The Notch driven long non-coding RNA repertoire in T-cell acute lymphoblastic leukemia

Abstract

Genetic studies in T-cell acute lymphoblastic leukemia have uncovered a remarkable complexity of oncogenic and loss-of-function mutations. Amongst this plethora of genetic changes, NOTCH1 activating mutations stand out as the most frequently occurring genetic defect, identified in more than 50% of T-cell acute lymphoblastic leukemias, supporting a role as an essential driver for this gene in T-cell acute lymphoblastic leukemia oncogenesis. In this study, we aimed to establish a comprehensive compendium of the long non-coding RNA transcriptome under control of Notch signaling. For this purpose, we measured the transcriptional response of all protein coding genes and long non-coding RNAs upon pharmacological Notch inhibition in the human T-cell acute lymphoblastic leukemia cell line CUTLL1 using RNA-sequencing. Similar Notch dependent profiles were established for normal human CD34⁺ thymic T-cell progenitors exposed to Notch signaling activity in vivo. In addition, we generated long non-coding RNA expression profiles (array data) from ex vivo isolated Notch active CD34⁺ and Notch inactive CD4⁺CD8⁺ thymocytes and from a primary cohort of 15 T-cell acute lymphoblastic leukemia patients with known NOTCH1 mutation status. Integration of these expression datasets with publicly available Notch1 ChIP-sequencing data resulted in the identification of long non-coding RNAs directly regulated by Notch activity in normal and malignant T cells. Given the central role of Notch in T-cell acute lymphoblastic leukemia oncogenesis, these data pave the way for the development of novel therapeutic strategies that target hyperactive Notch signaling in human T-cell acute lymphoblastic leukemia.

Introduction

The Notch pathway comprises a highly conserved signaling pathway that regulates various cellular processes in all meta-zoans, including stem cell maintenance, regulation of cell fate decisions, cellular proliferation, differentiation, cell death and adult tissue homeostasis.1 As such, Notch signaling is critically involved in many different tissues including epithelial, neuronal, blood, bone, muscle and endothelial cells.2 Precise regulation and duration of Notch signaling activity is of critical importance to ensure appropriate execution of the various developmental cues and cellular processes. Consequently, constitutive or acquired perturbation of Notch signaling frequently leads to human disease and cancer.41

Notch signaling plays multiple roles in hematopoiesis and is essential for the establishment of definitive hematopoiesis through the generation of hematopoietic stem cells,5 as well as for their subsequent differentiation in an expanding number of blood cell types.96 The role of Notch signaling has been particularly well documented in T-cell development where Notch1/Dll4 interactions are crucial to induce T-lineage differentiation at the expense of other hematopoietic lineages.1410 Subsequently, Notch signaling is implemented in TCR- rearrangements,1615 modulation of TCR-αβ versus -γδ development,2117 and in the support of proliferation during β-selection.2422 Sustained activation of Notch1 signaling beyond this developmental checkpoint has been shown to cause T-cell acute lymphoblastic leukemia (T-ALL) and NOTCH1 activating mutations are amongst the most frequently observed genetic alterations in T-ALL.2625 Importantly, γ-secretase inhibitors (GSIs) that block S3 cleavage of the Notch1 receptor and subsequent release of the intracellular signaling domain (ICN) are the subject of intensive investigation as novel drugs to combat T-ALL. However, single compound therapies almost invariably lead to resistance. Therefore, a deeper understanding of Notch signaling in normal thymocyte maturation27 and in Notch1 activated T-ALLs could yield novel insights that could make treatment more effective.

Activation of Notch1 converts the intracellular domain (ICN1) of the Notch1 receptor into a transcriptional activator and ICN1 subsequently acts as a direct regulator of multiple target genes.28 However, despite intensive investigation, the nature of these genes, as well as their context-dependent activation, remains largely elusive. In general, oncogenic Notch signaling promotes leukemic T-cell growth through direct transcriptional upregulation of multiple anabolic genes involved in ribosome biosynthesis, protein translation, and nucleotide and amino acid metabolism. Furthermore, Notch1 positively regulates G1/S cell cycle progression in T-ALL3129 and up-regulates several cyclins and CDKs,30 in addition to the recurrent oncogene MYC. Furthermore, Notch signaling regulates cell size, glucose uptake and PI3K-AKT activated glycolysis through HES1-mediated PTEN repression. Besides direct regulation of HES1, Notch1 is also implicated in the control of essential early T-cell genes such as pre-TCRα (PTCRA) and IL7R.3432 Taken together, these genes and pathways, as well as a further expanding list controlled by Notch1 in T-ALL and normal T-cell development, illustrate the complexity and vastness of the Notch1 controlled regulatory program.

Recent transcriptome-wide profiling efforts have uncovered an unanticipated pervasiveness of transcription of the human genome, most of which is not translated into protein.3835 Evidence is now emerging that more than 60% of the entire genome is transcribed.39 In addition to previously well-characterized untranslated RNA molecules such as tRNAs, snoRNAs and microRNAs (miRNAs), thousands of so-called long non-coding RNAs (lncRNAs) have been annotated to the human genome.4140 Although functional studies still need to be carried out on the vast majority of these lncRNA sequences, important cellular functions are rapidly being attributed to some of them, including roles in disease processes such as cancer.42 In contrast to microRNAs, a picture is emerging in which lncRNAs can exhibit a myriad of different functions. These include various regulatory mechanisms of gene transcription, splicing, post-transcriptional control, protein activity and nuclear architecture.4543 Despite this initial progress, mechanisms of upstream regulation of lncRNAs have so far remained largely unexplored.

In this study, we investigate the role of Notch in the control of lncRNA transcription in the context of normal T-cell development and T-ALL. To this end, lncRNA expression was measured following modulation of Notch signaling in the T-ALL cell line CUTLL1 as well as in normal human thymocytes, and the recently published data on genome-wide Notch1 binding sites was used to reveal the potential for direct regulation.34 Using this approach, we identified a total of 40 Notch-driven lncRNAs, thereby revealing a novel layer in the molecular machinery that mediates Notch signaling.

Methods

GSI treatment of T-ALL cell lines

HPB-ALL, TALL-1, ALL-SIL and CUTLL1 cells (see also Online Supplementary Methods) were seeded at a density of 1×10 cells/mL and treated with either DMSO or 1 μM of Compound E (Enzo Life Sciences). Cells were harvested 12 and 48 h after treatment.

Human thymocytes and OP9-DLL1 co-cultures

Pediatric thymus samples were obtained and used according to the guidelines of the Medical Ethical Commission of the Ghent University Hospital, Belgium. CD34 thymocytes were purified using magnetic activated cell sorting (MACS, Miltenyi Biotec) to a purity of more than 98% and seeded onto confluent OP9-GFP or OP9-DLL1 plates for 48 h in α-MEM media supplemented with 20% heat-inactivated FCS plus 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mM L-glutamine and the T-lineage supporting cytokines SCF, Flt3-L and IL-7 at 5 ng/mL each.20 Following 48 h of OP9 co-culture, cells were harvested by forceful pipetting and stained with CD45-PE (Miltenyi) to purify CD45 human thymocytes through sorting to remove contaminating OP9 stromal cells. For validation of selected lncRNAs, CD34 MACS purified thymocytes were labeled with CD34, CD1 and CD4 to sort CD34CD1CD4 uncommitted and CD34CD1CD4 committed early thymocytes, while CD4CD8CD3 and CD4CD8CD3 double positive thymocytes were sorted following CD4, CD8 and CD3 labeling of a total thymus suspension.20 Sorted cells were lysed in 700μl QIAzol (Qiagen) and stored at −70°C prior to RNA isolation.

Clinical samples

Diagnostic blood samples of 15 individuals with T-ALL were acquired after informed consent from the Department of Pediatric Hemato-Oncology at Ghent University Hospital, Belgium. This cohort includes 8 wild-type NOTCH1 cases and 7 mutant NOTCH1 cases (all FBXW7 wild type). Sequencing was performed as described by Mavrakis et al.46

Correlation analysis was performed on bone marrow lymphoblast samples from 64 T-ALL patients (unknown NOTCH1 mutation status), which were collected after informed consent according to the Declaration of Helsinki from Saint-Louis Hospital, Paris, France. The study was approved by the Institut Universitaire d’Hématologie Institutional Review Board. This primary T-ALL cohort had been previously investigated47 and the high-quality RNA samples from this cohort were used for lncRNA micro-array based expression profiling.

RNA sequencing

RNA samples from the CUTLL1 cells treated with GSI and thymocytes cultured on OP9-GFP/DLL1 were prepared (see also Online Supplementary Methods). RNA-seq was performed after unstranded poly-A library prep with an average coverage of 130×10 paired-end reads. Reads were mapped to the hg19 reference genome using Tophat and transcript assembly was performed with Cufflinks. Differential expression analysis was carried out with DESeq2 in R. The design formula was adjusted to take into account the paired nature of the data.

Micro-array based gene expression profiling

RNA samples (see also Online Supplementary Methods) were profiled on a custom designed Agilent micro-array covering all protein coding genes and 12,000 lncRNAs (23,042 unique lncRNA probes) as described by Volders et al.48 The data-analysis workflow can be found in the Online Supplementary Methods.

The data discussed in this publication have been deposited in the NCBI Gene Expression Omnibus49 and are accessible through GEO Series accession number GSE62006.

Complete details of study methods can be found in the Online Supplementary Appendix.

Results

Pharmacological Notch inhibition followed by RNA-sequencing reveals a set of Notch regulated lncRNAs in T-ALL

To identify lncRNAs that are regulated through Notch signaling activity in the context of T-ALL, we used the γ-secretase (GSI) inhibitor responsive T-ALL cell line CUTLL1 as a model system, since genome-wide information on this cell line is available with respect to Notch1 binding34 and the Notch dependent expression of coding genes.50 CUTLL1 T-ALL cells were treated with GSI for 12 h and 48 h in triplicate. Genome-wide transcriptional changes determined by performing differential expression analysis on the RNA-seq data (see alignment summary in Online Supplementary Table S1) with DESeq2 using Ensembl (release 75) as a reference, showed robust down-regulation of several of the canonical Notch1 protein coding target genes (e.g. DTX1, NRARP, NOTCH3) upon GSI treatment (Figure 1A). A decrease in ICN1 protein levels was shown by Western blot analysis and downregulation of the canonical Notch1 target gene DTX1 upon GSI treatment was further validated by RT-qPCR (Online Supplementary Figure S1). Amongst previously annotated lncRNAs48 we could detect significant differential expression (adjusted P-value <0.05) for 83 lncRNAs, using a base mean cut off of 100 (Figure 1B). In total, 50 out of the 83 differentially expressed lncRNAs were down-regulated after GSI treatment.

Figure 1.Pharmacological Notch inhibition followed by RNA-sequencing identifies a set of Notch regulated lncRNAs in T-ALL. (A) Volcano (left) and MA plot (right) representation of the differential expression of protein coding genes in CUTLL1 cells upon GSI treatment. Red dots represent the significant differentially expressed genes (adjusted P-value <0.05). (B) Volcano (left) and MA plot (right) representation of the differential expression of previously annotated lncRNAs in CUTLL1 cells upon GSI treatment. Red dots represent the significant differentially expressed genes (adjusted P-value <0.05; n=2). LncRNAs names depicted in the plots are the top differentially regulated lncRNAs.

Besides previously annotated lncRNAs, we also detected differential expression of non-coding transcripts that had not been previously annotated in other databases (Gencode, lncRNAdb, Broad Institute and Ensembl release 64).48 Differentially expressed lncRNA loci with a base mean higher than 100 and identified as “unknown, intergenic transcript” or “transfrag falling entirely within a reference intron” by Cuffcompare were retained for further analysis. This led to a selection of 134 lncRNA loci of which 74 were down-regulated upon GSI treatment.

Transcriptional regulation of Notch-regulated lncRNAs in immature normal human thymocytes

Physiological levels of Notch signaling are essential during the earliest stages of T-cell development, but no information is available on the Notch dependent expression of lncRNAs in these cells.27 Therefore, and in order to have an independent screening method in addition to the CUTLL1 cell line to identify Notch dependent lncRNAs, we used the in vitro OP9-DLL1 co-culture system (Figure 2A). Here, ex vivo purified CD34 thymocytes from healthy human donors (n=2) were cultured on a feeder layer of stromal OP9 cells either expressing GFP (as a negative control) or the Notch1 ligand DLL1 to trigger Notch signaling. CD34 progenitor cells were collected after 48 h of co-culture and deep RNA-sequencing was performed (see alignment summary in Online Supplementary Table S2). The set of differentially expressed protein coding genes and lncRNAs was defined as above for the CUTLL1 cells. Detection of differentially expressed protein coding genes known to be regulated by Notch signaling in early human thymocytes,2720 also validated our approach in this model system (Figure 2B). Differential expression analysis for previously annotated lncRNAs revealed 131 significantly up-regulated lncRNAs as a consequence of Notch activation (Figure 2C). From these 131 lncRNAs, 27 overlapped with the set of down-regulated lncRNAs upon GSI treatment of the CUTLL1 cell line (Figure 2D and Online Supplementary Table S3). Furthermore, we identified 156 unannotated lncRNA loci (base mean >100; adjusted P-value <0.05) in CD34 thymocytes that were up-regulated by the Notch ligand DLL1. In total, 13 unique unannotated lncRNA loci were identified to be positively regulated by Notch in both normal and malignant T-cell development (Figure 2E and Online Supplementary Table S4). Amongst the set of 13 overlapping Notch lncRNA loci, the recently described LUNAR151 was present, thus supporting the validity of our approach. In addition, we also identified 33 annotated lncRNAs to be up-regulated upon GSI treatment of CUTLL1 cells by RNA-seq, 18 of them overlapping with the set of lncRNAs down-regulated in CD34 T-cell progenitors upon DLL1 exposure in the OP9 in vitro culture system (366 in total) (Online Supplementary Table S5 and Figure S2A). In a similar manner, 7 of 57 previously unannotated lncRNAs up-regulated upon GSI treatment of CUTLL1 cells overlapped with the set of 320 unannotated lncRNAs down-regulated in CD34 thymocytes with DLL1 exposure (Online Supplementary Table S6 and Figure S2B). Furthermore, we hypothesize that the Notch dependent lncRNAs (both annotated and unannotated) that are not shared between CUTLL1 T-ALL cells and normal human thymocytes can be assumed to have very context-specific functions and should be regarded as potentially interesting for further exploration in future studies. For example, lncRNAs expressed exclusively in T-ALL cells could be restrictively connected to a malignant context. To evaluate the putative protein coding potential of all unannotated lncRNA loci identified by RNA-seq in CUTLL1 T-ALL cells and CD34 T-cell progenitors cultured on OP9 stromal cells, Phylogenetic Codon Substitution Frequency (PhyloCSF) scores for all loci were calculated and we could confirm that more than 90% of all unannotated lncRNA loci determined are truly ‘non-coding’ (Online Supplementary Figure S3). Putative unannotated lncRNA loci with a PhyloCSF score higher than the determined threshold score are listed and thus predicted to be ‘coding’ (Online Supplementary Tables S7 and S8) (see also Online Supplementary Methods).

Figure 2.Transcriptional regulation of Notch regulated lncRNAs in immature normal human thymocytes. (A) Schematic overview of the OP9-control and -DLL1 co-culture system used to manipulate Notch signaling in healthy human immature CD34⁺ thymocytes. (B) Volcano (left) and MA plot (right) representation of the differential expression of protein coding genes in CD34⁺ cells upon Notch signaling induction by an OP9-DLL1 feeder layer. Red dots represent the significant differentially expressed genes (adjusted P-value <0.05; n=2). (C) Volcano (left) and MA plot (right) representation of the differential expression of previously annotated lncRNAs in CD34⁺ cells upon Notch activation by an OP9-DLL1 feeder layer. LncRNA names depicted in the figure are the top differentially regulated lncRNAs. (D) Venn diagram depicting the overlap between previously annotated lncRNAs that are down-regulated upon GSI treatment of the CUTLL1 cell line and up-regulated upon co-culturing of CD34⁺ thymocytes on the OP9-DLL1 feeder layer. (E) Venn diagram depicting the overlap between previously unannotated lncRNAs that are down-regulated upon GSI treatment of the CUTLL1 cell line and up-regulated upon co-culturing of CD34⁺ thymocytes on OP9-DLL1 stromal cells.

Validation of Notch regulated lncRNAs in an extended panel of T-ALL cell lines, normal T-cell subsets and primary T-ALLs

To further validate our data, we used a custom designed Agilent micro-array48 developed in house that contains probes for 15 of the 27 previously annotated lncRNAs and the recently identified LUNAR1 lncRNA that were shown to be regulated by Notch in the above described RNA-seq data from the T-ALL and normal thymocyte models. First, we treated the T-ALL cell lines ALL-SIL, TALL-1, HPB-ALL and DND-41 with GSI (Online Supplementary Figure S4A and B) and carried out gene expression profiling after 12 h and 48 h. Inclusion of the GSI-treated CUTLL1 cell-line samples and the samples of 4 donors of CD34 thymocytes cultured on OP9 stromal cells, revealed that there was a significant overlap between the RNA-sequencing data and the micro-array data as validated by overlapping the protein-coding signatures derived from both datasets by Gene Set Enrichment Analysis (GSEA) (Online Supplementary Figure S4C and D). Nevertheless, few lncRNAs were significantly Notch dependent over all samples of the extended panel of T-ALL cell lines (ALL-SIL, HPB-ALL, DND-41 and TALL-1), probably related to the difference in the T-ALL genetic subgroup and concomitant differences in maturation arrest of the different cell lines evaluated (Figure 3A). From our selection, only lnc-PLEKHB2-1 and lnc-UBXN4-1 were differentially expressed at a significant level, while lnc-GSDMC-2 and lnc-CA7-2 narrowly failed to reach significance. As we were able to detect the previously unannotated and recently described lncRNA LUNAR151 on our custom designed micro-array platform, this lncRNA was one of the strongest overlapping and significantly differentially expressed lncRNAs amongst the four GSI-treated T-ALL cell lines screened. Secondly, we validated the Notch dependency of selected lncRNAs in normal thymocytes by analyzing their expression in the most immature Notch dependent CD34 stages in comparison to the Notch independent CD4CD8 double positive stages of human T-cell development. As is evident from the profiles of the Notch target gene DTX1, we could show that in these two T-cell subpopulations LUNAR1 follows the expression pattern of this canonical Notch target (Figure 3B). Remarkably, 9 out of 15 lncRNAs (and LUNAR1) from this selection significantly correlated with DTX1 expression (Spearman rho correlation), supporting their regulation by Notch during early stages of normal T-cell development (Figure 3C and Online Supplementary Table S9).

Figure 3.Screening expression of Notch regulated lncRNAs in an extended panel of T-ALL cell lines, normal T-cells subsets and primary T-ALLs. (A) Volcano plot representation of the differential expression of lncRNAs upon GSI treatment of ALL-SIL, TALL-1, HPB-ALL and DND-41 cells. Red dots represent the significant differentially expressed genes (adjusted P-value <0.05). LUNAR1 was amongst the top-differentially expressed lncRNAs across the panel of GSI-treated T-ALL cell lines. The other lncRNA names depicted in the figure are some of the selected lncRNAs from the CUTLL1 GSI treatment and the OP9-DLL1 co-culture system. (B) Plot representing the expression of lncRNAs in selected Notch-dependent and -independent stages of normal T-cell development for one healthy donor. LUNAR1 expression is significantly correlated with the expression of DTX1 (see also Online Supplementary Table S6) and the data are representative for 4 independent donors. (C) Similar analysis as in (B) for the other lncRNAs that are significantly correlated with the expression of DTX1 (see also Online Supplementary Table S6); data are representative for 4 independent donors. (D) Expression of LUNAR1, lnc-FAM120AOS-1 and lnc-UBXN4-1 in NOTCH1 wild-type (WT) versus NOTCH1 mutant (MUT) primary T-ALL samples. (E) Gene set enrichment analysis (GSEA) using the public gene set ‘VILIMAS_NOTCH1_TARGETS_UP’⁵² and the Spearman correlations between all protein coding genes and the set of 15 selected annotated candidate Notch lncRNAs was performed. This NOTCH1 signature was significantly enriched within the set of protein coding genes positively correlated to the expression of lnc-PGBD5-2, lnc-FAM120AOS-1 and lnc-c2orf55-1. This enrichment was also found for lncRNA LUNAR1.

Moreover, we also had access to 15 primary T-ALL samples of which 7 harbored activating NOTCH1 mutations while 8 were wild type (all cases are FBXW7 wild type). There was a significant difference in expression of LUNAR1 and lnc-FAM120AOS-1 between NOTCH1 wild type and mutant cases (Figure 3D and Online Supplementary Tables S10 and S11). By implying an additional dataset of 64 primary T-ALL patient samples, we could correlate the expression of lncRNAs lnc-PGBD5-2, lnc-FAM120AOS-1, lnc-c2orf55-1 and LUNAR1 with the Notch1 positively regulated gene set Vilimas_NOTCH1_targets_up52 by GSEA (Figure 3E).

Overall, these independent experiments confirm the Notch dependent regulation of the selected lncRNAs, thereby validating the RNA-seq data from the GSI treated CUTLL1 T-ALL cell line and the Notch perturbed normal human thymocytes.

Genome-wide analysis reveals direct Notch1 binding to selected lncRNAs

To further validate the direct regulation of selected lncRNAs by Notch, publicly available ChIP sequencing (ChIP-seq) data from the CUTLL1 cell line were analyzed for Notch1 binding at specific loci.34 From the Notch-driven annotated lncRNAs that overlapped between normal and malignant thymocytes (Figure 2D), 13 out of the 27 lncRNAs were bound by ICN1 (Online Supplementary Table S12) as illustrated for lnc-UBXN4-1 and lnc-PLEKHB2-1 (Figure 4). Remarkably, 12 out of the 13 lncRNAs with a Notch1 binding peak also show Brd4 and Med1 binding. Notably, from the putative Notch regulated lncRNAs that showed correlated expression with DTX1 in CD34 and CD4CD8 normal thymocytes, 6 out of 9 (Figure 4 and Online Supplementary Figure S5) showed binding of ICN1, suggesting that the majority are a direct Notch target.

Figure 4.Notch1 ChIP-seq reveals direct binding of Notch1 to a subset of regulated lncRNAs. Representation of ChIP-sequencing tracks³⁴ for Notch1, Brd4, Med1 and H3K27ac and representative RNA-sequencing tracks for CUTLL1 DMSO/GSI treatment and OP9-GFP/DLL1 of lnc-UBXN4-1 and lnc-PLEKHB2-1.

In addition, we evaluated the presence of H3K27 acetylation (H3K27ac) ChIP-seq signal at these lncRNA loci, a histone mark indicative for putative enhancer regions. For 18 out of the 27 selected lncRNAs, H3K27ac ChIP-seq signal was present in close proximity of the promoter region (Online Supplementary Table S12), suggesting the presence of enhancer sequences. Moreover, this public ChIP-seq data also showed LUNAR1 to be directly bound by ICN1, Brd4, Med1 and H3K27ac (Figure 5). We also evaluated ICN1 binding at annotated lncRNA loci up-regulated upon GSI treatment of CUTLL1 T-ALL cells and down-regulated in CD34 T-cell progenitors upon DLL1 exposure in the OP9 in vitro co-culture system. Only 4 out of the 18 overlapping annotated lncRNAs that are negatively regulated by Notch (Online Supplementary Figure S2A) showed direct binding by ICN1. The same analysis was performed on the set of 7 unannotated lncRNAs repressed by Notch1 signaling (Online Supplementary Figure S2B). Only 3 out of these 7 lncRNA loci showed ICN1 binding in the proximity of its promoter region. Given the established predominant role of Notch1 as a transcriptional activator, lncRNAs that are negatively affected by Notch1 signaling may actually be indirect targets.

Figure 5.Notch1 ChIP-seq reveals direct binding of Notch1 to LUNAR1. LUNAR1⁵¹ (lnc-IGF1R-1) was identified amongst the top differentially expressed novel, unannotated lncRNAs in both GSI-treated CUTLL1 cells and CD34⁺ thymic progenitor cells exposed to DLL1-triggered Notch signaling. Publically available ChIP-seq tracks³⁴ for ICN1, BRD4, MED1 and H3K27ac as well as representative in house generated RNA-seq data tracks for CUTLL1 DMSO/GSI treated cells and OP9-GFP/DLL1 are shown at the LUNAR1 locus.

Attributing functional annotation to Notch-regulated lncRNAs through guilt-by-association analysis

As described above, we defined a core set of 27 Notch-driven and previously annotated lncRNAs by considering only those differentially expressed and positively regulated by Notch signaling in the GSI perturbation model in CUTLL1 cells and the in vitro OP9-DLL1 co-culture system. As a next step, we aimed to assign potential functionalities to each of these candidates. To this end, we used the so-called ‘guilt-by-association’ approach (see also Online Supplementary Methods). As previously mentioned, 15 out of these 27 lncRNAs (together with LUNAR1) were detectable by a probe on our custom designed micro-array platform. In a first step, we calculated the Spearman correlation coefficients between the lncRNAs-of-interest and all protein coding genes using the expression data of a primary T-ALL cohort of 64 patients from which we profiled all samples on the custom designed Agilent array.48 These correlations were subsequently used as an input for a GSEA pre-ranked analysis. Next, the output of this GSEA analysis was further refined into functional clusters of enriched gene sets using the Cytoscape plug-in enrichment mapping. This analysis yielded markedly different functional clustering patterns for each of the 16 lncRNAs analyzed (including LUNAR1). Important putative functionalities were represented in each of the networks as exemplified by TCR-signaling and phospholipid metabolism for lnc-PLEKHB2-1, DNA replication and DNA repair for lnc-UBXN4-1 and splicing and cell cycle regulation for LUNAR1 (Figure 6A–C and Online Supplementary Figure S6A–M).

Figure 6.Attributing functional annotation to selected, annotated lncRNAs through guilt-by-association analysis. Enrichment maps of gene sets correlated with the expression of (A) lnc-PLEKHB2-1, (B) lnc-UBXN4-1 and (C) LUNAR1. Red nodes represent the positively correlated gene sets to the lncRNA of interest, blue nodes the negatively correlated gene sets. The size of the nodes depicts the size of the gene sets. Nodes that are clustered represent gene sets with the same or similar functional indication.

Discussion

Non-coding RNAs are emerging as important players in normal development and disease, including cancer. In previous studies, we investigated the role of miRNAs in T-cell acute lymphoblastic leukemia (T-ALL), thereby identifying a small set of miRNAs that is responsible for the cooperative suppression of several tumor suppressor genes.46 These miRNAs produced overlapping and co-operative effects with several bona fide T-ALL tumor suppressor genes including IKZF1, PTEN, BIM, PHF6, NF1 and FBXW7, and more recently this network was expanded further with PHF6.53 In order to provide some insight into the genetic components driving long non-coding RNAs in T-ALL formation, we performed an integrated analysis of lncRNA profiling data sets from GSI inhibited Notch-driven T-ALL cell lines and Notch-stimulated immature normal human thymocytes using the OP9 co-culture system, together with publicly available genome-wide data on Notch1 binding and specific chromatin marks. In addition, we correlated the expression of Notch-dependent lncRNAs with the Notch-dependent stages of normal thymocytes during T-cell differentiation. Overall, our work establishes a novel lncRNA network that acts downstream of Notch during normal and malignant thymocyte development.

Our study provides a number of fundamental new insights into Notch-dependent regulation of lncRNAs in TALL and normal developing thymocytes. First, we unambiguously demonstrate that a significant number of lncRNAs are directly regulated by Notch signaling activity. Through RNA-sequencing, we identified 40 lncRNAs that are positively regulated by Notch in both normal and malignant T lymphocytes (annotated as well as previously unannotated lncRNAs), supporting an important role for these lncRNAs in Notch-regulated T-cell biology. This could be related to various functions of Notch signaling, including T-cell lineage specification and commitment, proliferation and differentiation. Importantly, the recently identified lncRNA LUNAR151 was present amongst the most robustly Notch regulated long non-coding RNAs in our data sets. LUNAR1 was shown to be required for efficient T-ALL growth as a consequence of its role in enhancing IGF1R mRNA expression to sustain IGF1 signaling.51 As a prelude to assigning functional annotation to the newly assigned Notch-regulated lncRNAs in this study, we applied the so-called ‘guilt-by-association’ approach in which functions are predicted based upon correlation with known protein coding genes and subsequent gene set enrichment analysis. For the selected lncRNAs, various functions were predicted, several of which are linked to T-cell biology or processes that are perturbed in cancer. This marks these lncRNAs as prime targets for further functional studies in order to unravel their mode of action and assess to what extent they might serve as future therapeutic targets for treatment of T-ALL.

Not all of the 40 overlapping lncRNAs displayed ICN1 binding, as is evident from the publicly available ChIP-seq data.34 This may relate to the complexity of the chromosomal 3D-structures that are generated when lncRNAs act as cis-regulatory elements, as well as to the sensitivity of the Notch1 ChIP procedure. However, for the previously annotated lncRNAs directly bound by ICN1, all but one displayed Brd4 and Med1 binding. Those lncRNAs that are characterized by Brd4 and Med1 ChIP-seq signal are also characterized by the presence of H3K27ac ChIP-seq signal, which could be indicative of an enhancer activity of these loci.

A second aspect of our study involved the identification of novel, previously unannotated lncRNAs. Indeed, previous studies have shown that lncRNAs are often shown to be very restricted, but with biologically high relevant expression.4540 This includes expression during very specific time points during development and/or differentiation, as well as restriction to very specific cell subsets. Typically, these lncRNAs are expressed at significant levels in these cells whereas in other cell types their expression is very low or absent. Here, we identified novel lncRNAs in the CUTLL1 cell line and in the OP9-DLL1 co-culture system. Interestingly, 61 of these lncRNAs were present in T-ALL cells only, suggesting that their ectopic expression could be restricted to the malignant context. Likewise, lncRNAs only present in thymocytes may be implicated in differentiation of normal T cells which is disrupted in T-ALL cells, or may reflect differential Notch3 activity as the DLL1 ligand, to which CD34 progenitor T cells are exposed in the OP9-DLL1 co-culture system, can activate both Notch1 and Notch3 (Waegemans E, Van de Walle I and Taghon T, unpublished data on preferential Notch receptor-ligand interactions in human, 2011) while both receptors are implicated in modulating human T-cell development.21 Both subsets of lncRNAs may, therefore, serve as novel therapeutic targets for T-ALL treatment.

It is evident that our work, as well as the recent paper by Aifantis and colleagues,51 strongly favors an important role for lncRNAs in normal T-cell development and T-ALL oncogenesis. Moreover, we show that the Notch1 transcription factor directly controls the transcription of many of these long non-coding RNAs. Therefore, one can predict that other oncogenic transcription factors and drivers in T-ALL, such as TAL1, TLX1/TLX3, LMO1/2 and HOXA genes, as well as other transcriptional regulators of normal T-cell development, will also perform similar transcriptional control. Therefore, specific T-ALL subgroups and more distinct subsets of normal immature developing T cells need to be analyzed in human for detection of all lncRNAs. We predict that this will further dramatically expand the lncRNA landscape for T-ALL and thymocyte maturation, and thus provide an important regulatory framework for understanding some of the unique features that control human T-cell biology.27 Finally, given the central role of oncogenic Notch1 activation in most if not all T-ALLs, and the current limitations of targeted therapy, further exploration of the new therapeutic opportunities offered by these lncRNAs in the context of their specific functionality is strongly recommended.

Footnotes

• ↵* KD and AW contributed equally to this work.
• ↵** TT and FS contributed equally to this work.
• The online version of this article has a Supplementary Appendix.
• Funding The authors would like to thank the following funding agencies: the Fund for Scientific Research Flanders (‘FWO Vlaanderen’ research projects G.0202.09, G.0869.10N, 3G055013N, 3G056413N to FS; 3GA00113N, 3G065614, G.0C47.13N to PVV and G0B2913N, G037514N, 3G002711 to TT; PhD grant to AW; postdoctoral grants to IvdW, TT, PM, PVV, and PR; BP is a senior clinical investigator), IWT Vlaanderen (PhD grant to KD), the Belgian Foundation against Cancer (FS, JVds (SCIE 2010-177)); Ghent University (GOA grant 01G01910 to FS), the Cancer Plan from the Federal Public Service of Health (FS), the Children Cancer Fund Ghent (FS) and the Belgian Program of Interuniversity Poles of Attraction (IUAP P7/03 and P7/07). We also would like to thank Aline Eggermont for excellent technical assistance.
• Authorship and Disclosures Information on authorship, contributions, and financial & other disclosures was provided by the authors and is available with the online version of this article at www.haematologica.org.

• Received August 13, 2014.
• Accepted October 22, 2014.