MLL1 Promotes IL-7 Responsiveness and Survival during B Cell Differentiation

The mixed lineage leukemia (MLL, MLL1, KMT2A) gene is disrupted by chromosomal translocations that occur with >70% frequency in infant acute leukemia predominantly presenting with a pro-B/pre-B acute lymphocytic leukemia (ALL) phenotype (1). Genome-wide sequencing efforts have illustrated that this disease, even among childhood leukemia subtypes, exhibits remarkably low mutational burden (2), suggesting that the majority of the leukemogenic process is driven by the fusion oncoprotein. In adults, MLL1 translocations or rearrangements are more commonly associated with the myeloid lineage and tend to harbor other mutations, most commonly NRAS and FLT3 (3). Gain-of-function RAS mutations are also the most common mutation in MLL-rearranged pediatric pro-B cell ALL (B-ALL) and, within this poor-prognosis group, predict even worse outcome (4, 5).

MLL1 encodes a histone methyltransferase that is the ortholog of the Drosophila Trithorax protein. MLL1 and Trithorax function as positive epigenetic regulators of selective downstream target genes, such as the well-characterized clustered homeodomain (Hox) or Hom-C genes. This specific gene-regulatory role has been difficult to rationalize given the broadly acting histone-modifying activity and overlapping expression patterns of of related enzymes. In mammals, six histone H3, lysine 4 (H3K4) methyltransferases are responsible for mono-(me), di-(me2), and tri-(me3) methylation of H3K4. Although H3K4me3 enrichment at the transcriptional start site of genes is associated with transcriptionally active or poised genes, H3K4me1 enrichment is predominantly associated with enhancers (6, 7). H3K4me2 enrichment has a more nuanced relationship with regulatory elements but is closely linked to cell identity (8). Understanding which H3K4 methyltransferase performs which specific function is a major challenge, because all six enzymes are frequently coexpressed in tissues (9). Furthermore, how each enzyme is specifically targeted to tissue-specific gene networks is poorly understood. One of the better-characterized paradigms is represented by the recruitment of MLL3 and MLL4 by the sequence-specific PAX transcription activation domain interacting protein, which brings these complexes to IgH switch regions to control transcription and class-switching (10).

Inducible deletion of Mll1 in different hematopoietic populations demonstrated that this methyltransferase is nonredundant and uniquely required for hematopoietic stem cell maintenance in late embryogenesis and adult animals (11–13). Pan-hematopoietic Vav-cre Mll1 deletion resulted in anemia, bone marrow failure, and animal death ∼3 wk after birth (12). In these young Vav-cre;Mll1–deficient animals, the B cell lineage was more severely reduced than the T cell lineage; however, the gross defects in overall bone marrow cellularity confounded the assessment of MLL1 function in B cells specifically. In contrast, previous studies using the late-induced CD19-cre knock-in to delete Mll1 showed no impact on B cell numbers in adult animals (11). Therefore, to directly assess the normal role of MLL1 during early B cell specification and differentiation, we crossed Rag1-cre knock-in (14) with Mll1 floxed (f) allele animals and analyzed B cell differentiation from late gestation to adult animals. This early lymphocyte lineage-specific deletion strategy circumvented gross perturbations of the bone marrow environment and illuminated a B cell–intrinsic requirement for MLL1 for efficient B cell production in the bone marrow. This role was characterized by impaired survival, specifically at the pre-BCR checkpoint, as the result of downstream signaling deficits in the RAS/MAPK pathway. These data suggest that sufficient MLL1 is necessary to maintain effective pre-BCR signaling and that loss of MLL1 results in pressure on pre-B cells to enhance RAS signaling. The connection between wild-type MLL1 and RAS signaling is particularly intriguing, given that RAS pathway mutations are the most common genetic alteration to occur in MLL-rearranged pro/pre-B ALL (4).

Mice were maintained in compliance with the Dartmouth Center for Comparative Medicine and Research and the University of Colorado, Denver Institutional Animal Care and Use Committee policies. B1-8i knock-in mice (stock number 012642; The Jackson Laboratory) were crossed with Rag1-cre;Mll1^f mice. Rag1-cre animals were obtained from Dr. Rabbitts (University of Oxford) (14). Rag1-knockout mice were generated by intercrossing Rag1-cre animals, because the knock-in disrupts the Rag1 gene (14). Rag1-cre;Mll1^f mice were back-crossed to B6.SJL animals, as described (15). Female C57BL/6 animals between 6 and 12 wk of age (stock number 000664; The Jackson Laboratory) were used as recipients of transplanted cells and were sublethally (450 Rad) or lethally (950 Rad) irradiated using a [¹³⁷Cs] source and then maintained for 3 wk on 0.1 mg/ml Baytril (Bayer) in their drinking water.

Sorted pro-B or fraction B cells from the bone marrow were cultured in pro-B medium (Opti-MEM [Invitrogen] supplemented with 10% FBS, 50 μM 2-ME, 2 mM l-Glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin) and the indicated concentrations of recombinant murine (rm) IL-7 (PeproTech). Lineage-negative bone marrow cells were prepared with a lineage-depletion kit (Miltenyi Biotec) and cultured in IMDM with 10% FBS, 50 μM 2-ME, 10 mM HEPES, 10 μM nonessential amino acids, 2 mM l-Glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin (all from Mediatech), as well as 50 ng/ml rmSCF (R&D Systems), 40 ng/ml rmFlt3 ligand, and 5 ng/ml rmIL-7 (both from PeproTech).

Bone marrow cells were prepared and lineage staining was performed as previously described (15), with the exception that the F(ab′)₂ anti-rat conjugate was labeled with PE-Cy5.5 (Invitrogen). Peritoneal cells were collected by injecting 10 ml of FACS buffer (2% FBS in HBSS) into the peritoneal cavity and flushing back out into a collection tube. Flow cytometry was performed with a FACSCalibur or FACSAria (BD Biosciences), and data were analyzed using FlowJo software (TreeStar). Analyses used live cell gates that included most hematopoietic cells; however, for sorting purified populations, non-B cell lineages were depleted from the bone marrow cells prior to surface marker staining. For intracellular staining, cells were fixed for 20–30 min at room temperature with Cytofix/Cytoperm buffer and were washed and stained in Perm/Wash buffer (both from BD Biosciences). Bone marrow B cells were defined as follows: fraction A (pre/pro-B; CD43⁺CD24^negBP-1^neg), fraction B (early pro-B; CD43⁺CD24⁺BP-1^neg), fraction C (late pro-B; CD43⁺CD24⁺BP-1⁺), fraction C′ (large pre-B; CD43⁺CD24^hiBP-1⁺), fraction D (small pre-B; CD43^negIgM^negIgD^neg), fraction E (B220⁺IgM⁺IgD^neg), and fraction F (B220⁺IgM⁺IgD⁺).

B220⁺ or CD19⁺ bone marrow cells, enriched by Miltenyi Biotec MicroBeads, were cultured with 5–10 ng/ml IL-7 for 5–6 d to generate bone marrow–derived pro/pre-B cells. Cultured cells were incubated without IL-7 for 4–6 h and then treated with 30 μg/ml biotin-conjugated Igβ mAb HM79 (SouthernBiotech) for 20 min on ice. Stained cells were washed with cold Opti-MEM medium (Invitrogen), followed by addition of 30 μg/ml streptavidin (Sigma-Aldrich). Cells were incubated at 37°C for 0.5–10 min and centrifuged at 4°C, and total cellular proteins were extracted with E1a lysis buffer (50 mM HEPES [pH 7.4], 250 mM NaCl, 5 mM EDTA, 0.1% Nonidet P-40) containing cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail (Roche) plus Phosphatase Inhibitor Cocktails A and B (Santa Cruz Biotechnology). Cellular lysates were resolved using 12% SDS-PAGE gels and transferred to nitrocellulose. Blots were blocked in 5% nonfat milk in PBS containing 0.1% Tween-20 with the indicated Abs. For detection of p-ERK and p-AKT, the following Abs were used: rabbit total anti-ERK1/2 (9102), mouse anti–p-ERK (T202/Y204, E10), rabbit anti-AKT (C67E7), and rabbit anti–p-AKT (S473) (all from Cell Signaling Technology). U0126 was also obtained from Cell Signaling Technology.

Retroviral supernatants were produced by cotransfection of an MSCV plasmid and a Psi-minus Eco packaging plasmid (Dr. Witte, University of California, Los Angeles) in 293T cells using FuGENE 6 (Promega). Socs2 cDNA was amplified from C57BL/6 bone marrow cDNA using the following primers: 5′-GGGACGTGTTGACTCATCTCCCAT-3′ and 5′-CGAAAAAGAGAGAGAAATACTTA-3′. The cDNA was TA cloned and transferred into an MSCV–IRES–hCD4 vector (16). Supernatants were collected 48 h after transfection, filtered at 0.45 μm, and frozen at −80°C until use, as described (16). Lineage-negative bone marrow cells were spin infected on retronectin-coated plates (Takara), whereas fraction B cells were spin infected for 1.5 h in pro-B medium containing 5 μg/ml Polybrene (Sigma-Aldrich), washed, and transferred to new plates with fresh growth medium. To enumerate pre-B colonies, cells were counted using polystyrene beads (Polysciences) as a reference (15), and 3000 cells were plated onto pre-B (M3630) methylcellulose, according to the manufacturer’s recommendations (STEMCELL Technologies). Colonies were counted at days 7–8 and characterized by flow cytometry. The MSCV-Igμ plasmid encoding 383 cDNA was obtained from Dr. M. Mandal and Dr. M. Clark (University of Chicago).

Fraction B pro-B cells were first sorted from Rag1-cre;Mll1^F/F or Rag1-cre;Mll1^F/+ mouse bone marrow and infected with MSCV–BCR–ABL p210 viral supernatant (gift from Dr. S. Li and G. Raffel, University of Massachusetts Medical School) in pro-B cell medium supplemented with 5–10 ng/ml IL-7. Spin infection was performed as described above. Eight thousand transduced cells were mixed with 400,000 unfractionated C57BL/6 bone marrow cells, and the mixture was injected periorbitally into lethally irradiated female C57BL/6 recipients. Mice were closely monitored for any sign of disease and were sacrificed when moribund. A donor, B cell phenotype (GFP⁺/CD45.1⁺/CD19⁺/HSA⁺/Mac-1^neg) was confirmed for all moribund animals by flow cytometry.

RNA was prepared using TRIzol Reagent (Invitrogen), followed by RNeasy columns (QIAGEN). cDNA was prepared using Superscript III (Invitrogen), and quantitative real-time PCR (qRT-PCR) was performed with TaqMan PCR Master Mix or SYBR Green Mix (Bio-Rad) and analyzed on an ABI Prism 7500 (Applied Biosystems). Relative transcript abundance was calculated using the ΔΔCt method after normalization using a rodent Gapdh TaqMan assay (Applied Biosystems) or a Tbp or Hprt1 SYBR assay. Primers used for qRT-PCR were CD179a, 5′-CGTCTGTCCTGCTCATGCT-3′ and 5′-ACGGCACAGTAATACACAGCC-3′; CD179b, 5′-TGTGAAGTTCTCCTCCTGCTG-3′ and 5′-ACCACCAAAGTACCTGGGTAG-3′; Igμ, 5′-AAGGATGGGAAGCTCGTGGAATCT-3′ and 5′-TCAGGGTTTCATAGGTTGCCAGGT-3′; CD79a, 5′-CATCTTGCTGTTCTGTGCAGTG-3′ and 5′-TTCTCATTTTGCCACCGTTTC-3′; and CD79b, 5′-GCTGTTGTTCCTGCTGCTGC-3′ and 5′-CTTCACCATGGAGCTCCGCTTT-3′.

Genomic DNA from sorted pro-B cells was purified with a Wizard SV Genomic DNA Purification System (Promega) and used for SYBR-based PCR to detect unrearranged D_H-J_H products, as described previously (17). The Hprt genomic locus was used as a control for normalization using 5′-GCTGGTGAAAAGGACCTCT-3′ and 5′-CACAGGACTAGAACACCTGC-3′ primers.

DNA content was determined by staining sorted B cell fractions with Krishan’s reagent (0.05 mg/ml propidium iodide (PI), 0.1% sodium citrate, 0.02 mg/ml RNase A, 0.3% Nonidet P-40 [pH 8.3]). DNA content was analyzed by flow cytometry. For BrdU flux analysis, mice were injected with 0.6 mg BrdU (BD Biosciences) in 200 μl of PBS i.p. every 12 h and were sacrificed at 24, 48, or 72 h. Bone marrow was harvested and stained for surface phenotype and BrdU incorporation, as described (18). The percentage of BrdU-labeled cells in each subset was determined and plotted as a function of time, and least-squares regression analysis was performed to obtain the turnover and production rates, as previously described (19).

One cohort of age- and sex-matched Rag1-cre;Mll1^F/F and Mll1^F/F mice (n = 4 each genotype) was used to sort fraction B cells. RNA was purified as above and amplified using a MessageAmp II aRNA Amplification Kit (Ambion), labeled using a BIOARRAY HIGHYIELD RNA Transcript Labeling Kit (T7; Enzo Life Sciences), and fragmented and hybridized to Mouse 430 2.0 Arrays (Affymetrix) at the Dartmouth Genomics and Microarray Laboratory. Analysis was performed using Partek Genomics Suite (St. Louis, MO); intensities were determined and normalized using the GC robust multi-array average procedure, and differentially expressed probe sets were determined by class comparison using log₂ expression values (Supplemental Table I). An additional table was generated using a p value cutoff of 0.005, and probe sets were collapsed to genes by retaining the probe set with the greatest fold change difference (Supplemental Table I). Heat maps and numbers of genes discussed in the text were determined using this p value–filtered data with a fold-change cutoff of 1.2.

Total RNA was purified from sorted fraction B cells with an miRNeasy Micro Kit (QIAGEN). A 2100 Bioanalyzer (Agilent) was used to assess the concentration and integrity of the RNA. One hundred nanograms of amplified DNA was the input for microRNA profiling using an nCounter Mouse miRNA Expression Assay Kit (NanoString Technologies). Sample preparation and hybridization were performed following the manufacturer’s recommendations at the Dartmouth Molecular Biology Core facility. Data analysis was performed using nSolver Analysis Software 1.1. Expression counts were normalized to the sum of the counts of the 100 highest-expressed microRNAs in each sample (complete data in Supplemental Table II).

Chromatin immunoprecipitation sequencing libraries were prepared using sorted pro-B cells pooled from three mice per genotype, following the manufacturer’s recommendations (Ovation Ultralow Library Systems; NuGEN). Multiplexed sequencing was performed on a HiSeq 2000 instrument (Illumina). The resulting barcoded multiplexed sequences were debarcoded using Unix shell script and converted into Fastq format using qseq2fastq.pl perl script. Reads were mapped to the mouse genome (mm9) using Bowtie v0.12.7; only reads that aligned to a unique position in the genome with no more than two mismatches were retained for further analysis. Bowtie-outputted SAM files containing aligned reads were used to create bedGraph files using Homer, with default parameters converted to bigWig format using the bedGraphToBigWig program, and visualized on an Integrative Genomics Viewer (Broad Institute, Massachusetts Institute of Technology).

Unless otherwise indicated, unpaired Student t tests were used to calculate p values. All error bars represent 95% confidence intervals, unless otherwise indicated in the figure legends. Graphs were created and statistical tests were performed using Prism (GraphPad) or Excel (Microsoft).

Our previous studies suggested a function for Mll1 in B cells; however, the severe effects of Mll1 loss on multiple bone marrow populations made it difficult to rule out cell-extrinsic effects on B cell generation (12). To specifically examine the role of MLL1 in lymphocyte differentiation, we used a Rag1-cre knock-in strain to delete Mll1 exclusively in developing lymphocytes (14, 15). Bone marrow B cell stages of differentiation were identified using the Hardy scheme (fraction A through fraction F) (20). Cre expression from the Rag1 locus resulted in nearly 100% deletion of a Rosa-YFP reporter by fraction B, or pro-B cells (Fig. 1A), and in the T cell lineage by the double-negative 1 population (data not shown). Deletion of the floxed Mll1 gene paralleled what was observed using the YFP reporter (data not shown). Thus, phenotypic analyses focused on only those populations in which Mll1 deletion was 90–100%.

Rag1-cre;Mll1^F/F animals exhibited normal bone marrow cellularity, but B cell populations were reduced ∼2-fold in the bone marrow and were similarly reduced in peripheral organs (Fig. 1B, 1C). The B cell reduction was most profound in 2–3-wk-old animals but persisted as the animals aged (Fig. 1D). Peripheral T cell numbers and thymocyte subsets were not reduced, despite efficient Mll1 deletion (Fig. 1D).

To determine more specifically the stage of B cell differentiation affected in the bone marrow, we analyzed B cell populations using the strategy shown in Fig. 2A. Mll1-deficient B lymphocyte progenitors appeared to accumulate slightly in fraction C, but all stages from fraction C′ and beyond were consistently reduced (Fig. 2B). Because signals from the pre-BCR are critical for transitioning to large pre-B cells (C to C′ fraction), we considered whether reduced levels of the pre-BCR could explain this partial block in differentiation in Rag-cre;Mll1^F/F animals. Levels of the pre-BCR measured by the SL156 Ab were normal to high relative to controls (Fig. 2C, 2D), as was expression of most pre-BCR components, including Igμ, VpreB, λ5, and Igα/Igβ (Fig. 2E). Therefore, failure to express pre-BCR components does not account for the reduction in developing B cells. These data also demonstrate that MLL1 plays an important role in B lymphopoiesis at the C–C′ transition, in contrast to its minimal role in T cell differentiation.

Despite normal to high levels of pre-BCR components, we found that the Mll1-deficient pro-B cells had more unrearranged IgH D-J alleles than rearranged alleles (Fig. 3A), as well as higher levels of IgH germline transcripts (Fig. 3B). Because MLL1 is an H3K4 methyltransferase, and RAG-2 binding depends on H3K4me3 modification of its target chromatin (21), we tested whether H3K4me3 levels were reduced at the IgH locus in Mll1^−/− pro-B cells. Chromatin immunoprecipitation sequencing experiments revealed that there was no reduction in H3K4me3 along the entire IgH locus in Rag1-cre;Mll1^F/F pro-B cells, with the characteristic peak at the JH region appearing identical in control and Mll1-deficient cells (Supplemental Fig. 1A, arrow). To functionally test whether rearrangement of the IgH locus was limiting for B cell differentiation, we used two strategies to introduce prerearranged IgH loci to the Rag1-cre;Mll1^F/F cells or animals. First, we introduced a rearranged IgH cDNA by retroviral transduction into Mll1-deficient progenitors and enumerated pre-B colonies (22). GFP was also expressed from the bicistronic virus, providing a surrogate marker for retroviral IgH expression. Mll1^−/− cells transduced with the control virus exhibited reduced colonies in pre-B methylcellulose cultures, resulting in a lower percentage of B220⁺/GFP⁺ cells (Fig. 3C). This reduced B220⁺/GFP⁺ cell number was not rescued by expression of IgH cDNA (Fig. 3C), although control experiments using Rag1^−/− cells demonstrated Rag1 cDNA rescue (Supplemental Fig. 1B).

The second strategy used a prerearranged IgH knock-in (B1-i8) mouse (23), which was crossed to Rag1-cre;Mll1^F/F and Rag1-cre;Mll1^F/+ animals to ensure that sufficient IgH was provided at a developmentally accurate time and expression level. We found that introduction of the B1-i8 knock-in reduced pre-B cells, regardless of Mll1 status (Fig. 3D, gray symbols), which may reflect accelerated differentiation (23). Similar to the observations using the MSCV-IgH rescue strategy, B1-i8⁺;Rag1-cre;Mll1^F/F mice still exhibited significantly fewer CD43^neg/B220⁺ pre-B cells than B1-i8⁺;Rag1-cre;Mll1^F/+ animals (Fig. 3D), although this strategy clearly rescued Rag1-knockout pre-B cells (Supplemental Fig. 1C). Collectively, these results rule out inefficiencies in IgH recombination as a major explanation for the reduced B cell numbers in animals lacking MLL1 in the B cell lineage.

The amplification of B cells after successful IgH rearrangement occurs downstream of pre-BCR signaling and involves survival, proliferation, and, ultimately, differentiation of pro-B cells to pre-B cells. Therefore, we assessed whether the lack of expansion after pre-BCR signaling in Rag1-cre;Mll1^F/F mice was due to proliferation or cell survival defects. No difference in the number of cycling cells (S/G₂/M) in any of the pro/pre-B cell fractions were observed (Fig. 4A). Similarly, a more dynamic analysis using BrdU pulse-chase experiments revealed that flux through the pro-, pre-, and immature B cell populations was indistinguishable between control and Rag1-cre;Mll1^F/F animals (Fig. 4B). Together, these data suggest that altered proliferation does not account for the reduced output of B cells after the pro-B cell stage.

To determine whether cell survival was affected in Rag1-cre;Mll1^F/F developing B cells, we sorted pooled populations from the bone marrow and observed a slight increase in the annexin V⁺ cell percentage in Mll1-deficient B220⁺CD43^neg pre-B cells (fractions D–F, Fig. 4C). Given the difficulty in detecting annexin V–binding cells in vivo because of their rapid clearance, as well as the imperfect correlation with cells committed to apoptosis (24, 25), we also isolated fraction B pro-B cells (Lin^neg/CD19⁺/CD43⁺/BP-1^neg/CD24⁺) and assessed their survival in vitro using a range of IL-7 concentrations (26). IL-7 is critical for the survival of pro-B cells in vivo and in vitro, and it synergizes with signals from the pre-BCR to promote expansion of cells upon successful IgH rearrangement (27). Assessment of cell numbers after 7 d in culture revealed that Rag1-cre;Mll1^F/F cells exhibited an IL-7 concentration–dependent reduction in cell accumulation (Fig. 4D). Interestingly, the reduction in cell accumulation could be overcome by supraphysiologic levels of IL-7 (Fig. 4D, 4E), similar to Rag2^−/− pro-B cells (28). Furthermore, low (0–1 ng/ml) IL-7 cultures consistently contained more PI-permeable cells, yet they induced CD2 and CD22 similarly to wild-type cells (data not shown). Collectively, these data suggest that the reduced B cell output in Rag1-cre;Mll1^F/F animals is the result of reduced B cell survival, particularly downstream of IL-7, and not reduced proliferation or differentiation.

Given the defects in survival at or near the pre-BCR checkpoint (pro/pre-B transition), we explored further potential defects in pre-BCR signaling. The pre-BCR triggers a complex signaling network that promotes survival, proliferation, and differentiation of pre-B cells in synergy with IL-7 signaling. These signals converge at the level of ERK activation (28). We found that IL-7R levels were not reduced on Mll1-deficient pro-B cell populations, and also determined that STAT5 phosphorylation in response to 0.5 or 50 ng/ml IL-7 was not affected by Mll1 deficiency (Supplemental Fig. 2A–E). Furthermore, AKT basal phosphorylation or response to pre-BCR stimulation did not differ between control Rag1-cre;Mll1^F/+ and Rag1-cre;Mll1^F/F pro-B cells expanded in vitro (Supplemental Fig. 2F). In contrast, investigation of the RAS–MEK–ERK pathway revealed deficiencies in signaling upon pre-BCR stimulation. Although the levels of total ERK1/2 were similar to controls in cultured Rag1-cre;Mll1^F/F animals, p-ERK1/2 (detected with anti-pT202/Y204) was consistently attenuated (Fig. 5A). H-, K-, and N-RAS are all expressed in pro-B cells; however, K-RAS plays a dominant role downstream of the pre-BCR (29). Neither the level of K-RAS nor the phosphorylation status of MEK1/2 or B-RAF accounted for the attenuated p-ERK1/2 (Supplemental Fig. 2G, 2H). Furthermore, Mll1-deficient pro-B cells exhibited enhanced sensitivity to the MAPK inhibitor U0126, demonstrating that the impaired survival can be further inhibited (Fig. 5B). These data suggest that the poor survival in physiologic IL-7 and reduced B cell output in Rag1-cre;Mll1^F/F animals originates from a signaling inefficiency, but not block, downstream of the pre-BCR, culminating in insufficient phosphorylation of ERK1/2.

MLL1 translocations produce proteins with neomorphic functions, but fusion oncoproteins also lack normal MLL1 functions (e.g., histone methyltransferase activity at the C terminus). Thus, it is possible that gain- and loss-of-function activity of total MLL1 in the developing B cell occurs upon MLL1 translocation. The selective disruption in B cell differentiation observed in the loss-of-function model prompted us to examine whether the differentiation block could contribute to leukemogenesis, similar to the mechanisms by which reduction in PAX5, EBF, or IKAROS can contribute to leukemogenesis in human B-ALL (30). In contrast, if the signaling deficiencies cannot be overcome by a particular oncogenic signal, MLL1 loss may limit leukemogenesis. To determine the impact of MLL1 loss in a well-characterized B-ALL model system, we transduced Rag1-cre;Mll1^F/F and Rag1-cre;Mll1^F/+ pro-B cells with BCR-ABL (31) and transplanted cells into irradiated syngeneic recipients (Fig. 6A) or expanded these cells in IL-7 cultures (Fig. 6B). As shown in Fig. 6A, recipients of BCR-ABL/Mll1^−/− cells exhibited extended survival relative to their Mll1^+/− controls (p = 0.0152), demonstrating that MLL1 facilitates B-ALL driven by BCR-ABL.

To examine the status of pre-BCR signaling in the BCR-ABL–transformed cells, we used in vitro–expanded cultures of transformed cells. Surface marker expression demonstrated that Mll1-deficient and control Mll1^+/− cells were immortalized at the fraction C/C′ stage, and no phenotypic difference was observed between the two genotypes (Fig. 6B, data not shown). Upon anti-Igβ cross-linking, the level of p-ERK1/2 in BCR-ABL–transformed cells was significantly greater than that observed in primary pro-B cells. However, we still observed attenuated phosphorylation of ERK1/2 in Mll1-deficient BCR-ABL cells relative to controls (Fig. 6C), suggesting that the delayed latency of B-ALL obtained with Mll1-deficient B-ALL cells is due to a signaling defect at the level of ERK phosphorylation that cannot be overcome by BCR-ABL.

To gain insight into the phenotypic defects in Mll1^−/− pro-B cells, we performed genome-wide transcriptome and microRNA surveys using sorted fraction B pro-B cells. These analyses revealed that 224 genes were downregulated and 129 genes were upregulated in Mll1^−/− pro-B cells using a fold-change threshold of 1.2 and a p value cutoff of 0.005 (Fig. 7A, 7B, Supplemental Table I). Several well-described MLL1 direct target genes were among the most strongly downregulated genes (Hoxa9, Eya1, Meis1, Pbx1, red in Fig. 7A); however, surprisingly, many well-studied B cell regulators, including Ebf1, Pax5, Ikzf1, Spi1, Blk1, and Irf4, were unchanged (Supplemental Table I, data not shown). Several hypotheses were tested based on these findings.

First, Hoxa9 germline–knockout animals exhibit reduced pro-B populations and pre-B cell colony frequency, particularly in conjunction with Meis1 deficiency (32, 33), whereas Eya1 was consistently and significantly downregulated but is not known to play a role at the pro/pre-B cell transition. Thus, we attempted to rescue pre-B cell colony frequency by reintroducing Hoxa9 or Eya1 expression to Rag1-cre;Mll1^F/F progenitors, but neither of these individual genes restored pre-B cell colony numbers. In fact, Hoxa9 re-expression reduced pre-B colonies and tended to promote myeloid differentiation (Supplemental Fig. 3A, data not shown). Second, suppressor of cytokine signaling 2 (Socs2) is significantly increased in Rag1-cre;Mll1^F/F pro-B cells (Fig. 7B); given its potential in negatively regulating IL-7 signaling (34), we tested whether overexpression of SOCS2, similar to the level that we observed in Mll1^−/− pro-B cells, could recapitulate the IL-7–dependent survival defect. We found that neither high nor low (comparable to Mll1^−/− pro-B) levels of SOCS2 overexpression had any impact on IL-7–dependent survival (Supplemental Fig. 3B, 3D), suggesting that this single gene expression alteration is not sufficient to recapitulate the Mll1^−/− phenotype (or to reduce B cell viability). Thus, it is likely that the high Socs2 expression occurred as a result of attenuated RAS–MEK–ERK signaling and was not causative.

Analysis of mature microRNAs revealed that two were significantly decreased and three were significantly increased in Mll1-deficient pro-B cells relative to their wild-type counterparts, leading us to test two major hypotheses (Fig. 7C, 7D, Supplemental Table II). First, the increase in Let7a and the fact that LET7 family members can reduce K-RAS levels and activity (35, 36) prompted us to determine K-RAS protein levels. We found that K-RAS total protein was unaffected in Mll1^−/− pro-B cells (Supplemental Fig. 2H). Second, miR-155 and miR-146b are part of a feedback loop regulating NF-κB activity (37, 38). Because NF-κB activity can coordinate differentiation and survival downstream of the pre-BCR (39), we determined whether Mll1-deficient pro-B cells exhibit altered NF-κB activation downstream of the pre-BCR. As shown in Supplemental Fig. 3E and 3F, the turnover of IκBα and the nuclear translocation of p65/RelA (both are responses to NF-κB activation) were indistinguishable between control and Mll1-deficient pro-B cells. Thus, despite analysis of the earliest population to achieve 100% deletion of Mll1, it is likely that some of the gene expression changes observed were indirect or compensatory. Our data are consistent with the hypothesis that altered expression of multiple transcripts and microRNAs collectively impact the effectiveness of pre-BCR signaling. Overall, our data demonstrate that the role of wild-type MLL1 in B cell differentiation is to appropriately tune pre-BCR signals by impacting the RAS–MEK–ERK axis and, thereby, increase survival of pro-B cells in physiologic IL-7 concentrations during homeostasis.

In this study, we show that MLL1 plays a significant role in B cell differentiation, specifically at the pre-BCR checkpoint. Surprisingly, this role appears not to involve classic transcriptional regulators of B cell differentiation that also act as tumor suppressors in B-ALL; rather, it involves coordinating the response to pre-BCR signaling. The attenuated pre-BCR signaling observed in Mll1-deficient pro/pre-B cells results in reduced cell survival and inefficient, but not completely blocked, B cell differentiation that apparently cannot be compensated for at later stages of differentiation to restore peripheral B cell numbers. The exact molecular mechanism linking MLL1 to MEK/ERK signaling at this stage of differentiation appears to involve a complex network of mRNA and microRNAs, because no single candidate deregulated gene could account for the selective phenotype observed upon MLL1 loss. Interestingly, high IL-7, but not BCR-ABL–mediated transformation, could overcome the survival and p-ERK defects imparted by Mll1 deficiency.

IL-7R and pre-BCRs are critical for B cell differentiation and can be subverted in B cell malignancy (40). The requirement for IL-7 during B cell differentiation is highly dynamic, tuning survival and differentiation at the pro- to pre-B transition (41, 42). When a high level (ng/ml) of IL-7 is present in the environment, JAK-STAT5 and PI3K-AKT pathways activated solely by IL-7 are sufficient to sustain the expansion of pre-B cells. Developing B cells in a niche with limited IL-7 availability (pg/ml) exhibit attenuated STAT5 and AKT, thus requiring pre-BCR signaling to augment PI3K and RAS-ERK signaling to enable the survival of cells harboring functional pre-BCRs (27, 43). Using sorted fraction B pro-B cells, we found that MLL1 is specifically required for cells to respond to low IL-7, similar to observations using pre-BCR–deficient pro-B cells (43). Given the fact that Mll1-deficient cells still responded to high levels of IL-7, we speculate that these cells do not have an absolute defect associated with IL-7 signaling but, rather, exhibit inefficient signaling. Complete ERK1/2 double knockout in developing B cells results in severe proliferation and survival defects (44), whereas Mll1^−/− pre-B cells exhibit survival, but not proliferation, defects, suggesting that an effect on proliferation requires a more complete reduction in total p-ERK. In addition, one unique feature of the RAS-MEK-ERK pathway in B cell differentiation is that effects on survival are wired differently depending on the differentiation stage. For example, the survival of mature B cells during mitogenic stimulation requires MEK/ERK-mediated phosphorylation of Bim (45), whereas inhibition of RAS activity significantly reduced the survival of pre-B cells that was due, in part, to downregulation of Bcl-x_L (46). In this study, neither of these transcripts was consistently or significantly reduced in Mll1^−/− pro/pre-B cells (data not shown). Therefore, it remains to be determined through which pathways cell death is deregulated. Nonetheless, control of pre-BCR–dependent RAS-MAPK signaling by the proto-oncogene MLL1 is intriguing, given the propensity of infant ALL to acquire gain-of-function RAS mutations (4).

We thank Michael Farrar, Jing Zhang, Demin Wang, and our laboratory members for commenting on the manuscript. Dr. Malay Mandal, Dr. Marcus Clark, Dr. Shaoguang Li, and Dr. Glen Raffel provided critical reagents and advice. Reagents and mice donated by Prof. Klaus Rajewsky and Prof. Terence Rabbitts were important for this work.

P.E. owns Amgen stock and has consulted for Servier Oncology. The other authors have no financial conflicts of interest.