B-1a cells acquire their unique characteristics by bypassing the pre-BCR selection stage

B-1a cells are long-lived, self-renewing innate-like B cells that predominantly inhabit the peritoneal and pleural cavities. In contrast to conventional B-2 cells, B-1a cells have a receptor repertoire that is biased towards bacterial and self-antigens, promoting a rapid response to infection and clearing of apoptotic cells. Although B-1a cells are known to primarily originate from fetal tissues, the mechanisms by which they arise has been a topic of debate for many years. Here we show that in the fetal liver versus bone marrow environment, reduced IL-7R/STAT5 levels promote immunoglobulin kappa gene recombination at the early pro-B cell stage. As a result, differentiating B cells can directly generate a mature B cell receptor (BCR) and bypass the requirement for a pre-BCR and pairing with surrogate light chain. This ‘alternate pathway’ of development enables the production of B cells with self-reactive, skewed specificity receptors that are peculiar to the B-1a compartment. Together our findings connect seemingly opposing lineage and selection models of B-1a cell development and explain how these cells acquire their unique properties.


SUMMARY:
B-1a cells are long-lived, self-renewing innate like B cells that predominantly inhabit the peritoneal and pleural cavities. In contrast to conventional B-2 cells they have a receptor repertoire that is biased towards bacterial and self-antigens, promoting a rapid response to infection and clearing of apoptotic cells. Although B-1a cells are known to primarily originate from fetal tissues the mechanisms by which they arise has been a topic of debate for many years. Here we show that in the fetal liver (FL) versus bone marrow (BM) environment, reduced IL-7R/STAT5 levels promote immunoglobulin kappa (Igk) recombination at the early pro-B cell stage. As a result, B cells can directly generate a mature B cell receptor (BCR) and bypass the requirement for a pre-BCR and pairing with surrogate light chain (SLC). This 'alternate pathway' of development enables the production of B cells with self reactive, skewed specificity receptors that are peculiar to the B-1a compartment. Together our findings connect seemingly opposing models of B-1a cell development and explain how these cells acquire their unique properties.

INTRODUCTION:
B-CLL is the most common form of adult leukemia in the western world 1 . In the early 1980s, the T-cell antigen CD5 (Ly-1) was identified on the surface of cancerous B-cells in patients with B-CLL 2,3 . This observation led to the search for normal CD5 + B-cell counterparts to potentially determine the cancer cell of origin. As a result of these efforts, CD5 + B cells, otherwise known as B-1a cells, were discovered in mice 4,5 . Further characterization of CD5 + B-1a B cells revealed that they are long-lived, self-renewing cells that predominantly reside in the pleural and peritoneal cavities where they produce natural polyreactive IgM antibodies with a biased, autoreactive repertoire. In contrast to conventional B-2 cells, B-1a cells produce antibodies with reduced junctional diversity and less somatic hypermutation 6 . Furthermore, Igh V H gene rearrangements favor V H 12 segment usage 7 , generating antibodies that interact with phosphatidylcholine (PtC), a major lipid in the protective mucus layer of the gastrointestinal tract that is also present in the membranes of diverse bacteria. Thus, the B-1a receptor repertoire is biased towards bacterial and self-antigens, which is important for mounting a rapid immune response to infection and in the clearing of apoptotic cells [8][9][10] . Because B-1a cells are found in preimmune mice, they function as an important first line of defense against bacterial pathogens.
These characteristics distinguish B-1a cells from conventional B-2 cells, which have a highly diverse receptor repertoire that is important for mediating adaptive immunity.
Although B-1a cells were discovered in the early 1990s, their origin has been hotly debated since, and despite the efforts of numerous labs this remains an unresolved issue. The controversy has mainly been centered on two opposing models, the lineage model and the selection model. The lineage model proposes that a distinct B-1 progenitor cell gives rise to B-1a cells, while the selection model favors the idea that a common B cell progenitor can acquire a B-1a or a B-2 fate depending on the type of antigen it recognizes 9,11 . Support for the lineage model comes from early reconstitution experiments, which reveal that fetal tissues are much more efficient at generating B-1a cells in irradiated recipient mice than adult bone marrow counterparts 12 . Furthermore, the first wave of B-1a cells was shown to originate in early embryos in an HSC-independent manner [13][14][15][16][17] . However, cellular barcoding experiments demonstrate that a single progenitor cell can give rise to both B-1a and B-2 cells 18 challenging the notion that B-1a cells arise from a distinct lineage. Moreover, the finding that B-1a cells have a restricted and biased receptor repertoire provides support for a selection model 9,19 .
Investigations have also focused on the molecular switches that influence CD5 + B-1a cell fate. In this context, a lin28b/let7 axis has been characterized 20-22 which impacts B-1a development. The lin28b and let7 miRNAs respectively promote and inhibit the expression of the transcription factor, Arid3a, which in turn can drive the development of CD5 + B-1a cells 21,22 .
Nonetheless, the B cell receptor (BCR) repertoire of the resulting cells do not include PtC specific antibodies that are characteristic of a typical B-1a cell compartment 21 . Thus, enforced expression of Arid3a fails to fully explain how B-1a cells develop. Another transcription factor, BHLHE41 has also been shown to be important in B-1a cell biology 23 . Specifically, cells deficient in this transcription factor lose B-1a cells expressing V H 12/V k 4 PtC specific receptors, have impaired BCR signaling, increased proliferation and apoptosis. BHLHE41 therefore plays an important role in B-1a maintenance by regulating self-renewal and BCR repertoire; however, it is not known whether its forced expression can drive development of these cells.
In the fetus, B cell development takes place in the liver and moves to the bone marrow after birth. Each stage of development is marked by a particular rearrangement event that drives differentiation forward. These recombination events occur in a stage specific manner. The first step involves the joining of the immunoglobulin heavy-chain (Igh) D H and J H gene segments within pre-pro B cells. Rearrangement continues at the pro-B cell stage where V H -to-DJ H joining is both initiated and completed. Rearrangement of the immunoglobulin light-chain loci, Igk or Igl occurs at the subsequent pre-B cell stage of development. Igh and Igk rearrangement is separated by a proliferative burst of large pre-B cells that allows individual cells that have successfully rearranged their heavy-chain to clonally expand. At the following small pre-B cell stage, each B cell undergoes a distinct Igk recombination event 24 . Ultimately, this results in unique heavy and light-chain pairs that expand the antigen receptor repertoire. The successful pairing of immunoglobulin heavy-chain with surrogate light chain (SLC) forms the pre B cell receptor (pre-BCR), which is required for expansion of large pre-B cells and subsequent differentiation to the small pre-B cell stage where Igk recombination occurs. Since SLC pairs poorly with autoreactive heavy-chains, the pre-BCR provides a mechanism for negative selection of self-reactive B cells 25,26 .
However as noted early on, in a small fraction of B cell progenitors in the bone marrow Igk rearrangements occur prior to rearrangements at Igh and independent of SLCs [27][28][29][30] . In addition, in the absence of SLCs and thus pre-BCR expression, an autoreactive BCR can drive B cell development efficiently to the stage of the immature B cell, where BCR diversification and counterselection of autoreactivity is achieved through the process of receptor editing. This has led to a model in which early B cell development is driven by a positive signal from the pre-BCR (in the majority of progenitors) or an autoreactive BCR (in a minority of cells), with the pre-BCR having evolved as a surrogate autoreactive BCR 31,32 .
Downstream of pre-BCR signaling, IL-7 receptor (IL-7R) signaling is extinguished at the small pre-B cell stage. This is noteworthy because the IL-7R signaling pathway is responsible for directing the sequential ordering of recombination events, i.e. on heavy-chain gene followed by light-chain gene rearrangement in classical B cell development 33,34 . IL-7 is a cytokine secreted by stromal cells within the bone marrow where development takes place throughout the post-natal and adult life of the animal. In murine bone marrow, development past the prepro-B cell stage stringently requires IL-7R 35,36 . In contrast, B cell development within the fetal liver can occur independent of IL-7 37,38 . In terms of directing recombination, IL-7 and its downstream signaling component STAT5, have been shown to promote Igh accessibility and recombination [39][40][41] while actively inhibiting recombination of the Igk locus 33,42 . Activated STAT5 enters the nucleus and forms a complex with PRC2/EZH2 which binds to the intronic enhancer of Igk (iEκ) and induces H3K27me3-mediated repression to inhibit recombination of this locus 43,44 . Strikingly, conditional deletion of STAT5 within pro-B cells results in increased Igk recombination at the pro-B cell stage 34 .
It is known that B-1a cells originate primarily from fetal tissues, however it remains unclear what pathways drive B-1a cell development and lead to the acquisition of their unique characteristics. In addition, it has been shown that B-1a cells are efficiently generated in IL-7 -/and IL-7R -/mice albeit at reduced numbers compared to wild-type 37,38,45,46 . Given our finding that fetal liver (FL) pro-B cells have less active cytoplasmic pSTAT5 than bone marrow (BM) pro-B cells 47  Indeed, extending earlier work 27 we demonstrate that while SLC-independent development leads to a significant reduction in B-2 cell numbers, B-1a cells with their characteristic V H 12 antiphosphatidylcholine bias are still efficiently generated. Thus, reduced IL-7R/STAT5 signaling promotes an alternate pathway of development which favors the production of B-1a cells with self-reactive receptors characteristic of this B cell subset. Together these data connect opposing models of B-1a cell development and explain how these cells acquire their unique properties.

RESULTS: Early Igk recombination is increased in fetal liver versus bone marrow pro-B cells as a result of reduced STAT5-mediated repression
Fetal liver (FL) and bone marrow (BM) derived progenitor cells are differentially dependent on IL-7 for development and we have found that fetal liver derived pro-B cells have significantly lower levels of the downstream signaling component, phospho-STAT5 compared to their bone marrow derived pro-B cell counterparts 47 . Thus, we asked whether alterations in phospho-STAT5 levels correlate with altered regulation of Igk recombination in these two anatomic locations. To explore this, we analyzed recombination from ex vivo derived bone marrow (5-6wk adults) and fetal liver (E17-18) pro-B cells using semi-quantitative PCR analysis (Fig. 1a).
This assay uses a degenerate Vκ primer with a common reverse primer downstream of Jκ gene segments to measure recombination of a Vκ with each of the four Jκ segments 48 . As shown in Figure. 1a, FL pro-B cells have increased Igk recombination relative to BM pro-B cells.
To address the question of altered Igk recombination between these two cell types in a more quantitative manner, we made use of a real-time PCR assay that quantities Jκ1 rearrangement by assessing the retention of germline Igk 49 . In this assay, a product cannot be generated after recombination because the sequence to which the upstream primer binds is lost or inverted during the recombination process. Quantitation of the amplified product is determined as a ratio between the single copy gene β-actin and the germline Igk band. Control tail DNA, in which no Igk rearrangement occurs, is set at 100%. Consistent with the semiquantitative experiments described above, fetal liver pro-B cells have lower levels of Igk germline retention (61%) relative to bone-marrow pro-B cells (87%) (Fig. 1b).
To determine the proportion of cells that are actively undergoing recombination at a single cell level, we used three-dimensional (3-D) immuno-flourescent in situ hybridization (FISH) assay that visualizes recombination by the association of γ-H2AX foci with the individual alleles of antigen receptor loci. γ-H2AX DNA repair foci have been shown to be associated with antigen receptor loci undergoing recombination 50,51 . In these experiments, we used two Igk BAC DNA probes that hybridize to the distal Vκ24 (RP23-101G13) gene region and the Cκ region (RP24-387E13) in combination with an antibody against the phosphorylated form of γ-H2AX (Fig. 1c). We found significantly increased association of γ-H2AX foci with the Igk locus in fetal liver versus bone marrow derived pro-B cells ( Fig. 1d). These numbers reflect the combined data from two independent experiments (Supplementary Table S1).
Antigen receptor loci undergo large-scale contraction through chromatin looping which enables rearrangement between widely dispersed gene segments. Locus contraction is tightly linked both to recombination status and usage of distal gene segments 48,52 . To address the contraction status of the Ig  locus we measured the distance separating the distal Vκ24 and Cκ probes in interphase cells. Our analyses indicate that FL pro B cells are significantly more contracted than wild-type (WT) double positive (DP) T cells, a cell type in which Ig  rearrangement does not occur (Supplementary Fig. 1a,b). Indeed the level of contraction is comparable to that seen in wild-type bone marrow derived pre-B cells where light-chain rearrangement is known to occur.
It is known that activated STAT5 forms a complex with PRC2 and binds to the intronic enhancer of Igk, iEκ to inhibit Igk recombination 43,47,53 . To determine if the increase in early Igk rearrangement could be caused by reduced STAT5-mediated repression, we assessed how much STAT5 is bound to Igk via STAT5-ChIP qPCR. As shown in Fig. 1e, we found that FL pro-B cells have significantly less STAT5 bound to iEκ compared to BM pro-B cells and binding more closely resemble levels found in BM pre-B cells where Igk recombination normally occurs.
Taken together, these findings indicate that in the FL versus BM environment Igk rearrangement occurs at higher levels in pro-B cells due to a decrease in STAT5-mediated repression.

Pro-B cells that undergo early Igk rearrangement can bypass the pre-BCR checkpoint and make B cells that express a mature BCR.
Classical bone marrow B cell development starts with Igh recombination at the pre-pro to pro-B cell stages followed by Igk/Igl recombination at the small pre-B cell stage. However, JHT mice that lack J H segments as well as the Igh intronic enhancer, and are therefore unable to undergo Igh recombination, can rearrange Igk/Igl independent of Igh rearrangement ( Supplementary Fig. 2) 28,54,55 This is also true for SLC deficient mice and a fraction of B cells undergoing development in the bone marrow of wild-type mice 56  which are also IgK + . Importantly, B1.8 mice have a substantial pre-B cell compartment which is not detected in Igll1 -/-, B1.8 mice despite similar IgM + B cell output (Fig. 2). This shows that having a productive heavy and light-chain can bypass the requirement for pairing with SLC to form immature B cells that express the mature BCR. Together this data shows that in the fetal liver, expression of Igh and Igk at the pro-B cell stage can bypass the developmental block induced by surrogate light chain deficiency by forming B cells with a mature BCR.

B-1a cells are efficiently generated in surrogate light deficient mice.
B-1a cells are primarily generated from fetal tissues, and we have shown that FL derived B cells frequently undergo early Igk recombination allowing cells to bypass the pre-BCR selection stage. Thus we next asked whether an absence of surrogate light chain could impact the generation of B-1a cells 12,57 . Here we analyzed the B-2 as well as the B-1 compartment, which consists of both B-1a and B-1b cells. B-1b cells are distinct from B-1a cells in that they lack CD5 expression and have a memory function in protecting against bacterial infections 58 . As expected, peritoneal B-2 cells were significantly decreased compared to controls. Furthermore although B-1b cells were also significantly decreased, the B-1a cell compartment remained intact. This was reflected in cell percentage as well as total cell numbers (Fig. 3a,b). These data indicate that while B2 and B1b cells development is strongly promoted by SLC, B-1a cells do not depend on a pre-BCR checkpoint. It should be noted that B1.8 transgenic mice could not be used to analyze the B-1a cell compartment in surrogate light chain deficient mice because the B1.8 pre-rearranged heavy-chain does not support B-1a cell development 59 .
Two different subsets of B-1a cells that segregate different functions of the B-1a cell compartment, have been identified by plasma cell alloantigen 1 (PC1) 60,61 . In order to determine whether or not both of these subsets of B-1a cells were represented in B-1a cells generated in surrogate light chain deficient mice, we looked at PC1 expression. B-1a cells from Igll1 -/mice were found to have similar proportions of PC1 lo and PC1 hi B-1a cells as in wild-type mice ( Supplementary Fig. 3). This suggests that not only is the B-1a compartment size normal in surrogate light chain knockout mice, but PC1 lo and PC1 hi subsets are maintained.

Constitutive phospho-STAT5 signaling selectively inhibits B-1a cell development.
Our data suggest that low phospho-STAT5 signaling induces early Igk recombination in FL pro- In these mice we found that although B-2 and B-1b cell development were not significantly affected, all three independent littermate controlled Stat5b-CA mice exhibited a decrease in the B-1a cell compartment. This is reflected in cell proportions as well as total cell numbers (Fig. 4a,b). Thus, STAT5 signaling inhibits B-1a cell development underscoring the importance of reduced IL-7R/STAT5 signaling in driving the development of these cells.

SLC independent B-1a cells express receptors with phosphatidylcholine (PtC) specific V H 12 gene rearrangements
B-1a cells are known to have a strong bias for immunoglobulin rearrangements that confer specificity to phosphatidylcholine (PtC) (V H 12/V k 4). In addition, the pre-BCR is known to select against B-1a specific autoreactive heavy-chains as a result of a defect in pairing of the latter with surrogate light chain 25,26 . Given these facts we next asked whether B-1a cells that develop independent of SLC could lead to the generation of B-1a specific rearrangements. To address this question we analyzed peritoneal cavity B cells for V H 12 usage and specificity against PtC.
As shown in Fig. 5a B-1a cells from Igll1 -/mice express V H 12 rearranged receptors. Importantly, these V H 12 + B-1a cells bind PtC containing liposomes inferring that they have the canonical V H 12/V k 4 gene segment bias (Fig. 5a). Additionally, there is not a significant difference in the proportion of B-1a cells that have specificity against PtC (Fig. 5b). Thus, surrogate light chain independent B-1a cell development supports the generation of these B-1a cells.

B-1a cells have a similar gene segment usage to wild-type controls.
To determine how SLC deficiency affects other V H gene segment usage, we sequenced the BCR repertoires of B-1a cells from WT and SLC-deficient mice. Briefly, peritoneal cavity B-1a cells were sorted, lysed and the rearranged Igh products amplified by PCR. Illumina adapters and barcodes were added and the rearrangements sequenced by next-generation sequencing.
BCR repertoires were analyzed by IMGT/High-VQUEST and visualized using IMGT/StatClonotype [63][64][65][66] . These analyses revealed that B-1a cells from Igll1 -/mice have a similar overall V H usage compared to those of wild-type mice (Fig. 6). This shows that B-1a cells that develop in the absence of SLC have a wide distribution of gene segment usage rather than representing the progeny of a few rare progenitors. As expected V H 12 (V H 12-3) usage, was found to be high in wild-type B-1a cells and levels were matched in the cells from Igll1 -/mice ( Fig. 6) 7 . The rearrangement studies support the data from our FACs analysis and underscore the finding that B-1a cells generated in the absence of SLC maintain a V H 12 gene segment bias.

DISCUSSION:
Although it was known that B-1a cells predominantly develop in the fetal liver, the pathways driving development and the mechanisms underlying the generation of B cells with a repertoire skewed towards autoreactivity was not previously known. Indeed, this has been a subject of debate for many years. Here we now reveal that a reduction in IL-7 signaling in the fetal liver The idea that B-1 cells can develop in a SLC independent manner is not novel. In fact, by SLC deficiency, however the distinct effect on the B-1a versus B-1b cell compartment was not investigated 27 . Our analyses demonstrate that while B-2 and B-1b cell development is severely impaired in Igll1 -/mice, B-1a cells can be generated at wild-type levels. Furthermore, SLC independent B-1a development could occur through a mechanism that involves reduced IL-7/STAT5 signaling and early Igk rearrangement. We found that B-1a cell development is selectively impaired in Stat5b-CA mice and this provides additional support for the idea that SLC independent B-1a cell development is favored by a decrease in IL-7 signaling.

Immuno-DNA FISH
Combined detection of γ-H2AX and Ig loci was carried out on cells adhered to poly-L lysine coated coverslips as previously described 50 . Cells were fixed with 2% paraformaldehyde / PBS

Confocal microscopy and analysis
Cells were analyzed by confocal microscopy on a Leica SP5 Acousto-Optical

STAT5 ChIP-qPCR
STAT5 ChIP-qPCR was conducted as previously described 43 . Purified DNA was then analyzed by quantitative real-time PCR. Samples were analyzed in triplicate and represent three biological replicates.

Igh repertoire DNA library preparation and sequencing analysis
Cells were sorted and lysed in 0.5 mg/mL proteinase K at 55C for 4 hours, 85C 10 mins. Lysed cells were used as template for HotStarTaq (Qiagen) PCR amplification. Primers used to amplify rearrangements were adapted from V H FR3 and J H BIOMED-2 primers 74,75 . Additional primers designed to capture some V genes missing from the BIOMED-2 study, which include V H 12 genes (Supplementary Table S2). Amplified rearrangements were purified by gel extraction (Qiagen) followed by a standard end-repair reaction. Following purification steps were carried out by Ampure XP bead purification (Beckman Coulter). In order to attach Illuminacompatible adapters, samples were treated by standard dA tailing followed by adapter ligation using Quick Ligase (NEB) mixed with preannealed NEXTflex DNA Barcodes (Bioo Scientific). IMGT/StatClonotype was used to Analyze statistically significant gene segment usage between samples 63,65 .

ACKNOWLEDGMENTS:
We would like to thank members of the Skok lab for thoughtful discussions and critical comment on the study and manuscript. We would also like to thank the NYU Flow Cytometry and Cell Sorting Center as well as the NYU Genome Technology Center, both of which are supported in part by the Cancer Center Support Grant P30CA016087 at the Laura and Isaac Perlmutter Cancer Center. JAS is supported by NIH grants R35GM122515 and R21CA188968.
JBW was previously supported by the T32 CA009161 training grant (Levy) and is currently supported by the 2T32 AI100853-06 (Reizis) training grant. SLH was supported by an American

Society of Hematology (ASH) Scholar Award and by a Molecular Oncology and Immunology
Training Grant NIH T32. MAF is supported by NIH grants CA151845 and CA154998.