B-cell intrinsic and extrinsic signals that regulate central tolerance of mouse and human B cells

Summary

The random recombination of immunoglobulin V(D)J gene segments produces unique IgM antibodies that serve as the antigen receptor for each developing B cell. Hence, the newly formed B cell repertoire is comprised of a variety of specificities that display a range of reactivity with self-antigens. Newly generated IgM⁺ immature B cells that are non-autoreactive or that bind self-antigen with low avidity are licensed to leave the bone marrow with their intact antigen receptor and to travel via the blood to the peripheral lymphoid tissue for further selection and maturation. In contrast, clones with medium to high avidity for self-antigen remain within the marrow and undergo central tolerance, a process that revises their antigen receptor or eliminates the autoreactive B cell altogether. Thus, central B cell tolerance is critical for reducing the autoreactive capacity and avidity for self-antigen of our circulating B cell repertoire. Bone marrow cultures and mouse models have been instrumental for understanding the mechanisms that regulate the selection of bone marrow B cells. Here we review recent studies that have shed new light on the contribution of the ERK, PI3K, and CXCR4 signaling pathways in the selection of mouse and human immature B cells that either bind or do not bind self-antigen.

Introduction

The immune system has evolved complex and sophisticated cell biological processes in order to achieve a lymphocyte repertoire able to mount a broadly diverse response in the face of an extraordinarily varied and dynamic microbial world. The acquisition of the Recombination Activating Genes RAG1 and RAG2 in jawed fish, and activation-induced cytidine deaminase (AID, gene AICDA) even earlier in phylogeny, facilitated the development and evolution in higher vertebrates of an adaptive immune system that generates a remarkable diversity of B cell receptors (BCRs) and serum antibodies by B cells, and T cell receptors (TCRs) by T cells^1–4. Despite using the same DNA-recombining and mutating enzymes, higher vertebrates of different taxa and species have evolved slightly different processes that assemble antibody gene fragments encoding BCRs in developing B cells. Fortuitously, humans and mice employ a very similar process whereby B cell progenitors that develop within the bone marrow undergo RAG1/2-mediated rearrangement, initially at the immunoglobulin (Ig) heavy (H) chain locus (Igh) where a D gene segment is juxtaposed to a J gene segment and, subsequently, by the rearrangement of a V gene segment to the previously rearranged DJ immunoglobulin (Ig) gene segment. This is then followed by ordered rearrangements of V to J gene segments at the light (L) chain Igk and Igλ independent loci. This ordered V(D)J recombination process leads to the development of immature B cells (referred to as fraction E according to Hardy’s nomenclature⁵) that express BCRs (harboring an IgM) for the first-time during B cell poiesis and whereby about 95% of all cells bear a unique pair of H and L chains⁶. The genomes of mice and humans, moreover, possess roughly similar numbers of Ig V, D, and J gene segments⁷, suggesting they can achieve a similar combinatorial diversity of their antibody repertoire. While this diversity may be necessary for counteracting the variety of potentially pathogenic threats, this stochastic process also has the possible drawback of generating V(D)J sequence combinations encoding antibodies that bind self-antigens⁸ and that may, therefore, lead to immune responses against self and to autoimmunity.

Supporting the idea that a B cell repertoire that is too frequently autoreactive poses a danger to the individual or even an entire species, the immune system has evolved an efficient process of B cell tolerance that, via different checkpoints through B cell development, greatly reduces the overall autoreactive capacity of the circulating mature B cell repertoire^9–11. While it is unclear whether the B cell tolerance process exerts its function with comparable mechanisms and efficiencies and at analogous B cell developmental checkpoints in all higher vertebrates, or even in all mammals, the similarities in B cell development and antibody generation in mice (Mus musculus) and humans (Homo sapiens) suggest these species also share a similar process of B cell tolerance. Indeed, studies have shown that B cell tolerance in mice and humans occurs at similar checkpoints and with similar efficiencies^12,13.

The first checkpoint of B cell tolerance in adult mice and humans operates at the immature B cell stage (i.e., fraction E) in the bone marrow tissue when the BCR is expressed for the first time during B cell poiesis and has, thus, been named central B cell tolerance. At this checkpoint the newly generated BCR of immature B cells is tested for reactivity against local self-antigens. Immature B cells expressing BCRs with significant autoreactivity are precluded from continuing in development and undergo tolerance induction. About half of the autoreactive immature B cells in mice and humans (30–35% of all immature B cells) exhibit a sufficiently high level of autoreactivity to trigger central tolerance^10,14. The remaining immature B cells are licensed to exit the bone marrow and to further differentiate, and these include cells (about 35–40% of transitional B cells) that measurably bind self-antigen but with an avidity below the threshold for inducing central tolerance. Most of the low avidity autoreactive B cells that are permitted to enter the periphery are quickly subjected to a process of peripheral immunological tolerance that induces a state of anergy, prevents their differentiation into the mature cell stage, and drastically reduces their survival^13,15.

This review focuses on central B cell tolerance and on recent studies from our group and others that have shed new light on the molecular pathways that regulate this tolerance checkpoint in both mouse and human B cells, and on defects that lead high avidity newly generated autoreactive B cells to enter the circulating B cell repertoire.

Experimental models to investigate central B cell tolerance in mouse and human B cells

Central tolerance takes place in the bone marrow tissue after which newly-generated B cells emigrate to the periphery. Thus, it is possible to gauge the efficiency of central B cell tolerance by comparing the autoreactive capacity (i.e., the frequency of autoreactive B cells) of recent B cell emigrants to that of the fraction E IgM⁺IgD^– immature B cells within the bone marrow. Such analyses, which were first performed in wild-type mice⁹ and subsequently, and in more granular form, in humans¹⁰, have demonstrated central B cell tolerance has similar efficiencies in both species. Before the development of transgenic animals, early experimental models that aimed to investigate tolerance injected neonates with anti-Ig antibodies to engage the BCRs of developing B cells. These studies demonstrated that B cell tolerance could be achieved systemically by triggering the BCR of all or a fraction of B cells (reviewed in¹¹).

Conventional Ig H and L transgenic mice developed in the late 1980s and 1990s allowed central B cell tolerance to be investigated more specifically by generating mice in which all B cells express BCRs of defined antigen reactivity. This afforded the opportunity to evaluate differences in B cells that developed in either the presence or absence of their specific antigen and with antigens that varied in form, affinity, and avidity. These early important mouse models clearly demonstrated that central B cell tolerance operates via clonal deletion, which is the death of autoreactive immature B cells within the bone marrow tissue^16–20. They also highlighted the following characteristics of central B cell tolerance: i) self-antigen must be expressed by bone marrow cells ubiquitously and abundantly to trigger central tolerance; ii) immature B cells engaging self-antigen internalize their BCRs to a degree that correlates with the avidity to self-antigen; iii) cell death is not immediate, taking approximately 3 days during which autoreactive cells are arrested in development; iv) dying autoreactive cells are rapidly engulfed and cleared by macrophages and are difficult to identify even when all immature B cells are autoreactive and undergo clonal deletion.

In the last two decades, further in depth analyses of conventional Ig H (+L) transgenic mice as well as genetically-engineered Ig knock-in mice harboring rearranged VDJ and VJ gene segments at the endogenous Ig H and L loci, demonstrated that autoreactive B cells undergoing central tolerance do not necessarily undergo clonal deletion, but instead predominantly rely on an alternative mechanism of tolerance dubbed receptor editing (reviewed in^21,22). With this process, autoreactive B cells perform secondary V-to-J rearrangement across the rearranged VJ gene at the Igk or Igλ L chain loci. This new V to J recombination event displaces the VJ sequence encoding the original autoreactive L chain with a new VJ gene segment encoding a novel L chain. Pairing of the new L chain with the existing H chain forms a new BCR that is again tested for self-reactivity. The immature B cell ceases receptor editing when the new BCR is non-autoreactive or much less autoreactive²³. These studies further demonstrated that receptor editing is a very efficient process able to successfully edit the BCR of each immature B cell in a mouse model where every developing B cell is autoreactive, and thus rescuing the cells from clonal deletion. Notably, receptor editing operates efficiently whether autoreactive B cells represent 100% of all immature B cells, or when they are only a minority population developing in competition with non-autoreactive B cells^19,24. Furthermore, by analyzing the peripheral repertoire of hemizygous Igk mice it has been possible to demonstrate that about 30% of the peripheral mouse mature B cell population is composed of clones that were processed by receptor editing during their bone marrow development¹⁴. Overall, these studies have shown that clonal deletion is generally only activated in immature B cells that fail to significantly reduce their autoreactivity via receptor editing, a phenomenon prevalent in B cells with BCRs in which the H chain dominates the binding to self-antigen irrespective of the L chain, and in cells of Ig conventional transgenic mice in which random genomic integration of the transgene is not compatible with secondary V to J recombination, thus precluding receptor editing. Skewing central B cell tolerance toward clonal deletion can also be achieved in Ig knock-in mice by reducing the chance of secondary Ig gene recombination (e.g., by deleting the J elements 3’ of the rearranged VJ gene segment) or by deleting one of the RAG genes, which eliminates receptor editing altogether^24,25.

Subsequent to the generation of Ig knock-in mouse models, additional models of central B cell tolerance were constructed to study this process in B cells expressing a polyclonal wild-type repertoire. Specifically, mice were generated in which all tissues express transgenes encoding B cell ‘macroself’ superantigens: surface proteins displaying the variable regions and specificity of anti-mouse IgM or Igκ antibodies. With these transgenic mice it became possible to establish that even for mouse B cells expressing a wild-type polyclonal repertoire, central tolerance operates via receptor editing first, with clonal deletion serving as a reserve mechanism preventing the emigration of autoreactive B cells^26,27.

Ig transgenic mice, Ig knock-in mice, and macroself transgenic mice have been utilized over the last 30 years to understand the mechanism(s) by which BCR engagement of bone marrow immature B cells, or lack thereof, are translated into two distinct processes of tolerance induction and continued differentiation. More specifically, tolerance induction encompasses arrest in differentiation, retention in bone marrow tissue, continual Ig V(D)J recombination, and possible cell death, whereas continued differentiation entails stopping V(D)J gene recombination, maintaining cell survival, and exiting the bone marrow (Fig. 1).

Figure 1: Requisite biological activities for the selection of immature B cells out of the bone marrow.

The activities listed in this box are indispensable for the selection into the periphery of newly generated B cells that retain their selected antibody and have a chance to survival and participation in immune responses.

What triggers central B cell tolerance?

Results from multiple studies indicate that the BCR avidity for self-antigen, and not affinity per se, triggers induction of central B cell tolerance in immature B cells. Two Ig H+L transgenic and knock-in mouse models that demonstrate this tenet particularly well are the anti-hen egg lysozyme (HEL) mice and the anti-MHC class I (MHC-I) 3–83 mice. In the anti-HEL mice, B cells express a BCR and produce antibodies with a high affinity (~10⁹ M^–1) for HEL²⁸. All anti-HEL B cells that develop in the presence of the high avidity membrane form of HEL undergo central tolerance, via clonal deletion in conventional Ig transgenic mice¹⁷ and via receptor editing in knock-in mice²⁹. In contrast, most anti-HEL B cells bypass central tolerance and are licensed to leave the bone marrow when they develop in the presence of the low avidity soluble form of HEL^28,29, and despite possessing the same high affinity BCR. In the 3–83 mouse model, on the other hand, anti-MHC-I 3–83 B cells produce antibodies with low affinity for diverse MHC class I haplotypes, ranging from 10⁴ to 10⁶ M^–1 for H-2 K^bm3, K^b, and K^k MHC-I proteins³⁰. Expression of MHC-I antigens, however, is abundant on the surface of most cells, thus conferring high avidity to the interaction with anti-MHC-I reactive B cells. Indeed, all 3–83 developing B cells undergo central tolerance similarly (generally, clonal deletion in conventional transgenic and receptor editing in knock-in models;^16,24,31,32) whether they develop in the presence of K^bm3, K^b, or K^{k 30}. Moreover, 3–83Ig-expressing B cells are partially licensed to enter the periphery only upon lowering their avidity for self-antigen, achieved by artificially decreasing the concentration of self-antigen³³ or by introducing the co-expression of non-autoreactive BCRs that dilute and ultimately reduce the number of autoreactive receptors on the cell membrane³⁴. Reducing the concentration of a membrane-bound self-antigen prevents central tolerance also in wild-type B cells, as shown using low-expressing anti-Igκ macroself transgenic mice³⁵. Together, these data support a model in which a degree of BCR avidity for the self-antigen above a certain threshold triggers central tolerance (Fig. 2A).

Figure 2: Phenotype of mouse and human immature B cells in relation to self-antigen binding.

A) Schematic representation of BCR surface expression and internalization on immature B cells responding to self-antigen with various degree of avidity. B) Markers that distinguish mouse and human immature B cells that do not engage (or minimally engage) self-antigen from those that engage self-antigen with high avidity. Markers listed in the middle are expressed by all (mouse or human) immature B cells independent of self-reactivity. Markers listed on the left or right of the scheme are those that, by their lower or higher degree of expression, distinguish non-autoreactive from high-avidity autoreactive cells.

At present, we do not have a clear understanding of the precise threshold or range of avidity that triggers central tolerance. Avidity is difficult to measure but can be approximated by the extent to which the BCRs of autoreactive B cells are internalized. Based on this, it appears that the avidity that triggers central tolerance must be sufficient to induce internalization of >50% surface BCRs^34,36. In fact, integrating data from multiple studies suggests that whether an immature B cell undergoes central tolerance depends on the amount of BCRs (i.e., IgM molecules) that remains on the cell surface and the ability of these BCRs to transduce tonic signals^36–39. Importantly, the number of BCRs remaining on the cell surface after engagement with self-antigen will also be affected by the amount of total IgM expressed by the B cell, which is typically artificially higher in conventional Ig transgenic B cells owing to multiple transgene copies, compared with wild-type or knock-in B cells that bear one or maximum two copies of the rearranged Ig genes⁴⁰. In some developing B cells, endosomal Toll-like receptor (TLR) signaling by TLR7 and TLR9 via MyD88 synergizes with BCR signaling to reduce the avidity threshold of autoreactive B cells undergoing central tolerance and via a process in which the enzymes RAG1/2 and AID promote receptor editing and/or cell death (reviewed in⁴¹). The MyD88/AID pathway is relevant for reducing the peripheral entry of newly generated B cells reacting with nuclear self-antigens^42,43.

Phenotype of mouse and human B cells undergoing central selection

Developing autoreactive B cells encounter self-antigen for the first time upon transitioning from the pre-B cell stage following the productive VJ rearrangement of an Ig L chain gene and the expression of an (H+L) IgM BCR. Because B cells undergoing central tolerance often downregulate most of the surface BCR, these immature B cells appear phenotypically as surface IgM^– and can be mistaken for pre-B cells. What distinguishes surface IgM^– autoreactive immature B cells from pro-B and pre-B cells in mice is the presence of intracellular (IC) L chain, lower expression of the pro-B cell marker CD43, absence of the transitional and mature B cell markers IgD, CD21, and CD23, and the presence (or high expression) of the pre-B and immature B cell markers CD2, CD24 and CD93 (Fig. 2B). Immature B cells that do not bind self-antigen (i.e., non-autoreactive cells) display the same general phenotype of autoreactive B cells but are clearly surface IgM positive. After autoreactive immature B cells successfully edit their autoreactive BCR, they become surface IgM⁺ immature B cells indistinguishable from non-autoreactive B cells. In a polyclonal B cell population, however, IgM⁺ immature B cells display a range of surface IgM that inversely correlates with self-reactivity^44,45. B cell development is slightly altered in Ig H+L chain transgenic or knock-in mice because B cells already bear pre-rearranged VDJ and VJ gene segments without the need of Ig gene recombination events. In these genetically modified mice, therefore, B cell development proceeds from the pro-B cell stage, at which the Igh locus is first transcribed, to the immature B cell stage, through a brief large pre-B cell stage during which the Igk locus is opened for transcription⁴⁶. The normally substantial small pre-B cell population observed in wild-type mice is usually absent in mice whose B cells express pre-rearranged IgL chains.

The cell markers described above have been used to identify and further compare the phenotype of de facto autoreactive and non-autoreactive immature B cells, as well as non-autoreactive B cells that emerge after successful receptor editing (edited B cells). These studies have demonstrated that an additional marker that clearly differentiates mouse autoreactive immature B cells undergoing central tolerance is the reduced expression of CD19⁴⁷, a positive coreceptor for BCR signaling. Indeed, anti-MHC-I 3–83 immature B cells exhibit significantly lower surface CD19 levels when they develop in the presence of the self-antigens K^k or K^b relative to when they develop in H-2^d mice^40,47. Other differences that discriminate 3–83 autoreactive immature B cells from non-autoreactive B cells are lower amounts of BAFFR and slightly higher levels of IL-7R^48,49 (Fig. 2B). Differences in CD19, BAFFR and IL-7R have also been observed in wild-type immature B cell populations between cells that bind or do not bind self-antigen^27,44. Moreover, these differences in surface receptor expression have biological consequences, such that BAFF increases the survival of non-autoreactive B cells over autoreactive cells, while IL-7 promotes better survival and expansion of autoreactive B cells^48–50.

Our overall understanding of central B cell tolerance mechanisms is predominantly based on investigation of autoreactive immature B cells in a variety of mouse models. Mechanisms of central B cell tolerance in humans were ill-defined until technological advances in single-cell BCR cloning facilitated the ability to evaluate antibody specificities from BCRs derived from single developing immature B cells¹⁰. These challenging studies have not only revealed that central tolerance is induced in at least 30% of all newly generated immature B cells in healthy individuals¹⁰ (similar to mice), but they have also identified genes that contribute to this efficiency⁵¹. However, these studies have not provided insights into the phenotype of autoreactive immature B cells undergoing tolerance, which may prove useful in identifying individuals at-risk for developing autoimmune responses. To address this gap in knowledge, a few years ago our group developed a human immune system humanized mouse model (HIS hu-mouse) to study central tolerance in human B cells⁵². HIS hu-mice are immunodeficient mice transplanted with human hematopoietic stem cells that facilitate the development of functional human B cells and T cells in the mouse host^53,54. Our hu-mouse model of B cell tolerance, developed based on Igκ-specific macroself transgenic mice²⁶, relies on a ubiquitously expressed membrane super-antigen (Hcκ; anti-human Igκ) that binds the human Igκ light chain constant region. In Hcκ hu-mice, all developing human κ⁺ B cells (about half of all B cells) undergo central tolerance to the Hcκ self-antigen (mainly in the form of receptor editing) at the immature cell stage in the host bone marrow tissue⁵². Using these hu-mice we have been able to directly compare the phenotype of de facto human autoreactive immature B cells (κ⁺ cells in Hcκ hu-mice) to that of non-autoreactive cells, λ⁺ in the same animals or κ⁺ in non-transgenic (Hcκ⁻) hu-mice. These studies have shown that human immature B cells developing in the presence of self-antigen possess intracellular Ig L chains, but express low levels of surface IgM, CD19, and BAFFR relative to non-autoreactive cells and, thus, exhibit a phenotype similar to mouse autoreactive B cells^52,55 (Fig. 2B). More recent studies have highlighted additional markers that discriminate human autoreactive B cells from non-autoreactive cells and include a lower expression of the tetraspanin CD81, a molecule known to associate with CD19, and higher expression (about five-fold on average) of the CD69 activation receptor⁵⁵ (Fig. 2B). Interestingly, differences in CD81 and CD69 appear specific to human B cells as these proteins are expressed minimally, if at all, by mouse immature B cells in vivo⁵⁵. Furthermore, immature human B cells with a phenotype consistent with autoreactivity (positive for intracellular IgL chain and CD19^lowIgM^lowCD81^lowCD69^high) were also observed in pediatric bone marrow biopsies, which represent physiological tissue conditions, at around 10–20% of all immature B cells⁵⁵. We propose that the differential expression of the markers highlighted here can help identify autoreactive and non-autoreactive B cells within a wild-type mouse or human immature B cell populations.

Signaling pathways differentially activated during central B cell selection

BCR engagement results in the activation of an intracellular BCR signaling cascade. Initial consideration of tolerance induction by immature B cells led investigators to postulate that central B cell tolerance, including arrest in cell differentiation, receptor editing, and possibly cell death, was mediated by self-antigen induced BCR signaling. In contrast, receptor editing has been shown to correlate best with a reduced level of a signal referred to tonic BCR signaling: low basal level of BCR signaling mediated by surface BCR expression in the absence of antigen⁵⁶. Indeed, even the inducible deletion of the BCR in non-autoreactive immature B cells triggers RAG1/2 expression and Ig gene recombination at the L chain loci⁵⁷. Moreover, when expression of RAG1/2 has been reportedly induced following BCR stimulation, the kinetics of RAG expression correlate more closely with BCR internalization rather than rapid activation of the BCR signaling cascade^39,57. These findings suggested to us that modulating the activity of tonic BCR signaling mediators could alter the selection of immature B cells, potentially breaking central tolerance and causing the entry and incorporation of high avidity autoreactive B cells into the primary B cell repertoire. The remainder of this review focuses on summarizing and discussing the results of these studies. It is nevertheless acknowledged that other molecular pathways not discussed herein (e.g., PTPN22, MYD88, AID, GADD45α) have also been shown to profoundly impact central B cell tolerance^41,51,58.

Initial studies aimed to uncover the signaling molecules mediating tonic BCR signaling in non-autoreactive immature B cells were focused on understanding the mechanisms by which this signal suppresses receptor editing and/or promote cell differentiation. These studies were performed for the most part in vitro, using IL-7-spiked cultures of bone marrow cells from Ig transgenic or knock-in mice, a method that enriches for immature B cells²⁰. Adding pharmacologic inhibitors or dominant active signaling molecules to these cell cultures were used to identify mediators of tonic BCR signaling that blocked RAG1/2 expression and/or induced cell differentiation. Results from these studies highlighted three well-known BCR signaling pathways, the ERK MAP kinase, the RAS GTPase (a well-known activator of ERK), and the PI3K lipid kinase^{36,40,47,57,59}. These pathways were found to be modestly (1.5–2-fold) but significantly more activated in non-autoreactive immature B cells relative to autoreactive cells^40,47, a result initially considered puzzling. Indeed, how can non-autoreactive B cells exhibit higher activity of BCR signaling mediators than autoreactive cells that have just engaged antigen? The most likely explanation for this counterintuitive observation is that an acute BCR signaling cascade is strongly activated within minutes after antigen binding, but also subsides within a couple of hours as a result of feedback mechanisms that lead to RAS, ERK, and PI3K activity to return to baseline levels^40,60. Tonic BCR signaling sustains a low basal activity of these effector molecules, but also depends on the level of BCR on the cell surface³⁶, which is extremely low on high avidity autoreactive B cells after antigen binding. Thus, self-antigen recognition by bone marrow immature B cells leads to BCR endocytosis and a concomitant reduction in tonic BCR signaling. The IL-7 bone marrow in vitro culture proved to be an excellent model for screening and uncovering signaling pathways potentially implicated in central B cell selection. Follow-up in vivo studies were developed to establish whether RAS, ERK, and PI3K pathways were indeed necessary and/or sufficient for breaking central tolerance and to mediate the selection of newly generated B cells into the periphery.

The ERK pathway: contributing but not necessary?

The serine/threonine MAP kinase ERK (comprising the two isoforms ERK1 and ERK2, or ERK1/2) translates extracellular cues sensed by a variety of cell surface receptors to control fundamental cellular processes⁶¹. These kinases play a crucial role in translating BCR signals that are necessary for the development, activation, and differentiation of B cells⁶². The main means by which ERK1/2 are activated is from upstream RAS GTPases through RAF and MEK kinases. Like other BCR signaling mediators, ERK1/2 are rapidly phosphorylated within minutes after BCR stimulation, and then are dephosphorylated to baseline within a couple of hours⁶⁰. However, continual BCR stimulation, as occurs with autoreactive anergic peripheral B cells in the absence of costimulatory signals, maintains elevated phospho-ERK1/2 levels⁶³ that function to promote apoptosis⁶⁴ or to inhibit differentiation into plasma cells⁶⁵, depending on the context. Like mature B cells, immature B cells respond to BCR stimulation with a rapid and strong phosphorylation of ERK1/2, which subsides within two hours⁴⁰. However, in contrast to that observed in peripheral anergic B cells, autoreactive immature B cells have lower (not higher) phospho-ERK1/2 when compared to non-autoreactive B cells⁴⁰. Moreover, the elevated activity of ERK1/2 in non-autoreactive immature B cells extends to the entire RAS/RAF/MEK/ERK pathway (⁴⁰ and data now shown), is dependent on the amount of surface BCR and the activity of Src family kinases, and is independent from the activity of other common receptors such as type I and II IFN receptors and TLRs^36,40.

In our studies, bone marrow cultures of non-autoreactive immature B cells treated with small molecule inhibitors of ERK1/2 or MEK1/2 were severely impaired in differentiation into transitional B cells, manifesting with reduced upregulation of CD21 and CD23^36,40. This suggested the ERK pathway was important for the differentiation of immature B cells into transitional B cells, a hypothesis in line with the known contribution of ERK1/2 to early B cell development⁶². Indeed, we showed that the expression of the constitutively active protein N-RAS-D12 in 3–83 autoreactive immature B cell cultures increases phospho-ERK1/2 to levels seen in non-autoreactive B cells while concomitantly reducing receptor editing and improving the differentiation of transitional B cells⁴⁰. Moreover, when tested in vivo, N-RAS-D12 expression was similarly able to decrease receptor editing and to increase the frequency of autoreactive B cells in peripheral tissue, although it did not induce appropriate differentiation⁴⁰. This led to our initial conclusion that activation of ERK (via RAS/RAF/MEK) was able of, at least partially, breaking central B cell tolerance.

This hypothesis was further tested in vivo by using a Cre-regulated Rosa26 allele harboring a constitutively active MEK transgene (MEK1 S218D/S222D, or MEK1DD;⁶⁶). Surprisingly, although MEK1DD was able to elevate phospho-ERK1/2 levels in autoreactive immature B cells in vivo to the levels observed in non-autoreactive B cells, it failed to alter central B cell tolerance⁶⁷. We believe the reason for these discrepant results was twofold: 1) experiments using ERK1/2 pharmacologic inhibitors in vitro and those performed with the MEK1DD transgene in vivo showed that active ERK1/2 was unable to suppress RAG1/2 expression and, therefore, could not block receptor editing^40,67; 2) constitutively active MEK was also unable to induce the expression of BAFFR⁶⁷, a receptor that promotes the differentiation of immature B cells into transitional and mature B cells as well as cell survival⁴⁸.

In summary, while ERK1/2 kinases contribute, at least in vitro, to the differentiation of non-autoreactive immature B cells into transitional B cells, they are unable to override central tolerance when activated in autoreactive B cells in vivo.

The PI3K pathway: a major player in central B cell selection

PI3Ks are lipid kinases that phosphorylate position 3 of the inositol ring of phosphoinositides located on the cytosolic side of the plasma membrane (reviewed in⁶⁸). There are multiple isoforms of PI3Ks, and the most relevant to B cells are class IA PI3Kα and PI3Kδ. These PI3Ks are heterodimers composed of a catalytic subunit (P110α and P110δ) and a SH2-containing regulatory subunit (p85α), which directs the catalytic subunit to specific phosphorylated substrates. PI3Kα and PI3Kδ phosphorylate phosphatidylinositol-(4,5)-diphosphate (PI(4,5)P₂) generating phosphatidylinositol-(3,4,5)-triphosphate (PI(3,4,5)P₃, or PIP3), a lipid second messenger (PIP3) that coordinates multiple downstream signaling pathways that control a variety of essential cellular processes, such as cell survival, proliferation, differentiation, and metabolic fitness. Although PI3Kα and PI3Kδ can play a redundant role in B cells, PI3Kδ appears to be more critical as indicated by the fact that deletion of PI3Kδ results in a phenotype similar to p85α deletion, although depending on the differentiation stage of B cells and the specific process under examination^68,69. The activity of PI3K enzymes is mainly regulated by two lipid phosphatases, Phosphatase and Tensin Homolog deleted on Chromosome 10 (PTEN) and SH2-containing inositol polyphosphate 5-phosphatase (SHIP), which dephosphorylate PI(3,4,5)P₃ into PI(4,5)P₂ or PI(3,4)P₂, respectively. Indeed, deletion of either PTEN or SHIP-1 in B cells leads to activation of PI3K and dysregulated B cell development, maturation, and peripheral tolerance (reviewed in⁷⁰).

One of the critical functions mediated by the PI3K pathway in developing B cells is the suppression of RAG1 and RAG2 expression and, consequently, the inhibition of V(D)J recombination at the IgH and IgL chain loci at defined development checkpoints. This critical function of PI3K is carried out via AKT-mediated phosphorylation of FOXO1, a major transcription factor required for the expression of RAG1/2 genes^59,71,72. Phosphorylated FOXO1 is then excluded from the nucleus and targeted for degradation. Activation of PI3K can be mediated independently or synergistically by the pre-BCR, the IL-7R, the BCR, and additional specific coreceptors⁶⁹. Foundational studies have shown that in the absence of BCR engagement, immature B cells exhibit higher PI3K-AKT activity compared to the activity observed when the BCR is engaged for extended time^47,73. Further, inhibition or deletion of PI3K promotes RAG1/2 expression and subsequent V-J recombination of IgL chain genes even in the absence of BCR stimulation^47,57,74. Thus, diminished PI3K activity, and not (chronic) BCR engagement, induces IgL chain rearrangement. Moreover, the basal level of PI3K activity found in immature B cells depends on the expression of both the BCR and CD19, because deletion of either molecule^39,50 leads to RAG1/2 upregulation and receptor editing similar to that observed with loss of PI3K. This is particularly relevant when considering that autoreactive immature B cells greatly downmodulate both BCR and CD19 from the cell surface⁴⁷. Interestingly, PI3K does not appear to suppress RAG1/2 expression in thymocytes; in these cells, tonic TCR signaling suppresses RAG via the ERK pathway⁷⁵.

While investigating the mechanisms by which active RAS inhibits tolerance in bone marrow cultures of 3–83Igi,H-2^b autoreactive B cells, we observed that the pharmacologic inhibition of PI3K was able to diminish the ability of N-RAS-D12 to promote immature B cell differentiation into transitional B cells and to restrain receptor editing⁴⁰. The ability of PI3K to promote early pro-B to pre-B cell development and to inhibit RAG1/2-mediated V(D)J recombination had previously been demonstrated^47,74,76, however we were surprised to find an additional role for PI3K in advancing the differentiation of immature B cells. To investigate the extent and breadth to which PI3K inhibits central tolerance in vivo, we generated mice in which 3–83 IgH+L B cells developed in the presence or absence of self-antigen while expressing P110*, a constitutively active recombinant form of PI3Kα⁶⁶. A constitutively active PI3K expressed in developing B cells would be expected to strongly inhibit RAG1/2 expression thus preventing the development of most B cells due to an inability to undergo V(D)J rearrangement. However, in the 3–83 H+L mouse model all B cells carry already rearranged VDJ and VJ gene segments and therefore develop without the need of primary V(D)J recombination events, allowing to test the contribution of PI3K to central tolerance. Importantly, the expression of P110* in autoreactive 3–83 immature B cells (i.e., in H-2^b mice) increases phospho-AKT to levels seen in non-autoreactive cells⁷³. This allowed us to compare the extent of central B cell tolerance in the presence or absence of basal PI3K-AKT activity. We have recently established that P110* does not lead to excessive phospho-AKT, as might be expected, due to a feedback mechanism that increases the expression of the PTEN and SHIP-1 phosphatases (Fig. 3A), which can then reduce PIP3 amounts and the activity of the PI3K pathway downstream.

Figure 3: Expression of PTEN and SHIP-1 phosphatases in immature B cells.

A) Analysis of PTEN and SHIP in bone marrow immature B cells from 3–83Igi,H-2^b autoreactive (AUT) mice, expressing or not the active PI3Kα molecule P110*. Immature B cells were gated as B220⁺CD24^highCD21^– (and GFP⁺ for P110*-mb1Cre cells). Histogram plots show representative intracellular (IC) expression of PTEN and SHIP-1 in gated immature B cells. The bar graphs display mean and SD of PTEN and SHIP-1 geometric MFI form three mice per group. Symbols represent individual mice. B) Intracellular PTEN expression in 3–83 immature B cells (gated as B220⁺CD24^highIgD^–) from 3–83Igi,H-2^d (non-autoreactive, NA) and 3–83Igi,H-2^b (autoreactive, AUT) mice. Bar graph shows mean and SD from three mice per group. Symbols represent individual mice. Statistical analyses in all bar graphs were performed with a one-tailed Student’s t test. *p≤0.05; **p≤0.01.

Our analyses of central tolerance in 3–83Igi,P110* mice established that activation of PI3K prevents RAG1/2 expression and abrogates receptor editing even when B cells develop in the presence of a high avidity self-antigen in vivo. Strikingly, we did not observe the development of any receptor-edited B cells in 3–83Igi,H-2^b, P110* mice⁷³. While this result was somewhat predictable based on the known ability of PI3K to block FOXO1, we were surprised to find that P110* also promoted the efficient release of autoreactive B cells from the bone marrow, their relocation into the spleen, and their further differentiation into transitional and mature B cells, as demonstrated by the downregulation of the immature/transitional cell markers CD24 and CD93 in splenic B cells. The differentiation of P110* autoreactive immature B cells into mature cells was suboptimal, but this was similarly observed with the differentiation of P110* non-autoreactive B cells and, therefore, was unrelated to autoreactivity and tolerance. This defective maturation also appeared to be unrelated to BAFFR, which was actually upregulated by P110*⁷³.

Because P110* is a recombinant synthetic form of active PI3Kα⁷⁷, it could potentially introduce artifacts unrelated to the physiological processes under study. However, our findings and conclusions have been supported by an independent study⁷⁸ that showed loss of the lipid phosphatase PTEN similarly abolishes central tolerance in developing 3–83 B cells. Furthermore, genetic loss of Pten was also shown to break central tolerance in wild-type B cells developing in the presence of the anti-IgM^b macroself superantigen⁷⁹. These data overall support the conclusion that depression of the PI3K pathway maintains central B cell tolerance while its activation promotes positive selection of B cells even in the presence of severe self-antigen reactivity (Fig. 4). We have recently observed that PTEN is present at higher levels in 3–83 autoreactive B cells relative to non-autoreactive cells (Fig. 3 B). This supports a model in which autoreactive B cells physiologically dampen the activity of the PI3K-AKT pathway via increasing the expression of PTEN. A plausible scenario accounting for these results is that BCR engagement by self-antigen on immature B cells leads to strong and rapid PI3K activation that causes feedback upregulation of PTEN, which in turn inhibits the PI3K pathway to levels below those in cells that do not engage antigen, and thereby also increasing FOXO1 levels. The involvement of other pathways activating FOXO1 transcription is also likely, given that B cells lacking p85 or PI3Kδ can still further upregulate RAG1/2 expression in response to BCR stimulation^47,74. Of note, we observed increased phospho-ERK in autoreactive B cells expressing P110*⁷³. This suggests that the ERK pathway may still contribute to central B cell selection in vivo, but only in the context of PI3K activation.

Figure 4: Schematic of PI3K contribution to bone marrow selection of immature B cells.

This schematic illustrates the central role PI3K plays during the selection of immature B cells, regulating molecules that enforce or relax central tolerance. Autoreactive immature B cells must diminish PI3K activity to remain in the bone marrow (via CXCR4 upregulation) and undergo receptor editing (via FOXO1). Non-autoreactive immature B cells must activate PI3K to stop VJ recombination and preserve their selected BCR (via suppressing FOXO1), and to exit the bone marrow, enter the peripheral tissue, and further their differentiation (via downmodulating CXCR4 and upregulating BAFFR).

Rare individuals with gain of function (GOF) mutations in the PI3Kδ-encoding gene PIK3CD have been recently identified among immunodeficient patients with recurrent respiratory infections; their disease is referred to as activated PI3Kδ-syndrome (APDS⁸⁰). A third to half of APDS patients exhibit autoinflammatory manifestations and autoantibodies^81,82, which is in line with the ability of an active PI3K pathway to break peripheral B cell tolerance⁸³. Whether these individuals harbor also defects in central B cell tolerance has not yet been determined. To study the mechanisms by which PIK3CD GOF mutations promote immunodeficiency and autoimmunity, a few groups have independently generated mice expressing PI3Kδ-E1020K^82,84–86, which is analogous to the most common APDS GOF mutation (E1021K). These mice recapitulate well the phenotype of APDS, including the lymphopenia and phenotypic changes in B cells and T cells and the production of autoantibodies. Thus, these models represent terrific systems with which to tease apart the exact cause of the various clinical manifestations of APDS in patients.

The CXCR4-CXCL12 axis regulates the retention and release of mouse and human immature B cells during central B cell selection

The analysis of developing autoreactive B cells expressing P110* or carrying a deletion of PTEN have demonstrated that activation of PI3K not only abrogates receptor editing but also releases the autoreactive (unedited) B cells from the bone marrow where they relocate in the periphery. This raises the question: how is the bone marrow retention and egress of autoreactive and non-autoreactive immature B cells differentially regulated?

Numerous cell membrane receptors control the tissue location and retention of lymphocytes. These include the G protein-coupled chemokine receptors whose ligands form gradients that emanate from tissue-specific stromal cells. The CXCL12 chemokine is considered to play a major role in the attraction and retention of hematopoietic cells within the bone marrow parenchyma. CXCL12 (also known as SDF-1) is the canonical ligand for the chemokine receptor CXCR4, which is ubiquitously expressed in the hematopoietic system and in many non-hematopoietic cells⁸⁷. It has been shown that the expression of CXCR4 on developing B cells is mediated by the signaling of both the IL-7R and the pre-BCR, resulting in high CXCR4 levels in pro-B cells that attain even higher levels in pre-B cells (reviewed in⁸⁸). Membrane levels of CXCR4 are then sharply reduced in immature B cells^89–91, suggesting that the release of newly generated IgM⁺ B cells into the circulation is dependent on CXCR4 downmodulation. Indeed, analyses of mice in which B cells express either reduced amounts of CXCR4 or a hyper-active CXCR4 variant, demonstrates that CXCR4 surface levels and signaling regulate the bone marrow retention and release of immature/transitional B cells⁹². A revealing finding for us was that stimulating the BCR of bone marrow immature B cells increases CXCR4 levels (by about two-fold) within a few hours⁹², suggesting that developing autoreactive B cells may express higher levels of CXCR4 to remain in the bone marrow tissue.

Indeed, we have recently shown that CXCR4 is expressed 1.5-fold higher on ex-vivo 3–83 immature B cells that developed in the presence of self-antigen (i.e., autoreactive cells) relative to cells developing in the absence of antigen (i.e., non-autoreactive cells)⁷³. This difference in expression was also observed in human immature B cells developing in our central tolerance Hcκ hu-mouse model⁵⁵. Moreover, we found that expression of P110* in 3–83 autoreactive B cells leads to a reduction of CXCR4 to levels similar to those observed in cells developing in the absence of self-antigen⁷³. Finally, we reported that this small difference in the expression of CXCR4 is nevertheless functional, as it translates into differential migration of human autoreactive and non-autoreactive B cells toward CXCL12 in an in vitro transwell assay⁵⁵, and this is in line with published observations in B cells with comparable differences in CXCR4 expression^90–92. Given these findings, we set out to investigate whether retention of autoreactive B cells in the bone marrow microenvironment requires CXCR4 signaling. This question was addressed by the in vivo treatment of both 3–83Igi,H-2^b mice and Hcκ hu-mice with the well-known CXCR4 antagonist AMD3100⁹³. Results from these experiments were remarkably similar in the two models. Treating 3–83Igi,H-2^b mice and Hcκ hu-mice with AMD3100 leads to the release of autoreactive mouse and human B cells, respectively, from the bone marrow into the blood and the spleen. The release is quick and readily observable within one hour after treatment.

Our studies with AMD3100 demonstrate that the CXCR4-CXCL12 axis retains and restrains newly developed autoreactive B cells within the bone marrow tissue whereby they undergo central tolerance. Relatedly, CXCR4 signaling has also been shown to regulate accessibility of the Igk locus to the V(D)J recombination machinery⁹⁰ and, thus, could directly contribute to the receptor editing process. However, it would be surprising if CXCR4 represents the only safeguard between the bone marrow and the circulation. Joining the circulation from the bone marrow parenchyma requires first entering the sinusoids, thin walled and irregular vessels that lack a basal membrane and are highly permeable. In addition to this high permeability, sinusoids are enriched in chemoattractants such as sphingosine-1-phosphate (S1P), a lipid known to promote lymphocyte’s entry into the blood⁹⁴. Immature B cells express at least three of the five known S1P receptors, with the canonical S1PR1 showing the highest expression^95,96. Furthermore, deletion of S1PR1 has been shown to reduce bone marrow egress of IgM⁺ immature B cells⁹⁶, while deletion of S1PR3 neutralizes the migration of these cells toward S1P in vitro and reduces their entry into the sinusoids in vivo⁹⁵. In this latter study, it was also shown that non-autoreactive mouse immature B cells are able to migrate toward S1P, while autoreactive B cells are not. Finally, we were struck by the finding that in Hcκ hu-mice, human autoreactive immature B cells greatly upregulate CD69⁵⁵, a membrane protein whose surface expression is inversely correlated with surface expression of S1PR1⁹⁷. Overall, these data suggested that S1P receptors may contribute to the bone marrow egress of immature B cells in the context of CXCR4 downregulation or inhibition.

To test this possibility, we setup new experiments to treat autoreactive Hcκ hu-mice and 3–83Igi,H-2^b mice with either FTY720 (an antagonist for the S1PRs expressed by immature B cells) or the solvent DMSO, followed by the injection of AMD3100. Lymphocytes were analyzed in the blood of treated mice before, during, and after treatments (Fig. 5A). Autoreactive immature B cells were measured relative to the total circulating immature/transitional B cell population (Supplementary Fig. S1A, D) as done previously⁵⁵, and after establishing that the presence of immature/transitional B cells in blood is not significantly affected by FTY720 (Supplementary Fig. S1B, E). At the end of these in vivo treatments, we observed similar frequencies of mouse and human autoreactive immature B cells in circulation, whether the animals were treated or not with FTY720 (Fig. 5B, 5C). These findings indicate that AMD3100-mediated bone marrow release of (mouse and human) autoreactive immature B cells is independent of S1PR signaling. In addition, this was not unique to the robust CXCR4 inhibition mediated by AMD3100, because the bone marrow release of non-autoreactive 3–83 immature B cells was also unaffected by S1PR inhibition (Supplementary Fig. S1C).

Figure 5: Relative S1PR and CXCR4 contribution to the bone marrow retention of autoreactive mouse and human immature B cells.

A) Schematic of the longitudinal treatment of mice and hu-mice with either FTY720 or DMSO followed by AMD3100. Blood cells were analyzed before all treatments, after FTY720 or DMSO treatment, and one hour after the last AMD3100 injection. B) Top: Representative flow cytometric analysis of 3–83Igi,H-2^b mouse immature/transitional B cells (B220⁺CD19⁺CD2⁺CD24^highIgλ^–) gated as shown in Figure S1A. The plots show the gating of IgM^lowIgD^low autoreactive immature B cells. Bottom: Frequencies (mean and SD) of IgM^lowIgD^low (left graph) and IC-Igκ⁺IgM^lowIgD^low (right graph) mouse autoreactive immature B cells within all CD24^high immature/transitional B cells in PBMCs from 3–83Igi,H-2^b treated mice (N=4–5 mice per group). Statistical analyses were performed with a one-tailed Student’s t test. C) Left: Representative flow cytometric analysis of Hcκ hu-mice human immature/transitional B cells (hCD45⁺CD20⁺CD24^highCD38^high IgM⁺ or Igκ⁺) gated as shown in Figure S1D. The plots show the gating of IC-Igκ⁺IgM^low autoreactive immature B cells. Right: Frequencies (mean and SD) of IC-Igκ⁺CD20⁺ human autoreactive immature B cells within all (IgM⁺ or Igκ⁺) CD24^highCD38^high immature/transitional B cells in PBMCs from treated hu-mice (N=5–8 hu-mice per group). Statistical analyses were performed with one-way ANOVA. *p≤0.05; **p≤0.01; ***p≤0.001; ****p≤0.0001; ns=not significant.

In summary (Fig. 6), regulation of CXCR4 expression and signaling controls the bone marrow retention of autoreactive immature B cells and the bone marrow release of non-autoreactive immature B cells. In contrast, exit from the bone marrow and entry into the circulation does not significantly depend on S1PR signaling.

Figure 6: Contribution of CXCR4 signaling to the retention and release of bone marrow immature B cells.

Expression of CXCR4 progressively decreases from pre-B cells to IgM⁺ immature B cells. Immature B cells that engage self-antigen (i.e., autoreactive cells) do not downmodulate CXCR4, and signaling by this chemokine receptor is critical for their retention in the bone marrow tissue where they undergo receptor editing. CXCR4 antagonism by the drug AMD3100 releases autoreactive immature B cells into the circulation from where they relocate into the spleen.

Conclusions and future directions

Central B cell tolerance ensures our primary B cell repertoire is devoid of high avidity autoreactive B cells. As reviewed here, the PI3K-AKT pathway plays a deciding role in establishing whether high avidity autoreactive B cells remain in the bone marrow and undergo central tolerance induction or are released in the periphery where they might be recruited into an immune response (Fig. 4). The increase in PTEN and the downmodulation of CD19 that occur in developing autoreactive immature B cells following self-antigen binding, dampen the potency of the PI3K-AKT pathway, reducing FOXO1 phosphorylation and degradation and, thus, increasing its nuclear activity to promote RAG1/2 expression. Low PI3K also prevents CXCR4 downmodulation, ensuring that autoreactive cells stay in the bone marrow and undergo receptor editing. Moreover, reduced PI3K activity can also translate into higher expression of pro-apoptotic Bcl-2 family members, limiting the lifespan of autoreactive B cells.

The data we have discussed suggest that inhibition of PI3K is a major mechanism that prevents high avidity autoreactive B cells from bone marrow emigration. Indeed, releasing this brake either by introducing active PI3K or deleting PTEN, abrogates RAG1/2 expression and receptor editing and reduces CXCR4, fostering the release of autoreactive B cells into the circulation and the spleen. While the contribution of PTEN in enforcing central B cell tolerance is a foregone conclusion given its direct inhibition of PI3K, a better understanding of FOXO1 activity remains to be elucidated. Specifically, it remains unclear whether deletion of FOXO1 not only abrogates receptor editing, but also recapitulates the phenotype observed by the expression of P110* (or the deletion of PTEN), promoting the bone marrow egress and differentiation of high avidity autoreactive B cells. FOXO1 has been shown to control the expression of CXCR4 in germinal center B cells^98,99 and may therefore perform a similar function during B cell development. However, other Forkhead box family transcription factors, such as FOXO3 and FOXP1, could feasibly compensate for the lack of FOXO1 in this context^72,100. Another question to answer is whether the balance between PTEN levels and PI3K activity also controls the central selection of human B cells. Furthermore, it will be of interest to determine whether the enzyme AID, which also significantly contributes to establish central B cell tolerance in mouse and human B cells⁴¹, integrates with the PI3K pathway in this context. We envision that the expression of AID in immature B cells is (at least partly) negatively and positively regulated by PI3K-AKT1/2 and FOXO1, respectively, similarly to what described in germinal center B cells^71,101.

In this review we also discussed data demonstrating that the retention of high avidity autoreactive mouse and human B cells within the bone marrow tissue is dependent on CXCR4 signaling. In a way, it is surprising that inhibition of CXCR4 is sufficient to release bone marrow autoreactive B cells even before they are censored via editing. Pharmacologic inhibition of CXCR4 is utilized in some clinical settings⁹³, and genetic variants that modulate CXCR4 expression or signaling have also been described^93,102. Whether pharmacological inhibition or genetic alteration of CXCR4 leads to changes in the autoreactive capacity of the circulating B cell pool is not yet known and will be interesting to determine. Furthermore, whether the autoreactive B cells released from the bone marrow following inhibition of the CXCR4-CXCL12 axis are silenced in the periphery or survive and participate in (auto-) immune responses remains to be seen. The CXCR4-CXCL12 pathway has been positively associated with autoimmunity¹⁰³, but exceptions to this view have been reported¹⁰². As CXCR4-CXCL12 represents a chemokine receptor and gradient that attracts autoreactive B cells toward, and retains them within, the bone marrow tissue, a complementary gradient may also exist that attracts non-autoreactive B cells into the periphery. Our findings show that S1P does not represent this complementary gradient. Potential candidates for this task should include the chemokine receptors CXCR5 and CCR7, which are upregulated in immature B cells relative to their precursors^91,104.

Despite the existence of central B cell tolerance, one wonders whether preventing high avidity autoreactive B cells from entering the periphery is necessary to reduce the odds of developing autoimmunity, given peripheral B cell tolerance can efficiently reduce survival, activation, and differentiation of these autoreactive B cells. The fact that central B cell tolerance exhibits similar high stringency in healthy individuals while it is defective in most patients affected by autoimmune diseases⁵¹, argues that this process is beneficial. Patients with lupus, rheumatoid arthritis, type 1 diabetes, Sjogren’s syndrome, or myasthenia gravis, display a much higher frequency of autoreactive clones among their circulating immature/transitional B cells⁵¹. The data discussed in our review suggest these patients may suffer from dysregulated expression or activity of the PI3K and/or the CXCR4 signaling pathways. Indeed, lower PTEN has been documented in B cells from patients with lupus, type 1 diabetes, and autoimmune thyroid disease^105,106. Importantly, PI3K and PTEN play similar antagonistic roles in the maintenance of peripheral B cell anergy. Thus, genetic and epigenetic mechanisms that suppress PTEN expression and/or elevate PI3K activity would be expected to result in defects of both central and peripheral B cell tolerance, which is the case in the autoimmune diseases listed above.