Deletion of chromosome 6q is a well-recognized abnormality found in poor-prognosis T-cell acute lymphoblastic leukemia (T-ALL). Using integrated genomic approaches, we identified two candidate haploinsufficient genes contiguous at 6q14, SYNCRIP (encoding hnRNP-Q) and SNHG5 (that hosts snoRNAs), both involved in regulating RNA maturation and translation. Combined silencing of both genes, but not of either gene alone, accelerated leukemogeneis in a Tal1/Lmo1/Notch1-driven mouse model, demonstrating the tumor-suppressive nature of the two-gene region. Proteomic and translational profiling of cells in which we engineered a short 6q deletion by CRISPR/Cas9 genome editing indicated decreased ribosome and mitochondrial activities, suggesting that the resulting metabolic changes may regulate tumor progression. Indeed, xenograft experiments showed an increased leukemia-initiating cell activity of primary human leukemic cells upon coextinction of SYNCRIP and SNHG5. Our findings not only elucidate the nature of 6q deletion but also highlight the role of ribosomes and mitochondria in T-ALL tumor progression.
The oncogenic role of 6q deletion in T-ALL has remained elusive since this chromosomal abnormality was first identified more than 40 years ago. We combined genomic analysis and functional models to show that the codeletion of two contiguous genes at 6q14 enhances malignancy through deregulation of a ribosome–mitochondria axis, suggesting the potential for therapeutic intervention.
This article is highlighted in the In This Issue feature, p. 1494
The identification of somatic genome aberrations in cancer cells has been a highly fruitful strategy to identify cancer genes, especially in hematopoietic malignancies. With the availability of massive sequencing tools, it is now relatively easy to identify cancer genes from translocations, point mutations, or short insertion/deletions (1). However, the molecular targets of large chromosomal losses or gains can remain difficult to elucidate, and these types of abnormalities are often associated with a worse prognosis. This is exemplified by the considerable efforts that were developed over several decades to understand and model oncogenesis for abnormalities such as monosomy 7 in myeloid malignancies or deletion 6q in T-cell acute lymphoblastic leukemia (T-ALL). T-ALL is related to leukemic transformation of T-cell progenitors (2, 3). In addition to a founding rearrangement that deregulates transcription factor genes and determines the oncogenic subtype, such as TAL1/2, LMO1/2, TLX1/3, and HOXA, the T-ALL genome harbors a myriad of additional mutations, deletions, and duplications that collectively lead to overt leukemia (3–10). These have a particular impact on the cell cycle (CDKN2A/p16/ARF deletion), the Notch pathway (NOTCH1 and FBXW7 mutations), and the JAK/STAT and PI3K/AKT pathways (IL7R and PTEN mutations). Deletion of the long arm of chromosome 6 (del6q) is a frequent karyotypic abnormality in T-cell ALL and lymphoblastic lymphoma, where it has been associated with an unfavorable prognosis (11–13). Although first reported as a recurrent structural abnormality in lymphoblastic leukemia in 1976, the underlying molecular targets of this chromosomal event remain elusive (14).
In our study, we used a large range of integrated genomic and functional analyses and identified two genes simultaneously inactivated through del6q, the combined haploinsufficiency of which accelerates T-ALL progression in vivo. By using a genome-editing CRISPR/Cas9 approach, we precisely engineered a short del6q in human cells that enabled us to examine the proteomic and translatome profile of the deleted cells. We found that the haploinsufficiency of these two genes, one encoding a ribonucleoprotein and the other hosting noncoding small nucleolar RNA (snoRNA), deregulates cellular metabolism and ultimately affects the leukemia-initiating cell (LIC) activity of human T-ALL cells through modulation of ribosomal functions.
Del6q Is a Late Chromosomal Event in TAL1 Oncogene–Related T-ALL
To identify tumor suppressor genes in this deletion, we initially screened deletion 6q by array comparative genomic hybridization (aCGH) in a cohort of 78 primary T-ALL cases that were previously characterized by large-scale expression profiling (ref. 4; see flow chart of the study in Supplementary Fig. S1). Del6q was associated with the T-ALL subtype that is characterized by aberrant expression of the TAL1 oncogene (P = 0.032), suggesting oncogenic cooperation (Fig. 1A). We focused on this subtype and analyzed a total of 107 TAL1-related (TAL-R) cases from three patient cohorts (4, 5, 15), in which del6q was detected in 34 cases (32%). Del6q could be subclonal, and distinct del6q could be observed in the same sample, or in the diagnosis and relapse leukemia samples from the same patient, or in primary and patient-derived xenografts (PDX; Supplementary Fig. S2; Fig. 1B). Additional, co-occurring mutations in NOTCH1/FBXW7 or PTEN genes were often found, as previously reported in TAL-R cases, suggesting multistage oncogenesis (Fig. 1C; refs. 3, 7). Collectively, these data indicate that del6q is strongly associated with the TAL-R subtype in which it occurs as a late-stage chromosomal event associated with leukemia progression.
Two Contiguous Genes, SYNCRIP and SNHG5, Are Candidate Haploinsufficient Tumor Suppressors at 6q14
We then aimed to identify the molecular targets of chromosome 6q deletion. Del6q was interstitial in all cases, heterogeneous in size, with a common deleted region (CDR) of 4.9 Mb at 6q14 that included 18 genes (Fig. 2A; Supplementary Table S1). Located just beside the telomeric border, the MYB oncogene was never included in the deletion, suggesting that loss of one MYB copy would limit T-ALL cell fitness (Fig. 2A; the alternative hypothesis of a MYB positional activation driven by del6q was ruled out by allele expression analysis; see Supplementary Fig. S3). We then focused on the CDR using custom 6q aCGH and high-throughput sequencing tools (Supplementary Figs. S1 and S4). Although no deleterious point mutation or short insertion/deletions were discovered, focal somatic deletions of 52 kb and 38 kb, respectively, were found in two T-ALL cases without large del6q (Fig. 2B; Supplementary Table S2). Critically, only two genes, SYNCRIP and SNHG5, were involved in both short deletions (Fig. 2B and C).
In parallel, we analyzed gene-expression data in the T-ALL cases (including n = 107 TAL-R with n = 34 cases harboring a del6q) and found an overall decreased expression of 6q13–q22 genes consistent with a gene dosage effect (Fig. 2D). The downregulated transcripts in T-ALL cases with del6q included five genes from the CDR, and SYNCRIP and SNHG5 in particular (Fig. 2D; Supplementary Fig. S5A). We also analyzed healthy human thymic subset data and found that both genes were differentially regulated during normal T-cell development, with a peak at the immature single positive (ISP) stage and downregulation after β-selection (Fig. 3A). In addition, chromatin immunoprecipitation sequencing data showed that TAL1 can bind regulatory regions of both the SYNCRIP and SNHG5 genes, suggesting transcriptional regulation (Supplementary Fig. S5B). We sorted double-negative (DN3 and DN4) thymocytes from young pSil-TSCL (Tal1tg) and Lck-LMO1 (Lmo1tg) transgenic mice (Tal1tgLmo1tg) and did indeed find a significant increase of Syncrip and Snhg5 expression compared with age-matched wild-type littermates (Fig. 3B and C), suggesting that del6q counteracts TAL1-related sustained expression of these genes in the human TAL-R subtype, and thereby explaining why the deletion is found in that subtype.
SYNCRIP (also known as hnRNP-Q or NSAP1) encodes the heterogeneous nuclear ribonucleoprotein hnRNP-Q, whereas SNHG5 is a non–protein-coding gene hosting two C/D-box small nucleolar RNAs in its introns (U50A and U50B snoRNAs, also known as SNORD50A and SNORD50B). Both genes have known functions in the regulation of mRNA processing (alternative splicing, editing, transport, and degradation) and translation (16–19). Specifically, U50 snoRNAs regulate ribosomal biogenesis through site-specific recognition of preribosomal RNAs to mediate 2′-O-methylation (20). The expression of the two genes was correlated with pathways involved in mRNA processing and ribosome activities in normal human thymic populations (Fig. 3D and E). In both 6q short deletions from T-ALL patients, the ATG and most exons of SYNCRIP were lost, as were the snoRNA sequences in the introns of SNHG5, suggesting inactivation of both genes (Fig. 2C). Collectively, our integrated genomic and expression analysis of del6q in T-ALL converged on two candidate tumor suppressor genes differentially regulated during T-cell differentiation and whose haploinsufficiency may play a role at a late stage of TAL1-associated T-cell leukemia.
Syncrip–Snhg5 Cosilencing Accelerates Tumor Development in a Tal1/Lmo1/Notch1-Induced T-ALL Mouse Model
To functionally investigate the effects of single or combined SYNCRIP and SNHG5 haploinsufficiency, we developed a mouse model of tumor progression in the context of TAL1-related leukemogenesis (Fig. 4A). To mimic human TAL-R primary cases (Fig. 1C), we collected bone marrow progenitor cells from young Tal1tgLmo1tg mice (known to develop an overt T-ALL by the age of 6 months; ref. 21), transduced them with a Notch1 construct (22, 23), and transplanted them into RAG−/−γc−/− mice. This resulted in a fully penetrant lethal T-ALL with a median latency of 60 days (Fig. 4B), in which we tested whether additional silencing of murine Syncrip and/or Snhg5 by shRNAs might accelerate further the onset of overt leukemia. Notably, we chose an shRNA approach rather than a genetic inactivation of the Syncrip and Snhg5 mouse locus, considering that the snoRNAs from Snhg5 are duplicated several times in the mouse genome. As shown in Fig. 4B, recipients of cells transduced with individual Syncrip or Snhg5 shRNA developed leukemia with a comparable median latency to the control (59, 60, and 60 days, respectively). Strikingly, mice transplanted with progenitor cells silenced for both Syncrip and Snhg5 succumbed to T-ALL significantly earlier (median latency of 50 days; Global Cox likelihood ratio test, P = 0.00013; Fig. 4B). Therefore, combined Syncrip and Snhg5 silencing, but not individual silencing, significantly accelerates Tal1/Lmo1/Notch1-induced T-cell leukemogenesis in mice. Collectively, these data demonstrate that Syncrip and Snhg5 function as a tumor suppressor region, the silencing of which drives tumor acceleration in vivo.
6q14 Microdeletion Engineered by Genome Editing Affects Ribosomal Functions, Translation Programs, and Mitochondrial Respiration in Human T-ALL Cells
To investigate the mechanisms of T-ALL tumor progression induced by SYNCRIP and SNHG5 haploinsufficiency, we used the CRISPR/Cas9 genome-editing approach to engineer a 6q14 microdeletion in a human T-ALL cell line harboring a prototypical TAL1 rearrangement (TAL1d, i.e., SIL-TAL microdeletion) but no del6q, and then compared the resulting deleted and nondeleted clones. Briefly, we engineered the 38-kb deletion precisely, as seen in patient TL97, by using two sets of Cas9 target-flanking sequences simultaneously (Fig. 5A) and generated several SYNCRIP–SNHG5 microdeleted CCRF-CEM clones (Fig. 5B). Immunoblot showed reduced protein levels of hnRNP-Q (SYNCRIP gene product) in the Del6q/ΔSYNCRIP/ΔSNHG5 clones (clones A6 and C4) compared with the nondeleted clone (C2) and isogenic cell line (Fig. 5C). Considering the known function of hnRNP-Q and snoRNAs in regulating mRNA processing and translation (16–20), we undertook a comprehensive proteomic approach to examine biological functions that might be affected. Mass spectrometry–based label-free quantitative proteomic analysis identified a list of 253 proteins that were expressed differentially between the deleted versus nondeleted cells out of a total of 2,352 significantly quantified proteins (Fig. 5D; Supplementary Table S3). As expected, the SYNCRIP/hnRNP-Q protein level was lower in the deleted clones (Fig. 5D). Strikingly, pathway enrichment analysis found a global downregulation of a number of pathways, including protein metabolic process, translational initiation, protein/RNA complex assembly, and mRNA processing (Fig. 5E and F). In the light of these results, we examined the ribosome population fractions in deleted and nondeleted cells by profiling them on sucrose gradient. Although the overall profile was unchanged, indicating that translation remained active, we found that the 80S fraction was higher in the deleted cells, indicating a change in the translation machinery activity (Fig. 5G). We also quantified in an unbiased manner the 2′-O-methylation occupancy on N = 106 sites using a highly sensitive RiboMethSeq approach (24). Strikingly, only the two known SNORD50A target sites (28S-Cm2848 and 28S-Cm2863) were significantly hypomethylated in the deleted clones when compared with nondeleted and isogenic lines, whereas the 104 other sites remained unaffected (Fig. 5H), showing the specificity of the methylation change upon SNHG5 haploinsufficiency. To evaluate whether altered 2′-O-methylation and ribosome assembly might affect the global translation rate, we evaluated global protein synthesis by quantifying the L-azidohomoalanine incorporation into nascent proteins (Click-iT AHA assay), but did not find consistent changes in basal culture condition (not shown). However, we reasoned that hypoxic condition, mimicking the hematopoietic niche, might stress the system and reveal a differential response. Indeed, the protein neosynthesis was significantly lower in the deleted clones in hypoxic condition (Fig. 5I). Consistent results were also observed in hypoxia in independent experiments of 35S-labeled methionine incorporation into nascent peptides (Fig. 5J). To analyze whether ribosome changes can affect biological processes through differential translation efficiency (TE) of specific gene transcripts, we analyzed the actively translated mRNAs in deleted and nondeleted cells by sequencing the polysome-bound and total cytoplasmic mRNA fractions (translatome). Using dedicated translatome profiling tools (25, 26), we found a differential TE of transcripts from 475 genes, these being mainly linked to RNA binding (up) and mitochondrial function (down) using Gene Ontology (GO) analysis (Fig. 5K and L; Supplementary Table S4). Specifically, the differential TE of mitochondria transcripts suggested a potential downstream mechanism of leukemia cell regulation by the deletion (Fig. 5L). Notably, in normal thymic cells, the expression of SYNCRIP and SNHG5 genes correlates positively with oxidative phosphorylation pathway genes, in agreement with a functional link (Fig. 3D and E). We thus performed a direct analysis of mitochondrial respiration using real-time measurements of oxygen consumption rate (OCR) and did indeed find reduced basal and maximal mitochondrial respiration in the deleted clones compared with nondeleted clones (Fig. 5M), whereas glycolysis was increased (Fig. 5N). Collectively, these results point to deregulated ribosome activity arising from the 6q microdeletion in human T-ALL cells that results in metabolic changes especially in mitochondrial respiration.
SYNCRIP–SNHG5 Cosilencing Increases the In Vivo LIC Activity of Human T-ALL Cells along with a Deregulated Ribosomal and Mitochondrial Signature
Low metabolism including differential oxidative phosphorylation is thought to regulate normal and leukemic stem cell fate and functions (27–33). Given the known functions of SYNCRIP and SNHG5 in mRNA processing and translation (16–20), along with the consistent changes we observed when these genes were codeleted (Fig. 5E–N), we hypothesized that the del6q might increase malignancy by changing the metabolic state in the T-ALL cell population (Fig. 6A). We therefore aimed to directly test whether the SYNCRIP–SNHG5 deletion increases the LIC activity of human T-ALL cells. We took advantage of our engraftment model of primary T-ALL in immunodeficient NOD/SCID/IL2Rγ-null (NSG) mice (15) and used SYNCRIP–SNHG5 shRNA transductions to cosilence both human genes (Fig. 6B). To provide a faithful model of the deletion as a late-stage chromosomal event in the context of TAL1-related leukemogenesis, primary diagnosis T-ALL blast cells harboring a prototypical TAL1d rearrangement, but no del6q, were lentivirally silenced, and the sorted cells were injected into NSG mice as shown in Fig. 6B. Using this setting that includes human SYNCRIP and SNHG5 shRNAs, we found that silenced cells propagated leukemia significantly more efficiently than control cells in both parallel and competitive engraftment experiments, suggesting increased malignancy (Fig. 6C; Supplementary Fig. S6), consistent with the shorter leukemia latency seen in our mouse model (Fig. 4B). Moreover, the accelerated engraftment of the SYNCRIP–SHNG5-silenced leukemic cells persisted in dilution experiments, indicating greater LIC activity (Fig. 6D).
We then characterized the human leukemic cells that expanded in the xenografts. We measured the 2′-O-methylation level of the 28S rRNA at the Cm2848 site, known to be a target of SNORD50A (34), and found that it was lower in silenced than in control T-ALL cells that emerged in NSG mice (Fig. 6E). Comparing gene-expression data in these cells (Supplementary Table S5), relative SYNCRIP and SNHG5 levels were consistent with haploinsufficiency, suggesting in vivo selection of the optimal level of gene silencing and reinforcing the interest of these models. Moreover, the SYNCRIP–SNHG5-silenced T-ALL cells also exhibited downregulation of many genes related to ribosome, RNA degradation, and oxidative phosphorylation (Fig. 6F and G; Supplementary Table S5), i.e., pathways that are relevant not only to the known functions of both genes, i.e., regulating mRNA processing and ribosome biogenesis (16–20), but also to the translatome and proteomic changes of ΔSYNCRIP/ΔSNHG5 clones (Fig. 5E–N). To directly assess the effects of reduced oxidative phosphorylation in leukemia progression, we exposed PDX-derived TAL1d, non-del6q, human T-ALL cells to tigecycline, an antibiotic that inhibits mitochondrial protein translation (35), and subsequently performed ex vivo limiting dilution and immunophenotype experiments (Fig. 7A). We found that tigecycline conferred increased clonogenicity to T-ALL cells (Fig. 7B); in addition, tigecycline exposure partially reproduced PDX cell dedifferentiation that was seen after SYNCRIP–SNHG5 cosilencing using shRNA (Fig. 7C).
Finally, to confirm the relevance of the SYNCRIP–SNHG5-focused codeletion with the usually larger 6q deletions seen in patients, we analyzed the expression of the “SYNCRIP/SNHG5-silenced xenograft T-ALL” signature defined in silenced xenograft (Fig. 6F) in the cohort of TAL1-related patients. Strikingly, the downregulated signature was also enriched in primary T-ALL cases with large del6q versus nondeleted cases (Fig. 7D).
The oncogenic meaning of del6q in T-ALL has been a long-standing unresolved question since the original description of this structural chromosome rearrangement in leukemia (11–14). In this study, we addressed del6q by investigating large cohorts of highly annotated primary T-ALL samples, which allowed us to focus on the homogeneous TAL1-related subtype, and by using original in vitro and in vivo models. Integrated global genomic and expression analyses identified two contiguous downregulated genes at 6q14, SNHG5 and SYNCRIP. Both genes were found induced upon TAL1 overexpression in thymocytes, which provided an explanation as to why del6q is found specifically in the TAL-R subtype in the patients, i.e., to counteract abnormal (considering T-ALL differentiation stage) SNHG5- and SYNCRIP-sustained expression after β-selection (36, 37). Regarding del6q, short or large deletions, but no single point mutation of these two genes, were found, in line with the idea that the simultaneous inactivation by a single chromosomal deletion is much more probable than two or more concomitant mutations leading to combined haploinsufficiency of SNHG5 and SYNCRIP. Moreover, our mapping and expression results strongly suggest that the deletion size must be sufficiently large to include the two-gene tumor suppressor region, but interstitial rather than complete to prevent the loss of MYB. Indeed, MYB is a known oncogene of T-ALL that can activate TAL1, and its silencing has been shown to be deleterious for T-ALL cells (10, 38, 39). We then demonstrated that the codeletion was not a passenger lesion but truly a driver of tumor progression using a multievent T-ALL in vivo model—i.e., acceleration of Tal1tgLmo1tgNotch1IC mouse leukemia—in line with the human TAL-R subtype in which del6q occurs as a late-stage chromosomal event. Importantly, the acceleration was seen only when both genes were codeleted, demonstrating the tumor suppressor nature of the global SYNCRIP–SNHG5 region, but not of either gene alone.
To mechanistically investigate the role of the combined loss of the two genes, we made use of the CRISPR/Cas9 technology to engineer a precise 38 kb deletion in a T-ALL cell line as seen in patient cells. Proteomic analyses of del6q-engineered clones showed a decreased ribosomal signature, consistent with the known functions of these two genes, which both regulate mRNA processing and ribosome biogenesis. Such a downregulation of ribosomal functions may seem counterintuitive for the fitness of cancer cells. However, on the basis of recent reports that ribosomal functions can regulate normal and leukemic stem cells (27–30), we hypothesized that these changes might induce metabolic changes and confer an oncogenic gain at late stage of T-ALL progression. Interestingly, our data showing differential translation of mitochondria gene transcripts, along with reduced mitochondrial respiration, suggested direct oncogenic deregulation of the T-ALL LIC (31–33, 40). Because the nature of the LICs is not clearly identified in T-ALL, precluding consistent purification and characterization (41, 42), we performed a functional in vivo approach, i.e., xenograft of leukemia cells in immunodeficient mice. This enabled us to show that SYNCRIP–SNHG5 cosilencing did indeed confer greater LIC ability to primary TAL-R leukemia cells. In addition, ex vivo exposure to tigecycline, an antibiotic that inhibits mitochondrial protein translation, conferred to non-del6q primary human T-ALL cells greater properties associated with LIC activity and gain of malignancy, suggesting a direct driving role of the reduced mitochondrial respiration in T-cell oncogenesis. This differs from the inhibitory effect of tigecycline on myeloid leukemia (acute myeloid leukemia and chronic myeloid leukemia) LICs, which relies on enhanced basal oxygen consumption (43, 44). By contrast, T-ALL LICs would fit the classic figure of malignant cells relying more on glycolysis than on oxidative phosphorylation. Differential mitochondrial activity may also directly modulate T-ALL LIC self-renewal and cell fate, as was reported in normal HSCs and thymocytes (33, 45). Notably, although our studies clearly establish the existence of a SYNCRIP/SNHG5 → ribosome → mitochondria → LIC axis in T-ALL, we cannot exclude that SYNCRIP/SNHG5-dependent ribosome modifications also impinge on the translation of other key regulators of cellular self-renewal or proliferation. Importantly, we retrieved reduced ribosome and oxidative phosphorylation signatures, not only in the engineered deleted cells, but also in the SYNCRIP–SNHG5-silenced engrafted T-ALL cells, as well as in del6q cases from our TAL-R patient cohorts. Collectively, the data also confirmed the general relevance of the short SYNCRIP–SNHG5 deletion to the larger 6q deletions seen in most patients.
In conclusion, our study clarifies the role played by 6q deletion in T-ALL oncogenesis and identifies SYNCRIP/hnRNP-Q and snoRNAs as tumor suppressors in this deletion. Together with previous evidence of mutations in ribosomal genes in hematopoietic malignancies and other cancers (8, 46, 47), our findings provide new insights for a critical driver role of qualitative ribosome and mitochondria deregulation, in the context of multistage oncogenesis, consistent with the increasingly recognized regulatory functions of subtle metabolic changes on normal and leukemic stem cells (28–30, 40, 47, 48), and opening potential avenues for therapeutic targeting. For instance, therapies that change ribosomal activity or oxidative respiration may antagonize LICs in del6q T-ALL. In addition, ribosomal stress induced by del6q suggests that therapeutic intervention, such as synthetic lethality, may be used to target leukemic cells, thereby improving clinical outcomes in patients.
Patients and Collection of Biological Material
Three annotated cohorts of patients with T-ALL were analyzed in this study. The CIT and ALEK cohort included children and adults, whereas the Rotterdam cohort included only children (4, 5, 15). Informed written consent was obtained from the patients or relatives in all cases, and the study was conducted in accordance with the Declaration of Helsinki. The study was approved by the review board of the Institut Universitaire d'Hématologie, Hôpital Saint-Louis, Paris, France.
Genome-Wide DNA Array Copy-Number Analysis
A total of 107 T-ALL cases from the TAL1-related subtype were tested using 105K, 180K, 244K, and/or custom 8 × 15K Human Genome CGH Microarrays arrays (Agilent Technologies), allowing precise mapping of a CDR at 6q14. Hybridization was performed in accordance with the manufacturer's recommendations, and copy numbers were analyzed using VAMP tools (Curie Institute) or Genomic Workbench software with the help of the ADM-2 algorithm (Agilent Technologies). All data were visually inspected and cured manually by at least two investigators (D. Avran and J. Soulier).
Design of the 6q14 Custom Array
The custom oligonucleotide-based microarray contained 15,689 60-mer probes (including 1,391 manufacturer control probes) spanning both coding and noncoding genomic sequences of chromosome 6 (Supplementary Fig. S4A). The multistep custom design included 14,296 probes distributed among several groups with various probe densities. Group I comprised 2,574 probes spread along chromosome 6p (∼22-kb average probe spacing); group II comprised 3,370 probes spread along chromosome 6q (∼32-kb average probe spacing). Several intervals in chromosome 6q were made denser.
Next-Generation Sequencing at 6q14
We designed a custom SureSelect Agilent capture in solution to enrich genomic sequences from the CDR and flanking regions at 6q14 (Supplementary Fig. S4B). The Agilent eArray (https://earray.chem.agilent.com/earray/) was used to design and assess coverage across the target genomic regions of bait libraries. Different bait groups were established in accordance with various parameters (tiling, number of replicates, and orphan probes). Regions with repeats were eliminated with the “repeat masker” option. The final customized target capture contained 57,670 baits of 120 bp, corresponding to 5,676,317 bp. DNAs were sonicated to obtain 150- to 200-bp fragments using a Covaris S220 Focused ultrasonicator instrument (Covaris Inc.). 6q14 custom capture was performed using the SureSelect Target Enrichment System for Illumina Paired-End Sequencing Library protocol (version 1.2, May 2011) in accordance with the manufacturer's recommendations. Captured samples were sequenced using a HiSeq 2000 (Illumina) operated in the paired-end 2 × 100 bp mode. Somatic variants were identified by selecting the variants found in tumor reads that were absent in the nontumor reads. All somatic variants were visually inspected and manually cured by at least two investigators (S. Quentin and J. Soulier).
Large-Scale Gene-Expression Analysis
Microarray data files from the three T-ALL cohorts are available from ArrayExpress under accession number E-MEXP-313 (CIT cohort), E-MTAB-604 (ALEK cohort), and from Gene Expression Omnibus (GEO) accession number GSE26713 (Rotterdam cohort). Affymetrix GeneChip probe set summarization, background correction, and quantile normalization were performed using the RMA routine, as implemented in the “R” Affymetrix package. Differential gene expression arising from chromosomal gene localization was analyzed using MACAT (microarray chromosomal analysis tool) and the associated R Bioconductor packages (www.bioconductor.org) using the following parameters: permute = labels, nperms = 750, kernel = KNN, and step.with = 100,000. Biological gene-pathway changes were scored using the Gene Set Enrichment Analysis algorithm (GSEA; broadinstitute.org). Human T-ALL samples from xenografts were hybridized using the GeneChip Human Gene 1.0 ST (Affymetrix); gene-expression profiling and splice analysis were performed using the Easana and Fast DB tools (Genosplice Technologies; http://www.genosplice.com/).
Analysis of Syncrip and Snhg5 Expression Levels in Murine Thymocyte Subsets by Real-Time Quantitative PCR (RT-qPCR)
Murine thymocytes were extracted from thymi from 3- to 4-week-old Tal1tgLmo1tg and age-matched wild-type littermates. Briefly, CD3-, CD4-, CD8-, and Ter119-positive cells were isolated by magnetic activated cell sorting. DN3 and DN4 thymocytes were purified by FACS-mediated cell sorting from the enriched positive cells using murine antibodies against lineage markers CD3, CD4, CD8, CD25, and CD44. Total RNA from DN3 and DN4 subsets was extracted using the Direct-Zol RNA Mini Prep Plus kit (Zymo/Ozyme). The cDNA was synthesized using the SuperScript VILO Master Mix kit (Thermo Fisher). Syncrip and Snhg5 expression levels were measured by RT-qPCR performed on the LightCycler 480 (Roche) using the TaqMan Multiplex Master Mix (Thermo Fisher). Each sample was analyzed in duplicate, and gene expression was normalized relative to the expression of two housekeeping genes, Hprt and Gapdh. TaqMan probes are listed in Supplementary Methods.
Gene Silencing in Murine and Human Cells
Short hairpin RNA sequences are listed in Supplementary Methods. They were cloned under the H1 promoter in the pTRIP/ΔU3-MND-GFP lentivector. The vector that was used to construct tandem shRNAs was a kind gift from Bruno Verhasselt (Ghent University Hospital, Belgium). pTRIP/ΔU3-MND-Ctrl-GFP and pTRIP/ΔU3-MND-Ctrl-Cherry for shRNA Luc were previously described (15, 51). Silencing efficiency was checked by immunoblot (human or murine SYNCRIP; Abcam clone I8E4, ab10687) and RT-qPCR (SYNCRIP, SNHG5, and U50A). RT-qPCR was performed on RNA samples using TaqMan probes (Applied Biosystems); primers and probes are listed in Supplementary Methods. Gene expression was normalized relative to the expression of the reference genes GUS, Ywhaz, and/or Gapdh.
Short del6q Engineering by CRISPR/Cas9
Potential Cas9 target sites were searched within 2 kb of the two germline regions encompassing the chromosomal breakpoint so as to minimize potential off-target effects (52). The corresponding oligonucleotides were cloned in a derivative of the GeckoV2 with the puromycin resistance gene inactivated (52). Two sets of Cas9 target sequences bordering the inner edge of the intended deletion were selected for genome engineering (set 1: GGCACACTACAAAGCACACC, ATCAGACCTTGTATATGTACC; set 2: GATTGTATTAC TTTTTAACCT, GTATTATGTGTACATATACA). A 1.8-kb DNA fragment spanning the positions 86,349,297–86,347,424 of chromosome 6 (GRch37/hg19) and a 1.1-kb DNA fragment spanning the positions 86,387,432–86,388,539 were amplified from genomic DNA and cloned respectively in the upstream and downstream polylinkers of the HR110-PA1 homologous recombination vector (SBI). The linearized HR construct and the Cas9 vector set (2:1 ratio) were cotransfected by nucleofection in CCRF-CEM T-ALL cells with an Amaxa Nucleofector in accordance with the manufacturer's recommendations. Bulk transfected cells were puromycin selected and cloned by RFP+ flow-cytometric cell sorting. Single-cell–derived clones were expanded in RPMI medium (Life Technologies) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies) under constant puromycin (Life Technologies) selection at 2 μg/mL. The expanded individual RFP+ clones were screened for the microdeletion by PCR genotyping with two sets of primers (system 1: CGGGGGAGGGACGTAATTAC, GCCTTTACTAAAATGGCGAAG; system 2: GCTCTCATTGAAACTGTAAGCA, GTGGCGGCCGCTGTCTAGAT). The overall targeting efficiency was 8.3%, with 16 of 192 single-cell–derived clones harboring the simultaneous recombination events at SYNCRIP and SNHG5 loci (data not shown). Individual SYNCRIP–SNHG5 microdeleted clones were further validated by our custom 6q14 array (Fig. 5B; Supplementary Fig. S4). The level of hnRNP-Q protein was analyzed by immunoblot using human SYNCRIP Ab (clone I8E4, sc-56703, Santa Cruz Biotechnology).
The CCRF-CEM T-ALL cell line had been purchased from DSMZ in 2008, aliquoted, and authenticated by SIL-TAL1 testing and arrayCGH (not shown); for the current work, a fresh vial was defrozen, and the cell line was reauthenticated by arrayCGH. The isogenic CCRF-CEM cell line and SYNCRIP–SNHG5 microdeleted clones were routinely tested for Mycoplasma using a luminescence-based array (MycoAlert Mycoplasma Detection Kit, Lonza), most recently in May 2018. Cells were passaged no more than 20 times before thawing new low-passage batches that were also Mycoplasma tested before use.
Label-Free Proteomic Quantitative Analysis
Two deleted and two nondeleted clones were analyzed, each of them in duplicate (total 8 samples). Cells were lysed in 50 mmol/L Tris SDS 2% pH8.5 and boiled 10 minutes at 95°C, and peptides were prepared using the filter-aided separation method. Proteins were digested for 14 hours at 37°C with 1 μg trypsin (Promega), and were fractionated by strong cationic exchange (SCX) StageTips. Mass spectrometry analyses were performed on a U3000 RSLC nano-LC-system coupled to an LTQ Orbitrap-Velos mass spectrometer (Thermo Fisher Scientific). The data were analyzed using MaxQuant version 220.127.116.11 (53). The database used was a concatenation of human sequences from the Uniprot–Swissprot database (Uniprot, release 2015-02) and a list of contaminant sequences from MaxQuant. The false discovery rate (FDR) was kept below 1% on both peptides and proteins. Label-free protein quantification (LFQ) was carried out using both unique and razor peptides. At least two such peptides were required for LFQ. Data were imported into Perseus software (version 18.104.22.168). Protein copy numbers per cell were calculated by standardization on total histone MS signal as described (54).
Polysome fractionation was performed as described previously (55). Briefly, a total of 30 × 106 deleted and nondeleted CEM-CCRF cells were incubated with 50 μg/mL Emetin (Sigma) and cytoplasmic fractions were isolated by mechanical lysis of cells. One milligram of cytosolic proteins was separated on a 15% to 47% sucrose gradient by ultracentrifugation, and absorbance profiles were generated at 254 nm.
mRNAs associated with polysomal ribosomes were analyzed and compared with total cytoplasmic mRNAs in order to identify changes in TE between deleted and nondeleted CEM-CCRF cells. Briefly, polysomes were fractionated from 3 mg of cytosolic extract from each 6q deleted or control sample as described (55). Polysomal fractions were pooled, and mRNAs were extracted from the polysomal and total cytoplasmic fractions using TriPure reagents (Roche) and quality controlled using a BioAnalyzer (Agilent). Libraries were prepared from 0.5 ng of polyA+ fraction per sample using the Scriptseq v2 (Illumina Epicentre), and RNA sequencing was run on a HiSeq1000. Reads were mapped to the human reference genome GRCh38 using STAR v2.5 with the setting –quantMode gene counts to quantify the number of reads per gene. Differential expression analysis and identification of differences in TE were performed using the Anota2Seq program (25, 26). Samples from three independent cell extraction and polysome purification experiments were analyzed after cpm (counts per million) transformation, batch correction, quantile normalization, log2 (x + 10) transformation and selection of the expressed genes monitored by qqnorm fitting resulting in a final paired comparison between three deleted and three nondeleted samples (with polysome and cytoplasmic fraction for each). Default Annota2seq parameters were used (anota2seqRun). Validation of the specificity of the TE changes was checked using permutations between deleted and nondeleted labels (nine aberrant combinations, the cytoplasmic or polysome origin being kept for each sample). R version 3.4.2 (September 28, 2017) was used. The resulting list of 475 genes was analyzed using the GSEA algorithm (broadinstitute.org).
RNA 2′-O-Methylation Profiling by High-Throughput Sequencing (RiboMethSeq)
Site-specific rRNA methylation was determined by RiboMethSeq as previously described (56). Briefly, 100 ng of total cellular RNA was subjected to alkaline hydrolysis in 50 mmol/L bicarbonate buffer pH 9.2 for 10 to 12 minutes at 95°C. RNA fragments were purified and converted to cDNA libraries using the NEBNext Small RNA Library kit (New England Biolabs). Libraries were subjected to high-throughput sequencing using an Illumina HiSeq 1000 instrument with a 50-bp single-end read mode. Adapter sequence trimming was conducted using Trimmomatic-0.32. Alignment to the reference rRNA sequence was achieved by Bowtie2 (ver 2.2.4) in End-to-End mode. 5′-end counting was carried out directly on *.sam file using dedicated Unix script. Final analysis was performed by calculation of MethScore for quantification of 2′-O-methylated residues.
Quantification of Global Protein Neosynthesis
Total protein synthesis rates were measured with Click-iT AHA incorporation assay (Thermo Scientific) in accordance with the manufacturer's recommendations. Briefly, cells were seeded at 2.5 × 105 cells/mL per well in 24-well plates in RPMI medium (Life Technologies) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies) ± puromycin (Life Technologies) selection at 2 μg/mL and grown for 24 hours under basal (normoxia) or stress (3% O2 hypoxia) conditions. Before proceeding to Click-iT AHA incorporation, cells were washed once with warm Ca2+and Mg2+-free phosphate-buffered saline (PBS) and incubated for 60 minutes at 37°C in methionine-free medium to deplete methionine reserves. Click-iT AHA (L-azidohomoalanine; Thermo Scientific) was added to the culture methionine-free medium to a final concentration of 100 μmol/L for 2.5 hours, and then cells were washed twice with PBS. Cells were fixed in 0.5 mL of 4% paraformaldehyde in PBS for 15 minutes on ice, permeabilized in 0.5 mL 0.1% saponin (Sigma) in PBS supplemented with 1% BSA (Sigma) for 10 minutes on ice and washed once with PBS supplemented with 3% BSA. The azide–alkyne cycloaddition was performed using the Click-iT Cell Reaction Buffer Kit (Thermo Scientific), with the alkyne conjugated to Alexa Fluor 488 (Thermo Scientific) at 1 μmol/L final concentration. Cells were incubated for 30 minutes for the reaction and washed twice with PBS before proceeding to flow cytometry.
As independent experiments, 35>S-labeled methionine incorporation was measured in the deleted and nondeleted cells following 18-hour culture under hypoxia (2% O2). Cells were labeled 30 minutes with 75μCi of a [35S]-methionine and [35S]-cysteine mix as described in ref. 55.
Mitochondrial Respiration and Glycolytic Function Measurements
OCR was measured with the Seahorse XFp Cell Mito Stress Test on the Seahorse XFp Analyzer (Agilent, Seahorse Bioscience) in accordance with the manufacturer's recommendations. Briefly, cells were seeded at 5 × 104 cells per well (optimal cell seeding density was determined in preliminary experiments) in an XFp-well miniplate previously coated with Cell-Tak (Corning, Fisher Scientific) in 180 μL of XF Base Medium, supplemented with fresh sodium pyruvate (1 mmol/L), glutamine (2 mmol/L), and glucose (10 mmol/L; pH 7.4). Sequential compound injections of oligomycin A (1 μmol/L), FCCP (1 μmol/L), rotenone (0.5 μmol/L), and antimycin A (0.5 μmol/L) enabled measurements of basal respiration, ATP production, proton leak, maximal respiration, spare respiratory capacity, and nonmitochondrial respiration. Extracellular acidification rate (ECAR) was measured with the Seahorse XFp Glycolysis Stress Test in accordance with the manufacturer's recommendations. Cells (5 × 104 per well) were seeded in 180 μL of the glycolysis stress test medium [XF Base Medium supplemented with fresh glutamine (2 mmol/L) only; pH adjusted to 7.4] on a previously coated XFp-well miniplate. Sequential compound injections of glucose A (10 mmol/L), oligomycin (1 μmol/L), and 2-deoxy-glucose (2-DG, 50 mmol/L) enabled measurements of glycolysis, glycolytic capacity, glycolytic reserve, and nonglycolytic acidification. Oxygen consumption and glycolytic flux were recorded for 90 minutes after cells were placed into the analyzer. OCR (pmol/minute) and ECAR (mpH/minute) were measured simultaneously in all wells, three times at each step, by optical fluorescent O2 sensors, and a minimum of three replicates were analyzed per condition in any given experiment; all compounds and material were obtained from Agilent Seahorse Bioscience.
T-ALL Mouse Model
Animals were housed and handled in the IUH Département d'Expérimentation Animale in accordance with the guidelines of the Animal Care and Use Committee (IUH, Hôpital Saint-Louis). The pSil-TSCL (Tal1tg) and Lck-LMO1 (Lmo1tg) transgenic mouse lines were kind gifts from P. Aplan (NCI, Bethesda, MD; ref. 21). They were backcrossed onto the C57BL6/J background (The Jackson Laboratory) for more than 12 generations. Bone marrow was harvested from the femur of 3- to 4-week-old double-transgenic C57BL/6 mice, and Lin− cells were infected with lentiviral vectors and cultured overnight. The cells were then infected with a retroviral vector for a Notch1-mutant (Notch-ICN, intracellular cleaved Notch1) that was coupled to a truncated nerve growth factor receptor (tNGFR), kindly provided by J. Ghysdael (Institut Curie-Orsay, France) with the permission of W. Pear (ref. 57; University of Pennsylvania, Philadelphia, PA). Next, 2 × 105 cells were i.v. injected into the tail of sublethally irradiated (3 Gy) RAG−/−γc−/− mice. Recipient mice were monitored by FACS analysis of blood cells and clinical observation. Pathology analysis confirmed T-ALL diagnosis, with marrow and spleen infiltration by lymphoblasts expressing the Thy-1, CD4, and/or CD8 T-cell markers.
Xenograft of Transduced Primary Leukemic Cells from Patients
This experimental procedure has been described in detail previously (15). Briefly, 3 × 106 primary human leukemic cells were transduced using a lentivirus containing a shRNA Ctrl-GFP or SYNCRIP–SNHG5-GFP vector, cultured for 5 days, and sorted by FACS, and 5 × 103 cells were i.v. injected to sublethally irradiated (2.25 Gy) NSG mice. For competitive experiments, cells were transduced with a shRNA Ctrl-Cherry or SYNCRIP–SNHG5-GFP lentivirus, FACS-sorted, and mixed in equal number. The starting GFP+/Cherry+ ratio was rechecked by FACS, and 1 × 104 of these cells were i.v. injected into sublethally irradiated (2.25 Gy) NSG mice. Leukemia development was monitored by FACS analysis of blood cells.
In Vitro Limiting Dilution Experiment of Xenografted T-ALL Cells Exposed to Tigecycline
PDX T-ALL cells from a TAL1d non-del6q primary case were exposed to 2.5 μmol/L tigecycline (Sigma) or an equivalent amount of DMSO and cocultured on MS5-DL1 feeder cells as previously described (15) for 48 hours. CD45+CD7+GFP− human leukemic cells were then FACS sorted and seeded on irradiated (30 Gy) MS5-DL1 cells under limiting dilution conditions from 15,625 to 1 cell per well (24 wells per condition). FACS analysis of recovered human CD45+CD7+GFP− tigecycline or DMSO treated cells was performed after 2 to 4 weeks of culture upon hypoxia (1.5% O2). Human PDX T-ALL cells were stained with APC anti-human CD45 (clone 5B1; Miltenyi Biotec) and PE-Vio770 anti-human CD7 (clone 6B7; Miltenyi Biotec) antibodies. Flow cytometry analyses were performed on a FACS Canto II or a Fortessa Analyzer (Beckton Dickinson).
Immunophenotype Analysis of Xenografted T-ALL Cells Exposed to Tigecycline or SYNCRIP–SNHG5 shRNA-Mediated Silencing
PDX T-ALL cells from a TAL1d non-del6q patient were exposed to 2.5 μmol/L tigecycline (Sigma) as above or transduced with a lentivirus containing a shRNA Ctrl-GFP or SYNCRIP–SNHG5-GFP vector as indicated above for diagnosis cells, and cultured for 48 hours on MS5-DL1 or on coated DL4, respectively. Mean fluorescence intensity of cell-surface expression for CD34, CD8, and CD4 was measured by flow cytometry analysis and represented as ratios of cells expressing strongly each surface marker. Human PDX T-ALL were stained with Pacific Blue anti-human CD45 (Beckman Coulter), PE-Vio770 anti-human CD7 (clone 6B7; Miltenyi Biotec), APC anti-human CD34 (clone 561; Sony), APC anti-human CD8 (Beckman Coulter), and PE anti-human CD4 (Beckman Coulter) antibodies. Immunophenotype analyses were performed on a FACS Fortessa Analyzer (Beckton Dickinson).
Quantification of rRNA Methylation by RT-qPCR
This procedure, described by Belin and colleagues (34), was used on RNAs extracted from xenografted human T-ALL samples. The first step was reverse transcription (RT), performed at high (1 mmol/L) or low (10 μmol/L) dNTP concentration. RT is blocked by the presence of a methyl group at low dNTP concentration, but it is not affected when the reaction is performed at high dNTP concentration. The second step was to quantify the respective cDNAs by qPCR with primers surrounding the methylation site. The methylation level was evaluated by the ratio of Ct obtained from the RT performed in the two conditions. RT was performed using 50 ng total RNA in the presence of 200 units of M-MLV reverse transcriptase (Invitrogen), 10 mmol/L DTT, 10 μmol/L or 1 mmol/L dNTPs, and 5 ng/μL of random primers (Promega). Reactions were incubated at 37°C for 50 minutes and then stopped for 15 minutes at 70°C. Quantitative amplification of cDNAs was assessed by PCR using LightCycler FastStart DNA Master SYBR Green I and a LightCycler 480 (Roche Applied Science). The primers were 5′-AGGTAAGGGAAGTCGGCAAG-3′ and 5′-CAGCCCTTAGAGCCAATCCT-3′ and 5′-CCGTAAGGGAAAGTTGAAAAG-3′ and 5′-CCCCACCCGTTTACCTCTTA-3′ for the 28S-2858 and 28S-389 sites, respectively. The methylation level was calculated as Log2(Ctlo − Cthi), where the Ctlo and Cthi values corresponded to Ct obtained after RT performed at low and high dNTP concentrations, respectively.
Statistical analyses were performed using the Fisher, Mann–Whitney, log-rank, unpaired t test and binomial tests, as indicated in the figure legends.
Disclosure of Potential Conflicts of Interest
H. de The is a consultant/advisory board member for VectorLab. No potential conflicts of interest were disclosed by the other authors.
Conception and design: S. Gachet, T. El-Chaar, D. Avran, E. Genesca, S. Quentin, I. André-Schmutz, J.P. Meijerink, E. Clappier, C. Gazin, J. Soulier
Development of methodology: S. Gachet, T. El-Chaar, D. Avran, E. Genesca, S. Quentin, D. Briot, L. Hernandez, J.P. Meijerink, J.-J. Diaz, C. Gazin
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Gachet, D. Avran, E. Genesca, F. Catez, S. Quentin, G. Thérizols, G. Meunier, L. Hernandez, M. Pla, W.K. Smits, J.G. Buijs-Gladdines, T. Taghon, P. Van Vlierberghe, J.P. Meijerink, A. Baruchel, H. Dombret, E. Clappier, C. Gazin
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Gachet, T. El-Chaar, D. Avran, E. Genesca, F. Catez, S. Quentin, M. Delord, W. Van Loocke, G. Menschaert, T. Taghon, P. Van Vlierberghe, E. Clappier, J.-J. Diaz, C. Gazin, H. de Thé, F. Sigaux, J. Soulier
Writing, review, and/or revision of the manuscript: S. Gachet, T. El-Chaar, D. Avran, E. Genesca, S. Quentin, J.P. Meijerink, A. Baruchel, H. Dombret, J.-J. Diaz, H. de Thé, F. Sigaux, J. Soulier
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. El-Chaar, D. Avran, S. Quentin, J.P. Meijerink, H. Dombret, E. Clappier
Study supervision: S. Gachet, J. Soulier
We would like to acknowledge contributions from François Guillonneau, Marjorie Leduc, and Patrick Mayeux (3P5 proteomic facility, Université Paris Descartes, Institut Cochin, Paris), Valerie Igel-Bourguignon, Lilia Ayadi, Virginie Marchand, and Yuri Motorin (NGS Core Facility, FR3209 CNRS-UL, Université de Lorraine, Vandoeuvre-les-Nancy), Niclas Setterblad and Antonio Alberdi (Genomic Platform of Institut Universitaire d'Hématologie, University Paris Diderot, Paris, France), Carèle Fédronie (Institute of Hematology, Université Paris Diderot, Paris, France), Pierre De La Grange (Genosplice, Paris, France), and Olivier Alibert and Amélie Rondot (CEA/DSV/iRCM/LEFG Genopole, Evry, France). This work was supported by grants to J. Soulier from the ERC St Grant Consolidator #311660, the CIT program from the Ligue Contre le Cancer, the Cancéropole IDF, and the ANR-10-IBHU-0002 Saint-Louis Institute program. Work from the J.J. Diaz team was supported by ANR RiboMeth-13-BSV8-0012. T. El-Chaar, D. Avran, and G. Therizols were supported by fellowships from Fondation ARC, INCa, and the Ligue Contre le Cancer.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.