|Home | About | Journals | Submit | Contact Us | Français|
Anti-CD20 B cell depletion therapy (BCDT) is very effective for some patients with rheumatoid arthritis (RA), however the pathogenic role of B lymphocytes in RA and the primary targets of BCDT are unknown. The human TNF transgenic (hTNF-tg) mouse model of RA displays a chronic-progressive disease that spreads from distal to proximal joints, and is generally considered to be adaptive immune system-independent. We have previously reported that knee arthritis in hTNF-tg mice is accompanied by structural and functional changes of the adjoining popliteal lymph node (PLN), detectable by contrast-enhanced magnetic resonance imaging (CE-MRI). To better understand these changes, here we show that onset of knee synovitis and focal erosions are paralleled by PLN contraction and accumulation of large numbers of B cells in the lymphatic sinus spaces within the node. Flow cytometry from 2, 4-5, and 8-12 month old TNF-tg mice demonstrated that B cell accumulation in the PLN follows ankle arthritis, but commences before knee disease, and involves early expansion of CD21hi, CD23+, IgMhi, CD1d+, activation marker-negative, polyclonal B cells which are found to be specifically restricted to lymph nodes draining inflamed, arthritic joints. The same B cell population also accumulates in PLNs of K/BxN mice with autoantigen-dependent arthritis. Strikingly, we show that BCDT ameliorates hTNF-tg disease and clears follicular and CD21hi, CD23+ B cells from the PLNs. Based on these findings, we propose a model whereby B cells contribute to arthritis in mice, and possibly RA, by directly affecting the structure, composition and function of joint-draining lymph nodes.
RA is a chronic, progressive inflammatory-erosive autoimmune disease of the joints that affects as many as 1% of the population, predominantly females. Although major progress has been made in recent years in understanding the mechanisms of disease, many questions about RA pathogenesis remain unanswered. Clearly, autoimmunity and ultimately tissue destruction in RA are the result of the complex interaction of multiple contributing mechanisms. Pro-inflammatory cytokines, such as TNFα, IL-1 and IL6 play a critical, possibly primary role in disease (1). In particular, TNFα has emerged as a key cytokine exerting pleiotropic effects in driving the arthritis process by regulating other pro-inflammatory cytokines, promoting osteoclastogenesis, recruiting leukocytes to inflamed sites and directly driving expression of enzymes responsible for tissue damage such as metalloproteinases and oxygenases (2, 3). As a result, TNF antagonists have become common in the clinical treatment of RA (4). T cell involvement in RA is highlighted by strong genetic associations with MHC haplotypes, synovial and joint infiltration by activated T cells, and the recognized role of T cells in murine models of disease, such as collagen-induced arthritis (5). On the other hand, the contribution of B cells to disease has been more controversial: although production of autoantibodies (rheumatoid factor, anti-cyclic citrullinated peptide antibodies), accumulation of immune complexes and of ectopic germinal center-like structures in the joint and synovium are common in RA patients, they are not universal features of the disease (6-8).
Despite these uncertainties, B cell depletion therapy (BCDT) with anti-CD20 antibodies, originally developed for the treatment of B cell malignancies, has emerged in recent years as an effective strategy to ameliorate disease in patients who do not respond to more conventional therapy (9, 10). Although disease amelioration by BCDT underscores the importance of B cells in RA, clinical improvement does not always correlate with reduction of serum autoantibody levels, indicating that B cells may exert additional pathogenetic functions (reviewed in 11). B cells have the potential to promote autoimmune pathology by a number of antibody-independent effector mechanisms, in their role as antigen-presenting cells, or by secreting cytokines with pro-inflammatory activity (TNF, IFNγ, IL12p40, etc) and chemotactic factors (MIP1α and –β, CCL1, RANTES) (12, 13; reviewed in 14). B cells also play a critical role in the formation of ectopic lymphoid tissue structures, which are commonly observed in the inflamed synovium of RA patients and are thought to play a role in local pathogenesis (15, 16, reviewed in 17).
Human TNFα transgenic (hTNF-tg) mouse strains develop a disease that closely resembles RA, which although variable in features and timing based on transgene type and expression levels, is characterized by spontaneous, chronic, progressive inflammatory-erosive joint disease, generally starting in the hind paws and advancing cephalically to the knees and fore limbs (18). The initiating role of TNFα in this model, the lack of significant lymphocytic infiltration in the joints and synovium, absence of detectable serum rheumatoid factor and the finding that recombination activating gene 1 (RAG1)-deficient mice, in which the endogenous TNF gene was up-regulated by gene targeting, develop ankle arthritis indistinguishable from their RAG-competent counterparts, has led to the conclusion that TNF-induced arthritis in transgenic mice is likely B and T cell-independent (18-20; reviewed in 21, 22).
Using imaging techniques such as contrast-enhanced magnetic resonance imaging (CE-MRI) and micro-computerized tomography (μCT), we have previously shown that in the single-copy number hTNF-tg strain Tg3647 disease progression is paralleled closely by changes in the popliteal lymph node (PLN) (23, 24). These studies developed precise metrics that allow the longitudinal study of disease progression patterns in vivo. Among the most striking of the biomarkers identified is the observation that disease progression from the ankle (starting at around 1-2 months of age) to the knee (4-5 months) is accompanied by a dramatic increase in PLN size and contrast enhancement values (i.e., fluid content) (23, 24). Anti-TNF treatment in these mice reversed both arthritis and the associated changes in PLN structure, strongly suggesting a functional link between the two phenomena (23, 24).
Here we extend these findings by analyzing later changes in PLNs structure and organization associated with the progression of inflammatory-erosive disease in the knee; demonstrate an involvement of PLN B cells in these changes from the earliest stages of disease, and in particular of a CD21-high, CD23+, CD1d+ subset of B cells that accumulate specifically in inflamed nodes; identify similar B cell changes in the lymph nodes of K/BxN mice, and show that, unexpectedly, B cell depletion significantly ameliorates disease in the hTNF-tg model.
The 3647 line of TNF-transgenic mice in a C57BL/6 background were obtained from Dr. George Kollias (Institute of Immunology, Alexander Fleming Biomedical Sciences Research Center, Vari, Greece) (18). All animal studies were performed under University of Rochester Committee for Animal Resources approved protocols. Starting at 3 months of age hTNF-tg mice received CE-MRI bimonthly as described (23, 24 and below) until PLN collapse was detected, at which time they received baseline μCT (23, 24). Mouse anti-mouse CD20 monoclonal antibody (18B12 IgG2a) or non-specific placebo IgG2a antibody (2B8) (Biogen Idec, San Diego, CA), were dosed at 10mg/kg/i.p. every two weeks for 6 weeks, with continuous CE-MRI every 2 weeks, followed by post-treatment μCT scan. Knee joints were subjected to histology and cells from PLN and ILN were subjected to flow cytometry.
(KRNxNOD)F1 transgenic mice were obtained by crossing KRN transgenic males in a C57BL/6 genetic background (kindly provided by Dr. C. Benoist, Harvest University) (25) with female non-obese diabetic (NOD) mice (purchased from the Jackson Laboratory). Offspring were bled at day 21 of age, and those expressing the αVβ6-TCR KRN transgene were identified by flow cytometry. These TCR-transgene-positive mice were named K/BxN mice, and non-TCR transgenic littermates were used as controls. All K/BxN mice developed severe ankle joint inflammation around 1-month of age and the joint tissue damage progressed thereafter. The K/BxN mice and littermates used in this work were 1-year-old.
Detailed methods of CE-MRI have been previously described (23, 24). Briefly, anesthetized mice were positioned with the knee inserted in a custom-designed mouse knee coil. MR images were obtained on a 3 Tesla Siemens Trio MRI (Siemens Medical Solutions, Erlangen, Germany). Amira 3.1 (TGS, Mercury Computer Systems, Inc., San Diego, CA, USA) was used for analysis of high resolution CE-MRI data. For segmentation of the LN, Regions of Interest (ROIs) are manually drawn on postcontrast 3D stack of images and thresholded based on signal intensity ≥1500 arbitrary units (AU) to define the boundary between the LN and the fat pad surrounding the node. Tissue Statistics module is used to quantify the volume of the LN and the value of CE of this tissue and of surrounding muscle. LN capacity (LNCap) is defined as the LN contrast enhancement divided by muscle contrast enhancement and multiplied by LN volume.
Bone volume analysis was performed by scanning the knee joint in a Viva micro-CT 40 imaging system (Scanco Inc.). Patellar bone volume and 3D reconstruction of knee joint was performed using Amira 3.1 as previously described (23, 24).
Knee joints were fixed in 4.5% phosphate-buffered formalin and decalcified in 14% EDTA for 7 days. Histology sections were stained with Orange G/alcian blue (H&E). LNs were processed using two different protocols. For immunohistochemistry (IHC) PLNs were dissected and fixed in 10% neutralized formalin. Tissues were embedded in paraffin wax and deparaffined sections were quenched with 3% hydrogen peroxide and treated for antigen retrieval for 30 minutes. Sections were then stained with anti-B220 antibody (Biolegend). For multicolor immunofluorescence (IF) microscopy, fresh frozen PLNs were cut into 7μm thick sections. PLN sections were fixed with 4% paraformaldehyde, rehydrated in PBS, blocked with rat serum and stained with PE conjugated anti-IgM (eBioscience) and FITC-conjugated anti-CD3ε (Biolegend).
Single-cell suspensions were prepared from lymphoid organs at defined stages of disease, and were analyzed for expression of surface markers with combinations of the following fluorochrome-labeled antibodies: APC-Alexa 750 anti-B220 (clone RA3-6B2, eBioscience); phycoerythrin (PE) anti-IgM (clone II/41, eBioscience); Alexa Flour 700 anti-CD19 (clone 6D5, Biolegend); Alexa Flour 647 anti-IgD (clone 11-26c.2a, Biolegend); FITC anti-CD93, (clone AA4.1, eBioscience); Pacific Blure anti-CD21/35 (clone 7E9, BioLegend); PE-Cy7 anti-CD23 (clone B3B4, BioLegend); biotin anti-CD24 (clone M1/69, eBioscience); PE-anti-CD1d (clone 1B1, BD Pharmingen); PE-Cy5 anti-CD5 (clone 53-7.3, Biolegend); PE-Cy5 anti-CD80 (clone 16-10A1, Biolegend); Pacific Blue anti-CD86 (clone GL-1, Biolegend); biotin anti-CD69 (clone H1.2F3, BD Pharmingen); PerCP-Cy5.5 anti-CD25α (clone PC61, BioLegend); FITC anti-GL7 (clone GL-7, BD Pharmingen); PE-Cy7 anti-CD4 (clone GK1.5, eBioscience); FITC anti-CD3ε (clone 17A2, BioLegend); PE-Cy5 anti CD8α (clone 53-6.7, BioLegend); Alexa Flour 647 anti-CCR6 (clone 140706, BD Pharmingen); PE anti-CXCR3 (clone 220803, R&D Systems); biotin anti-CXCR5 (clone 2G8, BD Pharmingen); PE-Cy5 anti-CCR7 (clone 4B12, BioLegend); rabbit anti mouse Ki-67 (clone SP6, EPITOMICS), followed by secondary antibody PE-Goat anti rabbit Ig(H+L) (Invitrogen); and PE-Texas Red streptavidin (Invitrogen). Samples were run on LSRII cytometer and analyzed by FlowJo software (BD Pharmingen). To control for non-specific antibody binding, isotype control experiments were conducted and resulted in non-significant background stains.
Total RNA from the indicated sources was isolated using TRIZOL reagent (Invitrogen) and cDNAs were generated using random primers and Mu-MLV reverse transcriptase (Invitrogen). The cDNA samples were subjected to PCR to amplify the CDR3 region using a VHall primer (26) and the CμR primer listed below, which maps at the 3’ end of the Cμ1 exon. The PCR products were subjected to semi-nested PCR using VHall primer and an internal 6-carboxyfluorescein (6FAM)-labeled CμP primer. The semi-nested PCR products were run on an Applied Biosystems 3730 Genetic Analyzer at the University of Rochester Functional Genomic Center, and the resulting chromatograms were analyzed by Peak Scanner software version 1 (Applied Biosystems).
Primers sequences (5’-3’) are:
Linear mixed-effects regression models, with mouse as a random effect and time (treated as a continuous covariate) as a fixed effect, were used to assess changes over time based on longitudinal data. Differences between groups in the synovial volume, LN volume, LNCap and disease progression over time were tested by 2-sided t-test. P values less than 0.05 were considered significant.
We have previously shown how progression of knee synovitis in hTNF-tg mice can be followed non-invasively via CE-MRI to quantify synovial volume (SynVol), and correlated these findings with μCT and histology (23, 24). While this work largely corroborated findings from cross-sectional studies demonstrating that arthritis initiates at in the distal joints (i.e. ankle) and spreads to proximal joints (i.e. knee) over time (18-20), we discovered that increased knee SynVol is paralleled by an increase in volume and contrast enhancement (i.e. fluid content) of the adjacent PLN, yielding parameters LNvol and LN-CE, respectively, which can be combined in a single functional biomarker, lymph node capacitance (LNcap= LNvol × LN-CE) (23, 24), which correlates with lymphatic flow through a LN. However, to our surprise, we found that knee synovitis in some hTNF-tg mice is asymmetric. Moreover, this dichotomy was associated with distinct PLN phenotypes determined by CE-MRI in which the unaffected knee drains to an expanded-contrast enhancing PLN, while the contralateral knee with severe inflammatory-erosive arthritis is adjacent to a much smaller PLN that fails to take up Gd-DTPA (Figure 1A-H).
To further investigate these findings, we performed a prospective study in which hTNF-tg mice with bilateral ankle arthritis were followed with CE-MRI every 2-weeks until they presented with knee synovitis, which revealed two distinct phases of disease progression (Figure 1I). The first, characterized as the PLN “expansion” phase, is associated with increased, but relatively stable synovial volumes without bone erosions, and large LNcap values, which indicate an expanded, fluid-filled node (exemplified in Fig. 1A,C, E). Subsequently, a yet to be identified event triggers the PLN “collapse” phase, in which LNcap values decrease rapidly due to parallel reductions in both PLN volume and CE, while synovitis worsens, as highlighted by higher SynVol values (Fig. 1B, D, F). Consistent with synovitis presentation, knees that drain to an expanding PLN have no evidence of focal erosions (Fig. 1 G), whereas knees adjacent to collapsed PLN display extensive bone loss (Fig. 1H). Thus, we aimed to elucidate the cellular changes in hTNF-tg PLN, and hereafter refer to the initial PLN phase as “expansion”, and the later stage as “collapse”.
Since the close correlation of PLN changes and hTNF-tg arthritic progression suggests a direct link with pathogenesis, we examined the histological and cellular features of PLNs at both the expanded and collapsed stages. As we had previously reported, expanded PLNs display dramatically enlarged and mostly acellular paracortical sinusoidal spaces, which likely account for their increased fluid content and CE, and at least in part, in size; staining with anti-B220 indicates that most B cells reside in the follicles, although some clusters of B220-positive cells are present in the sinusoidal expansion area. (Fig.2, A,B). In contrast, collapsed PLNs display a strikingly different structure: the sinusoidal spaces are mostly completely filled with B220+ cells, although a portion of B cells are still clearly detectable in the follicular areas; B220+ cells also infiltrate more medullary areas of the node and the T cell areas (Fig. 2 C,D).
To better characterize these changes, we stained frozen sections from WT, expanding and collapsed PLN with fluorescently labeled antibodies to CD3ε and IgM (Fig.2E-H). Consistent with the immunohistochemical analysis above, we noticed relatively normal follicular and T cell zone areas in the expanded PLNs, with occasional clusters of IgM-bright cells in both the follicles and the paracortical area (Fig.2G-H). However, in collapsed PLNs, the node structure was completely disrupted, IgM-high cells have extensively invaded the central areas of the node, and the integrity of the T cell zone is lost (Fig.2F). Together, these findings strongly point to B cells as key participants in the dramatic structural and histological changes observed during PLN expansion and collapse, which accompany arthritis progression.
In order to elucidate the nature of the B cell populations involved in the observed PLN changes, we conducted an extensive analysis of PLNs and other peripheral lymphoid organs (spleen, as well as axillary, iliac and mesenteric lymph nodes – ALN, ILN, MLN, respectively) from hTNF-tg mice and wild-type controls by flow cytometry (summarized in Table 1). hTNF-tg mice were selected from several age groups corresponding to different stages of disease: “young”, 4-8 weeks old, displayed initial signs of ankle arthritis, but no detectable changes in PLNs or knees by CE-MRI; “expanded” samples were from mice with abnormally large (>5mm3) PLNs with high CE values (>3), as described above (in mice with asymmetrical PLNs, the ipsilateral ILNs draining the same leg were also included in the “expanded” group for statistical analysis); “collapsed” samples were PLNs from mice in which a remarkable decrease in LNvol (>1 mm3) and LNCap (>5) were observed over 2-weeks via CE-MRI, usually accompanied by exacerbation of knee arthritis (ipsilateral ILNs, spleens, MLNs and ALNs from mice with at least one collapsed PLN were also included in the “collapsed” category for statistical analysis); and “old” transgenics were 8-12 months of age, with advanced hind limb disease and detectable signs of ongoing arthritis in the forepaws.
The samples were analyzed by 11-color flow cytometry with a large panel of antibodies to B cell markers, as well as markers to other cell types (see Materials and Methods). Figure 3A shows the result of a representative set of flow cytometry plots for the key markers B220, IgM, CD21 and CD23 obtained from PLNs of a cohort of mice at the various age/disease groups. The complete set of data for these markers in all examined organs are summarized in Table I. The results indicate a clear expansion of B220+ B cells, the vast majority of which are IgM+, starting from the young transgenic PLN samples. The absolute numbers of PLN total B cells are on average 3- to 5-fold higher in hTNF-tgs compared to WT controls, accounting for an increase in total cellularity of the node from 1.5 to >2.2-fold. When the B220+ cells were analyzed for expression of CD23 and CD21, it became apparent that an abundant subset of B cells, co-expressing high levels of CD21 and CD23, were selectively expanded in the PLNs of hTNF-tg mice.
Analysis of the other lymphoid organs revealed a similar picture in the ILNs, which are known to also drain the posterior leg (27, and our unpublished observations), but not in the MLNs or spleens of hTNF-tg mice (Table I). Interestingly, ALNs showed significant accumulation of CD21-high, CD23+ B cells only in the older mice, in which disease had spread to the fore limbs, but not in younger hTNF-tgs, regardless of knee disease stage. Thus, CD21-high, CD23+ B cells appear to selectively accumulate in lymph nodes draining sites of arthritic inflammation, but not other nodes, and hereafter are referred to as B cells in inflamed nodes (Bin).
We then analyzed marker expression profiles on B cells gated according to CD21 and CD23 expression: CD21-low, CD23+ conventional follicular B cells (FoB), CD21-high, CD23-low marginal zone B cell (MZB)–like cells (this region was defined based on gating of MZB cells in the spleen, although these cells are virtually absent from normal LNs), and the expanded CD23+, CD21-high Bin population (Fig. 3B). The Bin population differs from FoB cells because of higher expression of CD1d, IgM, CD5 and CD24, and from MZB-like cells because of lower IgM and CD1d expression, but higher IgD (Fig. 3B). According to Allman's classification of peripheral B cell subsets (28), these cells do not match the phenotype of the T1-T3 transitional subsets, due to their lack of AA4.1/CD93 expression, and appear more similar, although not identical due to lower IgM levels, to the MZB-precursor population that is normally restricted to the spleen (28, 29, reviewed in 30).
Because of their extrafollicular localization and high IgM expression levels, we tested the possibility that Bin may correspond to an expanded, activated plasmablast-like population. However, we found that they do not express significant levels of any typical activation, germinal center or plasma cell markers, including CD80, CD86, CD69, GL7, CD138, CD27, CD25α, and they do not appear proliferative based on Ki-67 expression (Supplemental Figs. 1A,B and 2). In addition, no significant consistent increase in expression of Blimp-1 or AID is observed in PLNs of hTNF-tg mice, arguing against ongoing B cell activation (Supplemental Fig.1C). Analysis of IgH complementarity-determining region (CDR)-3 segment lengths using spectratyping also showed that no mono- or oligoclonal expansion is observed in mRNA from total B cells or sorted Bin from hTNF-tg PLNs, indicating that the rapid, early accumulation of these cells is unlikely to be driven by reactivity to one or a few auto-antigens (Fig.4).
An important question regarding significance of Bin is whether they represent a unique population restricted to the hTNF-tg model, or whether they are more generally associated with autoimmune inflammatory arthritis. To answer this question, we characterized the B cell populations in the PLNs, ILNs, MLNs and spleen of K/BxN mice, another mouse model that develops spontaneous B- and T-cell dependent arthritis within 2-months of age, due to expression of a self-reactive I-A-restricted TCR transgene to a peptide from the glucose-6-phosphate-isomerase enzyme (25, 31, 32). We analyzed samples from 4 arthritic K/BxN mice, together with 4 littermates with no detectable arthritis in their hind legs. As in the hTNF-tg mice, a very significant increase in cellularity, B cell and Bin absolute numbers and frequency was observed in the PLNs of diseased K/BxN mice compared to their healthy littermates (Fig.5 and Supplementary Table 1). A similar tendency was observed in the ILNs, but not in the spleen, while MLNs from arthritic animals displayed a small relative decrease in Bin cells compared to healthy littermates (Supplementary Table 1). Consistent with published data (33), comparison of the structure of PLNs in WT, diseased K/BxN and hTNF-tg mice by immunofluorescence also showed a significant expansion and distortion of the node's histology in K/BxN mice similar to that observed in hTNF-tg PLNs, although not as severe (not shown). Altogether, we conclude that the key observations regarding hTNF-tg PLN structure and cellular composition are shared with the K/BxN model.
It was previously reported that onset of ankle arthritis in another strain of TNF-overexpressing (gene-targeted TNF ΔARE × RAG1-/-) animals does not require the presence of B or T lymphocytes (19), although inflammatory arthritis in the proximal joints of these mice were not reported. The results discussed above however clearly implicate B cells in the dramatic PLN changes associated with disease progression in the hTNF-tg strain we used in our studies. We therefore tested the hypothesis that Bin cells are targets of anti-CD20 BCDT in hTNF-tg mice experiencing knee flare due to collapse of the draining PLN, and whether this treatment effectively ameliorates arthritic progression.
We first established that Bin cells indeed express CD20 based on flow cytometry analysis (Supplemental Fig.3). To test BCDT efficacy, a cohort of 10 hTNF-tg mice with established ankle arthritis and collapsed PLN were treated with anti-CD20 antibodies every 2 weeks for 6 weeks, and the progression of knee synovitis in these animals during the treatment period was compared to a cohort of 4 hTNF-tg animals treated with a placebo antibody. Figure 6A shows the extent of B cell depletion in PLN, ILN and spleen of a representative hTNF-tg mouse that completed BCDT. B cells were significantly depleted from the PLNs (>85% decrease in absolute numbers compared to controls), although at somewhat lower level than in spleen (>95% reduction). Interestingly, both the FoB and Bin populations were equally reduced, while the MZB-like CD23-/CD21-high cells represented the bulk of the residual cells after treatment. Strikingly, disease progression was essentially arrested in the BCDT cohort, with SynVol stabilizing over the 6 weeks. In contrast, we observed a significant increase in SynVol overtime (1 mm3/week; p<0.002) in the untreated control group, which culminated in a significant (p<0.05) increase vs. the BCDT cohort at 6-weeks (Fig. 6L). Figure 6C illustrates an extreme case in which SynVol actually decreased following BCDT; note that the increase in CE and LNcap for the PLN suggest a “reopening” of the lymphatic flow through the node. These results show that BCDT is effective in the treatment of arthritic flare in the hTNF-tg mouse, strongly suggesting a pathogenetic role for B cells in disease progression.
Although the identification of autoantibodies in the serum of RA patients dates back to the 1950s, the role that these autoantibodies and B cells may play in the disease's pathogenetic processes are still poorly understood. One of the main reasons for this uncertainty is the underlying heterogeneity of the human patient population, which provides a strong rationale for the use of genetically and etiologically homogeneous mouse models of disease to tease out possible contributing factors. Among the many available arthritis models, the hTNF-tg strain stands out for sharing several key characteristics with human RA, including the spontaneous, progressive nature of the disease, and the well-recognized key pathogenetic role of TNFα. Although experiments with the TNF-overexpressing TNF ΔARE strain in a RAG-deficient background indicated that B and T lymphocytes are not required for arthritis onset (19), the experiments detailed above highlight several key features that accompany arthritis progression in hTNF-tg mice which implicate B cells in at least some aspects of pathogenesis.
First, we show that onset of arthritic disease is paralleled by a dramatic increase in the B cell component of the draining lymph nodes, which involves most markedly a population with a unique CD23+, CD21-high, IgM-high, IgD+, CD1d+ phenotype. These B cells are preferentially restricted to LNs draining arthritic tissues, suggesting that their accumulation is dependent on signals coming from the affected joints. However, the expansion of the same population in K/BxN mice clearly indicates that this Bin population is not a unique feature of the TNF-transgenic microenvironment. The rapid and significant expansion of B cells, and particularly of the Bin population, in the early stages of disease in hTNF-tg mice does not appear to be dependent on one or a few autoantigens. Whether Bin cell expansion is equally polyclonal in K/BxN mice, in which a large proportion of lymph node B cells are known to be expressing anti-gpi antibodies (33), remains to be determined. Our clonality analysis cannot rule out the possibility that hTNF-tg PLN B cells are more broadly autoreactive (beyond the limit of detection of oligoclonal expansion by spectratyping), or that clonal populations may be selected as disease progresses, but the lack of expression of activation and plasma cell markers on these cells implies that, regardless of their antigen specificity, they are not directly involved in a conventional immune responses within the node. Thus, it seems more likely that Bin cells arise as a polyclonal, possibly antigen-independent population that is associated with arthritis, regardless of the primary etiology and nature of autoantigen, and may exert additional roles in the context of disease progression.
Interestingly, B cells with a range of phenotypes that resemble marginal zone precursors, are CD1d+, and in some cases CD21-high have been defined as a regulatory, anti-inflammatory subset in a number of murine autoimmune conditions, including arthritis models (34-37, reviewed in 38, 39). The common feature to these B-regulatory (B-reg) subsets is their ability to produce IL10, but according to our preliminary observations, hTNFtg CD23+ CD21-high B cells do not seem to be capable to prominently express this cytokine, or pro-inflammatory cytokines such as TNFα (human or mouse) and IFNγ (Supplemental Fig.3). Thus, while it is tempting to speculate that Bin cells may be a regulatory subset specifically recruited/differentiated at sites of ongoing inflammation, further analysis will be required to directly address this possibility. Interestingly, it was recently reported that T-reg cells are inherently unstable and can transition to a pathogenetic state and accumulate at inflammation sites (40), highlighting the fluid nature of regulatory populations and their potential to contribute to pathogenesis. If Bin cells do play a regulatory role in pathogenesis, however, a central function on the progression of inflammatory processes at the lymph node level seems more likely than local effects at the inflamed sites, because minimal if any lymphoid infiltrates are known to be present in the arthritic joints of these mice (20, 41, 42).
The second key observation we have made here is that a close correlation exists between the exacerbation of knee disease in the hTNF-tg strain, and significant changes in the structure of the ipsilateral PLN, with a marked reduction in the node capacitance and a massive migration of B cells into the expanded lymphatic spaces in the node (“collapse” phase). In support of this correlation we have observed hTNF-tg knees of 1-year-old mice with expanding PLN >20mm3 (10x WT) that never developed inflammatory-erosive arthritis. However, this correlation is not absolute, as we have also observed some hTNF-tg knees (~20%) with expanding PLN and inflammatory-erosive arthritis. Thus, PLN collapse appears to be a prominent, but not necessary component for the initiation of inflammatory-erosive arthritis of the knee in mice. To better understand lymphatics in this model, we have reported that arthritis in mice is accompanied by an increase in lymphoangiogenesis and that lymphangiogenesis and lymphatic drainage are reciprocally related to the severity of joint lesions during the development of chronic arthritis (43, 44). These results are consistent with expansion of the sinusoids in the draining lymph nodes and the higher PLNcap that we have shown by CE-MRI and histology, which is associated with the earlier stages of disease (23, and Figure 1). Thus, we hypothesize that the reduction in LNcap and an open sinusoidal spaces caused by B cell migration would correlate with a reduction in the lymphatic flow capacity of the draining lymph nodes, with resulting reduction in the clearance of inflammatory cells and factors from the drained sites. In the case of knee arthritis in the hTNFtg mice, this would result in the “flare” in synovitis and bone erosion that is observed in the PLN collapse phase. Because human RA is well known to alternate between moderate stages of inflammation and acute flares, the origin of which is yet unexplained, this may also represent an intriguing candidate for a more general mechanism of disease behavior.
Two critical questions with regard to the collapse process are the nature of the signals that induce migration of B cells from the follicular sites to the sinusoidal spaces, and whether B cell migration is causal to the collapse, or simply associated with it. Based on immunohistological analysis, the migrating cells are preferentially of an IgM-high phenotype, suggesting they may be the same CD23+, CD21-high cells that are observed accumulating during the expansion stage. However, only adoptive transfer experiments of purified, identifiable cells of the various subsets can answer the question of direct lineage relationship. Regardless, the distinct phenotype of the migrating population renders it amenable to specific functional studies aimed at identifying the potential chemotactic signals responsible for their unusual localization.
Finally, we have shown that BCDT is effective in ameliorating disease in hTNFtg mice. This is a startling observation, because of the commonly accepted paradigm that arthritis in TNF-overexpressing mouse models does not require adaptive immunity. Several possibilities can reconcile these findings. First, it is likely that the levels of TNF overexpression in the Tg3647 strain used here are lower than those in the TNF ΔARE mice used by Kontoyannis and co-workers, making disease in Tg3647 mice more dependent on additional mechanisms. Certainly, TNF ΔARE mice display a far more aggressive disease phenotype than Tg3647 mice, and die by 3 months of age (18, 19). An additional and more interesting possibility is that we are looking at two different stages of disease, with potentially different proximal causal mechanisms. Both Tg3647 and TNF ΔARE first develop arthritis in the ankle, where disease commences with infiltration of the tendon sheaths by granulocytes and macrophages, and the formation of osteoclasts next to the inflamed tendon sheaths (20). Then, the tenosynovitis rapidly progresses into pannus-like tissue largely void of lymphocytes, with osteoclasts mediating focal erosions. While this earlier stage is dominated by innate immunity components, we would like to suggest here that there exists a second stage, associated with knee “flare” and PLN collapse, which is B cell-dependent. If this is the case, the variability in clinical effectiveness of BCDT in RA patients may be in part due to the type/stage of disease primarily responsible for that patient's symptoms. Future pre-clinical and clinical studies prospectively designed to assess the cause-effect relationship of BCDT on lymphatic flow are warranted to test this hypothesis.
The authors would like to thank Dr. Edmund Kwok and Patricia Weber for technical assistance with the MRI, Michael Thullen for technical assistance with the μCT, Abbie Turner for assisting with animal breeding and genotyping, Ryan Tierney for technical assistance with the histology, and Dr. Jennifer Anolik for helpful discussions.
1This work was supported in part by Centocor Inc. and National Institutes of Health PHS awards AR46545, AR48697, AR54041 and AR56702.