NLR Nod1 signaling promotes survival of BCR-engaged mature B cells through up-regulated Nod1 as a positive outcome

Although appropriate T cell antigen receptor binding to self-ligands is a well-documented step in T cell maturation known as “positive selection”(Klein et al., 2009), a positive role for self-ligand engagement by the majority of B cells remains unclear. In mice, the majority of mature B cells form follicles in the lymphoid organs, hence their name, follicular (FO) B cells. Prior work has demonstrated that B cell antigen receptor (BCR) expression is essential for the survival of B cells (Kraus et al., 2004), and delivery of a “tonic” BCR signal in the absence of BCR ligand engagement is sufficient for progression to mature FO B cells (Pelanda et al., 1997; Rowland et al., 2010). In this process, availability of the tumor necrosis factor member BAFF (B cell activating factor), provided mainly by myeloid and stromal cells in the microenvironment, is critical for allowing mature B cell survival (Mackay and Schneider, 2009; Mackay et al., 2010).

Although maturation can occur without BCR ligand when BAFF is provided, self-antigen–dependent positive selection is known to occur for two minor B cell subsets in mice, B1 B (Hayakawa et al., 1999) and marginal zone (MZ) B cells (Martin and Kearney, 2000; Wen et al., 2005a). Both of these subsets contain autoreactive B cells that produce autoantibodies (Hayakawa et al., 1999; Wen et al., 2005a; Baumgarth, 2011; Ichikawa et al., 2015). Though B1 B cells are dominantly generated in early life as a unique Lin28⁺ fetal/neonatal B-1 development outcome (Yuan et al., 2012; Zhou et al., 2015), MZ B cells are generated from BM through Lin28⁻ B-2 development after the neonatal stage. In adults, FO B cells are the major mature IgM^med/lowIgD⁺ B cell type from B-2 development, and most have no clearly detectable autoreactivity. However, some FO B cells show autoreactivity, and mutations that handicap NF-κB activation induced by BCR signaling result in a decreased frequency of FO B cells, in particular IgM^loIgD⁺ FO B cells, together with a severe reduction of B1 B and MZ B cells (Thome, 2004). Furthermore, a large fraction of the FO B cell pool expresses Nur77, a gene rapidly up-regulated by BCR ligand signaling from the transitional stage, but not in B cells, where the BCR ligand is absent, and IgM^loIgD⁺ B cells express the highest levels of Nur77 among FO B cells, suggesting that antigen-experienced cells predominate in the FO B subset (Zikherman et al., 2012). Recent data indicate that IgD BCRs require polyvalent antigens for activation, whereas they are unresponsive to monovalent antigens, in contrast with IgM BCRs (Übelhart et al., 2015). These data argued that the majority of IgM^med/lowIgD⁺ FO B cells have experienced some level of BCR engagement, with a different extent and form of engagement. However, it remained unclear whether the BCR ligand engagement experience has a positive impact on FO B cells compared with ligand ignorance.

BCR deletion or BCR editing accomplished mainly by further rearrangement of the Ig light chain (IgL) locus (Wardemann et al., 2003, 2004; Halverson et al., 2004; Nemazee, 2006) was originally described as a major negative selection mechanism that eliminates dangerous autoreactive specificities during mature B cell generation. However, BCR editing also occurs in B cells that lack self-reactivity (Cascalho et al., 1997; Braun et al., 2000), for reasons that have been debated, arguing against an exclusive role in negative selection but, alternatively, the possibility of positive selection. Here, we show that L chain editing occurs in an anti-thymocyte/Thy-1 BCR knock-in mouse model lacking self-Thy-1 ligand, resulting in preferential survival of BCR edited B cells, including FO B and MZ B cells with natural autoreactivity, and IgM^loIgD⁺ FO B cells predominantly composed of edited B cells. Generation of mature B cells via BCR editing in this model is associated with up-regulation of the NOD (nucleotide-binding oligomerization domain)–like receptor (NLR) Nod1. Nod1 recognizes the iE-DAP (γ-D-glutamyl-meso-DAP) dipeptide found in peptidoglycan present in the cell wall of gram-negative commensal gut microbiota (Chamaillard et al., 2003). A portion of this DAP-containing bacterial product is constantly turned over and translocated from the intestine into the circulation (Clarke et al., 2010), thereby providing a continual signal and competitive survival advantage for BCR-engaged mature B cells.

BCR with V_H3609/D/J_H2 μ heavy chain together with V_k21-5/J_k2 κ light chain expression (as V_H3609μ/V_k21-5κ) recognizes a Thy-1/CD90 glycoform predominantly expressed by immature thymocytes and also secreted in serum (Hayakawa et al., 1999; Wen et al., 2005a). As previously shown (and summarized in Fig. 1), expression of V_H3609μ/V_k21-5κ at the immature stage drives positive selection of B cells as anti-thymocyte/Thy-1 autoreactive (ATA) B cells in the presence of normal WT self–Thy-1 (Thy^WT; Hayakawa et al., 1999). Such ATA B cells, which induce CD5 as B1a (CD5⁺ B1 B), are largely the product of fetal/neonatal B cell development (B-1), constituting a minor B cell subset in adults, as B1 B cells (Hardy and Hayakawa, 2001). This ATA B cell positive selection is independent of CD40-signal and does not require mature T cells (Hayakawa et al., 2016). In adult B cell development (B-2) from BM, expression of V_H3609μ/V_k21-5κ BCR drives negative selection, resulting in arrest. However, when V_H3609μ cells express V_k19-17/Jκ1 light chain (as V_H3609μ/V_k19-17κ), the generation of MZ B dominates, and some FO B cells are also generated. This BCR is a natural anti-intestinal goblet cell autoantibody (AGcA) binding to the mucin 2 glycoprotein (Ichikawa et al., 2015). This positive outcome of autoreactive BCRs by B1a, MZ B, and some FO B cells was confirmed in V_H3609 Igμκ transgenic mice (Hayakawa et al., 1999) and also the V_H3609/D/J_H2 targeted insertion (knock-in) mouse line V_H3609t (Ichikawa et al., 2015). Generation of ATAμκ transgenic (ATAμκTg) mice, allowing a dominant presence of V_H3609μ/V_k21-5κ ATA B cells at the immature B cell stage, provided the evidence that ATA B cells mature as FO B cells in the absence of self–Thy-1 in Thy-1–knockout mice (Thy^KO), in contrast to their arrest under Thy^WT (Hayakawa et al., 2003; Fig. 1).

In ATAμκ Tg mice, ATA B cells dominantly mature under Thy^KO, without BCR editing. However, this ATAμκ transgene was made without addition of the Igδ-constant region (Cδ), thus lacking ATA IgD. To further assess the fate of ATA B cells from adult B cell development, a V_H3609/D/J_H2 knock-in mouse line, V_H3609t, allowing IgD generation, was crossed with V_k21-5κ L chain transgenic mice. This V_H3609t × V_k21-5Tg mouse also generated an initial predominance of B cells with the complete ATA BCR in the immature BM B cell pool (Fig. 2 A). As expected, these V_H3609t × V_k21-5Tg mice (V_H3609^+/−V_k21-5^+/−) accumulated ATA B cells in neonates that were present in the adult peritoneal cavity (PerC) as B1a with IgM⁺IgD^lo/− under a normal background (Thy^WT; Fig. 2 A, PerC) and showed progressively increased ATA serum levels with age (Fig. 2 B), as found in V_H3609t mice (Fig. S1). In the same adult mice, spleen B cells, predominantly as an outcome of B-2 development, showed a developmental arrest by ATA B cells (Fig. 2 A, red diagonal region) at the T2 transitional stage (AA4/CD93⁺CD23⁺) under Thy^WT condition, with down-regulation of surface IgM and IgD, as a result of negative selection. In the absence of self-antigen (Thy^KO), ATA BCR⁺ cells became mature B cells (AA4⁻ CD23⁺) with an IgM^med/IgD⁺ phenotype. These ATA B cell data confirmed negative selection by the presence of a self-antigen and B cell maturation by BCR tonic signaling as previously found in ATAμκ transgenic mice. Surprisingly, however, the IgM^loIgD⁺ FO B cell compartment in the absence of Thy-1 antigen was predominantly composed of BCR-edited B cells (Fig. 2 A, Thy^KO). This editing was accelerated under Thy^WT conditions (Figs. 2 A and S2 A). As shown in Fig. 2 C, in the spleen of Thy^KO mice, ATA B cells predominated at the AA4⁺ transitional stage, whereas AA4⁻ mature B cells showed an increased number of cells with a higher ratio of kappa to ATA idiotype fluorescence staining intensity (“ATAid^lo” cells). These ATAid^lo cells used the VDJ engineered chromosome (IgM^a+), and most retained the J_H-targeted VDJ (they are V_H3609id⁺); however the BCRs in many B cells paired this engineered heavy chain with an endogenous Igκ L chain while still retaining the transgenic light chain at low levels on the surface, yielding ATAid^lo cells (Fig. 2 C, center; Table S1, Igκ usage; Fig. S3, endogenous and Tg Igκ coexpression). These L chain–edited B cells became either FO B or MZ B cells in the presence or absence of Thy-1 (Fig. 2 C, right, Thy^KO; and Fig. S2 A, Thy^WT). Significantly, L chain sequence analysis of the ATAid^lo-edited FO B cell population revealed inclusion of B cells with a natural autoreactive AGcA specificity (auto⁺) by expression of a V_k19-17/J_k1 light chain paired with the V_H3609 heavy chain. This auto⁺AGcA B generation was strongly represented in L chain–edited B cells in both V_H3609t × V_k21-5Tg and ATAμκTg mice on a Thy^WT background as an outcome of negative selection (Fig. S3). Under Thy^KO conditions in V_H3609t × V_k21-5Tg mice, AGcA B cell generation was also found in the edited FO B cell pool (3/23 [13%]; Fig. 2 D and Table S1) and further increased in MZ B cells.

This observation raised the possibility that B cell editing might involve a BCR ligand–mediated signal, even on a Thy^KO background. To identify genes altered between unedited ATA FO B and both edited and WT FO B cells, we performed a microarray analysis comparing these FO B cells. In parallel, FO B cells expressing auto⁻ ATA in ATAμκTg.Thy^KO versus auto⁺ AGcA in AGcAμκTg mouse lines (Ichikawa et al., 2015) were also compared. As shown in Table S2, our microarray results consistently revealed up-regulation of the NLR family member Nod1 in the edited B cell pool and auto⁺ AGcA B cells, reaching a level similar to that in WT FO B cells, as confirmed by quantitative PCR, including the ATAid^lowest cells (Fig. 2 E). In contrast, we observed no alteration of BAFF-R or TLR (TLR2, TLR4, TLR7, and TLR9) mRNA levels (not depicted). To assess the frequency of Nod1 up-regulated cells among edited B cells, we applied a flow cytometry assay, stimulating cells with Nod1 ligand (C12-iE-DAP [CiE-DAP]) in vitro and then assessing expression of the early activation marker CD69 6 h after stimulation (Fig. 2 E). This analysis revealed that levels of Nod1-responding cells were higher in the majority of both WT FO B cells and edited B cells than in un-edited ATA B cells, as were with Nod1 mRNA levels, indicating Nod1 up-regulation by edited B cells. These unedited and edited FO B cells were equally functionally mature and capable of responding to anti–IgM stimulation (Fig. 2 E). Strikingly, the frequency of unedited ATA B cells in Thy^KO mice declined further with age, and edited B cells (and B cells expressing endogenous IgH chains) became predominant (Fig. 2 F), suggesting that edited B cells have a survival advantage.

In normal WT mice, neonatal-generated B1a cells expressed up-regulated Nod1 mRNA (Fig. 3 A). During B-2 development from adult BM, Nod1 mRNA levels slightly increased from the transitional to mature FO B stages and progressively increased in MZ B cells. B1a cell Nod1 mRNA expression continued to be the highest (Fig. 3 A). In contrast to such mature B cell subsets in WT mice, mature ATA FO B cells in nonedited ATAμκTg.Thy^KO mice showed lower levels of both Nod1 mRNA and cytoplasmic protein (Fig. 3, A and B) and the lowest response to CiE-DAP stimulation (Fig. 3 C). Nod1 is a cytosolic protein containing a Card 4 domain that is found in diverse cell types (Fritz et al., 2006). Although Nod1 mRNA/protein levels are not substantially higher in FO B cells than other cell types in WT spleen, the majority of B cells in adult spleen showed sensitive activation responsiveness to CiE-DAP stimulation (Fig. 3 D). During normal mouse B cell maturation, a large fraction of mature cells became sensitive to CiE-DAP stimulation, with the greatest enrichment occurring among IgM^loIgD⁺ FO B cells (Fig. 3 E, top). These CiE-DAP–reactive mature FO B cells increased from young to older age (Fig. 3 F). In contrast to WT mice, the large reduction of FO B cell generation and a lack of IgM^loIgD⁺ B cells in Btk mutant Xid mice (Thomas et al., 1993) was associated with low Nod1 ligand responsiveness, and Nod1 mRNA remained low in aged Xid mice (Fig. 3 G). Collectively, these Nod1 data suggested the involvement of BCR signaling in Nod1 up-regulation and survival.

To test this concept, we first sought to verify whether ATA B cells under Thy^KO are capable of undergoing Rag (recombinase-activating gene)–dependent BCR editing at the transitional stage in the spleen. In WT mice, both Rag1 and Rag2 mRNAs (and proteins) are still detectable at the transitional stage, although at very low levels (Yu et al., 1999). This low level of Rag mRNA expression at the T1/T2 transitional stage was increased in ATA B cells under a Thy^KO compared with Thy^WT background in ATAμκTg mice (Fig. 4 A). The few already edited cells at the transitional stage in V_H3609t × V_k21-5Tg.Thy^KO mice showed lower Rag levels than nonedited ATA B cells, similar to ATAμκTg mice under Thy^WT conditions (Fig. 4 B). Thus, in addition to early Rag down-regulation after exiting from the BM and before entering the spleen, the BCR editing capacity continues in the spleen at the transitional stage.

To examine whether BCR-engaged signaling leads to Nod1 up-regulation, purified FO B cells from WT mice were stimulated with anti-IgM in vitro (Fig. 4 C). This in vitro culture data did not show direct Nod1 up-regulation by anti-IgM or by LPS or CpG TLR stimulation (Fig. 4 C), regardless of the presence or absence of BAFF. Also, naive ATA B cells in the spleen (transitional B and FO B cells) stimulated with Thy^M antigen did not show Nod1 up-regulation (unpublished data). In contrast, acute temporal induction of Ptpn22 (protein tyrosine phosphatase nonreceptor 22) mRNA occurred as an immediate consequence of BCR signaling (Fig. 4 C). For T cells, TCR stimulation also up-regulates Ptpn22, and Ptpn22 functions as a negative regulator of TCR signaling (Cohen et al., 1999; Mustelin et al., 2005). During B cell development, Ptpn22 mRNA was up-regulated at transitional stages and was higher in MZ B cells, with the highest levels in B1a B cells, including protein levels (Fig. 4 D). This indicated that BCR signaling does not directly lead to Nod1 up-regulation in vitro, in contrast to Ptpn22.

In vivo, B cells are constantly interacting with the microenvironment, and the generation of MZ B cells requires Notch2 signaling triggered by interaction with delta-like 1, which is dominantly expressed by splenic endothelium (Tan et al., 2009). Thus, Nod1 up-regulation may also have occurred by interaction with non–B cells as a consequence of increased cell adhesion by a BCR-initiated signal (Wen et al., 2005a; Arana et al., 2008). To provide an in vivo microenvironment, nonediting ATAμκTg.Thy^KO mice were bred with lck-Thy-1 transgenic mice that express self-antigen at lower than physiological levels (intermediate in “Thy 60” or lowest in “Thy 10”; Wen et al., 2005a; Fig. 4 E). Lower levels of Thy-1 prevent a maturation arrest and instead yield early maturating FO B cells (by Thy 60) or matured FO B cells together strongly with MZ B cells (by Thy 10), without BCR editing (Wen et al., 2005a). As shown in Fig. 4 E, the level of Nod1 in ATA FO B cells from such mice was higher than in self-antigen unexposed ATA B cells in littermates, paralleling Ptpn22 mRNA levels. This result indicated that BCR engagement directly induces Ptpn22 up-regulation, followed by Nod1 up-regulation in a physiological microenvironment in vivo (Fig. 4 E, right). The lowest (nearly undetectable) Ptpn22 protein level in WT FO B cells (Fig. 4 D) is likely caused by weak BCR signaling, transient Ptpn22 up-regulation, and loss by degradation. This secondary induction of Nod1 in vivo does not involve CD40 signaling or follicular dendritic cells (Fig. 4 F). In lymphotoxin α–deficient (LTα^−/−) mice, the generation of follicular dendritic cells and B cell follicle formation are defective, whereas mature B cells are generated and increase in the circulation (De Togni et al., 1994). Thus, BCR signal–initiated Nod1 up-regulation occurs independently of T cell–CD40 signaling and in the absence of follicle formation.

In edited V_H3609t × V_k21-5Tg mice, both Nod1 and Ptpn22 mRNA were expressed at higher levels by arrested ATA B cells under Thy^WT conditions when compared with transitional ATA B cells in Thy^KO mice (Fig. 4 G), confirming Nod1 and Ptpn22 induction by self-antigen exposure. In contrast to the peak expression of Ptpn22 at the arrested stage in Thy^WT mice, edited B cells that became mature FO B cells expressed increased Nod1 levels compared with transitional stage on both Thy^WT and Thy^KO backgrounds, suggesting additional cell stage context dependence after BCR editing. Based upon these data, we conclude that BCR editing first provides B cells with the capability to experience BCR ligand signaling, followed by Nod1 up-regulation, generating the mature FO B cell pool, together with nonedited naive ATA B cells (Fig. 4 H).

We next asked whether Nod1 up-regulation by mature B cells plays a positive function in their activation and survival. FO B cells in normal mouse spleen were sensitive to stimulation with CiE-DAP (Fig. 3 E), and their survival was promoted, synergizing with BAFF (Fig. 5 A), in the absence of proliferation. Nod1 and BCR coengagement also enhanced survival of all B cell subsets in normal mice in vitro, but not proliferation/CFSE dilution (Fig. 5 B). Simultaneous stimulation by CiE-DAP and anti-IgM led to increased FO B cell activation and survival. Stimulation by anti-IgM alone of MZ B or B1a B cells resulted in low or undetectable CD69 induction, respectively, and most MZ B cells died after such treatment in vitro (Oliver et al., 1999). In sharp contrast, addition of Nod1 ligand prevented MZ B death (Fig. 5 B, day 3) and further increased survival when combined with BAFF (not depicted). B1a cells, which express high Nod1 and Ptpn22 levels, showed the highest spontaneous survival in culture. Simultaneous Nod1 and BCR signaling also slightly enhanced survival relative to BCR signaling alone by B1a, without inducing proliferation.

To further assess the function of Nod1 by B cells experienced with BCR signaling, ATAμκTg.Thy^KO (auto⁻), WT, and AGcAμκTg (auto⁺) mice were compared (Fig. 5 C). In auto⁺ AGcAμκTg mice, FO and MZ AGcA B cells showed increased Nod1 and Ptpn22 expression (Fig. 5 C, 3 and 4) relative to FO B cells in auto⁻ ATA B and WT mice (Fig. 5 C, 1 and 2). Although auto⁻ FO B cells had the lowest levels of Nod1 expression, CiE-DAP still stimulated some increase in their survival, and WT FO B cells showed further increased cell survival (Fig. 5 D), with a slight increase in cell size, as seen in auto⁺ B cells (Fig. 5 D, day 3, thin lines), mediated through Nod1–Ripk2 signaling (Hasegawa et al., 2008; Fig. 5 E). Under BCR stimulation alone, although auto⁺ AGcA FO B cells were initially activated (Fig. 5 D, 6 h), activation was handicapped, resulting in diminished viability, and this deficit was more severe in MZ B cells (Fig. 5 D, day 3, thick blue lines) without BAFF. This BCR signaling handicap in auto⁺ B cells was partially rescued by Nod1 co-stimulation, resulting in activation (Fig. 5 D, day 3, blue dotted lines) and decreased cell death. Thus, Nod1 signaling exerts a positive influence on B cells by promoting antigen activation by FO B cells and survival of BCR ligand experienced B cells, including B cells with detectable autoreactivity.

NF-κB signals are essential for mature B cell persistence and function in both BCR- and BAFF-mediated survival pathways (Stadanlick et al., 2008; Cancro, 2009; Mackay and Schneider, 2009; Kurosaki et al., 2010; Mackay et al., 2010). PI3K signaling also plays a critical role. In absence of the BCR, provision of a PI3K–Akt–FoxO1 signal, rather than the IKKβ canonical NF-κB signal, promoted B cell survival (Srinivasan et al., 2009). Thus, Nod1-mediated NF-κB, MAPK, and PI3K signaling was examined. Consistent with nonlymphoid cell signaling through Nod1 (Chen et al., 2009), CiE-DAP stimulation of FO B cells in WT mice resulted in phosphorylation and degradation of IκBα (Fig. 6 A). Receptor interacting serine/threonine kinase 2 (Ripk2/Rip2) is a critical adaptor protein that is recruited by oligomerized Nod1, followed by ubiquitination and TAK1 recruitment and NF-κB activation (Hasegawa et al., 2008). In accordance, B cell Nod1-dependent NF-κB activation was Ripk2 dependent (Fig. 6 A). Nod1 stimulation of B cells also initiated MAPK activation (Chen et al., 2009; Fig. 6 B). Similar to BAFF treatment of B cells (Patke et al., 2006; Mackay and Schneider, 2009), Nod1-dependent Akt phosphorylation also occurred (Fig. 6 C) through a PI3K signaling pathway (Fig. 6 D), leading to FoxO1 and Gsk3β phosphorylation (Fig. 6 E).

To understand the functional outcome in Fig. 5 D, Nod1 signal–mediated IκBα degradation was compared in FO B and MZ B samples used in Fig. 5 D, including B1a (Fig. 6 F). Also, activation of NF-κB, Akt, and MAPK by Nod1 and/or anti-IgM stimulation was compared and summarized in Fig. 6 G (time-course analysis is shown in Fig. S4). CiE-DAP treatment alone increased NF-κB signaling with IκBα down-regulation and Akt phosphorylation in all Nod1–up-regulated auto⁺ B cells and WT FO B cells compared with auto⁻ FO B cells. CiE-DAP–treated auto⁻ FO B cells also lacked JNK and Erk phosphorylation but showed low-level p38 MAPK phosphorylation. Conversely, with BCR cross-linking alone, both auto⁻ and WT FO B cells induced NF-κB, Akt, and MAPK signals. In contrast, these BCR-induced signals were weaker or lacking in auto⁺ B cells, reflecting an anergic/tolerant state. Simultaneous BCR and Nod1 activation of FO B cells from WT mice consistently showed further increased NF-κB and JNK signals compared with auto⁻ FO B cells, leading to greater activation and survival as found in Fig. 5 B. For auto⁺ B cells, induction of NF-κB, Akt, and JNK activation was consistently high upon engagement of Nod1 alone or combined Nod1/BCR (Fig. 6 G), consistent with rescue from cell death or enhanced survival. Thus, after BCR signaling in vivo, the Nod1 signaling potentiates survival through a mechanism that includes NF-κB and PI3K signaling pathways, which are known for their role in B cell persistence.

To assess whether Nod1 signaling is playing a role for B cells in a normal environment, we first retrovirally transduced BM with Nod1-GFP. Nod1-GFP–transduced BM from nonedited auto⁻ATAμκTg.Thy^KO mice or edited V_H3609t × V_k21-5Tg.Thy^KO mice was transferred into Thy^KO recipients. After ATAμκTg.Thy^KO BM transfer, the representation of an initially small portion of Nod1-GFP⁺ immature ATA B cells was strongly increased among mature ATA B cells at 3 wk in PBL and spleen FO B cells at 6 wk compared with T1 stage cells (Fig. 7 A). Thus, comparing Nod1-GFP⁺ and GFP⁻ cells with identical ATA BCR in the same mouse confirmed that Nod1 plays a functional role in enriching the representation of mature B cells (Fig. 7 B). When Nod1-GFP transduced V_H3609t x V_k21-5Tg.Thy^KO BM was transferred, this also resulted in increased Nod1-GFP⁺ cells in FO B and MZ B cells (Fig. 7 C, left). Further, BCR editing was greatly reduced in mature Nod1-GFP⁺ B cells, compared with GFP^– B cells in the same mouse, and mature GFP⁺FO B cells were predominantly ATA B cells (Fig. 7 C). In conclusion, cells with a BCR that lacks self-ligand survived in the absence of BCR editing when Nod1 was constitutively provided by retroviral transduction.

To assess potential alteration with age, PBLs of mice transferred with Nod1-GFP V_H3609t × V_k21-5Tg.Thy^KO BM were monitored for 12 mo (Fig. 7 D). As expected, GFP⁻ cells had a higher number of edited B cells (as originally found; Fig. 2 F). In contrast, ATA B cells were consistently dominant among Nod1-GFP⁺ B cells. Although the number of edited B cells slightly increased among GFP⁺ cells with age, these cells expressed a lower level of GFP than nonedited GFP⁺ ATA B cells (Fig. 7 D, middle). Also, GFP^hi+ nonedited ATA B cells continued to be IgM^medIgD⁺ B cells, unlike edited B cells (GFP^lo/−), which included IgM^loIgD⁺ B cells. This confirmed an increased survival outcome of Nod1⁺ B cells in aged mice and suggested that IgM^loIgD⁺ FO B cells are predominantly BCR signal–experienced Nod1-up-regulated B cells (Fig. 7 E).

In contrast to Nod1-GFP, reduction of Nod1 by shNod1-GFP transduction in auto⁺ AGcATg BM resulted in decreased representation among AGcA FO B cells relative to transduction with empty-vector GFP⁺ control (Fig. 8 A). These GFP experiments confirmed that Nod1 signaling is playing a functional role in enriching or depleting the representation of mature B cells, depending on its level of expression. However, in Nod1-deficient (Nod1^−/−) mice, mature B cell subset generation occurs. Because NLR Nod2 was found in B1a cells and at lower levels in MZ B and FO B cells (Fig. S5 A), there was a possibility that Nod2 signaling was compensating in Nod1^−/− mice. Thus, we tested Nod1 and Nod2 double-knockout mice (Nod1/2^−/−) and Ripk2^−/− mice, both of which have defective Nod1 and Nod2 signaling. However, both Nod1/2^−/− (Fig. 8 B) and Ripk2^−/− (Fig. 8 C) mice showed B cell generation similar to WT and Nod1^−/− mice. Importantly, we found that mature FO B cells in Ripk2^−/− mice had Nod1⁺ and Ptpn22⁺ mRNA levels similar to those from Ripk2^+/+ littermates, and BCR signaling was normal (Fig. 8 C). These results indicated that the Nod1 signaling is not required in the initial generation and maturation of BCR-engaged B cells with Nod1 up-regulation.

Because both Nod1 up-regulated and non–up-regulated B cells mature and survive in the presence of BAFF, a possible role for Nod1 signaling in normal mice was competitive survival, considering our auto⁻ and auto⁺ B cell GFP transduction data. Thus, we performed 1:1 mixed BM chimeric transfer from WT (Ly5.1) and Nod1/2^−/−(Ly5.2) mice. Compared with B cells at the transitional stage (“1”; IgM^hi IgD^lo AA4⁺), the WT to Nod1/2^−/− B cell ratio was slightly but consistently increased in mature B cells in PBLs and in FO B and MZ B cells in the spleen (Fig. 8 D). Thus, the frequency of Nod1/2^−/− B cells decreased with maturation. When BM cells from Ripk2^−/− mice were cotransferred with WT, increased competitive survival of Ripk2⁺ (WT) FO and MZ B cells also occurred (Fig. 8 E), including a higher WT/Ripk2^−/− ratio by MZ B lineage.

In the absence of competitive B cells, Ripk2 deficiency in auto⁺AGcAμκTg mice also did not affect AGcA B cell maturation (Fig. 8 F). When WT BM cells were cotransferred at a three times greater frequency than AGcA BM cells, Ripk2^+/+AGcA B cells became the preferred CD21^hi B cell type entering the MZ B subset (Fig. 8 G, top), as previously observed (Ichikawa et al., 2015). In contrast, generation of Ripk2^−/− AGcA B cells was significantly reduced with limited MZ B cell generation (Fig. 8 G, bottom). LPS- and CpG TLR-mediated proliferation was normal in Ripk2^−/− AGcA B cells, whereas Ripk2 deficiency blocked rescue from BCR signal–mediated death upon Nod1 ligand addition (Fig. S5 B). Thus, it is clear that a Nod1–Ripk2 signaling is occurring in Nod1⁺ B cells, and an intact Nod1 signal confers a competitive advantage for both FO B and MZ B cell generation and maintenance.

Chimeric transfers of purified splenic B cells also revealed a gradual increase of Rip2^+/+ WT B cells over Ripk2^−/− B cells in PBLs (Fig. 8 H). 12 wk posttransfer, after 40–50% B cell loss, 20–30% of the surviving B cells in circulation and the spleen were IgM^hiIgD^lo AA4⁻ (with up-regulated Blimp1; not depicted) and predominantly Ripk2^+/+ (Fig. 8 H). This confirmed that a physiological Nod1 ligand interacts directly with responsive B cells, and the Nod1–Ripk2 signal promotes competitive survival for BCR-engaged Nod1–up-regulated B cells, including edited B cells (Fig. 8 I).

Our data demonstrate that Nod1 up-regulation has a positive impact on BCR-engaged mature B cells. Exposure of the BCR to self-ligands is a critical event that impacts B cell development. Through B-2 development, antigen-unexposed naive B cells can become mature FO B cells when sufficient BAFF is provided. Such naive B cell maturation without BCR ligand signaling was confirmed in our ATA knock-in mouse system. However, we also observed an increase in BCR-edited B cells that predominated with age. We show here that these edited mature B cells up-regulate Nod1 as an outcome of BCR engagement in vivo. Although Nod1 up-regulation also occurs at the arrested stage during negative selection, mature B cells express even higher Nod1 mRNA levels and Nod1 ligand responsiveness. This indicates that Nod1 up-regulation is cell stage context dependent and only reaches functional levels at the mature cell stage in vivo. A positive outcome by increased Nod1 was confirmed by transduction, and up-regulation of Nod1 in normal mice promotes competitive survival by mature B cells.

Nod1 signaling does not require initial BCR-engaged mature B cell generation. This BCR-engaged Nod1⁺ mature B cell generation might be due to either altered BCR sensitivity of mature B cells in sensing their microenvironment and/or increased simultaneous signaling by BCR ligand and BAFF as an initial positive outcome. It is after such BCR-engaged mature B cell generation that the NLR signal begins to play a role. The presence of Nod1 ligand, a commensal microbial product, promotes B cell survival, synergizing with BAFF and conferring an advantage relative to B cells lacking BCR-ligand exposure. Thus, the FO B and MZ B cell subsets become further enriched for B cells with NLR signaling capacity. Our finding of an initial BCR ligand signal, followed by a secondary Nod1 signal promoting competitive survival, explains the importance of canonical NF-κB and PI3K signal involvement for FO B cell generation and maintenance. Although BCR-engaged selection occurs, the immune system must cope with the unlimited diversity of antibodies that B cells can generate. Because a great degree of conformational flexibility exists in the unmutated germline structure of antibodies (Wedemayer et al., 1997), the mature FO B cell pool retains great diversity as a primary repertoire.

It is still unclear how BCR signaling is involved in Nod1 up-regulation, because BCR ligand–mediated signaling does not directly induce Nod1. In vivo, developing B-2 B cells circulate from BM, entering the spleen which is the main site filtering the blood. In this process, B cells encounter the microenvironment, including serum immunoglobulins, dendritic cells, and vascular endothelial cells. Considering that BCR ligand signaling promotes cell adhesion (Wen et al., 2005a; Arana et al., 2008), interaction with nonlymphoid cells and increase of Nod1 may occur. The accumulation of Nod1 up-regulated B cells appears to be the result of a complex interplay with the microenvironment. Initially, it is likely to be dependent on the amount, form, and valency of antigen engaging the BCR, including polyvalent antigen engagement to become IgM^loIgD⁺ FO B cells (Übelhart et al., 2015), followed by interaction with nonlymphoid cells.

The positive influence of coexisting microorganisms on the immune system has become increasingly apparent (Mazmanian and Kasper, 2006). After birth, microbiota-mediated Nod1 signaling by epithelia can contribute to the initial maturation of isolated lymphoid follicles with B cells in intestine (Bouskra et al., 2008). Nod1 signaling through nonhematopoietic cells also strongly contributes to induction of adaptive immunity together with antigen (Fritz et al., 2007). Now, our data show that mature B cells expressing Nod1 can directly respond to Nod1 ligands. Because stimulation of cytosolic Nod1 does not require physical contact with live bacteria, and because bacterial products can circulate (Clarke et al., 2010), mature Nod1⁺ B cells become sensitive to Nod1 signals. Increased sensitivity to the Nod1 ligand induces survival of B cells, but not proliferation. Thus, the increase in Nod1 in mature B cells promotes a positive survival outcome, increases BCR activation, and rescues cells from BCR-induced tolerance. Increased survival by Nod1 signaling in aged mice may also involve NLR in autophagy as a host defense by maintaining immune homeostasis (Philpott et al., 2014). An increase in autoreactive B cells with age in both mice and humans may be caused in part by continuous NLR signaling throughout life, a consequence of the presence of commensal microbiota. Although it is possible that nonautoreactive B cells also become high Nod1-expressing B cells, because the BCR signal does not directly induce Nod1, understanding the function of commensal microorganisms in enriching Nod1 up-regulated B cells that play a beneficial role and, conversely, pose a risk of autoimmune disease or cancer with aging is important and requires future investigation.

V_H3609/V_k21-5 ATA and V_H3609/V_k19-17 AGcAμκ transgenic mouse lines (3369 as ATAμκTg and EP67 as AGcAμκTg, respectively), Thy^KO, lck-Thy-1 transgenic lines, the V_H3609/D/J_H2 targeted insertion (V_H3609t) line, and the V_k21-5/J_k2 κ Tg mouse line have all been previously described (Hayakawa et al., 2003; Wen et al., 2005a; Ichikawa et al., 2015). All Ig transgenic and V_H3609t mouse lines were backcrossed onto the C.B17 background. Xid mice are on a BALB/c background, and CD40^−/− mice are on a C.B17 background. 2–3-mo-old C.B17 (or BALB/c) mice were used as WT mice in most experiments (except as noted). C57BL/6 (B6) mice, and B6.Ly5.1 mice for cotransfer, were used as WT controls in experiments with Nod1^−/−, Nod1/2^−/−, Ripk2^−/−, or LTα^−/− mice (all on a C57BL/6 background). The AGcAμκTg mouse line was also backcrossed to C57BL/6 (Ichikawa et al., 2015) and then was crossed with Ripk2^−/− mice for cotransfer experiments. Nod1^−/− mice were originally made by T.W. Mak (University of Toronto, Toronto, Ontario, Canada), and Nod1/2^−/− mice were provided by R. Tsolis (University of California at Davis, Davis, CA). All animal experiments were conducted under a protocol approved by the Fox Chase Cancer Center Institutional Animal Care and Use Committee.

Multicolor flow cytometry analysis, sorting, and monoclonal antibody reagents, including rat anti–mouse Ig idiotype antibodies, P9-19A4 (anti-ATAid), and P9-10C7 (anti-V_H3609^SM6C10 id), have been described previously (Hayakawa et al., 2003; Wen et al., 2005a; Ichikawa et al., 2015). APC-A20 and FL-104 antibodies were used for anti-CD45.1 and anti-CD45.2 staining. Analysis and sorting for B cell subsets: FO B (B220⁻AA4⁻CD21^medCD23⁺), MZ B (B220⁺AA4^–CD21^hiCD23⁻), B1a (CD19⁺B220^loCD5⁺).

Gene expression was quantitated by real-time PCR, with duplicate samples, using TaqMan assays from Applied Biosystems, an ABI 7500 real-time thermal cycler, and ABI software (Life Technologies). Relative gene expression levels were normalized using β-actin values as a standard. A minimum of three samples were used, with up to five samples (duplicate analysis for each), to determine mean, standard error, and p-value by t test. For B cell subset quantitative PCR, a minimum of five samples were used. Nod1 RT-PCR primer set was 5′-GATTGGAGACGAAGGGGCAA-3′ and 5′-TGGCTGTGTTCTTCTGCAGT-3′.

For CD69 analysis, cell sorter–purified B cells were cultured in a 96-well U-plate at 2 × 10⁵ cells in 150 µl medium per well, with or without stimulation. Stimulating reagents were C12-iE-DAP (InvivoGen) at 10 µg/ml, goat F(ab′)₂ anti-IgM (Jackson Laboratories) at 10 µg/ml, LPS (Sigma-Aldrich) at 20 µg/ml, CpG (InvivoGen) at 3 µg/ml, anti-CD40 (eBioscience) at 5 µg/ml, and muramyl dipeptide (InvivoGen) at 10 µg/ml. 6 h after stimulation, CD69 expression was analyzed using PE–anti-CD69 (eBioscience). Unseparated spleen cells were also stimulated using the same conditions, then costained with reagents to define various cell types together with anti-CD69. For survival analysis after culture, total cell number, viability, and cell size were measured by flow cytometry. BAFF (Alexis) was used at 0.2 µg/ml. For dye dilution cell division analyses, purified B cells were stained with CSFE (BioLegend), followed by stimulation in culture.

To determine cytoplasmic Nod1 levels, spleen cells were first surface stained to detect FO B cells, followed by cytoplasmic staining with anti-Nod1 (LSBio), or rabbit IgG (Biotech) as a control, at 1 µg/ml, detected by APC-goat anti-rabbit (Invitrogen). Relative medium stained levels in FO B cells were used for comparison.

For B cell subset purification for Western blot, FO B cells (CD21^medCD23⁺AA4⁻) and MZ B cells (CD21^hiCD23⁻AA4⁻) from spleen and B1a cells (B220^loCD5⁺) from the PerC were used. To purify μκTg (IgM^a)⁺ FO B and MZ B cells on a C.B17 mouse background, anti–IgM^b antibody was added to exclude endogenous B cells. B1a cells were sorted from both the spleen and PerC of V_H11t.CB17 mice (Wen et al., 2005b). B cell subsets purified by cell sorting (2 × 10⁶ cells per tube) were preincubated in RPMI 1640 plus 0.5% FCS medium in 0.5 ml for 6 h and then stimulated with C12-iE-DAP and/or anti-IgM. Cell lysates were subjected to 10% SDS-PAGE. The Cell signaling analysis reagents used were anti-IκBα, anti–P-IκBα (Ser32), anti-P-p44/p42 MAPK (Erk1/2; Thr202/Tyr204), anti-P-p38 MAPK (Thr180/Tyr182), anti–P-SAPK/JNK (Thr183/Tyr185), anti–P-Akt(Ser473 and Thr308), anti–P-FoxO1 (Ser256), and anti–P-GSK3β (Ser9); protein was detected using HRP anti–rabbit IgG antibody (all from Cell Signaling Technology). The inhibitor LY294002 (Cell Signaling Technology) was used. Anti–β actin (Bethyl Lab), anti-Nod1 (LSBio), and anti-Ptpn22 (Proteintech) Western blots were done without preincubation.

To produce the Nod1 pMIG-GFP construct, the mouse full-length Nod1 cDNA in the pLenti-C-mGFP lentiviral vector was obtained from OriGene (catalog no. MR211295L2). Nod1 cDNA was subcloned into the pMIG-EGFP retroviral vector by cutting the OriGene lentiviral vector with BamHI and MluI to release the Nod1 insert. The original clone had Nod1 fused to GFP, so a stop fragment containing MluI and XhoI restriction sites (synthesized by Integrated DNA Technologies; 5′-CGCGTTAAGCGGCCGCTCTCAAC-3′ and 5′-TCGAGTTGAGAGCGGCCGCTTAA-3′; both 5′-phosphorylated and duplexed) was ligated to the end of the Nod1 fragment and into the pMIG retroviral backbone, previously cut with BamHI and XhoI. The resulting retroviral construct was verified by sequencing and then used to generate retroviral supernatant by calcium phosphate–mediated transfection of Phoenix-E cells. Nod1 overexpression was verified in transduced pro–B cells by sorting GFP⁺ cells and determining mRNA levels using the TaqMan quantitative PCR assay. For production of shNod1-GFP, the retroviral vector LMP from Open Biosystems was used to express the shRNAmir construct from RNA polymerase II promoters. Hairpin sequences specific for Nod1 were from the Broad Institute public TRC portal. Oligos for hairpins were synthesized and cloned into LMP following a standard protocol, selected for appropriate size, and then verified by sequence analysis.

For cell transfer, B-lineage progenitors, including hematopoietic stem cells (through “immature” B cells in the case of μκTg or V_H3609t × V_k21-5Tg.Thy^KO BM transfer) were purified from BM, followed by overnight culture in medium containing OP9 stromal cell line with IL-7, and then transduced with retroviral supernatant. Medium was replaced after 2–3 h, and cells were cultured for an additional 24 h; GFP levels were checked (40–60% GFP⁺ in general), and then 1–3 × 10⁵ cells per recipient were injected i.v. into recipient mice. Recipients: 6.5 Gy irradiated C.B17.Thy^KO mice for ATAμκTg.Thy^KO BM transfer, 2 Gy irradiated C.B17 Rag1^KOThy^KO mice (Wen et al., 2005a) for V_H3609t ×V_k21-5Tg.Thy^KO BM transfer, and 2 Gy irradiated C.B17.scid mice for AGcAμκ Tg BM transfer. Initial PBL analysis was done 3 wk after transfer to check GFP levels.

Stem/progenitor cell–enriched fractions (CD19⁻ IgM⁻ CD5⁻ and low side scatter) were sorted from BM of B6.Ly5.1 (WT) and either Nod1/2^−/− or Ripk2^−/− mice and cotransferred to B6.scid mice (3 Gy, 1 d before) by i.v. injection (5 × 10⁵ cells each; total of 10⁶ cells per recipient). To monitor B cell origin in PBLs, a six-color antibody combination (CD45.1, CD45.2, CD19, IgM, IgD, and AA4.1) was used. No CD19⁺IgM⁺ B cells were detected in B6.scid recipient alone without BM transfer. After transfer, the WT/knockout mouse B cell ratio levels in spleen were based on the WT/knockout mouse B cell ratio level data of T1 (CD19⁺IgM^hiAA4⁺CD21⁻CD23⁻) ratio as 1.0. For B cell cotransfer (WT with Ripk2^−/−), B220⁺CD5⁻ cells were purified from spleen (5 × 10⁶ cells per recipient). For cotransfer of WT (B6) BM with AGcAμκTg (Ripk2⁺ or Ripk2⁻) BM, BM cells were used at a 3:1 ratio (4.5 × 10⁵ from WT and 1.5 × 10⁵ from AGcAμκTg per recipient). AGcA B cell generation was monitored by V_H3609id and 13H8k (recognizing V_k19-17 L chain) costaining to detect AGcA B cells (Ichikawa et al., 2015) and anti-V_H3609id together with kappa and CD19 to define total B cells. As a control, Ripk2^−/−AGcAμκTg mouse BM cells were injected at 5 × 10⁵ cells per recipient without cotransfer.

Original raw microarray data are available under GEO accession no. GSE101318.

Fig. S1 shows V_H3609t mouse analysis. Fig. S2 presents V_H3609t × V_k21Tg.Thy^WT mouse data showing strong BCR editing under Thy^WT and also with Nod1 up-regulation. Fig. S3 demonstrates that the Igκ-edited ATAid^lo cells coexpress with original ATA BCR. Fig. S4 is the time-course Western blot analysis, comparing different FO B cells. Fig. S5 shows predominant Nod1 expression over Nod2 in FO B and MZ B cells and Ripk2 deficiency effect in CiE-DAP stimulated auto⁺ AGcAμκTg mouse B cells. Table S1 lists endogenous Igκ gene usage in edited ATAid^lo FO B cells. Table S2 presents a comparison of genes between B cells with unedited and edited BCRs in Excel format.

We thank Li-Jun Wen for generation of V_H3609t knock-in mice, Alex Kowalczyk for assistance with serum ATA analysis, and several Fox Chase Cancer Center Facilities (Lab Animal, Flow Cytometry, Transgenic, and DNA Sequencing) for technical support. We also thank Renée Tsolis for providing Nod1 and Nod2 double-knockout mice and D. Kappes and K. Campbell for comments on the manuscript.

This work was supported by the National Institutes of Health (grant RO1 AI49335 to K. Hayakawa, grant AI026782 to R.R. Hardy, grant AI113320 to R.R. Hardy and K. Hayakawa, and grant DK61707 to G. Núñez) and the Fox Chase Cancer Center Blood Cell Development and Cancer Keystone program.

The authors declare no competing financial interests.

Author contributions: K. Hayakawa designed the study, performed the experiments, and wrote the manuscript. R.R. Hardy supervised mouse model generation, cloning and transduction of Nod- and shNod1-GFP BM for cell transfer, discussed the work with K. Hayakawa, and contributed to writing the manuscript. A.M. Formica performed BCR gene sequence, quantitative PCR, and Western blot analysis. Y. Zhou, supervised by Y.-S. Li, performed the microarray analysis that led to the discovery of the significance of Nod1. D. Ichikawa and M. Asano contributed original findings of AGcA B generation and determination of its specificity. J. Brill-Dashoff contributed technical help with mouse model generation, breeding, and screening. S.A. Shinton generated retroviral supernatant, transduced B cell precursors, and performed cell transfers. G. Núñez provided Nod1 knockout mice and discussed the work.