The TCF-1 and LEF-1 transcription factors have cooperative and opposing roles in T-cell development and malignancy

SUMMARY

The TCF-1 and LEF-1 transcription factors are known to play critical roles in normal thymocyte development. Unexpectedly, we found that TCF-1-deficient (Tcf7^−/−) mice developed aggressive T-cell malignancy, resembling human T-cell acute lymphoblastic leukemia (T-ALL). Surprisingly, LEF-1 was aberrantly upregulated in pre-malignant Tcf7^−/− early thymocytes and lymphoma cells. We further demonstrated that TCF-1 directly repressed LEF-1 expression in early thymocytes and that conditional inactivation of Lef1 greatly delayed or prevented T-cell malignancy in Tcf7^−/− mice. In human T-ALLs, an early thymic progenitor (ETP) subtype was associated with diminished TCF7 expression, and two of the ETP-ALL cases harbored TCF7 gene deletions. We also showed that TCF-1 and LEF-1 were dispensable for T-lineage commitment but instead were required for early thymocytes to mature beyond the CD4⁻CD8⁻ stage. TCF-1 thus has dual roles, i.e., acting cooperatively with LEF-1 to promote thymocyte maturation while restraining LEF-1 expression to prevent malignant transformation of developing thymocytes.

INTRODUCTION

Hematopoietic progenitors seed the thymus to initiate the T-cell differentiation program. The development of αβ T cells follows sequential maturation stages, starting from immature thymocytes that are double negative (DN) for CD4 and CD8 expression (Rothenberg and Taghon, 2005). The DN thymocytes are further divided into 4 distinct subsets based on CD25 and CD44 expression, and these are CD25⁻CD44⁺ DN1, CD25⁺CD44⁺ DN2, CD25⁺CD44⁻ DN3, and CD25⁻CD44⁻ DN4 cells. The c-Kit⁺ DN1 cells, defined as the most immature early thymic progenitors (ETPs) (Bhandoola and Sambandam, 2006), give rise to DN2 cells, a process known as T-cell specification (Rothenberg and Taghon, 2005). Whereas ETP and DN2 thymocytes retain multi-lineage potentials, DN3 cells are completely committed to T lineage, with the T cell receptor (TCR) β (Tcrb) gene locus rearranged. During β-selection, DN3 thymocytes with productive TCRβ rearrangements receive survival and proliferative signals and mature into the DN4 stage (Rothenberg et al., 2008). DN4 thymocytes then develop sequentially through CD8⁺ immature single positive (ISP) and CD4⁺CD8⁺ double positive (DP) stages. After vigorous selection processes, the DP thymocytes differentiate into either CD4⁺ or CD8⁺ single positive T cells.

The stage-wise T-cell development is guided by a large number of transcription factors with essential and recurrent roles (Rothenberg et al., 2008). Among these, the Wnt pathway effector transcription factors, T cell factor 1 (TCF-1) and lymphoid enhancer-binding factor (LEF-1), encoded by Tcf7 and Lef1 respectively, have been known to be critical for production of T cells (Staal and Sen, 2008; Xue and Zhao, 2012). TCF-1-deficient mice show a severe reduction in thymocyte numbers (Verbeek et al., 1995), with young Tcf7^−/− mice exhibiting incomplete blocks of T-cell development at the DN1, DN3, and ISP stages (Schilham et al., 1998). On the other hand, LEF-1-null mice have no apparent abnormalities in T-cell development (Okamura et al., 1998). In a fetal thymic organ culture study, thymocytes that are deficient for LEF-1 and hypomorphic for TCF-1 are completely arrested at the ISP stage (Okamura et al., 1998). These observations suggest that TCF-1 and LEF-1 have overlapping functions during thymocyte maturation and that TCF-1 appears to have a more dominant effect. Notch signaling is an indispensable positive regulator of T-lineage specification, commitment, and β-selection (Yashiro-Ohtani et al., 2010). Two recent studies demonstrated that TCF-1 is sharply induced upon activation of the Notch signaling in hematopoietic progenitors (Germar et al., 2011; Weber et al., 2011). Importantly, forced expression of TCF-1 is sufficient to specify a T-cell fate in the absence of active Notch signaling (Weber et al., 2011).

Developing T cells are susceptible to malignant transformation, because of their highly proliferative nature at several developmental stages and DNA double-strand breaks generated during TCR locus rearrangements. Juxtaposition of genes encoding oncogenic transcription factors next to the strong TCR enhancers and promoters is a known leading cause of human T-cell leukemia and lymphoma (Aifantis et al., 2008; Grabher et al., 2006). Deregulation of key pathways that are critical for T-cell development also constitutes either initiating or promoting events that ultimately lead to malignant transformation of thymocytes. For example, forced expression of dominant active forms of Notch1 was first demonstrated in mouse models to be causative of highly aggressive immature T-cell malignancy (Allman et al., 2001). Subsequently, somatic activating mutations in the NOTCH1 gene were found in more than 50% human T-cell acute lymphoblastic leukemia (T-ALL) cases (Weng et al., 2004).

Aberrant activation of the Wnt-β-catenin pathway has been found in various cancers including hematological malignancy (Reya and Clevers, 2005). β-catenin protein is post-translationally modulated by Wnt morphogen-initiated signaling. Glycogen synthase kinase-3β (GSK-3β), a component of a cytosolic multi-molecular “destruction complex”, phosphorylates four conserved serine and threonine residues in the N-terminus of β-catenin, marking it for proteosome-mediated degradation. Wnt-elicited signaling cascades ultimately leads to inactivation of GSK-3β and hence β-catenin stabilization (Xue and Zhao, 2012). The accumulated β-catenin then translocates into the nucleus where it interacts with the TCF-LEF transcription factors and a myriad of other factors to modulate gene expression (Mosimann et al., 2009). Albeit a causative involvement of activated Wnt-β-catenin pathway in human T-cell malignancy has not been established thus far, forced expression of stabilized forms of β-catenin in mice results in T-cell malignancy that resembles T-ALL (Guo et al., 2007). However, it remains unknown if TCF-1 and LEF-1 are involved in malignant transformation of developing thymocytes.

Here we have reported a surprising finding that Tcf7^−/− thymocytes are highly susceptible to malignant transformation, indicating a role of TCF-1 in tumor suppression. Mechanistically, the tumor suppressive effect of TCF-1 was mediated by directly restraining the expression of LEF-1, revealing an unexpected interplay between TCF-1 and LEF-1, beyond their redundant functions in promoting thymocyte maturation. This notion was further supported by in vivo evidence that elimination of LEF-1 greatly delayed or prevented malignant transformation of TCF-1-deficient thymocytes. In-depth analyses of TCF-1 and LEF-1 double deficient thymocytes revealed that TCF-1 and LEF-1 were not required for T-cell specification or commitment, but were rather indispensable for β-selection and maturation beyond the DN4 stage. These observations elucidate dual roles of TCF-1, i.e., acting cooperatively with LEF-1 to promote β-selection at the DN3 stage and restraining LEF-1 expression to prevent transformation of early thymocytes.

RESULTS

TCF-1-deficient mice develop T-cell lymphoblastic lymphomas

During our studies on the role of TCF-1 in regulating mature T cell responses (Zhou et al., 2010), we found that Tcf7^−/− mice developed progressive illness involving hunched posture and lethargy, with palpably enlarged spleen and peripheral lymph nodes, and all mice succumbed to death (Figure 1A). Necropsy revealed enlarged thymus and loss of its distinctive bilobed morphology. Histological analysis confirmed complete effacement of the cortical-medullary structure in the thymus and normal structure in the spleen by monotonous lymphoblastic cells (Figure 1B). To determine the origin of the neoplastic cells, we performed immunophenotypic analysis of lymphoma cells and found that they were negative for B220, CD19, NK1.1, and myeloid lineage markers, but were positive for proteins characteristic of T-lineage cells. Of 39 individual lymphomas analyzed, 31 (79.4%) were CD3⁺TCRβ⁺, and among these, 16 were CD25⁺CD44⁻ (corresponding to the DN3 stage), 11 were CD25⁻CD44⁻ (DN4 stage), and 4 were a mixture of both populations (Figure 1C). Interestingly, some of these lymphomas exhibited aberrant expression of CD8, reminiscent of ISP cells (Figure 1C). On the other hand, 8 out of 39 lymphomas (20.6%) were negative for CD3 and TCRβ but were CD4^high and expressed varied amounts of CD8 (Figure 1D), loosely corresponding to the CD4⁺CD8⁺ DP stage.These data indicate that the neoplastic cells in Tcf7^−/− mice are derived from T-cell origin.

Figure 1. Tcf7^−/− mice develop T-cell lymphoblastic lymphomas.

(A) Kaplan-Meier survival curve of Tcf7^−/− mice. A cohort of Tcf7^−/− mice (n = 24) along with WT and Tcf7^+/− littermates (each >10) were monitored for 40 weeks. None of the control mice developed lymphomas (not depicted).

(B) Histological analysis of thymi and spleens from WT or Tcf7^−/− mice that developed lymphomas. Representative images are shown (black bar = 100 μm). All Tcf7^−/− mice with lymphomas exhibited similar diffuse infiltration and effacement of normal tissue structure by neoplastic cells. Insets in right panels show neoplastic cells on higher magnification (yellow bar = 20 μm).

(C and D) Immunophenotypic analysis of Tcf7^−/− lymphomas. Lymphoma cells were harvested and surface-stained. Initial gating was on lymphoma blasts based on their increased forward and side scatters, and surface markers were analyzed on the blasts. Representative contour plots for 31 DN lymphomas (C) or 8 DP lymphomas (D) are shown. Marked in red on top of the contour plots is the frequency of each immunophenotype observed.

We further analyzed the clonality of TCRβ⁺ lymphomas. Six of 8 independent lymphoma lines expressed a known TCRβ subtype in the mouse TCRβ screen panel (Figure S1A), suggesting that Tcf7^−/− T-cell lymphomas could arise from a single clone. The lymphomas were propagated upon transplantation into C57BL/6 recipients without irradiation preconditioning, demonstrating their malignant properties (Figure S1B). Based on the “Bethesda proposals for classification of lymphoid neoplasms in mice” (Morse et al., 2002), Tcf7^−/− lymphomas can thus be defined as precursor T-cell neoplasm, resembling human T-ALL.

During T-cell development, TCR loci undergo extensive recombination events, and improper repair of the resulting DNA double-strand breaks may lead to chromosomal abnormalities that ultimately result in thymocyte malignancy. To address this possibility, we crossed Tcf7^−/− mice with a Rag1^−/− strain. The Tcf7^−/−Rag1^−/− mice still developed lymphomas (Figure S1C), indicating that Tcf7^−/− lymphomas are independent of Rag activity. This is in contrast to the thymic lymphomas found in mice expressing stabilized β-catenin, where deletion of Rag2 prevented thymocyte transformation (Guo et al., 2007). Quantitative cytogenetics revealed that 5 of 6 lymphomas were diploid with only rare tetraploid cells (Figure S1D), suggesting that chromosome instability may not be a major cause of Tcf7^−/− thymocyte transformation.

Aberrant upregulation of Id2 and LEF-1 in Tcf7^−/− early thymocytes and Tcf7^−/− T-cell lymphomas

Because Tcf7^−/− T-cell lymphomas can develop independent of Rag activity, gain or loss of chromosomes, we hypothesized that perturbations in gene expression due to loss of TCF-1 is a leading cause of malignant transformation of Tcf7^−/− thymocytes. We thus performed transcriptomic analysis of WT and premalignant Tcf7^−/− DN3 cells (each in triplicate) along with 6 primary lymphomas with DN3-like phenotypes. We reasoned that gene expression alterations that were causative in tumorigenesis should be TCF-1-dependent and thus first detected in Tcf7^−/− premalignant cells and that such alterations should be further exaggerated in lymphoma cells. By a direct comparison of WT and Tcf7^−/− DN3 thymocytes, we identified 138 unique genes that were upregulated ≥1.5 fold (p<0.05) in Tcf7^−/− cells and grouped these genes as “TCF-1-repressed gene set”. We then used this gene set in GSEA (Gene Set Enrichment Analysis) to compare lymphomas with DN3 cells. This analysis revealed that 63 TCF-1-repressed genes were positively enriched in lymphomas (Figure 2A and 2B). Functional annotation using DAVID resources identified several transcription regulators including inhibitor of DNA binding 2 (Id2) and LEF-1 (Figure S2A).

Figure 2. Aberrant upregulation of Id2 and LEF-1 in Tcf7^−/− thymocytes and T-cell lymphomas.

(A) Assessment of TCF-1-repressed genes in Tcf7^−/− lymphomas by GSEA. The gene set of TCF-1-repressed genes was analyzed by GSEA against rank-ordered data set of Tcf7^−/− lymphomas vs. WT control (Ctrl) and Tcf7^−/− DN3 cells. The enrichment plot is shown, with genes at the leading edge highlighted in a red rectangle.

(B) Heatmap of the 63 genes in the leading edge as in (A). Id2, Lef1, and Dtx1 are marked by green arrows. For functional annotation of these genes, see Figure S2A.

(C) Validation of increased Id2 and Lef1 transcripts in Tcf7^−/− DN3 thymocytes and lymphomas. Data are means ± s.d. (n ≥ 4 for control and Tcf7^−/−, and n =12 for lymphomas). *, p<0.05; ***, p<0.001 compared with control DN3 cells.

(D) Aberrant expression of Id2 and LEF-1 proteins in Tcf7^−/− T-cell lymphomas. Cell lysates were isolated from control DN thymocytes, independent lines of lymphomas from Tcf7^−/− or Tcf7^−/−Id2^−/− mice. The lysates were western-blotted for Id2 and LEF-1 with β-actin as an equal loading control.

Id2 is a natural dominant negative regulator of E proteins including E2A (Kee, 2009), and deficiency in E2A or transgenic expression of Id2 results in high incidence of T-cell lymphomas (Bain et al., 1997; Morrow et al., 1999; Yan et al., 1997). The elevated LEF-1 expression in Tcf7^−/− T-cell lymphomas appeared counter-intuitive given the known redundant roles of TCF-1 and LEF-1 in promoting thymocyte maturation. However, LEF-1 has been reported to have tumorigenic effects since its ectopic expression causes B lymphoblastic and acute myeloid leukemia (Petropoulos et al., 2008). Increased Id2 and Lef1 transcripts were validated in Tcf7^−/− DN3 thymocytes and lymphoma cells (Figure 2C). Interestingly, Id2 and Lef1expression was also elevated in Tcf7^−/− DN4 and preselect TCRβ^lowCD69⁻ DP thymocytes (Figure S2B and S2C), suggesting that TCF-1 is required for restrained expression of Id2 and LEF-1 throughout early T-cell developmental stages. Id2 protein is downregulated after the DN3 stage and below detection sensitivity in DN thymocytes (David-Fung et al., 2009); however, Id2 was detected in all Tcf7^−/− lymphomas tested, although the expression amounts varied (Figure 2D, lanes 1–4). LEF-1 was detected as a 57 kDa full-length and a 38 kDa short isoform in WT DN thymocytes, and its expression was drastically elevated in all the lymphoma cells (Figure 2D). Taken together, these data indicate that loss of TCF-1 causes abnormal upregulation of Id2 and LEF-1, which may lead to thymocyte transformation.

A role of TCF-1 in human T-ALLs

In human T-ALLs, a subtype that phenotypically resembles ETPs (ETP-ALLs) comprises approximately 15% of T-ALL cases and associates with poor prognosis (Coustan-Smith et al., 2009). The ETP-ALL cases also show increased genomic instability, harboring distinct sets of activating and inactivating mutations from the conventional non-ETP TALLs (Zhang et al., 2012). To determine if the Tcf7^−/− T-cell lymphomas bear resemblance to human disease on a molecular level, we compared gene expression between Tcf7^−/− lymphomas and WT DN3 thymocytes, and identified 299 downregulated and 309 upregulated genes in Tcf7^−/− lymphomas. We then analyzed these two gene sets by performing GSEA on the expression profiles of 12 ETP-ALLs and 40 non-ETP TALLs in a St Jude Children’s Research Hospital cohort (GEO accession GSE28703) (Zhang et al., 2012). Thirty-nine of the downregulated genes in Tcf7^−/− lymphomas were negatively enriched in the ETP-ALL cases (Figure 3A), and among these, TCF7 expression was the most diminished in ETP-ALLs compared with non-ETP T-ALL cases (Figure 3B). On the other hand, 78 of the upregulated genes in Tcf7^−/− lymphomas were positively enriched in ETP-ALL cases, including ID2 (Figure S3). These analyses reveal that the murine T-cell lymphomas caused by TCF-1 deficiency share common deregulated genes with human diseases, and that most importantly, ETP-ALL cases are consistently associated with decreased expression of TCF-1.

Figure 3. Molecular resemblance of Tcf7^−/− lymphomas to human ETP-ALLs.

(A) and (B) GSEA on downregulated genes in Tcf7^−/− lymphomas. The downregulated gene set was run against the gene expression profiles of 12 ETP-ALLs and 40 non-ETP T-ALLs. The enrichment plot is shown in (A), with the red rectangle marking negatively enriched genes in ETP-ALLs.

(B) shows heatmap of the top 10 most differentially regulated genes between ETP-ALLs and non-ETP T-ALLs.

(C) Identification of monoallelic TCF7 deletion in ETP-ALL cases. Genomic DNA from 15 ETP-ALL samples was hybridized to the Affymetrix SNP6 GeneChip Human Mapping Arrays. The signal strength was manually curated for the TCF7 and flanking gene loci on chromosome 5q and shown in the heatmap. The color-coded relative signal strength is displayed below the heatmap. The TCF7 locus was highlighted in green rectangle. The green vertical arrows mark the 2 ETP-ALL cases that harbor monoallelic loss of TCF7. The SNP6 microarray data were deposited in the dbGaP database, accession number phs000340.v1.p1.

In different sets of experiments, 12 of 19 ETP-ALL samples were analyzed with whole-genome sequencing, but no mutations were found in the TCF7 locus (Zhang et al., 2012). However, genomic single nucleotide polymorphism array analysis of 15 of the 19 samples revealed that 2 ETP-ALL cases showed single copy loss of the region flanking the TCF7 gene on chromosome 5q (Figure 3C). This observation suggests that loss of heterozygosity in the TCF7 genes might be an initiation and/or promoting genetic event in transformation of human thymocytes.

TCF-1 directly restrains LEF-1 expression in early thymocytes

Both TCF-1 and LEF-1 can interact with β-catenin coactivator or TLE/GRG (transducin-like enhancer/Groucho-related gene) corepressors to achieve balanced expression of their target genes (Hoverter and Waterman, 2008; Xue and Zhao, 2012). To investigate if TCF-1 directly modulates the expression of LEF-1 and/or Id2, we stimulated sorted DN3 thymocytes with 6-bromo-substituted indirubin-acetoxime (BIO), a specific inhibitor of GSK-3β, to stabilize β-catenin (Zhou et al., 2010). Whereas its inactive analogue N-methylated BIO (MetBIO) had little effect, BIO treatment induced the expression of Axin2, a known target of the Wnt-β-catenin pathway (Figure 4A). Interestingly, stimulation of DN3 thymocytes with BIO repressed the expression of both Id2 and Lef1 (Figure 4A). Because inhibition of GSK-3β may have off-target effects, we introduced a WT or mutant form of β-catenin into DN thymocytes by retroviral transduction. The mutant β-catenin is constitutively activated and stabilized due to an internal deletion of the GSK-3β phosphorylation sites (Tetsu and McCormick, 1999). Although WT β-catenin did not show an apparent effect compared with an empty vector, the mutant β-catenin induced Axin2 and repressed the expression of both Id2 and Lef1 in DN3 thymocytes (Figure 4B), highlighting a specific effect mediated by β-catenin activation. These observations are consistent with the current understanding that β-catenin can actively repress gene expression (Hoverter and Waterman, 2008).

Figure 4. TCF-1-mediated repression of Id2 and LEF-1 in early thymocytes.

(A) Inhibition of GSK-3β repressed Id2 and LEF-1 expression. Sorted DN3 thymocytes were treated with MetBIO or BIO for 6 hrs, and indicated transcripts were quantitatively determined.

(B) Activation of β-catenin repressed Id2 and LEF-1. Lin⁻ DN thymocytes were cultured on monolayer of OP9-DL1 cells overnight and then transduced with retrovirus expressing GFP only (pMIG), WT or mutant form of β-catenin. After 24 hrs, GFP⁺DN3 thymocytes were sorted and measured for expression of indicated genes. For (A) and (B), data are means ± s.d. from 3 independent experiments. **, p<0.01; ***, p<0.001.

(C) Schematic showing conserved TCF-1 binding motifs in the Lef1 locus. Shown on the top is the cross-species conservation of the Lef1 regulatory sequences at −10 to +5 kb region based on the UCSC genome browser. Conserved TCF-1 binding motifs and their relative locations are marked. Also see Figure S4C for sequence alignments at the −4.4 kb cluster (TBC).

(D) Direct binding of TCF-1 to the Lef1 locus. ChIP was performed on DN3 thymocytes, and all conserved motifs in (C) were assessed for enriched TCF-1 binding. The Axin2 T2-T3 regulatory region was used as a positive control, and the Gapdh gene body as a negative control. Data are pooled results from 2 independent experiments with each motif measured in triplicates.

(E) Schematics showing the retroviral reporter constructs. The retrovirus is self-inactivating due to mutations in its long terminal repeats (SIN). PGK, phosphoglycerate kinase promoter. LefPro, murine Lef1 promoter. Arrows denote the transcription initiation sites and orientations.

(F) The TCF-1 binding cluster confers LEF-1 repression. Lin⁻ DN thymocytes were infected with the reporter retroviruses as in (B). The expression of Thy1.1 reporter, in terms of frequency and MFI, was determined on GFP⁺ DN3 thymocytes. Similar results were obtained for other DN subsets (not shown). Gating of Thy1.1⁺ cells was based on background staining in corresponding GFP⁻ subset. Data are representative of 2 independent experiments with similar results (n = 4).

To determine if the Id2 and Lef1 loci are directly modulated by TCF-1, we scanned 15 kb regulatory regions (−10 kb to +5 kb) flanking the transcription initiation sites of both genes for consensus TCF binding motif, 5′-TCAAAG. We then performed chromatin immunoprecipitation (ChIP) with a TCF-1 antibody on DN3 thymocytes. Survey of the cross species-conserved motifs by quantitative polymerase chain reaction (PCR) revealed enriched binding of TCF-1 at a cluster of three motifs located around −4.4 kb in the Lef1 locus (designated as TCF1 binding cluster or TBC) (Figure 4C and 4D). Direct binding of TCF-1 was not found within the 15 kb region in the Id2 gene, suggesting that Id2 may not be directly regulated by TCF-1 or alternatively regulated through more distal elements. We also confirmed β-catenin-mediated repression of Id2 and LEF-1 and direct binding of TCF-1 to the TBC in the Lef1 genes in preselected DP thymocytes (Figure S4A and S4B).

To investigate if the TBC confers repression of Lef1 expression, we performed reporter assays using self-inactivating retroviruses to deliver reporter genes and their respective control elements into primary thymocytes (Rosenbauer et al., 2006). The retrovirus contains two reporters: the GFP reporter is driven by phosphoglycerate kinase promoter and serves as a marker for selection of virally infected cells, and the Thy1.1 reporter is driven by the murine Lef1 promoter (−333 to +81 bp) (Figure 4E). Upstream of the Lef1 promoter, we inserted either the WT TBC segment (~350 bp, −4637 to −4289) or a mutant TBC where all three TCF-1 binding motifs were mutated (Figure 4E). The Lef1 promoter led to expression of Thy1.1 in >50% GFP⁺ DN3 thymocytes, and insertion of the WT TBC greatly diminished the frequency of cells expressing Thy1.1 and the amount of Thy1.1 protein as measured by mean fluorescent intensity (MFI) (Figure 4F). Importantly, mutation of TCF-1 binding motifs in the TBC partly relieved the repressive effect, resulting increased frequency of Thy1.1⁺ cells as well as Thy1.1 protein expression (Figure 4F). These data thus strongly support that TCF1 directly restrains LEF-1 expression in early thymocytes.

TCF-1-mediated repression of Id2 and LEF-1 suppresses T-cell malignancy

Based on the observation that TCF-1 utilized β-catenin to repress LEF-1 and Id2 expression in early thymocytes, we next investigated if deregulation of Id2 or LEF-1 led to transformation of Tcf7^−/− thymocytes. By crossing Tcf7^−/− mice to an Id2^−/− background (Yokota et al., 1999), we found that elimination of Id2 in Tcf7^−/− mice delayed the onset of lymphomas and reduced the penetrance rate of the disease at 40 weeks (Figure 5A). Interestingly, the T-cell lymphomas in Tcf7^−/−Id2^−/− mice retained abnormal upregulation of LEF-1 (Figure 2D, lanes 5 and 6), suggesting that abnormal upregulation of LEF-1 may have contributed to transformation of Tcf7^−/−Id2^−/− cells.

Figure 5. Elimination of Id2 or LEF-1 delayed or prevented malignant transformation of Tcf7^−/− thymocytes.

(A) Elimination of Id2 delayed onset of lymphomas in Tcf7^−/− mice. Shown are Kaplan-Meier survival curves for Tcf7^−/−Id2^−/− and Tcf7^−/− littermates.

(B) Effective elimination of LEF-1 protein using Vav1-Cre. Cell lysates from bone marrow (BM) cells and thymocytes (Thy) of Lef1^−/− or control mice were western-blotted for LEF-1 expression.

(C) Elimination of Lef1 transcripts in DN3 thymocytes. Lef1 transcripts were measured by quantitative RT-PCR in sorted DN3 thymocytes and normalized to Hprt1. N.D., not detectable. Similar results were obtained for DN4 cells (not shown).

(D) Elimination of Lef1 transcripts in HSCs. Bone marrow cells were harvested and sorted for lineage-negative Sca1⁺c-Kit⁺Flt3⁻ subset, which contains both long-term and short-term HSCs, followed by measurements of Lef1 transcripts as in (C). Data in (C) and (D) are means ± s.d. (n = 4 from 2 independent experiments).

(E) Total thymic cellularity. Data are means ± s.d. (n = 5).

(F) DN3 thymocyte numbers. Data are means ± s.d. (n ≥ 7). ***, p<0.001. N.S., not statistically significant.

(G) Elimination of LEF-1 greatly delayed onset and partly protected Tcf7^−/− mice from T-cell lymphomas. Shown are Kaplan-Meier survival curves for Lef1^−/−Tcf7^−/− and Tcf7^−/− littermate controls.

Since germline disruption of both TCF-1 and LEF-1 was embryonic lethal, we conditionally targeted the Lef1 gene by inserting 2 LoxP sites to flank exons 7 and 8, which encode the DNA binding domain of LEF-1 (Figure S5A). Upon germline transmission (Figure S5B), we used a Flippase transgene to remove the neomycin-resistant gene to minimize disturbance of the native Lef1 gene locus, and the resulting allele was designated as Lef1-floxed allele (Lef1^FL/+) (Figure S5C). To ablate LEF-1 protein in thymocytes at early developmental stages, we used a Vav1-Cre transgene which initiates excision of the floxed sequences in early hematopoietic progenitor cells (Ogilvy et al., 1999). Western blotting confirmed efficient elimination of the LEF-1 protein in total bone marrow cells and thymocytes from Vav1-Cre⁺:Lef1^FL/FL mice (referred to as Lef1^−/−) (Figure 5B). We then used more sensitive quantitative RT-PCR to validate complete elimination of Lef1 transcripts in DN3 thymocytes from Lef1^−/− or Vav1-Cre⁺:Lef1^FL/FLTcf7^−/− mice (referred to as Lef1^−/−Tcf7^−/−) (Figure 5C). Lef1 transcripts were detected at much lower quantity in hematopoietic stem cells (HSCs), and deletion of Lef1 was complete in both Lef1^−/− and Lef1^−/−Tcf7^−/− HSCs (Figure 5D). TCF-1 deficiency alone substantially diminished total as well as DN3 thymocytes (Schilham et al., 1998; Verbeek et al., 1995). Elimination of both TCF-1 and LEF-1 did not cause further significant reduction in total thymic cellularity or the DN3 subset (Figure 5E and 5F), and definitive T lymphopoiesis did occur in Lef1^−/−Tcf7^−/− mice (detailed below). We have monitored a cohort of Lef1^−/−Tcf7^−/− mice along with their Tcf7^−/− littermates. All the Tcf7^−/− littermates succumbed to T-cell lymphomas by 40 weeks of age. In contrast, 10 out of 13 Lef1^−/−Tcf7^−/− mice survived the 45-week observation period without overt signs of lymphoma formation (Figure 5G). The other 3 Lef1^−/−Tcf7^−/− mice that died between 30 to 45 weeks were found to have enlarged thymi, and the malignant thymocytes showed a phenotype of CD4⁻CD8⁻CD3⁻TCRβ⁻, consistent with a complete block of T-cell development at the DN stage in the absence of TCF-1 and LEF-1 (detailed below). Thus, deletion of LEF-1 greatly delayed or prevented malignant transformation of Tcf7^−/− thymocytes.

The involvement of Notch signaling in Tcf7^−/− T-cell lymphomas

Somatic activating mutations of NOTCH1 have been found in over 50% of all T-ALL cases (Aifantis et al., 2008; Grabher et al., 2006) and in several mouse models of T-cell lymphomas (Lin et al., 2006; O’Neil et al., 2006). Deltex1 (encoded by Dtx1), a known Notch target gene, exhibited increased expression in both Tcf7^−/− DN3 thymocytes and T-cell lymphomas (Figure 2B). We validated these changes in Dtx1 and further found that Notch1 itself and other Notch target genes, including Ptcra (encoding pre-TCRα) and Hes1, all exhibited moderately increased expression in Tcf7^−/− and Lef1^−/−Tcf7^−/− DN3 thymocytes (Figure 6A). However, the expression of Notch1 and its other targets were not increased, but rather decreased, in Tcf7^−/− lymphomas (Figure 6A), suggesting deregulation of Notch signaling in transformed cells.

Figure 6. Notch1 in Tcf7^−/− thymocytes and T-cell lymphomas.

(A) Notch signaling was moderately enhanced in Tcf7^−/− DN3 thymocytes but was not sustained after transformation. Sorted DN3 thymocytes were measured for the expression of Notch1 and its target genes, in direct comparison with Tcf7^−/− lymphomas. Data are means ± s.d. from 3 independent experiments (n = 4~8).

(B) Inhibition of γ-secretase did not repress Lef1 expression in DN3 thymocytes. Lin⁻ DN thymocytes were cultured on OP9-DL1 stromal cells in the presence of DMSO or GSI for 24 hrs. DN3 thymocytes were sorted and measured for indicated transcripts. Data are means ± s.d. from 2 independent experiments (n = 4).

(C) Forced expression of ICN or DN-MAML did not detectably affect Lef1 expression. Lin⁻ DN thymocytes were infected with pMIG retrovirus or that expressing either ICN or DN-MAML. The GFP⁺ DN3 cells were sorted 24 hours post-infection and measured for indicated transcripts. Data are means ± s.d. from 3 independent experiments (n = 6). For (A)–(C), *, p<0.05; ***, p<0.001; N.S., not statistically significant.

(D) Cleaved Notch1 was expressed in Tcf7^−/− but not Lef1^−/−Tcf7^−/− T-cell lymphomas. Cell lysates from the lymphomas or total thymocytes were western-blotted to detect cleaved Notch1, with β-actin as a loading control.

(E) Notch1 was mutated in the PEST domain in Tcf7^−/− T-cell lymphomas. Notch1 cDNA was amplified from thymocytes or lymphomas and sequenced for marked regions. For Notch1, the first nucleotide in its open reading frame was designated as 1, and the coding sequence in exon 34 is within 6151 to 7597. Mutation sites were determined based on this numbering.

It was reported that Notch1 signaling is essential for survival of T-cell lymphomas caused by E2A deficiency and that LEF-1 acts downstream of Notch1 in this context (Spaulding et al., 2007). Since Tcf7^−/− thymocytes showed increased expression Notch1 and its target genes, we next investigated if elevated Notch signaling may induce LEF-1 expression in early thymocytes. To this end, we treated DN thymocytes with a γ-secretase inhibitor (GSI) to block the release of transcriptionally active intracellular domain of Notch1 (ICN). GSI treatment effectively diminished the expression of Dtx1 and Ptcra as expected, but did not repress Lef1 or Notch1 expression in DN3 cells (Figure 6B). We also retrovirally introduced a dominant-negative form of the coactivator mastermind-like (DN-MAML) into DN thymocytes to specifically disrupt the Notch1/CSL/MAML transactivation complex (Maillard et al., 2004). The DN-MAML led to suppression of Notch1 targets Dtx1 and Ptcra but did not affect Lef1 or Notch1 (Figure 6C). Conversely, forced expression of ICN in DN thymocytes induced Dtx1 and Ptcra expression, without an apparent effect on Lef1 or Notch1 (Figure 6C). Thus, the induction of LEF-1 by Notch signaling appears to be more specific to selected cell context and is not operating in T-lineage-committed early thymocytes.

Although the Notch signaling is deregulated in Tcf7^−/− lymphomas, cleaved Notch1 proteins were readily detected in all Tcf7^−/− lymphomas tested, with their molecular weights being smaller than the expected 120 kD (Figure 6D). The truncated forms of Notch1 protein in the lymphomas suggested possible mutations, which are frequently found in the heterodimerization (HD) domain (coded by exons 26 and 27) and proline, glutamic acid, serine, threonine-rich (PEST) domain (coded by exon 34) in human TALLs and murine T-cell lymphomas (Grabher et al., 2006; Lin et al., 2006; O’Neil et al., 2006). Mutations in the HD domain facilitate Notch1 cleavage by γ-secretase, and mutations in the PEST domain prolong Notch1 half-life (Grabher et al., 2006). By sequencing the Notch1 cDNA, we found mutations in the PEST domain in all 6 Tcf7^−/− lymphomas examined, but found no mutations in the HD domain in these samples (Figure 6E). Surprisingly, none of the 3 T-cell lymphomas in aged Lef1^−/−Tcf7^−/− mice expressed cleaved Notch1 (Figure 6D), and none of them had Notch1 mutations in the HD or PEST domains (Figure 6E). It is of interest to note that Notch1 was not mutated in the premalignant Tcf7^−/−or Lef1^−/−Tcf7^−/− DN3 thymocytes (Figure 6E), although Notch1 and its target genes were upregulated in these cells (Figure 6A). It is thus likely that Notch1 mutations are acquired after malignant transformation of Tcf7^−/− thymocytes and such mutations are dispensable for transformation in the context of TCF-1 and LEF-1 deficiency.

TCF-1 and LEF-1 are dispensable for T cell-lineage commitment but are required for β-selection and further maturation beyond the DN4 stage

Our studies thus far revealed an unexpected role of TCF-1 in restraining LEF-1 expression to prevent thymic tumorigenesis. A recent report demonstrated that forced expression of TCF-1 in bone marrow progenitors is sufficient to promote T-cell specification and activate the T-lineage differentiation program (Weber et al., 2011). To further explore the requirement of TCF-1 and LEF-1 in the T-cell developmental process, we performed in-depth analysis of the developmental state of Lef1^−/−Tcf7^−/− thymocytes. TCF-1 and LEF-1 double deficiency blocked T cell development at the DN stage (Figure 7A), and this block was earlier than the previously reported ISP block in fetal thymic organ culture of TCF-1-hypomorphic LEF-1-null fetal liver cells (Okamura et al., 1998). In some Lef1^−/−Tcf7^−/− mice, thymocytes that were positive for CD4 and/or CD8 could be detected at low frequencies, but they were devoid of cell surface TCRβ expression (Figure S6A), suggesting that Lef1^−/−Tcf7^−/− thymocytes cannot mature beyond the DN stage to acquire a functional TCR. Consistent with this, no mature T cells were detected in the Lef1^−/−Tcf7^−/− spleens (Figure S6B). Further analysis of DN subsets showed that TCF-1 and LEF-1 double deficiency did not cause more severe blocks at DN1 or DN3 stages compared with loss of TCF-1 alone (Figure 7B). Importantly, the Lef1^−/−Tcf7^−/− DN3 thymocytes expressed intracellular TCRβ at a similar frequency as control or Tcf7^−/− DN3 cells (Figure 7C and 7D). Furthermore, assessment of DJ and V to DJ recombination at the Tcrb locus revealed that similar to Tcf7^−/− cells, Lef1^−/−Tcf7^−/− thymocytes contained all the expected rearranged segments (Figure S7). Coupled with the observation that deletion of LEF-1 did not further reduce DN3 cell numbers in the context of TCF-1 deficiency (Figure 5F), our data indicate that TCF-1 and LEF-1 are not required for generation of DN3 thymocytes and TCRβ rearrangement, the definitive events for T-lineage commitment.

Figure 7. A cooperative function between TCF-1 and LEF-1 is required for β-selection at the DN3 stage.

(A) TCF-1 and LEF-1 double deficiency arrested T-cell development at the DN stage. Lin⁻ thymocytes were fractioned based on CD4 and CD8 expression.

(B) Ablation of LEF-1 did not further exacerbate T-cell developmental blocks caused by TCF-1 deficiency within the DN stage. Lin⁻CD4⁻CD8⁻ thymocytes were fractionated based on CD25 and CD44 expression. Data in (A) and (B) are representative of at least 3 independent experiments with similar results. The percentage of each subset is marked.

(C) TCF-1 and LEF-1 are required for efficient β-selection. DN3 and DN4 thymocytes were intracellularly stained for TCRβ expression, and the percentage of TCRβ⁺ population is marked in representative histograms.

(D) Cumulative frequency of intracellular TCRβ⁺ DN3 and DN4 thymocytes. Data are means ± s.d. (n = 3). *, p<0.05; **, p<0.01; and ***, p<0.001.

β-selection at the DN3 stage allows those with productive TCRβ rearrangements to mature to DN4 thymocytes, with majority of these cells expressing intracellular TCRβ protein (Rothenberg and Taghon, 2005). β-selection was not affected by LEF-1 deficiency only and was less efficient in Tcf7^−/− mice (Figure 7C and 7D). Strikingly, no increase in intracellular TCRβ⁺ cells was found in Lef1^−/−Tcf7^−/− DN4 cells (Figure 7C and 7D), suggesting impaired β-selection and/or pre-TCR signaling events in the absence of both TCF-1 and LEF-1. These data collectively indicate that whereas definitive T-lineage specification and commitment can occur in the absence of TCF-1 and LEF-1, these two factors are indispensable for β-selection and further maturation beyond the DN4 stage.

DISCUSSION

In this study, we demonstrated an unexpected interplay between TCF-1 and LEF-1. They belong to the same TCF-LEF subfamily of HMG box-containing transcription factors and recognize the same 5′-TCAAAG consensus sequence via their conserved HMG DNA binding domain. Existing genetic evidence indicates that TCF-1 and LEF-1 have cooperative roles in promoting thymocyte maturation (Okamura et al., 1998). Our new findings revealed that restraint of LEF-1 by TCF-1 is critical to fulfill their physiological function in T-cell development and avoid detrimental transformation. In the context of colon cancer cells where constitutively activated β-catenin is a known cause of colorectal malignancy, the TCF-4/β-catenin complex was reported to upregulate LEF-1 and TCF-1 (Hovanes et al., 2001; Roose et al., 1999). In developing thymocytes, TCF-1 and LEF-1 are both abundantly expressed from the DN3 stage onward, and TCF-1 utilizes β-catenin to actively repress LEF-1 expression. We showed direct binding of TCF-1 to a −4.4 kb cis-regulatory sequence in the Lef1 locus and further demonstrated that this sequence was sufficient to confer repression of Lef1 expression in primary thymocytes. Most importantly, elimination of LEF-1 greatly delayed or prevented malignant transformation of TCF-1-deficient early thymocytes. These findings thus elucidate a new form of genetic interaction among the TCF-LEF factors to ensure their normal functions. Forced expression of TCF-1 in hematopoietic progenitors was recently reported to lead to induction of LEF-1 (Weber et al., 2011). This induction might be secondary to activation of T-cell specification program by TCF-1, or alternatively specific to the cell context of hematopoietic progenitors. These possibilities merit future investigations.

Transformation of Tcf7^−/− thymocytes involves several genetic events. Firstly, in addition to loss of restraint of LEF-1, Tcf7^−/− thymocytes showed elevated Id2 expression. Overexpression of Id2 or deletion of E2A is known to cause T-cell lymphomas (Bain et al., 1997; Morrow et al., 1999; Yan et al., 1997). Deletion of Id2 in the Tcf7^−/− mice indeed delayed the onset of lymphomas. Secondly, PU.1 was reported to be negatively regulated by TCF-1 in early thymocytes through a −14 kb regulatory element, and deletion of this element caused T-cell lymphomas (Rosenbauer et al., 2006). Interestingly, upregulation of PU.1 was not observed in premalignant thymocytes or lymphomas lacking the −14 kb element, and thus a “hit-and-run” mechanism has been proposed for a role of PU.1 in thymocyte transformation (Rosenbauer et al., 2006). Deregulation of Id2 and/or PU.1 might be causative events to the few cases of T-cell lymphomas in mice lacking both TCF-1 and LEF-1. Lastly, improper activation of Notch signaling is considered to promote thymocyte transformation in cooperation with known predisposing genetic alterations (Grabher et al., 2006). Early thymocytes lacking TCF-1 or both TCF-1 and LEF-1 showed similar amounts of increased expression of Notch1 and its target genes; however, they were starkly different in transformation potentials, with all Tcf7^−/− mice succumbing to the disease by 40 weeks and >75% of Lef1^−/−Tcf7^−/− mice being lymphoma-free at 45 weeks of age. Although a role of enhanced Notch signaling cannot be entirely excluded, its contribution to transformation of Tcf7^−/− thymocytes appears to be small compared with aberrant upregulation of LEF-1. Because LEF-1 expression is relatively unaffected by gain or loss of Notch1 signaling in early thymocytes, the LEF-1 upregulation is unlikely secondary to Notch1 activation. Additionally, no activating mutations in Notch1 were found in pre-malignant Tcf7^−/− or Lef1^−/−Tcf7^−/− thymocytes and even the Lef1^−/−Tcf7^−/− lymphomas, suggesting that acquisition of Notch1 mutations is dispensable for transformation of thymocytes lacking TCF-1 or both factors. Notch1 mutations found in the PEST domain in Tcf7^−/− lymphomas appear to be acquired as “addiction” to Notch signaling, which may cooperate with early oncogenic hits to promote onset and/or propagation of transformed cells. This view is in line with the finding that common mutations found in human T-ALLs were insufficient to induce transformation on their own but can accelerate K-ras-initiated leukemia (Chiang et al., 2008).

Gene expression profiling of the Tcf7^−/− lymphomas revealed their molecular resemblance to the ETP subtype of human T-ALLs. ETP-ALLs were consistently associated with decreased expression of TCF-1 and increased expression of ID2, the key characteristics of Tcf7^−/− lymphomas. Of particular interest, two of the 15 ETP-ALL cases had monoallelic deletion of the TCF7 gene. These data suggest that loss of TCF-1, in forms of diminished expression or gene deletion, could be a critical contributing event in thymocyte transformation, adding to in-depth understanding of etiology of human T-cell malignancy. In the context of TCF-1 deficiency, increased expression of LEF-1 in both premalignant thymocytes and T-cell lymphomas is consistent with its reported tumorigenic potentials. Forced expression of LEF-1 in murine bone marrow progenitors results in B lymphoblastic and acute myeloid lymphoma in recipient animals (Petropoulos et al., 2008). In humans, overexpression of LEF-1 is strongly associated with B-cell chronic lymphocytic leukemia (Gutierrez et al., 2010a), and predicts unfavorable outcome in patients with B-cell ALL (B-ALL) (Kuhnl et al., 2011). Paradoxically, deletion of the LEF1 gene has been reported in B-ALLs (Mullighan et al., 2007) as well as T-ALLs (Gutierrez et al., 2010b). There might be several possible explanations for these seemingly contradictory observations. Firstly, many factors, such as Notch1 and Runx proteins, can function as oncogenes as well as tumor suppressors depending on the cellular context (Blyth et al., 2005; Lobry et al., 2011). LEF-1 could act in both capacities in different cell types. Secondly, in one given cell type (for example developing T cells), the role of LEF-1 in cancer may not be fixed but depend on other tumorigenic genetic events. Among T-ALL cases that were previously found to harbor LEF1 deletions, all had homozygous deletions of the CDKN2A gene, and some of them had activating mutations in NOTCH1 or mutations in the PTEN tumor suppressor gene (Gutierrez et al., 2010b). Lastly, LEF-1 can interact with both β-catenin coactivator and TLE/GRG corepressors. Gain in LEF-1 expression may acquire increased interaction with β-catenin and cause aberrant activation of its target genes, and loss in LEF-1 expression may diminish TLE/GRG recruitment and lead to improper gene de-repression. Both changes in LEF-1 could lead to similar detrimental effect in causing transformation of thymocytes.

Conditional targeting of the Lef1 gene also allowed us to further address the physiological requirements of TCF-1 and LEF-1 in normal T-cell development. Forced expression of TCF-1 was recently reported to be sufficient to specify hematopoietic progenitors to T-lineage and activate T-cell-specific gene expression program (Weber et al., 2011). Our study used Vav1-Cre to eliminate LEF-1 expression in hematopoietic progenitors (Ogilvy et al., 1999). Coupled with germline deletion of TCF-1, we found that in the absence of TCF-1 and LEF-1, definitive thymopoiesis did occur, as evidenced by detection of Tcrb rearrangements and intracellular TCRβ proteins in double deficient DN3 and DN4 thymocytes. Accumulation of DN1 and paucity of the DN2 subset were evident in TCF-1 deficiency alone (Goux et al., 2005), and these phenotypes were not further exacerbated by deletion of LEF-1. Thus, TCF-1 and LEF-1 are not required for T-cell specification and commitment, albeit both processes may have proceeded with diminished efficiency in the absence of these factors. The expression pattern of TCF-1 and LEF-1 is consistent with this conclusion. Although both TCF-1 and LEF-1 were expressed at low quantity at the DN1 or ETP stage, TCF-1 is upregulated at the DN2 and LEF-1 expression showed sharp increase at the DN3 stage (David-Fung et al., 2009). Both factors are maintained at high expression levels from DN3 stage onward, and this is also in line with the critical requirements of TCF-1 and LEF-1 in β-selection and maturation beyond the DN4 stage.

In summary, this study has revealed dual roles of TCF-1 in developing thymocytes. It cooperates with LEF-1 to promote β-selection and further thymocyte maturation. At the same time, the LEF-1 expression level must be kept in check by TCF-1 to prevent malignant transformation of early thymocytes. These findings highlight a pivotal interplay among transcription factors in the same family to achieve balanced regulation of thymopoiesis.

EXPERIMENTAL PROCEDURES

Mice

Tcf7^−/− and Id2^−/− mice were previously described (Verbeek et al., 1995; Yokota et al., 1999). Rag1^−/−, Vav1-Cre transgenic, Flippase transgenic mice were from the Jackson laboratory. Generation of Lef1-floxed mice was detailed in the Extended Experimental Procedures. All mouse experiments were performed under protocols approved by the Institutional Animal Use and Care Committee of the University of Iowa.

Microarray and bioinformatics analysis

Total RNA was extracted from sorted WT or Tcf7^−/− DN3 thymocytes and primary lymphoma cells, and GeneChip Mouse GENE 1.0 ST arrays (Affymetrix) were used to probe global gene expression changes (Zhou et al., 2010). Gene expression data have been deposited in the Gene Expression Omnibus database under accession number GSE33292.

Statistical analysis

For monitoring primary lymphoma formation in Tcf7^−/− mice or mice that were double deficient for TCF-1 and Rag1, TCF-1 and Id2, or TCF-1 and LEF-1, all death events were recorded and analyzed using the Statview5.0 software (SAS Institute Inc). The Kaplan-Meier survival curves were generated with the program, and statistical significance was assessed using log-rank test. For comparison of gene expression and β-selection among mice of different genotypes, student’s t test was used for statistical analysis.