Background
B cell precursor acute lymphoblastic leukemia (BCP-ALL) is the most prevalent disease in children but affects also adults. Despite improvements in treatment regimens such as chemotherapy and allogeneic hematopoietic stem cell transplantation, the prognosis remains poor for patients in high-risk groups and at relapse [
1]. Various risk subtypes have been established based on the cytogenetic analysis and molecular genetics studies. These subtypes are classified based on the presence of high hyperdiploidy (51–65 chromosomes) [
2], hypodiploidy (less than 44 chromosomes) [
3], and fusion genes, such as BCR-ABL, ETV6-RUNX, and MLL [
4]. About 70–80% of both adults and pediatric cases of BCP-ALL constitute these subtypes, although the frequency may differ [
5].
Recent efforts taking advantage of whole transcriptome sequencing (RNA-seq) have refined this classification by identifying novel BCP-ALL subtypes [
6]. RNA-seq analyses identified cytogenetically non-detectable recurrent rearrangements and gene fusions, which allowed characterization of additional subtypes based on distinct gene expression profiles [
7]. For example, the DUX4 subtype is defined mainly by the IGH-DUX4 [
8] gene fusions; the Ph-like subtype is a high-risk subtype with a gene expression profile similar to Ph-positive ALL, but lacking BCR-ABL1 fusion gene [
9]; and the near haploid/high hyperdiploid (NH-HeH) (51–67 chromosomes) subtype is a common subtype [
10,
11] comprising 30% of all pediatric BCP-ALL. These subtypes are clinically relevant with distinct gene expression profile and have been widely studied in the recent past.
Nevertheless, we are far from complete understanding of BCP-ALL subtypes and their heterogeneity and its associated molecular mechanisms, which pose a major challenge for improving diagnosis and therapy. Recent studies have suggested that long non-coding RNAs (lncRNAs) and small non-coding RNAs (e.g., microRNAs) play a key role in development and progression of leukemia [
11] and thus constitute as new biomarkers and potential targets for novel therapies [
12].
lncRNAs are arbitrarily defined as transcripts longer than 200 bp and lacking an extended protein-coding open reading frame (ORF). It has become apparent that lncRNAs are frequently spliced and polyadenylated and are mainly transcribed by RNA polymerase II [
13]. lncRNA expression has been reported as highly tissue-specific even though the expression abundance is generally lower compared to protein-coding genes [
14]. The expression specificity has been extended to a wide variety of physiological and pathological mechanisms like cancer development and pluripotency [
15]. lncRNAs can act either proximally (in the
cis region) or distally (in the
trans region) interfering in the transcriptional regulation of protein-coding genes [
16]. Like proteins, various lncRNAs are attributed to oncogenic or tumor-suppressive activities exerting various cellular functions [
17,
18]. In addition, lncRNAs regulate gene expression at the epigenetic [
19] and post-transcription levels [
20]. Genome-wide association studies in cancer have disclosed that 80% of cancer-associated single-nucleotide polymorphisms (SNPs) [
21] are in non-coding regions [
22], including lncRNAs, suggesting that a significant portion of the genetic etiology of cancer can be related to lncRNAs. Moreover, lncRNAs are reported to be useful for disease prognosis, exemplified by the lncRNA HOTAIR [
23] (HOX transcript antisense RNA), which is upregulated in acute myeloid leukemia (AML) patients.
So far, the majority of profiling studies explored the role of single lncRNAs in leukemia including AML [
24], chronic lymphocytic leukemia (CLL) [
25], and pediatric ALL [
21,
26]. Yet a comprehensive genomic and epigenetic delineation of lncRNA deregulations in BCP-ALL subtypes and their molecular and functional insights during the evolution of the disease are lacking.
In the present study, we explored lncRNA landscapes in DUX4, Ph-like, and NH-HeH BCP-ALL subtypes and extracted novel biological and functional insights of BCP-ALL subtype-specific lncRNAs and their epigenetic activity. On the basis of RNA-seq transcriptional and DNA methylation survey of lncRNAs, we have determined 1235 subtype-specific and relapse-specific lncRNAs. Interestingly, a subset of lncRNAs were epigenetically altered. From our in-depth analyses, we have inferred the potential functions of subtype-specific lncRNAs. Overall, this work provides a most comprehensive and integrative insight that highlight the impact of lncRNAs on relevant pathways that are dysregulated in the molecular subgroups of BCP-ALL and may provide new approaches for prognosis and treatment.
Discussion
Although previous studies have demonstrated the involvement of lncRNAs in acute leukemias [
21,
25], comprehensive characterization of the transcriptome, epigenetic regulation and functional contribution of lncRNAs in distinct BCP-ALL subtypes are lacking. LncRNAs, as the novel class of functional molecules, are involved in cancer biology and have previously been reported in different molecular subtypes in various cancers. However, their role in BCP-ALL subtypes has not been investigated. Here, we unravel the lncRNA landscape using transcriptome and methylome data from 45 (adult and pediatric) relapsed BCP-ALL patients focusing on the three molecular subtypes namely DUX4, Ph-like, and NH-HeH.
Our integrated transcriptomic analyses using RNA-seq and DNA methylation bring significant novel insights and advances: they provide the most comprehensive novel datasets so far for BCP-ALL subtypes. We provide a resource of subtype-specific and relapse-specific lncRNAs and potential lncRNA functions and uncover their epigenetic alterations within the BCP-ALL subtypes. We identified 1235 DE subtype-specific lncRNAs dysregulated in at least one of the three subtypes. These 1235 DE subtype-specific lncRNAs successfully stratified subtypes in our discovery cohort, an independent validation cohort.
Another important aspect of our study is the identification of relapse-specific dysregulated lncRNAs across three BCP-ALL subtypes. A closer look into the relapse-specific lncRNA signature identified lncRNAs previously described as oncogenic including
RP11-701P16.5 [
50],
SLC38A3 [
51], and
LINC00312, which are upregulated in relapsed samples within DUX4 subtype.
Importantly, apoptosis suppressor lncRNA in Myc-driven lymphomas
miR-17/92 cluster host gene (
MIR17HG) [
52] is upregulated in relapsed samples within the Ph-like subtype and downregulated in relapsed samples within DUX4 and NH-HeH subtypes. Overall, the relapse-specific lncRNAs highlight the oncogenic relevance in BCP-ALL subtypes: near haploid or high hyperdiploid (NH-HeH;
n = 16) and low hypodiploid. Besides the oncogenic properties, lncRNAs can act as prognostic markers [
53,
54] and aid at disease diagnosis and treatment. A subset of our relapse-specific lncRNAs (
n = 61, Additional file
4: Table S3) overlaps with the prognostic markers identified from 14 Pan-Cancer datasets [
42], including lung cancer-associated transcript 1 (
LUCAT1), which is previously reported for its drug resistance in solid cancer [
55]. Within the DUX4 subtype, we identified the upregulated expression of
LUCAT1 at relapse, potentially providing a novel insight into treatment resistance for BCP-ALL subtypes. Together, this illustrates the catalog of relevant lncRNAs in different subtypes of BCP-ALL as subtype-specific and relapse-specific markers with the potential of RNA-based approaches in the treatment of BCP-ALL subtypes.
The dissection of the regulatory pathways mediated by the molecular subtype-specific and relapse-specific lncRNAs revealed the activation of pivotal signaling pathways across three BCP-ALL subtypes. The functional analysis by means of the guilt-by-association approach highlights the subtype-specific and relapse-specific lncRNAs associated with activation of signaling pathways and metabolic pathways that are associated with leukemogenesis including TGF-Beta, hippo, P53, JAK-STAT, cytokine-cytokine receptor, endocytosis, mTOR, and metabolic pathways. Characterization of the lncRNAs involved in these pathways may potentially reveal novel targets in molecular therapies.
The functional insights of relapse-specific and subtype-specific lncRNAs revealed biological relevance to BCP-ALL subtypes including cell cycle functions, signal transduction, cell migration, and metabolic processes. Some of the functions predicted here corroborate previous studies emphasizing the strengths of the employed guilt-by-association. For example, lncRNA
AC002454.1, which we associated to the PIK3-AKT pathway in Ph-like subtype, was recently reported to regulate cyclin-dependent kinase (
CDK6) to participate in cell cycle disorder [
56]. The
CDK6 gene appears to be frequently dysregulated in hematopoietic malignancies [
45] and is hence attributed a critical role in tumorigenesis, also shown in ALL driven by mixed lineage leukemia (MLL) fusion proteins [
57]. In Ph-like subtype, both
CDK6 and
AC002454.1 are correlated and upregulated specifically in Ph-like samples, suggesting they displayed enhancer-like functions. We identified 8 relapse-specific lncRNAs (Additional file
4: Table S3) associated with metabolic pathways in the DUX4 subtype overlapping with lncRNAs [
58] reported to synergistically dysregulate metabolic pathways in multiple tumor contexts.
Besides known lncRNAs, we also identified novel lncRNAs associated with activation of key signaling pathways. For instance, in the DUX4 subtype, we identified a set of novel lncRNAs associated with TGF-beta pathway, including the antisense
RP11-224019.2, with a significant positive correlation to the
TGFB gene. Recently, a number of lncRNAs were documented to be associated with TGFß signaling pathway, including MEG3 regulating the TGFB2 pathway in breast cancer [
40]. However, lncRNAs associated with the TGFß pathway in BCP-ALL subtypes have not been reported. The co-expression of
RP11-224019.2 and
TGFB in DUX4 subtype may indicate their functional relatedness or regulatory relationships. In addition to that, a notable observation was a strong correlation between relapse-specific lncRNAs with genes involved in the activation of metabolic pathways in the DUX4 subtype. We identified 112 relapse-specific lncRNAs co-expressed with 29 (Additional file
4: Table S3) PC genes activated in metabolic pathways, including previously reported 8 biomarker lncRNAs. For example, we identified oncogenic lncRNA
LUCAT1 reported to be associated with poor prognosis in lung cancer [
59]. However, the
LUCAT1 has not yet been reported in the BCP-ALL context. The global co-expression analysis and gene expression profiling suggest important and previously unappreciated roles of lncRNAs in the BCP-ALL subtypes. Our analyses provide important functions of subtype-specific and relapse-specific lncRNA genes whose expression correlates tightly with oncogenic coding genes.
Although we observed that subtype-specific lncRNAs and subtype-specific protein-coding genes were predicted to activate or inhibit the same pathways, some important exclusivity was observed. For instance, the signaling pathways such as the PI3K and mTOR in Ph-like subtype was enriched only in the lncRNA-based enrichment analysis, whereas these pathways did not appear to be enriched/dysregulated in the mRNA-based analysis. The PI3K and mTOR signaling pathways control proliferation, differentiation, and survival of hematopoietic cells [
54]. Consistent with our results, other studies indicated the potency of lncRNAs facilitating the cancer cell growth through mTOR and PI3K signaling pathways [
33,
44,
55], yet reports on BCP-ALL subtypes have been lacking so far. Considering the functional nexus between Ph-like-specific lncRNAs and the activation of pathways such as mTOR and PI3K signaling pathways, targeting those lncRNAs may be a promising novel therapeutic target for BCP-ALL subtypes.
Our work also underscores the importance of epigenetic alterations in modulating lncRNA transcriptional activities. Although previous studies [
60‐
62] have demonstrated cross-talk between DNA methylation and transcriptional activities of lncRNAs, their role in the etiology of BCP-ALL subtypes has not been investigated. DNA methylation analyses of lncRNAs revealed that DNA methylation might underlie the differential expression of BCP-ALL subtype-specific lncRNAs. Some subtype-specific lncRNAs identified here have been reported by previous studies. For example,
SOX2-OT (67, [
63]),
LINC00312 [
46],
TCL6, and
PVT1 are onco-lncRNAs, which are promoter methylated in one of the three subtypes. The lncRNA,
PVT1, was reported for its MYC activity [
64,
65] and as oncogenic lncRNA with multiple roles in cell growth, dysfunction, and differentiation in AML [
66]. Both lncRNAs,
LINC00312 and
TCL6, have been extensively investigated on expression levels but not on the epigenetic level. The promoters of both
TCL6 and
LINC00312 were observed to be hyper-methylated with corresponding diminished expression in the DUX4 and NH-HeH samples. Notably, the DNA methylation analysis of lncRNAs revealed that DNA methylation might underlie the differential expression of subtype-specific lncRNAs.