Abstract
The transcription factors EBF1 and PAX5 are frequently mutated in B cell acute lymphoblastic leukemia (B-ALL). We demonstrate that Pax5+/- x Ebf1+/- compound heterozygous mice develop highly penetrant leukemia. Similar results were seen in Pax5+/- x Ikzf1+/- and Ebf1+/- x Ikzf1+/- mice for B-ALL, or in Tcf7+/- x Ikzf1+/- mice for T cell leukemia. To identify genetic defects that cooperate with Pax5 and Ebf1 compound heterozygosity to initiate leukemia, we performed a Sleeping Beauty (SB) transposon screen that identified cooperating partners including gain-of-function mutations in Stat5 (∼65%) and Jak1(∼68%), or loss-of-function mutations in Cblb (61%) and Myb (32%). These findings underscore the role of JAK/STAT5 signaling in B cell transformation and demonstrate unexpected roles for loss-of-function mutations in Cblb and Myb in leukemic transformation. RNA-Seq studies demonstrated upregulation of a PDK1>SGK3>MYC pathway; treatment of Pax5+/- x Ebf1+/- leukemia cells with PDK1 inhibitors blocked proliferation in vitro. Finally, we identified conserved transcriptional variation in a subset of genes between human leukemias and our mouse B-ALL models. Thus, compound haploinsufficiency for B cell transcription factors likely plays a critical role in transformation of human B cells and suggest that PDK1 inhibitors may be effective for treating patients with such defects.
Introduction
Heterozygous deletions or loss of function mutations in a number of B cell transcription factors are a common feature of human B cell acute lymphoblastic leukemia (ALL)[1]. This is clearly evident for three transcription factors - EBF1, PAX5 and IKZF1[1, 2]. Interestingly, alterations involving these transcription factors commonly occur together[1, 3]. This is particularly pronounced in BCR-ABL+ leukemia in which 50% of leukemias with IKZF1 deletions also have mutations affecting Pax5 expression or function[4]. Therefore, an important question is whether compound haploinsufficiency for these transcription factors drives transformation and which combinations of transcription factors can promote transformation. Finally, the genetic alterations that cooperate with haploinsufficiency for these transcription factors to drive transformation have also not been comprehensively elucidated.
To address the questions above, we generated a set of mice that exhibited compound haploinsufficiency for various combinations of Ebf1, Pax5, Ikzf1, Cebpa, and Tcf7. Herein, we demonstrate that Pax5+/- x Ebf1+/-, Pax5+/- x Ikzf1+/-, and Ebf1+/- x Ikzf1+/- mice all generated B cell leukemia, while Tcf7+/- x Ikzf1+/- mice generated T cell leukemia. Furthermore, we used a SB Transposon screen to identify mutations that cooperate with Pax5+/- x Ebf1+/- compound haploinsufficiency to promote transformation. Our findings document the key role that compound haploinsufficiency for critical transcription factors plays in leukemia transformation and identify mutations that cooperate with such alterations to initiate transformation.
MATERIALS AND METHODS
Mice and Cells
All mice have been previously described [5-10]; the University of Minnesota IACUC approved all animal experiments. Mice were monitored for up to 400 days for leukemia. Spleen, lymph nodes, and bone marrow were isolated from tumor-bearing mice and used for further experiments.
RNA-Seq Analysis
RNA-seq was performed on total RNA extracted from either column purified progenitor B control cells (C57Bl/6, Pax5+/-x Ebf1+/-, Pax5+/- or Ebf1+/-) or leukemic cells from lymph nodes of tumor-bearing mice using a RNeasy Plus kit (Qiagen). Fifty-three barcoded TruSeq RNA v2 libraries were created and sequenced on a HiSeq 2500. A second set of data were used for variant calling analysis. Eight barcoded libraries were sequenced on the HiSeq 2000 to produce 100 bp paired end reads.
Systematic identification of gene clusters
Phase_2 BCCA and St Jude (2016) and phase_3 St Jude ALL(2018) Human mRNA RNA-Seq data were downloaded from https://ocg.cancer.gov/programs/target/data-matrix on April 23 2019 (dbGaP accession phs000218.v22.p8). The Value of 0.1 was added to each value the data was mean centered and log transformed. A SD cutoff was used to identify ∼8500 genes in each of three datasets. Unsupervised hierarchical clustering was used to define sets of genes which were defined by average linkage correlation > 0.2 and more than 150 members. Statistical enriched Gene cluster memberships across clusters were defined by Fisher exact test to identify “common” clusters across datasets. For the mouse data, the tumors were treated in a similar fashion except an SD cutoff of 1.0 was used. Statistical enriched gene cluster memberships across clusters were defined by Fisher exact test to identify “common” clusters across datasets using gene name matching.
Data Analysis
Data was analyzed using Prism 8 (Graphpad). A Shapiro-Wilk test was used to determine normality of all data. Unpaired data that passed normality was analyzed by an ordinary one-way ANOVA with Holm-Sidak’s multiple comparison test or by an unpaired t-test; data that failed normality were analyzed using an unpaired Kruskal-Wallace test with Dunn’s multiple comparison test. Kaplan-Meier Survival curves were analyzed by Log-rank (Mantel-Cox) Test. Integrated Genomics Viewer was used to view aligned sequences (Broad Institute).
Accession Numbers
RNA-Seq data was deposited with GEO and is accessible through GEO Series accession number GSE148680 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE148680).
Supplemental methods
Supplemental Methods section includes detailed protocols of cell lines and culture conditions, NGS, flow cytometry, qPCR, western blotting, Gene Set Enrichment Analysis, SB Mutagenesis, Transposon Insertion Analysis, Reverse Phase Proteomics and Inhibitor Assay.
RESULTS
Reduced expression of transcription factors critical for lymphocyte development leads to leukemia
To explore whether compound haploinsufficiency for Ebf1 and Pax5 leads to B cell transformation, we generated Pax5+/- x Ebf1+/- mice. As shown in Figure 1A, ∼67% of Pax5+/- x Ebf1+/- mice develop leukemia with a mean survival of 296 days. Flow cytometry analysis from bone marrow, lymph nodes and spleens of these mice demonstrated that leukemias resemble progenitor-B cells with a B220+CD19+IgM- phenotype (Fig. 1B) and also express pre-BCR, CD43, IL7R, TSLPR, c-KIT, AA4.1 and CD25 confirming their progenitor-B cell like phenotype (Sup Fig. 1). Although both male and female mice developed leukemia in this model, female mice exhibited greater penetrance (97% versus 71% at 400 days) and reduced median survival (265 days vs 298 days, p= 0.005; Sup Fig. 2). The effect of reduced expression of Pax5 and Ebf1 on transformation was not limited to this combination of transcription factors as similar results were observed in Pax5+/- x Ikzf1+/- and Ebf1+/- x Ikzf1+/- mice (Fig. 1A). Compound haploinsufficiency for all three transcription factors (Pax5+/- x Ebf1+/- x Ikzf1+/- mice) resulted in 100% penetrance of leukemia and much shorter mean survival (202 days). We previously reported that Pax5+/- or Ebf1+/- mice do not develop leukemia[11]. In contrast, Ikzf1+/- mice do develop leukemia with low penetrance (Fig. 1A)[12, 13]; however, these were always T cell leukemias (Fig. 1C). Deleting one copy of Pax5 and Ebf1 not only increased the frequency of B cell leukemias in Ikzf1+/- mice (none to ∼40%), but surprisingly, also resulted in a dramatic increase in T cell leukemias (∼5% in Ikzf1+/- mice versus ∼35% in Pax5+/- x Ebf1+/- x Ikzf1+/- mice)(Fig. 1C,D). Thus, although PAX5 and EBF1 are only expressed in B cells, reduced expression of these two transcription factors paradoxically also promoted T cell leukemia.
We next examined whether compound haploinsufficiency for lineage determining transcription factors was a general mechanism that could promote transformation of multiple cell lineages. To this end, we generated Tcf7+/- x Ikzf1+/- mice, as Tcf7, which encodes TCF1, and Ikzf1, are both required for T cell development[8, 9]. In addition, we generated Cebpa+/- x Ikzf1+/- mice, as Cebpa and Ikzf1 are both involved in myeloid cell development[14]. Cebpa+/- x Ikzf1+/- mice did not develop myeloid leukemia and the rate of T cell leukemia in Cebpa+/- x Ikzf1+/- mice was no higher than that observed for Ikzf1+/- mice. Thus, not all combinations of transcription factor haploinsufficiency promote transformation. However, Tcf7+/- x Ikzf1+/- mice developed T cell leukemias with high penetrance, comparable to that seen for B cell leukemias in Pax5+/- x Ebf1+/- mice (Fig. 1A,C). Thus, compound haploinsufficiency for lineage defining transcription factors can promote transformation in multiple cell lineages and may underlie many types of leukemias.
Genetic mutations that cooperate with Pax5 and Ebf1 heterozygosity to induce leukemia
Previous reports by Prasad and colleagues suggested that Ebf1+/- mice have defects in DNA repair with decreased expression of Rad51 and increased γH2AX foci[15]. These studies further claimed that defects in DNA repair resulted in increased mutation rates in Pax5+/- x Ebf1+/- leukemias and that this accounts, in part, for progenitor B cell transformation in those mice[15]. This suggestion is difficult to reconcile with the relatively low frequency of somatic mutations reported in human in B-ALL[16]. We re-examined this issue using Ebf1+/- mice in our colony. In contrast to the previous study, we found no difference in Rad51, Rad51AP or γH2AX expression when examining log2 transformed FPKM values generated from two separate RNA-seq experiments (Sup. Fig. 3A). In fact, the low level of variation paralleled that observed for a panel of housekeeping genes (B2m, Hprt, and Actb; Sup. Fig. 3A). Since the previous studies compared progenitor-B cells from WT and Ebf1+/- mice that had been cultured extensively in vitro we examined γH2AX expression in long-term cultured progenitor-B cells from WT and Ebf1+/- mice; no significant difference was observed (Sup. Fig. 3B). Further, we examined γH2AX expression by flow cytometry in progenitor-B cells directly from the bone marrow of WT and Ebf1+/- mice. Again, we found no significant difference in expression (Sup. Fig. 3C). Next, we examined whether genes involved in DNA repair were enriched in Ebf1+/- cells by Gene Set Enrichment Analysis (GSEA) using our RNA-seq data. We saw a significant enrichment for DNA repair genes (Sup. Fig. 3D), although it is unclear whether this reflects a direct effect of EBF1 on genes involved in DNA repair or just a relative increase in cells stuck at a stage undergoing VDJ recombination, as there is significant overlap between genes involved in DNA damage response and VDJ recombination. Finally, we examined whether subsets of human B cell leukemias exhibited increased mutation rates and if so, whether they were enriched in those containing Ebf1 mutations. As shown in figure 2A, ∼16% of B-ALLs obtained from the NIH TARGET ALL database showed significant levels of missense mutations. We broke the total B-ALL samples down into smaller subsets, characterized by ETV6-RUNX1 or TCF3-PBX1 translocations, or those with missense mutations in CDKN2A, PAX5, EBF1 or IKZF1. Leukemias expressing the ETV6-RUNX1 or TCF3-PBX1 translocations, or PAX5 missense mutations, were not enriched in hypermutated leukemias (Fig. 2B,C). In contrast, leukemias with missense mutations in CDKN2A, EBF1 or IKZF1 showed an increased percentage of leukemias with high numbers of missense mutations (Fig.2B). The number of CDKN2A, EBF1, or IKZF1 samples was too small to assess whether the increased percentage of hypermutated leukemias was statistically significant. However, mutations in EBF1, IKZF1 and CDKN2A are all enriched in BCR-ABL-like leukemias and when we pooled samples with these three mutations together there was a clear enrichment in samples with high numbers of missense mutations (Fig. 2C). In conclusion, leukemias with EBF1 mutations may be preferentially found in hypermutated B-ALL, but this is not a feature restricted to EBF1 as mutations in CDKN2A and IKZF1 are also associated with this hypermutated phenotype.
To discover novel genes that cooperate with Pax5 and Ebf1 heterozygosity to induce B-ALL we employed a Sleeping Beauty (SB) transposon mutagenesis screen[17]. Pax5+/- x Ebf1+/- x Cd79a-cre mice were crossed to mice expressing the mutagenic transposon SB in a Cd79a-Cre-dependent, and hence B cell-specific, manner. We generated 34 mice that were heterozygous for both Ebf1 and Pax5 and expressed the mutagenic transposon. Mice were housed for up to 400 days to allow them to develop leukemia. We also included single heterozygous combinations (Pax5+/-x Cd79a-Cre x SB and Ebf1+/- x Cd79a-Cre x SB) but neither of these cohorts developed leukemia within 400 days (data not shown). As seen in figure 3A, all of Pax5+/-x Ebf1+/- x Cd79a-Cre x SB mice developed leukemia. Thus, the presence of the sleeping beauty transposon increased penetrance of leukemia from 67% to 100% and decreased the median age of death from 296 to 205 days. Thus, other genes mobilized or silenced by SB transposition clearly cooperate with Pax5 and Ebf1 heterozygosity to initiate transformation.
To identify which genes were targeted by the transposon, we performed RNA-seq analysis on 31 SB induced leukemias. The SB transposon contains a unique 5’ leader sequence with a splice donor site that allows for splicing into transcripts. In addition, the SB transposon also has a splice acceptor and SV40 polyA tail that allows for splicing of upstream exons to the SV40 poly A sequence, thereby allowing for premature termination. These unique 5’ SB sequences and 3’ SV40 polyA signal sequences can be identified by RNA-Seq as novel fusion proteins. This allowed us to map SB fusions and determine how transposon insertions altered specific gene expression[18]. We carried out RNA-Seq analysis on progenitor B cells (CD19+B220+Igκ/λ-) from WT, Ebf1+/-, Pax5+/-, and Pax5+/- x Ebf1+/- pre-leukemic mice (∼6-12 weeks of age), as well as spontaneous Pax5+/- x Ebf1+/- leukemias (Fig. 3B). Differential gene expression analysis showed that WT and pre-leukemic samples all clustered distinctly from the SB-induced and spontaneous leukemias. The spontaneous Pax5+/- x Ebf1+/- leukemias were interspersed among the SB-induced leukemias demonstrating that the SB-induced leukemias shared gene expression signatures with the spontaneous leukemias. Finally, the leukemias were clearly heterogenous with a number of distinct subsets harboring distinct gene signatures (Fig. 3C).
RNA fusion analysis defines genes that cooperate with Pax5 x Ebf1 heterozygosity to induce leukemia
To identify the targets of transposon mutagenesis, we performed an RNA-Seq-based analysis of transposon fusions to identify genes targeted in our screen. The fusion transcripts are detected either directly as unique gene fusions or can be imputed from paired end reads that have one end derived from SB and the second end derived from the target gene sequence (called bridging fusions)[18]. There were 758 unique gene fusions or bridging fusions that were used to identify recurrent fusion events in 27 of 31 leukemias. Figure 3D lists all the reoccurring RNA fusions identified in our screen. Consistent with the heterogeneity of the gene expression profiles in distinct B-ALL subsets (Fig 3C), many of the targeted genes were only found in a fraction of the leukemias (Fig. 3E). The most notable exception was that almost all leukemias had SB insertions involving either Jak1 or Stat5b (Fig. 3E). SB RNA fusion analysis demonstrated that the SB 5’ leader UTR sequence typically fused to the first 1-4 coding exons of Stat5b or Jak1 genes (Fig. 4A, Sup. 4A). This suggested that a full-length or nearly full-length coding transcript would be generated for both Jak1 and Stat5b. Stat5b mRNA was expressed at significantly higher levels in leukemic samples harboring a SB transposon insertion (Fig. 4B,C). Consistent with data for Stat5b mRNA, there was a clear increase in the expression of STAT5 protein in samples with an SB insertion in the Stat5b gene (Fig. 4D-E). In contrast, the spontaneous Pax5+/- x Ebf1+/- leukemias did not exhibit significant increases in Stat5b expression (Fig. 4B,C). However, when we examined levels of phosphorylated STAT5 we found that Pax5+/- x Ebf1+/- leukemic cells expressed significantly higher levels of activated STAT5 than WT control mice, either directly ex vivo (Fig. 4F), or following in vitro stimulation with IL7 (Fig. 4G). This change represents a significant increase in phospho-STAT5 (p-STAT5) as it was equal to or higher than seen in mice expressing a constitutively active form of STAT5b in progenitor B cells (Fig. 4F,G)[19]. This result illustrated that there were increased levels of pStat5 expression in our leukemic cells. We next looked at known targets of STAT5b - Cish and Socs2 - to determine if there is increased Stat5b activity in these leukemias. We saw increased expression of both Cish and Socs2 in leukemic cells from both Pax5+/- x Ebf1+/- and Pax5+/- x Ebf1+/- x Cd79a-Cre x SB leukemias, which suggests that pStat5 is active in the leukemic cells (Fig. 4H, I). Similar expression results were seen for Jak1. We detected a significant increase in Jak1 mRNA in mice harboring insertions in the Jak1 locus (Sup. Fig. 4B,C). This increase in Jak1 transcription significantly increased expression of JAK1 protein in leukemic samples with an SB transposon insertion in the Jak1 gene locus (Sup. Fig. 4D,E). Our findings are consistent with the high rate of STAT5 activation observed in both human and murine B-ALL[11, 20] and underscore the critical role of JAK/STAT5 signaling in B cell leukemia – particularly those with reduced expression of Pax5 and Ebf1.
Loss of Cblb cooperates with reduced expression of Pax5 and Ebf1 to more rapidly induce leukemia
The other top hit in our mutagenesis screen was Cblb, which was targeted in almost 2/3 of our leukemias. Transposon insertional analysis from RNA-seq suggested that Cblb expression would be reduced as the majority of the SB gene fusions detected involved splicing in exons 6-9 (Fig. 5A). Consistent with this idea, both spontaneous Pax5+/- x Ebf1+/- leukemias, and SB-induced Pax5+/- x Ebf1+/- leukemias with an SB insertion in Cblb, showed significantly reduced Cblb mRNA expression (4.6-fold) compared to WT controls (Fig. 5B). Similar results were seen for CBLB protein expression as SB-induced leukemias with an SB insertion in the Cblb gene exhibited significantly lower expression of CBLB protein (1.8-fold) than SB-induced leukemias without an insert (Fig. 5C,D). To determine the role of reduced Cblb expression in leukemic transformation, we crossed our Pax5+/- x Ebf1+/- mice to Cblb-/- mice. Pax5+/- x Ebf1+/- x Cblb-/- mice developed B-ALL and died significantly faster than Pax5+/- x Ebf1+/- mice, demonstrating that Cblb acts as a tumor suppressor in progenitor B cells (Fig. 5E).
Reduced levels of Myb cooperate with Pax5 and Ebf1 heterozygosity to more rapidly induce leukemia
Myb was another frequent target of our mutagenesis screen. SB transposon insertions were scattered throughout the Myb gene locus, suggesting that this would result in decreased expression of Myb (Fig. 6A). Spontaneous Pax5+/- x Ebf1+/- leukemias showed a clear trend towards reduced Myb expression. Consistent with this observation, in SB-induced Pax5+/- x Ebf1+/- leukemias we saw a decrease in Myb expression in leukemias that lacked an SB insertion in Myb (1.5-fold, Fig. 6B) and an additional significant decrease in leukemias with an SB insertion in the Myb (2.3-fold, Fig. 6C). Thus, downregulation of Myb appears to be a general feature of Pax5+/-x Ebf1+/- leukemias. In SB-induced Pax5+/- x Ebf1+/- leukemias with insertions in the Myb gene, there was also a significant reduction (2.8 fold, Fig. 6D, E) at the protein level. Importantly, we found that Myb expression as assessed by RNA-Seq correlated with age of death - leukemias with less Myb were more aggressive resulting in earlier lethality (Fig. 6F).
PDK1-signaling pathway is deregulated in Pax5+/-x Ebf1+/- leukemias
In addition, to gene alterations directly regulated by SB transposition, we also noted a number of genes whose expression was significantly altered in Pax5+/- x Ebf1+/- leukemias relative to non-transformed controls. These included genes such as the tumor suppressor Bach2, which was significantly reduced (Sup. Fig. 5A). Intriguingly, we also noted dramatic (5.7 fold) downregulation of Asparagine synthetase in these leukemias, which may explain their susceptibility to L-Asparaginse[21](Sup. Fig. 5B). Other genes were strikingly upregulated including Pdk1 (3.0-fold) and its downstream targets Sgk3, and Rhebl1 (Fig. 7A,B,C). Conversely Tsc2, which inhibits this pathway, was downregulated (Sup. Fig. 5C,D); this pathway has been previously shown to enhance mTORC1 function and ultimately MYC expression[22, 23]. To determine whether PDK1 plays a critical role in maintaining viability of Pax5+/- x Ebf1+/- leukemias, we treated two cell lines generated from Pax5+/- x Ebf1+/- leukemias in vitro with either vehicle control or the PDK1 inhibitor (GSK2334750). Treatment of these cell lines with the PDK1 inhibitor resulted in a dose-dependent decrease in the survival of these cell lines suggesting that this might be a useful treatment for B-ALL subsets with reduced Pax5 and Ebf1 expression (Fig. 7D).
To examine a possible role for PDK1 expression in human ALL, we examined ALL patient samples using a reverse phase proteomics approach[24]. PDK1 was expressed in five subsets of B-ALL but expression varied widely (Sup. Fig. 5E). We examined PDK1 expression in the two largest cohorts B-NOS and BCR-ABL+ leukemias. PDK1 levels did not correlate with overall survival in B-NOS patients (data not shown). In contrast, BCR-ABL+ patients with the highest levels of PDK1 expression did significantly better than those with lower PDK1 expression (Fig. 7E). The difference in overall survival appears to be driven most strongly by young adults, as they showed the most dramatic difference in overall survival (Fig. 7F). Finally, we examined PDK1 expression in patients with BCR-ABL and B-NOS leukemia based on relapse status. In both subsets of leukemia, lower levels of PDK1 correlated with relapse (Fig. 7G,H). Thus, PDK1 appears to play an important role in B-ALL survival or proliferation, but patients with the highest level of PDK1 expression respond better to therapy.
Pax5 x Ebf1 leukemia show common transcriptional variation patterns across mouse and human
To determine if the murine leukemias that developed in our sleeping beauty screen are similar to subsets of human B-ALL, we quantified inter-leukemia transcriptional variation using our newly developed gene cluster expression summary score (GCESS)[25]. Using this approach, we first examined inter-leukemia transcriptional variation in distinct human leukemia datasets. There was notable heterogeneity between human B-ALL subsets. However, we could identify clusters with similar variations in gene expression (Fig. 8A). We used the GCESS approach to establish transcriptional similarity between these human leukemias and our murine sleeping beauty transposon induced leukemias. As shown in figure 8B, there were two distinct transcriptional variants in our SB dataset. One of these SB leukemia subsets showed a similar gene expression signature to one of the conserved human leukemia subsets (fishers exact test, p=4.3e-07). Thus, the leukemias that developed in our SB system are similar to human leukemias, thereby validating our approach as a useful model of human leukemia.
To better characterize these leukemias, we utilized ENRICHR to examine gene lists from the GCESS of each of the datasets from the conserved murine and human leukemias (Sup. Table 1)[26, 27]. Consistent with other findings in this study, we found upregulation of cytokines and cytokine receptor genes, as well as genes involved in JAK/STAT5 signaling. In addition, NFκB signaling was significantly altered, which is consistent with work from multiple groups on NFκB in B cell development and leukemia [20, 28, 29]. A surprising observation was a strong myeloid gene signature in the human and murine leukemias. There are two potential reasons for this. First, these leukemias could be infiltrated with myeloid cells. Alternatively, the leukemias could have lost lineage fidelity and begun to express myeloid genes. Since PAX5, EBF1 and IKZF1 all play key roles in enforcing B cell lineage fidelity, and our murine B cell leukemias were relatively pure populations of leukemic blasts, we favor this later possibility. Thus, subsets of Pax5+/- x Ebf1+/- leukemia exhibit some myeloid characteristics.
Discussion
Genes encoding the transcription factors PAX5, EBF1 and IKZF1 frequently exhibit loss of one allele or express loss-of-function mutations in human B cell leukemia[2, 4, 30]. However, their role in B cell transformation is not entirely clear. It is likely that loss of function mutations in these transcription factors affect B cell differentiation. Previous studies using inducible Pax5 mutants in murine models of B cell leukemia suggest that this plays a role[31]. However, Pax5+/- and Ebf1+/- mice do not develop leukemia, while Ikzf1+/- mice have only been shown to develop T cell leukemia. This raises the question of how mutations in these genes promote transformation. We previously demonstrated that combining loss of one allele of either Pax5 or Ebf1 with a constitutively active Stat5 transgene (referred to as Stat5b-CA) led to rapid onset of B cell leukemia. A key feature of these leukemias is that Ebf1 expression was reduced ∼50% in Stat5b-CA x Pax5+/- leukemias, while Pax5 expression was comparably reduced in Stat5b-CA x Ebf1+/- leukemias[11]. These findings suggested that perhaps compound haploinsufficiency for these transcription factors might be key for promoting transformation. Herein we demonstrated that this is the case as Pax5+/- x Ebf1+/- mice develop B cell leukemia with high penetrance. Importantly, this was not a phenomenon restricted to this pair of transcription factors as we saw qualitatively similar onset of B cell leukemia in Pax5+/- x Ikzf1+/- and Ebf1+/- x Ikzf1+/- mice. Nor was this observation restricted to B cell leukemia as we observed that Tcf7+/- x Ikzf1+/- mice gave rise to highly penetrant T cell leukemia. Thus, compound haploinsufficiency for transcription factors that play key roles in either B cell or T cell development can promote transformation.
The mechanism by which compound haploinsufficiency promotes transformation remains unclear. Previous studies suggested that this may be due to defective DNA repair upon reduced Ebf1 expression. We were unable to validate defects in Rad51, Rad51ap expression or increased γH2AX expression in pre-B cells from Ebf1+/- mice. However, we did observe that a subset of human leukemias expressed a hypermutated phenotype and that leukemias with CDKN2A, EBF1 and IKZF1 missense mutations were enriched in this subset. It is possible that the newly described BCR-ABL-like subset of B-ALL (which is enriched in leukemias with mutations in CDKN2A, IKZF1 and EBF1) might be characterized by the hypermutated phenotype. This should be examined further and if confirmed suggests that these leukemias may be more susceptible to immunotherapy-based treatments.
An alternative mechanism by which compound haploinsufficiency promotes transformation could involve loss of lineage fidelity. Consistent with this hypothesis, EBF1 and PAX5 have both been shown to play important roles in restricting cells to the B cell lineage. Consistent with this observation, B cell progenitors in Pax5+/- x Ebf1+/- mice retain T cell lineage potential[32]. Moreover, we found that compound haploinsufficiency for Pax5 and Ebf1 promotes increased penetrance of B cell and T cell leukemia on an Ikzf1+/- background. Thus, decreased expression of transcription factors that are only expressed in B cells (PAX5 and EBF1) paradoxically enhance the development of T cell leukemia in Ikzf1+/- mice. Although the mechanism remains unclear, it is possible that in some cases T-ALL may emerge from a B cell progenitor. This may have implications for how such leukemias develop resistance to therapy if the key progenitor is more closely linked to B cell rather than T cell development. Finally, our finding that a subset of murine Pax5+/- x Ebf1+/- leukemias, as well as their similar human B-ALL counterparts, exhibit a strong myeloid gene signature also suggests that loss of lineage fidelity may be a key feature of this disease. Alternatively, the myeloid signature could arise due to preferential infiltration of this type of leukemia with myeloid cells. This is certainly a possibility, especially for leukemias in the human datasets. However, our murine leukemias are strongly enriched for B lineage cells. Thus, we favor an explanation in which the myeloid gene signature arises due to aberrant expression of myeloid genes in B cell leukemic blasts.
To gain a better understanding of the molecular alterations that cooperate with Pax5 and Ebf1 haploinsufficiency to promote transformation we carried out an unbiased SB transposon screen in Pax5+/- x Ebf1+/- mice. These studies identified two major pathways that cooperate with Pax5 and Ebf1 haploinsufficiency to drive transformation. First, we found gain-of-function mutations for Jak1 or Stat5b in almost all of our leukemias in this screen. This finding underscores in an unbiased way the critical role of JAK/STAT5 signaling in B cell transformation. In a previous SB mutagenesis study targeting the STAT5 pathway, we were able to induce more rapid leukemia onset than SB mice with only Pax5+/- x Ebf1+/- predisposing mutations (average onset of leukemia ∼72 versus 302 days, respectively) [33]. This suggests that changes needed to activate STAT5 may take longer to arise than secondary loss-of-function mutations to Pax5, Ebf1, or other transcription factors.
The second major pathway identified involves CBL-B, and to a much lesser extent the related family member CBL. The mechanism by which Cblb loss-of-function affects transformation is unclear. However, the fact that these are loss-of-function mutations is supported by the observation that crossing Cblb-deficiency onto the Pax5+/- x Ebf1+/- background accelerated the onset of leukemia and increased overall penetrance. Our SB screen suggest that CBLB mutations should be examined in greater detail in human B-ALL.
A number of other target genes were identified in our SB screen, although none were as prevalent as the mutations in Jak1, Stat5b or Cblb. These include SB insertions in several cytokine/receptor genes that have previously been shown to be involved in transformation including Il2rb, Gh, Csf2, as well as the histone acetyltransferase Ep300. Finally, we noted relatively frequent mutations in the transcription factor Myb. The finding that Myb was targeted by SB in our leukemias was initially not surprising as Myb has previously been identified as an oncogene. However, what was surprising is that the mutations in Myb were loss-of-function mutations resulting in reduced Myb expression. This leads to the somewhat surprising observation that Myb acts operationally as a tumor suppressor in this context and parallels our previous observation that NFκB also acts as a functional tumor suppressor in progenitor B cells[20, 34]. Since MYB plays a role in B cell differentiation[35] this likely reflects a role for MYB in blocking differentiation at the highly proliferative pre-B cell stage.
In addition to targets directly identified by SB integration, our RNA-Seq analysis of Pax5+/- x Ebf1+/- and SB x Pax5+/- x Ebf1+/- leukemias also identified other deregulated signaling pathways. One of the most prominently deregulated pathways involved the serine/threonine kinase PDK1. PDK1 expression was modestly but significantly upregulated in Ebf1+/- and Pax5+/- x Ebf1+/- pre-leukemic progenitor B cells, and significantly further elevated in Pax5+/- x Ebf1+/- leukemias. The mechanism by which increased PDK1 expression promotes transformation is unclear. However, previous studies have shown that PDK1 interacts with SGK1/3 to inhibit TSC2 function and expression[22]. This results in increased function of RHEB or RHEBL1, which in turn promote mTOR function and ultimately MYC expression[22]. Consistent with this model we found that Tsc2 expression levels were reduced in Pax5+/- x Ebf1+/- leukemias while Sgk3, Rhebl1 and Myc levels were increased. An alternative pathway that could also be affected by PDK1 involves PDK1-dependent activation of PLK1, which in turn has been shown to phosphorylate and activates MYC[23]. Thus, there are a number of potential mechanisms by which increased PDK1 expression could promote transformation. What is clear is that PDK1 inhibitors effectively blocked proliferation of Pax5+/- x Ebf1+/- primary leukemia cell lines in vitro. Since there are currently a number of PDK1 inhibitors available with some demonstrating efficacy in preclinical trials[36, 37], our findings suggest that PDK1 inhibition might be an effective strategy for treating B cell leukemias that exhibit reduced expression of Pax5 and Ebf1.
Authors’ Contribution
Conception and design
L. Heltemes-Harris and M. Farrar
Development of Methodology
L. Heltemes-Harris, A. Sarver, S. Kornblau and M. Farrar
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.)
L. Heltemes-Harris, G. Hubbard, A. Sarver, T. Knutson, S. Kornblau and M. Farrar
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis)
L. Heltemes-Harris, G. Hubbard, R. LaRue, T. Starr, S. Munro, T. Knutson, C. Henzler, A. R. Yang, A. Sarver, S. Kornblau and M. Farrar.
Writing, review, and/or revision of the manuscript
L. Heltemes-Harris, G. Hubbard, R. LaRue, T. Starr, S. Munro, C. Henzler, R. Yang, A. Sarver, S. Kornblau and M. Farrar.
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases)
L. Heltemes-Harris
Study Supervision
L. Heltemes-Harris, M. Farrar
Conflict of interest
The authors declare no conflicts of interest.
ACKNOWLEDGEMENTS
We thank A. Rost, for technical assistance with mouse breeding; the University of Minnesota’s Supercomputing Institute for providing computing and bioinformatic resources; Dr. Meinrad Busslinger (Pax5+/-), Dr. Rudolf Grosschedl (Ebf1-/-), Dr. Peter Johnson (Cebpa-/-), Dr. Andrew Wells (Ikzf1-/-) and Dr. David Largaespada (Rosa26LSL-SB11 T2/OncxRosa26LSL-SB11) for providing the indicated mouse strains. The results published here are in part based upon data generated by the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative, phs000218, managed by the NCI. This work was supported by a Cancer Research Institute Investigator award, a Leukemia and Lymphoma Society Scholar award, funding from the UMN Masonic Cancer center and grants from the NIH (RO1 CA232317) to MAF. TKS was supported by grants from the Randy Shaver Cancer Research and Community Fund, NIH NCI (R21 CA216652), and the Masonic Cancer Center. ALS was supported by NCI (CA211249) and Masonic Cancer Center Support Grant (CA077598). SMK was supported by CPRIT MIRA RP 160693 and NIH/NCI P50 CA100632-09.