Inflammatory lesions in the bone marrow of rheumatoid arthritis patients: a morphological perspective

The synovial tissue stands at the epicenter of joint pathology in rheumatoid arthritis (RA). As a primary target of the disease, studies on the synovium have provided invaluable insights into the mechanisms involved in disease pathogenesis. Recent work has, however, revealed the importance of a previously unseen anatomic compartment in direct contact with the joint space, namely the subchondral bone marrow. Bone marrow edema (BME) visible on magnetic resonance imaging (MRI) is clinically meaningful in both early and late RA as it associates with future development of bone erosions and poor functional outcomes. Although the histopathologic correlates of MRI-based BME in early RA remain obscure, studies in advanced disease are consistent in describing lymphocytic inflammatory infiltrates within the subchondral marrow cavity of affected joints. In this review, we discuss the nature of bone marrow lesions in patients with RA, analyze their relationship with synovitis, and explore their potential contribution to the pathological processes of the disease.

Rheumatoid arthritis (RA) is traditionally regarded as an archetypal disease of the synovial tissue. Th e synovial membrane indeed undergoes early infl ammatory changes, which include increased vascularity as well as intimal lining layer hyperplasia and accumulation of macrophages, plasma cells, T cells, B cells, dendritic cells, natural killer cells and mast cells in the sublining, among other changes [1]. Collectively, localization of infl ammation to the synovium is primarily responsible for the dysregulated cellular and molecular mechanisms that ultimately lead to the typical signs and symptoms of RA, including joint pain, stiff ness, swelling and structural changes. Although the pathogenic and clinical importance of synovitis is beyond question, it is becoming increasingly apparent that a 'synovio-centric' model of RA could be limiting. Indeed, other anatomic compartments appear to be involved at all stages of the disease. Th e most important example comes from the recognition that clinical arthritis and subclinical synovitis are anticipated by a pre-articular immunologic phase possibly developing in lymphoid tissues as well as in the lungs [2,3]. Full-blown joint disease itself spreads well beyond synovial tissue infl ammation. Established arthritis can be characterized by the involvement of at least two other compartments that are in direct contact with the joint space. Th ese include the draining lymph nodes (LNs) [4,5] and, of special relevance, the subchondral bone marrow (BM) [6].
As dysregulated B cell responses are central pathogenic events in RA and B-cell autoreactivity originates, at least in part, from defective checkpoints within the BM (a primary lymphoid organ) [7], the role of this compartment in the immunopathogenesis of RA has attracted great attention over the years. Data supporting the existence of BM abnormalities were mainly derived from studies on marrow aspirates/biopsies from the sternum and the iliac crest, which are primary hematopoietic and immunologic sites in adult life. Th e role that the 'red marrow' might play in the immunologic disturbances of RA thus appears interesting but not completely unexpected. Less predictably, evidence has accumulated supporting the additional involvement of the BM adjacent to infl amed joints at peripheral sites. Th is is normally a fat-rich tissue ('yellow marrow') where hematopoiesis is not prominent. In RA, the fat can be replaced by a vascularized, cell-rich infl ammatory tissue. Combined with imaging data coming from magnetic resonance imaging (MRI) studies, histopathologic changes described within the subchondral BM have been postulated to be intimately involved in the pathological processes producing local infl ammation and tissue

Abstract
The synovial tissue stands at the epicenter of joint pathology in rheumatoid arthritis (RA). As a primary target of the disease, studies on the synovium have provided invaluable insights into the mechanisms involved in disease pathogenesis. Recent work has, however, revealed the importance of a previously unseen anatomic compartment in direct contact with the joint space, namely the subchondral bone marrow. Bone marrow edema (BME) visible on magnetic resonance imaging (MRI) is clinically meaningful in both early and late RA as it associates with future development of bone erosions and poor functional outcomes. Although the histopathologic correlates of MRI-based BME in early RA remain obscure, studies in advanced disease are consistent in describing lymphocytic infl ammatory infi ltrates within the subchondral marrow cavity of aff ected joints. In this review, we discuss the nature of bone marrow lesions in patients with RA, analyze their relationship with synovitis, and explore their potential contribution to the pathological processes of the disease.
remodeling in RA joints [6,[8][9][10]. Here we will discuss the most relevant fi ndings related to BM alterations in RA, focusing on their morphological characteristics and patho logic signifi cance. In view of the clinical relevance of MRI fi ndings, special emphasis is given to the subchondral BM.

Systemic bone marrow
Alongside its hematopoietic function and its role in the early selection of lymphocytes, the BM is an immune regulatory organ involved in migration, selective retainment and function of innate and adaptive immune cells [11]. A variety of morphological, immunophenotypic and functional abnormalities in BM cells in RA have been detected as a consequence of augmented local production of infl ammatory cytokines and cell-cell interactions. Both myeloid and lymphoid lineage cells appear aff ected at some stage of the disease.
Th e absolute number of mononuclear cells was found markedly increased in iliac BM aspirates in RA patients compared to controls [12]. In addition, the spontaneous generation of CD14+ myeloid cells from BM progenitors in vitro as well as their maturation towards HLA-DR+ cells were accelerated in RA patients, indicating that alterations in the regenerative potential of the BM may occur [13,14]. Th e fi nding that levels of IL-6 and IL-8 were markedly elevated in BM aspirates from RA patients (compared with controls) [15] is suggestive of ongoing pathology in the BM region. DNA microarray analysis also demonstrated that BM cells from RA patients have abnormal functional networks in immune response and cell cycle when compared to those from osteoarthritis patients, with overexpression of genes that take part in the antigen presentation pathway and interferon signaling [16]. Studies in experimental arthritis have confi rmed a generalized pattern of red marrow conversion with enhanced myelopoiesis in conjunction with either increased IL-1 and IL-6 activity [17] or TNF overexpression [18]. Interestingly, high cellularity throughout the marrow seen on histologic examination in TNF-transgenic mice appears associated with a diff use bone marrow edema (BME) pattern on MRI, distinct from sub chondral BME coexisting with focal erosions [18]. Together with the observed alterations in the myeloid compartment, there is evidence that the BM environment in patients with RA may harbor aberrant B-and T-cell immune responses. Marrow plasmacytosis and hyperglobulinemia were already described in the 1950s in a number of independent studies (reviewed by Duthie and colleagues [19]). B cells derived from RA BM were later found to be capable of producing RA-associated autoantibodies, such as all rheumatoid factor (RF) isotypes [20] as well as IgM anti-citrullinated peptide antibodies (ACPAs) [21]. Supporting a pathogenic role of T cells in this disease, an abnormal accumulation of HLA-DR+CD8+ and recently activated CD3+CD4+ T lymphocytes has been detected in the iliac BM of RA patients (a site remote from infl amed joints) and also in femoral heads [12,22,23].
From a morphological perspective, it is worth noting that, in chronic infl ammatory conditions such as RA, the expanded lymphoid component of the BM may form follicle-like structures and undergo a localized process of ectopic lymphoid neogenesis. In a series of 65 marrow trephines, Engels and colleagues [24] reported morphologically benign lymphoid aggregates in 15 cases, 5 of which were diagnosed with RA or other autoimmune disorders. More recently, Kuca-Warnawin and colleagues [23] confi rmed a follicular pattern of infi ltrating lymphocytes in 9 of 15 BM trephine biopsy specimens from RA patients. Of these, fi ve exhibited features of compartmentalized lymphoid follicles, with a core of CD20+ B cells surrounded by CD4+ T cells and histologic evidence of germinal center (GC)-like structures. Function ality has yet to be demonstrated at these sites, but B-cell clonality was recognized in 9 of 15 patients (4 with an autoimmune disorder) in the series described by Engels and colleagues [24]. Local expression of the activation marker CD69 on follicular cells as well as increased levels of the T-cell growth factor IL-15 [23] are further clues that local immune cell activation may take place.
Collectively, the above-mentioned studies suggest that the systemic BM compartment may be involved in the pathogenesis of RA, though currently available data on its morphological alterations and functional properties have yet to be integrated into a coherent model of disease. Interest in the BM compartment as a potential source of biomarkers and information on drug mode of action has, however, been renewed following the successful introduction of B-cell-targeted therapies in RA. Compared to the almost complete suppression of B cells in circulation, relatively high numbers of B cells and B-cell precursors persist in the BM after rituximab treatment [25][26][27][28]. Th e remaining B cells mostly consist of memory B cells [27]. Patients exhibiting a less complete B-cell depletion and high proportion of more mature B cell precursors in their BM after rituximab therapy show a clinical response of relatively short duration [25]. Clinical response has also been shown to be preceded by a decrease in the number of memory B cells in the BM [28]. Disease relapse after rituximab could thus be attributable to several mechanisms related to BM partial depletion, involving B cell subsets at diff erent stages of diff erentiation and activation. Incomplete suppression of B cells in the BM may be due to pharmacokinetic factors, local expression of anti-apoptotic molecules, patientrelated diff erences, or others. Further investigations on the mode of action of rituximab in the BM and in other compartments of the immune system are awaited in order to gain better understanding of the pathogenetic and clinical variability of RA.

Subchondral bone marrow
Infl ammatory changes occurring within the juxta-articular BM in RA were reported over three decades ago. Barrie [29] observed cartilage damage, reparative changes and fi brovascular tissue within the subchondral marrow of surgically removed metatarsal and metacarpal heads in advanced RA. In similar surgical material, Wyllie [30] further described proliferation of fi broblasts, deposition of collagen, formation of thin-walled vascular channels, infi ltrates of macrophages, lymphocytes and plasma cells. In RA knee joints, Bromley and colleagues [31] observed the presence of blood vessels and infl ammatory cell infi ltrates including mononuclear/macrophage cells in subchondral regions remote from the synovial pannuscartilage junction. Th ese cellular aggregates were adjacent to erosions of the articular cartilage. In a similar study of RA subchondral bone removed from a patient undergoing knee replacement surgery, Watson and colleagues [32] described a local infl ammatory reaction with nodular lymphocytic infi ltrates accompanied by immuno globulin deposits near areas of destroyed cartilage. More recently, in a series of 164 hip arthroplasties performed for various joint diseases, O'Connell and colleagues [33] found a total of ten cases (six with RA) with subchondral infl ammation variably characterized by diff use and nodular distribution. Altogether, descriptions of a subchondral infl ammatory reaction in damaged RA joints gave rise at the end of the 1980s to the hypothesis of a 'bi-compartmental' model of RA, in which infl ammation of the BM of aff ected joints actively contributed to the establishment of local damage [8,31].
Th e almost complete inaccessibility of the BM com partment to either clinical examination or traditional imaging (radiography) has contributed to a lack of interest in local marrow pathology until recent years. It has only been since the advent of MRI that the subchondral bone has gained recognition as a site of pivotal importance when considering the nature of the infl ammatory process in RA. MRI-based BM involvement in the form of BME is a common feature in late and early RA, where it associates with disease activity and, most importantly, with future development of bone erosions and poor functional outcomes (for a review on BME in RA, see [9,10]

and the manuscript by FM McQueen in this edition of Arthritis Research and Th erapy).
While the clinical importance of MRI-based BM involvement has been demonstrated consistently, its exact nature and pathologic signifi cance remain largely obscure. Whilst the synovial membrane is accessible using minimally invasive techniques, which can be applied even in the earliest phases of the disease [34], histologic verifi cation of the bone edema seen on MRI is virtually impossible in early disease because of diffi culties in obtaining appropriate biopsy tissue. Th e few published studies comparing the histopathology of subchondral bone with MRI appearances include patients with longstanding disease undergoing major joint surgery, in which erosive and secondary degenerative changes could complicate the picture. Taking into account these limitations, Jimenez-Boj and colleagues [35] in 2007 significantly contributed to this fi eld of research by providing the fi rst description of the histologic correlates of BME. Th e authors had the chance to examine 12 small joints of the hand obtained from 3 RA patients scheduled for joint replacement surgery and undergoing MRI on the day before surgery. Disease duration was 8, 14 and 24 years, respectively. Despite the limited number of patients, it was consistently shown that BME corresponded to regions in which the BM fat had been replaced by an infl ammatory infi ltrate rather than true edema. Such infi ltrate was dense and highly vascularized in superfi cial regions of the marrow adjacent to synovial penetration through cortical bone (bone erosions), whilst was milder (<50% infi ltrates/tissue area) in more central areas. In the same year, McQueen and colleagues [36] independently confi rmed the presence of osteitis in seven bone specimens with BME at MRI obtained from four diff erent patients with longstanding disease. On immunohistochemical evaluation of these plus additional bone samples (n = 28) from 11 RA patients, the same authors later showed that marrow infl ammatory lesions corresponding to BME were characterized by a variable proportion of macrophages, plasma cells, CD8+ T cells and B cells [37]. Th e nature and cellular composition of BM involve ment in RA has been defi ned in detail by two further independent studies using modern immunohistochemical techniques [38,39]. Although direct comparison with MRI images was not available, our group could demonstrate that the marrow infl ammatory infi ltrate in both large and small joints of patients with long-standing RA was mainly organized into follicles of B and T lymphocytes (Figure 1) [38]. Th ese structures, which were highly variable in size and density distribution among diff erent bone samples, had acquired, in a proportion of cases, lymphoid-like morphological features, includ ing a peripheral LN addressin-positive vascular appa ratus, CD21+ follicular dendritic cell (FDC) networks and expression of the lymphoid chemokines CXCL13 and CCL21 [38]. Jimenez-Boj and colleagues [39] described a similar pattern, with preponderance of B cells within marrow aggregates. A more detailed analysis of B-cell markers revealed an abundance of CD27+ mature B cells. Plasma cells located at the periphery of marrow aggregates contributed 8 ± 4% of total cells [39]. A follicular organization of the lymphocytic infi ltrate with histologic evidence of GC-like structures has been recently confi rmed in three out of six subchondral bone specimens obtained from hip replacement surgery in patients with advanced RA [23].
Although no direct comparison is available with the histologic pattern of BME in other joint diseases, some morphological peculiarities of RA osteitis could refl ect disease-specifi c mechanisms. BME in the spondyloarthritides equally corresponds to a mononuclear infi ltrate possibly organized into follicles. Here, however, T cells largely predominate [40][41][42][43], in keeping with the limited role attributable to B cells in these conditions [44]. On the other hand, the marrow infl ammatory tissue in osteoarthritis is mainly fi brovascular with scarce cellular infi ltration [45][46][47][48]. Th is pattern could result from biomechanical (excessive loading) rather than infl ammatory alterations [49].
In summary, notwithstanding the relatively small case series analyzed and the unavailability of data on early arthritis, it appears indisputable that the subchondral aspect of the RA joint may harbor regions of bone infl ammation (osteitis) dominated by lymphocytes. Whether this process is quantitatively or qualitatively comparable to synovial infl ammation (synovitis) and to what extent it aff ects specifi c disease outcomes will be discussed in the next two paragraphs. Table 1 summarizes the main fi ndings of modern studies addressing the histologic picture of subchondral BM involvement in RA.

Morphological relationship
A detailed description of the histopathology of the infl amed synovium in RA goes beyond the scope of this review. Some specifi c aspects of the lymphoid infl am matory infi ltrate, however, deserve brief discussion in light of the similarities to BM features.
It has long been recognized that lymphocytes infi ltrating the infl amed synovium may become organized into follicular clusters in a variable but yet considerable proportion of tissue specimens [4,50,51]. Although these structures frequently lack complex organizational features [52], some may acquire specifi c lymphoid-like aspects. Th ese include the formation of high endothelial venules, the enrichment of T and B cells in partially separated areas, the diff erentiation of local stromal cells with development of FDC networks and vestiges of the fi broblastic reticular network, and the local synthesis of homeostatic chemokines [4,52,53]. Whether these structures may promote local immunological activity and to what extent they contribute to RA pathology are critical issues. Th e fi ndings of Schröder and colleagues [54] are pertinent with respect to this. After microdissection of B lymphocytes from FDC-rich lymphoid aggregates within RA synovial membrane, they performed mole cular analysis of the V-gene repertoire. Th ey found evidence for an oligoclonal repertoire with highly mutated V regions, compatible with a local antigen-driven GC-like reaction [54]. Furthermore, a large amount of plasma cells infi ltrating the synovial tissue appeared to be generated by locally activated B cells [55]. Th ese fi ndings are consistent with the observed local expression of AID (activation-induced cytidine deaminase), an enzyme critically involved in somatic hypermutation and classswitching of the immunoglobulin genes within conventional GCs of secondary lymphoid tissues [56]. Synovial lymphoid follicles may represent an histologic feature associated with disease severity and target organ damage [57]. Ectopic lymphoid aggregates are indeed associated with more severe synovial and systemic infl ammation [58,59] and B-cell-rich infi ltrates appear a marker of erosive disease [60][61][62]. Whether these relationships are linked causally or represent bystander events remains unknown. Supporting a direct link between B-cell autoimmunity and tissue damage, human autoantibodies against citrullinated vimentin have been recently shown to induce osteoclastogenesis and bone loss [63].
Although studies directly comparing marrow and synovial follicles are limited [39], data presented herein indicate that the overall morphological picture mainly coincides. Slight diff erences in cellular composition may be related to higher proportions of B cells and plasma cells and less macrophages in subchondral BM aggregates [39]. Consistent with the considerable heterogeneity observed at the synovial tissue level [52], characteristics of the lymphoid-like features may vary considerably among diff erent BM samples. Th us, subchondral BM CD21+ FDC networks and histological features of GClike structures have been described but appear rare [23,32,38] or even absent [39]. Similarly, vascular addressins are found in a variable proportion of specimens [38,39]. Other relevant aspects, such as the local organization of T-cell area fi broblastic reticular networks, have been assessed at the synovial tissue level [64] but their eventual correlates within the BM are so far unknown. Functionality at these sites has yet to be demonstrated. Th e possibility of a link between subchondral BM follicles and the pathological processes of the disease is suggested by the fi nding that these structures are more prominent in patients with large numbers of swollen joints, a high acute phase response and high titers of RF [39]. Histologically, enrichment in marrow follicles has been consistently observed in areas of cortical bone destruction [38,39,57].

Hierarchical relationship
A critical aspect for our understanding of the actual signi fi cance of marrow pathology in RA is the identifi cation of the hierarchical relationship between synovitis and subchondral osteitis. Two separate scenarios can theoretically be hypothesized: one is that RA is a disease starting in the synovial tissue, which then invades the subchondral bone cavity through established bone erosions; the other is that of a primary BM disease, which then spreads to the synovial membrane. In the absence of histopathologic data in the very early and possibly preclinical phases of the disease, both theories are based on assumptions rather than facts.
Analyses of tissue samples obtained from patients with long-standing erosive RA have consistently shown that marrow follicles are preferentially located in superfi cial areas of the subchondral bone at sites of synovial tissue penetration through cortical bone erosions ( Figure 1A) [38,39]. Such a picture appears strongly evocative of an 'outside-in' model in which marrow infl ammation originates from propagation of adjacent synovial infl amma tion. On the other hand, although no clues at present indicate that early osteitis morphologically corresponds to the osteitis of advanced disease, MRI studies in the earliest stages of RA reveal BM alterations before any obvious communication with the synovium is detectable [65]. Extensive BME has been also documented in patients with undiff erentiated arthritis later developing RA [66] as well as in patients with ACPA-positive palindromic rheumatism, a proposed model of preclinical RA [67]. Histologically, mild infl ammatory infi ltrates have been described also in deeper areas of the subchondral bone remote from the synovial-marrow junction [38,39]. More intriguingly, lymphoid reactions occurring at the synovial-marrow boundary are conspicuous whilst the corresponding synovial tissue penetrating through cor tical erosions is mainly fi brous [38]. Altogether, these data would suggest the autonomy (at least in part) of osteitis from synovitis. Th e picture is, however, further complicated by recent evidence demonstrating that the synovium and the BM can physiologically communicate through microscopic bone canals (<0.5 mm in width or depth) visible using highresolution computed tomo graphy [68]. Th ese canals could allow the transduction of infl ammation from the outside (synovium) to the inside (BM) and vice versa. Both routes are theoretically viable. In collagen-induced arthritis, mesenchymal cells ori ginat ing from the juxtaarticular BM have been shown to travel to the synovium through enlarged bone canals in the pre-arthritic phase [69]. On the other hand, in the same experimental model, expression of infl ammatory cytokines and osteoclastogenic factors has been almost exclusively detected at the synovial tissue level, suggesting that subchondral BM reactions are in fact driven by signals propagating from the synovium [70]. Exactly how the fi ndings obtained in animal models fi t into the spectrum of human RA awaits further defi nition.
Th e 'inside-out' concept may be more applicable in other forms of arthritis, such as the spondyloarthritides. In anterior spondylitis, infl ammation dominates the BM in areas with no adjacent synovium [41,42]. In contrast, the centrality of synovitis in RA remains incontestable. Imaging and histologic studies have, however, added another layer of complexity to models of pathogenesis in the infl ammatory arthritides, by demonstrating that joint infl ammation is not exclusive to the synovial membrane but also extends to the neighboring BM.

Are additional structures involved in the local infl ammatory process of RA? The possible contribution of the draining lymph nodes
As long as our perspective on joint infl ammation in RA has expanded from the synovial compartment to the subchondral BM, the draining LN has emerged as an additional player involved in multiple aspects of the disease. Th ese include the generation of local immunological responses as well as the control of cell effl ux from the joint.
LN involvement in experimental models appears early and possibly antedates clinical arthritis. An increase in the percentage of B lymphocytes as well as a high proliferation of CD8+ cells were observed in regional LNs in the latency period of adjuvant arthritis [71]. In the K/ BxN model of spontaneous autoimmunity, the LNs draining the distal joints were found to be essential for the amplifi cation of the arthritogenic B-cell response [72]. Similarly, changes in the popliteal LN have been reported prior to disease onset in TNF-transgenic mice [73,74], with accumulation of CD23+CD21highCD1high B cells [75][76][77]. Such a cell population was recently shown to diff erentiate locally, to display an enhanced ability to capture and process antigens and to exhibit a GC phenotype during T-cell-dependent immune responses [78]. Th e specifi c relevance of these fi ndings to autoimmune responses in RA is currently unknown. Also, early LN involvement in humans remains to be demonstrated. Former histologic studies on LN biopsies from diff erent anatomic sites in established RA described follicular hyperplasia and interfollicular plasmacytosis [79] as well as increased GCs with high B-cell activity [80]. In RF and/or ACPA-positive individuals at risk of developing RA and in early arthritis patients, an increase of activated CD69+ cells and a signifi cant change in the CD4/CD8 distribution was reported in inguinal LNs [81]. Ultrasound-guided biopsy of inguinal LNs appears feasible and safe [82] and promises to yield important information in the near future.
Alongside their potential role in modulating autoimmunity, the draining LNs may be also critically associated with the severity of joint involvement. Longitudinal studies in TNF-transgenic mice demonstrated a negative correlation between LN contrast enhancement and LN volume on MRI and local progression of synovitis, suggest ing that reduced LN drainage capacity may determine worst arthritis outcomes [73,74]. A similar relationship between local disease activity in the joints and ultrasonographic signs of axillary LN involvement has been recently reported in humans [5]. Histologically, arthritic fl ares in experimental models are associated with ipsilateral LN collapse due to lymphatic obstruction and diminished lymphatic fl ow [77]. Accordingly, inhibition of lymphatic drainage increases the severity of joint infl ammation in TNF-transgenic mice [75].

Relationship between subchondral bone marrow infl ammation and bone remodeling
Clinically, BME on MRI is one of the strongest predictors of bone erosions in RA [9,10,65]. However, whether such an association implies a causative link remains to be determined.
Chondroclasts and osteoclasts were morphologically described in the subchondral bone of RA knee joints over 30 years ago [83]. More recently, using specifi c immunostaining we demonstrated tartrate-resistant acid phospha tase (TRAP)-and cathepsin K-positive multi nucleated osteoclasts on the marrow side of RA samples, which were associated (in terms of density) with the extent of subchondral marrow infl ammation [38]. Others have later extended these fi ndings by providing evidence of local expression of molecules involved in osteoclastogenesis and tissue destruction, such as receptor activator of NF-κB ligand (RANKL), cathepsins and metalloproteinases [37,84]. Together with MRI studies, these data would suggest an active participation of the subchondral compartment to joint remodeling processes. RA erosions, however, anatomically and radiologically remain 'outside-in' processes. Th is pattern could be ascribed to diff erences in the local balance of erosive and reparative mechanisms between the synovial and the subchondral side of RA joints. Whilst bone-resorbing osteoclasts can be detected on both sides, bone-forming osteoblasts and osteoid deposition are recognized only on bone surfaces adjacent to the marrow (Figure 2), as suggested by studies in experimental arthritis and human RA [38,39,85,86]. Th e role of the BM in RA might thus not be entirely negative. Accordingly, it has been recently shown that repair of bone erosions in RA patients treated with TNF inhibitors, although rare, is based on bone apposition at the base of erosion and probably involves the BM [87].

Conclusion
RA remains a chronic autoimmune disease mainly involv ing the synovial tissue. However, evidence is mounting that the initiation and perpetuation of infl ammation may involve additional anatomic compartments, especially at certain stages of the disease. Following the recognition that RA-specifi c autoantibodies may arise years before the onset of clinical disease [88] and that the production of autoantibodies may persist despite eff ective suppression of synovial infl ammation [89], the question of the specifi c extra-synovial sites contributing to RA immunopathology is attracting increasing attention. In view of its role in the maintenance of B-cell central tolerance and as a survival niche for long-lived plasma cells, the systemic BM is a plausible candidate, although its involvement in the very early, pre-clinical phases of the disease remains speculative. Fascinating but still merely hypothetical is the role of secondary lymphoid organs, including LNs. Besides their potential role in systemic autoimmunity, the same compartments (BM and LNs) could be also involved in local pathology at the sites of joint infl ammation. From this perspective, the subchondral BM of aff ected joints appears tightly linked to the processes of local infl ammation and tissue remodeling, as suggested by (many) imaging studies and (few) histopathologic data. Analogous evidence is emerging for the draining LNs, whose drainage capacity appears inversely related to local arthritis severity in experi mental models.
We have just started unraveling the mystery of a 'multicompartmental' model of RA, but preliminary results encourage further research aimed at identifying novel pathogenic and clinical targets of the disease that may go beyond the well established synovial tissue immunopathologic environment.

Competing interests
The authors declare that they have no competing interests.