This study reports on the characterization of B cells of germinal center (GC)-like structures infiltrating the salivary glands (SGs) of patients with Sjögren’s syndrome. Eight two-color combinations were devised to characterize the phenotype of these B cells in 11 SG specimens selected from biopsies obtained from 40 Sjögren’s syndrome patients and three normal tonsils. The 9G4 mAb, which recognizes V4.34-encoded autoAbs, enabled us to identify autoreactive B cells. Quantitative RT-PCR was used to determine the level of mRNAs for activation-induced cytidine deaminase (AICDA), repressors and transcription factors. CD20+IgD−CD38+CD21+CD24− B cells, similar to those identified in tonsil GCs, were seen in the SGs of four patients and, and since they expressed AICDA, they were termed “real GCs”. CD20+IgD+CD38−CD21+CD24+ B cells, seen in aggregates from the remaining seven samples, were characteristically type 2 transitional B cells and marginal zone-type B cells. They lacked AICDA mRNAs and were termed “aggregates”. Real GCs from SGs contained mRNAs for Pax-5 and Bcl-6, like tonsil GC cells, whereas aggregates contained mRNAs for Notch-2, Blimp-1, IRF-4, and BR3, similar to marginal zone B cells. Further experimental data in support of this dichotomy included the restriction of CXCR5 expression to real GC cells, while sphingosine 1-phosphate receptor 1 was expressed only in aggregates. In contrast, both types of B cell clusters expressed the idiotype recognized by the 9G4 mAb. Our data indicate that, in SGs, a minority of B cell clusters represent genuine GC cells, while the majority manifest features of being type 2 transitional B cells and marginal zone cells. Interestingly, both types of B cell aggregates include autoreactive B cells.
Epithelial structures in inflamed tissues of patients with Sjögren’s syndrome (SS)3 are wrapped in a sheath of lymphocytes (1). Their influx is initiated by autoimmune epithelitis (2). The aggregated cells are predominating T lymphocytes with some B lymphocytes (3). These are thought to be initially restricted to Ab production, but then take a prominent role in sustaining the immune response (4).
One key approach to unraveling the function of the different B cell subsets has been to analyze their ontogeny. Immature B cells migrate from the bone marrow to populate peripheral lymphoid organs. During their maturation in the periphery, they progress through a number of transitional stages (5). They first constitute a population termed transitional type 1 B lymphocytes (BT1). Phenotypically, they are CD21±CD23±IgM++IgD± (6), and undergo apoptosis upon BCR engagement (7). Cells that are protected from negative selection in the absence of their Ag become CD21+++CD23±IgM+++IgD++ and evolve to BT2 cells (8). The first stage of this progression is spontaneous, but subsequent progression requires BAFF (B cell-activating factor of the TNF family) (9) binding to its third receptor (BR3). Autoreactive BT2 cells are ineffective relative to nonautoreactive B cells in their ability to enter the follicles (FO), where their maturation is helped by follicular dendritic cells (FDCs) and copious amounts of BAFF (10, 11). The possibility thus emerges that autoreactive B cells, including those harboring the idiotype recognized by the 9G4 mAb (12), circumvent this checkpoint.
Once B cells have progressed beyond the pre-germinal center (GC) checkpoint, their fate is determined by the affinity of the BCRs to local Ags (13) before the induction of transcription factors (TFs). A weak signal induces B cells to express Notch-2 receptor, increases their expression of BR3 (14), and confines them to the marginal zone (MZ) of the spleen. As highlighted by Sanz et al. (reviewed in Ref. 15), MZ equivalents have been found in other lymphoid organs, and such cells might accumulate in inflamed tissue as pinpointed by Dörner and colleagues (16). Locally, they suppress Bcl-6, activate IRF-4, and reside where their sphingosine 1-phosphate receptor 1 (S1P1) occupancy (17) prevents their migration to the FOs under the effect of CXCL13 (18). Should the affinity of the BCR be strong, BT2 cells migrate to FO compartment where Bcl-6 is up-regulated. There, they initiate a GC and bring Pax-5 into play (19). As a result, the activation-induced cytidine deaminase (AID) gene is transcribed, which promotes somatic hypermutation and class-switch recombinations. The enzyme was originally defined as specific for GC B cells (20), but this statement was reappraised when AID was detected in the interfollicular large B cells (21) and GC founder cells (22). Simultaneously with AID induction, BR3 expression is lost during the GC traversal (23), and functional inhibition of Pax-5 permits plasma cell (PC) maturation. Once unleashed from Bcl-6 repression, Blimp-1 silences Bcl-6 such that PC generation is sustained (24).
AutoAb-producing B cells may escape pre-GC regulatory events, or result from subsequent somatic hypermutation. Because there are currently no means to identify them in tissue sections, one approach has been to track V4.34-expressing B cells, using the 9G4 mAb (25). Its expression serves as a surrogate for autoreactive B cells (26). These are excluded from the GCs of normal controls (27), but not from those of systemic lupus erythematosus (SLE) patients (26). Relevant to the issue of how autoreactive B cells escape this control in SS patients are increased serum levels of BAFF (28, 29), its aberrant production by B cells (30), and the high density of CXCL13 (31) in the salivary glands (SGs). All these characteristics generate the possibility of ectopic GC initiation in affected tissues. We (32) and others (33, 34, 35) have since confirmed this scenario by identifying GC-like structures surrounded by a follicular mantle (FM) in patients’ SGs. Their presence does not imply that they are functional. Even worse, their relevance has been questioned on the basis that most of the SG-infiltrating B cells in BAFF-transgenic mice exhibit a MZ-like phenotype (28). For example, these cells lack the GC B cell-associated CD10 and CD38 markers (36), and they constitute the first B lymphocytes to repopulate the SGs after B cell depletion (37).
This uncertainty prompted us to analyze the phenotype of B cells in GC-like clusters through the sequential expression of their TFs. Given the phenotypic complexity of B cell subsets in their lymphoid organs and inflammatory organs (15, 16, 36, 38), we chose to microdissect different areas of SGs to enable phenotyping of their B cells through RT-PCR identification of these TFs. We also provide clues into the integrity of GC tolerance mechanisms by tracing the emergence of B cells expressing the idiotype targeted by the 9G4 mAb. Most of the SG samples contained MZ-like areas, while few contained ectopic GCs. Nonetheless, our results indicate that, even in the absence of the necessary signals for GC-like reaction, a default mechanism that promotes autoAb production exists in these MZ equivalents.
Materials and Methods
Patients and controls
SG biopsies specimens were collected from 40 female patients, ranging in age from 31 to 72 years, and fulfilling the American-European Consensus Group criteria for the diagnosis of SS (39). Clinical details were available for all of the patients, the duration of their disease was recorded, and a particular note of the presence of extraglandular complications was made. None of them suffered from associated lymphoma, nor were they taking steroids or immunosuppressive drugs. Three children undergoing routine tonsillectomy, aged 5, 9, and 14 years, supplied the tonsils used as lymphoid organs for comparisons with the SG biopsies. All patients and donors’ parents gave informed consent, and the study was approved by our Ethics Committee.
The isotype-specific in-house ELISAs for IgM, IgG, and IgA rheumatoid factor ELISAs have previously been described (40), as well as our assay for BAFF (32). The antinuclear Abs were screened by an in-house immunofluorescence test, and anti-Ro/SSA and anti-La/SSB were identified by using commercial ELISA kits (BioMedical Diagnostics). C3 and C4 were measured by nephelemetry.
Immunofluorescence staining of SGs and tonsils
Biopsy specimens were embedded in OCT (Miles Laboratories), snap-frozen in isopentane, and stored at −80°C. Serial 10 μm-thick sections from patients’ SGs and controls’ tonsils were mounted onto poly-l-lysine-coated slides. To locate GC-like structures, the sections were stained with a HistoGene microdissection kit. This device enables distinguishing MZ B cells, which appear as dark blue, from GC-like B cells, which appear as light blue (see Fig. 2A, left). As previously explained (36), 11 of 40 SG specimens were selected for further analysis based on observing FITC-conjugated anti-CD19 mAb (Beckman Coulter) binding to B cell clusters (see Fig. 2A, right), and the absence of clonal expansion by RT-PCR of the rearranged IgV genes. GC-like structures were then acquired using Veritas laser capture microdissection system (Arcturus).
For phenotype analysis of cells in the mounted sections, mAbs were purchased from Beckman Coulter unless specified otherwise. The first step of the two-color staining was with rabbit anti-CD20 Ab (Interchim), plus FITC-conjugated anti-CD10, anti-CD21, anti-CD24, or anti-CD27 (BD Biosciences) murine mAbs. After a 40-min incubation, a second step was applied, consisting of tetramethylrhodamine isothiocyanate (TRITC)-conjugated donkey anti-rabbit Ab (Jackson ImmunoResearch Laboratories) in donkey serum-containing PBS. Given that the definition of GC usually relies on the presence of FDCs (41), these were sought and identified as such by the expression of the CD21 and CD35 membrane markers.
FITC-conjugated rabbit anti-human IgD (Dakopatts) was combined with anti-CD38 mAb, and the second Ab was developed with Alexa Fluor 594-conjugated donkey anti-mouse Ab (Jackson ImmunoResearch Laboratories). In parallel, anti-CD19 mAb was combined with rabbit anti-BR3, anti-Bcl-6 (both from ProSci), or anti-IRF-4 (Abcam) Abs. They were developed either with TRITC-conjugated donkey anti-mouse plus FITC-conjugated donkey anti-rabbit Ab, or with FITC-conjugated donkey anti-mouse plus TRITC-conjugated donkey anti-rabbit Ab (Jackson ImmunoResearch Laboratories).
Autoreactive B cells inside GCs and throughout the MZ-like areas (12, 25, 26) were tracked with the 9G4 mAb (kindly provided by Dr. Freda K. Stevenson), and the proportion of other VH4 gene family-positive B cells were evaluated with anti-idiotype LC1 mAb (kindly provided by Dr. Rizgar A Mageed). The tissues were fixed with 4% p-formaldehyde, washed again, and covered with coverslips (Vector Laboratories). Images were obtained using an Ar/Kr laser and analyzed with a TCS-NT confocal imaging system (Leica). Control rabbit IgG plus TRITC-conjugated donkey anti-mouse Ab and control mouse IgG plus FITC-conjugated donkey anti-mouse Ab did not induce background fluorescence.
A threshold for each channel (red and green) was set using ImageJ 1.41o freeware (National Institutes of Health). Pixels below this threshold are ignored for the purposes of the colocalization quantification. A scatterplot was then generated and colocalization coefficients were calculated depending on the threshold for each channel tM1 (red) and tM2 (green) as described by Manders et al. (42).
All infiltrates of interest were microdissected using the Veritas system. Total mRNA was extracted with a PicoPure RNA isolation kit (Arcturus) and amplified with a TransPlex whole transcriptome amplification kit (Sigma-Aldrich). Transcripts for AID (Hs 00221068), Pax-5 (Hs 00277134), BR3 (Hs 00175536), Bcl-6 (Hs 00153368), Blimp-1 (Hs 00153357), and Notch-2 (Hs 00225747) were quantified by RT-PCR using the TaqMan gene expression master mix (Applied Biosystems). Relative levels of gene expression of IRF-4, S1P1, CXCR4, CXCR5, lymphotoxin (LT)-α, LT-β, and BR3 were quantified by RT-PCR with SYBR Green master mix reagent in an ABI PRISM 7000 sequence detection system according to the manufacturer’s instructions. The number of threshold cycles (Ct) was counted using the ΔΔCt method, with GAPDH transcripts as an internal control. The primer sets used are reported in Table I.
|Oligonucleotide .||Sequence .|
|Oligonucleotide .||Sequence .|
The phenotype of B cells correlates with expression of the AID enzyme
Immunofluorescence analyses revealed that phenotypes of the B cell infiltrates of SGs from four patients were similar to tonsillar GC B cells (Fig. 1, left). Thus, the lymphocytes expressed CD10, CD38, and CD21+ on their B cells, and they resided within a FM of IgD+ B cells. The B cell phenotype of the remaining seven samples was CD10−CD38+CD21+, and their B cell clusters were not enclosed in a FM. They represented characteristic BT2 cells (Fig. 1, right). Of note, some cells expressed CD27, suggesting that they might be memory B cells (15). These results were substantiated by quantifying fluorescence above the tM1 and tM2 thresholds for the red and the green, respectively.
In fact, functional GCs are sites of somatic hypermutation that are dependent on AID. To be confirmed as such, the microdissected GC-like structures (Fig. 2,A) were subjected to quantitative RT-PCR to detect mRNA for AID. Interestingly, the four GC-resembling clusters contained AID mRNAs (Fig. 2,B), whereas the seven GC-nonresembling clusters did not. In brief, of 11 GC-like structures dissected for analysis, 4 contained AID+ B cells (hereafter referred to as “real GCs”), and 7 did not (hereafter referred to as “aggregates”). As previously described by us (36) and others (28, 31, 41), CD21+CD35+ FDCs were identified in real GCs as well as in aggregates (Fig. 2 A, right).
Profile of TFs in B cells from real GCs and aggregates
To substantiate the proposed dichotomy of the examined B lymphocytes, cells from real GCs and aggregates were compared for the expression of transcripts for IRF-4, Bcl-6, Blimp-1, Pax-5, and CD3 used as a control (Fig. 3,A). The latter agent is indispensable for maintaining B cell identity, Bcl-6 prevents untimely maturation, while IRF-4 enables differentiation of B cells into PCs. Naive and memory B cells express low levels of Bcl-6, in contrast to FO B cells that up-regulate Bcl-6 throughout the GC passage. Pax-5 and Bcl-6 transcripts (Fig. 3, B and C) were mainly expressed in the four real GC-containing infiltrates, thus supporting their classification.
In contrast, the seven aggregate-containing infiltrates contained more mRNAs for Blimp-1 and IRF-4 (Fig. 3, D and E) than for Pax-5 and Bcl-6 (Fig. 3, B and C). However, among these specimens, only infiltrates 2, 3, 4, and 7 showed high levels of mRNAs for Blimp-1 and IRF-4, suggesting that PCs were not prevented from emergin in the remaining three specimens. In the seven aggregates, the Notch-2 TF predominated in the former four infiltrates over the other three (Fig. 3 F). This TF is involved in the final stages of MZ B cell maturation. The implication from its weak expression is that, because of the low Bcl-6 and Pax-5 expression, as opposed to the high Blimp-1 and IRF-4 expression, these B cells are ready to differentiate into MZ-like B cells and/or into PCs.
Expression of BR3
The conversion of BT1 into BT2 cells and the requirement of BT2 cells for BAFF led us to determine the expression level of BR3. This was mainly detected in infiltrates with aggregates (Fig. 3 G), confirming our prediction of a BT2 phenotype for the cells. Of note, BR3 was highly expressed in infiltrates 8, 10, and 11, which contained low amounts of Notch-2, Blimp-1, and IRF-4. Given that BR3 is increased in BT2 cells and decreased in PCs (23, 43), differential modulation of BR3 would be relevant to BT2 cell survival and to their differentiation into PCs.
The results of mRNA measurements were confirmed at the protein level by immunofluorescence staining. BR3 was strongly expressed in aggregates and FM of tonsil GCs (Fig. 4, upper right), but not in the real GC cells. Importantly, Bcl-6 was only found in real GC cells, as in tonsil GC cells (Fig. 4, second row), and IRF-4 was only found in aggregate cells, as in tonsil cells of MZ equivalents (Fig. 4, bottom row).
Chemokine expression in B cells from the real GCs and aggregates
The presence of Notch-2 in some aggregate cells was the first clue to their MZ-like B cell nature. Accordingly, we predicted that S1P1 was expressed in this very group of infiltrates, while CXCL13 would be restricted to the other. The rationale was based on the finding that S1P1 helps retain B cells in the MZ or its equivalent by repressing CXCR5, which helps counteract CXCL13-mediated recruitment to FOs (17). Aggregate cells from the Notch-2+ infiltrates 2, 3, 4, and 7 exhibited increased levels of S1P1 transcripts and reduced those of CXCR5 transcripts (Fig. 5, A and B). Incidentally, the cells also express high levels of CXCR4 transcripts (Fig. 5 C).
In contrast, real GCs contained high levels of mRNAs for CXCR5, which helps direct B cells to the FO, unlike those for S1P1 (Fig. 5, A and B). The aggregate cells were also associated with increased amounts of LT-α and LT-β transcripts (Fig. 5, D and E). This elevated expression is consistent with the prevailing role ascribed to these two cytokines in the organization of conventional and ectopic GCs (44).
Autoreactive B cells are not excluded from ectopic real GCs
Patients with autoimmune disease fail to properly censor their transitional autoreactive B cells (45). For example, B cells in SLE patients that express the V4.34 gene participate in the GC reaction (12). To verify if a similar abnormality exists in SS, we examined real GC and aggregate cells for the expression of the V4.34 gene. In tonsillar lymphoid tissue, V4.34+ B cells were confined to the FM (Fig. 6, upper left), and thus were precluded from entering the GCs. In contrast, aggregates in patients’ SGs included B cells expressing the V4.34-associated idiotype recognized by the 9G4 mAb (Fig. 6, upper center).
Other VH4 Ig gene products recognized by the LC1 mAb were present in both pathological real GC as well as aggregate cells (Fig. 6, bottom row). These findings indicate that the potential tolerance checkpoints that have evolved to restrict autoreactive B cells entering GCs could be defective also in SS. This is a scenario akin to what has previously been shown to exist in patients with SLE (26).
Serologic and clinical parameters of the patients were reviewed to determine whether there were any correlations with the presence of real GCs or aggregates, and with the RT-PCR findings. There were indeed trends for longer disease duration and more frequent extraglandular complications in the patients with (17.3 ± 9.9 years duration and three of four patients) than in those without (12.5 ± 11.2 years duration and three of seven patients). Antinuclear Abs, including anti-Ro/SSA and anti-La/SSB Abs, and rheumatoid factor, irrespective of their isotypes, were not associated with the findings in the SGs. Low C4 levels, which are associated with ensuing lymphoma (46), were not restricted to the real GC+ patients. Whether these findings are significant requires a larger series of patients.
Our study of SGs of patients provides evidence at mRNA and protein levels that only a proportion of B cell clusters found in these sites fulfill the requisites for an ectopic GC (47). These B lymphocytes have mRNA for AID, Pax-5, and Bcl-6, all recognized as signature genes expressed during the GC reaction. Additionally, a panel of chemokines and cytokines are essential to provide necessary signals for the homing of B cells to GCs and to induce lymphoid organogenesis. There is evidence that, once recruited through the chemotactic effects of CXCL13, B cells take over the production of LT-α1β2 (48). In this respect, key roles have been assigned to LT-α and LT-β, which induce and maintain the FDC networks (49). Data generated in the present study suggest that CXCL13 could be involved in maintaining B cells in the FOs. These are the high transcription levels of CXCR5 in real GCs, but not in aggregates.
Furthermore, our results on AID expression in ectopic GCs support the proposition that the humoral immune response involving autoreactive B cells mature in such ectopic GCs. Pathogenic autoAbs would be promoted in chronically inflamed tissues. The continued recruitment of V4.34+ B cells within real GCs indicates that autoreactive B cells are not excluded from ectopic GCs. This abnormality appears to influence the ectopic GC reaction, before the acquisition of a centroblast phenotype by B cells. Similar abnormalities have previously been noted in SLE patients (26). In contrast, V4.34+ B cells that proceed through conventional GC reactions expand into post-GC IgG memory B cells and PCs, but lose expression of the idiotype recognized by the 9G4 mAb (27).
The immune system has evolved censoring tolerance mechanisms to protect the body from dangerous autoimmune reactions. These include sequestration of autoreactive B lymphocytes in the MZ or its equivalents, and preventing these cells from participating in productive GC reactions. Despite the key role assigned to ectopic GCs in generating autoreactive B cells, and our finding of genuine GCs in the SGs, most ectopic GC-like structures were revealed to be aggregates. Indeed, cells within these structures primarily include BT2- and MZ-like B cells. However, our data indicate also that some autoreactive B cells may slip through these pre-GC checkpoints to initiate genuine GC reactions.
In previous studies, we observed that rituximab-induced B cell depletion in SS patients culminates with the return of B cells into the SGs and that the repopulating cells are primarily BT1 cells (37). In normal mice, BT1 cells mature into BT2 cells. Therefore, the prediction from the path of BT1 maturation is that BT1 cells give rise to BT2 cells in the SGs. The production of high levels of BAFF to engage the numerous BR3 on SG BT2 cells supports this inference. BT2 cells found in the SGs would thus be at the crossroads of B cell development into MZ-like B cells forming MZ equivalents or GC-forming B cells.
One key issue concerns what determines the developmental decision of BT2 cells. In normal circumstances, strength of the signal generated following BCR engagement drives BT2 cells to become FO cells (51). In contrast, upon TLR engagement, BT2 cells become MZ-like B cells (52). Relevant in this respect is our observation that mRNA for TLR9 is highly expressed in microdissected aggregates (not shown). The simultaneous up-regulation of mRNAs for two key genes for the differentiation of BT2 cells in aggregates raises the possibility of BT2 cells evolving into MZ B cells that eventually differentiate into PCs: Notch-2, which controls the development of MZ B cells (10), is expressed in MZ-like B cells; and S1P1, which is coexpressed with Notch-2+ infiltrates, is needed for maintaining B cells within the MZ (50).
Our results suggest that S1P1 could play additional roles in B cell localization in patients’ SGs. Their roles could include enhancement of proliferation and IFN-γ production by CD4+ T cells, up-regulation of Fas expression by epithelial cells, and subsequent Fas-mediated apoptosis. AID−, Notch-2+, and S1P1+ B cell infiltrates also contained mRNA for Blimp-1 and IFR-4 akin to PCs. S1P1 was previously found in Ab-secreting cells, and differential regulation of S1P1 expression in PCs determines whether these cells stay in secondary lymphoid organs or home back to the bone marrow (53). This decision depends on the expression of CXCR4 (54), since S1P1 is expressed in the absence of CXCR4 (55). Thus, S1P1 expression in B cells within aggregate cells is indispensable to counteract the attraction from CXCL15.
In contrast to FO B cells, MZ B cells (and possibly MZ-like B cells) respond to T-independent Ags. The available evidence indicates that early PCs and GC B cells derive from different precursors in the MZ, although they could share part of the V gene repertoire, as well as HSM. Hence, BT2 cells and MZ-like B cell infiltrates represent the product of different clones, with some making an immediate Ab response, while others form GCs. A functional distinction is indeed illustrated by coexistence of real GC and aggregate cell clusters in one and the same SG biopsy samples (our manuscript in preparation).
Finally, our study suggests that ectopic GC formation is likely be involved in SS pathogenesis. These structures act as a conduit to recruit and expand autoreactive B cells. AID, which helps in the generation of high-affinity Igs to exogenous Ags, may also contribute to the emergence of high-affinity autoAbs in the SGs (56). Subsequent to these initial events, self-Ag presentation by autoreactive B cells, as well as by professional APCs and epithelial cells (57), would be crucial in sustaining the autoimmune process in SS.
We thank Dr. Freda K. Stevenson (University of Southampton, Southampton, U.K.) and Dr. Rizgar A. Mageed (William Harvey Research Institute, London, U.K.) for providing reagents. Dr. Rizgar A. Mageed is also acknowledged for helping us with editing the manuscript. Finally, thanks are due to Cindy Séné and Simone Forest for secretarial assistance.
The authors have no financial conflicts of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by the French Ministry for Education and Research.
Abbreviations used in this paper: SS, Sjögren’s syndrome; BT1 and BT2, type 1 and type 2 transitional B cells; BR3, BAFF receptor 3; BAFF, B cell-activating factor of the TNF family; FO, follicle/follicular; FDC, follicular dendritic cell; GC, germinal center; TF, transcription factor; MZ, marginal zone; S1P1, sphingosine 1-phosphate receptor 1; AID, activation-induced cytidine deaminase; PC, plasma cell; SLE, systemic lupus erythematosus; SG, salivary gland; FM, follicular mantle; TRITC, tetramethylrhodamine isothiocyanate; LT, lymphotoxin; AICDA, activation-induced cytidine deaminase.