A role for gut-associated lymphoid tissue in shaping the human B cell repertoire

We initially asked if GALT might recruit lymphocytes from the immature B cell pool. Subsets of B cells were analyzed in suspensions of mononuclear cells isolated from human Peyer’s patches that were selectively sampled at endoscopy. The T2 subset of transitional B cells was consistently enriched in isolated Peyer’s patch cell suspensions compared with the T1 subset (Fig. 1, A and B; and Fig. S1). Consistent with this, cells with transitional B cell phenotype were identified on the periphery of GALT germinal centers (Fig. 1, C and D), and as scattered CD10⁺ cells around the periphery of the mantle zone where it merges with the mucosal marginal zone (Fig. 1, D and E).

Recruitment of T2 cells into GALT would logically remove them from the blood circulation so that they would be selectively depleted from the venous blood that drains to the liver via the hepatic portal vein; most of the blood in the hepatic portal vein has previously passed through the gut. To confirm the selective recruitment of T2 cells to the gut, we analyzed the mononuclear cells obtained by vascular perfusion of normal liver before liver transplant. The T2 subset of transitional B cells was significantly reduced in the mononuclear cell population obtained by perfusion compared with paired donor blood samples (Fig. 1, F and G). The ratio of T2 to T1 transitional cells was higher in Peyer’s patch cells and lower in liver perfusates compared with control and paired blood samples (Fig. 1 H), further confirming the selective recruitment of T2 transitional cells to the gut.

The interaction between integrin α4β7 on lymphocytes and the endothelial ligand MAdCAM-1 that is expressed by endothelium in GALT (Brandtzaeg et al., 1999; Fig. 1 I) mediates the localization of lymphoid cells in the gut. We therefore asked if the selective recruitment of T2 transitional B cells in the gut might be associated with a higher expression of α4β7. A greater percentage of blood transitional B cells expressed high levels of α4β7 integrin (β7 expression considered representative for α4β7; Rott et al., 1996) than naive B cells, with the highest proportion of expression being within the T2 subset (Fig. 1, J and K). Residual T2 cells observed in liver perfusates were the T2 cells with lowest expression of α4β7 integrin, consistent with prior sequestration of T2 cells with higher α4β7 integrin expression in GALT (Fig. 1, J and K).

Having observed that T2 cells are selectively recruited into GALT, we considered what the consequences of this might be. GALT is rich in exogenously derived bacterial antigens from the luminal microbiota. We therefore asked whether transitional B cells might be activated in the GALT microenvironment. Freshly isolated GALT transitional B cells were observed to have phosphorylated Bruton’s tyrosine kinase (BTK), Erk, and Syk (Fig. 2 A), indicative of recent activation in vivo (Aoki et al., 1994). This frequency was significantly higher than that observed in other B cell subsets from GALT or blood (Barone et al., 2011; Fig. 2, A and B). The relatively low frequency of kinase phosphorylation in GC B cells is consistent with previous studies (Barone et al., 2011; Khalil et al., 2012).

The different percentages of cells detected by the different anti-kinase antibodies (Fig. 2 A) probably reflects the different signaling pathways that the kinases may contribute to and the different potential routes that might be involved in the activation of individual B cells, including activation through the BCR and TLRs. Higher frequencies of transitional and naive B cells with pBTK, pErk, and pSyk could be induced by polyclonal activation of peripheral blood mononuclear cell (PBMC) used as a positive control (Fig. 2 B).

To mimic the phenomena observed in vivo, T1, T2, and naive B cells were FACS sorted from blood (gating strategy shown in Fig. S2 A) and coincubated with washed heat-inactivated isolates of bacteria from the intestinal lumen. Phosphorylation of BTK in blood transitional B cells could be initiated in vitro by coincubation with washed heat-inactivated isolates of bacteria from the intestinal lumen (Fig. 2 C). Therefore antigens derived from the intestinal microbiome could potentially induce the activated state of transitional B cells in vivo. The relatively high activity of the T1 cells might be related to the expression of TLR9 and comparatively high frequency of cells expressing polyspecific receptors (Yurasov et al., 2005; Aranburu et al., 2010). Supernatant of T1, T2, and naive B cells isolated from Peyer’s patches or blood that had been incubated with washed heat-inactivated isolates of bacteria from the intestinal lumen gave positive results for presence of autoantibody, according to cut off controls provided with the assay for clinical diagnoses (Fig. 2, D and E). Such studies of cells isolated from human GALT involved pooling of samples and, therefore, it was not possible to assess variability in these experiments. The production of autoantibodies may relate to a previously described ability of some antibodies produced by plasma cells in the gut, which are known to be derived from GALT, to cross react between autoantigens and gut bacterial antigens (Benckert et al., 2011).

Blood transitional B cells have immunoglobulin heavy chain variable region gene (IGHV) sequences that are unmutated from germline (Aranburu et al., 2010; Wu et al., 2010), consistent with their naive status. However, transitional B cells can acquire somatic mutations in vitro on activation (Aranburu et al., 2010). Because GALT transitional B cells are activated in vivo (Fig. 2 A), we asked whether they might have somatic mutations in their IGV genes. Transitional and germinal center B cells were isolated from GALT by a method that yielded populations with >99% purity (Fig. S2 B). Although the majority of transitional B cells had unmutated IGHV genes, 30% had three or more mutations in IGHV. The frequency of mutation was significantly lower than that observed in the GC (Fig. 2 F). This observation was consistent between seven individual donors (Fig. 2 G). We identified examples of clonal expansion in the GC B cells by the observation of the same IGHV rearrangements in different PCR reactions (Fig. 2, H and I). However, these clones were not observed in the transitional B cell samples from the same individual. No clones were observed in the transitional B cell samples. The different population profiles of germinal center cells and transitional B cells in terms of both population sort purity and features of the immunoglobulin gene sequences confirmed that we analyzed these two CD10-expressing populations independently and that the mutations identified in the transitional B cells were highly unlikely to be a result of germinal center cell contamination.

Because T2 transitional B cells are selectively recruited to GALT where they acquire an activated state, we hypothesized that this could have a major impact on B cell repertoire development, as activation of transitional B cells by antigens in GALT would by definition remove them from the naive circulating pool. This could constitute a novel specificity checkpoint, failure of which may result in pathology, and might constitute a checkpoint that is not functional in an autoimmune disease such as SLE, which is reported to be B cell driven. We tested our hypothesis by studying SLE (Table S1; Yurasov et al., 2005, 2006; Meffre and Wardemann, 2008; Mietzner et al., 2008). We observed that lymphocytes in SLE, including transitional B cells, had significantly lower expression of α4β7 compared with healthy controls, implying a reduced potential to localize in the gut in SLE and reduced potential to access GALT (Fig. 3 A). Rheumatoid arthritis (RA) was studied as a disease control because it is also a chronic inflammatory autoimmune disease treated with immunosuppressive drugs. Expression of α4β7 by B cells in RA was the same as in healthy controls (Fig. 3 A). To exclude the possibility that lower expression of α4β7 in SLE was a consequence of treatment with hydroxychloroquine (HCQ), which is known to interfere with TLR9 signaling, we divided patients according to whether they took HCQ medication or not. Expression of α4β7 directly ex vivo was not associated with HCQ and was not associated with any other immunosuppressive treatment (Fig. 3 B). Furthermore, we observed a failure to up-regulate β7 expression on SLE blood B cells compared with normal blood B cells after in vitro stimulation with CpG and retinoic acid, and this was irrespective of HCQ treatment (Fig. 3 C).

Having observed reduced expression of α4β7 by circulating B cell subsets in patients with SLE, we reasoned that if this is relevant to immune physiology by reducing immune access to the gut, then GALT should be depleted in SLE. We analyzed random histologically normal biopsies, including 25 from healthy controls and 38 from patients with SLE. Aggregates of lymphoid cells that did not show features of GALT were observed in biopsies from healthy and SLE donors (Fig. 3 D). However, whereas multiple examples of GALT were observed in healthy donors, no examples of GALT were observed in the sample of SLE biopsies (Fig. 3, D and E). A significantly lower density of IgA plasma cells was observed in the lamina propria in SLE compared with controls (Fig. 3 F). Reduced expression of α4β7 was also detected on blood T lymphocytes from SLE patients (Fig. 3 G). The intraepithelial T cell compartment was significantly depleted in the large and small intestines of patients with SLE (Fig. 3, H and I) irrespective of treatment. CD4 and CD8 cells were present in the lamina propria, but CD8 cells did not tend to localize in the epithelium in SLE as they do in health (Fig. 3 I).

The receptors for retinoic acid and estrogen are nuclear receptors, the signaling pathways of which have been observed to interact (Hua et al., 2009). RAR binding is highly coincident and overlaps with estrogen receptor α binding. Further analysis showed that these two nuclear receptor–signaling systems could antagonistically regulate their target genes (Hua et al., 2009). Estrogens can also alter B cell behavior (Grimaldi et al., 2002; Hill et al., 2011). We added β-estradiol to lymphocytes at high and low physiological plasma concentrations. A consistent dose-dependent interference with the retinoic acid induction of α4β7 by β-estradiol was observed in healthy blood B cells (Fig. 3, J and K). We saw no evidence of differences in expression of retinoic acid receptor α, or activity of retinaldehyde dehydrogenase between SLE and health (Fig. 3, L and M; and Fig. S3), and no evidence of elevated estrogens in the small sample of the SLE patients analyzed. The mechanism of dysregulation of α4β7 integrin in SLE is not known but may be related to signals downstream of estrogen binding (Grimaldi et al., 2005).

The fate of the T2 cells activated in GALT remains unknown, but we propose that they could be precursors of either marginal zone B cells, intestinal plasma cells, or they could die. The origin of human splenic marginal zone B cells has been a tantalizing unknown for many years (Weill et al., 2009). In mice, transitional B cells mature into either naive B cells or marginal zone B cells in the spleen. This does not happen in humans where the splenic anatomy is different and the marginal zone B cell population, in contrast to murine marginal zone B cells, has somatic mutation in IGV. Data presented here suggests that human splenic marginal zone B cells could arise from transitional B cells activated in GALT, two zones suggested to share a common pool of cells from studies of infection and malignancy (Harris et al., 1996; Du et al., 1997). Recruitment to the marginal zone by activation in response to mucosal microbiota could provide the explanation for the responsiveness of this population to bacterial antigens and also their prior acquisition of mutations in IGHV. It would also explain the appearance of the splenic marginal zone that is responsive to bacterial antigens over the first 2 years of life and splenic marginal zone expansion after bacterial colonization of the mucosa (Timens et al., 1989; Harris et al., 1996; Kruschinski et al., 2004).

The intestinal IgA-producing plasma cell population that is derived from B cells activated in GALT (Barone et al., 2011), while mostly antigen specific, includes a relatively high frequency of cells secreting autoreactive and polyspecific antibodies (Benckert et al., 2011; Scheid et al., 2011). This contrasts with the bone marrow IgG plasma cell population where autoreactivity is relatively rare. It is possible that recruitment of transitional B cells to GALT could enhance the antigen-binding spectrum of IgA antibody produced in the gut. IgA is transported into the intestinal lumen where antibody diversity is paramount but where autoreactivity is not a relevant issue. Continuous recruitment of T2 B cells could also contribute to the scale of IgA production that exceeds the production of all other isotypes in the body.

We propose that selective entry and activation of T2 cells in human GALT, as demonstrated here, would affect the developing B cell repertoire by removing a proportion of B cells from the circulating pool before differentiation to mature naive B cells. We therefore suggest that this could constitute a novel checkpoint. Failure to access GALT efficiently in SLE may contribute to disease pathogenesis. SLE is a complex condition with diverse manifestations. It is not yet clear how our data relate to each individual phenotype associated with this disease.

Reagents were purchased from Sigma-Aldrich if not stated otherwise. CpG DNA was purchased from Thermo Fisher Scientific. The retinoic acid receptor (RAR) alpha antagonist Ro41 was purchased from Enzo Life Sciences.

Mononuclear cells from human peripheral blood and GALT were isolated as previously described (Vossenkämper et al., 2010). An overview of the lupus patients is given in Table S1. Blood cells used as controls were obtained from healthy female individuals between the ages of 20 and 45. Peyer’s patch biopsies were obtained during routine endoscopy. All patients gave informed written consent. The study was preformed according to human experimental guidelines and with approval of the NRES Committee London-City, and East and KCL infectious diseases BioBank, Local Research Ethics Committee working under the authority of the Southampton and South West Hampshire Research Ethics Committee. Cells were cultured in RPMI medium supplemented with 10% FBS, 100 U/ml penicillin, 100 µg/ml streptomycin, 2 mM l-Glutamine, and 1% nonessential amino acids.

Liver perfusates were collected from human liver grafts. During procurement of the donor liver, the aorta was clamped and the liver was perfused through the portal vein with 2 liters of University of Wisconsin (UW) solution to remove residual blood from the vasculature. After resection, the liver was placed in a bag and perfused with a further 1–2 liters of UW solution. After implantation of the liver, the ex vivo perfusate was retrieved. Mononuclear cells from the perfusate were isolated within 12 h by density gradient centrifugation using Ficoll Paque Plus (GE Healthcare). PBMCs were obtained from organ donors. All samples were stored in 10% DMSO-containing medium at −180°C. The study protocol was approved by the Ethics Committee at King’s College Hospital.

Staining of cells for flow cytometry, intracellular phospho-staining, and FACS was performed as previously described (Barone et al., 2011). Phospho-staining in PBMCs was performed after 30 min of culture with heat-killed bacteria from human feces. Bacteria had been grown under aerobic and anaerobic conditions on blood-agar plates for 48 h and were then washed and heat-killed for 10 min at 95°C. Flow cytometry and FACS were performed using the LSR II analyzer and FACS Aria (BD). Antibodies were used at concentrations recommended by the manufacturers: phospho-BTK, phospho-syk, phospho-ERK, CD79b, CD10, IgD, HLA-DR, and CD11c (all BD); CD19, CD38, CD27, CD24, and IgM (all BioLegend); CD20 efluor 450 and β7 (eBioscience); and dendritic cell exclusion (“lineage”) cocktail (anti-CD3/CD14/CD16/CD19/CD34; AbD Serotec). The specificities of phospho-BTK, phospho-syk, and phospho-ERK were confirmed by the manufacturers by methods including Western blotting.

Cryosections were fixed with 4% paraformaldehyde, followed by antigen retrieval in citrate buffer, pH 9, for 10 min at 95°C. For immunofluorescence, sections were incubated with the primary Abs (CD10 and IgD; Dako) for 1 h, and then washed and incubated with Alexa-fluorochrome–conjugated secondary Abs (Invitrogen) and DAPI. Specific staining was confirmed using isotype control Abs. Confocal images were acquired with the LSM 510 Meta, Plan-Neofluar 40× oil/0.50 objective (Carl Zeiss). Immunohistochemistry of paraffin sections was performed by UCL Advanced Diagnostics, London as previously described (Barone et al., 2009). CD20 was purchased from Dako and MAdCAM-1 from AbD Serotec.

Quantitative PCR was done as described previously (Barone et al., 2009). Primers and probes to quantify retinoic acid receptor expression were purchased from Applied Biosystems and data standardized to GAPDH expression.

IGHV gene rearrangements in DNA from sorted cells were PCR amplified with primers to each of the IgVH families and a nested protocol used as described previously (Dunn-Walters et al., 1995). PCR products were cloned into the pGEM-T vector (Promega) and sequenced by Beckman Coulter Sequencing Services (Beckman Coulter Genomics). Alignments were made using the IMGT/V-quest database.

B cell subsets were sorted by FACS from Peyer’s patches or PBMCs from healthy individuals. Purified B cell subsets were cultured as previously described with minor alterations (Huggins et al., 2007). Cells were cultured at a density of 5 × 10⁴ cells/200-µl well in 96-well U-bottomed plates. The base cell culture media was IMDM (Invitrogen) with 10% human AB serum (Biosera), 50 µg/ml human transferrin, 5 µg/ml human insulin, 15 µg/ml gentamicin, and 100 U/ml penicillin-streptomycin. For the initial 5 d of culture, the base media was supplemented with 20 U/ml IL-2, 50 ng/ml IL-10, 10 ng/ml IL-15, and 5 µl/well of heat-killed bacteria from human feces. After 5 d, supernatants were removed and replaced with base cell culture media supplemented with 20 U/ml IL-2, 50 ng/ml IL-6, 50 ng/ml IL-10, and 10 ng/ml IL-15. Cells were then transferred to 200-µl wells in U-bottomed plates lined with irradiated mouse fibroblasts (L-cells) expressing human CD40L. Supernatants were collected after 3 d. Qualitative detection of autoantibodies (i.e., immunoglobulin against Hep2 cell antigens) in cell supernatants was performed using the ANA-Hep2 Screen ELISA (IBL-America). To assess IgA content, the kit’s anti–IgG-enzyme conjugate was substituted with an anti–IgA-peroxide conjugate (Sigma-Aldrich). Results (in optical densities) of the Hep2 ELISA were standardized to immunoglobulin concentrations in the supernatants (TIG ELISA kit; BlueGene) and were adjusted to eliminate any contribution from the human AB serum.

Aldehyde dehydrogenase (ALDH) activity in blood DCs (identified as CD3/CD14/CD16/CD19/CD34⁻, HLA-DR⁺, CD11c⁺ population) was assessed by using the Aldefluor reagent system (STEMCELL Technologies). Aldefluor is a fluorescent substrate for ALDH. Cells expressing high levels of ALDH appear brightly fluorescent and were identified during flow cytometry by comparing the fluorescence in test samples to that in controls containing diethylaminobenzaldehyde (DEAB), a specific inhibitor of ALDH (Fig. S3).

Instat or Prism (GraphPad Software) was used to perform statistical tests. A two-tailed Student’s t test, Mann-Whitney-test, and χ² test were used as appropriate. The one-way ANOVA test was used to compare normally distributed multiple groups. Kruskal-Wallis was used for non-normally/skewed distributed multiple groups. P < 0.05 was considered significant.

Fig. S1 shows full gating strategy for identifying transitional 1 (T1), transitional 2 (T2), and mature naive B cells based on their expression of CD19, CD20, CD27, CD24, and CD38. Fig. S2 shows FACS sort strategies for isolation of transitional, mature naive, and germinal center B cells from peripheral blood or Peyer’s patches. Fig. S3 shows a B cell β7 expression after in vitro stimulation ALDH enzyme activity in blood myeloid DCs. Table S1 shows demographic and clinical details of the SLE and RA patients enrolled in this study.

We thank Sandra Martins, Dr. Gary Warnes, and Rowa Alhabbab for help with cell sorting. We thank Dr. Valerie Corrigall from Academic Rheumatology at KCL, and Dr. Jessica Manson from University College Hospital for providing SLE and RA samples. We thank Drs. Peter D. Fairclough and Andrew Rochford for providing Peyer’s patch biopsies, and Drs. Francesca Barone and Mats Bemark for valuable discussions.

A. Vossenkämper and N. Safinia are funded by the Medical Research Council, UK. A. Vossenkämper receives financial support from GlaxoSmithKline. P.A. Blair and G. Lombardi are funded by the British Heart Foundation. L.D. Fraser is funded by the Oliver Bird Rheumatism Program. L.D. Fraser, J. Spencer, and D.P. D’Cruz are funded by the St Thomas’ Lupus Trust. The authors acknowledge financial support from the Department of Health via the National Institute for Health Research (NIHR) comprehensive Biomedical Research Centre award to Guy’s & St Thomas’ NHS Foundation Trust in partnership with King’s College London and King’s College Hospital NHS Foundation Trust. The authors acknowledge the support of the MRC Centre for Transplantation.

The authors declare no competing financial interests.

Author contributions: A. Vossenkämper, P.A. Blair, N. Safinia, L.D. Fraser, T.J. Sanders, and A.J. Stagg performed experiments and acquired data. L.D. Fraser, J.D. Sanderson, K. Taylor, F. Chang, L.M. Choong, and D.P. D’Cruz analyzed data and provided clinical samples. A. Vossenkämper, P.A. Blair, J. Spencer, T.T. MacDonald, and G. Lombardi designed the study, analyzed data, and wrote the manuscript. All authors approved the final version of the manuscript.