Mice were backcrossed onto an FvB/NJ background for at least 8 generations. All experiments were performed in compliance with the University of Chicago Institutional Animal Care and Use Committee. E2a^-/- , Lef1^f/f , and Lck-Cre mice and genotyping protocols were described previously (38). FvBn/J mice and Lck-Cre mice were purchased from the Jackson Laboratory.

Thymocytes or splenocytes were dissected and dispersed using frosted glass slides followed by filtration through a 100 mM cell strainer. Cells were stained at a concentration of 2 x 10⁷ cells/ml in FACS buffer (PBS + 5% FCS +.02% azide) after incubation with FcBlock. Intracellular staining was performed using the FoxP3/Transcription Factor Staining Kit. Antibodies were purchased from BioLegend, eBiosciences or Fischer Scientific and specific antibody clones and fluorochromes are available upon request. Antibodies used: Lef1, CD4, CD8, CD117, CD25, CD44, TCRb, CD11b, CD11c, DX5, B220, Gr1. Data was acquired on an LSRII or Fortessa and analyzed using FlowJo (TreeStar).

Leukemia cell lines were generated by culturing thymic cells from moribund mice in OPTI-MEM media containing 10% FCS, 2-mercaptoethanol and Pen/Strep/Glu for greater than 2 weeks. All established lines were frozen in 5% DMSO/50%FCS in liquid nitrogen for long term storage.

OP9-DL1 stromal cells were maintained in OPTI-MEM and plated 1 day before used to achieve a near confluent monolayer of cells. Multipotent progenitors were isolated as Ter119-Gr1- cells from e13 fetal liver and cultured on OP9-DL1 in the presence of 5 ng/ml Flt3 ligand, IL-7 and CD117 ligand.

The retroviral vectors MigR1, MigR1-Cre, MigR1-DNMAML were described previously (39). Retroviral plasmid DNA was isolated using CsCl. Retroviral supernatants were produced by transfecting plasmid DNA into Phoenix cells using Ca₂PO₄ precipitation and cells were transduced with retrovirus as previously described (40).

RNA was extracted using Trizol Reagent. For microarray analysis, RNA was converted to cDNA that was used to probe Affymetrix MOE 430_2 arrays as previously described (41). Raw array data were normalized with RMAexpress (http://rmaexpress.bmbolstad.com/) and analyzed by dChip (http://www.biostat.harvard.edu/complab/dchip/). Probe set annotation was obtained from Affymetrix. For RNA-seq analysis, RNA was sequenced on a Next-Seq500 and analyzed as described (42). Raw sequence reads were trimmed using Trimmomatic v0.33 (TRAILING:30 MINLEN:20) and then aligned to mouse genome assembly mm10 with TopHat v 2.1.0. Reads were assigned to genes using the htseq-count tool from HTSeq v 0.6.1 and gene annotations from Ensembl release 78. The R package EdgeR was used to normalize the gene counts and to calculate differential expression statistics for each gene for each pairwise comparison of sample groups. Metascape analysis was performed on differential gene expression lists (https://metascape.org) (43). Genes were considered differentially expressed at Log2FC with an adj. p-valule of <0.01.

RNA was reverse transcribed with Superscript III (Invitrogen). Quantitative RT-PCR was performed in an iCycler (Bio-Rad Laboratories) with SYBR Green (Bio-Rad Laboratories). Expression values were normalized to Hprt and were calculated by the ΔC_T method. Primer sequences are available on request.

To characterize the cytogentic pattern of T cell leukemias from E2a^-/-Lef1 ^Δ/Δ mice, SKY analysis was performed using the ASI SkyPaintTM assay for mouse chromosomes as described previously (44) on cell lines or fresh leukemic cells from moribund mice with leukemia (10 metaphase cells were analyzed per case). Karyotype results are in Table S2 .

Total protein extracts were prepared and analyzed by Western blot analysis as described previously (30). Primary antibodies used were anti-Notch1 antibody (V1744) reactive with the cleaved cytoplasmic domain (Cell Signaling Technology) and anti-actin (Abcam).

For original data please contact bkee@bsd.uchicago.edu. RNA-sequencing data can be accessed in the Gene Expression Omnibus under GSE186420. Microarray data is under GSE196391.

Our previous studies revealed that Notch1 is mutated in E2a^-/- T cell leukemias and required for their survival (26). We identified Lef1 as a target of the Notch pathway in these cells and demonstrated that siRNA directed against Lef1 reduced the viability of these leukemic cells (30). Here, using flow cytometry, we found that LEF1 and TCF1 protein are detected in E2a^-/- leukemias and that LEF1, but not TCF1, was reduced after treatment of cells with a γ-secretase inhibitor, which antagonizes Notch signaling ( Figures 1A, B ). To rigorously demonstrate that LEF1 was required for growth of these leukemias, we created E2a^-/- mice that were homozygous for alleles of Lef1 with loxp sites flanking the DNA binding domain (32). We generated 2 lines from the leukemias arising in these mice and used a retrovirus producing Cre to delete the DNA binding domain of LEF1 ( Figure 1C ). The retrovirus produced GFP in addition to Cre and therefore we could track Cre expressing cells by their expression of GFP ( Figure 1C, D ). LEF1 was lost from a subset of cells after transduction with MigR1-Cre and the frequency of LEF1 negative cells mirrored the frequency of GFP⁺ cells ( Figure 1D ). QPCR analysis of sorted GFP⁺ cells from MigR1 or MigR1-Cre transduced cells revealed decreased Lef1 mRNA after transduction of E2a^-/-Lef1^f/f but not E2a^-/- leukemias with MigR1-Cre ( Figure 1E ). We tracked the fate of cells with deletion of Lef1 by following the frequency of GFP⁺ cells. GFP⁺ cells in MigR1-Cre expressing E2a^-/-Lef1^f/f cells declined steadily over time consistent with the loss of cells that lacked LEF1 ( Figure 1F ). An E2a^-/- leukemia line that was heterozygous for the Lef1^f allele (E2a^-/-Lef1^f/+) and transduced with MigR1-Cre also showed reduced representation of GFP⁺ cells with time in culture, although not to the degree of E2A^-/-Lef1^f/f leukemias ( Figure 1F ). In contrast, the same cell lines transduced with MigR1, which does not promote deletion of Lef1, demonstrated a stable frequency of GFP⁺ cells over time ( Figures 1D–F ). Similarly, E2a^-/- lines transduced with either MigR1 or MigR1-Cre showed stable GFP expression ( Figure 1F ). To further examine the impact of LEF1-deletion on these leukemic cells, we analyzed the transcriptome of GFP⁺ cells 72 hours after transduction with MigR1 or MigR1-Cre. We found decreased expression of multiple signaling-associate genes (Id3, Syk, Sgk, Rasgrp1) and increased expression of Cdkn1a, encoding the cell cycle inhibitor p21 ( Figure 1G and Table S1 ). The increased expression of p21 could contribute to the reduced expansion of LEF1-deleted leukemias. These data demonstrate that LEF1 is required for the maintenance of E2a^-/- T cell leukemia lines in vitro and are consistent with our previous studies using Lef1 siRNA (30).

To determine whether the increased expression of LEF1 occurs in pre-leukemic mice, we examined the Lineage^- population of the thymus for expression of LEF1. As expected, there were fewer DN3 thymocytes in E2a^-/- mice compared to control with an increased frequency of CD117^loCD25^int cells, previously shown to be innate lymphoid cells ( Figure 2A ) (45, 46). The E2a^-/- DN3 cells expressed substantially more LEF1 than control DN3 thymocytes ( Figure 2B ). By qRT-PCR we found increased expression of Lef1 mRNA in E2a^-/- DN3 thymocytes ( Figure 2C ). DN3 thymocytes fail to develop from E2a^-/- multipotent progenitors cultured in vitro on OP9-DL1 (45, 47), indicating that these cells are compromised with respect to T cell differentiation. To determine whether LEF1 might provide an advantage to these cells, we used a retrovirus to force multipotent progenitor (MPP) cells to ectopically express LEF1. E2a^-/- MPPs transduced with the MigR1 retrovirus, which produce GFP only, failed to generate DN3 cells in vitro, as expected ( Figures 2D, E ). In contrast, E2a^-/- MPPs transduced with LEF1 producing retrovirus generated DN3 cells and an increased frequency of DN2 cells. Notably, even control MPPs transduced with LEF1 producing retrovirus showed increased generation of DN2 and DN3 cells ( Figures 2D, E ). This propensity to promote differentiation of control and E2a^-/- MPPs was not dependent on the presence of the β-catenin interaction domain (CAT) as a retrovirus producing a mutant form of LEF1 lacking the CAT domain also supported differentiation ( Figures 2D, E ). These data lead us to hypothesize that the increased expression of LEF1 in E2a^-/- DN3 thymocytes aids in their differentiation from more immature progenitors and that LEF1’s essential functions are independent of its interaction with β-catenin.

To test the hypothesis that LEF1 is essential for the development of E2a^-/- DN3 cells and for leukemic transformation, we generated Lck-Cre E2a^-/-Lef1^f/f mice, which express Cre starting at the DN2 stage. Lck-Cre⁺ E2a^-/-Lef1^f/f mice (DKO) and E2a^-/- mice had similar numbers of thymocytes and Lineage^- thymocytes ( Figures 3A–C ). However, the Lineage^- population of DKO mice showed a reduced frequency of DN3 thymocytes ( Figures 3B, D ). The number of DN3 thymocytes was also significantly decreased in DKO compared to E2a^-/- mice ( Figure 3E ). In contrast Lck-Cre⁺ Lef1^f/f (Lef1^Δ/Δ ) mice had thymocyte and Lineage^- thymocyte numbers, as well as DN3 frequencies that were similar to Control mice ( Figure 3D ), indicating that Lef1 deletion did not have a major impact on these cells.

Since the number of thymocytes was similar in E2a^-/- and DKO mice we examined the phenotype of more mature thymocytes. As previously reported, E2a^-/- thymocytes had a reduced frequency of CD4⁺CD8⁺ (DP) thymocytes with an increased frequency of CD4⁺ and CD8⁺ cells. The CD4 by CD8 profile of DKO thymocytes revealed a slightly increased frequency of CD4⁺CD8⁺ (DP) cells and a decreased frequency of CD4⁺ of CD8⁺ single positive thymocytes compared to E2a^-/- mice ( Figure 4A ). E2a^-/- mice also have an increased frequency of TCRβ⁺ thymocytes, although the number of these cells is lower than in Ctrl mice ( Figures 4A, B ). The frequency and number of TCRβ⁺ cells was reduced in DKO mice compared to E2a^-/- mice, although it was still higher than in Ctrl or Lef1^Δ/Δ mice ( Figures 4A, B ). These data suggest that deletion of LEF1 had a subtle but significant impact on the differentiation of thymocytes in the absence of E2a. Interestingly, a substantial portion of DKO Lineage⁺ thymocytes, which are primarily DP cells, expressed CD25 ( Figures 4C, D ). The frequency of Lineage⁺CD25⁺ cells was also higher in E2a^-/- thymocytes than in Ctrl or Lef1^Δ/Δ thymocytes but was not significantly elevated in number ( Figures 4C, D ). Indeed, a direct comparison of DP thymocytes revealed a substantial increase CD25 on DKO compared to E2a^-/- thymocytes ( Figure 4E ). Given that CD25 is a known Notch1 target gene (48), we tested the hypothesis that Notch1 was expressed in DP thymocytes from DKO mice. By QPCR analysis we observed mRNA for Notch1 and the Notch1 target gene Nrarp in DP thymocytes from DKO but not E2a^-/- or Ctrl mice ( Figure 4F ). These data indicate that Notch1 is activated in DKO DP thymocytes.

To determine whether deletion of Lef1 impacted the transformation potential of E2a^-/- thymocytes, we allowed the mice to age and monitored them for signs of leukemia. Surprisingly, DKO mice became moribund with an average latency of 100 days (range 80-130 days) whereas the average latency for E2a^-/- mice was 130 days (range 100-170 days), and notably, all mice that we followed developed disease ( Figure 5A ). At sacrifice, leukemia was confirmed by counting thymocyte numbers and by flow cytometry of multiple tissues. Cell numbers were similar in the thymus of moribund E2a^-/- or DKO mice, although the range was greater for the DKO cells ( Figure 5B ). However, splenic lymphocyte numbers were elevated in the moribund DKO as compared to E2a^-/- mice ( Figure 5C ). Primary leukemic cells in DKO mice were frequently low for CD4 and CD8 whereas E2a^-/- leukemias were CD4^hiCD8^hi or contained SP cells ( Figures 5D, E ). The DKO leukemias also expressed CD25 without CD44, a phenotype that is distinct from that of the majority of E2a^-/- leukemias ( Figure 5E ). Taken together, these data demonstrate that, despite the reduced number of DN3 cells, DKO mice developed T cell leukemia with reduced latency compared to E2a^-/- mice and these leukemias had a distinct surface receptor phenotype.

E2a^-/- leukemias have mutations in the Notch1 gene and they are dependent on Notch signaling for their survival (26). To determine whether DKO leukemias also had mutations in the Notch1 gene, we PCR amplified the 3’ portion of Notch1 and performed sequencing on 3 DKO cell lines that we established in culture. Notably, we found insertions that resulted in out of frame translation of the PEST domain of Notch1 in all of the DKO lines ( Figure 6A ). These mutations are predicted to stabilize ICN1 and, indeed, ICN1 protein could be detected in these leukemias by western blot analysis ( Figure 6B ). To determine whether these leukemias were dependent on Notch1 signaling we used retroviral transduction to ectopically express a dominant negative version of the Notch1 co-activator MAML in these cells (39). The frequency of cells expressing DN-MAML, identified by their expression of GFP, declined over time in culture regardless of whether they were E2a^-/- or DKO leukemias ( Figure 6C ). In contrast, the frequency of GFP⁺ cells remained stable in cultures transduced with the control MigR1 retrovirus ( Figure 6C ). Cytogenetic analysis using spectral karyotyping of three DKO primary leukemias revealed trisomy for chromosome 15 containing the c-Myc locus, a feature that is also observed in E2a^-/- leukemias ( Figures 6C, D and Table S2 ) (24). Consistent with this observation, E2a^-/- and DKO cell lines had similar levels of c-Myc mRNA ( Figure 6E ). These data demonstrate that DKO leukemias, like E2a^-/- leukemias, required Notch signaling for their survival and had mutations impacting the stability of ICN1 expression of c-Myc.

To gain further insight into the differences between E2a^-/- and DKO leukemias we performed RNA-sequencing on multiple E2a^-/- and DKO leukemia lines. There was a large amount of variation in the gene programs of these leukemias but they were resolved into distinct populations by principle component analysis through PC1 and PC3 ( Table S3 and Figure 7A ). We identified 89 genes that were generally decreased in DKO as compared to the E2a^-/- lines and 70 genes that were increased (Log2FC, adj. p-value <0.01) ( Figure 7B ). Tcf7 mRNA appeared to be increased in DKO leukemias but TCF1 protein, unlike LEF1 protein, was expressed similarly in E2a^-/- and DKO lines when evaluated by flow cytometry ( Figure S1 ). Analysis of the differentially expressed genes by Metascape revealed that the genes that decreased in the DKO leukemias were enriched for genes involved in IL-4 production whereas those that increased were enriched for genes in the monocarboxylic acid transport (MCT) and the Hedgehog signaling pathways ( Figures 7C, D ). Genes in the MCT pathway included Fabp5, Pla2g12a, Syk and Hoxa13 ( Figures 7B, E ). Genes in the Hedgehog signaling pathway included Axin2, Cdkn1a, Prkch and Rasgrp1 ( Figures 7B, F ). Taken together, these data indicate that DKO and E2a^-/- leukemia lines share many common gene expression features but differ in a few key genes that could impact their metabolic requirements or response to external signals.