Long non-coding RNAs as novel therapeutic targets in juvenile myelomonocytic leukemia

Juvenile myelomonocytic leukemia (JMML) treatment primarily relies on hematopoietic stem cell transplantation and results in long-term overall survival of 50–60%, demonstrating a need to develop novel treatments. Dysregulation of the non-coding RNA transcriptome has been demonstrated before in this rare and unique disorder of early childhood. In this study, we investigated the therapeutic potential of targeting overexpressed long non-coding RNAs (lncRNAs) in JMML. Total RNA sequencing of bone marrow and peripheral blood mononuclear cell preparations from 19 untreated JMML patients and three healthy children revealed 185 differentially expressed lncRNA genes (131 up- and 54 downregulated). LNA GapmeRs were designed for 10 overexpressed and validated lncRNAs. Molecular knockdown (≥ 70% compared to mock control) after 24 h of incubation was observed with two or more independent GapmeRs in 6 of them. For three lncRNAs (lnc-THADA-4, lnc-ACOT9-1 and NRIR) knockdown resulted in a significant decrease of cell viability after 72 h of incubation in primary cultures of JMML mononuclear cells, respectively. Importantly, the extent of cellular damage correlated with the expression level of the lncRNA of interest. In conclusion, we demonstrated in primary JMML cell cultures that knockdown of overexpressed lncRNAs such as lnc-THADA-4, lnc-ACOT9-1 and NRIR may be a feasible therapeutic strategy.

www.nature.com/scientificreports/ methylome studies encouraged the use of the DNA methyltransferase-inhibiting agent azacitidine as a bridge to HSCT, with promising results in single case reports and the first clinical trial [13][14][15][16] . Long non-coding RNAs (lncRNAs) are RNA transcripts with a minimum length of 200 nucleotides and little to no evidence of protein-coding capacity. LncRNAs have emerged as an additional layer of gene expression regulation and were shown to play a role in normal and malignant hematopoiesis 17,18 . Dysregulation of key lncRNAs involved in processes such as maintenance and differentiation of hematopoietic stem cells can result in the development of hematological malignancies 19 . Interestingly, compared to protein-coding messenger RNAs (mRNAs), lncRNAs generally have a more tissue or cell-type specific expression pattern, making them ideal candidates for targeted therapies 20,21 .
Recently, we documented the lncRNA landscape of JMML using microarray profiles from 44 untreated JMML patients and 7 healthy controls, and correlated lncRNA expression with clinical and molecular characteristics 7 . In addition, we discovered a subset of lncRNAs specifically overexpressed in JMML patients that could potentially serve as new therapeutic targets 7 .
GapmeRs, potent antisense oligonucleotides (ASOs), represent a unique tool to perturbate lncRNA expression, even more efficiently than RNA interference 22 . LNA GapmeRs were shown earlier to act in hematological and solid tumors. For example, we have shown that silencing of lncRNA lnc-RTN4R-1 induces cell death in ETV6-RUNX1 childhood B-cell precursor leukemia 23 . Likewise, Leucci et al. recently illustrated that silencing of the lncRNA SAMMSON, in Patient-Derived Tumor Xenograft models of melanoma, strongly decreases melanoma cell survival 24 .
Here, we further refined the lncRNA transcriptome in JMML patients using RNA sequencing (RNA-seq) and quantitative reverse transcriptase PCR (qPCR) and evaluated GapmeR ASOs as a therapeutic strategy targeting overexpressed lncRNAs in JMML.

Results
Patients characteristics. To identify differentially expressed lncRNAs in JMML, we profiled bone marrow (BM) (n = 15) or peripheral blood (PB) (n = 4) mononuclear cell preparations (MNCs) from 19 JMML samples and 3 BM samples from age-matched pediatric haematological normal patients (PNs) using total RNA-seq. The patient cohort consisted of 8 PTPN11, 4 KRAS, 5 NRAS and 2 NF1 mutated patients. Karyotypic abnormalities were detected in 6/19 patients, including monosomy 7 in four cases. The median age at diagnosis was 26 months (range 1-159 months). Seventeen out of 19 patients (89.5%) received HSCT. Relapse after HSCT occurred in 4/17 patients and one patient died of transplant-related toxicity. Detailed patient characteristics are summarized in Supplemental Table S1. Differential RNA expression between JMML patients and normal bone marrow controls. In total, 167,569 genes were identified of which 58,096 corresponded to known Ensembl genes. Of these, 19,979 were coding, and 38,117 non-coding genes. Using edgeR, we identified 227 and 419 differentially expressed coding and non-coding genes, respectively.
The LNCipedia database was used to further identify 8559 known lncRNAs showing expression of ≥ 1 count per million in at least 2/3 of the patients or PN BM samples 21 . Principal component analysis (PCA), using this subset of lncRNAs showed clustering of PN BM samples and divergent of JMML patients (Fig. 1A). Within the JMML samples, heterogeneity can be observed. Moreover, a lncRNA signature was established, consisting of 185 differentially expressed lncRNAs genes (131 up-and 54 downregulated) between JMML and PN BM (Fig. 1B, Supplemental Table S2). In silico analysis with pre-ranked gene set enrichment analysis and lncRNA pathway analysis (see Supplementary methods) identified pathways associated with JAK-STAT-MAPK signaling, cancer development, and systemic inflammation to be synergistically regulated by these differentially expressed lncR-NAs in JMML patients.

Identification and validation of lncRNAs overexpressed in JMML. The top-25 overexpressed
lncRNAs together with other lncRNAs, based on log2 fold change (logFC) and previously discovered differential expression 7 , were selected for validation with quantitative reverse-transcriptase polymerase chain reaction (qPCR). In total, in house designed and validated primers were available for 45 lncRNAs (Supplemental Table S4). In our independent validation cohort (12 JMML patients and 6 PN), significant and near-significant (P ≤ 0.1) overexpression was confirmed for 10 and 4 lncRNAs respectively (31%) (Supplemental Figure S1). However, due to the small sample size of this cohort, not all molecular subgroups were represented in sufficient numbers. After expansion of the validation cohort with patients from the discovery cohort, more lncRNAs showed significant or near significant overexpression ( www.nature.com/scientificreports/ plemental Table S5). For 16/18 validated lncRNAs, sufficient data was present to perform a correlation analysis between RNA-seq and qPCR, and for 10/16 lncRNAs a significant and positive correlation could be observed (Supplemental Figure S2). Although upregulation was noted for both lnc-ACOT9-1 and MKLN1-AS in all mutational subgroups in our RNA-seq data, this was only validated for lnc-ACOT9-1 in the qPCR validation. Detailed analysis of our qPCR results indicated that lnc-THADA-4 was low in all of the KRAS mutated samples from our cohort (P = 0.006), whereas lnc-EGR3-1 was overexpressed (> 3 SD above the average expression in PN) in all KRAS and NF1 mutated patients (P = 0.044), and LINC02217 in NRAS mutated patients (P = 0.02) (Supplemental Figure S3). None of the lncRNAs validated was associated with outcome in our cohort. However, MIR222HG overexpression was significantly associated with normal HbF levels (P = 0.032), overexpression of SH3RF3-AS1 with low platelet counts (P = 0.045), overexpression of LINC02218 with high platelet counts (P = 0.037) and overexpression of lnc-SPOPL-18 with low blast counts (P = 0.049) at diagnosis. lncRNA knockdown in hematopoietic cell lines. Next, LNA GapmeRs were designed for each lncRNA with validated overexpression in JMML, to study the therapeutic potential and biological functions. To minimize off-target effects of individual GapmeRs, we aimed to evaluate a minimum of four different GapmeRs per lncRNA. For 8 lncRNAs it was not possible to design four LNA GapmeRs of sufficient predicted quality. For the remaining 10 lncRNAs (lnc-THADA-4, lnc-ACOT9-1, NRIR, lnc-ACSL1-1, lnc-EGR3-1, MIR222HG, LINC02217, lnc-CYP4F22-4, lnc-RUNX3-4 and MEG3), GapmeRs could be designed and subsequently tested.
Since no JMML cell lines are available to evaluate GapmeRs, we screened publicly available RNA-seq data from 934 human cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE) to identify hematopoietic cell lines with overexpression of the lncRNAs validated. For all these lncRNAs, at least one cell line from another pediatric or adult hematopoietic malignancy was selected and expression verified with qPCR (Supplemental Figure S4; Supplemental Table S6).
After 24 h incubation of cells with individual GapmeRs at a concentration of 5 µM, expression of the targeted lncRNA was determined with qPCR and compared with cells treated with empty mock control. For six lncRNAs (lnc-THADA-4, lnc-ACOT9-1, NRIR, MIR222HG, LINC02217, and MEG3) a downregulation ≥ 70% of the level of expression of the targeted lncRNA for two or more GapmeRs could be observed (Supplemental Figure S5).

Discussion
JMML is a very rare pediatric disease in which HSCT remains the main option for cure, but is associated with major morbidity. In addition, many patients still experience relapse. Thus, the identification of novel therapeutic targets is warranted. Consensus is growing that lncRNAs are involved in solid and hematopoietic human cancers and that cancer-associated lncRNAs might reveal new prognostic biomarkers. Moreover, using RNA-based www.nature.com/scientificreports/ therapies, cancer-associated lncRNAs may serve as specific therapeutic targets 25 . With this study, we demonstrate that lncRNA perturbation using LNA GapmeRs may serve as a novel therapeutic option in JMML in vitro.
Using total RNA-seq in a cohort of 19 JMML patients, we identified 185 differentially expressed lncRNAs. Despite the generally lower expression of lncRNAs, we were still able to validate approximately 45% of overexpressed lncRNAs with qPCR. Of note, 5/13 lncRNAs, i.e. lnc-THADA-4, previously identified as upregulated in JMML 7 , could be confirmed in this study. RNA profiles may differ between BM and PB samples, and both BM (n = 15) and PB (n = 4) samples were present in our RNA-seq cohort. Nevertheless, differential expression analyses between BM and PB samples did not reveal differentially expressed lncRNAs.
As JMML is a very rare disease it was not possible to establish an independent validation cohort of sufficient size for all molecular subgroups. However, when using 12 independent samples as validation cohort we found comparable although less powered results. In addition to the scarcity of diagnostic samples, no JMML specific cell lines are available for research. Therefore, it is still difficult to study novel therapies for this disease. We tackled this problem by using cell lines of other hematopoietic malignancies as surrogates and by generating primary JMML-cell cultures. We demonstrated that gymnotic delivery of LNA GapmeRs targeting overexpressed lncRNAs in JMML can decrease cell proliferation in both hematopoietic cell lines and primary MNC cultures of JMML patients. LNA GapmeRs against lnc-THADA-4, NRIR and lnc-ACOT9-1 could impact cell viability with effects of -30 to − 50%, 72-96 h after a single dose. Moreover, we could demonstrate a correlation between GapmeR effect and the level of molecular expression of the lncRNA of interest.
For most lncRNAs functional characterization is still lacking. It is known that some lncRNAs can regulate local chromatin structure and the expression of neighboring genes through its ability to recruit regulatory factors to the locus and/or modulate their function. Interestingly, NRIR is located close to the protein-coding gene www.nature.com/scientificreports/ CMPK2, which is overexpressed in chronic myeloid leukemia K-562 and lymphoblastic leukemia MOLT-4 cell lines, implicating a role in the pathophysiology of these diseases 26 . Our data demonstrated a high Spearman correlation coefficient between NRIR and CMPK2 (p value = 0.002; R = 0.87). Other mRNA genes with high correlation with NRIR are alpha interferon-induced genes and PKR (protein kinase R; R = 0.90 and p value < 0.001). Increased expression of the latter gene has been reported in acute leukemia and solid tumors and correlates with worse survival and shortened remission duration in AML 27 . Lnc-ACOT9-1 is surrounded by the protein-coding genes SAT1 and ACOT9, both of which are overexpressed in JMML patients based on our previously generated microarray data. Noteworthy, lnc-THADA-4 was found to be essential for cellular growth of chronic myeloid leukemia K-562 cells in a genome-wide screening for functional lncRNAs by Cas9 targeting of splice sites 28 . Our finding of decreased cell viability of overexpressing cell lines and primary culture of JMML cells after knockdown of lnc-THADA-4 further corroborates the postulated essential role of this lncRNA. Chemical modifications, like LNAs, could increase potency and simultaneously increase affinity to unintended targets with partial sequence complementarity 29 . As we did not observe a decrease in cell viability in cells without molecular expression of the lncRNA, against which the GapmeR is targeted, severe off-target effects seem improbable. GapmeRs are already employed for other diseases and are currently in phase 1 and 2 clinical trials for centronuclear myopathy. However, the safety and efficacy of the GapmeRs we designed in this study need to be validated in in vivo models. In particular, hepatotoxicity is being observed with some LNA-modified Gap-meRs and thus requires extensive testing 30,31 . Recently, two in vivo models have been described and can be used for this purpose, one in Rag2 −/− gamma(c) −/− mice and one in NOD-scid IL2Rgnull-3/GM/SF (NSGS) mice [32][33][34] .
As only part of cells were killed, we propose to use sequential dosings or to combine different GapmeRs against key lncRNAs or combine GapmeRs with other treatment modalities (e.g. chemotherapy, hypomethylating agents, etc.) to increase killing efficacy of JMML cells both in vitro as in vivo. It is still difficult to study novel therapies in this rare disease, due to limited samples and the absence of JMML cell lines. Therefore, the generation of stable in vivo models as described above will be crucial.
In summary, our study further refines the lncRNA transcriptome in JMML and illustrates that lncRNA targeting through LNA GapmeRs could be a promising new approach in the treatment of JMML. We demonstrated in primary JMML cell cultures that knockdown of overexpressed lncRNAs such as lnc-THADA-4, lnc-ACOT9-1 and NRIR may be a feasible therapeutic strategy.

Methods
Patient samples. Diagnostic BM (n = 15) and PB (n = 4) samples from 19 children with JMML and BM of 3 healthy children selected for sibling graft donation (PN) were used for RNA-seq. Sequencing results were verified with qPCR in a validation cohort consisting of 12 additional JMML patients (9 BM and 3 PB) and 6 PN BM samples. Additionally, living cells from 3 JMML patients, derived from spleen or PB, were used for primary cell culture generation. JMML samples were obtained from different institutions throughout Europe (Belgium, the Netherlands, France, the Czech Republic and Germany) (Supplemental Table S1). Out of 33 JMML patients, 17 are registered in the European Working Group of Myelodysplastic Syndromes in Childhood (EWOG-MDS) studies EWOG-MDS98 and EWOG-MDS2006 (National Institutes of Health trials registered as #NCT00047268 and #NCT00662090 at www.clini caltr ials.gov), and 16 in the French national JMML biobank. PN samples were collected at the Ghent University Hospital. Written informed consent was obtained from the parents of the patients or the healthy children before sample collection and approval for the study was granted from institutional review committees at each participating center. All experiments and methods were performed in accordance with the relevant guidelines and regulations.
Mononuclear cell preparations derived from spleen and PB from 3 JMML patients were primary cultured in StemSpan SFEM II medium (Stemcell Technologies, Vancouver, Canada) supplemented with recombinant human IL3 (0.01 µg/mL; PeproTech, London, UK), FLT3L (0.01 µg/mL; PeproTech, London, UK), TPO (0.01 µg/ mL; PeproTech, London, UK) and SCF (0.025 µg/mL; PeproTech, London, UK) and incubated at 37 °C in 5% CO 2 incubators. RNA isolation, RNA sequencing, differential gene expression analysis, functional lncRNA analysis, complementary cDNA synthesis and quantitative reverse transcriptase PCR. Detailed information can be found in Supplementary Methods. Briefly, MNCs from JMML and PN BM were isolated and total RNA was extracted using TRIzol. Samples in the discovery cohort were used for total paired-end RNA-seq, whereas samples in the validation cohort were used for cDNA synthesis and qPCR.
LncRNA perturbation. For each lncRNA of interest, a minimum of four antisense LNA GapmeRs (Qiagen) were designed using the LNA GapmeR designer (Qiagen). GapmeRs were resuspended in Tris-EDTA buffer and stored at − 20 °C. LncRNA expression of selected cell lines and primary cells was verified with qPCR. Cells from cell lines and MNCs from primary patients were plated in 96-well plates and incubated at 37 °C with