U/ml) prior to addition of stimuli. Proliferation was measured with a 6 h pulse of ³H-thymidine 24 h post-stimulation. The level of ³H-thymidine incorporation induced by PS-ODN 1826 ranged from ∼190,000 cpm to 330,000 cpm, depending on the experiment. Due to this variability, proliferation data averaged from multiple experiments are reported as the percent proliferation induced by PS-ODN 1826.

IC Binding

0.5×10⁶ purified B cells were suspended in 3% FBS/PBS containing 5 μg/ml of 2.4G2 to block FcγRII receptor activity. 100 μL of ICs consisting of 100 ng/ml of biotin-labeled DNA and 5 μg/ml of 1D4 antibody were incubated with the cells for 1 h on ice. Binding of 1D4 antibody was detected with goat anti-mouse IgG2a-FITC at 5 μg/ml and analyzed by flow cytometry.

ELISAs

For 3H9R/Vκ8R specificity analysis, plates were coated with goat anti-mouse IgM (Southern Biotech) followed by 3H9R/Vκ8R antibody. Unlabeled ssDNA was added along with 125 ng/ml biotin-labeled ssCG50 and incubated overnight (O/N). ELISAs were developed with Streptavidin-horseradish peroxidase (Southern Biotech) and TMB substrate (Sigma)

Western blotting

For phosphotyrosine blots, lysates from purified B cells were separated by SDS-PAGE, blotted to PVDF membranes, and probed with anti-phosphotyrosine 4G10 (Upstate). Blots were stripped and reprobed for β-actin (Cell Signaling). For phospho-Syk analysis, cell lysates were immunoprecipitated with an anti-Syk antibody coupled to agarose according to the manufacturer's protocol (Santa Cruz) at 4°C for 2 h. Precipitated proteins were analyzed by western blot with the anti-phosphotyrosine, 4G10 (Upstate). The blot was stripped and reprobed with anti-Syk (Cell Signaling) as a loading control. PLCγ2 and phospho-PLCγ2 (Tyr1217) specific mAbs were from Cell Signaling.

Calcium flux

B cells were loaded with Indo-1AM (1 μM) (Molecular Probes) for 30 min at 37°C washed and resuspended to a final concentration of 5 × 10⁶ cells/ml. The cells were incubated for 20 m at RT, stimulated with ICs, and intracellular Ca²⁺ was evaluated by measuring fluorescence at 405 nm and 485 nm after excitation at 355 nm with a Mo-Flo flow cytometer (DakoCytomation). Data analysis was performed using FlowJo software (Treestar).

Results

Endogenous CpG-rich DNA sequences activate autoreactive B cells

To determine whether CpG-rich dsDNA fragments derived from mammalian cells can activate murine B cells, dsDNA fragments ranging in size from approximately 200 bp to 1 kb were isolated from a mouse CpG island library (Fig. 1A), and compared to digested total mouse genomic DNA or E.coli genomic DNA for their ability to stimulate proliferation of BCR transgenic DNA-reactive B cells. Previously characterized dsDNA fragments were also included in the assay; CG50, a 607 bp dsDNA fragment containing 50 CpG motifs optimized for mouse TLR9, and CGneg, a 629 bp dsDNA fragment containing no CpG dinucleotides (19). Pairing of the 3H9 heavy chain with a Vκ8 light chain generates a BCR specific for ssDNA (34). Consequently, all studies with 3H9 × Vκ8 were performed with ssDNA. These 3H9 × Vκ8 B cells proliferated in a dose-dependent fashion to denatured ssDNAs including E.coli DNA and CG50 fragments, but did not respond to total mouse DNA or the CGneg fragments, as detected by ³H-thymidine incorporation after 24 h. Remarkably, the CpG island DNA fragments triggered a response comparable to that induced by the E.coli DNA and CG50 (Fig. 1B). The E.coli DNA, CG50, and CpG island responses could all be blocked by inhibitors of TLR9 (Fig. 1C).

Figure 1. Endogenous CpG-rich DNA sequences activate autoreactive B cells and B cells present in the normal repertoire.

(A) Size profile of DNA fragments from CpG island library. 3H9R/Vκ8R B cells were stimulated with ssDNA (B) in the presence or absence of the TLR9-inhibitor S-ODN 2088 (C). AM14 B cells were stimulated with dsDNA fragment ICs (5 μg/ml Hy1.2 + TNP-dsDNA) (D) or with dsDNA fragment ICs (5 μg/ml Hy1.2 + 10 ng/ml TNP-dsDNA) containing mouse genomic DNA, PCR-amplified unmethylated mouse genomic DNA (average length 400 bp), or PCR-amplified CpG island DNA (E). WT B6 or anti-HEL/RAG^−/− B cells were stimulated with dsDNA fragments (F). DNA was from E coli ■, CG50 △, CpG islands ○, mouse genome ▲, or CGneg □. Proliferation was assessed by ³H-thymidine incorporation and results are standardized as percent proliferation induced by PS-ODN 1826. The data shown are the mean ± SEM of 3 experiments (B, D) 4 experiments (E), or are representative of 2 experiments (C, F).

To determine whether the ability of CpG island DNAs to activate proliferation extended to other autoreactive B cells, E.coli genomic DNA, mouse genomic DNA, CG50, CGneg and CpG island dsDNA fragments were haptenated and then combined with an IgG2a anti-hapten mAb to form ICs. These ICs were then evaluated for their ability to stimulate IgG2a-reactive AM14 BCR transgenic B cells. As reported previously, the CG50 and E.coli dsDNA fragments induced a robust response (19). Importantly, the CpG island fragment ICs again induced a response that was comparable to that of the experimentally designed CG50 ICs (Fig. 1D). The fragments did not stimulate in the absence of the anti-hapten mAb (data not shown). Overall these data demonstrate that certain endogenous DNA sequences are essentially comparable to bacterial DNA with regard to their ability to activate autoreactive B cells through a BCR/TLR9 dependent mechanism.

The mouse genomic DNA and the CpG island DNA were differentially methylated due to their purification from tissue (mouse DNA, methylated) or PCR products (CpG island DNA, unmethylated). To determine whether the low immunostimulatory activity of mouse DNA could be due to its methylation status, we made dsDNA fragments by ligation-mediated PCR of total mouse genomic DNA (29). These fragment were labeled with TNP, and used to stimulate AM14 B cells in combination with anti-TNP antibody. The PCR-amplified genomic dsDNA showed only a modest increase in activity relative to tissue-derived DNA , suggesting that overall sequence composition, rather than methylation status, was the primary reason for lack of immunostimulatory activity (Fig. 1E).

B cells present in the normal repertoire proliferate in response to endogenous CpG-rich DNA fragments

B cells capable of recognizing autoantigens are present in the normal B cell repertoire of mice and humans (35, 36). To determine whether the normal repertoire includes B cells that respond to CpG-island DNA fragments, B cells from WT mice were stimulated with E.coli and mouse genomic DNA as well as with the CG50, CGneg and CpG island dsDNA fragments. As in the case of the autoreactive B cells, the WT B cells showed a measurable response to the CpG-containing fragments, including the CpG island fragments. However, the response was markedly less robust than the autoreactive B cell responses, when standardized to the response to 1826, a PS-ODN that does not require the BCR for access to a TLR9-containing compartment (Fig. 1F). We assumed that this level of response reflected the relatively low frequency of DNA-responsive cells in the normal repertoire. To confirm this premise, we tested B cells expressing a transgenic BCR specific for hen egg lysozyme (HEL). HEL-reactive RAG2-deficient B cells do not express any endogenously rearranged BCRs. These cells completely failed to respond to any of the CpG-rich dsDNA fragments (Fig. 1F), indicating that B cells that express autoreactive DNA-specific receptors can respond to endogenous CpG-rich, but not CpG-poor, DNA sequences.

Endogenous DNA sequences differentially activate 3H9R/Vκ8R and AM14 B cells

To more carefully examine the sequence characteristics of the CpG island fragments, 24 individual clones were isolated from the CpG island library. 17 of the 24 clones were CpG islands as defined by having a GC-content > 50% and a ratio of observed to expected CpGs greater than 0.5, similar to the reported percentage of CpG islands in the library (28). Other sequences included 2 long interspersed nuclear elements (LINEs), 1 satellite, 1 mitochondrial, 1 coding, and 2 unknown sequences (Table 1). DNA fragments were made from each of the clones, and ssDNA was used to stimulate 3H9R/Vκ8R B cells. Unexpectedly, the activity of the individual sequences was highly variable. Some of the clones induced remarkably low levels of proliferation, while others induced responses comparable to E.coli DNA (Fig. 2A). Both CpG island and non-CpG island sequences induced proliferation, but all stimulatory clones contained at least 1 optimal (i.e. PuPuCGPyPy) CpG motif, or multiple CpG dinucleotides. The level of proliferation did not directly correlate with the number of optimal motifs or the total number of CpG dinucleotides, suggesting that additional sequence or structural properties could influence the activity of the fragments.

Figure 2. Ability of endogenous DNA sequences to activate 3H9R/Vκ8R B cells correlates with ability to bind the BCR.

(A) 3H9R/Vκ8R B cells were stimulated with ssDNA (300 ng/ml). (B) AM14 B cells were stimulated with fragment ICs (5 μg/ml Hy1.2 or 1D4 + 100 ng/ml TNP-dsDNA or bio-dsDNA). Results are presented as percent proliferation relative to either clone 5 (A) or PS-ODN 1826 (B) and are the mean ± SEM of 3 experiments. (C, D) Increasing concentrations of unlabeled ssDNA were added to plate-bound 3H9R/Vκ8R IgM along with biotinylated CG50 fragment. Results are representative of 3 experiments.

Because many of the non-stimulatory clones contained potentially active CpG motifs, we considered the possibility that their low activity was due to their inability to efficiently bind to the 3H9R/Vκ8R BCR. To test this possibility, we labeled the fragments with TNP or biotin and mixed them with IgG2a anti-TNP (Hy1.2) or anti-biotin (1D4) mAbs to form ICs that could be recognized by the AM14 receptor independent of DNA sequence. ICs containing dsDNA fragments representative of CpG-island fragments with low, intermediate, or high capacity to activate 3H9R/VκR8 B cells (Table 2) were then used to stimulate AM14 B cells. Despite their variable activity on 3H9R/Vκ8R B cells, all the clones tested strongly stimulated AM14 B cells (Fig. 2B). Importantly, AM14 B cells failed to respond to DNA fragments added without antibody (right panel), confirming that DNA preparations lacked endotoxin. Even though all of the CpG island library clones were strong ligands in the AM14 assay, ICs containing dsDNA fragments derived from CpG-poor mammalian genomic regions and containing low numbers of CpG dinucleotides (Sumo, Senp1), or from a plasmid lacking CpG dinucleotides (pCpG) (Table 2), failed to stimulate AM14 proliferation (Fig. 2B). 3 CpG dinucleotides derived from plasmid sequences were present in each of the CpG island clones due to the primers used to generate the fragments. Although these CpG dinucleotides are not in an optimal sequence context we asked whether they might contribute to TLR9 activation. To do so, the CpG-poor pCpG fragment was made with primers incorporating these non-optimal CpG sequences at the 5′ ends. ICs containing this fragment showed no increase in activity relative to the original pCpG fragment (data not shown). Overall the data confirmed that CpG-rich endogenous DNA sequences were capable of engaging TLR9, and suggested that only some DNA sequences could be efficiently bound and internalized by the 3H9R/Vκ8R receptor.

Table 2.

DNA fragments tested on AM14 B cells

DNA fragment	Sequence element	Size (bp)	CpG	Optimal CpGa	Activity on 3H9R/Vκ8R
3	CpG islandb	586	17	0	Low
20	LINE	121	1	1	Low
10	LINE	353	3	1	Low
15	CpG island	587	34	3	Low
7	CpG island	167	17	1	Intermediate
16	CpG island	791	30	2	Intermediate
23	CpG island	574	40	0	High
11	CpG island	573	42	2	High
Sumo	cDNA	619	1	0	-
Senp1	cDNA	557	4	0	-
pCpG	Plasmid	573	0	0	-

^aOptimal CpG – PuPuCGPyPy

^bDefined as GC content > 50% and ratio of observed to expected CpGs greater than 0.5

Ability of endogenous DNA sequences to activate 3H9R/Vκ8R B cells correlates with ability to bind the BCR

To further evaluate the relative binding affinity of the 3H9R/Vκ8R receptor for the individual DNA fragments, we used a competition ELISA. The various CpG-island ssDNA fragments were compared for their ability to compete with biotin-labeled CG50 for binding to plate-bound 3H9R/Vκ8R IgM mAb. 3H9R/Vκ8R bound CGneg, CG50, mouse DNA, E.coli DNA, and the CpG island library comparably. However, differences in binding were observed among clones derived from the CpG island library (Fig. 2C, 2D). For example, clone 20, which stimulated 3H9R/Vκ8R B cells very poorly, exhibited the lowest levels of binding to the 3H9R/Vκ8R antibody. Clones 23 and 11, which stimulated high levels of proliferation of 3H9R/Vκ8R B cells, exhibited the highest levels of binding to the 3H9R/Vκ8R antibody. Intermediate clones exhibited intermediate levels of binding. These data demonstrate that the 3H9R/Vκ8R receptor has a binding preference for certain DNA fragments, and that recognition by the 3H9R/Vκ8R receptor correlates with the ability of these DNA fragments to presumably traffic to TLR9 and induce proliferation.

BCR crosslinking by multivalent ICs cannot enhance autoreactive B cell responses to non-optimal DNA fragments

The previous data suggested that the BCR was critical for delivery of DNA ligands to TLR9, but whether BCR crosslinking contributed to this activation was unknown. Prior studies from other investigators had shown that crosslinking the BCR with anti-IgM could enhance B cell responses to non-optimal TLR9 ligands (37). The dsDNA ICs used above were only haptenated at their ends and therefore were of limited valency. To determine whether more extensive crosslinking of the BCR would permit DNA fragments lacking CpG dinucleotides, or containing low numbers of non-optimal CpG dinucleotides, to more effectively trigger B cell proliferation, we constructed larger multivalent ICs. The CGneg, Senp1, and CpG island clone 11 fragments were chosen as representative non-CpG, non-optimal CpG, and CpG-rich DNA fragments respectively, and increasing percentages of biotin-conjugated dUTP were incorporated into these fragments by PCR. ICs were then formed by mixing the biotinylated dsDNA fragments with an IgG2a anti-biotin mAb, 1D4. As expected, in the presence of a fixed level of antibody, higher ratios of biotin led to the formation of larger ICs, as indicated by an electrophoretic mobility shift (Fig. 3A). These larger complexes bound more avidly to the surface of AM14 B cells (Fig. 3B). In addition, at comparable ratios of biotinylation, the ICs formed with the CpGneg, Senp1 and clone 11 fragments bound comparably to the AM14 B cells.

Figure 3. BCR crosslinking by multivalent ICs cannot enhance autoreactive B cell responses to non-optimal DNA fragments.

(A) Samples containing a fixed amount of clone 11 dsDNA prepared with increasing percentage of biotinylated bases were incubated alone in the presence or absence of the 1D4 anti-biotin mAb and then electrophoresed on a 1% agarose gel. Similar results were obtained with CGneg and Senp1 (data not shown). (B) The same DNA-bio/1D4 ICs were added to AM14 B cells for 1 hr at 4°C and level of IC bound to the cell was determined by flow cytometry by staining with goat anti-mouse IgG2a-FITC. Results are representative of 2 experiments. (C) AM14R B cells were stimulated with 15 μg/ml F(ab')₂ goat anti-mouse IgM, 1 μg/ml PS-ODN 1826, 5 μg/ml PA4, or dsDNA ICs (5 μg/ml 1D4 + 100 ng/ml biotinylated dsDNA) for 10 min and cell lysates were assayed for phospho-tyrosine by western blot. β-actin is shown as a loading control. (D) AM14R B cells were stimulated as above and total cell lysates were immunoprecipitated with anti-Syk followed by western blot with anti-phosphotyrosine. (E) AM14R B cells were stimulated with 5 μg/ml 1D4, or fragment ICs (5 μg/ml 1D4 + 100 ng/ml biotinylated dsDNA) and Ca²⁺ flux was measured by flow cytometry. (F) AM14R B cells were stimulated with ICs (5 μg/ml 1D4 + 100 ng/ml biotinylated dsDNA) and proliferation was measured. Results are representative of 3 experiments (C, D), 2 experiments (E), or are the mean ± SEM of 3 experiments (F).

Ligation of the BCR leads to rapid tyrosine phosphorylation of the BCR-associated signaling molecules Igα/β, and a number of intracellular signaling molecules such as Lyn, Syk, and B cell linker protein (BLNK) (38). In order to confirm that our multivalent ICs could trigger the BCR signaling cascade, we analyzed the tyrosine phosphorylation pattern of AM14 B cells stimulated with ICs. The anti-biotin antibody 1D4 alone failed to induce a detectable increase in tyrosine phosphorylation while increasing amounts of phosphorylation were observed when AM14 B cells were stimulated with ICs composed of either clone 11, or Senp1 fragments of increasing valency (Fig. 3C). PS-ODN 1826 failed to induce tyrosine phosphorylation as previously reported (39), indicating that IC induced tyrosine phosphorylation resulted from engagement of the BCR and not TLR9 (Fig. 3C). Importantly, multivalent complexes of 1D4 and clone 11 or Senp1 induced the accumulation of comparable levels of phospho-Syk (Fig. 3D).

As a second criterion for effective BCR engagement, these ICs were also assessed for their ability to induce increased levels of intracellular calcium. The larger complexes induced a more rapid and extensive level of calcium flux (Fig. 3E). Importantly, the optimal (clone 11) and non-optimal (Senp1) ICs elicited essentially identical responses. We next asked whether the highly derivatized ICs induced a stronger proliferative response in AM14 B cells. As expected, the ICs containing the more extensively biotinylated CpG island clone 11 fragments, stimulated AM14 B cells more robustly, suggesting that these ICs induced the B cells to proliferate in response to signals from both the BCR and TLR9. In contrast, ICs containing the highly derivatized non-CpG fragment CGneg, or the non-optimal CpG fragment Senp1, failed to activate AM14 B cells (Fig. 3F). Activation by the CpG island clone 11 was completely abolished by a TLR9 inhibitor (data not shown). These data indicate that despite their ability to comparably engage the BCR, ICs containing non-CpG-rich DNA are unable to activate low affinity autoreactive B cells.

IFN-α can enhance autoreactive B cell responses to non-optimal DNA fragment ICs

Both patients and mice afflicted with SLE frequently express an IFN signature, and IFN-α has been shown to promote plasma cell differentiation (40). We had previously shown that IFN-α dramatically enhanced the proliferative response elicited by RNA ICs, most likely due to increased TLR7 expression (41). To determine the effect of IFN-α on the response to dsDNA fragment ICs, AM14 B cells were primed with IFN-α and then stimulated with ICs containing the non-optimal CpG fragment Sumo, or CpG island clone 11. IFN-α priming had a minimal effect on the response to the clone 11 ICs, but led to a 6-fold increase in the response to the Sumo ICs (Fig. 4A). Responses to saturating concentrations of the BCR crosslinking agent anti-IgM or the TLR7 ligand CLO97 were unaffected by IFN-α priming. Importantly, IFN-α primed AM14 TLR9^−/− B cells failed to respond to either clone 11 or Sumo ICs, demonstrating a consistent requirement for TLR9 (Fig. 4B).

Figure 4. IFN-α can enhance autoreactive B cell responses to non-optimal DNA fragment ICs.

Proliferation of AM14 (A) and AM14/TLR9^−/− (B) B cells in response to 15 μg/ml F(ab')₂ goat anti-mouse IgM, 100 ng/ml CLO97, 1 μg/ml PS-ODN 1826, or fragment ICs (5 μg/ml 1D4 + 100 ng/ml of biotinylated dsDNA) in the presence or absence of 1000 U/ml IFN-α. (C, D) Proliferation of AM14R B cells in response to 5 μg/ml 1D4, 1 μg/ml OVA-bio, protein IC (5 μg/ml 1D4 + 300 ng/ml OVA-bio), or DNA ICs (5 μg/ml 1D4 + 100 ng/ml biotinylated mouse dsDNA or clone 11) in the presence or absence of 1000 U/ml IFN-α. Results are the mean ± SEM of 3 experiments. (E) Proliferation of AM14R B cells in response to increasing concentrations of PS-ODN 1826 or fragment ICs (1.5 μg/ml 1D4 + biotinylated dsDNA) in the presence or absence of 1000 U/ml IFN-α; results are representative of 3 experiments. AM14R B cells were incubated for 24 hrs in the presence or absence of 1000 U/ml IFN-α and surface levels of CD69 and IgM were measured by flow cytometry (F), or they were stimulated with fragment ICs (1.5 μg/ml 1D4 + 30 ng/ml biotinylated dsDNA), 5 μg/ml F(ab')₂ goat anti-mouse IgM, or 1 μg/ml PS-ODN 1826 and Ca²⁺ flux was measured by flow cytometry (G), or cell lysates were assayed for phospho-PLCγ2 (H); results are representative of 2 experiments.

To examine the relative involvement of the BCR and TLR9 to IFN-α priming, we measured the responsiveness of AM14 B cells to protein ICs composed of the anti-biotin antibody 1D4 and biotinylated OVA. As previously reported, protein ICs alone induced only a weak proliferative response from AM14 B cells (6). In contrast, priming AM14 B cells with IFN-α led to an approximately 4-fold increase in the response to the protein ICs (Fig. 4C), suggesting that IFN-α augments signals from the BCR. In addition, IFN-α priming resulted in an approximately 5-fold increase in the ability of ICs containing mouse genomic dsDNA to stimulate AM14 proliferation, while the response to ICs containing CpG island clone 11 remained unchanged (Fig. 4D).

The concentrations of clone 11-containing ICs used above induced maximum B cell proliferation, perhaps masking the effects of IFN-α priming on the response of B cells to suboptimal BCR signals. We therefore asked whether IFN-α priming could enhance B cell responses to lower concentrations of clone 11-containing ICs. Indeed, we found that priming led to a 2-3 fold increase in the proliferation of AM14 B cells to these suboptimal concentrations of clone 11-containing ICs. Importantly, priming had no effect on the proliferation of B cells to suboptimal concentrations of PS-ODN 1826 (Fig. 4E), supporting the idea that the effects of IFN-α priming are mediated through the BCR and not TLR9. Flow cytometry demonstrated that this was not due to an IFN-α dependent increase in the surface expression of the BCR, even though there was a substantial increase in the expression of the early activation antigen CD69 (Fig. 4F).

To further examine the effects of IFN-α on BCR signaling, we measured calcium flux in IC stimulated AM14 B cells. IFN-α priming enhanced the level of calcium flux in response to suboptimal concentrations of ICs containing either the clone 11 fragment or the CpG-negative fragment, Sumo (Fig. 4G, top), or a suboptimal concentration of anti-IgM (fig. 4G, bottom). As expected, PS-ODN 1826 did not induce a calcium flux in either untreated or IFN-α primed B cells. Cytosolic calcium levels are regulated by multiple mechanisms. BCR engagement initially triggers the release of Ca²⁺ from the endoplasmic reticulum, through a process dependent on the activation of PLCγ2. We found that IFN-α alone induced more extensive phosphorylation of PLCγ2 Tyr 1217, than either anti-IgM F(ab)'₂ or dsDNA ICs (Fig 4H). We found no effect of IFN-α on the levels of pSyk (data not shown). Collectively, these data suggest that environmental signals such as IFN-α can augment signaling through the BCR through a process that involves priming of PLCγ2, and the ensuing increase in BCR signal strength relaxes the selectivity of autoreactive B cells for CpG-rich DNA.

Discussion

We have examined the role of endogenous sequence elements, BCR binding, and BCR engagement in the activation of low-affinity autoreactive B cells. Our data support a model in which low-affinity autoreactive B cells that survive negative selection and tolerance induction can become activated by nucleic acid-containing autoantigens in the periphery through TLR9 engagement. We find the presence of CpG motifs to be essential for engagement of TLR9. The BCR plays a critical role in binding to ICs and/or native DNA and delivering nucleic acid to TLR9, but crosslinking of the BCR by ICs itself is not sufficient for activation as defined by ³H-thymidine incorporation, and thus cell-cycle entry and proliferation.

Although previous studies from our group had established that mammalian DNA could effectively engage TLR9, the specific DNA elements responsible for this activation were unknown. The current report confirms the essential role of CpG-rich DNA in the TLR9-dependent activation of autoreactive B cells and identifies for the first time, specific endogenous DNA sequences with adjuvant activity. These include both CpG islands and LINE sequences. Notably, six of the CpG island clones (clones 5, 7, 8, 9, 11 and 12) correspond to regions of the mouse 45S pre-rRNA gene that are hypomethylated in vivo (42, 43). The pre-rRNA genes are repeated ∼200 times in the mouse genome, which may help to explain their high frequency in the CpG island library. The 45S pre-rRNA genes are the sites of assembly for the nucleolar RNA/protein complex involved in ribosome synthesis. Interestingly, nucleolar components are frequent autoantibody targets in patients afflicted with the systemic autoimmune disease systemic sclerosis (44). Four of the clones (clones 4, 18, 22, and 23), while meeting the definition of a CpG island, were not homologous with the sequenced mouse genome. This may be due to genetic differences between the MF1 strain of mouse used to generate the CpG library and the sequenced B6 genome. While the majority of CpG islands are unmethylated, LINE sequences are highly methylated in human DNA (45), and therefore are unlikely to be an important source of TLR9 ligands in vivo, unless changes in their methylation status occur. Although we observe a dependence on CpG dinucleotides, the specific base context of the motifs in PD-DNA that effectively engage TLR9 remains unknown. Optimal motifs for TLR9 have been defined using large panels of PS-ODNs, but this specificity is not necessarily maintained in PD-ODNs (46). Although our conclusions are limited due to the large size and small number of sequences examined, our data are consistent with a requirement for the presence of at least one optimal CpG motif, or large numbers of non-optimal CpG dinucleotides for engagement of TLR9. Identification of optimal stimulatory CpG motifs for dsDNA will require further investigation.

Ligation of the BCR by multivalent ICs has been shown to promote more efficient BCR activation and antigen presentation (47). Previous work in our lab had shown that protein ICs are able to activate 20.8.3 high affinity RF transgenic B cells, but fail to activate AM14 low affinity RF B cells (19), suggesting that BCR crosslinking is sufficient for activation of high affinity B cells by protein antigens, but is insufficient to activate low-affinity B cells. Both transgenic lines could be activated by chromatin ICs, but the contribution of BCR crosslinking to activation by chromatin ICs was unknown. Importantly, multivalent crosslinking of membrane Ig has been shown to sensitize B cells to activation by non-optimal ODNs and methylated CpG ODNs (37, 48), raising the possibility that the observed dependence on CpG motifs was due to failure of our experimentally designed ICs to mimic the BCR-crosslinking ability of natural ICs. In the current study, we clearly demonstrate that ICs capable of crosslinking the BCR still failed to induce proliferation if they lacked immunostimulatory motifs for TLR9 (Fig. 3). The inability of these ICs to activate AM14 B cells may be due to the low affinity of the BCR for its antigen and could reflect a relative increased importance for TLR stimulation, or crosstalk between BCR and TLR pathways, in the activation of low-affinity B cells.

How CpG-rich DNA becomes available to the immune system in the course of autoimmune disease is unknown. It is possible that defects in cell death processes could lead to enrichment for certain sequences that can engage TLR9. CpG islands make up a small percentage of the total genome, but they have an open chromatin configuration, due to their association with actively transcribed housekeeping genes (45). Actively transcribed genes have been shown to be cleaved early during apoptosis (49), suggesting the possibility that CpG islands, which are predominantly located in promoter regions may be released during apoptosis, and therefore enriched in certain apoptotic bodies. In addition, adding back total genomic DNA to anti-DNA antibodies reduces proliferation observed in response to DNA present in the culture (data not shown), supporting the idea that immunostimulatory DNA is enriched in the cell debris present in culture.

Plasmacytoid dendritic cells (pDCs), but not conventional dendritic cells (cDCs) have been shown to retain the TLR9 ligand CpG-A in endosomal compartments, resulting in IFN-α induction. cDCs, which do not normally produce high levels of IFN-α, can be induced to produce IFN-α if CpG-A is retained in the endosome by complexing it with a cationic lipid (50). In addition, recent evidence has suggested that association of human DNA with the cationic anti-microbial peptide LL37 allows for uptake and retention of DNA in an early endosomal compartment, where it can activate pDC IFN-α production (16). The DNA-binding protein high mobility group box 1 (HMGB1) has also been shown to associate with CpG ODNs and TLR9, leading to a faster redistribution of TLR9 to early endosomes (51). It is possible that association of total mammalian DNA with anti-chromatin or anti-DNA antibodies, or with other DNA-binding proteins such as LL37 or HMGB1, could result in preferential trafficking to, or retention of complexes in an endosomal compartment, resulting in B cell activation by CpG-poor DNA. Immune complex activation of dendritic cells may promote autoreactive B cell survival through the production of survival factors such as BAFF (20).

This work also highlights the critical importance of the BCR in delivery of DNA ligands to TLR9. Although B cells expressing non-DNA-reactive BCRs are able to respond to ODNs, they are unable to respond to low doses of native dsDNA. This difference is presumably due to the ability ODNs to be taken up through endocytosis, while large native DNA is unable to reach the TLR9 compartment without delivery through the BCR. The endosomal localization of TLR9 is thought to have evolved to prevent engagement by self-DNA (15), but access to this location by delivery through an autoreactive BCR overrides this protection.

Importantly, we have found that DNA-reactive BCRs can mediate selective binding to particular DNA sequences. Whether the differences in DNA binding are due to secondary structure of the DNA, or to sequence, is unknown. Previous studies have found that sequence-specific antibodies can be generated (52, 53), and 3H9R single-chain Fv has been reported to have a binding preference for poly(dG)·poly(dC) over poly(dA)·poly(dT) (54).

SLE patients produce increased levels of type 1 IFNs that are thought to contribute to both disease initiation and progression (23). Our findings suggest that one way IFN-α may contribute to disease is by lowering the threshold of activation for autoreactive B cells. Previous studies from our lab as well as the current report clearly demonstrate that low affinity autoreactive B cells proliferate in response to a combination of signals emanating from both the BCR and either TLR9 or TLR7 (6, 19). Therefore it is reasonable to assume that a stronger signal through the BCR can synergize with a weaker signal from a TLR and still reach the activation threshold that might be obtained by a weaker BCR/stronger TLR combination, and thereby appear to relax the stringency of TLR9 for specific DNA sequences. A recent study reported that CpG-poor DNA can bind TLR9 (17), even though in our (physiologically relevant) experimental system this level of engagement is not sufficient to promote cell cycle entry.

As shown above, IFN-α priming enhanced signaling through the BCR by protein ICs, but did not enhance signaling through TLR9 by PS-ODNs (Fig 4E), it seems likely that priming is primarily due to the ability of IFN-α to reprogram the BCR signaling cascade. In fact, type 1 IFNs have been previously shown to prime B cells and enhance their responsiveness to BCR crosslinking (24). We now demonstrate that this priming involves phosphorylation of PLCγ2 at Tyr 1217. We assume that PLCγ2 is not fully activated because full activation of PLCγ2 would lead to the release of intracellular Ca²⁺ stores and the IFN-α-primed B cells maintain basal levels of cytosolic Ca²⁺. The status of IFN-primed B cells resembles that reported for anti-IgM stimulated B1 cells where it was found that PLCγ2 was phosphorylated at Tyr 1217 (the PLCγ2 pTyr detected by the commercially available p-PLCγ2 antibody), but unable to cleave its normal substrates (55). Full activation of PLCγ2 requires phosphorylation of 3 additional tyrosine residues, Tyr 753, Tyr 759, and Tyr 1197; all 4 tyrosines are phosphorylated following BCR engagement, most likely by Btk (56, 57). We speculate that IFN-α leads to partial phosphorylation of PLCγ2, and as a result less Btk or other upstream kinases are required to completely activate PLCγ2. Studies are in progress to confirm this hypothesis. In addition, we cannot rule out an IFN-α effect on TLR9 signaling, since crosstalk between IFNAR, TLRs, and immunoreceptor tyrosine-based activation motif (ITAM)-containing receptors has been described in other cell types (58).