PAX5A and PAX5B isoforms are both efficient to drive B cell differentiation

Pax5 is the guardian of the B cell identity since it primes or enhances the expression of B cell specific genes and concomitantly represses the expression of B cell inappropriate genes. The tight regulation of Pax5 is therefore required for an efficient B cell differentiation. A defect in its dosage can translate into immunodeficiency or malignant disorders such as leukemia or lymphoma. Pax5 is expressed from two different promoters encoding two isoforms that only differ in the sequence of their first alternative exon. Very little is known regarding the role of the two isoforms during B cell differentiation and the regulation of their expression. Our work aims to characterize the mechanisms of regulation of the expression balance of these two isoforms and their implication in the B cell differentiation process using murine ex vivo analyses. We show that these two isoforms are differentially regulated but have equivalent function during early B cell differentiation and may have functional differences after B cell activation. The tight control of their expression may thus reflect a way to finely tune Pax5 dosage during B cell differentiation process.


INTRODUCTION
The commitment of hematopoietic stem cells to each cell lineage is strictly controlled. Specific extracellular stimuli and transcription factors play important roles in the development of B cells from hematopoietic stem cells. The transcription factors SPI1 (PU.1) and IKZF1 (Ikaros) work together during the early developmental stages, while TCF3 (E2A), Early B cell Factor 1 (EBF1) and Paired box 5 (PAX5) are crucial from pro-B cells to mature B cell stages PAX5 also enhances the transcription of B cell specific genes and participates in the chromatin-remodeling of the immunoglobulin heavy chain (IGH) locus, ensuring its contraction during VDJ recombination [8]. At later stages, PAX5 regulates the IGH 3ʹ regulatory region (3ʹRR). The 3ʹRR is a 30 kb-long cis-acting regulation element of the immunoglobulin heavy chain (IGH) locus containing four enhancers in mice (hs1,2, hs3a, hs3b and hs4) with a strict B lineage specificity. They have been implicated in the late stages of B cell differentiation with a crucial role in class switch recombination (CSR) and somatic hypermutation (SHM) [9][10][11][12].
Pax5 homozygous inactivation in mouse leads to a blockade at the pro-B cell stage [6]. Pax5-deficient pro-B cells are able to transdifferentiate into other cell types, such as T, natural killer, and dendritic cells [1,13] while mature B cells can dedifferentiate into lymphoid precursor cells upon Pax5 loss even at late stages of B cell differentiation as shown by Pax5 conditional inactivation [14].
In vertebrates, PAX5 expression is controlled by two distinct promoters: a distal P1a and a proximal P1b [15] which initiate transcription from two alternative 5′ first exons (exons 1A and 1B respectively) leading to the expression of two isoforms, PAX5A and PAX5B. The only structural difference between PAX5A and PAX5B consists in a short different N-terminal sequence. Exon 1A encodes 15 amino acids that differ totally from the 14 amino acids encoded by the exon 1B sequence [16]. Despite high sequence similarities between these two isoforms, they show different patterns of expression: PAX5B is transcribed in B cells, central nervous system and testis, while PAX5A, whose product is also named B cell-specific activator protein (BSAP) is restricted to the B cell lymphoid lineage [4]. However their respective roles in B cell differentiation have never been determined. Our work aims to detail the mechanisms that control the expression of PAX5A and PAX5B isoforms along B cell development and their effect on B cell differentiation.

Pax5 expression in B cell differentiation is independent of adjacent genes
The murine Pax5 gene encompasses a region of 392 kb of chromosome 4 from the end of its upstream neighbor gene, Melk, to the start of its downstream neighbor gene, Zcchc7 ( Figure 1A). Pax5 has a reverse orientation compared to its two neighbors, from telomere to centromere ( Figure 1A). The human PAX5 gene has a similar organization covering a slightly larger region of 444 kb on chromosome 9. In order to clarify the transcriptional activities within the Pax5 locus, quantitative RT-PCR (QPCR) was performed to measure the overall expression of Pax5 transcripts and its neighboring genes (Melk and Zcchc7), together with Abl1 as a widely expressed control gene and Ebf1 as a transcriptional target of Pax5. Their expression were measured in a series of murine B cell lines representing different stages of B cell differentiation (from the less to the most differentiated: Ba/F3, 70Z3, 38B9, 18.81, A20 and WEHI-231) along with murine primary tissues (T and B cells, Figure 1B). Since Pax5 expression is regulated by Ebf1, Pax5 expression is highly correlated to the expression of Ebf1, as expected (Pearson correlation, r 2 = 0.85). In contrast, expression of Pax5 is independent of the expression of its two neighboring genes, Melk and Zcchc7 (Pearson correlation, r 2 = 0.40 and r 2 = 0.54 respectively), suggesting that the regulatory elements of Pax5 are not shared by Melk and Zcchc7.

Pax5 isoforms are differentially expressed during B cell differentiation
Two major 5ʹ isoforms of Pax5 are expressed during B cell differentiation. Pax5A expression is driven by the promoter 1A and Pax5B by the promoter 1B [using alternative first exons (1A and 1B respectively, Figure 1A)]. We detailed the expression pattern of the two Pax5 isoforms during murine B cell differentiation using specific primers of these two isoforms on sorted B cell subsets from bone marrow. Pax5B has a low expression which does not vary during B cell differentiation. In contrast, Pax5A expression is strongly modulated during B cell differentiation with a higher expression in immature B cells (Figure 2A, left panel).

Expression of Pax5 isoforms is correlated with the histone H3 acetylation of their respective promoter
We defined the expression pattern of Pax5 isoforms on B cell lines. The 38B9, 18.81, A20 cell lines expressed both Pax5 isoforms. The early blocked 70Z3 B cell line expressed only Pax5A isoform. The Ba/F3 cell line did not express any of the two isoforms ( Figure 2A, right panel).
Using the fact that B cell lines differentially expressed Pax5 isoforms, we detailed the transcriptional regulation within the 392 kb genomic region based on the acetylation of histone H3, a major mark of active regulatory elements [17] at low resolution over the 50 most conserved regions in vertebrates. We observed a peak of acetylation covering Pax5 promoters A and B in pro-B cells and the B cell lines expressing at least Pax5A in contrast to Ba/F3 ( Figure 2B). We increased the resolution of the ChIP analysis to 1 kb over this region on five B cell lines ( Figure 2C). The three cell lines expressing both Pax5 isoforms (38B9, 18.81 and A20) exhibited two acetylation peaks located upstream of exon 1A (defined as the location origin, 0 kb) and 1B (+7 kb). 70Z3 expressing only Pax5A isoform had peaks at 0 kb and -4kb, but none at the +7 kb while Ba/F3 displayed no acetylation peak ( Figure 2C). www.oncotarget.com Decker et al. identified an enhancer region located in the intron 5-6 of Pax5 [24]. We analyzed the acetylation status of this region and showed that the acetylation is weak compared to the promoter regions. Interestingly, only the cell lines expressing both isoforms have an H3 acetylation mark on the enhancer region ( Figure 2D).

Both Pax5 isoforms restore early B cell differentiation
We took advantage of the Pax5 -/cells which do not express any of the two isoforms and are subsequently blocked at a pro-B cell stage but still proliferating in presence of IL7. PAX5 isoforms were individually transduced in Pax5 -/pro-B cells and assessed for their ability to reinitiate B cell differentiation, using the expression of Cd19, a specific target of PAX5, and the appearance of IgM pos IgΚ pos B cells as hallmarks of B cell progression. Cells infected with the empty vector (MIE, only expressing eGFP) did not induce expression of Cd19. In contrast, both isoforms induced the expression of surface Cd19 in presence or absence of IL7 ( Figure 3A, respectively upper and lower panels). Furthermore, both isoforms were also able to initiate the expression of IgM and IgΚ at the cell surface after removal of IL7 ( Figure 3B, lower panel).
Transduced Pax5 -/cells were seeded at the same concentration and cultured for 3 days in presence of IL7. Although the percentage of Cd19 pos cells is the same after transduction with PAX5A or PAX5B ( Figure 3C, left panel), the absolute number of pro-B cells transduced by PAX5 isoforms was increased compared to the MIE with the most potent proliferation being driven by PAX5B ( Figure 3C, right panel). Furthermore, after 10 days of culture in presence of IL7, Pax5 -/pro-B cells complemented with PAX5B have a quarter less in spontaneous cell death rate compared with PAX5A ( Figure 3D, left panel). This difference of cell survival may explain the highest percentage of IgM pos IgΚ pos B cells after IL7 withdrawal with PAX5B complementation ( Figure 3B).
We also stably transduced Ba/F3 cells with either MIE, PAX5A or PAX5B and stressed the cells by serum deprivation for 6 hours. Ba/F3 cells are less sensitive to serum deprivation when they express PAX5B compared to PAX5A whereas no difference was observed in 10% serum condition ( Figure 3D, right panel). This difference is not linked to a difference of PAX5 isoform stability since half-life of the two proteins are comparable ( Figure 3E). Consequently, the alternative use of PAX5 exon 1 has no qualitative effect on early B cell differentiation but may have an impact on cell death.  Both PAX5A and PAX5B enhance the expression of B cell specific genes such as Cd19 and repress the expression of non-B cell specific markers such as Cd7 (a T cell marker) or Gp49a (a myeloid marker) ( Figure 4D). Furthermore, out of the 24,184 genes, no expression differs more than 5-fold between PAX5A and PAX5B target genes ( Figure 4C). By performing a Gene Ontology (GO) analysis [18,19] on the genes whose expression is the most discriminant, we found a significant enrichment in genes implicated in the regulation of apoptosis (Supplementary Figure 1A, 1B). These data strongly suggest that the two PAX5 isoforms globally induce B cell differentiation by activating the same targets during early B cell differentiation and differences in their target gene expression may explain the differences observed in cell death rate and cell growth ( Figure 3C and 3D).

Pax5 isoforms have a distinct pattern of expression during late B cell differentiation
We also investigated the functional specificities of PAX5 isoforms during late B cell differentiation. We quantified Pax5 isoforms in murine sorted splenic resting B cells (Cd43 neg B220 pos ), in vitro LPS-activated B cells (B220 pos Cd138 neg ) and plasmablasts (B220 pos Cd138 pos ). Overall, Pax5 expression is decreased after B cell   Figure 5A). In details, Pax5A was the most expressed isoform and downregulated in activated B cells and plasmablasts. In contrast, Pax5B was significantly upregulated in activated B cells and almost undetectable in plasmablasts. Therefore, the expression patterns of the two isoforms are distinct during late B cell differentiation.

Pax5B is inactive in plasma cells
The differential regulation of Pax5 isoforms at late B cell differentiation stages may reflect different properties of these isoforms. To assess this possibility, we evaluated the transactivation of Cd19 at an early B cell stage using transduced Ba/F3; both PAX5A and PAX5B can efficiently transactivate Cd19 ( Figure 5B). We also evaluated the transactivation potential of PAX5A and PAX5B at later stage of B cell differentiation using S194, a plasmacytoma cell line, or in a non-lymphoid context using 293T, a human kidney cell line, using a luciferase reporter system containing the Cd19 promoter region. Compared to AID, a negative control, both PAX5A and PAX5B activate the Cd19 promoter ( Figure 5C). The activity of PAX5B is significantly lower than PAX5A. In S194, only PAX5A displays transcriptional activity suggesting that PAX5B transactivation capacity on Cd19 promoter is altered in those cells ( Figure 5C). To characterize the binding of Pax5 on the Cd19 promoter at late stages of the B cell differentiation, we precipitated chromatin (ChIP) using an anti-Pax5 antibody in splenic B cells ( Figure 5D). We detected a specific binding of Pax5 to the Cd19 promoter sequence in resting B cells compared to the amplification of an irrelevant sequence (Upf1 intronic sequence) devoid of Pax5 binding sites ( Figure 5D, left panel). After 4 days of LPS stimulation, we could detect the same enrichment of Pax5 on Cd19 promoter ( Figure 5D, right panel).

Neither PAX5A nor PAX5B are able to silence the 3ʹRR in plasma cells
The 3ʹRR is an important region regulating the expression of the heavy immunoglobulin chain at late B cell differentiation stages. It has been reported that Pax5 could bind hypersensitive sites (hs)1,2 (3ʹαE) and hs4 elements ( [20,21]). A Pax5-specific ChIP using an antibody that does not discriminate between the two isoforms, identified a 3.8-fold enrichment at the hs4 element site in resting B cells ( Figure 5E, left panel). Upon activation, Pax5 is enriched both at the hs4 and to a much lesser extend to hs1,2 sites (6-fold enrichment for hs4 and a 2.4-fold for hs1,2; Figure 5E, right panel).
We then analyzed whether both PAX5 isoforms have a similar transactivation activity on the 3ʹRR in S194 plasma cells ( Figure 5F). PAX5 isoforms are physiologically downregulated in plasma cells, in contrast the 3ʹRR activity is increased. We evaluated the effect of the PAX5 isoforms on the 3ʹRR activity using a reporter assay depending on the VH promoter (pVH) active in the B cell lineage. Its transcription level is upregulated by the addition of a synthetic 3ʹRR containing the four enhancers in their palindromic structure (hs3a, hs1.2, hs3b, hs4). We did not observe any significant change upon the addition of both PAX5 isoforms, suggesting that 3ʹRR is insensitive to Pax5 ( Figure 5F) despite the accumulation of Pax5 at the 3ʹRR after activation of B cells ( Figure 5E, right panel).

DISCUSSION
PAX5 is a major regulator of B cell differentiation promoting the expression of B cell specific genes and repressing commitment to other cell lineages. Consequently, to allow an efficient B cell differentiation, the expression of PAX5 has to be tightly controlled. An alteration of PAX5 dosage induced B cell malignancies: its overexpression, secondary to the IGH-PAX5 rearrangement, is associated with B cell lymphomas [16] and loss of heterozygosity is associated with B cell progenitor acute lymphoblastic leukemia [22,23]. The regulation of PAX5 expression, in vertebrates, is complex with two 5ʹ isoforms, PAX5A and PAX5B, initiated from two different promoters. These isoforms differ only by the sequence of their first coding exon encoding no specific protein motif. PAX5B isoform is related to PAX2, since the sequence encoded by their first exon are 65% homologous and is present as early as in the Fugu ( Table 1). The PAX5A isoform appears later, with mammals and by extension B cells.
Very little is known about the differential regulation of these two isoforms and their functional differences. The aim of this study was to determine if the presence of these two isoforms translate a functional difference or are merely a differential regulation of expression. In B cells, PAX5 isoforms are differentially expressed suggesting that different sets of transcription factors enhance or repress the activity of these promoters. ChIP analyses on cell lines and pro-B cells confirm a correlation of histone H3 membrane markers (the percentage of IgM + /Igk + cells is indicated on the right panel graph) in two different conditions: with (+IL7, upper line) or without IL7 (-IL7, lower line) for three days. The first column show Pax5 -/cells infected with the eGFP-expressing vector (MIE), the second column with PAX5A retroviral construct and the third column with the PAX5B retroviral construct. The mean and the standard deviation of percentage of each population compared to the parental gate are indicated, * p < 0.05, **** p < 0.001. (C) Mean and standard deviation of percentage of Cd19 + cells (left panel); Absolute number of GFP + Pax5 -/pro-B complemented with eGFP (MIE), PAX5A or PAX5B after 3 days of culture for three independent infections, **** p < 0.001 (right panel). (D) Percentage of 7-AAD pos AnnexinV pos cells after 5 or 10 days in presence of IL7 (left panel) for three independent infections with PAX5A-or PAX5B-expressing vectors. **** p < 0.001. (E) Quantification of PAX5A and PAX5B proteins by Western-blot using an anti-Pax5 antibody after cycloheximide treatment for the indicated period of time on Ba/F3 infected cells demonstrating a similar half-life. acetylation on both promoter regions and expression of the two isoforms. Interestingly, we also detected histone H3 acetylation on the enhancer region [24] only when the two isoforms are expressed.
Functionally, PAX5A and PAX5B can act similarly in early B cell differentiation. Both are able to complement the invalidation of Pax5 in Pax5 -/cells and resume ex vivo the B cell differentiation until the Cd19 pos /mIgΚ pos stage demonstrating a similar capacity to transactivate or repress target genes during this process. This is shown by the transcriptome since no gene expression differs more than five times in Pax5 -/complemented by PAX5A or PAX5B. Nonetheless, we detected differences between the two isoforms. PAX5B confers a growth advantage of the rescued Pax5 -/cells compared to PAX5A-rescued cells. The physiological up-regulation of Pax5B during B cell activation may prevent apoptosis of activated B cells since suppression of PAX5B increases susceptibility of human acute lymphocytic leukemia REH cells to apoptotic death [25]. Interestingly GO term analysis on discriminant genes for which the expression is at least 1.5 fold higher for PAX5A or PAX5B condition shows that they are involved in cell death pathways.
As described by Revilla-i-Domingo et al., although Pax5 interacts with a similarly high number of genomic binding sites defining 8,000 target genes in pro-B and mature B cells, only 13 target genes are commonly activated and 18 target genes are commonly repressed in pro-B and mature B cells (cited from [26]) suggesting that Pax5A and Pax5B could have different transcriptional activities in early and late B cell differentiation stages.
At later stages of B cell differentiation, we observed that Pax5B and Pax5A are expressed simultaneously in primary B cells at different levels. After B cell activation,  (D) ChIP was performed on Cd19 promoter (pCD19) and Upf1 intron 9 (as a PAX5 non-target) and using an irrelevant IgG or an anti-PAX5 antibody. Results are expressed as mean of fold change compared to IgG condition ± SEM of independent experiments. (E) ChIP was performed on the hypersensitive sites (hs)1,2 and hs4 elements of the 3ʹRR in resting splenic B cells or activated B cells (left and right panel respectively) using an irrelevant IgG or a anti-PAX5 antibody. Results are expressed as mean ± SEM of independent experiments. (F) A Luciferase expression B cell specific vector (pVH-luc) containing the IGH 3ʹRR minilocus (+3ʹRR) or not (-3ʹRR) was used in a reporter assay to determine the regulatory function of PAX5A and PAX5B in S194 cells. Results are expressed as mean of fold change compared to IgG condition ± SEM of independent experiments. Ctl: negative control, * p < 0.05; ** p < 0.01; *** p < 0,001 (unpaired two-tailed t test). www.oncotarget.com they exhibit a distinct pattern of expression: Pax5A decreases in activated B cells and plasmablasts while Pax5B increases in activated B cells and is virtually absent in plasmablasts.
PAX5A is able to transactivate Cd19 promoter in both early (Ba/F3) and late B cells (S194) and in a nonlymphoid 293T whereas PAX5B is able to transactivate Cd19 only at an early stage of B cell differentiation and in a non-lymphoid context. These results point out that PAX5B isoform cannot transactivate the Cd19 promoter in a late lymphoid context (S194) or may function in a different context as transcriptional repressor as proposed by Robichaud et al. The authors have shown that PAX5B negatively regulates CD19 expression in REH cell line since the specific suppression of PAX5B expression using ribozymes leads to an increase of CD19 expression [25].
PAX5 has been described as a negative regulator of the 3ʹRR activity. Nonetheless, we observed an increased binding of PAX5 to the 3ʹRR upon B cell activation. The accumulation of PAX5 binding on the 3ʹRR in activated B cells does not seem to reflect a role of PAX5 in the 3ʹRR activity regulation but rather a function in chromatin remodeling permitting accessibility to the IGH locus. Recently, it has been shown that PAX5 recruits PTIP, a molecule involved in class switch recombination [27,28] which induces H3K4me3 and stabilizes the PAX5 binding on the IGH locus [28]. Alternatively, this could be explained by the fact that the activation of B cells results in the induction or increase of the expression of other factors counteracting PAX5 activity.
PAX5 isoforms are physiologically downregulated in plasma cells in contrast to the 3ʹRR activity which is strongly increased. We investigated whether these isoforms could modify the 3ʹRR activity. Our data showed that 3ʹRR is insensitive to a PAX5A or PAX5B regulation in the S194 plasmacytoma cell line, in agreement to the observation that an hs1,2-dependent transgene is not inhibited whereas Pax5 expression is still detected in activated B cells [29]. Consequently, repression of the 3ʹRR activity by Pax5 [20,30,31] is likely more complex.
Our results reveal only discrete differences of action between PAX5A and PAX5B functions and demonstrate that these two isoforms may be interchangeable in early B cell differentiation. Their very tight regulation may reflect different binding of alternative transcription factors according to the cellular context (testis, brain, B cells) leading to a fine control of the Pax5 dosage which is required for a zero-default B cell differentiation.

Quantitative RT-PCR
RNA was isolated using the AllPrep™ DNA/RNA Micro Kit (QIAGEN) and cDNA was synthesized using SuperScript ® VILO™ cDNA Synthesis Kit (Invitrogen™) according to the manufacturer's instructions. Quantitative PCR were performed on a LightCycler ® 480 II System (Roche). Quantitative SYBR Green PCR was performed to quantify relative mouse Cd19 cDNA expression in transduced Ba/F3 using LightCycler ® 480 SYBR Green I Master (Roche Diagnostics GmbH) according to the manufacturer's instructions, and primers 10 µM primers for Cd19 (forward: 5ʹ-AAGGAACAGGTCCTCTGGGA-3ʹ and reverse: 5ʹ-GATTCAAACTGCTCCCCCGA-3ʹ) and 10 µM primers for Abl1 (forward: 5ʹ-GGAGAA GGTCTACGAGCTCAT-3ʹ and reverse: 5ʹ-ATCTGAGATA CTGGATTCCTGGA-3ʹ). Relative gene expression in cell lines or sorted bone marrow B cells was quantified using LightCycler ® 480 Probes Master (Roche Diagnostics GmbH) and the following TaqMan gene expression assays from Applied Biosystems™ Life Technologies: Mm00802038_ m1 for Abl1, Mm00487680_m1 for Melk, Mm00435501_ m1 for Pax5 (both isoforms), Mm00621713_m1 for Zcchc7, Mm00395519_m1 for Ebf1, Mm00435494 for specific expression of Pax5A. Primers/TaqMan probe mix specific for PAX5B transcripts was synthetized by Applied Biosystems (5ʹFAM-ATG CAT AGA CAT GGA GGA GTG AAT CAG CTT G-Tamra3ʹ as probe; 5ʹ-CAC TGT AAG CAC GAC CCG TT-3ʹ as forward primer; 5ʹ-AGT GGC CGT CCA TTC ACA AA-3ʹ as reverse primer). All QPCR programs were carried out as follows in a 20 µl volume: 5 min at 95° C, followed by 45 cycles of 10 s at 95° C, 10 s at annealing temperature of 60° C and 10 s at 72° C. All signals were quantified using either the ΔCt method or the Relative Quantification (ΔΔCt) study both including a normalization to the ΔCt values of Abl1 gene expression levels. Data were analyzed using the LC480 software (Roche Diagnostics) followed by statistical analyses.

Retroviral particles production
Full-length human PAX5A and PAX5B cDNA were amplified, sequenced and inserted into the retroviral vector pMSCV-IRES-eGFP (MIE), a bicistronic vector allowing the expression of eGFP under the control of the long terminal repeat (LTR) promoter. Retroviral supernatants were produced by Phoenix retrovirus producer line (Orbigen) using Lipofectamine 2000 (Invitrogen) reagent for transfection and proceeding according to the manufacturer's instructions. After overnight incubation of the transfection mix, cells were treated with 10 mM sodium butyrate (Millipore) and washed with PBS before pro-B cell medium was added. Viral supernatants were harvested after 24 h of incubation at 32° C, passed through a 0.45 µm filter, aliquoted and frozen at -80° C.
The pro-B cells, cultured on irradiated OP9 cells in the presence of 2 ng/ml of IL7, are infected during exponential growth. Retroviral supernatant was diluted 1:3, supplemented with 4 µg/ml polybrene (Sigma) and 2 ng/ml of IL7 (Peprotech); spinoculation was performed at 1000 × g at 32° C for 90 min. After centrifugation, 1ml of fresh medium supplemented with IL7 was added and the cells were incubated overnight at 37° C. After infected pro-B cell expansion for few days, cells were washed out of IL-7 and plated in IMDM, 5% FBS, 0.05 mM β-mercaptoethanol and 2 mM L-glutamine on irradiated OP9-derived stromal cells with or without IL7. At day 2.5, single-cell suspensions were stained using standard protocols for flow cytometry using the antibodies listed above. Dead cells were excluded from FACS analysis as 7-amino-actinomycin D (7-AAD; BD Biosciences #559763)) positive (BD Biosciences). Stained cells were analyzed with a FACS Calibur flow cytometer (BD). Data were analyzed with the BD FACSDiva™ software (BD Biosciences) or FlowJo (TreeStar, Ashland, OR). All antibodies are rat monoclonal antibodies from BD biosciences: anti-B220 (clone RA3-6B2) conjugated with pacific blue, anti-CD19 (clone 1D3) conjugated with Alexa Fluor 700, anti-mouse Igk light chain (clone 187.1) conjugated with Phycoerythrin-Cyanin7, anti-IgM (clone R6-60.2) conjugated with Peridinin chlorophyll protein-cyanin5.5.

Cycloheximide treatment and Western blot
Stability assay was performed in the presence of cycloheximide (100 μg/mL, Sigma) for the indicated periods of time. The day before, 1.2 × 10 6 transduced Ba/F3 cells were incubated overnight at 37° C, 5% CO 2 . At each time point, treated cells were washed once with ice-cold Phosphate buffer saline (PBS) and then cell pellets were lysed in 40 μl of lysis buffer composed of 150 mM NaCl, 1 mM EDTA, 50 mM Tris, 1% Triton and protease inhibitor cocktail (cOmplete™, Roche). Protein concentrations of cell lysates were determined using the BCA1-Kit for Protein Determination (Sigma-Aldrich) to load 30 µg in every well. Samples were subjected to 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to PVDF membranes. Blots were blocked 1 hour in the presence of 5% milk and incubated with 0.3 μg/mL of primary antibody overnight at +4° C (mouse anti-Pax-5 mAb (A-11), Santa Cruz Biotechnology). After incubation with secondary antibodies conjugated to horseradish peroxidase (Cell Signaling Technology), the immunoreactive bands were visualized by using the enhanced chemiluminescence lighting system (Amersham™ ECL™ Prime Western Blotting Detection Reagent, GE Healthcare) and data was acquired with Chemi-Smart 5000 and Bio-1D software (Vilber Lourmat). Bands were quantified by using ImageJ software. www.oncotarget.com

Transcriptomic analysis
After 5 days in culture with 2 ng/ml of IL7, infected cells were sorted according to their positivity for GFP fluorescence with a FACSAria II (BD Biosciences) cell sorter. Total RNA from GFP pos sorted pro-B infected cells were extracted using the Trizol method according to manufacturer's instruction (Invitrogen). RNA quality was assessed using the RNA 6000 Nano Assay on the Agilent 2400 Bioanalyser (Agilent Technologies, Massy, FRANCE). For each condition (PAX5A, PAX5B and MIE), three independent infection experiments were performed. RNA samples were purified and prepared according to the manufacturer's protocol with the Affymetrix's GeneChip Whole Transcript Sense Target Labeling Assay Kit (Affymetrix, UK) and hybridized on Affymetrix GeneChip Mouse Gene 2.0 ST arrays. Probe-signal intensities were normalized and summarized by the Robust Multiarray Average (RMA) method using Affymetrix's Expression Console software.
The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus [43] and are accessible through GEO Series accession number GSE104890 (https://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc=GSE104890).
For the volcano plots, permutation analyses were performed with dChip software [44] allowing the calculation of false discovery rate p-value and the fold difference volcano plots were generated using GraphPad prism software.
The gene ontology analysis has been performed with the AutoCompare ZE software [45] using Gene Ontology database [18,19], p-values have been calculated with the ZE test.

Luciferase reporter assay
Experiments were performed by cell transfections with 2µg DNA at 1:7:1 (pcDNA3-PAX5 vector: reporter vector: GFP-CMV) vector ratios. GFP-CMV was used to evaluate transfection efficiency. The S194 line was cultured in RPMI as above and for each condition 2 × 10 6 cells were electroporated according to the manufacturer's instructions (Lonza). The 293T line was cultured in DMEM (with 10% FBS, sodium pyruvate, nonessential amino acids, penicillin and streptomycin) and subconfluent cells in 6-well dishes were transfected by lipofection with Exgen500 (Euromedex). Firefly luciferase-CD19-2(A ins) [46] was kindly provided by Pr M. Busslinger was used to measure PAX5 activity on the PAX5 binding site of the Cd19 promoter. PAX5 activity on the 3ʹRR was evaluated with firefly luciferase-pV H enhanced or not by the combination of core enhancers of 3ʹRR [47]. After 48 h transfection, cells were collected and split for GFP fluorescence analysis by flow cytometry and luminescence detection using the Luciferase Reporter Assay System and the 20/20 n Luminometer (Promega). For each sample, luminescence was normalized to the measured GFP fluorescence.

Statistical analysis
Experiments have been performed at least as three independent conditions. All statistical analyses were performed using Excel (Microsoft) and GraphPad Prism software, version 7 (GraphPad Software). Statistical differences were determined using a 2-tailed Student's t test or RM one-way ANOVA with Tukey's correction or one-way ANOVA with Sidak's correction for multiple comparisons or Pearson correlation, as mentioned in legends. All data are presented as mean ± SD or SEM. A p-value of less than 0.05 was considered statistically significant, the null hypothesis being rejected at the 0.05 level ( *** p < 0.0005, ** p < 0.005, * p < 0.05).