Dicer1 imparts essential survival cues in Notch-driven T-ALL via miR-21–mediated tumor suppressor Pdcd4 repression

MicroRNAs (miRNAs) are small noncoding RNA molecules, which posttranscriptionally regulate gene expression.^1,2  Multiple miRNAs have been implicated in the regulation of hematopoiesis and leukemia.^3,4  A global assessment for miRNA requirement within the hematopoietic system using tissue and stage-specific Cre-recombinase-mediated inactivation of Dicer1 has been performed.⁵  These studies revealed that the absolute requirement for Dicer1-processed miRNAs differs between hematopoietic lineages and/or developmental stages. Dicer1 inactivation in hematopoietic stem cells (HSCs) results in the induction of apoptosis and loss of HSC self-renewal and reconstitutive capacity, suggesting that Dicer1 is essential for HSC survival.⁵  Whereas, the requirement of Dicer1 for other blood lineages such as T cells seems to be less stringent. Loss of Dicer1 during T-cell development results in rather mild phenotypes.^6,7 

Aberrant expression of miRNAs has been observed in several cancers including hematologic malignancies. Using Dicer1 loss-of-function (LOF) studies, the global requirement for miRNA biogenesis in B-cell acute lymphoblastic leukemia (B-ALL), myelodysplastic syndrome, and acute myeloid leukemia (AML) was recently assessed.^7-9  It became clear that miRNA dependency varies widely between different types of leukemia.^7-9  Whereas the global requirement of miRNAs in T-cell acute lymphoblastic leukemia (T-ALL) remains elusive, various studies have been carried out identifying several miRNA candidates relevant to T-ALL onset.^10-14  A comprehensive miRNA expression study using human T-ALL samples revealed that 5 highly abundantly expressed miRNAs function in a protumorigenic manner in a mouse model of Notch1-driven T-ALL. Adoptive transfer experiments using bone marrow (BM) cells that express individual oncomiRs together with the dominant active Notch1-intracellular domain (N1ICD) resulted in accelerated leukemia development, demonstrating cooperation with Notch to drive disease development.¹⁰  These miRNAs exhibited overlapping and cooperative effects on various known tumor suppressor genes. Downregulation of individual miRNAs had therefore only limited effects.¹⁰  Other studies identified miR-181a-1/b-1 as a modulator of oncogenic Notch signaling¹²  and showed that repression of miR-451, miR-709, and miR-193b-3p cooperates with N1ICD-driven T-cell leukemia.^11,13  These studies clearly demonstrate that miRNAs can influence disease onset in Notch-driven T-ALL. However, it is unclear whether these effects reflect a critical role of miRNAs during the development/progression of Notch-driven T-ALL, or if they merely represent a potentiating effect following disease onset.

We demonstrate that development, progression, and maintenance of Notch1-driven T-ALL is entirely dependent on Dicer1-mediated miRNA biogenesis. Cell fate–tracing experiments show that Dicer1 is essential for survival of T-ALL cells but does not affect cell cycle. Moreover, microarray-based expression analysis on mouse and human T-ALL samples identified miR-21 as an important, previously unrecognized miRNA in T-ALL that regulates survival, in part, via repression of the tumor suppressor Pdcd4.

Dicer1^lox/lox,¹⁵  CD4-Cre,¹⁶  and Mx1-Cre¹⁷  mice were previously described. Mice were crossed to generate Dicer1^lox/lox CD4-Cre, Dicer1^wt/lox CD4-Cre, and Dicer1^lox/lox Mx1-Cre mice and backcrossed into C57BL/6 for at least 6 generations. C57BL/6 CD45.1 (Jackson Laboratories, ME) and Rag2γc^−/− (Rag2^−/−γc^−/−; Taconic Europe, Denmark) were initially purchased, then bred and maintained at the Ecole Polytechnique Fédérale de Lausanne animal facility. Mx1-Cre induced gene inactivation was achieved by intraperitoneal injections of 2 μg/g body weight polyI:polyC (p-I:C; Invivogen) at 2-day intervals. Genomic deletion of Dicer1 was assessed by polymerase chain reaction (PCR; primer sequences are listed in supplemental Methods; see the Blood Web site).

Single cell suspensions were prepared from spleen, thymus, BM, or peripheral blood lymphocyte (PBL) of corresponding mice. T-cell subsets were analyzed as previously described.¹⁸  Expansion of T-ALL cells was monitored by flow-cytometric analysis of nerve growth factor receptor + (NGFR⁺) cells in lymphoid organs. For clonality studies and miRNA expression assays, NGFR⁺ T-ALL cells were sorted using the FACSAria II cytometer (Becton Dickinson). The purity of all sorted subsets was >97%. Flow-cytometric data were acquired on a Cyan flow cytometer (Beckman Coulter). Analysis of flow-cytometric data were carried out using the FlowJo software (version 9.2, Tree Star). A detailed list of primary and secondary antibody conjugates is listed in the supplemental Methods section.

Late-stage primary Notch1-driven T-ALL tumor cells were isolated from spleens of T-ALL-bearing mice and expanded in vitro (RPMI 1640–Glutamax supplemented with 10% fetal bovine serum, 25 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid, 50 μM β-mercaptoethanol, and 50 μg/mL gentamicin) to generate 2 independent T-ALL cell lines, M295 and R112 (both of Dicer1^lox/lox hN1ICD-IRES-NGFR genotype), for use in in vitro studies.

A lentiviral lineage-tracing system based on the conditional expression of enhanced green fluorescent protein (eGFP) flanked by loxP sites was generated to monitor the fate of Dicer1-deficient T-ALL cells. eGFP is expressed in T-ALL cells in which Cre^ERT is inactive (Dicer1-proficient cells). Upon Cre activation, eGFP is deleted concomitantly with Dicer1. Fluorescence-activated cell sorter (FACS) sorting of GFP^hi vs GFP^lo cells allows for differential assessment of Dicer1^lox/lox vs Dicer1^−/− cell fate.

LNA transfection solution was prepared mixing 100 μL OptiMEM medium (Gibco) with 40 pg Fluorescein amidite (FAM)-LNA and 9 µL HiPerFect (Qiagen). Following incubation at room temperature for 10 minutes, 100 µL transfection solution (100 nM final LNA concentration) was added dropwise to 250 μL supplemented RPMI 1640 medium containing 50 000 T-ALL cells. Following incubation for 72 hours at 37°C, effects of miRNA inhibition were assessed by comparing anti-miR LNAs with control. Typically, T-ALL cell transfection yielded 20% to 60% of FAM⁺ cells 3 days posttransfection. For double knockdown (KD) experiments of miR-21 and Pdcd4, T-ALL cells (0.2 × 10⁶) were transfected with FAM-LNA-Ctr or FAM-LNA-21 and either Cy5-siRNA-Ctr or a mixture of 2 siRNAs (100 nM final) targeting human or mouse programmed cell-death protein 4 (Pdcd4), respectively (Cy5-siRNA-Pdcd4-1 and Cy5-siRNA-Pdcd4-2; see supplemental Methods). FAM⁺Cy5⁺ cells were assessed 3 days posttransplantation by flow cytometry using Annexin-V for detection of apoptotic cells. Pdcd4 KD efficiency was assessed by quantitative reverse-transcription (qRT) PCR on FACS-sorted FAM⁺Cy5⁺ cells. B-cell lymphoma–extra large (BCL-xL) expression was assessed using fixed cells (BD Cytofix) and followed by incubation with the primary anti-BCL-xL antibody (rabbit monoclonal α-BCL-xL) or rabbit immunoglobulin G isotype control. For detection, cells were incubated with the secondary anti-rabbit immunoglobulin G antibody (AlexaFluor405 coupled). Antibody details are listed in supplemental Methods.

T-cell clonality of NGFR⁺ primary T-ALL cells or T-ALL cell lines was assessed by T-cell receptor β rearrangement PCR.¹⁹  miRNA expression studies were performed on total RNA from primary T-ALL cells or T-ALL cell lines prepared using the miRNeasy extraction kit (Qiagen). Isolated RNA was polyadenylated and reverse transcribed according to manufacturer’s instructions using the miRNA first-strand cDNA Synthesis Kit (Agilent Technologies). All primer sequences used are listed in supplemental Methods.

All animal work was carried out in accordance with Swiss national guidelines. This study was reviewed and approved by the cantonal veterinary service (Service vétérinaire cantonal de Vaud).

To investigate whether Dicer1-mediated miRNA processing is essential for the development of Notch1-driven T-ALL, we used a genetic LOF approach and crossed conditional Dicer1 mice with CD4-Cre transgenic mice^15,16  (Dicer1^Δ/ΔCD4-Cre). In this model, Dicer1 is inactivated as thymocytes or leukemogenic cells transit from the double-negative (DN) to the double-positive (DP) stage. Lineage negative (Lin⁻) CD45.2⁺ BM cells from Dicer1^lox/lox and Dicer1^Δ/ΔCD4-Cre mice were infected with a bicistronic retroviral (RV) vector expressing hN1ICD-NGFR or NGFR alone and subsequently transplanted into lethally irradiated CD45.1⁺ congenic hosts (Figure 1A). As expected, analysis of thymocytes 8 weeks posttransplantation revealed efficient recombination of the floxed Dicer1 allele in chimeric animals reconstituted with transduced Dicer1^Δ/ΔCD4-Cre cells (Figure 1B). Interestingly, CD4-Cre-mediated inactivation of Dicer1 prevented tumor development (Figure 1C) but did not prevent the generation of Dicer1 mutant NGFR⁻ peripheral T cells or other lymphoid lineages (supplemental Figure 1).^20,21  In contrast, Dicer1-proficient chimeras developed Notch-driven T-ALL, with all animals dying of disease by 91 days posttransplantation with an MST of 80 days (Figure 1C and supplemental Figure 1A). NGFR⁺ CD4⁺CD8⁺ (DP) cells accumulated over time in the peripheral blood cells (PBLs) of Dicer1^lox/lox chimeras, whereas these cells were only transiently present in Dicer1^Δ/ΔCD4-Cre chimeras (Figure 1D-G). This suggests that loss of Dicer1 does not prevent the development of NGFR⁺ preleukemic DP cells, but does impair their tumorigenic potential and/or survival. To ensure that the transient presence of preleukemic NGFR⁺ DP cells in Dicer1 mutant chimeras was not because of the persistence of Dicer1-proficient escaper cells, Dicer1 deletion status and the expression levels of Dicer1-dependent miRNAs were confirmed on sorted early stage NGFR⁺ polyclonal (pre)leukemic cells (Figure 1H-I). The expression levels of Dicer1-dependent miRNAs were significantly reduced in cells sorted from Dicer1-mutant chimeras (Figure 1H), concomitant with efficient Dicer1 deletion (Figure 1I), whereas expression of miR-451, a Dicer1-independent miRNA, was unaffected (Figure 1H). Dicer1-dependent miR-19b miR-20a, miR-92a, and miR-26a have been previously shown to function as oncomiRs in Notch1-driven T-ALL.^10,22 

Previous studies using solid tumor models (eg, lung cancer²³  and retinoblastoma²⁴ ) or a mouse AML model⁷  have demonstrated that Dicer1 can function as a haploinsufficient tumor suppressor.^7,23,24  Therefore, BM chimeras reconstituted with Dicer1^wt/lox or Dicer1^wt/ΔCD4-Cre expressing hNICD-NGFR were established to assess whether Dicer1 may play a similar role in Notch-driven T-ALL (supplemental Figure 2A). Dicer1^wt/ΔCD4-Cre and Dicer1^wt/lox developed T-ALL at a similar rate (supplemental Figure 2B-E), with an MST of 56 days posttransplantation. This suggests that Dicer1 haploinsufficiency does not influence Notch-driven T-ALL development.

Our results so far clearly indicate that Dicer1 is essential for the induction of Notch-driven T-ALL. However, whether established T-ALL cells also depend on Dicer1 function is currently unclear. To assess whether Dicer1 is required for the maintenance of early and late-stage T-ALL cells, we generated Dicer1^{lox/loxMx1-Cre} mice. Chimeras reconstituted with Dicer1^{lox/loxMx1-Cre} BM cells expressing either hN1ICD-NGFR or NGFR alone (control) were subsequently established (Figure 2A). Once ∼20% NGFR⁺ cells were detected in PBLs of both cohorts, Cre-mediated inactivation of Dicer1 was induced. Strikingly, this resulted in an overall survival rate of 50% in Dicer1^ΔMx1-Cre chimeras (Figure 2B). Remaining mice in this cohort had an MST of 82.5 days, whereas Dicer1-proficient chimeras all died of T-ALL by day 75 posttransplantation with an MST of 52 days (Figure 2B and supplemental Figure 3A). Although NGFR⁺ T-ALL cells continuously accumulated in PBLs of Dicer1^lox/lox chimeras, the frequency of these cells dropped to 4% in surviving Dicer1^ΔMx1-Cre chimeras before rising again to 10% at day 60 posttransplantation (Figure 2C), presumably because of cells that escaped Cre-mediated Dicer1 inactivation. To counteract T-ALL relapse, all BM chimeras were subjected to a second p-I:C cycle, which led to the eradication of NGFR⁺ leukemic cells and survival of the remaining mice (Figure 2C-D). Analysis of NGFR⁺ leukemic cells sorted from spleens of leukemic Dicer1^ΔMx1-Cre chimeras (day 75 posttransplantation) revealed that the floxed Dicer1 allele was retained by the majority of tumor cells (Figure 2E). This indicated that T-ALL relapse in Dicer1^ΔMx1-Cre chimeras was because of inefficient deletion of Dicer1, resulting in the persistence and expansion of leukemic Dicer1-proficient cells. As expected, analysis of the BM compartment of surviving Dicer1^ΔMx1-Cre chimeras at 90 days posttransplantation revealed the complete loss of the CD45.2⁺ transplant confirming previous reports (supplemental Figure 3B-E).⁵ 

To investigate whether Dicer1 is essential for late-stage T-ALL, we established BM chimeras using donor hN1ICD-NGFR transformed Dicer1^lox/lox and Dicer1^{lox/loxMx1-Cre} BM cells. Once T-ALL was established, clonal NGFR⁺ leukemic cells were transplanted into secondary Rag2^−/−γc^−/− hosts, and continuous p-I:C administration was initiated 12 days posttransplantation (supplemental Figure 4A-C). Ablation of Dicer1 significantly prolonged the survival (2.8-fold increase; P < .001) of corresponding animals (Figure 3A). Despite the attempts to continuously inactivate Dicer1, NGFR⁺ T-ALL cells rapidly accumulated and recipients died of disease (Figure 3B and supplemental Figure 4D). To determine whether late-stage Dicer1-deficient T-ALL cells contributed to the tumor bulk but had impaired tumorigenic potential or continuous counterselection and elimination of late-stage Dicer1-deficient T-ALL cells occurred in recipients, Dicer1 recombination status in PBLs of secondary recipients was analyzed at different time points. Incomplete deletion of Dicer1 was detected in Dicer1^ΔMx1-Cre recipient PBLs at day 23 posttransplantation with undeleted cells accumulating over time (day 31 posttransplantation; Figure 3C), eventually becoming the only detectable tumor fraction by PCR in splenocytes at end point analysis (day 43 posttransplantation; Figure 3C). At this stage, tumor burden in the spleen was comparable between control and Dicer1 mutant cohorts (Figure 3D), indicating complete counterselection against Dicer1-deficient T-ALL cells. To confirm the hypothesis of counterselection of Dicer1-deficient T-ALL cells, 2 independent monoclonal T-ALL cell lines carrying floxed alleles of Dicer1 (M295 and R112) were established (Figure 4 and supplemental Figure 5A-C). Following infection with a lentivirus encoding tamoxifen-inducible Cre^ERT to induce Dicer1 deletion, the cells were serially passaged in the absence of tamoxifen and the maintenance of Dicer1 pro- and deficient cells was monitored over time (Figure 4A). Approximately 50% Dicer1-deficient cells were present 3 days post–tamoxifen treatment. However, Dicer1-deficient cells were progressively lost with serial passages in vitro, indicated by the gradual disappearance of the Dicer1 deletion PCR band 3 and 7 days post–tamoxifen treatment (Figure 4B and supplemental Figure 5D). These results demonstrate that Dicer1-deficient T-ALL cells have a growth disadvantage compared with Dicer1-proficient T-ALL cells.

To investigate the fate of Dicer1-deficient T-ALL cells, mouse T-ALL cell lines were infected with a lentivirus encoding a floxed eGFP reporter gene allowing lineage tracing of tamoxifen-activated Cre^ERT infected cells (Figure 4C-D and supplemental Figure 5E-F). Following tamoxifen treatment, the parental cell lines remained GFP^hi, whereas eGFP^hi and eGFP^lo cells were present in Cre^ERT-expressing cultures. Efficient Dicer1 deletion in sorted eGFP^lo was confirmed by PCR analysis (Figure 4E). The cell cycle status of Dicer1-deficient T-ALL cells (GFP^lo) was comparable to that of undeleted Dicer1-proficient (GFP^hi) T-ALL cells (Figure 4F). However, a significant increase in apoptotic cells was detected in the GFP^lo fraction compared with the GFP^hi fraction and parental T-ALL cells (Figure 4F-G and supplemental Figure 5F-H), strongly suggesting that Dicer1 is necessary for T-ALL cell survival in vitro.

Prosurvival factors provided by the cellular microenvironment in vivo could potentially influence T-ALL cell survival. Therefore, equal numbers of M295^CreERTDicer1-proficient and Dicer-deficient cells (ΔMix 4-OHT) or fully Dicer1-proficient M295^CreERT cells (CTR Et-OH) were transplanted into Rag2^−/−γc^−/− recipients to assess Dicer1-deficient T-ALL cell maintenance in vivo (Figure 4H). Recipients of a mixture of Dicer1-proficient and mutant cells survived slightly, but not significantly, longer than those transplanted with fully Dicer1-proficient cells (MST 20.5 days vs 16 days) (Figure 4I). Importantly, only floxed Dicer1 alleles were detected in leukemic cells derived from moribund recipients, demonstrating that Dicer1 ablated T-ALL cells were also outcompeted by Dicer1-proficient cells in vivo (Figure 4J), strongly suggesting that Dicer1-mediated miRNA biogenesis is required for survival of T-ALL cells.

Induction of apoptosis following conditional loss of Dicer1 in the context of Notch1-driven T-ALL suggested that miRNAs exert an essential prosurvival function. Thus, we reasoned that T-ALL induction and/or maintenance may rely on abundantly expressed miRNAs. Furthermore, miRNAs are differentially regulated during T-cell development^25,26  as well as in cancer.²⁷  We hypothesized that certain prosurvival miRNA candidates may be abundantly expressed at early compared with later stages of T-cell development, and that these may be hijacked in T-ALL cells. The miRNA expression profiles of mouse DN and DP thymocytes were compared with those of primary mouse early stage and late-stage T-ALL samples as well as the M295 mouse T-ALL cell line in order to identify potential deregulated miRNAs in Notch1-driven T-ALL (Figure 5A). The miRNA profile of the human T-ALL cell line CUTLL1 was also assessed. Interestingly, out of 409 detected candidates, miR-21 was found to be the third most abundantly expressed miRNA in mouse T-ALL samples (Figure 5A-B). miR-21 expression was downregulated 15-fold in DP thymocytes compared with DN thymocytes (Figure 5B). Moreover, miR-21 levels were increased 30-fold in both early and late-stage primary murine T-ALL cells compared with DP thymocytes. miR-21 was also abundantly expressed in 2 primary human T-ALL samples as well as in CUTLL1, T-ALL1, DND41, and KOPTK1 cell lines suggesting that miR-21 expression is conserved in both mouse and human T-ALL (Figure 5C-D). Treatment of NOTCH1-dependent CUTLL1, T-ALL1, DND41, and KOPTK1 cells^28,29  for 3 days with the γ-secretase inhibitor DAPT revealed no significant change in the most abundantly expressed miRNAs including miR-21 indicating that these are unlikely to be under the control of Notch signaling (Figure 5C,E). This is in agreement with publically available chromatin immunoprecipitation sequencing data for NOTCH1 and CSL,³⁰  which do not show any binding of NOTCH1 or CSL 10 kb up- or downstream of the transcription start side of the MIR-21 gene (data not shown).

We next decided to assess the biological function of miR-21 in mouse and human T-ALL by miR-21 KD using fluorescein (FAM)–coupled LNAs.³¹  KD of miR-21 in both murine and human T-ALL cell lines did not affect cell cycle regulation (supplemental Figure 6A,C-D) but instead induced apoptosis (Figure 5F and supplemental Figure 6B). These results indicate that miR-21-controlled target genes are involved in regulating the survival of T-ALL cells.

In various solid tumors, miR-21 has been suggested to function as an oncomiR.³²  A comparison of predicted target genes between mouse and human identified 149 conserved miR-21 target genes (supplemental Figure 6E), among which apoptotic regulators have been identified.³³  Using Leukemia Gene Atlas,³⁴  the expression pattern of the 149 conserved miR-21 target genes in BM aspirates of 174 T-ALL patients compared with 73 normal controls or compared with patient derived CD4CD8⁺ DP thymocytes was reanalyzed^34,35  (supplemental Table 1 and supplemental Figure 7). This analysis allowed clustering the samples into 2 groups, which correlated with disease status (T-ALL vs normal BM, and T-ALL vs CD4⁺CD8⁺ DP thymocytes). miR-21 target-gene repression correlated mostly with patient samples from T-ALL compared with control including DP thymocytes. Among these miR-21 target genes, several candidates implicated in apoptosis were identified, including PDCD4^36-40  (Figure 6A). To determine whether miR-21 regulates the expression of some of these conserved genes in T-ALL cells, expression of several miR-21 target genes was assessed by qRT-PCR in mouse and human T-ALL cell lines following miR-21 KD. In murine and human T-ALL cell lines, Pdcd4 was found the most consistently derepressed gene after miR-21 KD (Figure 6B-C and supplemental Figure 8A-C). To assess whether derepression of Pdcd4 is functionally involved in the induction of apoptosis, Pdcd4⁴¹  was ectopically expressed in T-ALL cell lines using an eGFP-traceable Pdcd4 expression vector (Figure 6D). Interestingly, Pdcd4 expression was sufficient in both mouse and human T-ALL cells to induce apoptosis (Figure 6E and supplemental Figure 8D).

To ensure that miR-21 KD-driven induction of apoptosis is mediated by derepression of Pdcd4, we performed double KD experiments for miR-21 and Pdcd4 in T-ALL cell lines. Simultaneous KD of miR-21 and Pdcd4 rescued to a large extent the miR-21 KD-induced apoptotic phenotype in T-ALL cell lines (Figure 6F-G and supplemental Figure 8E). Pdcd4 has been proposed to act at least in part through translational inhibition of BCL-xL,^39,42  which has been implicated in the survival of human T-ALL lines and primary tumors.⁴³  Interestingly, BCL-xL protein levels were downregulated in miR-21 KD T-ALL cells and increased in cells in which miR-21 and Pdcd4 were knocked down simultaneously (supplemental Figure 8F). This strongly suggests that miR-21, through suppression of its downstream target Pdcd4, acts in a prosurvival manner in T-ALL at least in part through stabilizing the protein levels of BCL-xL.

Although several miRNA candidates have been shown to modulate the onset of Notch1-driven T-ALL in mouse models,^8,10,11,13  it remains unclear whether miRNAs are essential for onset and/or maintenance of this disease. Here, we show that CD4-Cre-mediated LOF of Dicer1 prevents the development of Notch-induced T-ALL in BM chimeras in vivo. Although Dicer1 mutant mice did not develop disease, short-lived preleukemic DP T cells were transiently detected. These findings demonstrate that aberrant Notch1 signaling can induce the development of preleukemic cells in the absence of Dicer1, but complete cellular transformation and/or survival is strictly dependent on Dicer1-mediated miRNA biogenesis. Whereas a similar essential role for Dicer1 has been reported for the development of myc-induced B-cell lymphomas,⁹  its requirement in myeloid leukemia seems to differ. miRNAs in myeloid progenitors appear to be involved in regulating the switch of self-renewal vs differentiation,⁷  whereas in ALL they regulate proliferation and/or survival.⁹  Although Dicer1 mutant myeloid progenitors exhibited increased self-renewal, Dicer1-deficient dysplastic cells did not give rise to AML. Interestingly, mono- but not biallelic Dicer1 deletion in myeloid committed progenitors in a p53 mutant background led to the development of AML-like leukemia.⁷  These data suggest that Dicer1 in AML acts as a haploinsufficient tumor suppressor, and loss of both alleles of Dicer1 is sufficient to induce a preleukemic disease stage. Similar observations have also been reported in a Kras-driven model of lung cancer²³  and in a mouse model of retinoblastoma.²⁴  In both models, heterozygous deletion of Dicer1 led to accelerated tumor formation. However, in this study, mice with heterozygous deletion of Dicer1 were found to develop T-ALL like Dicer1-proficient animals. This suggests that Dicer1 does not function as a haploinsufficient tumor suppressor in Notch-driven T-ALL. These results are consistent with the notion that miRNAs in N1ICD-driven T-ALL predominantly act in an oncogenic fashion during progenitor cell transformation, whereas development of AML might rely more on tumor-suppressive miRNAs.

Analysis of global miRNA levels in multiple human tumors revealed a general trend of decreased miRNA content.^44-46  Thus, it is conceivable that a certain number of Dicer1-processed miRNAs are needed for transformation of progenitor cells but may no longer be required in established tumors. Indeed, the requirement for Dicer1 in tumor maintenance appears to vary depending on the type of cancer.²³  A recent study showed that Dicer1^−/− sarcoma cells, completely devoid of miRNAs, continue proliferating in vitro and maintain tumorigenicity in vivo, suggesting that in some types of cancer, Dicer1 activity is not required for tumor maintenance.^47,48  We assessed the role of Dicer1 in the maintenance of early and late-stage T-ALL. Successful ablation of Dicer1 in early established leukemic cells resulted in complete regression of T-ALL cells, and relapse did not occur, in contrast to incomplete Dicer1 inactivation. These Dicer1-proficient T-ALL cell clones continued to proliferate and animals died of disease. Corresponding splenic tumors were primarily composed of Dicer1-proficient cells, indicating that they quickly outcompeted Dicer1-deficient cells. These results imply that Dicer1 LOF may be deleterious for the progression of established T-ALL. This hypothesis was confirmed by studying Mx1-Cre-mediated ablation of Dicer1 in oligoclonal T-ALL cells following serial transplantation into RAG2γc^−/− mice. Animals died of T-ALL because of a strong selective pressure for cells escaping Mx1-Cre-mediated Dicer1 deletion.

The fact that Dicer1-deficient T-ALL cells are rapidly outcompeted indicated an important role for miRNAs in processes relevant for T-ALL maintenance.^49,50  Therefore, we established N1ICD-induced T-ALL cell lines and performed linage-tracing experiments to follow the fate of Dicer1-deficient T-ALL cells. These experiments revealed that Dicer1 has no detectable influence on cell cycle control but is absolutely necessary for survival. This observation is again similar to findings in B-ALL in which Dicer1 ablation in vitro resulted in induction of apoptosis in the majority of B-ALL cells, whereas residual Dicer1-proficient escaper cells continued to proliferate.⁹ 

The findings suggested that certain abundantly expressed miRNAs in T-ALL may execute important prosurvival functions. Thus, we aimed to identify novel prosurvival miRNAs potentially deregulated in T-ALL. We focused on miRNAs that are expressed in DN thymocytes, downregulated in DP thymocytes but abundant in T-ALL cells. Interestingly, only miR-21 fulfilled these criterions. To date there are no reports on the role of miR-21 in thymocyte development or T-ALL; miR-21 has been described as an oncogene in several solid tumors.^51-53  Importantly, miR-21 function has been linked to cancer-related processes in hepatocellular, glioblastoma, and breast cancer.^33,54-56  Furthermore, microarray analysis of miR-21 target gene expression in BM aspirates of T-ALL patients and normal controls revealed that many miR-21 targets were repressed in T-ALL patient samples, indicative of a clinical relevance of miR-21 target gene repression in T-ALL. KD of miR-21 in T-ALL cells led to the induction of apoptosis. Although this largely mimics the Dicer1 LOF effects in T-ALL, miR-21 is certainly not the only miRNA exerting important survival functions. Other miRNAs such as the miR-17-92 gene cluster (data not shown) may be of similar importance. Nevertheless, our data imply that miR-21 functions as an important oncomiR in T-ALL promoting survival of T-ALL cells.

Microarray analysis of miR-21 target gene expression led to the identification of several candidate target genes implicated in survival and apoptosis pathways. Among these, we focused on Pdcd4 because it was consistently downregulated in all murine and human T-ALL samples examined. Pdcd4 was originally characterized as an inhibitor of cellular transformation in mouse epidermal cells.⁵⁷  Subsequent studies showed that Pdcd4^−/− mice spontaneously develop B-cell lymphomas⁵⁸  and are more sensitive to carcinogen-induced skin cancer.^59,60  Moreover, Pdcd4 is a direct miR-21 target and is downregulated in tumors with high miR-21 expression, thus contributing to tumorigenesis including suppression of apoptosis.^39,41,61-64  To test whether derepression of Pdcd4 is sufficient to induce apoptosis, Pdcd4 was ectopically expressed in T-ALL cell lines. Pdcd4 was sufficient to induce apoptosis in T-ALL cells, suggesting that apoptosis of T-ALL cells in which miR-21 expression is downregulated is at least partly caused by derepression of Pdcd4. Recent structural studies on PDCD4 revealed that it inhibits initiation of cap-dependent protein translation by binding to the RNA helicase elongation initiation factor 4A (eIF4A) and the scaffold protein elF4G⁶⁵  or by blocking internal ribosomal entry site–mediated translation of the prosurvival genes Bcl-XL and Xiap.⁴²  Interestingly, our correlative studies on BCL-xL protein levels in murine and human T-ALL cell lines confirms that miR-21 is regulating BCL-xL through the repression of Pdcd4.

Collectively, we demonstrate that miRNAs are essential for development and maintenance of Notch1-driven T-ALL. Furthermore, we identified miR-21 as an important regulator of T-ALL cell survival, which mediates its function in part via repression of the tumor suppressor Pdcd4.

The authors acknowledge Christelle Dubey, Laure Bardouillet, and Marianne Nkosi for technical assistance; Michele de Palma and Mario Squadrito for fruitful discussions; the Flow Cytometry Core Facility for cell sorting; and the DNA array facility of the University of Lausanne for microarray preparation and technical analysis. The authors would also like to thank Marion Leleu from the Bioinformatics Department (Ecole Polytechnique Fédérale de Lausanne) for assistance in bioinformatic data analysis and Dr Jean-Pierre Bourquin, Kinderspital Zürich, and Viktoras Frismantas for providing primary human T-ALL samples.

This work was supported in part by the Swiss National Science Foundation and the Swiss Cancer League (F.R.).

Contribution: F.J. performed experiments, analyzed data, and wrote the manuscript; A.C. performed experiments and analyzed data; U.K. performed experiments, analyzed data, and wrote the manuscript; and F.R. conceived the study, analyzed data, and wrote the manuscript.

Conflict of interest disclosure: The authors declare no competing financial interests.

Correspondence: Freddy Radtke, Ecole Polytechnique Fédérale de Lausanne (EPFL), Swiss Institute for Experimental Cancer Research (ISREC), Station 19, 1015 Lausanne, Switzerland; e-mail: freddy.radtke@epfl.ch.