Regulation of B-cell development and tolerance by different members of the miR-17∼92 family microRNAs

The molecular mechanisms that regulate B-cell development and tolerance remain incompletely understood. In this study, we identify a critical role for the miR-17∼92 microRNA cluster in regulating B-cell central tolerance and demonstrate that these miRNAs control early B-cell development in a cell-intrinsic manner. While the cluster member miR-19 suppresses the expression of Pten and plays a key role in regulating B-cell tolerance, miR-17 controls early B-cell development through other molecular pathways. These findings demonstrate differential control of two closely linked B-cell developmental stages by different members of a single microRNA cluster through distinct molecular pathways.

A defining feature of B-cell development is the process of gene rearrangements in the B-cell receptor (BCR) loci, through which B cells acquire the capacity to express a BCR of a given specificity on the cell surface 1 . These rearrangements occur in an orderly manner over time, sequentially involving the immunoglobulin heavy (IgH) and light (IgL) chain genes during the pro-B and the pre-B stages of B-cell development, respectively. This is thought to depend on an orderly accessibility of the corresponding loci to the RAG1 and 2 recombinases, which mediate rearrangements of variable (V), diverse (D), and joining (J) gene segments through the process called V(D)J recombination. At the end of this process, each immature B cell expresses a single IgH and a single IgL chain, with a single and, most times, unique antigen specificity. The combinatorial and stochastic nature of gene rearrangements leads to the generation of immature B cells with self-reactive receptors. At this immature stage, the newborn B cells undergo the first checkpoint for self-reactivity, to eliminate potentially autoreactive cells by central tolerance mechanisms. Once a newborn B cell encounters a self-antigen for which its BCR is specific, it attempts to escape autoreactivity by continuing V(D)J recombination at the IgL locus (receptor editing) or dying by apoptosis (clonal deletion). When the cell has passed this developmental checkpoint, it differentiates into a mature B cell. Self-reactive B cells can be further regulated in the periphery through peripheral tolerance mechanisms, including the induction of anergy 2,3 . Previous studies of human B cells showed that self-reactivity is progressively diminished during normal B-cell development, consistent with the idea that several tolerance mechanisms are at work at different stages of the life of B cells 4 .
Despite intensive study, our understanding of molecular pathways regulating B-cell development and tolerance is still incomplete. Specifically, the function of individual microRNAs (miRNAs) in B-cell development and tolerance remains poorly understood. miRNAs are endogenously encoded single-stranded RNAs of B22 nucleotides in length. To date, B2,500 human and B1,900 mouse miRNAs have been identified and many of them play essential roles in the immune system [5][6][7] . They regulate gene expression by pairing with messenger RNAs through imperfect sequence complementarity, resulting in reduced protein output by mRNA cleavage, translational repression or promotion of mRNA decay 8 . It has been estimated that 25B40% miRNA precursors are located in close proximity (o10 kb) of other miRNA precursors, constituting miRNA clusters [9][10][11] . The majority of miRNA clusters are first transcribed into single polycistronic primary transcripts (pri-miRNAs) and then cleaved by Drosha into individual hairpins (pre-miRNAs), which are further processed by Dicer to produce mature miRNAs. Gene expression profiling studies have shown that the expression of different miRNAs in a cluster is generally co-regulated 12,13 , suggesting that they may cooperate with each other to accomplish common functions. Furthermore, comparative genomics show that miRNA clusters are evolutionarily stable and conserved across species, suggesting functional importance of such organization 14,15 . Some clusters consist of miRNAs with identical seed regions (termed homogeneous miRNA clusters), probably a result of gene duplication. The regulatory effect of homogeneous miRNA clusters may simply be an increase in gene dosage. Other clusters are composed of miRNAs with different seed regions (termed heterogeneous miRNA clusters). It remains unclear how members of heterogeneous miRNA clusters operate together to accomplish common functions 15 .
In this study we dissected the roles of the miRNA-17B92 family miRNAs at different stages of B-cell development. The miR-17B92 family consists of three miRNA clusters: miR-17B92, miR-106aB363 and miR-106bB25. Together, these three clusters contain 15 miRNA stem loops that give rise to 13 distinct mature miRNAs, which fall into four miRNA subfamilies (miR-17, miR-18, miR-19 and miR-92 subfamilies), with members in each subfamily sharing the same seed sequence 16 . The genomic organization and mature miRNA sequences of this family are conserved in all vertebrates 17 . During lymphocyte development, these miRNAs are highly expressed in progenitor cells, with expression levels decreasing two-to threefold on maturation 13,18 . Mouse genetic studies showed that miR-17B92deficient mice were runted and died at birth due to lung and heart hypoplasia. In the haematopoietic system, there was a partial block of early B-cell development at the pro-to pre-B transition. Deletion of miR-106aB363 and miR-106bB25 has no obvious phenotypic consequences, whereas compound mutant embryos lacking both miR-17B92 and miR-106bB25 died at midgestation, suggesting functional redundancy between these miRNA clusters 19 . Conversely, overexpression of miR-17B92 family miRNAs occurs frequently in a broad spectrum of human cancers and in lymphocytes from patients with autoimmune diseases 16,[20][21][22][23] . These observations suggest that miR-17B92 is involved in lymphoma development and autoimmune diseases. We have generated a conditional miR-17B92 transgenic allele (termed miR-17B92 Tg) whose expression can be turned on by Cre recombinase 24 . When miR-17B92 Tg was turned on specifically in B cells using CD19-Cre, transgenic mice exhibited premature death 25 . About 80% of those mice developed lymphomas, demonstrating that miR-17B92 is a powerful cancer driver 25 . Notably, the other 20% miR-17B92 transgenic mice died of autoimmune diseases. We speculate that miR-17B92 might play a critical role in the control of B-cell tolerance. In this study, we use two newly generated in vivo models to investigate the function and mechanism of miR-17B92 family miRNAs in regulating B-cell development and tolerance, and demonstrate functional specificity of different members of this cluster at two closely linked developmental stages of B cells.
Results miR-17B92 regulates B-cell central tolerance. To analyse the effect of transgenic miR-17B92 expression on B-cell central tolerance, we used the recently established IgM b -macroself mouse model 26 . In this model, mice are engineered to ubiquitously express a superantigen reactive to the heavy chain constant region of IgM, the first BCR expressed on the surface of immature B cells. As receptor editing changes only the light chain, it fails to eliminate superantigen reactivity and all developing B cells undergo cell death by clonal deletion. As a consequence, these mice have almost no mature B cells in the spleen and lymph nodes. We analysed B-cell development in IgM b -macroself recipients (with surface marker CD45.1 þ ) reconstituted with bone marrow from CD19-Cre;miR-17B92 Tg/Tg (TG) or wild-type (WT) mice (with surface marker CD45.2 þ ) (Fig. 1a). The WT-IgM b -macroself chimeras exhibited a severe B-cell lymphopenic phenotype recapitulating that of the IgM bmacroself mice, characterized by a complete developmental block at the immature B-cell stage in the bone marrow that results in the absence of mature B cells in the spleen. Strikingly, B-cell development in the TG-IgM b -macroself chimeras was similar to that in WT-WT chimeras (Fig. 1b,c). Consistent with the restoration of peripheral B cells, serum IgM levels were also substantially restored in the TG-IgM b -macroself chimeras ( Supplementary Fig. 1a). Moreover, these rescued B cells appeared to be functional, as TG-IgM b -macroself chimeras mounted a close-to-normal antibody response upon NP-CGG immunization ( Supplementary Fig. 1b). This is the first time that we observed such a complete restoration of B-cell development in the IgM b -macroself central tolerance model by a miRNA gene.
Central tolerance, when functionally intact, protects against autoimmunity by purging a large majority of self-reactive B cells from the B-cell compartment 2,3 . As B cells with elevated miR-17B92 expression were able to escape from central tolerance in IgM b -macroself model, in which all B cells are selfreactive, we reasoned that the BCR repertoire in TG mice might be significantly different from that of WT mice. To assess this, we analysed the BCR repertoire by performing deep sequencing of Ig heavy chain V genes (IGHV) of WT and TG mice. As predicted, TG mice exhibited an IGHV landscape different from that of WT mice, with drastically increased usage in TG mice (43-fold) of three IGHV genes (IGHV12-3, IGHV4-1 and IGHV11-2) (Fig. 1d). Among these, V H 11 and V H 12 gene families are predominantly associated with germline-encoded autoantibodies of phosphatidylcholine specificities 27,28 . A large numbers of V H 11 þ B cells, which are rarely detected in the spleen of WT mice, were present in the spleen of TG mice, as detected by using an anti-idiotypic monoclonal antibody for V H 11 ( Supplementary  Fig. 1c). Conversely, we found reduced usage of IGHV11-2 and IGHV12-3 in B cells deficient for the miR-17B92 miRNA family (CD19tKO) (Supplementary Fig. 1d). In addition, we measured the presence of anti-double-stranded DNA autoantibodies in the serum of WT and TG mice at terminal analysis. High titres of autoantibodies were detected in 4 out of 45 TG mice, but were not found in any of the WT mice ( Supplementary Fig. 1e).
We next investigated the function of physiological levels of miR-17B92 family miRNAs in B-cell central tolerance. We analysed receptor editing in WT mice and mice with B-cellspecific deletion of this miRNA family (by either Mb1-Cre or CD19-Cre, termed Mb1tKO and CD19tKO mice, respectively). Deletion of the miR-17B92 family miRNAs in B cells impaired receptor editing, as indicated by increased k/l-light chain ratios and reduced percentages of l-light chain-positive (Igl þ ) immature B cells ( Supplementary Fig. 2a-d). Conversely, transgenic miR-17B92 expression promoted receptor editing ( Supplementary Fig. 2c-d). These results show that endogenous levels of miR-17B92 family miRNAs exquisitely regulate receptor editing at the B-cell central tolerance checkpoint. We further analysed the effect of transgenic miR-17B92 expression on B-cell central tolerance in the IgHEL;mHEL mouse model 29   are deleted in the bone marrow and few mature B cells appear in the peripheral lymphoid organs 30 . Consistent with the previous report, we detected very small numbers of B cells bearing IgHEL (IgD a þ ) in the spleens of mHEL mice reconstituted with bone marrow cells from WT IgHEL mice (WT; IgHEL). The number of splenic IgHEL (IgD a þ ) B cells increased by fivefold when mHEL mice were reconstituted with bone marrow cells from IgHEL mice with B-cell-specific transgenic miR-17B92 expression (TG;IgHEL) ( Supplementary Fig. 3). These data, together with previous results from the IgM b -macroself model (Fig. 1b,c), demonstrate a general effect of elevated miR-17B92 expression on B-cell selection against self-antigens. Therefore, our data show that miR-17B92 is a critical regulator of B-cell central tolerance at the immature B-cell stage.
miR-19 plays a key role in B-cell central tolerance. We next investigated the functional contribution of individual miR-17B92 members in regulating B-cell central tolerance. The six miRNAs of miR-17B92 fall into four miRNA subfamilies (the miR-17, miR-18, miR-19 and miR-92 subfamilies), with members in each subfamily sharing the same seed region and probably similar functions. We generated two groups of lentiviral vectors expressing either the miR-17B92 cluster with one miRNA subfamily deleted, or four tandem copies of individual miRNAs (Fig. 2a). Northern blot analyses confirmed that these lentiviral vectors expressed encoded miRNAs at levels comparable to a lentiviral vector encoding the full-length (FL) miR-17B92 cluster (Fig. 2b). Highly enriched CD45.2 þ WT haematopoietic stem cells (HSCs) were transduced with lentiviruses encoding various combinations of miR-17B92 miRNAs and were used to reconstitute lethally irradiated IgM b -macroself recipients (CD45.1 þ ), together with unmanipulated WT bone marrow cells (CD45.1 þ ) (Fig. 2c). Twelve weeks after reconstitution, the bone marrow chimeras were analysed for splenic B cells. Although chimeras reconstituted with control lentivirus-transduced HSCs exhibited the same B-cell lymphopenic phenotype as unmanipulated IgM b -macroself mice, lentiviral expression of miR-17B92 restored the B-cell compartment in IgM b -macroself recipient mice, similar to that by TG bone marrow cells (Figs 1b,c and 2d,e). The contribution of each miR-17B92 miRNAs to the regulation of B-cell tolerance was determined by quantifying the numbers of splenic B cells that escaped IgM b -macroself-mediated deletion. As shown in Fig. 2e, deletion of the miR-19 subfamily almost completely abrogated the escape of B cells, whereas miR-17 subfamily deletion caused a modest reduction in splenic B-cell numbers. Deletion of miR-18 or miR-92 did not alter the ability of miR-17B92 to restore B-cell development in IgM bmacroself recipient mice. Consistently, lentiviral expression of miR-19 alone restored B-cell development substantially, whereas other individual miR-17B92 miRNAs did not have any significant effect (Fig. 2e). Based on these observations, we conclude that miR-19 subfamily miRNAs are the key members of the miR-17B92 cluster in regulating B-cell central tolerance, whereas miR-17 subfamily miRNAs play a supporting role in this process.

miR-19 controls B-cell central tolerance by suppressing Pten.
Our previous photoactivatable-ribonucleoside-enhanced crosslinking and immunoprecipitation analyses of mature B cells identified 868 protein-coding genes that contain miR-17B92 miRNA-binding sites conserved in humans and mice 25 . To gain new insights into the molecular mechanisms by which miR-17B92 regulates B-cell central tolerance, we focused on target genes containing binding sites for miR-19, the critical member of miR-17B92 in this process. The miR-19 targetome contains 386 protein-coding genes with 404 binding sites. These genes function in a broad spectrum of biological processes. Among them, regulators of the phosphatidylinositol-3 kinase (PI3K) pathway (phosphatase and tensin homologue (PTEN) and Phlpp2) and apoptosis pathway (Bim) are of special interest. The PI3K pathway plays a critical role in supporting B-cell survival at the mature B-cell stage 31 , whereas the apoptosis pathway is thought to determine the cell fate during B-cell central tolerance 32,33 .
To assess whether miR-17B92 regulates the expression of these target genes in immature B cells, we purified B220 þ IgM þ CD93 þ IgD À immature B cells from an in vitro culture of bone marrow B-cell precursors ( Supplementary Fig. 4) and measured protein expression by immunoblot analysis. Expression of PTEN and Phlpp2 in TG immature B cells was reduced to 70% and 80% of WT levels, respectively. The expression levels of Bim were reduced marginally and this reduction did not reach statistical significance (Fig. 3a). To determine whether miR-19 directly binds to the Pten and Phlpp2 mRNAs, we performed reporter assays with a Renilla luciferase (Rluc) reporter gene containing the Pten and Phlpp2 3 0 -untranslated region (UTR) fragments encompassing the predicted miR-19 binding sites. Overexpression of miR-19 decreased the Pten reporter activity by 440%, but did not reduce the Phlpp2 reporter activity (Fig. 3b). Mutation of the miR-19 binding site in the Pten 3 0 -UTR abrogated the inhibitory effect of miR-19, further confirming that miR-19 directly regulates Pten expression. Consistently, miR-19 overexpression significantly reduced endogenous PTEN, but not Phlpp2, protein levels in HeLa cells (Fig. 3c).
To evaluate whether the reduction in PTEN protein levels contributes to miR-17B92 regulation of B-cell tolerance, we reconstituted lethally irradiated IgM b -macroself mice with bone marrow from Pten-deficient mice. Heterozygous deletion of Pten (Pten fl/ þ ;CD19-Cre) resulted in significant restoration of the B-cell compartment in IgM b -macroself recipients (Fig. 3d,e). Remarkably, homozygous deletion of the Pten gene (Pten fl/fl ; CD19-Cre) completely restored the B-cell compartment, which is comparable to TG-IgM b -macroself chimeras (Fig. 3d,e). We next used the lentiviral expression system to co-express miR- 19 and Pten in HSCs, as illustrated in Fig. 3f, to test whether restoring Pten expression can prevent the break of B-cell central tolerance by miR-19 in the IgM b -macroself recipients. The miR-19-expressing lentivirus encodes GFP, whereas the Pten-expressing lentivirus encodes Ametrine, therefore allowing distinguishing cells transduced by a single lentivirus from those transduced by both. Terminal analysis of the reconstituted mice showed no difference in the total numbers of splenic B cells that escaped the central tolerance checkpoint, whether Pten or its vector control was introduced to the miR-19 overexpression scheme. However, we observed a drastic difference in the compositions of these escaped B cells. In the miR-19 þ control group, the number of B cells expressing miR-19 alone (GFP þ Ametrine À ) was comparable to that of B cells co-expressing miR-19 and control (GFP þ Ametrine þ ). In the miR-19 þ Pten group, the escaped B cells were predominantly those expressing miR-19 alone (GFP þ Ametrine À ), whereas the number of B cells co-expressing miR-19 and Pten (GFP þ Ametrine þ ) was close to the IgM b -macroself background level (Fig. 3g,h). Taken together, these results suggest a miR-19-Pten pathway that regulates B-cell central tolerance.

Impaired B-cell development in the absence of miR-17B92.
A previous study showed that germline deletion of miR-17B92 led to a partial block of early B-cell development at the pro-to pre-B transition and Bim was suggested to be a key functional  target of miR-17B92 in this process 19 . However, another study reported that transgenic mice with ubiquitous overexpression of miR-17, a member of the miR-17B92 cluster, exhibited overall growth retardation and severely reduced numbers of B lineage cells in both the bone marrow and the spleen 34 . Therefore, the cell autonomous function of the miR-17B92 family, as well as the functional contribution of individual miRNAs of this family, in B-cell development remains to be elucidated. To investigate the cell-intrinsic role of miR-17B92 family miRNAs in B-cell development, we generated Mb1-Cre;miR-17B92 fl/fl ; miR-106aB363 À / À ;miR-106bB25 À / À mice (termed Mb1tKO). In these mice, miR-17B92 is deleted specifically in the B-cell lineage by Mb1-Cre, a Cre allele expressed from the earliest stage of B-cell development 35 , whereas the two paralogue clusters, miR-106aB363 and miR-106bB25, are deleted in the germline, resulting in a complete absence of this family in the B-cell lineage. Rluc Pten 3′ UTR (wild type or binding site mutated)   Analysis of Mb1tKO mice revealed a twofold reduction in B-cell numbers in the bone marrow (Fig. 4a). Consistent with the previous study of miR-17B92 germline knockout mice, Mb1tKO mice showed an accumulation of pro-B cells (B220 int cKit þ ) and a reduction of pre-B cells (B220 int CD25 þ ) (Fig. 4b,c). The splenic B-cell number in Mb1tKO mice was also reduced by threefold (Fig. 4d). Apoptosis of pro-B and pre-B cells of Mb1tKO mice was determined by flow cytometry analysis of Annexin V and active caspase 3. We observed an increased apoptotic rate in pre-B cells of Mb1tKO mice when compared with their counterparts in WT mice (Fig. 4e). In conclusion, the miR-17B92 family miRNAs control the pro-to pre-B transition during B-cell development in a cell-autonomous manner and the complete absence of this miRNA family leads to increased apoptosis of pre-B cells.
A central role of miR-17 in early B-cell development. We next analysed which individual miRNA subfamily of the miR-17B92 cluster plays an essential role in the pro-to pre-B-cell transition.
To evaluate the functional contribution of each miRNA subfamily of the miR-17B92 cluster to the regulation of early B cell development, we restored the expression of miR-17B92 miR-NAs, either individually or in combination (Fig. 2a), in Mb1tKO HSCs by lentiviral transduction and determined their ability to undergo the pro-to pre-B-cell transition. A mixed bone marrow reconstitution approach, in which Mb1tKO B cells were competing with their WT counterparts during B-cell development, was performed to highlight the B-cell developmental block of Mb1tKO mice (Fig. 5a). In this experimental setting, the changes in the ratio of virus-transduced CD45.2 þ cells (GFP þ ) versus WT competitors (CD45.1 þ ) during the pro-to pre-B transition provided a measurement for the developmental defect. In the control WT:WT mix group, as no major difference existed between these two populations, the GFP þ /CD45.1 þ ratios remained unchanged during the pro-to pre-B transition, resulting in a pre-B/pro-B ratio close to 1 (Fig. 5b upper panels and Fig. 5c). In contrast, cells derived from the Mb1tKO HSCs, when transduced with control virus, underwent a severe developmental block in competition with WT cells. Therefore, the GFP þ /CD45.1 þ ratio shifted drastically as WT cells became dominant in the pre-B population, resulting in a pre-B/pro-B ratio below 0.2 (Fig. 5b middle panels and Fig. 5c). As expected, when Mb1tKO cells were transduced with virus expressing the FL miR-17B92, they became fully competitive with WT cells and the GFP þ /CD45.1 þ ratio in pre-B cells was comparable to that in pro-B cells, resulting in a pre-B/pro-B ratio close to 1 (Fig. 5b  lower panels and Fig. 5c). Interestingly, among the group of lentiviral vectors expressing the miR-17B92 cluster with one miRNA subfamily deleted, deletion of the miR-17 subfamily almost completely abrogated the ability of this cluster to rescue the pro-to pre-B transition defect of Mb1tKO cells (Fig. 5c and Supplementary Fig. 5). Conversely, lentiviral expression of miR-17 alone significantly restored the pro-to pre-B-cell transition ( Fig. 5c and Supplementary Fig. 5). Deletion of miR-19 or miR-92 subfamily had a minor effect on the rescue and expression of miR-92 and miR-19 family had no effect or minor effect in this process, respectively. Therefore, we conclude that the miR-17 subfamily plays a central role in regulating the pro-to pre-B transition during early B-cell development, with additional support from the miR-19 and miR-92 subfamilies. Our functional dissection of the miR-17B92 cluster reveals a shift of power among these co-expressed miRNAs in two sequential and closely linked stages of B-cell development: although miR-17 plays a central role in controlling pro-to pre-B transition, miR-19 is critical for regulating B-cell central tolerance at the immature B-cell stage.
PTEN, Phlpp2 and Bim do not control early B-cell development. The PI3K pathway has been shown to control early B-cell development 36 . Our results showed that the miR-17B92 family miRNAs suppress the expression of PTEN and Phlpp2 in immature B cells (Fig. 3a) and mature B cells 25 . Moreover, flow cytometry analysis of pAkt in developing B cells of Mb1tKO mice showed decreased Akt phosphorylation in pro-B cells when compared with their WT counterparts (Fig. 6a). Previous studies have also suggested Bim as a key functional target mediating miR-17B92 regulation of B-cell development 19,37 . We found slightly increased protein Bim levels in Mb1tKO pro-B cells compared with WT cells (Fig. 6a). For those reasons, we speculated that miR-17B92 family miRNAs regulate early B-cell development by suppressing the expression of PTEN, Phlpp2 and Bim.
To test whether decreased PTEN and Phlpp2 expression rescues the B-cell development defect of Mb1tKO mice, we generated a panel of mouse strains by introducing Pten and/or Phlpp2 deletion to Mb1tKO mice. As shown in Fig. 6b,c and Supplementary Fig. 6a,b, heterozygous and homozygous deletions of Pten and Phlpp2, either individually or in combination, were not able to rescue the pro-to pre-B transition block in Mb1tKO mice. Interestingly, homozygous deletion of Pten was detrimental to pro-B cells on the Mb1tKO background, as shown by the complete loss of B220 int cKit þ pro-B cells when both Pten alleles were deleted.
Previous studies showed that Mb1-Cre-mediated deletion of Dicer led to a severe block of early B-cell development at the pro-to pre-B transition 37 , which is similar to but stronger than ARTICLE that caused by germline deletion of miR-17B92 (ref. 19), or by the complete deletion of the miR-17B92 family miRNAs specifically in the B-cell lineage (that is, Mb1tKO mice in this study). An Em-Bcl-2 transgene (Bcl2Tg) was able to partially rescue the pro-to pre-B transition block caused by Dicer deletion 37,38 . We therefore tested whether the same Bcl2Tg was able to restore B-cell development in Mb1tKO mice. As shown in Supplementary Fig. 6c,d, although Bcl2Tg did restore the splenic B-cell number, probably by prolonging the survival of mature B cells in the periphery 39 , it did not have any measurable effect on the pro-to pre-B transition block in Mb1tKO mice. Taken together, these results exclude PTEN, Phlpp2 and Bim as major mediators of the regulation of early B-cell development by the miR-17B92 family miRNAs. Thus, other target genes and molecular pathways must play more important roles in this process.

Discussion
The present study revealed critical roles of the miR-17B92 family miRNAs in two sequential events in the development of B cells: To our knowledge, miR-17B92 is the first miRNA cluster that has been discovered to control B-cell tolerance. During early B-cell development, miR-17B92 miRNAs are relatively abundant in pro-and pre-B cells, but their expression goes down drastically when developing B cells transit from pre-B to immature B cells 13,18 . As miR-17B92 is generally thought to play prosurvival and pro-proliferation roles, keeping its expression level low renders immature B cells sensitive to self antigen-induced apoptosis, when the opportunity of receptor editing is exhausted. of the PI3K-Akt pathway. This PI3K-Akt pathway was previously shown to be essential for BCR tonic signalling, which is essential for the survival of mature B cells 31 . Our results demonstrate that this pathway also plays critical roles in supporting the survival of immature B cells.
Several previous studies suggested that the miR-17B92 family miRNAs are essential for the pro-to pre-B transition during early B-cell development, but discrepancy exists among those reports. In a study employing Mb1-Cre-mediated deletion of Dicer, which abolished the expression of all mature miRNAs, B-cell development was severely blocked at the pro-to pre-B transition. miR-17B92-mediated suppression of Bim expression was proposed to be the major underlying mechanism 37 . Indeed, germline deletion of miR-17B92 caused a partial B-cell development block at the pro-to pre-B transition and the Bim protein level was higher in miR-17B92-deficient pre-B cells 19 . The pro-to pre-B transition block caused by Dicer deletion was partly rescued by a Em-Bcl2 transgene, further supporting an important role of Bim in mediating miR-17B92 control of early B-cell development 37 . However, another study reported that transgenic mice with ubiquitous overexpression of miR-17, a member of the miR-17B92 cluster, exhibited a severe reduction in the number of B lineage cells in both the bone marrow and the spleen 34 . The latter study casted doubt on the cell autonomous function of the miR-17B92 family miRNAs in controlling early B-cell development. In the present study, we generated and analysed mutant mice harbouring B-cell-specific deletion of the miR-17B92 family miRNAs (Mb1tKO mice). These mutant mice exhibited a pro-to pre-B transition block that is similar to that caused by Mb1-Cre-mediated deletion of Dicer and germline deletion of miR-17B92 (refs 19,37). Therefore, the miR-17B92 family miRNAs do play a cell-intrinsic role in controlling early B-cell development. Our functional dissection experiments revealed that, among the six miRNAs encoded by miR-17B92, miR-17 plays a central role in controlling the pro-to pre-B-cell transition, whereas miR-19 and miR-92 probably synergize with miR-17 to exert this function. Surprisingly, although PTEN is an important mediator of the miR-17B92 function in immature B (shown in this study) and mature B cells 25 , it does not seem to play critical roles in miR-17B92 regulation of early B-cell development. In addition, our results excluded an important role of miR-17B92 regulation of Bim in controlling early B-lymphocyte development, which was suggested by previous studies 19,37 . Thus, other unknown molecular pathways must mediate miR-17B92 family miRNA control of the pro-to pre-B-cell transition. The elucidation of these pathways warrants future investigation. Our experimental approach provided a unique opportunity to investigate how individual members of a heterogeneous miRNA cluster work together to exert their functions in a dynamic manner. Previous studies of miR-17B92 highlighted the complexity of this issue. In a c-Myc-driven lymphomagenesis model and a Notch-driven leukemogenesis model, miR-19 was shown to be the key oncogenic member of the miR-17B92 cluster [40][41][42] . A recent study reported that miR-92 negatively regulated the oncogenic cooperation between miR-19 and c-Myc, suggesting functional antagonism among members of this cluster 43 . In another study employing retroviral overexpression of miR-17B92 and its individual members in HSCs, overexpression of miR-19a and miR-92a resulted in B-cell hyperplasia and erythroleukemia, respectively. Interestingly, miR-92a-induced erythroleukemia development was abrogated by co-overexpressing miR-17, suggesting functional antagonism between miR-17 and miR-92a 44 . Our study focused on functional dissection of the miR-17B92 cluster at two closely linked developmental stages of the B-cell lineage. While miR-17 plays a central role in controlling the pro-to pre-B transition, miR-19 is critical for regulating B-cell central tolerance at the immature B-cell stage. In both processes, other miRNAs in this cluster play supporting, instead of antagonistic, roles (that is, miR-19 and miR-92 contribute to the regulation of pro-to pre-B transition, whereas miR-17 partly controls B-cell central tolerance). Therefore, there is a shift of power among members of the miR-17B92 cluster at these two closely linked developmental stages of the B-cell lineage. It is conceivable that individual members of a heterogeneous miRNA cluster regulate different, yet overlapping, sets of target genes and might have different effects on the molecular pathways controlled by the cluster. It is also possible that at different developmental stages of a single cell lineage, or in different cellular contexts, the functional relevance of these molecular pathways may differ to a large degree. That would explain the differential contribution of individual members of a miRNA cluster to its roles in different cellular contexts. Thus, miR-19 plays a critical role in regulating B-cell central tolerance through suppressing the expression of Pten, whereas miR-17 controls the pro-to pre-B transition through other unknown molecular pathways.
In summary, our study identified critical roles of miR-17B92 in two closely linked developmental stages of the B-cell lineage, pro-to pre-B transition and immature B cells. Interestingly, different members of this cluster play central roles in these two processes by regulating different molecular pathways. These findings illustrate dynamic functional specificity of individual members of a heterogeneous miRNA cluster in B-cell development.
Lentiviral vector generation and virus production. Lentiviral plasmids pWPXLd, pMD2.G and psPAX2 were gifts from Dr Didier Trono laboratory (Addgene plasmids #12258, 12259 and 12260). The expression vectors for the FL miR-17B92, miRNA subfamily deletion mutants and individual miRNA components were constructed by stepwise cloning of various combinations of individual miRNAs (precursor sequences with 30-50 nucleotide flanking regions) into pWPXLd. The Ametrine expression vectors were constructed by replacing the GFP cassette of pWPXLd with the IRES-Ametrine 1.1, a violet-excitable yellowfluorescing GFP variant. The Pten-coding sequence was subsequently cloned into this vector for producing Pten-expressing lentiviruses. Recombinant lentivirus was produced in 293T cells by co-transfecting the expression vectors with packaging constructs pMD2.G and psPAX2 by the calcium phosphate method. The virus-containing supernatant was harvested 48 h after transfection, filtered and concentrated using PEG-it Virus Precipitation Solution (System Biosciences).
Bone marrow transduction and reconstitution experiments. HSCs were purified by enrichment of Sca-1 þ cells using magnetic cell separation followed by isolation of bone marrow side population cells using FACS based on the capacity of these cells to actively exclude the vital dye Hoechst 33342 (ref. 47). Briefly, bone marrow cells were prepared from the femurs and tibias. After red blood cell lysis, Sca-1 þ cells were enriched by anti-Sca-1 MACS MicroBeads according to manufacturer's instruction (Miltenyi Biotec). Sca-1 þ cells were re-suspended in HBSS containing 2% fetal bovine serum (FBS), penicillin-streptomycin and 10 mM Hepes buffer, and stained with 8.8 mg ml À 1 Hoechst 33342 (Invitrogen) at 5 Â 10 6 cells per ml. After incubation at 37°C for 90 min, the Hoechst 33342-negative side population cells were isolated using a FACSAria cell sorter (BD Biosciences).
Recipient mice were irradiated with two doses of 5 Gy, 3 h apart and subjected to bone marrow transplantation 2 h later by tail vein injection with: (i) 5 Â 10 6 bone marrow cells from the indicated donors or (ii) 10-15 Â 10 3 lentivirally transduced HSCs mixed with 1 Â 10 6 congenic, unfractionated bone marrow cells. Recipient mice were maintained with antibiotics-containing food or water for 30 days before switching to normal food and analysed at indicated time points.
Cell surface staining and flow cytometry analysis were performed following established protocols. Intracellular staining was performed following fixation and permeabilization using BD Phosflow Perm Buffer II. Stained cells were analysed on FACSCalibur or LSR II (BD Biosciences). Data were analysed with FlowJo software (Tree Star). Antibodies and reagents with the following specificities were used for staining: anti-B220 ( In vitro immature B-cell culture and enrichment. Total bone marrow cells were isolated from the femur and tibia. B lineage cells were enriched by magnetic cell sorting using anti-CD19 MACS MicroBeads (Miltenyi Biotech). CD19 þ cells were cultured in Advanced DMEM-reduced serum medium (Gibco) supplemented with 10% FBS, 55 mM b-mercaptoethanol, penicillin-streptomycin, 4 mM glutamine and 5 ng ml À 1 recombinant human IL-7 for 5 days. Expanded immature B cells were enriched by incubating with biotin anti-IgM antibodies (RMM-1, 406503; dilution 1/500; BioLegend) at 4°C followed by magnetic cell sorting using Streptavidin MicroBeads (Miltenyi Biotech).
Immunization and ELISA assay. Antigens for immunization were prepared by mixing NP 36 -CGG (Biosearch Technologies) dissolved in PBS and 10% KAl(SO4) 2 at 1:1 ratio and adjusting pH to 7, to form precipitate. Five micrograms of NP-CGG precipitated in alum was injected intraperitoneally for NP-specific antibody responses.
Northern blotting. Total RNA was extracted from miRNA-encoding lentivirustransduced HeLa cells using TRIzol Reagent (Life Technologies) following the manufacturer's instructions. Ten micrograms of total RNA was used to detect miRNAs. DNA oligonucleotides antisense to mature miRNAs were used as probes. U6 small nuclear RNA was used as internal control for normalization. Northern blotting results were acquired on a Typhoon 9410 imager (GE Healthcare) and analysed using the ImageQuant software.
Luciferase reporter assays. The 3 0 -UTR fragments of Pten and Phlpp2 containing predicted miR-19-binding sites (Supplementary Data set 1) were cloned into the psiCHECK2 vector (Promega), to generate Pten and Phlpp2 reporter plasmids. The miR-19-binding site was subsequently mutated to generate the Pten-3 0 -UTR-mut construct. HeLa cells were plated into 24-well plates at 4 Â 10 4 cells per well 24 h before transfection with 0.1 mg reporter plasmid and 0.4 mg miR-19-expressing pWPXLd plasmid using the Lipofectamine 2000 transfection reagent (Invitrogen). Luciferase assays were performed 48 h after transfection using the Dual-Luciferase Reporter Assay System (Promega) following the manufacturer's protocol. The Rluc activity was normalized by the firefly luciferase activity (Luc) and expression is presented as Rluc/Luc ratio, which was arbitrarily set as 1 for the empty pWPXLd plasmid for each reporter.
Sample preparation and sequencing of mouse antibody libraries were performed based on a similar procedure that has been described previously 49 . Briefly, total RNA was extracted from 10 to 20 million in vitro-expanded B cells into 30 ml of water with RNeasy Mini Kit (Qiagen). For unbiased antibody repertoire analysis, 5 0 -rapid amplification of cloned/complementary DNA ends (RACE) was performed with SMARTer RACE cDNA Amplification Kit (Clontech). The immunoglobulin PCRs were set up with Platinum Taq High-Fidelity DNA Polymerase (Life Technologies) in a total volume of 50 ml, with 5 ml of cDNA as template, 1 ml of 5 0 -RACE primer and 1 ml of 10 mM reverse primer. The 5 0 -RACE primer contained a PGM P1 adaptor, whereas the reverse primer contained a PGM A adaptor. We adapted the 3 0 -Cg1-3 inner primers and 3 0 -Cm inner primers, 3 0 -mCk outer primer and 3 0 -Cml outer primer as reverse primers for 5 0 -RACE PCR processing of the H, k and l chains, respectively 50 . Twenty-five cycles of PCRs were performed and the expected PCR products (500-600 bp) were gel purified (Qiagen). The antibody heavy (H) and light (k and l) chain libraries were quantified using Qubit 2.0 Fluorometer with Qubit dsDNA HS Assay Kit and then used at a ratio of 1:1:1 for all the PGM sequencing experiments. The dilution factor required for Ion Torrent PGM template preparation was determined such that the final concentration was 50 pM. Template preparation was performed with the Isothermal Amplification Kit obtained from the Early Access Program. Before PGM sequencing, quality control of the template was determined by the Qubit 2.0 Fluorometer with the Ion Sphere Quality Control Kit. Sequencing was performed on the Ion Torrent PGM sequencer with the PGM Hi-Q 400 Kit using an Ion 314 v2 chip for a total of 1,100 nucleotide flows. Raw data were processed without the 3 0 -end trimming in base calling, to extend the read length.
Bioinformatics analysis of antibody sequence data followed the human antibodyomics pipeline 49,[51][52][53][54][55] . Specifically, the mouse H, k and l germline genes from IMGT (http://www.imgt.org) including the V, D and J segments were incorporated into the pipeline where such information is required for gene assignment (step 2), error correction (step 3) and determination of H/LCDR3 and variable region boundaries (step 5). For heavy chains, 313 VH genes along with 39 DH and 9 JH genes were compiled into three libraries, whereas for light chains, 151 VK genes along with 5 JK genes and 19 VL genes with 7 JL genes were used in library construction, respectively. The mouse antibodyomics pipeline consists of five consecutive steps. Given a data set of NGS-derived mouse antibody sequences, each sequence was (1) reformatted and labelled with a unique index number; (2) assigned to V, D (for heavy chain only) and J gene families using the current rhesus macaque germline gene database and an in-house implementation of IgBLAST, and sequences with E-value410 À 3 for V gene assignment were removed from the data set; (3) subjected to a template-based error-correction procedure, in which insertion and deletion (indel) errors in the V gene segment were detected based on the alignment to their respective germline gene sequences. It is noteworthy that only indels of less than three nucleotides were corrected; (4) compared with the template antibody sequences at both the nucleotide level and the amino acid level using a global alignment module in CLUSTALW2 (ref. 56); (5) subjected to a multiple sequence alignment-based procedure to determine the complementarity determining region 3 (CDR3), which was further compared with the template CDR3 sequences at the nucleotide level and to determine the sequence boundary of the V(D)J coding region. After FL variable region sequences were obtained, a bioinformatics filter was applied to detect and remove erroneous sequences that may contain swapped gene segments due to PCR errors. Specifically, a FL read would be removed if the V gene alignment was o220 bp. The processed and annotated antibody chain sequences were then subjected to germline gene frequency analysis.
Data availability. The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information files, or from the authors upon a reasonable request. The IGHV deep-sequencing data sets are available in SRA database with the accession code SRP075587.