|Home | About | Journals | Submit | Contact Us | Français|
Increasing evidence suggests that the excessive accumulation of apoptotic or necrotic cellular debris may contribute to the pathology of systemic autoimmune disease. HMGB1 is a nuclear DNA-associated protein, which can be released from dying cells thereby triggering inflammatory processes. We have previously shown that IgG2a-reactive BCR transgenic AM14 B cells proliferate in response to endogenous chromatin immune complexes (ICs), in the form of the anti-nucleosome antibody PL2-3 and cell debris, in a TLR9-dependent manner, and that these ICs contain HMGB1. Activation of AM14 B cells by these chromatin ICs was inhibited by a soluble form of the HMGB1 receptor, RAGE-Fc, suggesting HMGB1/RAGE interaction was important for this response . To further explore the role of HMGB1 in autoreactive B cell activation, we assessed the capacity of purified calf thymus HMGB1 to bind dsDNA fragments and found that HMGB1 bound both CG-rich and CG-poor DNA. However, HMGB1/DNA complexes could not activate AM14 B cells unless HMGB1 was bound by IgG2a and thereby able to engage the BCR. To ascertain the role of RAGE in autoreactive B cell responses to chromatin ICs, we intercrossed AM14 and RAGE-deficient mice. We found that spontaneous and defined DNA ICs activated RAGE+ and RAGE− AM14 B cells to a comparable extent. These results suggest that HMGB1 promotes B cell responses to endogenous TLR9 ligands through a RAGE-independent mechanism.
HMGB1 is a non-histone, chromatin-associated protein expressed in almost all eukaryotic cells, which contains two DNA-binding domains denoted A-box and B-box [2, 3]. Release of HMGB1 by necrotic  or apoptotic [5, 6] cells alerts the immune system of tissue damage. Therefore HMGB1 has been designated an alarmin . This proinflammatory function depends on the interaction of HMGB1 with target cell receptors, such as RAGE , and possibly TLR2 and TLR4 [9, 10]. Interestingly, HMGB1 function is inhibited by addition of antagonist recombinant A-box , although the mechanism for this suppression remains unclear.
High levels of HMGB1 and anti-HMGB1 antibodies have been found in the serum of patients with autoimmune disorders, including systemic autoimmune diseases such as Systemic Lupus Erythematosus (SLE) [6, 12–15]. DNA and other chromatin-associated proteins are prevalent autoantibody targets in SLE, in part because of their capacity to activate TLR9 . Intracellular TLR9 was originally described as a pattern recognition receptor (PRR) specific for unmethylated CpG motifs common to bacterial but not mammalian DNA . However, our studies in B cells, in addition to reports in macrophages, myeloid and plasmacytoid denditic cells (pDCs) have also implicated TLR9 in the detection of endogenous mammalian DNA [18–20].
Mammalian DNA contains relatively few CpG motifs and a remarkably low frequency of CpG dinucleotides, and is highly methylated. Nevertheless, dsDNA fragments which incorporate unmethylated CpG islands are much more potent ligands for TLR9 than dsDNA fragments representative of total genomic DNA . Thus, we considered the possibility that mammalian DNA-bound cofactors, that preferentially bind CG-rich DNA, could promote TLR9 responses through a mechanism dependent on a third receptor. CpG islands are commonly found in gene promoters  and active gene promoters comprise regions of “naked” DNA . These areas can also include kinks or adopt a cruciform conformation, and HMGB1 has been reported to preferentially bind such DNA structures [23, 24]. Thus, CpG-rich regions in mammalian DNA may represent optimal sites for HMGB1 binding, and HMGB1 may thereby promote TLR9 functions. In support of this hypothesis, TLR9-dependent IL-6 secretion by BMDM and BMDC in response to CpG-B ODN was enhanced by HMGB1 . However, binding studies that have depended on synthetic oligonucleotides have given inconsistent results. In one report, HMGB1 preferred CpG-rich ODNs regardless of DNA structure , while in another study HMGB1 bound to CpG-A but did not bind CpG-B ODNs . Surprisingly, the possibility that HMGB1 may preferentially bind more physiologically relevant CpG-rich mammalian dsDNA, and thus promote TLR9 responses in autoreactive B cells has not been adequately explored. In support of the hypothesis that HMGB1 may promote TLR9-dependent responses in B cells, we previously showed that activation of IgG2a-autoreactive AM14 B cells by chromatin ICs was inhibited by the HMGB1 antagonist, A-box. Moreover, we demonstrated that HMGB1 was present in AM14 B cell-bound chromatin ICs, and that its presence enhanced the extent to which chromatin ICs bound AM14 B cells .
Our data also suggested that the HMGB1 receptor RAGE was involved in TLR9-dependent effects in B cells, as soluble RAGE blocked the AM14 B cell response to chromatin ICs. Moreover, IFNα production by pDCs in response to CpG-A ODN was enhanced by HMGB1, and this effect was blocked in RAGE−/− pDCs . RAGE is a high affinity (KD of 6.4×10−9 M) cell surface receptor for HMGB1  and mediates a range of HMGB1-dependent immune responses in macrophages, neurons and DCs [8, 26–28]. In addition, RAGE-deficiencies have been linked to HMGB1-dependent and independent abnormal cell specific functions, such as decreased DC maturation  and migration , T cell activation and differentiation . In contrast to a well-established role in other cell types, the role of RAGE in various aspects of B cell development and activation has not been fully examined.
To address the potential contribution of HMGB1 engagement of RAGE in the activation of autoreactive B cells, we have now evaluated the relative capacity of spontaneous and defined dsDNA ICs to activate RAGE-sufficient and RAGE-deficient AM14 B cells.
AM14 transgenic mice and AM14 Tg TLR9−/− mice have been described [31, 32]. AM14 heavy chain knock-in KI mice were generated by insertion of the AM14 H chain into the endogenous H chain locus (Christensen and Shlomchik, manuscript in preparation). These mice were then intercrossed to the Vk8 KI line  to produce H/L AM14 KI mice. RAGE−/− mice  were kindly provided by Angelika Bierhaus (University of Heidelberg, Germany). AM14 H/L KI were then intercrossed with the RAGE−/− mice to obtain experimental mice for the current studies. All mice were maintained at the Laboratory Animal Sciences Center at Boston University School of Medicine. Animal procedures were performed under the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care and approved by Boston University School of Medicine Institutional Animal Care and Use Committee.
ODN 1826 (CpG class B) was obtained from Coley Pharmaceuticals (Wellesley, MA). F(ab)′2 fragment of goat anti mouse IgM was from Jackson Immunoresearch (West Grove, PA). Pam3CysSK4 was from EMC Microcollections (Tuebingen, Germany) and R848 was from Invitrogen (Carlsbad, CA). CG50 and CGneg fragments were prepared as described previously . Synthesis and biotinylation of Clone 11 and SUMO dsDNA fragments has been described . The IgG2a mAbs PA4 (anti-DNA) and PL2-3 (anti-histone) were kindly provided by Dr. Marc Monestier (Temple University School of Medicine) and 1D4 (anti-biotin) has been described . Calf thymus HMGB1 and anti-HMGB1 IgG2a were provided by Medimmune Inc (Gaithersburg, MD).
Mouse splenic B cells were purified using B220+ magnetic beads (BD Biosciences, San Jose, CA). Primary B cells were stimulated with the specified reagents for 24 h, and then pulsed with 5 μCi/mL of [methyl 3H] thymidine for an additional 6 h. HMGB1/anti-HMGB1 immune complexes (ICs) containing CG50, Clone 11, CGneg and SUMO were preformed by combination of HMGB1 and DNA in RPMI and incubation for 30 min at 37°C, followed by addition of anti-HMGB1 and incubation for 30 min at 37°C. PA4-DNA fragment ICs were preformed by combination of PA4 IgG2a with CG50 or CGneg for 30 min at 37°C. 1D4-biotin DNA ICs were preformed by incubation at room temperature for 10 min.
Anti-mouse RAGE antibody was from Chemicon–Millipore (Billerica, MA). Fluorophore-labeled CD21, CD23, CD93 and B220 antibodies were from BD Biosciences (San Jose, CA). Staining was performed by addition of antibodies to cells in 3%FBS/PBS for 1 h on ice, washed three times using 2–3 volumes of 3%FBS/PBS, collected using LSRII Flow Cytometer (Beckton Dickinson) and analyzed using FlowJo software (Tree Star Inc, Ashland, OR).
CG-rich regions are found in gene promoters of mammalian DNA, and may represent favored sites of binding to HMGB1. To determine whether HMGB1 binds dsDNA with a sequence preference, we pre-incubated calf thymus HMGB1 with dsDNA fragments, approximately 600 bp in length, that incorporate CG-rich (CG50 or Clone 11) or CG-poor (CGneg and Sumo) sequences. CG50 is a synthetic dsDNA fragment including 50 repeats of the canonical CpG motif GACGTT while Clone 11 is derived from a murine CpG island. By contrast, Sumo and CGneg are mammalian or viral sequences more representative of the CG content found in total genomic mammalian DNA. The capacity of HMGB1 to bind dsDNA fragments was determined in a gel mobility shift assay. We found that HMGB1 bound comparably to all DNA fragments (Fig 1A), demonstrating that under the conditions of this assay calf thymus HMGB1 associates with dsDNA in a sequence independent manner.
In contrast to the standard TLR9 experimental ligands such as the CpG ODN 1826, the dsDNA fragments described above cannot activate B cells by themselves, even though the CG-rich fragments are potent activating moieties when incorporated into an IC that can be taken up by the BCR. Since HMGB1 has been implicated in TLR9 responses to CpG ODN in pDCs, mDCs and macrophages [1, 25, 26] and these responses in some cases were dependent on direct interaction between HMGB1 and RAGE [1, 26], it was important to determine whether RAGE, or any other HMGB1-reactive cell surface receptor, had the capacity to deliver dsDNA/HMGB1 ICs to TLR9 in B cells. To test this possibility, HMGB1 was premixed with CG50 and these complexes were used to stimulate purified AM14 B cells. We found that HMGB1/CG50 complexes completely failed to induce a proliferative response (Fig 1B). This inability to induce even a modest response was not due to the failure of HMGB1 to bind the dsDNA fragment because the addition of an IgG2a anti-HMGB1 mAb to either a premix of CG50/HMGB1 or clone 11/HMGB1 led to the formation of an IC that proved to be a potent AM14 B cell ligand. ICs formed with CG-poor dsDNA fragments did not have stimulatory activity (Fig 1B). As expected, the CG-rich/HMGB1/anti-HMGB1 response was TLR9-dependent (Fig 1C). Together these results demonstrate that HMGB1 can effectively bind dsDNA and, as previously shown for histone-bound DNA or biotinylated DNA, when complexed with both a CG-rich dsDNA fragments and an IgG2a mAb, stimulate AM14 B cells in a TLR9-dependent manner.
Even though dsDNA/HMGB1 complexes per se did not activate AM14 B cells, the question still remained as to whether an HMGB1-reactive receptor would enhance the response of AM14 B cells to either spontaneous or defined DNA-associated ICs. To address this issue, we used the mAbs PL2-3 and PA4. PL2-3 recognizes histones and PA4 directly binds dsDNA and both antibodies have been previously shown to bind cell debris present in B cell cultures to form DNAse sensitive, HMGB1-associated, ICs that activate AM14 B cells [16, 21]. Purified calf thymus DNA was then added to AM14 B cell cultures along with increasing concentrations of PL2-3 or PA4. We found that the addition of exogenous HMGB1, even up to a 30 μg/ml dose, did not enhance AM14 B cell responses to PL2-3 or PA4 at any of the concentrations tested (Fig 2A and data not shown).
We had previously determined that supernatants collected from 24 h B cell cultures contained HMGB1 at a concentration of ~150 ng/ml. Thus, one explanation for our inability to observe an enhancement of PL2-3 or PA4-dependent proliferation by exogenous HMGB1 was the cell debris DNA already incorporated sufficient B cell-derived HMGB1 to engage a putative HMGB1 receptor. As an alternate strategy, we utilized our defined dsDNA fragments ICs, where the dsDNA fragments were preincubated with calf thymus HMGB1 before combination with antibody and addition to AM14 B cells. We utilized two types of defined ICs: PA4 and dsDNA, or 1D4 (IgG2a anti-biotin mAb) and biotinylated-dsDNA. However, under these conditions we were unable to demonstrate an effect of HMGB1 (Fig 2B), even when the concentration of HMGB1 was brought up to 30 μg/ml (not shown). Thus, our results clearly show that the addition of calf thymus HMGB1 does not appear to enhance AM14 B cell responses to dsDNA-containing ICs. However, again it is difficult to completely rule out a role for B cell-derived HMGB1.
Given the difficulty is controlling the HMGB1 content of our cultures, we decided to focus on its likely receptor, RAGE. To definitively determine whether RAGE contributed to the activation of autoreactive B cells, we first verified that resting B cells express RAGE on their surface by flow cytometry. This analysis was somewhat complicated by the fact that RAGE gene had been targeted with a TK-driven GFP module such that all cells in these mice expressed medium to high levels of GFP . Several RAGE specific antibodies that were initially used for this analysis were unable to detect B cell RAGE expression, however we did find one reagent that gave reproducible staining patterns. By using a mouse IgG anti-RAGE mAb and a PE-conjugate anti-mouse IgG, we were able to to show that RAGE is expressed by a subset of splenic B cells from RAGE+/− (RAGE+) but not RAGE−/− (RAGE−) mice (Fig 3A). In support of these observations, RAGE mRNA was also detected in RAGE+ but not in RAGE− B cells (data not shown).
It has been previously shown that RAGE deficiencies affected maturation and migration of DCs [27, 29], and activation and differentiation of T cells . To determine whether RAGE-deficiency affected splenic B cell development, we stained RAGE+ and RAGE− B cells for CD93, a marker of immature B cells. In addition, to examine a possible effect on marginal zone (CD21hi CD23lo) or follicular (CD21lo CD23hi) splenic B cell distribution, we stained RAGE+ and RAGE− B cells with antibodies specific for CD21 and CD23. We found no differences in the relative proportion of immature and mature B cells (Fig 3B) or marginal zone and follicular B cells in the spleen (Fig 3C). Thus, RAGE deficiency does not appear to affect B cell development. To evaluate whether RAGE− B cell responses to BCR and TLR9 B cell mitogens were also normal, we stimulated RAGE+ and RAGE− B cells with suboptimal concentrations of F(ab)′2 fragment of goat anti-mouse IgM and the TLR ligands CpG ODN 1826, R848 (TLR7), LPS (TLR4) or Pam3CysK4 (TLR2). We found that RAGE+ and RAGE− B cells proliferated comparably to these mitogens both at optimal and suboptimal concentrations (Fig 3D and data not shown). These results indicate that RAGE deficiency does not affect BCR- or TLR- signaling cascades.
To unequivocally address the role of RAGE in the response of AM14 B cells to chromatin ICs, we then intercrossed AM14 knock-in (KI) and RAGE−/− mice. These mice had a comparable distribution of splenic immature and mature B cells (Fig 4A) and of follicular B cells (Fig 4B) to AM14 RAGE+ mice. The absence of marginal zone B cells is typical of the AM14 splenic B cell population.
To determine whether HMGB1 engagement of RAGE could possibly play a role in BCR/TLR9-dependent responses, we stimulate AM14 RAGE+ and AM14 RAGE− B cells with both spontaneous (PL2-3) and defined (PA4 + CG50) ICs. We found that these responses were completely comparable at suboptimal (Fig 4C) and optimal (not shown) concentrations of these ICs. These results indicate that RAGE is not required for autoreactive B cell proliferation by chromatin or DNA fragment IC.
In this report we showed that HMGB1 binds mammalian DNA fragments in a sequence-independent manner and that this binding can promote TLR9-dependent responses of autoreactive B cells. These responses appear analogous to those previously reported for chromatin ICs; both anti-HMGB1 and anti-histone IgG2a mAbs can direct an IC to the BCR, thereby promoting delivery of CG-rich DNA to TLR9. In the context of autoimmune disease, this process can be extrapolated to DNA-, histone- and HMGB1-reactive B cells, where again activation is likely to be facilitated by BCR-mediated uptake of a TLR9 ligand. Moreover, circulating HMGB1 in these patients  can most likely form ICs that further activate pDCs and other innate immune effector populations.
Surprisingly however, our results do not support a role for HMGB1 in the targeting of CG-rich DNA to TLR9. Based on previous data demonstrating that: (1) the AM14 response to PL2-3 was inhibited by HMGB1 antagonist A-box and soluble RAGE-Fc; (2) HMGB1 preferentially bound CpG (as opposed to GpC) ODNs; and (3) ICs formed between the anti-chromatin mAb PL2-3 and spent culture fluids incorporated HMGB1, we had hypothesized that HMGB1 preferentially bound CG-rich DNA and enhanced the uptake of these complexes through a mechanism that depended on both the BCR and the HMGB1 receptor RAGE. However, as mentioned above, we could not find any evidence that HMGB1 selectively targeted CG-rich DNA, nor could we find any defect in the ability of RAGE-deficient AM14 B cells to respond to PL2-3 ICs. While it is possible that soluble RAGE-Fc sterically hindered the interaction of PL2-3 ICs and the AM14 BCR, this explanation is unlikely to account for the inhibitory effect of HMGB1 A-box on the PL2-3 response. Interaction of RAGE-Fc with inhibitory FcγRIIB may also be discarded as an explanation since RAGE-Fc inhibited comparably the AM14 and AM14 FcγRIIB−/− B cell responses to PL2-3 ( and data not shown).
HMGB1 is a DNA binding protein that can serve as a transcriptional regulator. However, in activated macrophages, HMGB1 becomes acetylated, exits the nucleus and can be actively secreted. Soluble HMGB1 can serve as an alarmin and induce the release of inflammatory cytokines by binding to surface receptors such as RAGE, TLR2 and TLR4 [10, 28, 36]. Structure function analyses have shown that HMGB1 contains two DNA binding domains, A-box and B-box, and that the alarmin activity is associated with B box. Although A-box has no demonstrable cytokine activity, it can compete for B-box binding to the cell surface and thereby attenuate cytokine activity . A-box can therefore be considered an HMGB1 antagonist. HMGB1 can also be released from late apoptotic cells in tight association with nucleosomes. These HMGB1/nucleosomes complexes can be found in the serum of SLE patients and have been shown to recapitulate the activity of soluble HMGB1 and induce the secretion of proinflammatory cytokines through a RAGE-independent mechanism. . Thus the ability of A-box to effectively and selectively block PL2-3 activation of AM14 B cells  suggests that an HMGB1-binding receptor other than RAGE may be involved in the activation of autoreactive B cells. It is also possible that post-translational modifications associated with apoptosis or other forms of cell death may regulate the capacity HMGB1 to mediate its proinflammatory effects, and that the nuclear calf thymus-derived HMGB1 used in the current study lacked such modifications. HMGB1 has also the capacity to bind proteins of dissimilar structure and sequence [36, 38], thus suggesting that binding of cofactors may modify the biological functions displayed by HMGB1. Future studies will be needed to evaluate the impact of the various HMGB1 structures and/or HMGB1 interacting proteins on both innate and adaptive immune functions.
This work was supported by the Alliance for Lupus Research.