Research ArticleLEUKEMIA

MAFB enhances oncogenic Notch signaling in T cell acute lymphoblastic leukemia

See allHide authors and affiliations

Science Signaling  14 Nov 2017:
Vol. 10, Issue 505, eaam6846
DOI: 10.1126/scisignal.aam6846

New targets, better mouse model for leukemia

T cell acute lymphoblastic leukemias (T-ALLs) are often caused by mutations in the gene encoding Notch1, which mediates cell-cell contact signaling in embryonic development and adult tissue maintenance. However, mice expressing these mutants frequently fail to develop T-ALL. Pajcini et al. found that the transcription factors MAFB and ETS2 increased the expression of Notch1 target genes in mouse and human T-ALL cells by recruiting histone acetyltransferases. Expressing MAFB enhanced the development of Notch1-mutant T-ALL in mice. Because Notch1 is critical for the maintenance of various healthy adult tissues, developing a way to inhibit MAFB or its interacting partners may be a more targeted therapy for leukemia patients.


Activating mutations in the gene encoding the cell-cell contact signaling protein Notch1 are common in human T cell acute lymphoblastic leukemias (T-ALLs). However, expressing Notch1 mutant alleles in mice fails to efficiently induce the development of leukemia. We performed a gain-of-function screen to identify proteins that enhanced signaling by leukemia-associated Notch1 mutants. The transcription factors MAFB and ETS2 emerged as candidates that individually enhanced Notch1 signaling, and when coexpressed, they synergistically increased signaling to an extent similar to that induced by core components of the Notch transcriptional complex. In mouse models of T-ALL, MAFB enhanced leukemogenesis by the naturally occurring Notch1 mutants, decreased disease latency, and increased disease penetrance. Decreasing MAFB abundance in mouse and human T-ALL cells reduced the expression of Notch1 target genes, including MYC and HES1, and sustained MAFB knockdown impaired T-ALL growth in a competitive setting. MAFB bound to ETS2 and interacted with the acetyltransferases PCAF and P300, highlighting its importance in recruiting coactivators that enhance Notch1 signaling. Together, these data identify a mechanism for enhancing the oncogenic potential of weak Notch1 mutants in leukemia models, and they reveal the MAFB-ETS2 transcriptional axis as a potential therapeutic target in T-ALL.


T cell acute lymphoblastic leukemia (T-ALL) is an aggressive malignancy of immature T cell blasts that occurs in both children and adults (1). Although current treatments are relatively successful, especially in children (2), ~20% of patients are not cured by current therapy (1). Activating mutations in NOTCH1 are frequent, occurring in >60% of T-ALLs (3).

The intercellular Notch receptor–ligand interaction activates a conserved signaling pathway that regulates multiple cellular functions (4). Physiologic Notch signaling occurs when a Notch receptor interacts with a ligand of the Jagged or Delta-like family in a neighboring cell. This initiates a series of metalloproteinase- and γ secretase–mediated cleavages that release the intracellular Notch (ICN) domain from the plasma membrane. The ICN then translocates to the nucleus where it forms the Notch transcriptional complex (NTC) with the DNA binding protein RBPJ and a member of the Mastermind-like (MAML) family. This complex activates transcription of Notch target genes (5). The NTC is short-lived and is tagged for degradation by proteins including the ubiquitin ligase Fbxw7, which recognizes PEST sequences residing in the ICN C terminus (6).

Activating NOTCH1 mutations in T-ALL occur in the extracellular Notch1 regulatory region (NRR) and/or the C-terminal PEST domain (7). NRR mutations lead to ligand-independent signaling, whereas PEST mutations limit ICN degradation. Thus, these mutations either increase nuclear Notch1 and/or inhibit NTC turnover, which ultimately leads to dysregulated (increased) Notch1-mediated transcription. The enhanced Notch1 signaling resulting from either of these mutations can be inhibited by γ-secretase inhibitors (GSIs), which prevent release of the ICN from the plasma membrane (3).

Expressing ICN1 in murine T-ALL models rapidly induces leukemia in all mice; however, this form of Notch1 rarely occurs in human T-ALL (3). In contrast, expressing mutated NOTCH1 alleles commonly associated with human T-ALL results in T-ALL with a much longer latency that is incompletely penetrant (8). These findings suggest that events that synergize with the weak Notch1 mutants and/or increase their signaling strength likely contribute to T-ALL. In support of this idea, coexpressing weak Notch1 alleles with mutations found in human patients, such as oncogenic Ras, decreases T-ALL latency and increases penetrance (8).

To identify potentially oncogenic hits that directly modify Notch1 signaling, we performed a gain-of-function complementary DNA (cDNA) screen to discover molecules that enhance the ability of weak oncogenic NOTCH1 mutants. Among the hits were the transcription factors MAFB and ETS2. MAFB belongs to the Maf family of AP1 transcription factors, all of which contain a basic leucine zipper (bZIP) domain that binds two long palindromic sequences referred to as “MARE” sequences (9). MAFB also contains an N-terminal transcriptional activation domain, which optimally induces Maf target gene transcription by recruiting other transcriptional regulators (10, 11). MAFB serves many roles in embryonic development, including hematopoiesis, where it exerts important functions in macrophage differentiation. In cancer, translocations of MAFB and its closely related homolog, MAF, are frequent in multiple myeloma (12). ETS2 belongs to a large and nearly ubiquitously expressed family of transcription factors that function in a wide variety of developmental roles (13). In hematopoiesis, ETS2, like ETS1, with whom it shares close homology, is important for cortical thymocyte proliferation and survival (1416). ETS transcription factors are also dysregulated in multiple cancers, including Ewing’s sarcoma (17) and prostate cancer (18).

To determine the relevance of the findings of our screen, we investigated the function of MAFB and ETS2 in multiple assays relevant to T-ALL pathogenesis. We found that although both MAFB and ETS2 individually enhanced the signal strength of the Notch1 gain-of-function alleles, their coexpression resulted in synergistic activity that was comparable to the core NTC. Furthermore, MAFB alone increased the penetrance and decreased the latency of T-ALL induced in mice with weak Notch1 gain-of-function mutants, whereas inhibiting MAFB in both human and murine T-ALL cells inhibited their growth and decreased the expression of a subset of Notch target genes. We found that MafB interacts with ETS2 and, in doing so, cooperates with the NTC to augment Notch1-mediated transcription. We thus propose that MAFB functions to amplify the signaling output of naturally occurring Notch1 mutants, a finding consistent with its expression in a high percentage of human T-ALLs. Not only do these data identify new proteins that enhance Notch1 function in T-ALL but they also identify MAFB as a potential new therapeutic target in T-ALL.


A cDNA screen detects enhanced Notch signaling

Our Notch1 gain-of-function screen relies on three genetic components transfected into U2OS cells: (i) a sensitive, synthetic Notch signaling reporter (TP1) containing 12 iterated RBPJ-binding sites driving luciferase expression in a pSP72 backbone; (ii) a cytomegalovirus (CMV) vector (pcDNA3) expressing a NOTCH1 mutant; and (iii) the cDNA library, containing 18,000 open reading frames in the Sport6 CMV vector, individually preplated in 384-well plates (Fig. 1A). U2OS cells were used because of their high transfectability and low basal Notch signaling. We assayed several Notch mutants previously identified in primary patient samples. We chose NOTCH1 L1601P-DeltaPest (LPΔP) (8), which contains both heterodimerization domain (HD) and PEST mutations, because it consistently yielded a luciferase signal above background and produced a significant z score of 0.602 above background (fig. S1A). In contrast, the weaker NOTCH1 L1601P mutant (8), which lacks a PEST mutation, only marginally increased background Notch signaling in U2OS cells. Both mutants were considerably weaker transcriptional activators than ICN1 (fig. S1A).

Fig. 1 Screen to identify novel enhancers of Notch signaling.

(A) Schematic of the methodology used in the cDNA gain-of-function screen. (B) List of candidate genes potentiating NOTCH1 LPΔP activity. MAML1control is the MAML1 positive control. Black dots indicate candidates that were independently verified in reporter assays. (C) Luciferase induction using the Notch-responsive TP1 reporter and LPΔP to validate candidates identified in the screen. Data were quantified relative to the empty vector (EV) control. (D) Luciferase assay using the TP1 reporter (40 ng per well) and LPΔP (40 ng per well) and assaying dose dependency of Notch activation by ETS2 (40, 80, and 120 ng per well) or MAFB (40 and 80 ng per well). (E) Luciferase TP1 reporter assay with LPΔP when MAFB and ETS2 were singly or cotransfected (at 40 ng each). To normalize the amount of plasmid per well, we added EV so that the total amount of DNA was 80 ng per well (not including reporters). Data are from three independent experiments. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.0005, ****P ≤ 0.0001 by unpaired t tests. NS, not significant.

From the screen, we identified the top 30 candidate genes, ranked by relative fold induction of reporter activity (Fig. 1B and table S1). As expected, the positive control MAML1, a critical NTC component, produced the greatest enhancement. Additional evidence that the screen was specific was the identification of ATP2A3, which encodes SERCA3, and ERO1L, both of which are known to function in the Notch signaling pathway (19, 20).

MAFB induces dose-dependent enhancement of Notch signaling

We independently validated the ability of several top candidate hits to activate the TP1 reporter (Fig. 1C). Verified hits comprised several transcription factors including MAFB and ETS2. Of the confirmed candidates that we assayed, only MAFB showed a significant dose-dependent enhancement of Notch reporter activity in combination with LPΔP (Fig. 1D). This was not a general effect of MAFB on transcriptional reporters because MAFB had minimal effects on a nuclear factor κB (NF-κB) reporter (fig. S1B). These findings led us to focus on MAFB. Although MAFB was not expressed in U2OS cells, it was expressed in murine and human T-ALL cell lines and primagrafts (fig. S1, D and E). Furthermore, analysis of two independent patient data sets (21, 22) identified a range of MAFB expression in primary human T-ALL samples (fig. S1, F and G).

MAFB and ETS2 combine to enhance Notch signaling

ETS2 was also identified as an enhancer in our screen (Fig. 1B), and its closely related homolog, ETS1, is known to interact with MAFB (23). We used Notch reporter assays to test the potential cross-talk between MAFB and ETS2. The two genes were transfected into U2OS cells along with LPΔP. Not only did MAFB and ETS2 synergistically increase reporter activity but the signal was also comparable to that induced by the MAML1 positive control (Fig. 1E). Although MAFB and ETS2 synergized in this reporter assay, MAFB appeared to be the limiting component. A twofold decrease in the amount of MAFB decreased the reporter signal by ~50%, whereas a fourfold drop in ETS2 DNA concentration resulted in ~25% decrease in Notch1 reporter activity (fig. S1C). These results support a robust, two-component enhancement of the LPΔP mutant, with MAFB playing the major role. Consistent with MafB being the limiting component, both Ets1 and Ets2 are highly expressed in after β-selection thymocytes, whereas MafB expression is much lower in these populations [ (24)].

MAFB increases penetrance and decreases latency of Notch-induced T-ALL

To assess the effect of MAFB in vivo, we assayed its ability to influence the leukemogenic activity of LPΔP in a retroviral transduction/bone marrow transplant (BMT) model (Fig. 2A) (25). Notch LPΔP was expressed from a murine stem cell virus retroviral vector (MigR1) coexpressing green fluorescent protein (GFP) (26), whereas MAFB retroviral particles coexpressed truncated nerve growth factor receptor (NGFR) (8), thus providing the ability to distinguish between them.

Fig. 2 MAFB enhances T-ALL onset of a weak Notch gain-of-function allele.

(A) Schematic representation of experimental design for retroviral transduction of 5-fluorouracil (5-FU)–treated BM progenitors. When only a single cDNA was expressed (MAFB-only, ICN-only, or LPΔP-only), an equivalent dose of the retroviral vector (either MigR1 or MigR1-NGFR) was used to normalize the total retroviral titers. (B) Peripheral blood analysis of the four cohorts of transplanted mice (ICN, n = 8; MAFB n = 9; LPΔP, n = 14; and LPΔP+MAFB, n = 15) 4 weeks after BMT. Representative populations shown were gated for live, singlet, and Lin cells. NGFR is the surrogate marker for MAFB transduction, and GFP is the surrogate marker for Notch-mutant transduction. (C) Kaplan-Meier plot indicating percent T-ALL–free mice (that is, insert space between free and mice) after BMT. Additional animals that succumbed to non–T-ALL conditions, such as irradiation poisoning (MAFB-only, n = 1; LPΔP-only, n = 1) and BM failure or anemia (LPΔP-only, n = 1), were excluded from the analysis. (D) WBC counts in the peripheral blood of the mice in each of the experimental cohorts. A diagnostic threshold for tumor onset of 40 million/ml of WBCs and circulating double-positive (DP) T cells in the peripheral blood was used as a benchmark for T-ALL, as previously described (8, 25). ***P ≤ 0.0005 by Mantel-Cox test (C) or unpaired t test (D).

Reconstituted mice were monitored at various times after BMT. At 4 weeks after BMT, CD4+CD8+ T cells accumulated in the peripheral blood of ICN and LPΔP+MAFB mice (Fig. 2B). This population also expressed the GFP+ (ICN or LPΔP) or GFP+NGFR+ (LPΔP+ MAFB+) surrogate markers, indicating that they originated from the transduced progenitors. Consistent with previous data, mice reconstituted with ICN1 succumbed to T-ALL with a 100% penetrance within 80 days after BMT, with a median survival time of 58 days, whereas mice receiving LPΔP alone developed T-ALL with a penetrance of 36% and latency greater than 100 days (Fig. 2C) (8, 25, 27). None of the mice receiving MAFB alone succumbed to leukemia. In contrast, 16 of 16 mice reconstituted with LPΔP+MAFB succumbed to T-ALL with a median survival of 72 days. Consistent with the Kaplan-Meier plots, the white blood cell counts (WBCs) of the LPΔP+MAFB mice were significantly higher than the LPΔP-only transduced mice (Fig. 2D), as were the spleen weights and percentage of infiltrating tumor cells in tissues compared to the LPΔP-only mice (fig. S2). Together, these data show that MAFB strongly enhances the ability of weakly activating Notch1 mutants to induce T-ALL.

Suppressing MafB inhibits Notch signaling and T-ALL cell growth

To understand the function of MafB in T-ALL, we suppressed MafB in T6E cells with fluorophore-conjugated small interfering RNAs (siRNAs). T6E cells were chosen because they express high levels of MafB (fig. S1D). At 48 hours after treatment, >90% of T6E cells expressed the fluorophore (fig. S3A), and both siRNAs reduced MafB expression by ~50% (Fig. 3A). There was a concomitant decrease in the direct Notch1 targets Hes1, Myc, Notch3, and Deltex1 as well as the Myc target CAD (Fig. 3B). In contrast, neither the expression of Hey1 nor that of the negative control, GAPDH, changed (Fig. 3B and fig. S3B).

Fig. 3 MafB-deficient T-ALL cells down-regulate Notch targets and perform poorly in competitive culture conditions.

(A) MafB expression measured by quantitative reverse transcription polymerase chain reaction (qRT-PCR) in T6E cells 48 hours after treatment with indicated siRNAs relative to that in cells cultured with transfection reagent (untreated). *P ≤ 0.05, **P ≤ 0.005 by unpaired t test. (B) Relative expression analysis by qRT-PCR of Notch signaling and Myc signaling (CAD) targets and expression controls (GAPDH) in MafB-siRNA#1–treated cells relative to each in Scrmb-siRNA–treated cells. (C) ChIP assay measuring the occupancy of P300 at the Hes1 promoter and at the NDME in MafB-shRNA#3–treated T6E cells. Genomic DNA sequence of GAPDH is used as the nonspecific control. ***P < 0.0005 by unpaired t test. (D) Pie chart representation of Notch-dependent gene expression stratified on the basis of differential H3K27Ac abundance measured by ChIP-seq analysis of T6E cells transfected with shScrmb or shMafB#3. Of the 59 Notch-dependent transcripts, as determined by a GSI-washout experiment, 37 have nearby MafB-dependent H3K27Ac regions, which display a fold change of ≤−1.5 (dark gray portion of the pie chart). (E) Genome tracts of individual examples of MafB-dependent H3K27Ac for Notch target genes (Dtx1, top; Myc, bottom) in T6E cells transfected with shScrmb or shMafB#3. kb, kilobase. (F) Schematic representation of the experimental design for MafB shRNA–treated T-ALL cells. The cells were sorted for their respective markers and subsequently cultured in equal numbers directly afterward. Every 3 days, the percentage of the cellular population bearing each marker was determined in an aliquot of the competition culture. FACS, fluorescence-activated cell sorting. (G) Left: Competition assay of T6E cells treated with GFP+ or NGFR+ shScrmb plated in equal numbers (2.5 × 105) at day 0 (no significant difference; n = 3). Middle and right: Competition assay of T6E cells treated with GFP+ shScrmb or one of two NGFR+ MafB shRNAs plated in equal numbers (2.5 × 105) at day 0 (n = 3). P = 0.02 and 0.03, respectively.

To persistently inhibit MafB, we transduced T6E cells with retroviral vectors expressing MafB short hairpin RNAs (shRNAs). Both shRNAs suppressed MafB as well as Hes1 and Myc expression (fig. S3, C to E). Because NTCs recruit p300 to activate transcription (28, 29), we performed chromatin immunoprecipitation (ChIP)–PCR for this protein on T6E cells treated with MafB shRNAs to determine p300 occupancy at the 5′ promoter region of Hes1 and at the 3′ Notch-dependent Myc enhancer (NDME), both of which are known to recruit p300 in a Notch-dependent manner (30). When compared to scrambled shRNA (shScrmb)–treated cells, the loss of MafB significantly reduced p300 interaction at both the Hes1 promoter and the NDME (Fig. 3C).

To gain further insights into the effect of MafB on Notch1-dependent transcription, we performed ChIP sequencing (ChIP-seq) for the histone acetylation mark H3K27Ac on T6E cells treated with shScrmb or MafB shRNAs. We chose H3K27Ac because changes in Notch occupancy produce dynamic alterations in H3K27Ac levels at both enhancers and promoters of Notch1-dependent genes in T-ALL (31). To identify Notch1-dependent genes in T6E cells, we performed a GSI-washout experiment (32) to compare expressed genes in the Notch-off to Notch-on states (31, 32). Using stringent criteria (see Materials and Methods), we identified 59 Notch1 positively regulated transcripts (table S2). We then analyzed the effect of MafB knockdown on the putative enhancers, marked by H3K27Ac of these 59 Notch positively regulated transcripts. Suppression of MafB decreased H3K27Ac by at least 1.4-fold in 37 of 59 (62%) Notch1-responsive genes (Fig. 3D and table S2). These included the well-known direct Notch targets Dtx1, Myc, Notch3, NRarp, Notch1, and IL2RA (Fig. 3E, fig. S3F, and table S2). Although Hes1 expression significantly decreased in the GSI-washout condition, the decrease in H3K27Ac abundance did not meet our 1.4-fold threshold, because it decreased by only 1.2-fold; nevertheless, there was a decrease in H3K27Ac loading at the 5′ site of the Hes1 transcription start site (fig. S3F). Furthermore, shMafB is not a global inhibitor of transcription, given that many genes, such as Trib1, which is not a Notch1 target, showed no change in H3K27Ac (fig. S3F). Together, these data suggest that inhibiting MafB decreases Notch-mediated transcription at a significant number of direct Notch target genes (binomial P value = 0.06).

To test the fitness of T-ALL cells treated with MafB-shRNAs, we performed an in vitro competition assay comparing the shMafB-transduced T6E cells to cells receiving the shScrmb control (Fig. 3F). Unlike cells transduced with scrambled control vectors, the shMafB-transduced cells were outcompeted by the shScrmb cells (Fig. 3G). These findings suggest that inhibiting MafB represses T-ALL cell growth by inhibiting the expression of crucial Notch1 targets, such as Myc, Dtx1, and Hes1.

Increased MAFB expression promotes growth and subverts the effects of GSI in T-ALL cells

To better understand the effect of increased MAFB expression in T-ALL cells, we retrovirally expressed human-MAFB in T6E cells. Because of the poor quality of commercial anti-MafB antibodies, we appended 5′ HA or FLAG molecular tags to the MAFB coding sequence. We selected cell lines with different amounts of MAFB overexpression; HA-MAFB was slightly higher, whereas FLAG-MAFB was much higher than endogenous murine MafB protein (Fig. 4A). In both cell lines, the expression of the Notch1 target genes Hes1 and Myc was significantly increased (Fig. 4, B and C), and this was accompanied by increased cell proliferation (fig. S4A).

Fig. 4 Ectopic MAFB expression in T-ALL cells enhances Notch target gene expression and limits the effect of GSI treatment.

(A) Western blotting for MafB (with anti-MafB antibody; Abcam) and loading control (GAPDH) in whole-cell lysates (10 μg) from T6E cells expressing Flag-MAFB or HA-MAFB. MigR1, empty vector. (B and C) Hes1 (B) and Myc (C) expression measured by qRT-PCR in T6E cells overexpressing MAFB. EF1a, elongation factor 1α. (D and E) Hes1 (D) and Myc (E) expression measured by qRT-PCR in T6E cells treated with dimethyl sulfoxide (DMSO) or GSI (0.1 μM) for 24 hours and subsequently washed out for 6 hours. Black bars, control cells; gray bars, MAFB-overexpressing cells. (A to E) n = 3 to 6 experiments. (F) Proliferation of MigR1-transfected control or MAFB-overexpressing T6E cells (2.5 × 105) plated in a medium containing DMSO or GSI (10 nM). n = 2 experiments with three technical replicates per experiment. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.0005, ****P ≤ 0.00001 by unpaired t test.

We tested whether increased MAFB expression could compensate for decreased Notch1 signaling in T-ALL cells treated with GSIs. To determine the appropriate dose of GSI treatment for T6E cells, we titrated the concentration of GSI and determined that the range 0.1 to 0.01 μM reduced Hes1 expression by ~50% (fig. S4B). We next performed a GSI-washout experiment (33). As expected, GSI treatment down-regulated the Notch targets Hes1 and Myc, which were up-regulated after GSI washout (Fig. 4, D and E, black bars). However, the effects of both the GSI treatment and the washout were blunted in T6E cells expressing HA-MAFB (Figs. 4, D and E, gray bars). We next assayed cell growth under GSI treatment. As anticipated, MigR1 cells treated with DMSO continued to proliferate, whereas those treated with GSI experienced growth arrest as early as 48 hours after treatment. On the other hand, the MAFB-overexpressing cells continued to expand (Fig. 4F). These findings support a role for MAFB to enhance Notch1 function and suggest that T-ALL tumors expressing high amounts of MAFB may be more resistant to GSI-based therapies.

MAFB enhances Notch signaling by interacting with ETS

MAFB and ETS2 were initially detected by our luciferase Notch reporter screen; thus, we tested whether MAFB directly binds to the Rbpj/Notch1 sites in the TP1 reporter. For this, we used oligo-IP (OIP) assays, where a biotinylated oligonucleotide containing the consensus Rbpj/Notch1 (CSL) binding sequence (CGTGGGAA) found in the TP1 reporter was added to the lysate of U2OS cells transfected with vector control, MAFB, ETS2, or MAFB and ETS2 (Fig. 5A). Of particular relevance, we identified the core DNA sequence for ETS factor binding (C/A)GGAA(G/A) (34) interspaced between the Rbpj binding sequences of the TP1 reporter (underlined in Fig. 5A). The OIP assay revealed ETS2 binding to the TP1 oligo; however, MAFB bound only in the presence of exogenous ETS2, which is only marginally expressed in U2OS cells (Fig. 5B). This same dependency of MAFB on ETS factors was confirmed for ETS1 (fig. S5A).

Fig. 5 MAFB interacts with ETS proteins and recruits cofactors to Notch/RBPj binding sites.

(A) Schematic representation of the experimental design for the OIP assay in U2OS cells. A biotin-tagged oligo from the TP1 Notch reporter was used as the bait to immunoprecipitate proteins interacting at the CSL or EBS regions. (B) Representative results of OIP assay in U2OS cells transfected with MigR1, ETS2, Flag-MAFB, or ETS2 + Flag-MAFB and probed with ETS2 and Flag antibodies after OIP. Ten percent loading of nuclear lysates indicates protein abundance; GAPDH is the loading control. (C) Schematic representation of the tagged MafB and MafB ΔbZIP mutant. Right: Western blotting (WB) on mixed lysates of HA-MAFB and HA–MAFB ΔbZIP immunoprecipitated from doubly transduced T6E cells to show relative protein size and HA-pulldown efficiency. (D) Representative results of HA- or FLAG-IP assessing MAFB interactions with ETS1, ETS2, P300, and PCAF in T6E T-ALL cells. (E) Representative HA-IP in T6E cells transduced with MigR1, HA-MAFB, or HA–MAFB ΔbZIP and probed for HA, ETS2, PCAF, and P300. The background bands in the HA 10% loading control likely result from the blot being probed and restriped multiple times. (F) Representative HA-IP in T6E cells transduced with HA-MafB and probed for cleaved-Notch1 (Val1744). (G) Representative FLAG-IP in T6E cells transduced with MigR1, FLAG–MAFB ΔbZIP, or FLAG-MAFB and probed for RBPJ and FLAG. (H) Notch TP1 reporter assay in T6E cells transfected with LPΔP and the indicated plasmids. EV is the vector control. **P ≤ 0.005 by unpaired t test; n = 3. (I) Kaplan-Meier plot comparing percent of T-ALL–free survival after BMT with progenitors transduced with LPΔP-only, LPΔP+MAFB, or LPΔP + MAFB ΔbZIP domain deletion mutant. LPΔP, n = 6; LPΔP+MAFB, n = 12; LPΔP+MAFB ΔbZIP, n = 10. Animals that succumbed to non–T-ALL conditions, such as BM failure/anemia (LPΔP-only, n = 1), were excluded from the analysis. ***P ≤ 0.0001 by Mantel-Cox test.

These data also suggest that lack of Ets factors will limit MafB-dependent enhancement of Notch signaling and delay T-ALL onset in vivo. We tested this using the BMT model described in Fig. 2; however, for these studies, we transplanted the BM from ETS1−/− donor hematopoietic progenitors (35) that were transduced with ICN1 or LPΔP+MAFB. Although Ets1 deficiency had minimal effects on T-ALL onset in cells transduced with ICN1, the loss of a single Ets factor significantly decreased T-ALL penetrance in recipient mice, with median survival increasing from 72 days in wild-type (WT) BM donors transduced with LPΔP+MAFB to 160 days in ETS1−/− BM donors transduced with LPΔP+MAFB (fig. S5B). These results suggest that the amount of Ets may influence disease latency, even in the presence of MafB.

MAFB interacts with ETS2 through its bZIP domain and functions to recruit transcriptional coactivators

To determine whether MAFB directly binds ETS factors in T-ALL cells, we transduced T6E cells with epitope-tagged MAFB and performed IPs to identify binding partners (Fig. 5C). MAFB strongly bound both ETS1 and ETS2, which are highly homologous (16). Previous studies of the MAFB homolog, MAFA, showed that it forms complexes with important transcriptional coactivators with histone acetyltransferase activity, including the P300/CBP-associated factor (PCAF) complex (36) and p300 (37), both of which are recruited to NTCs where they activate transcription (38, 39). To test whether MAFB interacts with PCAF and P300, we performed IPs in T6E cells expressing FLAG-tagged MafB. Our results indicate that MafB interacts with both PCAF and P300 in T-ALL cells (Fig. 5D), suggesting a mechanism through which it enhances Notch signaling.

We next set out to determine how MAFB interacts with ETS2. We used site-directed mutagenesis to remove the bZIP domain from MAFB (Fig. 5C) (9, 11, 23). Using T6E cell lysates, we first verified the deletion and the size of the mutant (Fig. 5C, right). We then performed IP experiments (Fig. 5E), which showed that deleting the MAFB bZIP domain markedly diminished the MAFB-ETS2 interaction. MAFB ΔbZIP also markedly decreased p300 binding; however, the bZIP domain was not required for the PCAF interaction (Fig. 5E). Previous ChIP-seq analysis showed that ETS binding sites are enriched within 250 base pairs (bp) of Notch1/RBPJ binding sites, thus placing ETS factors in the immediate vicinity of the NTC (40). We tested whether MAFB can interact with components of the NTC, specifically the cleaved ICN1, as detected by the Val1744 antibody and RBPj. We found that MAFB coimmunoprecipitated in T6E cells with cleaved Notch1 (Fig. 5F) and RBPj (Fig. 5G) and that this interaction with the NTC was lost when the MAFB bZIP domain was deleted (Fig. 5G). These findings further highlight the ability of MAFB-ETS complexes to cooperate with the NTC to activate transcription.

To determine the effect of the MAFB ΔbZIP mutant on Notch transcriptional activity, we tested its effect in T6E cells using the TP1 reporter. The MAFB ΔbZIP mutant decreased reporter activity by more than 60% (Fig. 5H). When ETS2 was cotransfected in the assay with the MAFB ΔbZIP mutant, the overall Notch activity was increased, but the loss of the bZIP domain did not significantly enhance Notch signaling more than when ETS2 or MAFB were added alone, thus suggesting that MAFB-mediated enhancement of Notch transcription occurs through its bZIP domain. We next assayed the MAFB deletion mutant in vivo using the BMT model. As before, MAFB greatly decreased the latency and increased the penetrance of T-ALL development. In contrast, the MAFB ΔbZIP mutant markedly limited the ability of MAFB to enhance LPΔP leukemogenicity (Fig. 5I). Together, these data show that the bZIP domain of MAFB is required for MAFB to enhance LPΔP activity both in vitro and in vivo.

MAFB enhances Notch signaling in human T-ALL cells

To determine whether MafB served a similar role in human T-ALL, we determined the levels of MAFB in three T-ALL cell lines (Fig. 6A). Of these, KOPT-K1 had the highest levels of MAFB expression. KOPT-K1 cells were treated with shRNAs designed against human MAFB, which decreased MAFB expression by more than 50% (Fig. 6B). The effect of suppressing MAFB in human KOPT-K1 cells was similar to murine T6E cells as the expression of the direct Notch target genes HES1, DTX1, and MYC was suppressed (Fig. 6C). Suppression of MAFB did not affect the expression of ETS factors or ZMIZ1 nor did it alter the occupancy of ETS2 at the 5′ HES1 promoter in T-ALL cells (fig. S6, A to D). Likewise, KOPT-K1 cells treated with anti–MAFB-HU-shRNAs were at a disadvantage in cell growth assays when compared to cells treated with shScrmb (Fig. 6D).

Fig. 6 Loss of MAFB suppresses Notch target gene expression and cell proliferation in human T-ALL.

(A) MAFB mRNA expression in human T-ALL cell lines (U2OS cells used as negative control). (B) MAFB mRNA expression after MAFB knockdown in KOPT-K1 cells by each of the two shRNAs. (C) Notch target gene expression in sorted KOPT-K1 cells 72 hours after transduction with shRNA against MAFB relative to KOPT-K1 cells treated with Scrmb-shRNA. (D) Proliferation of sorted KOPT-K1 cells after transduction with shRNAs against MAFB. Growth is compared to cells treated with Scrmb-shRNA (n = 3). (E) Representative Western blotting for ETS1 and ETS2 in human T-ALL cell lines. GAPDH is the loading control. (F) Representative FLAG-IP in KOPT-K1 cells transduced with MigR1 or FLAG-MAFB and probed for FLAG, ETS2, and RBPJ. Ten percent loading of lysates indicates RBPJ and ETS2 abundance. GAPDH is the loading control. **P ≤ 0.005, ***P ≤ 0.001.

We next tested whether ETS factors also played a similar role in human T-ALL cells. First, we confirmed that three human cell lines expressed ETS1 and ETS2 (Fig. 6E). Next, we tested whether MAFB binds ETS2 in human T-ALL cells. We transduced KOPT-K1 cells with FLAG-tagged MAFB and performed IPs to detect interactions and found that MAFB strongly bound both ETS2 and RBPJ in KOPT-K1 cells (Fig. 6F). This suggests that a similar MAFB-ETS2 pathway enhances Notch signaling in human T-ALL.


The finding that human and murine T-ALLs with activating NOTCH1 mutations frequently require persistent Notch signaling for growth and survival led to the idea that Notch is a druggable target in this disease. Although multiple studies support this idea, one paradox is that the common NOTCH1 mutations in T-ALL patients are weak oncogenic drivers in murine T-ALL models. This raises the question of whether there are additional molecules that potentiate Notch signaling, which could identify additional therapeutic targets.

Here, we set out to identify proteins that enhance the activity of the common Notch1 mutations in T-ALL patients. Our screen identified multiple genes, some of which were previously verified to be important in Notch signaling (19, 20). We chose to focus on MAFB and ETS2. These transcription factors have well-established roles in myeloid and lymphoid development (15, 41), respectively; however, their involvement in T-ALL is poorly understood. In reporter assays, MAFB and ETS2 demonstrated strong synergy and enhanced Notch activity to a level on par with MAML1, a core component of the NTC. In our functional studies, MAFB seems particularly important because expressing MAFB along with the weak oncogenic NOTCH1 mutant LPΔP in BM progenitors accelerated T-ALL onset and increased T-ALL penetrance. Inhibiting MAFB, by either siRNA or shRNA, down-regulated the expression of multiple direct Notch targets, including MYC, DTX1, and HES1, in both murine and human T-ALL cells and also decreased the proliferation of murine and human T-ALL cells. In contrast, overexpressing MAFB enhanced the expression of these Notch targets and promoted T-ALL growth. Thus, we propose that the combination of MAFB and ETS2 has the potential to amplify the activity of weak activating Notch1 mutations. Although multiple Notch1 targets are inhibited by blocking MAFB, additional analyses are needed to determine whether MAFB influences all direct Notch1 target genes. It is unlikely that withdrawal of MafB broadly represses transcription because the expression of many genes, such as Trib1, was not affected by MafB knockdown.

How MAFB boosts the signaling of mutant Notch receptors was not immediately obvious. A previous work characterized MafB as a differentiation factor, with roles in monocyte-to-macrophage differentiation (42). In T cells, MAFB expression is low (43); however, data from published databases and our own analysis of cell lines, patient primagrafts, and primary tumors show that MAFB is frequently expressed in T-ALL cells and primary patient samples. There was no correlation with Notch1 mutation subtypes, and there was no evidence of MAFB or ETS1/2 genetic amplification in primary T-ALLs (21, 22). Thus, we hypothesize that in these tumors, MAFB expression, driven by unknown mechanisms, potentiates both Notch signaling and its oncogenic capacity. Our findings also suggest that in other tumors with weakly activating Notch1 mutations, such as chronic lymphocytic leukemia, additional cofactors analogous to MAFB may be required to potentiate oncogenic Notch1 signals.

MAFB is a multidomain transcription factor that can bind DNA directly and form protein-protein interactions (9, 44). In our T-ALL studies, MAFB did not appear to bind DNA independently; instead, it appears to either bind cooperatively with ETS transcription factors or be tethered to DNA through the ETS transcription factors. ETS binding sites are common and in close proximity with RBPJ binding sites (40). Because MAFB is known to directly interact with ETS1 (23), we believe that MAFB is recruited near sites of NTC formation through a similar interaction with ETS2 or ETS1. We found that MafB interacts with both P300 and PCAF, which are important for Notch-induced transcriptional activation (38), and that the loss of MafB decreases the occupancy of p300 at the Hes1 and Myc Notch-dependent enhancers. The bZIP domain of MafB is required for this activity, thus bringing PCAF and P300 and by connection, the Notch/RBPj/MAML complex into a larger protein complex capable of high transcriptional output. This MAFB-ETS2 synergy and its ability to enhance Notch signaling are also conserved in human T-ALL cells. A recent work showed that the T cell transcription factor Zmiz1 interacts with Notch1 and regulates the expression of oncogenic Notch targets such as Myc (45), indicating that enhancing Notch signaling by cofactors is an important step in T-ALL onset. In our studies, suppressing MafB did not affect Zmiz1 expression. Circulating double-positive T cells, which express high levels of ETS factors, are observed in the BMT model of NOTCH1 LPΔP–induced T-ALL (8); however, these mice develop T-ALL at a low frequency. From the results of our study, we predict that when MAFB is expressed at sufficient levels, it creates an environment where transcriptional cofactors are recruited at a higher-density near sites of Notch transcriptional targets, thus raising Notch1 signaling output to levels that exceed the threshold for oncogenic transformation.

In summary, our findings show that the MAFB-EST2 interaction enhances Notch1 signaling in a leukemic setting by supporting the higher expression of Notch targets. These findings suggest that the MAFB-ETS2 axis may serve as a therapeutic target in T-ALL. Our findings also raise issues regarding the efficacy of GSI therapy in tumors with increased MAFB abundance, because our data show that cells expressing high amounts of MAFB were less sensitive to GSI treatment. Thus, strategies that decrease MAFB expression or formation of the MAFB-ETS2 complex may enhance the sensitivity of Notch-dependent T-ALL cells to GSI therapy.


Cell culture

T-ALL cell lines were cultured in RPMI 1640 (Invitrogen) supplemented with 10% fetal bovine serum (FBS) (HyClone), 2 mM l-glutamine, 1% nonessential amino acids (Gibco), 1% sodium pyruvate (Gibco), and 2-mercaptoethanol [0.0005% (v/v); Sigma), with antibiotics. U2OS cells were maintained in Dulbecco’s modified Eagle’s medium (Invitrogen) with 10% FBS (Gibco) and antibiotics. Cells were grown at 37°C in 5% CO2. Retroviral transduction, sorting, and expression analysis of T-ALL cells were performed as described previously (3). The GSI, compound E (Calbiochem, 565790), was titrated on T6E cells for 12 hours at concentrations ranging from 1 μM to 1 nM. GSI-washout experiments were performed as described previously (32) after 24 hours of culture in 0.1 μM GSI.

Constructs and retroviruses

MigR1 (46), MigR1-NGFR, MigR1-Notch1-L1601P, and MigR1-Notch1 LPΔP are described previously (8). Sport6-MafB (780 bp; MafB coding sequence) and Sport6-ETS2 (822 bp; ETS2 coding sequence) were obtained from the library screen. The 5′ HA or FLAG tags added to MafB were generated by PCR. MafB deletions were generated using QuikChange II (Agilent Technologies), and primer design was based on the QuikChange Primer Design program (

Luciferase screen and reporter assays

The cDNA screening strategy involved the use of three key components: (i) a pcDNA3 plasmid encoding a modestly strong NOTCH1 gain-of-function mutant, LPΔP, driven from a CMV promoter (40 ng of cDNA per well), (ii) a Notch firefly luciferase reporter (TP1) containing 12 CSL binding sites (50 ng of cDNA per well), and (iii) a preplated cDNA library cloned into the Sport6 plasmid (40 ng of cDNA per well). A MAML1 cDNA was the positive control for each screen plate, whereas empty vector and a DTX1 cDNA were background and negative controls (40 ng of cDNA per well), respectively. DNA spotting was performed using a Matrix PlateMate (Thermo Fisher Scientific) for the control wells. The Matrix WellMate (Thermo Fisher Scientific) was used to dispense the transfection mix containing the reporter and Notch1 mutant plasmids in combination with transfection reagent (FuGENE6, Promega), which were added to the wells (4000 U20S cells per well) after a 30-min incubation. Luminescence was measured 48 hours after plating using Britelite plus (PerkinElmer) luciferase reagent with LJL BioSystems Analyst HT96-384. Methods for screen validation are provided in the Supplementary Materials.

Luciferase screen validation

For the validation of screen candidates and independent luciferase assays, 4000 U2OS cells per well were seeded on Corning opaque 384-well plates, and FuGENE6 transfection mix was prepared in Opti-MEM (Gibco) serum-free medium with three plasmid components: (i) 50 ng per well of Notch TP1 firefly reporter plasmid, (ii) 40 ng per well of pcDNA3-LPΔP plasmid, and (iii) 40 ng per well of screen-component cDNAs cloned into pCMVSport6 (for example, MafB or ETS2). pcDNA3-MAML1 and Flag-CMV-DTX1 plasmids (40 ng per well) served as positive and negative controls, respectively. Five nanograms of pRL-TK Renilla luciferase was used as internal control plasmid. After 20 min at room temperature, 20 μl of the transfection reaction mix was added to the cells by a multichannel pipette. Twenty-four wells were analyzed for each individual transfection sample set. After a 48-hour incubation with the transfection mix, 35 μl per well of Britelite plus luciferase reagent was added by a multichannel pipette; luminescence was measured by LJL BioSystems Analyst HT96-384 (LJL BioSystems Inc.). Stop & Glo buffer and Renilla luciferase reagent (Promega) were used to assess transfection efficiency. At least three individual repeats were performed for each experiment.

Human T-ALL data analysis and statistical analysis

Normalized microarray expression analysis for the Haferlach data set was obtained from the Oncomine database, and Zuurbier data set was provided by J. P. P. Meijerink. Each set was separately analyzed for the expression of MafB, Notch1 receptor, and Notch1 target genes: Hes1, Deltex, and Myc. Each sample that displayed higher than twofold Notch receptor and Notch target expression was parsed from the total data set. MafB expression was cross-referenced for any individual sample that showed a Notch enhancement signature. The comparative values of Notch1, Notch targets, and Mafb were plotted as a log2 median value for all the samples that show Notch enhancement in the Haferlach and Zuurbier data sets. Statistical analysis was performed using Prism 6 (GraphPad). Survival curves were computed using Kaplan-Meier analysis, and comparison of survival curves was performed using Mantel-Cox test provided through Prism. Predicted experimental mouse numbers for the Notch1-LPΔP and MafB experiment were powered (0.8). z score for Notch1 gain-of-function screen was determined by the following formula: z score = 1 − [(3SD + 3SD+)/(Av+ − Av)]. An unpaired t test P value of less than 0.05 was considered to be significant in all experiments, unless noted otherwise; *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.0005, ****P ≤ 0.00001.

Quantitative PCR

RNA was extracted using Qiagen RNeasy Mini and Micro Kits. cDNA was synthesized from RNA with the SuperScript III kit (Invitrogen). Transcripts were amplified with SYBR Green PCR reagent (Applied Biosystems), and quantitative PCR (qPCR) was performed on the ABI Prism 7900HT system (Applied Biosystems). mRNA quantities, either in absolute or relative quantification, were normalized to elongation factor 1α and GAPDH, respectively. Primer3 software was used for primer design, and sequences are provided in table S3.

siRNA and shRNA

RNA interference was carried out with siRNA duplexes designed and then screened for specific and effective knockdown of target genes. Fluorescence-conjugated duplexes (n = 3) targeting MafB were ordered directly as ON-TARGETplus siRNA from Thermo Fisher Scientific/Dharmacon. siRNA methods are provided in the Supplementary Materials. For shRNA suppression, two different MafB-targeting shRNAs and a shScrmb were purchased from Open Biosystems (sequences are provided in table S4). For siRNA control transfections, nontargeting siRNA and siGLO Green were purchased from Thermo Fisher Scientific/Dharmacon. For transfections of cells, siRNA duplexes were resuspended in siRNA buffer (Dharmacon) and added to the cell growth medium for 12 hours with siIMPORTER transfection reagent (Millipore) as per the manufacturer’s instructions. Forty-eight to 72 hours after transfection, RNA was harvested from cells with RNeasy Mini kit (Qiagen), and 500 ng of total RNA was used in qPCR analysis. For shRNA suppression, two different MafB-targeting shRNAs and a shScrmb were purchased from Open Biosystems (sequences are provided in table S3 and the Supplementary Materials) and were subcloned into the pLMP-GFP or pLPM-NGFR vector. For viral transduction, cells were centrifuged with viral supernatant and hexadimethrine bromide (8 μg/ml) (Sigma) at 2500 rpm for 90 min at 25°C. FACS-purified shRNA-treated cells were analyzed for growth or gene expression at the indicated times after transduction. For growth curves or competition assays, 5 × 105 sorted cells were seeded.

ChIP-seq analysis

ChIP-seq reads were aligned to mm10 using Burrows-Wheeler Aligner (47) and filtered to remove PCR duplicates and multimapped reads. All reads were postfiltered by known ENCODE blacklist regions. Peak calling was performed using MAC2 (version 2.0.9) (48) with the following parameters: --no model, --shift size = (1/2 estimated fragment length), --keep-dup = 1, and a false discovery rate (FDR) threshold of 1 × 10−6. ChIP-seq display files were generated using SAMtools, BEDTools, and UCSC utilities. Scaling for all ChIP-seq tracks in figures is equal to local fragment coverage × (1,000,000/total count).

The peaks were associated to their most proximal gene as defined in Ensembl GRCm38 transcript model using BEDTools and HOMER-annotatePeaks (version 4.8) (49). H3K27Ac signal was compared between the GSI and GSI-washout conditions on the 1500 base pair (bp) regions flanking the peak summits. Regions (1500 kb) flanking the summits of peaks were merged across GSI and GSI-washout conditions using BEDTools-merge (version 2.25.0). In each library, peak-filtered H3K27Ac signal was quantified and normalized to fragment per kilobase per million reads (FPKM). The logarithmic fold change of H3K27Ac load on with nonzero read counts in at least one condition was calculated as log2FPKM of the GSI versus GSI-washout condition with a pseudocount of 1, and the regions with absolute fold change greater than or equal to log2(0.5) were indicated in table S2 and MafB-dependent.

Mice and BM transduction

BM transductions and transplantation into lethally irradiated recipients were performed as described previously (26, 50). Recipient mice used in these experiments were 6- to 8-week-old C57BL/6 female mice obtained from Charles River Laboratories. In cases where two retroviruses (for example, MigR1-LPΔP and MigR1-NGFR-MafB) were used to transduce the BM progenitors, the total viral titer and multiplicity of infection were adjusted to reflect previously used T-ALL Notch-mutant virus titers (8). Mice were maintained on antibiotics in the drinking water for 2 weeks after BMT, and peripheral blood was drawn every 2 weeks to monitor blood counts and evaluate the presence of circulating immature T cells by flow cytometry. Mice with WBC counts of >4.0 × 106/ml and a body condition score of ≤2 were euthanized, and tissues were harvested for flow cytometry and histology (hematoxylin and eosin) analysis. All mice were housed in specific pathogen–free facilities at the University of Pennsylvania. Experiments were performed according to the guidelines from the National Institutes of Health with approved protocols from the University of Pennsylvania Animal Care and Use Committee.

Flow cytometry

Flow cytometry was performed on BD LSR II, cell sorting was performed on BD Aria II, and FlowJo software was used for analysis. The peripheral blood (bimonthly), spleen, thymus, and BM (terminal analysis) were harvested, and cell suspensions were generated from mice transduced as described above. Antibodies used in staining of the tissues include: CD45.2 (104, BD Biosciences), CD25 (PC61, BioLegend), CD44 (IM7, eBioscience), CD3 (17A2, eBioscience), CD4 (RM4-5, eBioscience), CD8 (53-6.7, eBioscience), CD19 (1D3, eBioscience), CD11b (M1/70, eBioscience), F480 (BM8, eBioscience), and Gr-1 (RB6-8C5, eBioscience). 4′,6-Diamidino-2-phenylindole (DAPI) was used for Live/Dead determination. Transduced T6E cell lines were sorted on the basis of internal GFP fluorescence or surface staining with anti-NGFR–biotinylated antibody generated in-house from the 8737 hybridoma line.

Western blotting and IP analyses

Whole-cell lysates were prepared with radioimmunoprecipitation assay (RIPA) buffer or FLAG-IP lysis buffer (50 mM tris, 150 mM NaCl, 1 mM EDTA, 0.5% NP-40, and 10% glycerol), with protease inhibitor tablets (cOmplete, Roche). Protein concentration was determined with the Bio-Rad protein assay dye reagent (Bio-Rad). Proteins were separated using SDS–polyacrylamide gel electrophoresis and wet-transferred to polyvinylidene difluoride membranes. Blots were visualized with SuperSignal West Pico Chemiluminescent or SuperSignal West Femto Chemiluminescent substrate (Thermo Fisher Scientific). Antibodies used for Western blotting were cleaved Notch1 (Val1744) antibody (Cell Signaling Technology, no. 2421), RBPJSUH (Cell Signaling Technology, D10), MafB (Santa Cruz Biotechnology, P20), rabbit monoclonal antibody (Abcam, ab66506), GAPDH (Santa Cruz Biotechnology, FL-335), ETS1 (Santa Cruz Biotechnology, C-20), ETS2 (Sigma, E3783), PCAF (Santa Cruz Biotechnology, E-8), P300 (Thermo Fisher Scientific, RW109), HA (Covance, MMS-101P), FLAG (Sigma, M2), and secondary anti-mouse horseradish peroxidase (HRP) (Pierce) or anti-rabbit HRP (Pierce).

IP experiments were performed using HA-probe agarose beads (Santa Cruz Biotechnology) or anti-FLAG agarose beads (Sigma, M1). Samples were loaded, and Western blot analysis was performed as described above. OIP experiments were performed as described previously (51). Oligonucleotide sequences for the OIP experiments are provided in table S3. For each pulldown, 0.5 to 1 mg of fresh protein lysate was used and combined with 25 to 30 μl of beads for incubation overnight at 4°C with rotation. Beads were washed four times in lysis buffer with increasing concentration of NaCl [200 nM (2×) and 300 nM (2×)]. The IP experiments depicted in Figs. 5 and 6 were conducted in similar fashion; however, the cells were lysed with FLAG-IP lysis buffer and washed only three times, once with lysis buffer and twice with lysis buffer with higher NaCl concentration (150 nM). Protein was released from washed beads by addition of 2× Laemmli buffer and boiled for 10 min at 95°C.

Microarray analysis

Mouse Gene 2.0 ST Affymetrix array CEL files were imported, normalized, and summarized using robust multichip average and median-polish algorithms, respectively, using the Bioconductor “oligo” package (52). Limma (53) was used for differential gene expression on triplicate samples of GSI and GSI-washout conditions. Package mogene20sttranscriptcluster.db in Bioconductor was used for annotation. Genes with greater than twofold change and a Benjamini-Hochberg FDR of <0.1 were called as differentially expressed.


Fig. S1. Validation of Notch1 gain-of-function screen and MAFB expression in murine and human T-ALL.

Fig. S2. Analysis of spleen and peripheral blood after BMT.

Fig. S3. Validation of MafB siRNA and shRNA reagents.

Fig. S4. Ectopic MAFB expression enhances T6E cell growth.

Fig. S5. MAFB interacts with DNA through ETS factors, and the loss of ETS1 delays onset of T-ALL.

Fig. S6. Zmiz1 expression in T6E cells is not affected by the loss of MafB.

Table S1. Notch signaling enhancement of Notch-mutant LPΔP by candidate genes from the cDNA library.

Table S2. MafB suppression affects a subset of Notch signaling gene targets in T6E cells.

Table S3. List of oligo sequences and primer sets for qPCR and OIP.


Acknowledgments: We thank K. Toscano and S. Yu for technical assistance; P. Gimmoty for statistical consultation; J. Aster, M. Chiang, A. Ferrando, and D. Gerhardt for insightful comments; J. Aster for reagents; J. Soulier and C. Mullighan for the advice on patient T-ALL data sets; and the following cores at the University of Pennsylvania that contributed to this study: Mouse husbandry (University Laboratory Animal Resources), the Abramson Cancer Center Flow Cytometry Core (P30-CA016520), and the Abramson Family Cancer Research Institute Cores. This work benefited from data assembled by the Immunological Genome (Immgen) Project Consortium. Funding: This work was supported by a Leukemia and Lymphoma Society Fellow Award and T32HL007843 (K.V.P.) and the NIH [grants P01CA119070 (to W.S.P. and S.C.B.), R01HL134971 (to K.V.P.), and R01AI047833 (to W.S.P.)]. This work was also supported by a grant from the KiKa foundation Stichting Kinderen Kankervrij [KIKA-2010-082 (to Y.L.)]. Author contributions: K.V.P. designed and performed most of the experiments, interpreted the results, wrote the manuscript, and assembled the figures. L.X. performed in vivo mouse experiments and BMTs. L.S. and K.P. performed the experiments and generated the data for Fig. 6. J.P. performed the H3K27Ac experiments for Fig. 3. Y.O. and W.B. performed the array shown in table S2. C.L. performed the experiments shown in fig. S4 (B and C). G.B.W. analyzed the histology and tumor infiltration in leukemic transplants. R.M. and N.M. provided the ETS1 knockout BM and helped design the experiments shown in fig. S5. Y.L. and J.P.P.M. analyzed the human T-ALL data sets shown in fig. S2. S.C.B. helped design the experiments and edited the manuscript. R.B.F. helped design the ChIP-seq and analyzed the results for the experiment in Fig. 3. S.C. helped design the Notch gain-of-function screen, provided the cDNA library, and assembled and analyzed the data shown in Fig. 1. W.S.P. designed the experiments in this study, interpreted all the data, and helped write the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The H3K27Ac ChIP-seq and Affymetrix data have been deposited to Gene Expression Omnibus (accession no. GSE104993).

Stay Connected to Science Signaling

Navigate This Article